[llvm] 949476f - [X86] matchShuffleAsBlend - keep per-lane blend masks 'whole' if only one input is used
Simon Pilgrim via llvm-commits
llvm-commits at lists.llvm.org
Sun Jan 29 05:34:07 PST 2023
Author: Simon Pilgrim
Date: 2023-01-29T13:30:00Z
New Revision: 949476f533838a0d724d184c9f9db856f52fb68d
URL: https://github.com/llvm/llvm-project/commit/949476f533838a0d724d184c9f9db856f52fb68d
DIFF: https://github.com/llvm/llvm-project/commit/949476f533838a0d724d184c9f9db856f52fb68d.diff
LOG: [X86] matchShuffleAsBlend - keep per-lane blend masks 'whole' if only one input is used
For 32/64-bit element 256-bit X86ISD::BLENDI shuffle masks, see if each sub-lane only references (plus undefs) a single input - in which case ensure the blend mask component for that lane ONLY references that input.
This helps later subvector extraction/insertion nodes to peek through the blend node for that lane.
Added:
Modified:
llvm/lib/Target/X86/X86ISelLowering.cpp
llvm/test/CodeGen/X86/any_extend_vector_inreg_of_broadcast.ll
llvm/test/CodeGen/X86/any_extend_vector_inreg_of_broadcast_from_memory.ll
llvm/test/CodeGen/X86/avx512-shuffles/partial_permute.ll
llvm/test/CodeGen/X86/pr38639.ll
llvm/test/CodeGen/X86/pr51615.ll
llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-5.ll
llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-6.ll
llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-3.ll
llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-5.ll
llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-6.ll
llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-7.ll
llvm/test/CodeGen/X86/vector-interleaved-load-i64-stride-3.ll
llvm/test/CodeGen/X86/vector-interleaved-store-i32-stride-7.ll
llvm/test/CodeGen/X86/vector-interleaved-store-i64-stride-5.ll
llvm/test/CodeGen/X86/vector-interleaved-store-i64-stride-6.ll
llvm/test/CodeGen/X86/vector-interleaved-store-i64-stride-7.ll
llvm/test/CodeGen/X86/vector-interleaved-store-i64-stride-8.ll
llvm/test/CodeGen/X86/vector-shuffle-256-v4.ll
llvm/test/CodeGen/X86/vector-shuffle-256-v8.ll
llvm/test/CodeGen/X86/vector-shuffle-combining-avx.ll
llvm/test/CodeGen/X86/vector-shuffle-combining-avx2.ll
llvm/test/CodeGen/X86/zero_extend_vector_inreg_of_broadcast.ll
llvm/test/CodeGen/X86/zero_extend_vector_inreg_of_broadcast_from_memory.ll
Removed:
################################################################################
diff --git a/llvm/lib/Target/X86/X86ISelLowering.cpp b/llvm/lib/Target/X86/X86ISelLowering.cpp
index 2df82b5976973..85d4855b1df08 100644
--- a/llvm/lib/Target/X86/X86ISelLowering.cpp
+++ b/llvm/lib/Target/X86/X86ISelLowering.cpp
@@ -12942,7 +12942,7 @@ static SDValue getVectorMaskingNode(SDValue Op, SDValue Mask,
const X86Subtarget &Subtarget,
SelectionDAG &DAG);
-static bool matchShuffleAsBlend(SDValue V1, SDValue V2,
+static bool matchShuffleAsBlend(MVT VT, SDValue V1, SDValue V2,
MutableArrayRef<int> Mask,
const APInt &Zeroable, bool &ForceV1Zero,
bool &ForceV2Zero, uint64_t &BlendMask) {
@@ -12955,37 +12955,67 @@ static bool matchShuffleAsBlend(SDValue V1, SDValue V2,
ForceV1Zero = false, ForceV2Zero = false;
assert(Mask.size() <= 64 && "Shuffle mask too big for blend mask");
+ int NumElts = Mask.size();
+ int NumLanes = VT.getSizeInBits() / 128;
+ int NumEltsPerLane = NumElts / NumLanes;
+ assert((NumLanes * NumEltsPerLane) == NumElts && "Value type mismatch");
+
+ // For 32/64-bit elements, if we only reference one input (plus any undefs),
+ // then ensure the blend mask part for that lane just references that input.
+ bool ForceWholeLaneMasks =
+ VT.is256BitVector() && VT.getScalarSizeInBits() >= 32;
+
// Attempt to generate the binary blend mask. If an input is zero then
// we can use any lane.
- for (int i = 0, Size = Mask.size(); i < Size; ++i) {
- int M = Mask[i];
- if (M == SM_SentinelUndef)
- continue;
- if (M == i ||
- (0 <= M && M < Size && IsElementEquivalent(Size, V1, V1, M, i))) {
- Mask[i] = i;
- continue;
- }
- if (M == (i + Size) ||
- (Size <= M && IsElementEquivalent(Size, V2, V2, M - Size, i))) {
- BlendMask |= 1ull << i;
- Mask[i] = i + Size;
- continue;
- }
- if (Zeroable[i]) {
- if (V1IsZeroOrUndef) {
- ForceV1Zero = true;
- Mask[i] = i;
+ for (int Lane = 0; Lane != NumLanes; ++Lane) {
+ // Keep track of the inputs used per lane.
+ bool LaneV1InUse = false;
+ bool LaneV2InUse = false;
+ uint64_t LaneBlendMask = 0;
+ for (int LaneElt = 0; LaneElt != NumEltsPerLane; ++LaneElt) {
+ int Elt = (Lane * NumEltsPerLane) + LaneElt;
+ int M = Mask[Elt];
+ if (M == SM_SentinelUndef)
+ continue;
+ if (M == Elt || (0 <= M && M < NumElts &&
+ IsElementEquivalent(NumElts, V1, V1, M, Elt))) {
+ Mask[Elt] = Elt;
+ LaneV1InUse = true;
continue;
}
- if (V2IsZeroOrUndef) {
- ForceV2Zero = true;
- BlendMask |= 1ull << i;
- Mask[i] = i + Size;
+ if (M == (Elt + NumElts) ||
+ (NumElts <= M &&
+ IsElementEquivalent(NumElts, V2, V2, M - NumElts, Elt))) {
+ LaneBlendMask |= 1ull << LaneElt;
+ Mask[Elt] = Elt + NumElts;
+ LaneV2InUse = true;
continue;
}
+ if (Zeroable[Elt]) {
+ if (V1IsZeroOrUndef) {
+ ForceV1Zero = true;
+ Mask[Elt] = Elt;
+ LaneV1InUse = true;
+ continue;
+ }
+ if (V2IsZeroOrUndef) {
+ ForceV2Zero = true;
+ LaneBlendMask |= 1ull << LaneElt;
+ Mask[Elt] = Elt + NumElts;
+ LaneV2InUse = true;
+ continue;
+ }
+ }
+ return false;
}
- return false;
+
+ // If we only used V2 then splat the lane blend mask to avoid any demanded
+ // elts from V1 in this lane (the V1 equivalent is implicit with a zero
+ // blend mask bit).
+ if (ForceWholeLaneMasks && LaneV2InUse && !LaneV1InUse)
+ LaneBlendMask = (1ull << NumEltsPerLane) - 1;
+
+ BlendMask |= LaneBlendMask << (Lane * NumEltsPerLane);
}
return true;
}
@@ -13013,7 +13043,7 @@ static SDValue lowerShuffleAsBlend(const SDLoc &DL, MVT VT, SDValue V1,
uint64_t BlendMask = 0;
bool ForceV1Zero = false, ForceV2Zero = false;
SmallVector<int, 64> Mask(Original);
- if (!matchShuffleAsBlend(V1, V2, Mask, Zeroable, ForceV1Zero, ForceV2Zero,
+ if (!matchShuffleAsBlend(VT, V1, V2, Mask, Zeroable, ForceV1Zero, ForceV2Zero,
BlendMask))
return SDValue();
@@ -38722,7 +38752,7 @@ static bool matchBinaryPermuteShuffle(
uint64_t BlendMask = 0;
bool ForceV1Zero = false, ForceV2Zero = false;
SmallVector<int, 8> TargetMask(Mask);
- if (matchShuffleAsBlend(V1, V2, TargetMask, Zeroable, ForceV1Zero,
+ if (matchShuffleAsBlend(MaskVT, V1, V2, TargetMask, Zeroable, ForceV1Zero,
ForceV2Zero, BlendMask)) {
if (MaskVT == MVT::v16i16) {
// We can only use v16i16 PBLENDW if the lanes are repeated.
diff --git a/llvm/test/CodeGen/X86/any_extend_vector_inreg_of_broadcast.ll b/llvm/test/CodeGen/X86/any_extend_vector_inreg_of_broadcast.ll
index 5ed68056069e5..bb2e790362e48 100644
--- a/llvm/test/CodeGen/X86/any_extend_vector_inreg_of_broadcast.ll
+++ b/llvm/test/CodeGen/X86/any_extend_vector_inreg_of_broadcast.ll
@@ -4398,7 +4398,7 @@ define void @vec384_i32_widen_to_i96_factor3_broadcast_to_v4i96_factor4(ptr %in.
; AVX2-NEXT: vmovdqa 32(%rdi), %ymm1
; AVX2-NEXT: vpaddb (%rsi), %ymm0, %ymm0
; AVX2-NEXT: vpaddb 32(%rsi), %ymm1, %ymm1
-; AVX2-NEXT: vpblendd {{.*#+}} ymm1 = ymm0[0,1],ymm1[2,3,4,5,6,7]
+; AVX2-NEXT: vpblendd {{.*#+}} ymm1 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-NEXT: vmovdqa {{.*#+}} xmm2 = [0,5,6,0]
; AVX2-NEXT: vpermd %ymm1, %ymm2, %ymm1
; AVX2-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
@@ -4518,20 +4518,17 @@ define void @vec384_i32_widen_to_i128_factor4_broadcast_to_v3i128_factor3(ptr %i
; AVX-NEXT: vmovdqa 16(%rdi), %xmm1
; AVX-NEXT: vmovdqa 48(%rdi), %xmm2
; AVX-NEXT: vpaddb 16(%rsi), %xmm1, %xmm1
-; AVX-NEXT: vpaddb 48(%rsi), %xmm2, %xmm2
; AVX-NEXT: vpaddb (%rsi), %xmm0, %xmm0
-; AVX-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm3
-; AVX-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0],ymm2[1,2,3],ymm3[4],ymm2[5,6,7]
-; AVX-NEXT: vextractf128 $1, %ymm2, %xmm3
-; AVX-NEXT: vpaddb 16(%rdx), %xmm3, %xmm3
+; AVX-NEXT: vpaddb 48(%rsi), %xmm2, %xmm2
+; AVX-NEXT: vpblendw {{.*#+}} xmm2 = xmm0[0,1],xmm2[2,3,4,5,6,7]
; AVX-NEXT: vpaddb (%rdx), %xmm2, %xmm2
; AVX-NEXT: vpaddb 48(%rdx), %xmm1, %xmm1
-; AVX-NEXT: vpaddb 32(%rdx), %xmm0, %xmm0
-; AVX-NEXT: vmovdqa %xmm0, 32(%rcx)
+; AVX-NEXT: vpaddb 32(%rdx), %xmm0, %xmm3
+; AVX-NEXT: vpaddb 16(%rdx), %xmm0, %xmm0
+; AVX-NEXT: vmovdqa %xmm0, 16(%rcx)
+; AVX-NEXT: vmovdqa %xmm3, 32(%rcx)
; AVX-NEXT: vmovdqa %xmm1, 48(%rcx)
; AVX-NEXT: vmovdqa %xmm2, (%rcx)
-; AVX-NEXT: vmovdqa %xmm3, 16(%rcx)
-; AVX-NEXT: vzeroupper
; AVX-NEXT: retq
;
; AVX2-LABEL: vec384_i32_widen_to_i128_factor4_broadcast_to_v3i128_factor3:
@@ -4541,7 +4538,7 @@ define void @vec384_i32_widen_to_i128_factor4_broadcast_to_v3i128_factor3(ptr %i
; AVX2-NEXT: vmovdqa 48(%rdi), %xmm1
; AVX2-NEXT: vpaddb 48(%rsi), %xmm1, %xmm1
; AVX2-NEXT: vpermq {{.*#+}} ymm2 = ymm0[0,1,0,1]
-; AVX2-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0],ymm1[1,2,3],ymm2[4],ymm1[5,6,7]
+; AVX2-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0],ymm1[1,2,3],ymm2[4,5,6,7]
; AVX2-NEXT: vpaddb (%rdx), %ymm1, %ymm1
; AVX2-NEXT: vpaddb 32(%rdx), %ymm0, %ymm0
; AVX2-NEXT: vmovdqa %ymm0, 32(%rcx)
@@ -4586,8 +4583,8 @@ define void @vec384_i32_widen_to_i128_factor4_broadcast_to_v3i128_factor3(ptr %i
; AVX512BW-LABEL: vec384_i32_widen_to_i128_factor4_broadcast_to_v3i128_factor3:
; AVX512BW: # %bb.0:
; AVX512BW-NEXT: vmovdqa64 (%rdi), %zmm0
+; AVX512BW-NEXT: vmovdqa {{.*#+}} ymm1 = [0,13,14,15,0,1,2,3]
; AVX512BW-NEXT: vpaddb (%rsi), %zmm0, %zmm0
-; AVX512BW-NEXT: vmovdqa {{.*#+}} xmm1 = [0,13,14,15]
; AVX512BW-NEXT: vpermd %zmm0, %zmm1, %zmm1
; AVX512BW-NEXT: vinserti64x4 $1, %ymm0, %zmm1, %zmm0
; AVX512BW-NEXT: vpaddb (%rdx), %zmm0, %zmm0
@@ -4656,7 +4653,7 @@ define void @vec384_i32_widen_to_i192_factor6_broadcast_to_v2i192_factor2(ptr %i
; AVX2-NEXT: vmovdqa 32(%rdi), %ymm1
; AVX2-NEXT: vpaddb (%rsi), %ymm0, %ymm0
; AVX2-NEXT: vpaddb 32(%rsi), %ymm1, %ymm1
-; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4,5,6,7]
+; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-NEXT: vmovdqa {{.*#+}} xmm1 = [0,5,6,7]
; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
; AVX2-NEXT: vpaddb (%rdx), %ymm0, %ymm0
@@ -4706,8 +4703,8 @@ define void @vec384_i32_widen_to_i192_factor6_broadcast_to_v2i192_factor2(ptr %i
; AVX512BW-FAST-LABEL: vec384_i32_widen_to_i192_factor6_broadcast_to_v2i192_factor2:
; AVX512BW-FAST: # %bb.0:
; AVX512BW-FAST-NEXT: vmovdqa64 (%rdi), %zmm0
+; AVX512BW-FAST-NEXT: vmovdqa {{.*#+}} ymm1 = [0,13,14,15,0,1,2,3]
; AVX512BW-FAST-NEXT: vpaddb (%rsi), %zmm0, %zmm0
-; AVX512BW-FAST-NEXT: vmovdqa {{.*#+}} xmm1 = [0,13,14,15]
; AVX512BW-FAST-NEXT: vpermd %zmm0, %zmm1, %zmm0
; AVX512BW-FAST-NEXT: vpermq {{.*#+}} ymm0 = ymm0[0,1,2,2]
; AVX512BW-FAST-NEXT: vpaddb (%rdx), %zmm0, %zmm0
diff --git a/llvm/test/CodeGen/X86/any_extend_vector_inreg_of_broadcast_from_memory.ll b/llvm/test/CodeGen/X86/any_extend_vector_inreg_of_broadcast_from_memory.ll
index ee53111c7a895..6037b9a68c056 100644
--- a/llvm/test/CodeGen/X86/any_extend_vector_inreg_of_broadcast_from_memory.ll
+++ b/llvm/test/CodeGen/X86/any_extend_vector_inreg_of_broadcast_from_memory.ll
@@ -3569,7 +3569,7 @@ define void @vec384_i32_widen_to_i96_factor3_broadcast_to_v4i96_factor4(ptr %in.
; AVX2: # %bb.0:
; AVX2-NEXT: vmovdqa 32(%rdi), %ymm0
; AVX2-NEXT: vpshufd {{.*#+}} xmm1 = mem[0,0,1,1]
-; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = mem[0,1],ymm0[2,3,4,5,6,7]
+; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = mem[0,1,2,3],ymm0[4,5,6,7]
; AVX2-NEXT: vmovdqa {{.*#+}} xmm2 = [0,5,6,0]
; AVX2-NEXT: vpermd %ymm0, %ymm2, %ymm0
; AVX2-NEXT: vpaddb (%rsi), %ymm0, %ymm0
@@ -3673,7 +3673,7 @@ define void @vec384_i32_widen_to_i128_factor4_broadcast_to_v3i128_factor3(ptr %i
; AVX2-NEXT: vmovdqa (%rdi), %ymm0
; AVX2-NEXT: vmovdqa 48(%rdi), %xmm1
; AVX2-NEXT: vpermq {{.*#+}} ymm2 = ymm0[0,1,0,1]
-; AVX2-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0],ymm1[1,2,3],ymm2[4],ymm1[5,6,7]
+; AVX2-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0],ymm1[1,2,3],ymm2[4,5,6,7]
; AVX2-NEXT: vpaddb (%rsi), %ymm1, %ymm1
; AVX2-NEXT: vpaddb 32(%rsi), %ymm0, %ymm0
; AVX2-NEXT: vmovdqa %ymm0, 32(%rdx)
@@ -3763,7 +3763,7 @@ define void @vec384_i32_widen_to_i192_factor6_broadcast_to_v2i192_factor2(ptr %i
; AVX2-LABEL: vec384_i32_widen_to_i192_factor6_broadcast_to_v2i192_factor2:
; AVX2: # %bb.0:
; AVX2-NEXT: vmovdqa 32(%rdi), %ymm0
-; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = mem[0,1],ymm0[2,3,4,5,6,7]
+; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = mem[0,1,2,3],ymm0[4,5,6,7]
; AVX2-NEXT: vmovdqa {{.*#+}} xmm1 = [0,5,6,7]
; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
; AVX2-NEXT: vpaddb (%rsi), %ymm0, %ymm0
diff --git a/llvm/test/CodeGen/X86/avx512-shuffles/partial_permute.ll b/llvm/test/CodeGen/X86/avx512-shuffles/partial_permute.ll
index fb2210366146d..2d4c2df704173 100644
--- a/llvm/test/CodeGen/X86/avx512-shuffles/partial_permute.ll
+++ b/llvm/test/CodeGen/X86/avx512-shuffles/partial_permute.ll
@@ -2045,7 +2045,7 @@ define <4 x i64> @test_8xi64_to_4xi64_perm_mask3(<8 x i64> %vec) {
; CHECK-FAST-PERLANE-LABEL: test_8xi64_to_4xi64_perm_mask3:
; CHECK-FAST-PERLANE: # %bb.0:
; CHECK-FAST-PERLANE-NEXT: vextractf64x4 $1, %zmm0, %ymm1
-; CHECK-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4,5,6,7]
+; CHECK-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; CHECK-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[2,0,0,3]
; CHECK-FAST-PERLANE-NEXT: retq
%res = shufflevector <8 x i64> %vec, <8 x i64> undef, <4 x i32> <i32 6, i32 0, i32 0, i32 7>
@@ -2063,7 +2063,7 @@ define <4 x i64> @test_masked_8xi64_to_4xi64_perm_mask3(<8 x i64> %vec, <4 x i64
; CHECK-FAST-PERLANE-LABEL: test_masked_8xi64_to_4xi64_perm_mask3:
; CHECK-FAST-PERLANE: # %bb.0:
; CHECK-FAST-PERLANE-NEXT: vextracti64x4 $1, %zmm0, %ymm3
-; CHECK-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1],ymm3[2,3,4,5,6,7]
+; CHECK-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5,6,7]
; CHECK-FAST-PERLANE-NEXT: vptestnmq %ymm2, %ymm2, %k1
; CHECK-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm1 {%k1} = ymm0[2,0,0,3]
; CHECK-FAST-PERLANE-NEXT: vmovdqa %ymm1, %ymm0
@@ -2086,7 +2086,7 @@ define <4 x i64> @test_masked_z_8xi64_to_4xi64_perm_mask3(<8 x i64> %vec, <4 x i
; CHECK-FAST-PERLANE-LABEL: test_masked_z_8xi64_to_4xi64_perm_mask3:
; CHECK-FAST-PERLANE: # %bb.0:
; CHECK-FAST-PERLANE-NEXT: vextracti64x4 $1, %zmm0, %ymm2
-; CHECK-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1],ymm2[2,3,4,5,6,7]
+; CHECK-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5,6,7]
; CHECK-FAST-PERLANE-NEXT: vptestnmq %ymm1, %ymm1, %k1
; CHECK-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm0 {%k1} {z} = ymm0[2,0,0,3]
; CHECK-FAST-PERLANE-NEXT: retq
@@ -2370,7 +2370,7 @@ define <4 x i64> @test_masked_8xi64_to_4xi64_perm_mem_mask1(ptr %vp, <4 x i64> %
; CHECK-FAST-PERLANE-LABEL: test_masked_8xi64_to_4xi64_perm_mem_mask1:
; CHECK-FAST-PERLANE: # %bb.0:
; CHECK-FAST-PERLANE-NEXT: vmovdqa 32(%rdi), %ymm2
-; CHECK-FAST-PERLANE-NEXT: vpblendd $3, (%rdi), %ymm2, %ymm2 # ymm2 = mem[0,1],ymm2[2,3,4,5,6,7]
+; CHECK-FAST-PERLANE-NEXT: vpblendd $15, (%rdi), %ymm2, %ymm2 # ymm2 = mem[0,1,2,3],ymm2[4,5,6,7]
; CHECK-FAST-PERLANE-NEXT: vptestnmq %ymm1, %ymm1, %k1
; CHECK-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm0 {%k1} = ymm2[0,3,2,0]
; CHECK-FAST-PERLANE-NEXT: retq
@@ -2394,7 +2394,7 @@ define <4 x i64> @test_masked_z_8xi64_to_4xi64_perm_mem_mask1(ptr %vp, <4 x i64>
; CHECK-FAST-PERLANE-LABEL: test_masked_z_8xi64_to_4xi64_perm_mem_mask1:
; CHECK-FAST-PERLANE: # %bb.0:
; CHECK-FAST-PERLANE-NEXT: vmovdqa 32(%rdi), %ymm1
-; CHECK-FAST-PERLANE-NEXT: vpblendd $3, (%rdi), %ymm1, %ymm1 # ymm1 = mem[0,1],ymm1[2,3,4,5,6,7]
+; CHECK-FAST-PERLANE-NEXT: vpblendd $15, (%rdi), %ymm1, %ymm1 # ymm1 = mem[0,1,2,3],ymm1[4,5,6,7]
; CHECK-FAST-PERLANE-NEXT: vptestnmq %ymm0, %ymm0, %k1
; CHECK-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm0 {%k1} {z} = ymm1[0,3,2,0]
; CHECK-FAST-PERLANE-NEXT: retq
diff --git a/llvm/test/CodeGen/X86/pr38639.ll b/llvm/test/CodeGen/X86/pr38639.ll
index bea6c84279f71..c4a085b0b32a8 100644
--- a/llvm/test/CodeGen/X86/pr38639.ll
+++ b/llvm/test/CodeGen/X86/pr38639.ll
@@ -6,7 +6,7 @@ define <8 x double> @test(<4 x double> %a, <4 x double> %b) {
; CHECK: # %bb.0:
; CHECK-NEXT: vmovaps {{.*#+}} ymm1 = <u,8.2071743224100002E-1,8.2071743224100002E-1,8.2071743224100002E-1>
; CHECK-NEXT: vblendps {{.*#+}} ymm2 = ymm0[0,1,2,3],ymm1[4,5,6,7]
-; CHECK-NEXT: vblendps {{.*#+}} ymm1 = ymm0[0,1],ymm1[2,3],ymm0[4,5,6,7]
+; CHECK-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; CHECK-NEXT: vunpckhpd {{.*#+}} ymm1 = ymm1[1],ymm2[1],ymm1[3],ymm2[3]
; CHECK-NEXT: vmovaps {{.*#+}} xmm2 = [8.2071743224100002E-1,8.2071743224100002E-1]
; CHECK-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5,6,7]
diff --git a/llvm/test/CodeGen/X86/pr51615.ll b/llvm/test/CodeGen/X86/pr51615.ll
index d279deb03d4cc..8b2cea73fd36c 100644
--- a/llvm/test/CodeGen/X86/pr51615.ll
+++ b/llvm/test/CodeGen/X86/pr51615.ll
@@ -13,7 +13,7 @@ define void @volatile_load_2_elts() {
; AVX-NEXT: vmovaps g0(%rip), %xmm0
; AVX-NEXT: vmovddup {{.*#+}} xmm1 = xmm0[0,0]
; AVX-NEXT: vxorps %xmm2, %xmm2, %xmm2
-; AVX-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5],ymm2[6,7]
+; AVX-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3],ymm2[4,5,6,7]
; AVX-NEXT: vinsertf128 $1, %xmm1, %ymm1, %ymm1
; AVX-NEXT: vperm2f128 {{.*#+}} ymm0 = zero,zero,ymm0[0,1]
; AVX-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5],ymm2[6,7]
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-5.ll b/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-5.ll
index 772060af4f1a4..27c81b579dae9 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-5.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-5.ll
@@ -1332,7 +1332,7 @@ define void @load_i16_stride5_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4],ymm0[5,6],ymm1[7]
; AVX2-SLOW-NEXT: vpshufb {{.*#+}} ymm0 = ymm0[u,u,u,u,u,u,u,u,u,u,u,u,4,5,14,15,24,25,18,19,28,29,22,23,u,u,u,u,u,u,u,u]
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm2[0,1,2],ymm0[3,4,5],ymm2[6,7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm2[0,1,2],ymm0[3,4,5,6,7]
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm1 = xmm6[3,1,2,3]
; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} xmm1 = xmm1[0,1,2,1,4,5,6,7]
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm2 = xmm4[0,2,2,3]
@@ -1424,7 +1424,7 @@ define void @load_i16_stride5_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm2 = <1,3,u,u,6,0,3,5>
; AVX2-FAST-NEXT: vpermd %ymm1, %ymm2, %ymm1
; AVX2-FAST-NEXT: vpshufb {{.*#+}} ymm1 = ymm1[u,u,u,u,u,u,u,u,u,u,u,u,0,1,6,7,16,17,22,23,24,25,30,31,u,u,u,u,u,u,u,u]
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm3[0,1,2],ymm1[3,4,5],ymm3[6,7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm3[0,1,2],ymm1[3,4,5,6,7]
; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm2 = <u,u,u,u,0,2,5,7>
; AVX2-FAST-NEXT: vpermd %ymm0, %ymm2, %ymm0
; AVX2-FAST-NEXT: vpshufb {{.*#+}} ymm0 = ymm0[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,16,17,22,23,24,25,30,31]
@@ -1510,7 +1510,7 @@ define void @load_i16_stride5_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4],ymm0[5,6],ymm1[7]
; AVX2-FAST-PERLANE-NEXT: vpshufb {{.*#+}} ymm0 = ymm0[u,u,u,u,u,u,u,u,u,u,u,u,4,5,14,15,24,25,18,19,28,29,22,23,u,u,u,u,u,u,u,u]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm2[0,1,2],ymm0[3,4,5],ymm2[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm2[0,1,2],ymm0[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpshufb {{.*#+}} xmm1 = xmm4[12,13,14,15,4,5,14,15,u,u,u,u,u,u,u,u]
; AVX2-FAST-PERLANE-NEXT: vpshufb {{.*#+}} xmm2 = xmm5[0,1,2,3,0,1,10,11,u,u,u,u,u,u,u,u]
; AVX2-FAST-PERLANE-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
@@ -1596,7 +1596,7 @@ define void @load_i16_stride5_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX512F-SLOW-NEXT: vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4],ymm0[5,6],ymm1[7]
; AVX512F-SLOW-NEXT: vpshufb {{.*#+}} ymm0 = ymm0[u,u,u,u,u,u,u,u,u,u,u,u,4,5,14,15,24,25,18,19,28,29,22,23,u,u,u,u,u,u,u,u]
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm2[0,1,2],ymm0[3,4,5],ymm2[6,7]
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm2[0,1,2],ymm0[3,4,5,6,7]
; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} xmm1 = xmm6[3,1,2,3]
; AVX512F-SLOW-NEXT: vpshuflw {{.*#+}} xmm1 = xmm1[0,1,2,1,4,5,6,7]
; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} xmm2 = xmm4[0,2,2,3]
@@ -1687,7 +1687,7 @@ define void @load_i16_stride5_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm1 = <1,3,u,u,6,0,3,5>
; AVX512F-FAST-NEXT: vpermd %ymm0, %ymm1, %ymm0
; AVX512F-FAST-NEXT: vpshufb {{.*#+}} ymm0 = ymm0[u,u,u,u,u,u,u,u,u,u,u,u,0,1,6,7,16,17,22,23,24,25,30,31,u,u,u,u,u,u,u,u]
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm2[0,1,2],ymm0[3,4,5],ymm2[6,7]
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm2[0,1,2],ymm0[3,4,5,6,7]
; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm1 = <u,u,u,u,0,2,5,7>
; AVX512F-FAST-NEXT: vpermd %ymm4, %ymm1, %ymm1
; AVX512F-FAST-NEXT: vpshufb {{.*#+}} ymm1 = ymm1[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,16,17,22,23,u,u,u,u,24,25,30,31,u,u,u,u]
@@ -2735,7 +2735,7 @@ define void @load_i16_stride5_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-SLOW-NEXT: vpshufb %ymm5, %ymm2, %ymm2
; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} xmm6 = [8,9,2,3,12,13,6,7,0,1,10,11,0,1,6,7]
; AVX2-SLOW-NEXT: vpshufb %xmm6, %xmm4, %xmm4
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm4[0,1,2],ymm2[3,4,5],ymm4[6,7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm4[0,1,2],ymm2[3,4,5,6,7]
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm1 = xmm15[3,1,2,3]
; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} xmm1 = xmm1[0,1,2,1,4,5,6,7]
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm4 = xmm11[0,2,2,3]
@@ -2753,7 +2753,7 @@ define void @load_i16_stride5_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-SLOW-NEXT: vextracti128 $1, %ymm3, %xmm4
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm3 = xmm4[0,1,2],xmm3[3,4],xmm4[5,6,7]
; AVX2-SLOW-NEXT: vpshufb %xmm6, %xmm3, %xmm3
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3,4,5],ymm3[6,7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3,4,5,6,7]
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm3 = xmm13[3,1,2,3]
; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} xmm3 = xmm3[0,1,2,1,4,5,6,7]
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm4 = xmm12[0,2,2,3]
@@ -2953,7 +2953,7 @@ define void @load_i16_stride5_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-NEXT: vpshufb %ymm4, %ymm0, %ymm0
; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm5 = [8,9,2,3,12,13,6,7,0,1,10,11,0,1,6,7]
; AVX2-FAST-NEXT: vpshufb %xmm5, %xmm1, %xmm1
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5],ymm1[6,7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm1 = [0,2,1,3,0,2,5,7]
; AVX2-FAST-NEXT: vpermd %ymm13, %ymm1, %ymm6
; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm7 = <u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,16,17,22,23,24,25,30,31>
@@ -2967,7 +2967,7 @@ define void @load_i16_stride5_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm5 = ymm12[0,1],ymm14[2],ymm12[3],ymm14[4],ymm12[5,6],ymm14[7],ymm12[8,9],ymm14[10],ymm12[11],ymm14[12],ymm12[13,14],ymm14[15]
; AVX2-FAST-NEXT: vpermd %ymm5, %ymm2, %ymm2
; AVX2-FAST-NEXT: vpshufb %ymm4, %ymm2, %ymm2
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3,4,5],ymm3[6,7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3,4,5,6,7]
; AVX2-FAST-NEXT: vpermd %ymm11, %ymm1, %ymm1
; AVX2-FAST-NEXT: vpshufb %ymm7, %ymm1, %ymm1
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
@@ -3172,7 +3172,7 @@ define void @load_i16_stride5_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm5, %ymm1, %ymm1
; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm6 = [8,9,2,3,12,13,6,7,0,1,10,11,0,1,6,7]
; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm6, %xmm4, %xmm4
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm4[0,1,2],ymm1[3,4,5],ymm4[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm4[0,1,2],ymm1[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
@@ -3187,7 +3187,7 @@ define void @load_i16_stride5_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm3, %xmm4
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm3 = xmm4[0,1,2],xmm3[3,4],xmm4[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm6, %xmm3, %xmm3
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm3[0,1,2],ymm1[3,4,5],ymm3[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm3[0,1,2],ymm1[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm2, %xmm8, %xmm2
; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm13, %xmm9, %xmm3
; AVX2-FAST-PERLANE-NEXT: vpunpckldq {{.*#+}} xmm2 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
@@ -3370,8 +3370,8 @@ define void @load_i16_stride5_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm2 = ymm3[0,1],ymm2[2],ymm3[3],ymm2[4],ymm3[5,6],ymm2[7],ymm3[8,9],ymm2[10],ymm3[11],ymm2[12],ymm3[13,14],ymm2[15]
; AVX512F-SLOW-NEXT: vpermq {{.*#+}} ymm3 = ymm2[2,3,0,1]
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4],ymm2[5,6],ymm3[7]
-; AVX512F-SLOW-NEXT: vpshufb {{.*#+}} ymm2 = ymm2[u,u,u,u,u,u,u,u,u,u,u,u,4,5,14,15,24,25,18,19,28,29,22,23,u,u,u,u,u,u,u,u]
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm5[0,1,2],ymm2[3,4,5],ymm5[6,7]
+; AVX512F-SLOW-NEXT: vpshufb {{.*#+}} ymm2 = ymm2[u,u,u,u,u,u,u,u,u,u,u,u,4,5,14,15,24,25,18,19,28,29,22,23,20,21,22,23,20,21,30,31]
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm5[0,1,2],ymm2[3,4,5,6,7]
; AVX512F-SLOW-NEXT: movb $7, %al
; AVX512F-SLOW-NEXT: kmovw %eax, %k1
; AVX512F-SLOW-NEXT: vinserti64x4 $0, %ymm2, %zmm0, %zmm7 {%k1}
@@ -3536,8 +3536,8 @@ define void @load_i16_stride5_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm8 = ymm9[0,1],ymm8[2],ymm9[3],ymm8[4],ymm9[5,6],ymm8[7],ymm9[8,9],ymm8[10],ymm9[11],ymm8[12],ymm9[13,14],ymm8[15]
; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} ymm9 = <1,3,u,u,6,0,3,5>
; AVX512F-FAST-NEXT: vpermd %ymm8, %ymm9, %ymm8
-; AVX512F-FAST-NEXT: vpshufb {{.*#+}} ymm8 = ymm8[u,u,u,u,u,u,u,u,u,u,u,u,0,1,6,7,16,17,22,23,24,25,30,31,u,u,u,u,u,u,u,u]
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm8 = ymm10[0,1,2],ymm8[3,4,5],ymm10[6,7]
+; AVX512F-FAST-NEXT: vpshufb {{.*#+}} ymm8 = ymm8[u,u,u,u,u,u,u,u,u,u,u,u,0,1,6,7,16,17,22,23,24,25,30,31,16,17,22,23,16,17,22,23]
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm8 = ymm10[0,1,2],ymm8[3,4,5,6,7]
; AVX512F-FAST-NEXT: vpblendd {{.*#+}} xmm2 = xmm2[0,1],xmm3[2],xmm2[3]
; AVX512F-FAST-NEXT: vpshufb {{.*#+}} xmm2 = xmm2[8,9,2,3,12,13,u,u,u,u,u,u,u,u,u,u]
; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm3 = ymm4[0],ymm5[1,2],ymm4[3],ymm5[4],ymm4[5],ymm5[6,7],ymm4[8],ymm5[9,10],ymm4[11],ymm5[12],ymm4[13],ymm5[14,15]
@@ -5708,7 +5708,7 @@ define void @load_i16_stride5_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-SLOW-NEXT: vpshufb %ymm8, %ymm0, %ymm0
; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} xmm3 = [8,9,2,3,12,13,6,7,0,1,10,11,0,1,6,7]
; AVX2-SLOW-NEXT: vpshufb %xmm3, %xmm1, %xmm1
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5],ymm1[6,7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm1 = xmm5[3,1,2,3]
; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} xmm1 = xmm1[0,1,2,1,4,5,6,7]
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm2 = xmm6[0,2,2,3]
@@ -5727,7 +5727,7 @@ define void @load_i16_stride5_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm2 = xmm4[0,1,2],xmm2[3,4],xmm4[5,6,7]
; AVX2-SLOW-NEXT: vpshufb %ymm8, %ymm1, %ymm1
; AVX2-SLOW-NEXT: vpshufb %xmm3, %xmm2, %xmm2
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4,5],ymm2[6,7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4,5,6,7]
; AVX2-SLOW-NEXT: vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
; AVX2-SLOW-NEXT: # xmm2 = mem[3,1,2,3]
; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[0,1,2,1,4,5,6,7]
@@ -5747,7 +5747,7 @@ define void @load_i16_stride5_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm4 = xmm5[0,1,2],xmm4[3,4],xmm5[5,6,7]
; AVX2-SLOW-NEXT: vpshufb %ymm8, %ymm2, %ymm2
; AVX2-SLOW-NEXT: vpshufb %xmm3, %xmm4, %xmm4
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm4[0,1,2],ymm2[3,4,5],ymm4[6,7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm4[0,1,2],ymm2[3,4,5,6,7]
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm4 = xmm13[3,1,2,3]
; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} xmm4 = xmm4[0,1,2,1,4,5,6,7]
; AVX2-SLOW-NEXT: vpshufd $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
@@ -5766,7 +5766,7 @@ define void @load_i16_stride5_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-SLOW-NEXT: vextracti128 $1, %ymm5, %xmm6
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm5 = xmm6[0,1,2],xmm5[3,4],xmm6[5,6,7]
; AVX2-SLOW-NEXT: vpshufb %xmm3, %xmm5, %xmm3
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2],ymm4[3,4,5],ymm3[6,7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2],ymm4[3,4,5,6,7]
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm4 = xmm11[3,1,2,3]
; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} xmm4 = xmm4[0,1,2,1,4,5,6,7]
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm5 = xmm7[0,2,2,3]
@@ -6145,7 +6145,7 @@ define void @load_i16_stride5_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-NEXT: vpermd %ymm2, %ymm7, %ymm2
; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm13 = <u,u,u,u,u,u,u,u,u,u,u,u,0,1,6,7,16,17,22,23,24,25,30,31,u,u,u,u,u,u,u,u>
; AVX2-FAST-NEXT: vpshufb %ymm13, %ymm2, %ymm2
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5],ymm1[6,7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm6 = [0,2,1,3,0,2,5,7]
; AVX2-FAST-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm2 # 32-byte Folded Reload
; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm0 = <u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,16,17,22,23,24,25,30,31>
@@ -6159,7 +6159,7 @@ define void @load_i16_stride5_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-NEXT: vpshufb %xmm12, %xmm2, %xmm2
; AVX2-FAST-NEXT: vpermd %ymm11, %ymm7, %ymm11
; AVX2-FAST-NEXT: vpshufb %ymm13, %ymm11, %ymm11
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2],ymm11[3,4,5],ymm2[6,7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2],ymm11[3,4,5,6,7]
; AVX2-FAST-NEXT: vpermd %ymm3, %ymm6, %ymm3
; AVX2-FAST-NEXT: vpshufb %ymm0, %ymm3, %ymm3
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm3[6,7]
@@ -6171,7 +6171,7 @@ define void @load_i16_stride5_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-NEXT: vpshufb %xmm12, %xmm3, %xmm3
; AVX2-FAST-NEXT: vpermd %ymm11, %ymm7, %ymm11
; AVX2-FAST-NEXT: vpshufb %ymm13, %ymm11, %ymm11
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2],ymm11[3,4,5],ymm3[6,7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2],ymm11[3,4,5,6,7]
; AVX2-FAST-NEXT: vpermd %ymm8, %ymm6, %ymm8
; AVX2-FAST-NEXT: vpshufb %ymm0, %ymm8, %ymm8
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm8[6,7]
@@ -6186,7 +6186,7 @@ define void @load_i16_stride5_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-NEXT: vextracti128 $1, %ymm8, %xmm10
; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm8 = xmm10[0,1,2],xmm8[3,4],xmm10[5,6,7]
; AVX2-FAST-NEXT: vpshufb %xmm12, %xmm8, %xmm4
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm4 = ymm4[0,1,2],ymm7[3,4,5],ymm4[6,7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm4 = ymm4[0,1,2],ymm7[3,4,5,6,7]
; AVX2-FAST-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
; AVX2-FAST-NEXT: vpshufb %ymm0, %ymm6, %ymm5
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5],ymm5[6,7]
@@ -6608,7 +6608,7 @@ define void @load_i16_stride5_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm3, %ymm0, %ymm0
; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm9 = [8,9,2,3,12,13,6,7,0,1,10,11,0,1,6,7]
; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm9, %xmm1, %xmm1
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5],ymm1[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm11 = [12,13,14,15,4,5,14,15,8,9,10,11,12,13,14,15]
; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm11, %xmm5, %xmm1
; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm2 = [0,1,2,3,0,1,10,11,8,9,10,11,12,13,14,15]
@@ -6628,7 +6628,7 @@ define void @load_i16_stride5_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm4 = xmm5[0,1,2],xmm4[3,4],xmm5[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm3, %ymm1, %ymm1
; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm9, %xmm4, %xmm4
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm4[0,1,2],ymm1[3,4,5],ymm4[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm4[0,1,2],ymm1[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm11, %xmm4, %xmm4
; AVX2-FAST-PERLANE-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
@@ -6646,7 +6646,7 @@ define void @load_i16_stride5_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm5 = xmm6[0,1,2],xmm5[3,4],xmm6[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm3, %ymm4, %ymm4
; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm9, %xmm5, %xmm5
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm4 = ymm5[0,1,2],ymm4[3,4,5],ymm5[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm4 = ymm5[0,1,2],ymm4[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm11, %xmm13, %xmm5
; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm2, %xmm12, %xmm6
; AVX2-FAST-PERLANE-NEXT: vpunpckldq {{.*#+}} xmm5 = xmm6[0],xmm5[0],xmm6[1],xmm5[1]
@@ -6666,7 +6666,7 @@ define void @load_i16_stride5_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm5, %xmm6
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm5 = xmm6[0,1,2],xmm5[3,4],xmm6[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm9, %xmm5, %xmm5
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm3 = ymm5[0,1,2],ymm3[3,4,5],ymm5[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm3 = ymm5[0,1,2],ymm3[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm2[6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
@@ -7090,7 +7090,7 @@ define void @load_i16_stride5_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX512F-SLOW-NEXT: vpshufb %ymm10, %ymm3, %ymm3
; AVX512F-SLOW-NEXT: vmovdqa {{.*#+}} xmm11 = [8,9,2,3,12,13,6,7,0,1,10,11,0,1,6,7]
; AVX512F-SLOW-NEXT: vpshufb %xmm11, %xmm9, %xmm9
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm9[0,1,2],ymm3[3,4,5],ymm9[6,7]
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm9[0,1,2],ymm3[3,4,5,6,7]
; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm5 = ymm5[0],ymm4[1,2],ymm5[3],ymm4[4],ymm5[5],ymm4[6,7],ymm5[8],ymm4[9,10],ymm5[11],ymm4[12],ymm5[13],ymm4[14,15]
; AVX512F-SLOW-NEXT: vextracti128 $1, %ymm5, %xmm6
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm5 = ymm6[0],ymm5[1],ymm6[2],ymm5[3,4,5,6,7]
@@ -7142,7 +7142,7 @@ define void @load_i16_stride5_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX512F-SLOW-NEXT: vextracti128 $1, %ymm6, %xmm7
; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} xmm6 = xmm7[0,1,2],xmm6[3,4],xmm7[5,6,7]
; AVX512F-SLOW-NEXT: vpshufb %xmm11, %xmm6, %xmm6
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm6[0,1,2],ymm4[3,4,5],ymm6[6,7]
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm6[0,1,2],ymm4[3,4,5,6,7]
; AVX512F-SLOW-NEXT: vinserti64x4 $0, %ymm4, %zmm0, %zmm3 {%k1}
; AVX512F-SLOW-NEXT: vmovdqa64 %ymm27, %ymm0
; AVX512F-SLOW-NEXT: vmovdqa64 %ymm28, %ymm4
@@ -7572,7 +7572,7 @@ define void @load_i16_stride5_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX512F-FAST-NEXT: vpblendw {{.*#+}} xmm5 = xmm7[0,1,2],xmm5[3,4],xmm7[5,6,7]
; AVX512F-FAST-NEXT: vmovdqa {{.*#+}} xmm7 = [8,9,2,3,12,13,6,7,0,1,10,11,0,1,6,7]
; AVX512F-FAST-NEXT: vpshufb %xmm7, %xmm5, %xmm5
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm5[0,1,2],ymm2[3,4,5],ymm5[6,7]
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm5[0,1,2],ymm2[3,4,5,6,7]
; AVX512F-FAST-NEXT: movb $7, %al
; AVX512F-FAST-NEXT: kmovw %eax, %k1
; AVX512F-FAST-NEXT: vinserti64x4 $0, %ymm2, %zmm0, %zmm6 {%k1}
@@ -7595,7 +7595,7 @@ define void @load_i16_stride5_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm6 = ymm15[0,1],ymm3[2],ymm15[3],ymm3[4],ymm15[5,6],ymm3[7],ymm15[8,9],ymm3[10],ymm15[11],ymm3[12],ymm15[13,14],ymm3[15]
; AVX512F-FAST-NEXT: vpermd %ymm6, %ymm14, %ymm3
; AVX512F-FAST-NEXT: vpshufb %ymm4, %ymm3, %ymm3
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm5[0,1,2],ymm3[3,4,5],ymm5[6,7]
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm5[0,1,2],ymm3[3,4,5,6,7]
; AVX512F-FAST-NEXT: vinserti64x4 $0, %ymm3, %zmm0, %zmm0 {%k1}
; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm3 = ymm11[0],ymm9[1],ymm11[2],ymm9[3],ymm11[4,5],ymm9[6],ymm11[7,8],ymm9[9],ymm11[10],ymm9[11],ymm11[12,13],ymm9[14],ymm11[15]
; AVX512F-FAST-NEXT: vextracti128 $1, %ymm3, %xmm4
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-6.ll b/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-6.ll
index 49383d8aba42f..c32461f762fce 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-6.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-6.ll
@@ -1412,7 +1412,7 @@ define void @load_i16_stride6_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm11 = xmm2[0,1,2,3,4,6,6,7]
; AVX1-ONLY-NEXT: vpunpckhdq {{.*#+}} xmm11 = xmm11[2],xmm6[2],xmm11[3],xmm6[3]
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm10 = xmm11[0,1,2],xmm10[3,4,5],xmm11[6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm10[0,1,2],ymm5[3,4,5],ymm10[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm10[0,1,2],ymm5[3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovdqa 176(%rdi), %xmm10
; AVX1-ONLY-NEXT: vpslld $16, %xmm10, %xmm5
; AVX1-ONLY-NEXT: vmovdqa 160(%rdi), %xmm11
@@ -1449,7 +1449,7 @@ define void @load_i16_stride6_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: vpsrld $16, %xmm6, %xmm9
; AVX1-ONLY-NEXT: vpunpckhdq {{.*#+}} xmm2 = xmm2[2],xmm9[2],xmm2[3],xmm9[3]
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0,1,2],xmm3[3,4,5],xmm2[6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2],ymm0[3,4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2],ymm0[3,4,5,6,7]
; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm1 = xmm1[0,1,1,3,4,5,6,7]
; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm2 = xmm12[1,1,1,1]
; AVX1-ONLY-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
@@ -1609,8 +1609,8 @@ define void @load_i16_stride6_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-SLOW-NEXT: vperm2i128 {{.*#+}} ymm6 = ymm2[2,3],ymm5[2,3]
; AVX2-SLOW-NEXT: vperm2i128 {{.*#+}} ymm7 = ymm2[0,1],ymm5[0,1]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm5 = ymm7[0],ymm6[1],ymm7[2,3,4,5],ymm6[6],ymm7[7]
-; AVX2-SLOW-NEXT: vpshufb {{.*#+}} ymm2 = ymm5[u,u,u,u,u,u,u,u,u,u,u,u,8,9,4,5,16,17,28,29,24,25,28,29,u,u,u,u,u,u,u,u]
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm13[0,1,2],ymm2[3,4,5],ymm13[6,7]
+; AVX2-SLOW-NEXT: vpshufb {{.*#+}} ymm2 = ymm5[u,u,u,u,u,u,u,u,u,u,u,u,8,9,4,5,16,17,28,29,24,25,28,29,24,25,28,29,24,25,20,21]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm13[0,1,2],ymm2[3,4,5,6,7]
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm10 = ymm2[0,1,2],ymm10[3,4,5,6,7],ymm2[8,9,10],ymm10[11,12,13,14,15]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm10[4,5,6,7]
; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} xmm8 = xmm8[0,1,2,3,5,5,5,5]
@@ -1620,8 +1620,8 @@ define void @load_i16_stride6_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-SLOW-NEXT: vpshufb {{.*#+}} xmm9 = xmm12[u,u,u,u,10,11,u,u,2,3,14,15,u,u,u,u]
; AVX2-SLOW-NEXT: vpshufb {{.*#+}} xmm10 = xmm11[2,3,14,15,u,u,6,7,u,u,u,u,12,13,14,15]
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm9 = xmm10[0,1],xmm9[2],xmm10[3],xmm9[4,5],xmm10[6,7]
-; AVX2-SLOW-NEXT: vpshufb {{.*#+}} ymm5 = ymm5[u,u,u,u,u,u,u,u,u,u,u,u,10,11,6,7,18,19,30,31,26,27,26,27,u,u,u,u,u,u,u,u]
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm5 = ymm9[0,1,2],ymm5[3,4,5],ymm9[6,7]
+; AVX2-SLOW-NEXT: vpshufb {{.*#+}} ymm5 = ymm5[u,u,u,u,u,u,u,u,u,u,u,u,10,11,6,7,18,19,30,31,26,27,26,27,30,31,26,27,26,27,22,23]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm5 = ymm9[0,1,2],ymm5[3,4,5,6,7]
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm8 = ymm5[0,1,2],ymm8[3,4,5,6,7],ymm5[8,9,10],ymm8[11,12,13,14,15]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm8[4,5,6,7]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm8 = ymm0[0,1],ymm1[2],ymm0[3,4],ymm1[5],ymm0[6,7]
@@ -1724,8 +1724,8 @@ define void @load_i16_stride6_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-NEXT: vperm2i128 {{.*#+}} ymm6 = ymm2[2,3],ymm5[2,3]
; AVX2-FAST-NEXT: vperm2i128 {{.*#+}} ymm7 = ymm2[0,1],ymm5[0,1]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm5 = ymm7[0],ymm6[1],ymm7[2,3,4,5],ymm6[6],ymm7[7]
-; AVX2-FAST-NEXT: vpshufb {{.*#+}} ymm2 = ymm5[u,u,u,u,u,u,u,u,u,u,u,u,8,9,4,5,16,17,28,29,24,25,28,29,u,u,u,u,u,u,u,u]
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm13[0,1,2],ymm2[3,4,5],ymm13[6,7]
+; AVX2-FAST-NEXT: vpshufb {{.*#+}} ymm2 = ymm5[u,u,u,u,u,u,u,u,u,u,u,u,8,9,4,5,16,17,28,29,24,25,28,29,24,25,28,29,24,25,20,21]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm13[0,1,2],ymm2[3,4,5,6,7]
; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm10 = ymm2[0,1,2],ymm10[3,4,5,6,7],ymm2[8,9,10],ymm10[11,12,13,14,15]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm10[4,5,6,7]
; AVX2-FAST-NEXT: vpshufhw {{.*#+}} xmm8 = xmm8[0,1,2,3,5,5,5,5]
@@ -1735,8 +1735,8 @@ define void @load_i16_stride6_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-NEXT: vpshufb {{.*#+}} xmm9 = xmm11[2,3,14,15,u,u,6,7,u,u,u,u,12,13,14,15]
; AVX2-FAST-NEXT: vpshufb {{.*#+}} xmm10 = xmm12[u,u,u,u,2,3,u,u,10,11,14,15,u,u,u,u]
; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm9 = xmm9[0,1],xmm10[2],xmm9[3],xmm10[4,5],xmm9[6,7]
-; AVX2-FAST-NEXT: vpshufb {{.*#+}} ymm5 = ymm5[u,u,u,u,u,u,u,u,u,u,u,u,10,11,6,7,18,19,30,31,26,27,26,27,u,u,u,u,u,u,u,u]
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm5 = ymm9[0,1,2],ymm5[3,4,5],ymm9[6,7]
+; AVX2-FAST-NEXT: vpshufb {{.*#+}} ymm5 = ymm5[u,u,u,u,u,u,u,u,u,u,u,u,10,11,6,7,18,19,30,31,26,27,26,27,30,31,26,27,26,27,22,23]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm5 = ymm9[0,1,2],ymm5[3,4,5,6,7]
; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm8 = ymm5[0,1,2],ymm8[3,4,5,6,7],ymm5[8,9,10],ymm8[11,12,13,14,15]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm8[4,5,6,7]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm8 = ymm0[0,1],ymm1[2],ymm0[3,4],ymm1[5],ymm0[6,7]
@@ -1832,8 +1832,8 @@ define void @load_i16_stride6_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-PERLANE-NEXT: vperm2i128 {{.*#+}} ymm6 = ymm2[2,3],ymm5[2,3]
; AVX2-FAST-PERLANE-NEXT: vperm2i128 {{.*#+}} ymm7 = ymm2[0,1],ymm5[0,1]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm5 = ymm7[0],ymm6[1],ymm7[2,3,4,5],ymm6[6],ymm7[7]
-; AVX2-FAST-PERLANE-NEXT: vpshufb {{.*#+}} ymm2 = ymm5[u,u,u,u,u,u,u,u,u,u,u,u,8,9,4,5,16,17,28,29,24,25,28,29,u,u,u,u,u,u,u,u]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm13[0,1,2],ymm2[3,4,5],ymm13[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpshufb {{.*#+}} ymm2 = ymm5[u,u,u,u,u,u,u,u,u,u,u,u,8,9,4,5,16,17,28,29,24,25,28,29,24,25,28,29,24,25,20,21]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm13[0,1,2],ymm2[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm10 = ymm2[0,1,2],ymm10[3,4,5,6,7],ymm2[8,9,10],ymm10[11,12,13,14,15]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm10[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpshufhw {{.*#+}} xmm8 = xmm8[0,1,2,3,5,5,5,5]
@@ -1843,8 +1843,8 @@ define void @load_i16_stride6_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-PERLANE-NEXT: vpshufb {{.*#+}} xmm9 = xmm11[2,3,14,15,u,u,6,7,u,u,u,u,12,13,14,15]
; AVX2-FAST-PERLANE-NEXT: vpshufb {{.*#+}} xmm10 = xmm12[u,u,u,u,2,3,u,u,10,11,14,15,u,u,u,u]
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm9 = xmm9[0,1],xmm10[2],xmm9[3],xmm10[4,5],xmm9[6,7]
-; AVX2-FAST-PERLANE-NEXT: vpshufb {{.*#+}} ymm5 = ymm5[u,u,u,u,u,u,u,u,u,u,u,u,10,11,6,7,18,19,30,31,26,27,26,27,u,u,u,u,u,u,u,u]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm5 = ymm9[0,1,2],ymm5[3,4,5],ymm9[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpshufb {{.*#+}} ymm5 = ymm5[u,u,u,u,u,u,u,u,u,u,u,u,10,11,6,7,18,19,30,31,26,27,26,27,30,31,26,27,26,27,22,23]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm5 = ymm9[0,1,2],ymm5[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm8 = ymm5[0,1,2],ymm8[3,4,5,6,7],ymm5[8,9,10],ymm8[11,12,13,14,15]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm8[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm8 = ymm0[0,1],ymm1[2],ymm0[3,4],ymm1[5],ymm0[6,7]
@@ -1934,25 +1934,25 @@ define void @load_i16_stride6_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX512F-SLOW-NEXT: vperm2i128 {{.*#+}} ymm6 = ymm1[2,3],mem[2,3]
; AVX512F-SLOW-NEXT: vinserti128 $1, 96(%rdi), %ymm1, %ymm7
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm10 = ymm7[0],ymm6[1],ymm7[2,3,4,5],ymm6[6],ymm7[7]
-; AVX512F-SLOW-NEXT: vpshufb {{.*#+}} ymm1 = ymm10[u,u,u,u,u,u,u,u,u,u,u,u,8,9,4,5,16,17,28,29,24,25,28,29,u,u,u,u,u,u,u,u]
+; AVX512F-SLOW-NEXT: vpshufb {{.*#+}} ymm1 = ymm10[u,u,u,u,u,u,u,u,u,u,u,u,8,9,4,5,16,17,28,29,24,25,28,29,24,25,28,29,24,25,20,21]
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm11 = ymm3[0],ymm4[1],ymm3[2,3],ymm4[4],ymm3[5,6],ymm4[7]
; AVX512F-SLOW-NEXT: vpshufb {{.*#+}} xmm12 = xmm11[0,1,12,13,u,u,4,5,u,u,u,u,12,13,14,15]
; AVX512F-SLOW-NEXT: vextracti128 $1, %ymm11, %xmm13
; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} xmm14 = xmm13[0,2,0,3]
; AVX512F-SLOW-NEXT: vpshufhw {{.*#+}} xmm14 = xmm14[0,1,2,3,4,6,6,7]
; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} xmm12 = xmm12[0,1],xmm14[2],xmm12[3],xmm14[4,5],xmm12[6,7]
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm12[0,1,2],ymm1[3,4,5],ymm12[6,7]
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm12[0,1,2],ymm1[3,4,5,6,7]
; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm9 = ymm1[0,1,2],ymm9[3,4,5,6,7],ymm1[8,9,10],ymm9[11,12,13,14,15]
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm9[4,5,6,7]
; AVX512F-SLOW-NEXT: vpshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,5,5,5,5]
; AVX512F-SLOW-NEXT: vpshufb {{.*#+}} xmm8 = xmm8[6,7,2,3,4,5,u,u,2,3,14,15,u,u,6,7]
; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} xmm5 = xmm8[0,1,2],xmm5[3],xmm8[4,5],xmm5[6],xmm8[7]
; AVX512F-SLOW-NEXT: vinserti128 $1, %xmm5, %ymm0, %ymm5
-; AVX512F-SLOW-NEXT: vpshufb {{.*#+}} ymm8 = ymm10[u,u,u,u,u,u,u,u,u,u,u,u,10,11,6,7,18,19,30,31,26,27,26,27,u,u,u,u,u,u,u,u]
+; AVX512F-SLOW-NEXT: vpshufb {{.*#+}} ymm8 = ymm10[u,u,u,u,u,u,u,u,u,u,u,u,10,11,6,7,18,19,30,31,26,27,26,27,30,31,26,27,26,27,22,23]
; AVX512F-SLOW-NEXT: vpshufb {{.*#+}} xmm9 = xmm13[u,u,u,u,10,11,u,u,2,3,14,15,u,u,u,u]
; AVX512F-SLOW-NEXT: vpshufb {{.*#+}} xmm10 = xmm11[2,3,14,15,u,u,6,7,u,u,u,u,12,13,14,15]
; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} xmm9 = xmm10[0,1],xmm9[2],xmm10[3],xmm9[4,5],xmm10[6,7]
-; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm8 = ymm9[0,1,2],ymm8[3,4,5],ymm9[6,7]
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm8 = ymm9[0,1,2],ymm8[3,4,5,6,7]
; AVX512F-SLOW-NEXT: vpblendw {{.*#+}} ymm5 = ymm8[0,1,2],ymm5[3,4,5,6,7],ymm8[8,9,10],ymm5[11,12,13,14,15]
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm5 = ymm8[0,1,2,3],ymm5[4,5,6,7]
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} ymm8 = ymm0[0,1],ymm2[2],ymm0[3,4],ymm2[5],ymm0[6,7]
@@ -2046,25 +2046,25 @@ define void @load_i16_stride6_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX512F-FAST-NEXT: vperm2i128 {{.*#+}} ymm6 = ymm2[2,3],mem[2,3]
; AVX512F-FAST-NEXT: vinserti128 $1, 96(%rdi), %ymm2, %ymm7
; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm10 = ymm7[0],ymm6[1],ymm7[2,3,4,5],ymm6[6],ymm7[7]
-; AVX512F-FAST-NEXT: vpshufb {{.*#+}} ymm2 = ymm10[u,u,u,u,u,u,u,u,u,u,u,u,8,9,4,5,16,17,28,29,24,25,28,29,u,u,u,u,u,u,u,u]
+; AVX512F-FAST-NEXT: vpshufb {{.*#+}} ymm2 = ymm10[u,u,u,u,u,u,u,u,u,u,u,u,8,9,4,5,16,17,28,29,24,25,28,29,24,25,28,29,24,25,20,21]
; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm11 = ymm3[0],ymm4[1],ymm3[2,3],ymm4[4],ymm3[5,6],ymm4[7]
; AVX512F-FAST-NEXT: vpshufb {{.*#+}} xmm12 = xmm11[0,1,12,13,u,u,4,5,u,u,u,u,12,13,14,15]
; AVX512F-FAST-NEXT: vextracti128 $1, %ymm11, %xmm13
; AVX512F-FAST-NEXT: vpshufd {{.*#+}} xmm13 = xmm13[2,1,0,3]
; AVX512F-FAST-NEXT: vpshufb {{.*#+}} xmm14 = xmm13[u,u,u,u,0,1,u,u,8,9,12,13,u,u,u,u]
; AVX512F-FAST-NEXT: vpblendw {{.*#+}} xmm12 = xmm12[0,1],xmm14[2],xmm12[3],xmm14[4,5],xmm12[6,7]
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm12[0,1,2],ymm2[3,4,5],ymm12[6,7]
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm12[0,1,2],ymm2[3,4,5,6,7]
; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm9 = ymm2[0,1,2],ymm9[3,4,5,6,7],ymm2[8,9,10],ymm9[11,12,13,14,15]
; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm9[4,5,6,7]
; AVX512F-FAST-NEXT: vpshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,5,5,5,5]
; AVX512F-FAST-NEXT: vpshufb {{.*#+}} xmm8 = xmm8[6,7,2,3,4,5,u,u,2,3,14,15,u,u,6,7]
; AVX512F-FAST-NEXT: vpblendw {{.*#+}} xmm5 = xmm8[0,1,2],xmm5[3],xmm8[4,5],xmm5[6],xmm8[7]
; AVX512F-FAST-NEXT: vinserti128 $1, %xmm5, %ymm0, %ymm5
-; AVX512F-FAST-NEXT: vpshufb {{.*#+}} ymm8 = ymm10[u,u,u,u,u,u,u,u,u,u,u,u,10,11,6,7,18,19,30,31,26,27,26,27,u,u,u,u,u,u,u,u]
+; AVX512F-FAST-NEXT: vpshufb {{.*#+}} ymm8 = ymm10[u,u,u,u,u,u,u,u,u,u,u,u,10,11,6,7,18,19,30,31,26,27,26,27,30,31,26,27,26,27,22,23]
; AVX512F-FAST-NEXT: vpshufb {{.*#+}} xmm9 = xmm11[2,3,14,15,u,u,6,7,u,u,u,u,12,13,14,15]
; AVX512F-FAST-NEXT: vpshufb {{.*#+}} xmm10 = xmm13[u,u,u,u,2,3,u,u,10,11,14,15,u,u,u,u]
; AVX512F-FAST-NEXT: vpblendw {{.*#+}} xmm9 = xmm9[0,1],xmm10[2],xmm9[3],xmm10[4,5],xmm9[6,7]
-; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm8 = ymm9[0,1,2],ymm8[3,4,5],ymm9[6,7]
+; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm8 = ymm9[0,1,2],ymm8[3,4,5,6,7]
; AVX512F-FAST-NEXT: vpblendw {{.*#+}} ymm5 = ymm8[0,1,2],ymm5[3,4,5,6,7],ymm8[8,9,10],ymm5[11,12,13,14,15]
; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm5 = ymm8[0,1,2,3],ymm5[4,5,6,7]
; AVX512F-FAST-NEXT: vpblendd {{.*#+}} ymm8 = ymm0[0,1],ymm1[2],ymm0[3,4],ymm1[5],ymm0[6,7]
@@ -2831,7 +2831,7 @@ define void @load_i16_stride6_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm2 = xmm8[0,1,2,3,4,6,6,7]
; AVX1-ONLY-NEXT: vpunpckhdq {{.*#+}} xmm2 = xmm2[2],xmm4[2],xmm2[3],xmm4[3]
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm1 = xmm2[0,1,2],xmm1[3,4,5],xmm2[6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovdqa 176(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vpslld $16, %xmm1, %xmm1
@@ -2885,7 +2885,7 @@ define void @load_i16_stride6_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vpunpckhdq {{.*#+}} xmm10 = xmm10[2],xmm1[2],xmm10[3],xmm1[3]
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm2 = xmm10[0,1,2],xmm2[3,4,5],xmm10[6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm0[3,4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm0[3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovdqa 368(%rdi), %xmm0
; AVX1-ONLY-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vpslld $16, %xmm0, %xmm0
@@ -2925,7 +2925,7 @@ define void @load_i16_stride6_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: vpsrld $16, %xmm15, %xmm8
; AVX1-ONLY-NEXT: vpunpckhdq {{.*#+}} xmm7 = xmm7[2],xmm8[2],xmm7[3],xmm8[3]
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm2 = xmm7[0,1,2],xmm2[3,4,5],xmm7[6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2],ymm0[3,4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2],ymm0[3,4,5,6,7]
; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm2 = xmm6[0,1,1,3,4,5,6,7]
; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm6 = xmm11[1,1,1,1]
; AVX1-ONLY-NEXT: vpunpckldq {{.*#+}} xmm2 = xmm2[0],xmm6[0],xmm2[1],xmm6[1]
@@ -2956,7 +2956,7 @@ define void @load_i16_stride6_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: vpsrld $16, %xmm9, %xmm4
; AVX1-ONLY-NEXT: vpunpckhdq {{.*#+}} xmm3 = xmm3[2],xmm4[2],xmm3[3],xmm4[3]
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm2 = xmm3[0,1,2],xmm2[3,4,5],xmm3[6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2],ymm0[3,4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2],ymm0[3,4,5,6,7]
; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm1 = xmm1[0,1,1,3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm2 = xmm3[1,1,1,1]
@@ -3316,10 +3316,10 @@ define void @load_i16_stride6_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,6,6,7]
; AVX2-SLOW-NEXT: vpshufb %xmm15, %xmm3, %xmm1
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm0 = xmm1[0,1],xmm0[2],xmm1[3],xmm0[4,5],xmm1[6,7]
-; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} ymm13 = <u,u,u,u,u,u,u,u,u,u,u,u,8,9,4,5,16,17,28,29,24,25,28,29,u,u,u,u,u,u,u,u>
+; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} ymm13 = <u,u,u,u,u,u,u,u,u,u,u,u,8,9,4,5,16,17,28,29,24,25,28,29,24,25,28,29,24,25,20,21>
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm6 = ymm4[0],ymm8[1],ymm4[2,3,4,5],ymm8[6],ymm4[7]
; AVX2-SLOW-NEXT: vpshufb %ymm13, %ymm6, %ymm4
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2],ymm4[3,4,5],ymm0[6,7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2],ymm4[3,4,5,6,7]
; AVX2-SLOW-NEXT: vmovdqa 352(%rdi), %ymm1
; AVX2-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovdqa 320(%rdi), %ymm9
@@ -3357,7 +3357,7 @@ define void @load_i16_stride6_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-SLOW-NEXT: vpblendd $66, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm15 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: # ymm15 = ymm14[0],mem[1],ymm14[2,3,4,5],mem[6],ymm14[7]
; AVX2-SLOW-NEXT: vpshufb %ymm13, %ymm15, %ymm13
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm12 = ymm12[0,1,2],ymm13[3,4,5],ymm12[6,7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm12 = ymm12[0,1,2],ymm13[3,4,5,6,7]
; AVX2-SLOW-NEXT: vinserti128 $1, %xmm8, %ymm0, %ymm8
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm8 = ymm12[0,1,2],ymm8[3,4,5,6,7],ymm12[8,9,10],ymm8[11,12,13,14,15]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm8 = ymm12[0,1,2,3],ymm8[4,5,6,7]
@@ -3367,9 +3367,9 @@ define void @load_i16_stride6_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} xmm13 = <2,3,14,15,u,u,6,7,u,u,u,u,12,13,14,15>
; AVX2-SLOW-NEXT: vpshufb %xmm13, %xmm3, %xmm3
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm3 = xmm3[0,1],xmm7[2],xmm3[3],xmm7[4,5],xmm3[6,7]
-; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} ymm7 = <u,u,u,u,u,u,u,u,u,u,u,u,10,11,6,7,18,19,30,31,26,27,26,27,u,u,u,u,u,u,u,u>
+; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} ymm7 = <u,u,u,u,u,u,u,u,u,u,u,u,10,11,6,7,18,19,30,31,26,27,26,27,30,31,26,27,26,27,22,23>
; AVX2-SLOW-NEXT: vpshufb %ymm7, %ymm6, %ymm6
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2],ymm6[3,4,5],ymm3[6,7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2],ymm6[3,4,5,6,7]
; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} xmm6 = <6,7,2,3,4,5,u,u,2,3,14,15,u,u,6,7>
; AVX2-SLOW-NEXT: vpshufb %xmm6, %xmm5, %xmm5
; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,5,5,5]
@@ -3385,7 +3385,7 @@ define void @load_i16_stride6_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-SLOW-NEXT: vpshufb %xmm13, %xmm11, %xmm2
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm1 = xmm2[0,1],xmm1[2],xmm2[3],xmm1[4,5],xmm2[6,7]
; AVX2-SLOW-NEXT: vpshufb %ymm7, %ymm15, %ymm2
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5],ymm1[6,7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
; AVX2-SLOW-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7],ymm1[8,9,10],ymm0[11,12,13,14,15]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
@@ -3604,8 +3604,8 @@ define void @load_i16_stride6_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-NEXT: vpshufb %xmm15, %xmm10, %xmm3
; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1],xmm3[2],xmm1[3],xmm3[4,5],xmm1[6,7]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm7 = ymm5[0],ymm8[1],ymm5[2,3,4,5],ymm8[6],ymm5[7]
-; AVX2-FAST-NEXT: vpshufb {{.*#+}} ymm3 = ymm7[u,u,u,u,u,u,u,u,u,u,u,u,8,9,4,5,16,17,28,29,24,25,28,29,u,u,u,u,u,u,u,u]
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm3[3,4,5],ymm1[6,7]
+; AVX2-FAST-NEXT: vpshufb {{.*#+}} ymm3 = ymm7[u,u,u,u,u,u,u,u,u,u,u,u,8,9,4,5,16,17,28,29,24,25,28,29,24,25,28,29,24,25,20,21]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm3[3,4,5,6,7]
; AVX2-FAST-NEXT: vmovdqa 352(%rdi), %ymm0
; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovdqa 320(%rdi), %ymm11
@@ -3640,8 +3640,8 @@ define void @load_i16_stride6_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
; AVX2-FAST-NEXT: vpblendd $66, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm15 # 32-byte Folded Reload
; AVX2-FAST-NEXT: # ymm15 = ymm14[0],mem[1],ymm14[2,3,4,5],mem[6],ymm14[7]
-; AVX2-FAST-NEXT: vpshufb {{.*#+}} ymm13 = ymm15[u,u,u,u,u,u,u,u,u,u,u,u,8,9,4,5,16,17,28,29,24,25,28,29,u,u,u,u,u,u,u,u]
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2],ymm13[3,4,5],ymm3[6,7]
+; AVX2-FAST-NEXT: vpshufb {{.*#+}} ymm13 = ymm15[u,u,u,u,u,u,u,u,u,u,u,u,8,9,4,5,16,17,28,29,24,25,28,29,24,25,28,29,24,25,20,21]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2],ymm13[3,4,5,6,7]
; AVX2-FAST-NEXT: vinserti128 $1, %xmm8, %ymm0, %ymm8
; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm8 = ymm3[0,1,2],ymm8[3,4,5,6,7],ymm3[8,9,10],ymm8[11,12,13,14,15]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm8[4,5,6,7]
@@ -3651,9 +3651,9 @@ define void @load_i16_stride6_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm4 = <u,u,u,u,2,3,u,u,10,11,14,15,u,u,u,u>
; AVX2-FAST-NEXT: vpshufb %xmm4, %xmm10, %xmm10
; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm3 = xmm3[0,1],xmm10[2],xmm3[3],xmm10[4,5],xmm3[6,7]
-; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm10 = <u,u,u,u,u,u,u,u,u,u,u,u,10,11,6,7,18,19,30,31,26,27,26,27,u,u,u,u,u,u,u,u>
+; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm10 = <u,u,u,u,u,u,u,u,u,u,u,u,10,11,6,7,18,19,30,31,26,27,26,27,30,31,26,27,26,27,22,23>
; AVX2-FAST-NEXT: vpshufb %ymm10, %ymm7, %ymm7
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2],ymm7[3,4,5],ymm3[6,7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2],ymm7[3,4,5,6,7]
; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm7 = <6,7,2,3,4,5,u,u,2,3,14,15,u,u,6,7>
; AVX2-FAST-NEXT: vpshufb %xmm7, %xmm5, %xmm5
; AVX2-FAST-NEXT: vpshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,5,5,5,5]
@@ -3669,7 +3669,7 @@ define void @load_i16_stride6_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-NEXT: vpshufb %xmm4, %xmm12, %xmm2
; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1],xmm2[2],xmm1[3],xmm2[4,5],xmm1[6,7]
; AVX2-FAST-NEXT: vpshufb %ymm10, %ymm15, %ymm2
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5],ymm1[6,7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
; AVX2-FAST-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7],ymm1[8,9,10],ymm0[11,12,13,14,15]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
@@ -3882,8 +3882,8 @@ define void @load_i16_stride6_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm15, %xmm10, %xmm3
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1],xmm3[2],xmm1[3],xmm3[4,5],xmm1[6,7]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm7 = ymm5[0],ymm8[1],ymm5[2,3,4,5],ymm8[6],ymm5[7]
-; AVX2-FAST-PERLANE-NEXT: vpshufb {{.*#+}} ymm3 = ymm7[u,u,u,u,u,u,u,u,u,u,u,u,8,9,4,5,16,17,28,29,24,25,28,29,u,u,u,u,u,u,u,u]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm3[3,4,5],ymm1[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpshufb {{.*#+}} ymm3 = ymm7[u,u,u,u,u,u,u,u,u,u,u,u,8,9,4,5,16,17,28,29,24,25,28,29,24,25,28,29,24,25,20,21]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm3[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqa 352(%rdi), %ymm0
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovdqa 320(%rdi), %ymm11
@@ -3918,8 +3918,8 @@ define void @load_i16_stride6_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vpblendd $66, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm15 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # ymm15 = ymm14[0],mem[1],ymm14[2,3,4,5],mem[6],ymm14[7]
-; AVX2-FAST-PERLANE-NEXT: vpshufb {{.*#+}} ymm13 = ymm15[u,u,u,u,u,u,u,u,u,u,u,u,8,9,4,5,16,17,28,29,24,25,28,29,u,u,u,u,u,u,u,u]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2],ymm13[3,4,5],ymm3[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpshufb {{.*#+}} ymm13 = ymm15[u,u,u,u,u,u,u,u,u,u,u,u,8,9,4,5,16,17,28,29,24,25,28,29,24,25,28,29,24,25,20,21]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2],ymm13[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm8, %ymm0, %ymm8
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm8 = ymm3[0,1,2],ymm8[3,4,5,6,7],ymm3[8,9,10],ymm8[11,12,13,14,15]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm8[4,5,6,7]
@@ -3929,9 +3929,9 @@ define void @load_i16_stride6_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm4 = <u,u,u,u,2,3,u,u,10,11,14,15,u,u,u,u>
; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm4, %xmm10, %xmm10
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm3 = xmm3[0,1],xmm10[2],xmm3[3],xmm10[4,5],xmm3[6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} ymm10 = <u,u,u,u,u,u,u,u,u,u,u,u,10,11,6,7,18,19,30,31,26,27,26,27,u,u,u,u,u,u,u,u>
+; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} ymm10 = <u,u,u,u,u,u,u,u,u,u,u,u,10,11,6,7,18,19,30,31,26,27,26,27,30,31,26,27,26,27,22,23>
; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm10, %ymm7, %ymm7
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2],ymm7[3,4,5],ymm3[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2],ymm7[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm7 = <6,7,2,3,4,5,u,u,2,3,14,15,u,u,6,7>
; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm7, %xmm5, %xmm5
; AVX2-FAST-PERLANE-NEXT: vpshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,5,5,5,5]
@@ -3947,7 +3947,7 @@ define void @load_i16_stride6_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm4, %xmm12, %xmm2
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1],xmm2[2],xmm1[3],xmm2[4,5],xmm1[6,7]
; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm10, %ymm15, %ymm2
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5],ymm1[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7],ymm1[8,9,10],ymm0[11,12,13,14,15]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
@@ -4134,8 +4134,8 @@ define void @load_i16_stride6_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm9[0],ymm12[1],ymm9[2,3,4,5],ymm12[6],ymm9[7]
; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 %ymm9, %ymm29
; AVX512F-ONLY-SLOW-NEXT: vmovdqa %ymm12, %ymm15
-; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} ymm9 = ymm3[u,u,u,u,u,u,u,u,u,u,u,u,8,9,4,5,16,17,28,29,24,25,28,29,u,u,u,u,u,u,u,u]
-; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2],ymm9[3,4,5],ymm0[6,7]
+; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} ymm9 = ymm3[u,u,u,u,u,u,u,u,u,u,u,u,8,9,4,5,16,17,28,29,24,25,28,29,24,25,28,29,24,25,20,21]
+; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2],ymm9[3,4,5,6,7]
; AVX512F-ONLY-SLOW-NEXT: vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512F-ONLY-SLOW-NEXT: vmovdqa 352(%rdi), %ymm9
; AVX512F-ONLY-SLOW-NEXT: vmovdqa 320(%rdi), %ymm10
@@ -4173,8 +4173,8 @@ define void @load_i16_stride6_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX512F-ONLY-SLOW-NEXT: vpshufb %xmm10, %xmm5, %xmm2
; AVX512F-ONLY-SLOW-NEXT: vpshufb %xmm12, %xmm1, %xmm1
; AVX512F-ONLY-SLOW-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1],xmm2[2],xmm1[3],xmm2[4,5],xmm1[6,7]
-; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} ymm2 = ymm3[u,u,u,u,u,u,u,u,u,u,u,u,10,11,6,7,18,19,30,31,26,27,26,27,u,u,u,u,u,u,u,u]
-; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm13 = ymm1[0,1,2],ymm2[3,4,5],ymm1[6,7]
+; AVX512F-ONLY-SLOW-NEXT: vpshufb {{.*#+}} ymm2 = ymm3[u,u,u,u,u,u,u,u,u,u,u,u,10,11,6,7,18,19,30,31,26,27,26,27,30,31,26,27,26,27,22,23]
+; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm13 = ymm1[0,1,2],ymm2[3,4,5,6,7]
; AVX512F-ONLY-SLOW-NEXT: vpshufb %xmm7, %xmm9, %xmm1
; AVX512F-ONLY-SLOW-NEXT: vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,5,5,5,5]
; AVX512F-ONLY-SLOW-NEXT: vpblendw {{.*#+}} xmm0 = xmm1[0,1,2],xmm0[3],xmm1[4,5],xmm0[6],xmm1[7]
@@ -4420,8 +4420,8 @@ define void @load_i16_stride6_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm5 = ymm11[0],ymm2[1],ymm11[2,3,4,5],ymm2[6],ymm11[7]
; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %ymm11, %ymm22
; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %ymm2, %ymm28
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} ymm2 = ymm5[u,u,u,u,u,u,u,u,u,u,u,u,8,9,4,5,16,17,28,29,24,25,28,29,u,u,u,u,u,u,u,u]
-; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2],ymm2[3,4,5],ymm0[6,7]
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} ymm2 = ymm5[u,u,u,u,u,u,u,u,u,u,u,u,8,9,4,5,16,17,28,29,24,25,28,29,24,25,28,29,24,25,20,21]
+; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2],ymm2[3,4,5,6,7]
; AVX512F-ONLY-FAST-NEXT: vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512F-ONLY-FAST-NEXT: vmovdqa 352(%rdi), %ymm2
; AVX512F-ONLY-FAST-NEXT: vmovdqa 320(%rdi), %ymm11
@@ -4459,8 +4459,8 @@ define void @load_i16_stride6_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX512F-ONLY-FAST-NEXT: vpshufb %xmm10, %xmm1, %xmm1
; AVX512F-ONLY-FAST-NEXT: vpshufb %xmm12, %xmm6, %xmm3
; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1],xmm3[2],xmm1[3],xmm3[4,5],xmm1[6,7]
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} ymm3 = ymm5[u,u,u,u,u,u,u,u,u,u,u,u,10,11,6,7,18,19,30,31,26,27,26,27,u,u,u,u,u,u,u,u]
-; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm3[3,4,5],ymm1[6,7]
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} ymm3 = ymm5[u,u,u,u,u,u,u,u,u,u,u,u,10,11,6,7,18,19,30,31,26,27,26,27,30,31,26,27,26,27,22,23]
+; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm3[3,4,5,6,7]
; AVX512F-ONLY-FAST-NEXT: vmovdqu64 %zmm1, (%rsp) # 64-byte Spill
; AVX512F-ONLY-FAST-NEXT: vpshufb %xmm9, %xmm8, %xmm1
; AVX512F-ONLY-FAST-NEXT: vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,5,5,5,5]
@@ -4696,8 +4696,8 @@ define void @load_i16_stride6_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX512DQ-SLOW-NEXT: vinserti128 $1, 96(%rdi), %ymm3, %ymm8
; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm8[0],ymm12[1],ymm8[2,3,4,5],ymm12[6],ymm8[7]
; AVX512DQ-SLOW-NEXT: vmovdqa64 %ymm8, %ymm29
-; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} ymm9 = ymm3[u,u,u,u,u,u,u,u,u,u,u,u,8,9,4,5,16,17,28,29,24,25,28,29,u,u,u,u,u,u,u,u]
-; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm9 = ymm1[0,1,2],ymm9[3,4,5],ymm1[6,7]
+; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} ymm9 = ymm3[u,u,u,u,u,u,u,u,u,u,u,u,8,9,4,5,16,17,28,29,24,25,28,29,24,25,28,29,24,25,20,21]
+; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm9 = ymm1[0,1,2],ymm9[3,4,5,6,7]
; AVX512DQ-SLOW-NEXT: vmovdqa 352(%rdi), %ymm0
; AVX512DQ-SLOW-NEXT: vmovdqa 320(%rdi), %ymm1
; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm14 = ymm1[0,1],ymm0[2],ymm1[3,4],ymm0[5],ymm1[6,7]
@@ -4739,8 +4739,8 @@ define void @load_i16_stride6_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX512DQ-SLOW-NEXT: vpshufb %xmm0, %xmm5, %xmm0
; AVX512DQ-SLOW-NEXT: vpshufb %xmm10, %xmm2, %xmm2
; AVX512DQ-SLOW-NEXT: vpblendw {{.*#+}} xmm0 = xmm2[0,1],xmm0[2],xmm2[3],xmm0[4,5],xmm2[6,7]
-; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} ymm2 = ymm3[u,u,u,u,u,u,u,u,u,u,u,u,10,11,6,7,18,19,30,31,26,27,26,27,u,u,u,u,u,u,u,u]
-; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm0[0,1,2],ymm2[3,4,5],ymm0[6,7]
+; AVX512DQ-SLOW-NEXT: vpshufb {{.*#+}} ymm2 = ymm3[u,u,u,u,u,u,u,u,u,u,u,u,10,11,6,7,18,19,30,31,26,27,26,27,30,31,26,27,26,27,22,23]
+; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm0[0,1,2],ymm2[3,4,5,6,7]
; AVX512DQ-SLOW-NEXT: vpshufb %xmm8, %xmm1, %xmm0
; AVX512DQ-SLOW-NEXT: vpshufhw {{.*#+}} xmm1 = xmm14[0,1,2,3,5,5,5,5]
; AVX512DQ-SLOW-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1,2],xmm1[3],xmm0[4,5],xmm1[6],xmm0[7]
@@ -4975,8 +4975,8 @@ define void @load_i16_stride6_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX512DQ-FAST-NEXT: vinserti128 $1, 96(%rdi), %ymm9, %ymm12
; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm6 = ymm12[0],ymm1[1],ymm12[2,3,4,5],ymm1[6],ymm12[7]
; AVX512DQ-FAST-NEXT: vmovdqa64 %ymm1, %ymm28
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} ymm2 = ymm6[u,u,u,u,u,u,u,u,u,u,u,u,8,9,4,5,16,17,28,29,24,25,28,29,u,u,u,u,u,u,u,u]
-; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm9 = ymm0[0,1,2],ymm2[3,4,5],ymm0[6,7]
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} ymm2 = ymm6[u,u,u,u,u,u,u,u,u,u,u,u,8,9,4,5,16,17,28,29,24,25,28,29,24,25,28,29,24,25,20,21]
+; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm9 = ymm0[0,1,2],ymm2[3,4,5,6,7]
; AVX512DQ-FAST-NEXT: vmovdqa 352(%rdi), %ymm0
; AVX512DQ-FAST-NEXT: vmovdqa 320(%rdi), %ymm2
; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1],ymm0[2],ymm2[3,4],ymm0[5],ymm2[6,7]
@@ -5017,8 +5017,8 @@ define void @load_i16_stride6_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX512DQ-FAST-NEXT: vpshufb %xmm0, %xmm3, %xmm0
; AVX512DQ-FAST-NEXT: vpshufb %xmm8, %xmm7, %xmm3
; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1],xmm3[2],xmm0[3],xmm3[4,5],xmm0[6,7]
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} ymm3 = ymm6[u,u,u,u,u,u,u,u,u,u,u,u,10,11,6,7,18,19,30,31,26,27,26,27,u,u,u,u,u,u,u,u]
-; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm0[0,1,2],ymm3[3,4,5],ymm0[6,7]
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} ymm3 = ymm6[u,u,u,u,u,u,u,u,u,u,u,u,10,11,6,7,18,19,30,31,26,27,26,27,30,31,26,27,26,27,22,23]
+; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm0[0,1,2],ymm3[3,4,5,6,7]
; AVX512DQ-FAST-NEXT: vpshufb %xmm14, %xmm4, %xmm0
; AVX512DQ-FAST-NEXT: vpshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,5,5,5]
; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1,2],xmm1[3],xmm0[4,5],xmm1[6],xmm0[7]
@@ -6555,7 +6555,7 @@ define void @load_i16_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,6,6,7]
; AVX1-ONLY-NEXT: vpunpckhdq {{.*#+}} xmm2 = xmm2[2],xmm4[2],xmm2[3],xmm4[3]
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm1 = xmm2[0,1,2],xmm1[3,4,5],xmm2[6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovdqa 176(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovdqa %xmm1, (%rsp) # 16-byte Spill
; AVX1-ONLY-NEXT: vpslld $16, %xmm1, %xmm1
@@ -6611,7 +6611,7 @@ define void @load_i16_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vpunpckhdq {{.*#+}} xmm3 = xmm3[2],xmm1[2],xmm3[3],xmm1[3]
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm2 = xmm3[0,1,2],xmm2[3,4,5],xmm3[6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2],ymm0[3,4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2],ymm0[3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovdqa 560(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vpslld $16, %xmm1, %xmm2
@@ -6665,7 +6665,7 @@ define void @load_i16_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vpunpckhdq {{.*#+}} xmm3 = xmm3[2],xmm1[2],xmm3[3],xmm1[3]
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm2 = xmm3[0,1,2],xmm2[3,4,5],xmm3[6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2],ymm0[3,4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2],ymm0[3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovdqa 368(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vpslld $16, %xmm1, %xmm2
@@ -6719,7 +6719,7 @@ define void @load_i16_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vpunpckhdq {{.*#+}} xmm0 = xmm0[2],xmm2[2],xmm0[3],xmm2[3]
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1,2],xmm14[3,4,5],xmm0[6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm0[0,1,2],ymm1[3,4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm0[0,1,2],ymm1[3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovdqa 752(%rdi), %xmm0
; AVX1-ONLY-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vpslld $16, %xmm0, %xmm0
@@ -6761,7 +6761,7 @@ define void @load_i16_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: vpsrld $16, %xmm5, %xmm15
; AVX1-ONLY-NEXT: vpunpckhdq {{.*#+}} xmm14 = xmm14[2],xmm15[2],xmm14[3],xmm15[3]
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm1 = xmm14[0,1,2],xmm1[3,4,5],xmm14[6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
; AVX1-ONLY-NEXT: vpshuflw $212, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm1 = mem[0,1,1,3,4,5,6,7]
; AVX1-ONLY-NEXT: vpshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload
@@ -6796,7 +6796,7 @@ define void @load_i16_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: vpsrld $16, %xmm13, %xmm13
; AVX1-ONLY-NEXT: vpunpckhdq {{.*#+}} xmm12 = xmm12[2],xmm13[2],xmm12[3],xmm13[3]
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm1 = xmm12[0,1,2],xmm1[3,4,5],xmm12[6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm1 = xmm11[0,1,1,3,4,5,6,7]
; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm11 = xmm6[1,1,1,1]
; AVX1-ONLY-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm1[0],xmm11[0],xmm1[1],xmm11[1]
@@ -6829,7 +6829,7 @@ define void @load_i16_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: vpsrld $16, %xmm6, %xmm9
; AVX1-ONLY-NEXT: vpunpckhdq {{.*#+}} xmm8 = xmm8[2],xmm9[2],xmm8[3],xmm9[3]
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm1 = xmm8[0,1,2],xmm1[3,4,5],xmm8[6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm1 = xmm7[0,1,1,3,4,5,6,7]
; AVX1-ONLY-NEXT: vpshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm7 = mem[1,1,1,1]
@@ -6863,7 +6863,7 @@ define void @load_i16_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: vpsrld $16, %xmm6, %xmm4
; AVX1-ONLY-NEXT: vpunpckhdq {{.*#+}} xmm3 = xmm3[2],xmm4[2],xmm3[3],xmm4[3]
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm1 = xmm3[0,1,2],xmm1[3,4,5],xmm3[6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm1 = xmm2[0,1,1,3,4,5,6,7]
; AVX1-ONLY-NEXT: vpshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm2 = mem[1,1,1,1]
@@ -7595,14 +7595,14 @@ define void @load_i16_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} xmm9 = <0,1,12,13,u,u,4,5,u,u,u,u,12,13,14,15>
; AVX2-SLOW-NEXT: vpshufb %xmm9, %xmm6, %xmm1
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm0 = xmm1[0,1],xmm0[2],xmm1[3],xmm0[4,5],xmm1[6,7]
-; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} ymm2 = <u,u,u,u,u,u,u,u,u,u,u,u,8,9,4,5,16,17,28,29,24,25,28,29,u,u,u,u,u,u,u,u>
+; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} ymm2 = <u,u,u,u,u,u,u,u,u,u,u,u,8,9,4,5,16,17,28,29,24,25,28,29,24,25,28,29,24,25,20,21>
; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-SLOW-NEXT: vpblendd $66, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: # ymm1 = ymm1[0],mem[1],ymm1[2,3,4,5],mem[6],ymm1[7]
; AVX2-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vpshufb %ymm2, %ymm1, %ymm1
; AVX2-SLOW-NEXT: vmovdqa %ymm2, %ymm5
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5],ymm0[6,7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
; AVX2-SLOW-NEXT: vmovdqa 544(%rdi), %ymm1
; AVX2-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovdqa 512(%rdi), %ymm2
@@ -7633,7 +7633,7 @@ define void @load_i16_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vpshufb %ymm5, %ymm1, %ymm1
; AVX2-SLOW-NEXT: vmovdqa %ymm5, %ymm10
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5],ymm0[6,7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
; AVX2-SLOW-NEXT: vmovdqa 352(%rdi), %ymm1
; AVX2-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovdqa 320(%rdi), %ymm2
@@ -7663,7 +7663,7 @@ define void @load_i16_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vpshufb %ymm10, %ymm0, %ymm12
; AVX2-SLOW-NEXT: vmovdqa %ymm10, %ymm14
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm12 = ymm13[0,1,2],ymm12[3,4,5],ymm13[6,7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm12 = ymm13[0,1,2],ymm12[3,4,5,6,7]
; AVX2-SLOW-NEXT: vmovdqa 736(%rdi), %ymm0
; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovdqa 704(%rdi), %ymm5
@@ -7700,7 +7700,7 @@ define void @load_i16_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-SLOW-NEXT: vpblendd $189, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm9 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: # ymm9 = mem[0],ymm9[1],mem[2,3,4,5],ymm9[6],mem[7]
; AVX2-SLOW-NEXT: vpshufb %ymm14, %ymm9, %ymm15
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2],ymm15[3,4,5],ymm0[6,7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2],ymm15[3,4,5,6,7]
; AVX2-SLOW-NEXT: vinserti128 $1, %xmm7, %ymm0, %ymm7
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm7 = ymm0[0,1,2],ymm7[3,4,5,6,7],ymm0[8,9,10],ymm7[11,12,13,14,15]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm7[4,5,6,7]
@@ -7710,10 +7710,10 @@ define void @load_i16_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} xmm15 = <2,3,14,15,u,u,6,7,u,u,u,u,12,13,14,15>
; AVX2-SLOW-NEXT: vpshufb %xmm15, %xmm6, %xmm6
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm6 = xmm6[0,1],xmm7[2],xmm6[3],xmm7[4,5],xmm6[6,7]
-; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} ymm7 = <u,u,u,u,u,u,u,u,u,u,u,u,10,11,6,7,18,19,30,31,26,27,26,27,u,u,u,u,u,u,u,u>
+; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} ymm7 = <u,u,u,u,u,u,u,u,u,u,u,u,10,11,6,7,18,19,30,31,26,27,26,27,30,31,26,27,26,27,22,23>
; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-SLOW-NEXT: vpshufb %ymm7, %ymm0, %ymm14
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm6 = ymm6[0,1,2],ymm14[3,4,5],ymm6[6,7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm6 = ymm6[0,1,2],ymm14[3,4,5,6,7]
; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} xmm14 = <6,7,2,3,4,5,u,u,2,3,14,15,u,u,6,7>
; AVX2-SLOW-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX2-SLOW-NEXT: vpshufb %xmm14, %xmm0, %xmm0
@@ -7730,7 +7730,7 @@ define void @load_i16_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm0 = xmm3[0,1],xmm0[2],xmm3[3],xmm0[4,5],xmm3[6,7]
; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX2-SLOW-NEXT: vpshufb %ymm7, %ymm3, %ymm3
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2],ymm3[3,4,5],ymm0[6,7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2],ymm3[3,4,5,6,7]
; AVX2-SLOW-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
; AVX2-SLOW-NEXT: vpshufb %xmm14, %xmm3, %xmm3
; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,5,5,5,5]
@@ -7745,7 +7745,7 @@ define void @load_i16_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm0 = xmm1[0,1],xmm0[2],xmm1[3],xmm0[4,5],xmm1[6,7]
; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-SLOW-NEXT: vpshufb %ymm7, %ymm1, %ymm1
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5],ymm0[6,7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
; AVX2-SLOW-NEXT: vpshufb %xmm14, %xmm8, %xmm1
; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} xmm2 = xmm13[0,1,2,3,5,5,5,5]
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1,2],xmm2[3],xmm1[4,5],xmm2[6],xmm1[7]
@@ -7760,7 +7760,7 @@ define void @load_i16_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-SLOW-NEXT: vpshufb %xmm15, %xmm12, %xmm2
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm1 = xmm2[0,1],xmm1[2],xmm2[3],xmm1[4,5],xmm2[6,7]
; AVX2-SLOW-NEXT: vpshufb %ymm7, %ymm9, %ymm2
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5],ymm1[6,7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
; AVX2-SLOW-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7],ymm1[8,9,10],ymm0[11,12,13,14,15]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
@@ -8206,13 +8206,13 @@ define void @load_i16_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-NEXT: vpshufb %xmm3, %xmm1, %xmm2
; AVX2-FAST-NEXT: vmovdqa %xmm3, %xmm8
; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1],xmm2[2],xmm0[3],xmm2[4,5],xmm0[6,7]
-; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm2 = <u,u,u,u,u,u,u,u,u,u,u,u,8,9,4,5,16,17,28,29,24,25,28,29,u,u,u,u,u,u,u,u>
+; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm2 = <u,u,u,u,u,u,u,u,u,u,u,u,8,9,4,5,16,17,28,29,24,25,28,29,24,25,28,29,24,25,20,21>
; AVX2-FAST-NEXT: vpblendd $66, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm1 # 32-byte Folded Reload
; AVX2-FAST-NEXT: # ymm1 = ymm6[0],mem[1],ymm6[2,3,4,5],mem[6],ymm6[7]
; AVX2-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vpshufb %ymm2, %ymm1, %ymm3
; AVX2-FAST-NEXT: vmovdqa %ymm2, %ymm4
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2],ymm3[3,4,5],ymm0[6,7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2],ymm3[3,4,5,6,7]
; AVX2-FAST-NEXT: vmovdqa 544(%rdi), %ymm1
; AVX2-FAST-NEXT: vmovdqu %ymm1, (%rsp) # 32-byte Spill
; AVX2-FAST-NEXT: vmovdqa 512(%rdi), %ymm2
@@ -8245,7 +8245,7 @@ define void @load_i16_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovdqa %ymm4, %ymm2
; AVX2-FAST-NEXT: vpshufb %ymm4, %ymm1, %ymm14
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm14 = ymm0[0,1,2],ymm14[3,4,5],ymm0[6,7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm14 = ymm0[0,1,2],ymm14[3,4,5,6,7]
; AVX2-FAST-NEXT: vmovdqa 352(%rdi), %ymm0
; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovdqa 320(%rdi), %ymm1
@@ -8277,7 +8277,7 @@ define void @load_i16_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vpshufb %ymm2, %ymm0, %ymm12
; AVX2-FAST-NEXT: vmovdqa %ymm2, %ymm14
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm12 = ymm13[0,1,2],ymm12[3,4,5],ymm13[6,7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm12 = ymm13[0,1,2],ymm12[3,4,5,6,7]
; AVX2-FAST-NEXT: vmovdqa 736(%rdi), %ymm0
; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovdqa 704(%rdi), %ymm1
@@ -8312,7 +8312,7 @@ define void @load_i16_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-NEXT: vpblendd $189, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm8 # 32-byte Folded Reload
; AVX2-FAST-NEXT: # ymm8 = mem[0],ymm0[1],mem[2,3,4,5],ymm0[6],mem[7]
; AVX2-FAST-NEXT: vpshufb %ymm14, %ymm8, %ymm15
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm15[3,4,5],ymm1[6,7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm15[3,4,5,6,7]
; AVX2-FAST-NEXT: vinserti128 $1, %xmm9, %ymm0, %ymm9
; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm9 = ymm1[0,1,2],ymm9[3,4,5,6,7],ymm1[8,9,10],ymm9[11,12,13,14,15]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm9[4,5,6,7]
@@ -8322,10 +8322,10 @@ define void @load_i16_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX2-FAST-NEXT: vpshufb %xmm9, %xmm0, %xmm15
; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm7 = xmm7[0,1],xmm15[2],xmm7[3],xmm15[4,5],xmm7[6,7]
-; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm15 = <u,u,u,u,u,u,u,u,u,u,u,u,10,11,6,7,18,19,30,31,26,27,26,27,u,u,u,u,u,u,u,u>
+; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm15 = <u,u,u,u,u,u,u,u,u,u,u,u,10,11,6,7,18,19,30,31,26,27,26,27,30,31,26,27,26,27,22,23>
; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FAST-NEXT: vpshufb %ymm15, %ymm0, %ymm14
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm7 = ymm7[0,1,2],ymm14[3,4,5],ymm7[6,7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm7 = ymm7[0,1,2],ymm14[3,4,5,6,7]
; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm14 = <6,7,2,3,4,5,u,u,2,3,14,15,u,u,6,7>
; AVX2-FAST-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX2-FAST-NEXT: vpshufb %xmm14, %xmm0, %xmm0
@@ -8343,7 +8343,7 @@ define void @load_i16_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1],xmm1[2],xmm0[3],xmm1[4,5],xmm0[6,7]
; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FAST-NEXT: vpshufb %ymm15, %ymm1, %ymm1
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5],ymm0[6,7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
; AVX2-FAST-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX2-FAST-NEXT: vpshufb %xmm14, %xmm1, %xmm1
; AVX2-FAST-NEXT: vpshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,5,5,5,5]
@@ -8358,7 +8358,7 @@ define void @load_i16_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1],xmm1[2],xmm0[3],xmm1[4,5],xmm0[6,7]
; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FAST-NEXT: vpshufb %ymm15, %ymm1, %ymm1
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5],ymm0[6,7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
; AVX2-FAST-NEXT: vpshufb %xmm14, %xmm6, %xmm1
; AVX2-FAST-NEXT: vpshufhw {{.*#+}} xmm3 = xmm13[0,1,2,3,5,5,5,5]
; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1,2],xmm3[3],xmm1[4,5],xmm3[6],xmm1[7]
@@ -8373,7 +8373,7 @@ define void @load_i16_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-NEXT: vpshufb %xmm9, %xmm12, %xmm2
; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1],xmm2[2],xmm1[3],xmm2[4,5],xmm1[6,7]
; AVX2-FAST-NEXT: vpshufb %ymm15, %ymm8, %ymm2
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5],ymm1[6,7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
; AVX2-FAST-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7],ymm1[8,9,10],ymm0[11,12,13,14,15]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
@@ -8806,13 +8806,13 @@ define void @load_i16_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm3, %xmm1, %xmm2
; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm3, %xmm8
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1],xmm2[2],xmm0[3],xmm2[4,5],xmm0[6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} ymm2 = <u,u,u,u,u,u,u,u,u,u,u,u,8,9,4,5,16,17,28,29,24,25,28,29,u,u,u,u,u,u,u,u>
+; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} ymm2 = <u,u,u,u,u,u,u,u,u,u,u,u,8,9,4,5,16,17,28,29,24,25,28,29,24,25,28,29,24,25,20,21>
; AVX2-FAST-PERLANE-NEXT: vpblendd $66, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm1 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # ymm1 = ymm6[0],mem[1],ymm6[2,3,4,5],mem[6],ymm6[7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm2, %ymm1, %ymm3
; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm2, %ymm4
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2],ymm3[3,4,5],ymm0[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2],ymm3[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqa 544(%rdi), %ymm1
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm1, (%rsp) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovdqa 512(%rdi), %ymm2
@@ -8845,7 +8845,7 @@ define void @load_i16_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm4, %ymm2
; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm4, %ymm1, %ymm14
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm14 = ymm0[0,1,2],ymm14[3,4,5],ymm0[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm14 = ymm0[0,1,2],ymm14[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqa 352(%rdi), %ymm0
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovdqa 320(%rdi), %ymm1
@@ -8877,7 +8877,7 @@ define void @load_i16_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm2, %ymm0, %ymm12
; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm2, %ymm14
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm12 = ymm13[0,1,2],ymm12[3,4,5],ymm13[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm12 = ymm13[0,1,2],ymm12[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqa 736(%rdi), %ymm0
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovdqa 704(%rdi), %ymm1
@@ -8912,7 +8912,7 @@ define void @load_i16_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-PERLANE-NEXT: vpblendd $189, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm8 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # ymm8 = mem[0],ymm0[1],mem[2,3,4,5],ymm0[6],mem[7]
; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm14, %ymm8, %ymm15
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm15[3,4,5],ymm1[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm15[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm9, %ymm0, %ymm9
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm9 = ymm1[0,1,2],ymm9[3,4,5,6,7],ymm1[8,9,10],ymm9[11,12,13,14,15]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm9[4,5,6,7]
@@ -8922,10 +8922,10 @@ define void @load_i16_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-PERLANE-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm9, %xmm0, %xmm15
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm7 = xmm7[0,1],xmm15[2],xmm7[3],xmm15[4,5],xmm7[6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} ymm15 = <u,u,u,u,u,u,u,u,u,u,u,u,10,11,6,7,18,19,30,31,26,27,26,27,u,u,u,u,u,u,u,u>
+; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} ymm15 = <u,u,u,u,u,u,u,u,u,u,u,u,10,11,6,7,18,19,30,31,26,27,26,27,30,31,26,27,26,27,22,23>
; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm15, %ymm0, %ymm14
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm7 = ymm7[0,1,2],ymm14[3,4,5],ymm7[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm7 = ymm7[0,1,2],ymm14[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm14 = <6,7,2,3,4,5,u,u,2,3,14,15,u,u,6,7>
; AVX2-FAST-PERLANE-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm14, %xmm0, %xmm0
@@ -8943,7 +8943,7 @@ define void @load_i16_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1],xmm1[2],xmm0[3],xmm1[4,5],xmm0[6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm15, %ymm1, %ymm1
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5],ymm0[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm14, %xmm1, %xmm1
; AVX2-FAST-PERLANE-NEXT: vpshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,5,5,5,5]
@@ -8958,7 +8958,7 @@ define void @load_i16_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1],xmm1[2],xmm0[3],xmm1[4,5],xmm0[6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm15, %ymm1, %ymm1
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5],ymm0[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm14, %xmm6, %xmm1
; AVX2-FAST-PERLANE-NEXT: vpshufhw {{.*#+}} xmm3 = xmm13[0,1,2,3,5,5,5,5]
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1,2],xmm3[3],xmm1[4,5],xmm3[6],xmm1[7]
@@ -8973,7 +8973,7 @@ define void @load_i16_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm9, %xmm12, %xmm2
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1],xmm2[2],xmm1[3],xmm2[4,5],xmm1[6,7]
; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm15, %ymm8, %ymm2
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5],ymm1[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7],ymm1[8,9,10],ymm0[11,12,13,14,15]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
@@ -9331,7 +9331,7 @@ define void @load_i16_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX512F-ONLY-SLOW-NEXT: vmovdqa {{.*#+}} ymm7 = [0,1,12,13,8,9,12,13,8,9,12,13,8,9,4,5,16,17,28,29,24,25,28,29,24,25,28,29,24,25,20,21]
; AVX512F-ONLY-SLOW-NEXT: vpshufb %ymm7, %ymm2, %ymm0
; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 %ymm2, %ymm28
-; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5],ymm1[6,7]
+; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
; AVX512F-ONLY-SLOW-NEXT: vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512F-ONLY-SLOW-NEXT: vmovdqa 736(%rdi), %ymm0
; AVX512F-ONLY-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -9398,7 +9398,7 @@ define void @load_i16_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm8 = ymm1[0],ymm4[1],ymm1[2,3,4,5],ymm4[6],ymm1[7]
; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 %ymm4, %ymm31
; AVX512F-ONLY-SLOW-NEXT: vpshufb %ymm7, %ymm8, %ymm2
-; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2],ymm2[3,4,5],ymm0[6,7]
+; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2],ymm2[3,4,5,6,7]
; AVX512F-ONLY-SLOW-NEXT: vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512F-ONLY-SLOW-NEXT: vmovdqa 352(%rdi), %ymm1
; AVX512F-ONLY-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -9439,7 +9439,7 @@ define void @load_i16_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX512F-ONLY-SLOW-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1],xmm3[2],xmm1[3],xmm3[4,5],xmm1[6,7]
; AVX512F-ONLY-SLOW-NEXT: vmovdqa {{.*#+}} ymm3 = [2,3,14,15,10,11,10,11,14,15,10,11,10,11,6,7,18,19,30,31,26,27,26,27,30,31,26,27,26,27,22,23]
; AVX512F-ONLY-SLOW-NEXT: vpshufb %ymm3, %ymm8, %ymm5
-; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm5[3,4,5],ymm1[6,7]
+; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm5[3,4,5,6,7]
; AVX512F-ONLY-SLOW-NEXT: vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512F-ONLY-SLOW-NEXT: vpshufb %xmm13, %xmm4, %xmm1
; AVX512F-ONLY-SLOW-NEXT: vpshufhw {{.*#+}} xmm0 = xmm7[0,1,2,3,5,5,5,5]
@@ -9470,7 +9470,7 @@ define void @load_i16_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX512F-ONLY-SLOW-NEXT: vpshufb %xmm11, %xmm2, %xmm2
; AVX512F-ONLY-SLOW-NEXT: vpshufb %xmm15, %xmm12, %xmm3
; AVX512F-ONLY-SLOW-NEXT: vpblendw {{.*#+}} xmm2 = xmm3[0,1],xmm2[2],xmm3[3],xmm2[4,5],xmm3[6,7]
-; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm2[0,1,2],ymm0[3,4,5],ymm2[6,7]
+; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm2[0,1,2],ymm0[3,4,5,6,7]
; AVX512F-ONLY-SLOW-NEXT: vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 %ymm16, %ymm0
; AVX512F-ONLY-SLOW-NEXT: vpshufb %ymm1, %ymm0, %ymm0
@@ -9959,7 +9959,7 @@ define void @load_i16_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX512F-ONLY-FAST-NEXT: vmovdqa {{.*#+}} ymm8 = [0,1,12,13,8,9,12,13,8,9,12,13,8,9,4,5,16,17,28,29,24,25,28,29,24,25,28,29,24,25,20,21]
; AVX512F-ONLY-FAST-NEXT: vpshufb %ymm8, %ymm2, %ymm1
; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %ymm2, %ymm24
-; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5],ymm0[6,7]
+; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
; AVX512F-ONLY-FAST-NEXT: vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512F-ONLY-FAST-NEXT: vmovdqa 736(%rdi), %ymm0
; AVX512F-ONLY-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -10025,7 +10025,7 @@ define void @load_i16_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm10 = ymm4[0],ymm1[1],ymm4[2,3,4,5],ymm1[6],ymm4[7]
; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %ymm1, %ymm31
; AVX512F-ONLY-FAST-NEXT: vpshufb %ymm8, %ymm10, %ymm4
-; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2],ymm4[3,4,5],ymm0[6,7]
+; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2],ymm4[3,4,5,6,7]
; AVX512F-ONLY-FAST-NEXT: vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512F-ONLY-FAST-NEXT: vmovdqa 352(%rdi), %ymm0
; AVX512F-ONLY-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -10066,7 +10066,7 @@ define void @load_i16_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0,1],xmm5[2],xmm2[3],xmm5[4,5],xmm2[6,7]
; AVX512F-ONLY-FAST-NEXT: vmovdqa {{.*#+}} ymm5 = [2,3,14,15,10,11,10,11,14,15,10,11,10,11,6,7,18,19,30,31,26,27,26,27,30,31,26,27,26,27,22,23]
; AVX512F-ONLY-FAST-NEXT: vpshufb %ymm5, %ymm10, %ymm6
-; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2],ymm6[3,4,5],ymm2[6,7]
+; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2],ymm6[3,4,5,6,7]
; AVX512F-ONLY-FAST-NEXT: vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512F-ONLY-FAST-NEXT: vpshufb %xmm0, %xmm8, %xmm2
; AVX512F-ONLY-FAST-NEXT: vpshufhw {{.*#+}} xmm1 = xmm4[0,1,2,3,5,5,5,5]
@@ -10096,7 +10096,7 @@ define void @load_i16_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %xmm23, %xmm4
; AVX512F-ONLY-FAST-NEXT: vpshufb %xmm13, %xmm4, %xmm4
; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} xmm3 = xmm3[0,1],xmm4[2],xmm3[3],xmm4[4,5],xmm3[6,7]
-; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm3[0,1,2],ymm1[3,4,5],ymm3[6,7]
+; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm3[0,1,2],ymm1[3,4,5,6,7]
; AVX512F-ONLY-FAST-NEXT: vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %ymm29, %ymm1
; AVX512F-ONLY-FAST-NEXT: vpshufb %ymm2, %ymm1, %ymm1
@@ -10579,7 +10579,7 @@ define void @load_i16_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX512DQ-SLOW-NEXT: vmovdqa {{.*#+}} ymm6 = [0,1,12,13,8,9,12,13,8,9,12,13,8,9,4,5,16,17,28,29,24,25,28,29,24,25,28,29,24,25,20,21]
; AVX512DQ-SLOW-NEXT: vpshufb %ymm6, %ymm3, %ymm1
; AVX512DQ-SLOW-NEXT: vmovdqa64 %ymm3, %ymm20
-; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm2[0,1,2],ymm1[3,4,5],ymm2[6,7]
+; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm2[0,1,2],ymm1[3,4,5,6,7]
; AVX512DQ-SLOW-NEXT: vmovdqa 736(%rdi), %ymm1
; AVX512DQ-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512DQ-SLOW-NEXT: vmovdqa 704(%rdi), %ymm2
@@ -10649,7 +10649,7 @@ define void @load_i16_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX512DQ-SLOW-NEXT: vmovdqa64 %ymm2, %ymm31
; AVX512DQ-SLOW-NEXT: vmovdqa64 %ymm4, %ymm30
; AVX512DQ-SLOW-NEXT: vpshufb %ymm6, %ymm8, %ymm2
-; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm6 = ymm0[0,1,2],ymm2[3,4,5],ymm0[6,7]
+; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm6 = ymm0[0,1,2],ymm2[3,4,5,6,7]
; AVX512DQ-SLOW-NEXT: vmovdqa 352(%rdi), %ymm0
; AVX512DQ-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512DQ-SLOW-NEXT: vmovdqa 320(%rdi), %ymm1
@@ -10690,7 +10690,7 @@ define void @load_i16_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX512DQ-SLOW-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1],xmm6[2],xmm1[3],xmm6[4,5],xmm1[6,7]
; AVX512DQ-SLOW-NEXT: vmovdqa {{.*#+}} ymm6 = [2,3,14,15,10,11,10,11,14,15,10,11,10,11,6,7,18,19,30,31,26,27,26,27,30,31,26,27,26,27,22,23]
; AVX512DQ-SLOW-NEXT: vpshufb %ymm6, %ymm8, %ymm5
-; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm5 = ymm1[0,1,2],ymm5[3,4,5],ymm1[6,7]
+; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm5 = ymm1[0,1,2],ymm5[3,4,5,6,7]
; AVX512DQ-SLOW-NEXT: vpshufb %xmm12, %xmm4, %xmm1
; AVX512DQ-SLOW-NEXT: vpshufhw {{.*#+}} xmm0 = xmm7[0,1,2,3,5,5,5,5]
; AVX512DQ-SLOW-NEXT: vpblendw {{.*#+}} xmm0 = xmm1[0,1,2],xmm0[3],xmm1[4,5],xmm0[6],xmm1[7]
@@ -10722,7 +10722,7 @@ define void @load_i16_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX512DQ-SLOW-NEXT: vmovdqa64 %ymm22, %ymm4
; AVX512DQ-SLOW-NEXT: vpshufb %xmm14, %xmm4, %xmm4
; AVX512DQ-SLOW-NEXT: vpblendw {{.*#+}} xmm3 = xmm4[0,1],xmm3[2],xmm4[3],xmm3[4,5],xmm4[6,7]
-; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm3[0,1,2],ymm2[3,4,5],ymm3[6,7]
+; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm3[0,1,2],ymm2[3,4,5,6,7]
; AVX512DQ-SLOW-NEXT: vmovdqa64 %ymm18, %ymm2
; AVX512DQ-SLOW-NEXT: vpshufb %ymm1, %ymm2, %ymm1
; AVX512DQ-SLOW-NEXT: vmovdqa64 %xmm19, %xmm2
@@ -11183,7 +11183,7 @@ define void @load_i16_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX512DQ-FAST-NEXT: vmovdqa {{.*#+}} ymm6 = [0,1,12,13,8,9,12,13,8,9,12,13,8,9,4,5,16,17,28,29,24,25,28,29,24,25,28,29,24,25,20,21]
; AVX512DQ-FAST-NEXT: vpshufb %ymm6, %ymm4, %ymm2
; AVX512DQ-FAST-NEXT: vmovdqa64 %ymm4, %ymm20
-; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm4 = ymm1[0,1,2],ymm2[3,4,5],ymm1[6,7]
+; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm4 = ymm1[0,1,2],ymm2[3,4,5,6,7]
; AVX512DQ-FAST-NEXT: vmovdqa 736(%rdi), %ymm1
; AVX512DQ-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512DQ-FAST-NEXT: vmovdqa 704(%rdi), %ymm2
@@ -11252,7 +11252,7 @@ define void @load_i16_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX512DQ-FAST-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm11 = ymm4[0],ymm1[1],ymm4[2,3,4,5],ymm1[6],ymm4[7]
; AVX512DQ-FAST-NEXT: vpshufb %ymm6, %ymm11, %ymm4
-; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm6 = ymm0[0,1,2],ymm4[3,4,5],ymm0[6,7]
+; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm6 = ymm0[0,1,2],ymm4[3,4,5,6,7]
; AVX512DQ-FAST-NEXT: vmovdqa 352(%rdi), %ymm0
; AVX512DQ-FAST-NEXT: vmovdqu %ymm0, (%rsp) # 32-byte Spill
; AVX512DQ-FAST-NEXT: vmovdqa 320(%rdi), %ymm1
@@ -11293,7 +11293,7 @@ define void @load_i16_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0,1],xmm6[2],xmm2[3],xmm6[4,5],xmm2[6,7]
; AVX512DQ-FAST-NEXT: vmovdqa {{.*#+}} ymm6 = [2,3,14,15,10,11,10,11,14,15,10,11,10,11,6,7,18,19,30,31,26,27,26,27,30,31,26,27,26,27,22,23]
; AVX512DQ-FAST-NEXT: vpshufb %ymm6, %ymm11, %ymm7
-; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm7 = ymm2[0,1,2],ymm7[3,4,5],ymm2[6,7]
+; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm7 = ymm2[0,1,2],ymm7[3,4,5,6,7]
; AVX512DQ-FAST-NEXT: vpshufb %xmm0, %xmm4, %xmm2
; AVX512DQ-FAST-NEXT: vpshufhw {{.*#+}} xmm1 = xmm8[0,1,2,3,5,5,5,5]
; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} xmm1 = xmm2[0,1,2],xmm1[3],xmm2[4,5],xmm1[6],xmm2[7]
@@ -11324,7 +11324,7 @@ define void @load_i16_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX512DQ-FAST-NEXT: vmovdqa64 %xmm21, %xmm5
; AVX512DQ-FAST-NEXT: vpshufb %xmm10, %xmm5, %xmm5
; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} xmm4 = xmm4[0,1],xmm5[2],xmm4[3],xmm5[4,5],xmm4[6,7]
-; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm4 = ymm4[0,1,2],ymm3[3,4,5],ymm4[6,7]
+; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm4 = ymm4[0,1,2],ymm3[3,4,5,6,7]
; AVX512DQ-FAST-NEXT: vmovdqa64 %ymm18, %ymm3
; AVX512DQ-FAST-NEXT: vpshufb %ymm2, %ymm3, %ymm2
; AVX512DQ-FAST-NEXT: vmovdqa64 %xmm19, %xmm3
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-3.ll b/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-3.ll
index 9779b4904d848..8a98177af0191 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-3.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-3.ll
@@ -177,7 +177,7 @@ define void @load_i32_stride3_vf4(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm0[0,1],ymm1[2,3],ymm0[4,5,6,7]
; AVX2-ONLY-NEXT: vpermps %ymm4, %ymm3, %ymm3
; AVX2-ONLY-NEXT: vmovaps {{.*#+}} xmm4 = [2,5,0,3]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2],ymm1[3,4],ymm0[5],ymm1[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2],ymm1[3],ymm0[4,5,6,7]
; AVX2-ONLY-NEXT: vpermps %ymm0, %ymm4, %ymm0
; AVX2-ONLY-NEXT: vmovaps %xmm2, (%rsi)
; AVX2-ONLY-NEXT: vmovaps %xmm3, (%rdx)
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-5.ll b/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-5.ll
index 07a51fb5a6ea8..b644b92e778aa 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-5.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-5.ll
@@ -444,7 +444,7 @@ define void @load_i32_stride5_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm11, %xmm11
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm11 = xmm11[0,1],mem[2],xmm11[3]
; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} xmm11 = xmm11[1,0]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm11[0,1,2],ymm8[3,4,5],ymm11[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm11[0,1,2],ymm8[3,4,5,6,7]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm11 = ymm0[2,3,0,1]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm12 = ymm0[1,0],ymm11[0,0],ymm0[5,4],ymm11[4,4]
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} ymm12 = ymm12[0,1,2,0,4,5,6,4]
@@ -453,7 +453,7 @@ define void @load_i32_stride5_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm12 = ymm12[2,0],ymm1[2,2],ymm12[6,4],ymm1[6,6]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm9 = xmm10[0,1],xmm9[2,3]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm9 = xmm9[3,0],mem[1,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm9 = ymm9[0,1,2],ymm12[3,4,5],ymm9[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm9 = ymm9[0,1,2],ymm12[3,4,5,6,7]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm10 = ymm0[2,0],ymm11[1,0],ymm0[6,4],ymm11[5,4]
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} ymm10 = ymm10[0,1,2,0,4,5,6,4]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm9 = ymm9[0,1,2,3,4,5],ymm10[6,7]
@@ -462,7 +462,7 @@ define void @load_i32_stride5_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2,3],ymm4[4,5],ymm3[6,7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm2, %xmm2
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm2[0],mem[1],xmm2[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4,5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm11[0,1,2,3,4,5,6],ymm0[7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX1-ONLY-NEXT: vmovaps %ymm5, (%rsi)
@@ -504,7 +504,7 @@ define void @load_i32_stride5_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
; AVX2-ONLY-NEXT: vinserti128 $1, 96(%rdi), %ymm0, %ymm8
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm8 = ymm3[0,1,2,3,4,5,6],ymm8[7]
; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm8 = ymm8[2,3,0,1,6,7,4,5]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm7 = ymm7[0,1,2],ymm8[3,4,5],ymm7[6,7]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm7 = ymm7[0,1,2],ymm8[3,4,5,6,7]
; AVX2-ONLY-NEXT: vpbroadcastq {{.*#+}} ymm8 = [21474836480,21474836480,21474836480,21474836480]
; AVX2-ONLY-NEXT: vpermd %ymm0, %ymm8, %ymm8
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5],ymm8[6,7]
@@ -514,14 +514,14 @@ define void @load_i32_stride5_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm9 = ymm9[3,0,2,2,7,4,6,6]
; AVX2-ONLY-NEXT: vpalignr {{.*#+}} ymm10 = ymm1[12,13,14,15],ymm2[0,1,2,3,4,5,6,7,8,9,10,11],ymm1[28,29,30,31],ymm2[16,17,18,19,20,21,22,23,24,25,26,27]
; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm10 = ymm10[0,3,2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm9 = ymm10[0,1,2],ymm9[3,4,5],ymm10[6,7]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm9 = ymm10[0,1,2],ymm9[3,4,5,6,7]
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm8 = ymm9[0,1,2,3,4,5],ymm8[6,7]
; AVX2-ONLY-NEXT: vperm2i128 {{.*#+}} ymm3 = ymm3[0,1],ymm4[0,1]
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3,4],ymm4[5],ymm3[6,7]
; AVX2-ONLY-NEXT: vmovdqa {{.*#+}} xmm4 = <4,1,6,u>
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5],ymm2[6,7]
; AVX2-ONLY-NEXT: vpermd %ymm1, %ymm4, %ymm1
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm3[3,4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm3[3,4,5,6,7]
; AVX2-ONLY-NEXT: vpbroadcastq {{.*#+}} ymm2 = [30064771074,30064771074,30064771074,30064771074]
; AVX2-ONLY-NEXT: vpermd %ymm0, %ymm2, %ymm0
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
@@ -887,7 +887,7 @@ define void @load_i32_stride5_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm3, %xmm3
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm3 = xmm3[0,1],mem[2],xmm3[3]
; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} xmm3 = xmm3[1,0]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2],ymm1[3,4,5],ymm3[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2],ymm1[3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps 128(%rdi), %ymm1
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm4 = ymm1[2,3,0,1]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm13 = ymm1[1,0],ymm4[0,0],ymm1[5,4],ymm4[4,4]
@@ -901,7 +901,7 @@ define void @load_i32_stride5_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm3, %xmm3
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm3 = xmm3[0,1],mem[2],xmm3[3]
; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} xmm3 = xmm3[1,0]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm13 = ymm3[0,1,2],ymm0[3,4,5],ymm3[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm13 = ymm3[0,1,2],ymm0[3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps 288(%rdi), %ymm0
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm3 = ymm0[2,3,0,1]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm14 = ymm0[1,0],ymm3[0,0],ymm0[5,4],ymm3[4,4]
@@ -911,7 +911,7 @@ define void @load_i32_stride5_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm14 = ymm6[0,0],ymm5[3,0],ymm6[4,4],ymm5[7,4]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm14 = ymm14[2,0],ymm5[2,2],ymm14[6,4],ymm5[6,6]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm10 = xmm10[3,0],mem[1,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm10 = ymm10[0,1,2],ymm14[3,4,5],ymm10[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm10 = ymm10[0,1,2],ymm14[3,4,5,6,7]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm14 = ymm1[2,0],ymm4[1,0],ymm1[6,4],ymm4[5,4]
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} ymm14 = ymm14[0,1,2,0,4,5,6,4]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm10 = ymm10[0,1,2,3,4,5],ymm14[6,7]
@@ -919,7 +919,7 @@ define void @load_i32_stride5_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm9 = ymm11[0,0],ymm2[3,0],ymm11[4,4],ymm2[7,4]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm9 = ymm9[2,0],ymm2[2,2],ymm9[6,4],ymm2[6,6]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm8 = xmm8[3,0],mem[1,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2],ymm9[3,4,5],ymm8[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2],ymm9[3,4,5,6,7]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm9 = ymm0[2,0],ymm3[1,0],ymm0[6,4],ymm3[5,4]
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} ymm9 = ymm9[0,1,2,0,4,5,6,4]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5],ymm9[6,7]
@@ -930,7 +930,7 @@ define void @load_i32_stride5_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: # ymm6 = ymm6[0,1,2,3],mem[4,5],ymm6[6,7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm6, %xmm6
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm6 = xmm6[0],mem[1],xmm6[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm6[0,1,2],ymm5[3,4,5],ymm6[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm6[0,1,2],ymm5[3,4,5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm4[0,1,2,3,4,5,6],ymm1[7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm5[0,1,2,3,4,5],ymm1[6,7]
; AVX1-ONLY-NEXT: vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 16-byte Folded Reload
@@ -938,7 +938,7 @@ define void @load_i32_stride5_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm7[0,1,2,3],ymm12[4,5],ymm7[6,7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm4, %xmm4
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm4 = xmm4[0],mem[1],xmm4[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm4[0,1,2],ymm2[3,4,5],ymm4[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm4[0,1,2],ymm2[3,4,5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3,4,5,6],ymm0[7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3,4,5],ymm0[6,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
@@ -1014,7 +1014,7 @@ define void @load_i32_stride5_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-ONLY-NEXT: vinserti128 $1, 96(%rdi), %ymm0, %ymm15
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm15 = ymm8[0,1,2,3,4,5,6],ymm15[7]
; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm15 = ymm15[2,3,0,1,6,7,4,5]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm13 = ymm13[0,1,2],ymm15[3,4,5],ymm13[6,7]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm13 = ymm13[0,1,2],ymm15[3,4,5,6,7]
; AVX2-ONLY-NEXT: vpbroadcastq {{.*#+}} ymm15 = [21474836480,21474836480,21474836480,21474836480]
; AVX2-ONLY-NEXT: vpermd %ymm0, %ymm15, %ymm7
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm13 = ymm13[0,1,2,3,4,5],ymm7[6,7]
@@ -1023,7 +1023,7 @@ define void @load_i32_stride5_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-ONLY-NEXT: vinserti128 $1, 256(%rdi), %ymm0, %ymm14
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm14 = ymm1[0,1,2,3,4,5,6],ymm14[7]
; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm14 = ymm14[2,3,0,1,6,7,4,5]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm7 = ymm7[0,1,2],ymm14[3,4,5],ymm7[6,7]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm7 = ymm7[0,1,2],ymm14[3,4,5,6,7]
; AVX2-ONLY-NEXT: vmovdqa 288(%rdi), %ymm14
; AVX2-ONLY-NEXT: vpermd %ymm14, %ymm15, %ymm15
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5],ymm15[6,7]
@@ -1031,7 +1031,7 @@ define void @load_i32_stride5_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm15 = ymm15[3,0,2,2,7,4,6,6]
; AVX2-ONLY-NEXT: vpalignr {{.*#+}} ymm10 = ymm3[12,13,14,15],ymm5[0,1,2,3,4,5,6,7,8,9,10,11],ymm3[28,29,30,31],ymm5[16,17,18,19,20,21,22,23,24,25,26,27]
; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm10 = ymm10[0,3,2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm10 = ymm10[0,1,2],ymm15[3,4,5],ymm10[6,7]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm10 = ymm10[0,1,2],ymm15[3,4,5,6,7]
; AVX2-ONLY-NEXT: vpbroadcastq {{.*#+}} ymm15 = [25769803777,25769803777,25769803777,25769803777]
; AVX2-ONLY-NEXT: vpermd %ymm0, %ymm15, %ymm11
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm10 = ymm10[0,1,2,3,4,5],ymm11[6,7]
@@ -1039,7 +1039,7 @@ define void @load_i32_stride5_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm11 = ymm11[3,0,2,2,7,4,6,6]
; AVX2-ONLY-NEXT: vpalignr {{.*#+}} ymm12 = ymm6[12,13,14,15],ymm4[0,1,2,3,4,5,6,7,8,9,10,11],ymm6[28,29,30,31],ymm4[16,17,18,19,20,21,22,23,24,25,26,27]
; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm12 = ymm12[0,3,2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm11 = ymm12[0,1,2],ymm11[3,4,5],ymm12[6,7]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm11 = ymm12[0,1,2],ymm11[3,4,5,6,7]
; AVX2-ONLY-NEXT: vpermd %ymm14, %ymm15, %ymm12
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm11 = ymm11[0,1,2,3,4,5],ymm12[6,7]
; AVX2-ONLY-NEXT: vperm2i128 {{.*#+}} ymm8 = ymm8[0,1],ymm9[0,1]
@@ -1047,7 +1047,7 @@ define void @load_i32_stride5_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm3 = ymm5[0,1,2,3],ymm3[4,5],ymm5[6,7]
; AVX2-ONLY-NEXT: vmovdqa {{.*#+}} xmm5 = <4,1,6,u>
; AVX2-ONLY-NEXT: vpermd %ymm3, %ymm5, %ymm3
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2],ymm8[3,4,5],ymm3[6,7]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2],ymm8[3,4,5,6,7]
; AVX2-ONLY-NEXT: vpbroadcastq {{.*#+}} ymm8 = [30064771074,30064771074,30064771074,30064771074]
; AVX2-ONLY-NEXT: vpermd %ymm0, %ymm8, %ymm0
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm3[0,1,2,3,4,5],ymm0[6,7]
@@ -1055,7 +1055,7 @@ define void @load_i32_stride5_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-ONLY-NEXT: vpermd %ymm3, %ymm5, %ymm3
; AVX2-ONLY-NEXT: vperm2i128 {{.*#+}} ymm1 = ymm1[0,1],ymm2[0,1]
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5],ymm1[6,7]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm3[0,1,2],ymm1[3,4,5],ymm3[6,7]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm3[0,1,2],ymm1[3,4,5,6,7]
; AVX2-ONLY-NEXT: vpermd %ymm14, %ymm8, %ymm2
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6,7]
; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
@@ -1854,7 +1854,7 @@ define void @load_i32_stride5_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm2, %xmm2
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm2[0,1],mem[2],xmm2[3]
; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} xmm2 = xmm2[1,0]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2],ymm0[3,4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2],ymm0[3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps 288(%rdi), %ymm6
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm10 = ymm6[2,3,0,1]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm6[1,0],ymm10[0,0],ymm6[5,4],ymm10[4,4]
@@ -1872,7 +1872,7 @@ define void @load_i32_stride5_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm2, %xmm2
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm2[0,1],mem[2],xmm2[3]
; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} xmm2 = xmm2[1,0]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2],ymm0[3,4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2],ymm0[3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps 608(%rdi), %ymm8
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm7 = ymm8[2,3,0,1]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm4 = ymm8[1,0],ymm7[0,0],ymm8[5,4],ymm7[4,4]
@@ -1887,7 +1887,7 @@ define void @load_i32_stride5_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm4, %xmm4
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm4 = xmm4[0,1],mem[2],xmm4[3]
; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} xmm4 = xmm4[1,0]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm4[0,1,2],ymm0[3,4,5],ymm4[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm4[0,1,2],ymm0[3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps 128(%rdi), %ymm12
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm3 = ymm12[2,3,0,1]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm4 = ymm12[1,0],ymm3[0,0],ymm12[5,4],ymm3[4,4]
@@ -1904,7 +1904,7 @@ define void @load_i32_stride5_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm4, %xmm4
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm4 = xmm4[0,1],mem[2],xmm4[3]
; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} xmm4 = xmm4[1,0]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm4[0,1,2],ymm0[3,4,5],ymm4[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm4[0,1,2],ymm0[3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps 448(%rdi), %ymm4
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm9 = ymm4[2,3,0,1]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm4[1,0],ymm9[0,0],ymm4[5,4],ymm9[4,4]
@@ -1918,7 +1918,7 @@ define void @load_i32_stride5_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm1[0,0],ymm13[3,0],ymm1[4,4],ymm13[7,4]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm15[2,0],ymm13[2,2],ymm15[6,4],ymm13[6,6]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm0 = xmm0[3,0],mem[1,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm15[3,4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm15[3,4,5,6,7]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm6[2,0],ymm10[1,0],ymm6[6,4],ymm10[5,4]
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} ymm15 = ymm15[0,1,2,0,4,5,6,4]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm15[6,7]
@@ -1930,7 +1930,7 @@ define void @load_i32_stride5_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm2[0,0],ymm11[3,0],ymm2[4,4],ymm11[7,4]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm15[2,0],ymm11[2,2],ymm15[6,4],ymm11[6,6]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm0 = xmm0[3,0],mem[1,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm15[3,4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm15[3,4,5,6,7]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm8[2,0],ymm7[1,0],ymm8[6,4],ymm7[5,4]
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} ymm15 = ymm15[0,1,2,0,4,5,6,4]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm15[6,7]
@@ -1942,7 +1942,7 @@ define void @load_i32_stride5_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm10[0,0],ymm14[3,0],ymm10[4,4],ymm14[7,4]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm15[2,0],ymm14[2,2],ymm15[6,4],ymm14[6,6]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm0 = xmm0[3,0],mem[1,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm15[3,4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm15[3,4,5,6,7]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm12[2,0],ymm3[1,0],ymm12[6,4],ymm3[5,4]
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} ymm15 = ymm15[0,1,2,0,4,5,6,4]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm15[6,7]
@@ -1954,7 +1954,7 @@ define void @load_i32_stride5_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm1[0,0],ymm5[3,0],ymm1[4,4],ymm5[7,4]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm15[2,0],ymm5[2,2],ymm15[6,4],ymm5[6,6]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm0 = xmm0[3,0],mem[1,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm15[3,4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm15[3,4,5,6,7]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm4[2,0],ymm9[1,0],ymm4[6,4],ymm9[5,4]
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} ymm15 = ymm15[0,1,2,0,4,5,6,4]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm15[6,7]
@@ -1966,7 +1966,7 @@ define void @load_i32_stride5_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: # ymm15 = mem[0,1,2,3],ymm3[4,5],mem[6,7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm15, %xmm15
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm15 = xmm15[0],mem[1],xmm15[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm15[0,1,2],ymm7[3,4,5],ymm15[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm15[0,1,2],ymm7[3,4,5,6,7]
; AVX1-ONLY-NEXT: vblendps $127, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm3 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm3 = mem[0,1,2,3,4,5,6],ymm6[7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm7[0,1,2,3,4,5],ymm3[6,7]
@@ -1977,7 +1977,7 @@ define void @load_i32_stride5_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: # ymm7 = ymm7[0,1,2,3],mem[4,5],ymm7[6,7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm7, %xmm7
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm7 = xmm7[0],mem[1],xmm7[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm7[0,1,2],ymm6[3,4,5],ymm7[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm7[0,1,2],ymm6[3,4,5,6,7]
; AVX1-ONLY-NEXT: vblendps $127, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm2 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm2 = mem[0,1,2,3,4,5,6],ymm8[7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm6[0,1,2,3,4,5],ymm2[6,7]
@@ -1988,7 +1988,7 @@ define void @load_i32_stride5_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: # ymm6 = mem[0,1,2,3],ymm6[4,5],mem[6,7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm6, %xmm6
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm6 = xmm6[0],mem[1],xmm6[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm6[0,1,2],ymm5[3,4,5],ymm6[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm6[0,1,2],ymm5[3,4,5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm9[0,1,2,3,4,5,6],ymm4[7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm5[0,1,2,3,4,5],ymm1[6,7]
; AVX1-ONLY-NEXT: vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm4 # 16-byte Folded Reload
@@ -1998,7 +1998,7 @@ define void @load_i32_stride5_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: # ymm5 = ymm5[0,1,2,3],mem[4,5],ymm5[6,7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm5, %xmm5
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm5 = xmm5[0],mem[1],xmm5[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm5[0,1,2],ymm4[3,4,5],ymm5[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm5[0,1,2],ymm4[3,4,5,6,7]
; AVX1-ONLY-NEXT: vblendps $127, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm5 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm5 = mem[0,1,2,3,4,5,6],ymm12[7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5],ymm5[6,7]
@@ -2164,7 +2164,7 @@ define void @load_i32_stride5_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm5[0,1,2,3,4,5,6],ymm2[7]
; AVX2-ONLY-NEXT: vmovdqa %ymm5, %ymm7
; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm2 = ymm2[2,3,0,1,6,7,4,5]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
; AVX2-ONLY-NEXT: vmovdqa 288(%rdi), %ymm2
; AVX2-ONLY-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vpbroadcastq {{.*#+}} ymm0 = [21474836480,21474836480,21474836480,21474836480]
@@ -2176,7 +2176,7 @@ define void @load_i32_stride5_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-ONLY-NEXT: vinserti128 $1, 576(%rdi), %ymm0, %ymm2
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm6[0,1,2,3,4,5,6],ymm2[7]
; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm2 = ymm2[2,3,0,1,6,7,4,5]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
; AVX2-ONLY-NEXT: vmovdqa 608(%rdi), %ymm9
; AVX2-ONLY-NEXT: vpermd %ymm9, %ymm0, %ymm4
; AVX2-ONLY-NEXT: vmovdqu %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -2191,7 +2191,7 @@ define void @load_i32_stride5_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-ONLY-NEXT: vmovdqu %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm4 = ymm14[0,1,2,3,4,5,6],ymm4[7]
; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm4 = ymm4[2,3,0,1,6,7,4,5]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm4[3,4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm4[3,4,5,6,7]
; AVX2-ONLY-NEXT: vmovdqa 128(%rdi), %ymm14
; AVX2-ONLY-NEXT: vpermd %ymm14, %ymm0, %ymm10
; AVX2-ONLY-NEXT: vmovdqu %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -2204,7 +2204,7 @@ define void @load_i32_stride5_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-ONLY-NEXT: vmovdqu (%rsp), %ymm1 # 32-byte Reload
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm8 = ymm1[0,1,2,3,4,5,6],ymm8[7]
; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm8 = ymm8[2,3,0,1,6,7,4,5]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm8 = ymm4[0,1,2],ymm8[3,4,5],ymm4[6,7]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm8 = ymm4[0,1,2],ymm8[3,4,5,6,7]
; AVX2-ONLY-NEXT: vmovdqa 448(%rdi), %ymm4
; AVX2-ONLY-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vpermd %ymm4, %ymm0, %ymm0
@@ -2216,7 +2216,7 @@ define void @load_i32_stride5_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-ONLY-NEXT: vpalignr $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm8 # 32-byte Folded Reload
; AVX2-ONLY-NEXT: # ymm8 = mem[12,13,14,15],ymm12[0,1,2,3,4,5,6,7,8,9,10,11],mem[28,29,30,31],ymm12[16,17,18,19,20,21,22,23,24,25,26,27]
; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm8 = ymm8[0,3,2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm8[0,1,2],ymm0[3,4,5],ymm8[6,7]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm8[0,1,2],ymm0[3,4,5,6,7]
; AVX2-ONLY-NEXT: vpbroadcastq {{.*#+}} ymm8 = [25769803777,25769803777,25769803777,25769803777]
; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
; AVX2-ONLY-NEXT: vpermd %ymm11, %ymm8, %ymm10
@@ -2229,7 +2229,7 @@ define void @load_i32_stride5_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX2-ONLY-NEXT: vpalignr {{.*#+}} ymm10 = ymm6[12,13,14,15],ymm4[0,1,2,3,4,5,6,7,8,9,10,11],ymm6[28,29,30,31],ymm4[16,17,18,19,20,21,22,23,24,25,26,27]
; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm10 = ymm10[0,3,2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm10[0,1,2],ymm0[3,4,5],ymm10[6,7]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm10[0,1,2],ymm0[3,4,5,6,7]
; AVX2-ONLY-NEXT: vpermd %ymm9, %ymm8, %ymm10
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm10[6,7]
; AVX2-ONLY-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -2238,7 +2238,7 @@ define void @load_i32_stride5_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[3,0,2,2,7,4,6,6]
; AVX2-ONLY-NEXT: vpalignr {{.*#+}} ymm10 = ymm15[12,13,14,15],ymm5[0,1,2,3,4,5,6,7,8,9,10,11],ymm15[28,29,30,31],ymm5[16,17,18,19,20,21,22,23,24,25,26,27]
; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm10 = ymm10[0,3,2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm10[0,1,2],ymm0[3,4,5],ymm10[6,7]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm10[0,1,2],ymm0[3,4,5,6,7]
; AVX2-ONLY-NEXT: vpermd %ymm14, %ymm8, %ymm10
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm10[6,7]
; AVX2-ONLY-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -2249,7 +2249,7 @@ define void @load_i32_stride5_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-ONLY-NEXT: vpalignr {{.*#+}} ymm10 = ymm3[12,13,14,15],ymm15[0,1,2,3,4,5,6,7,8,9,10,11],ymm3[28,29,30,31],ymm15[16,17,18,19,20,21,22,23,24,25,26,27]
; AVX2-ONLY-NEXT: vmovdqa %ymm3, %ymm14
; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm10 = ymm10[0,3,2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm10[0,1,2],ymm0[3,4,5],ymm10[6,7]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm10[0,1,2],ymm0[3,4,5,6,7]
; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-ONLY-NEXT: vpermd %ymm1, %ymm8, %ymm8
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm8[6,7]
@@ -2261,7 +2261,7 @@ define void @load_i32_stride5_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm7 = ymm7[0,1,2,3,4],ymm13[5],ymm7[6,7]
; AVX2-ONLY-NEXT: vmovdqa {{.*#+}} xmm8 = <4,1,6,u>
; AVX2-ONLY-NEXT: vpermd %ymm0, %ymm8, %ymm0
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2],ymm7[3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2],ymm7[3,4,5,6,7]
; AVX2-ONLY-NEXT: vpbroadcastq {{.*#+}} ymm7 = [30064771074,30064771074,30064771074,30064771074]
; AVX2-ONLY-NEXT: vpermd %ymm11, %ymm7, %ymm3
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm3[6,7]
@@ -2270,7 +2270,7 @@ define void @load_i32_stride5_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-ONLY-NEXT: # ymm9 = mem[0,1],ymm12[0,1]
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm9 = ymm9[0,1,2,3,4],ymm12[5],ymm9[6,7]
; AVX2-ONLY-NEXT: vpermd %ymm3, %ymm8, %ymm3
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2],ymm9[3,4,5],ymm3[6,7]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2],ymm9[3,4,5,6,7]
; AVX2-ONLY-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm2 # 32-byte Folded Reload
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm2[6,7]
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm3 = ymm15[0,1,2,3],ymm14[4,5],ymm15[6,7]
@@ -2278,7 +2278,7 @@ define void @load_i32_stride5_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-ONLY-NEXT: # ymm9 = mem[0,1],ymm5[0,1]
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm9 = ymm9[0,1,2,3,4],ymm5[5],ymm9[6,7]
; AVX2-ONLY-NEXT: vpermd %ymm3, %ymm8, %ymm3
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2],ymm9[3,4,5],ymm3[6,7]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2],ymm9[3,4,5,6,7]
; AVX2-ONLY-NEXT: vpermd %ymm1, %ymm7, %ymm1
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm3[0,1,2,3,4,5],ymm1[6,7]
; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
@@ -2289,7 +2289,7 @@ define void @load_i32_stride5_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-ONLY-NEXT: vpblendd $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm5 # 32-byte Folded Reload
; AVX2-ONLY-NEXT: # ymm5 = ymm4[0,1,2,3],mem[4,5],ymm4[6,7]
; AVX2-ONLY-NEXT: vpermd %ymm5, %ymm8, %ymm5
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm3 = ymm5[0,1,2],ymm3[3,4,5],ymm5[6,7]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm3 = ymm5[0,1,2],ymm3[3,4,5,6,7]
; AVX2-ONLY-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm4 # 32-byte Folded Reload
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm4[6,7]
; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
@@ -3875,7 +3875,7 @@ define void @load_i32_stride5_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm2, %xmm2
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm2[0,1],mem[2],xmm2[3]
; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} xmm2 = xmm2[1,0]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps 288(%rdi), %ymm0
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm2 = ymm0[2,3,0,1]
@@ -3893,7 +3893,7 @@ define void @load_i32_stride5_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm2, %xmm2
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm2[0,1],mem[2],xmm2[3]
; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} xmm2 = xmm2[1,0]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps 608(%rdi), %ymm0
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm2 = ymm0[2,3,0,1]
@@ -3911,7 +3911,7 @@ define void @load_i32_stride5_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm2, %xmm2
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm2[0,1],mem[2],xmm2[3]
; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} xmm2 = xmm2[1,0]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps 928(%rdi), %ymm0
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm14 = ymm0[2,3,0,1]
@@ -3930,7 +3930,7 @@ define void @load_i32_stride5_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm2, %xmm2
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm2[0,1],mem[2],xmm2[3]
; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} xmm2 = xmm2[1,0]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps 1248(%rdi), %ymm0
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm2 = ymm0[2,3,0,1]
@@ -3948,7 +3948,7 @@ define void @load_i32_stride5_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm1, %xmm1
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1],mem[2],xmm1[3]
; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} xmm1 = xmm1[1,0]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps 1088(%rdi), %ymm2
; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm2[2,3,0,1]
@@ -3966,7 +3966,7 @@ define void @load_i32_stride5_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm1, %xmm1
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1],mem[2],xmm1[3]
; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} xmm1 = xmm1[1,0]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm0[3,4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm0[3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps 128(%rdi), %ymm4
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm10 = ymm4[2,3,0,1]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm4[1,0],ymm10[0,0],ymm4[5,4],ymm10[4,4]
@@ -3982,7 +3982,7 @@ define void @load_i32_stride5_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm2, %xmm2
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm2[0,1],mem[2],xmm2[3]
; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} xmm2 = xmm2[1,0]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm1[3,4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm1[3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps 448(%rdi), %ymm6
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm12 = ymm6[2,3,0,1]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm5 = ymm6[1,0],ymm12[0,0],ymm6[5,4],ymm12[4,4]
@@ -3998,7 +3998,7 @@ define void @load_i32_stride5_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm5, %xmm5
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm5 = xmm5[0,1],mem[2],xmm5[3]
; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} xmm5 = xmm5[1,0]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm5[0,1,2],ymm2[3,4,5],ymm5[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm5[0,1,2],ymm2[3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps 768(%rdi), %ymm8
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm3 = ymm8[2,3,0,1]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm8[1,0],ymm3[0,0],ymm8[5,4],ymm3[4,4]
@@ -4014,7 +4014,7 @@ define void @load_i32_stride5_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm9[0,0],ymm1[3,0],ymm9[4,4],ymm1[7,4]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm15[2,0],ymm1[2,2],ymm15[6,4],ymm1[6,6]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm0 = xmm0[3,0],mem[1,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm15[3,4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm15[3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX1-ONLY-NEXT: vshufps $18, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm15 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm15 = ymm1[2,0],mem[1,0],ymm1[6,4],mem[5,4]
@@ -4029,7 +4029,7 @@ define void @load_i32_stride5_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm1[0,0],ymm2[3,0],ymm1[4,4],ymm2[7,4]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm15[2,0],ymm2[2,2],ymm15[6,4],ymm2[6,6]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm0 = xmm0[3,0],mem[1,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm15[3,4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm15[3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX1-ONLY-NEXT: vshufps $18, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm15 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm15 = ymm1[2,0],mem[1,0],ymm1[6,4],mem[5,4]
@@ -4044,7 +4044,7 @@ define void @load_i32_stride5_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm2[0,0],ymm1[3,0],ymm2[4,4],ymm1[7,4]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm15[2,0],ymm1[2,2],ymm15[6,4],ymm1[6,6]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm0 = xmm0[3,0],mem[1,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm15[3,4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm15[3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm1[2,0],ymm14[1,0],ymm1[6,4],ymm14[5,4]
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} ymm15 = ymm15[0,1,2,0,4,5,6,4]
@@ -4058,7 +4058,7 @@ define void @load_i32_stride5_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm15[0,0],ymm1[3,0],ymm15[4,4],ymm1[7,4]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm15[2,0],ymm1[2,2],ymm15[6,4],ymm1[6,6]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm0 = xmm0[3,0],mem[1,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm15[3,4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm15[3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX1-ONLY-NEXT: vshufps $18, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm15 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm15 = ymm1[2,0],mem[1,0],ymm1[6,4],mem[5,4]
@@ -4073,7 +4073,7 @@ define void @load_i32_stride5_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm14[0,0],ymm1[3,0],ymm14[4,4],ymm1[7,4]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm15[2,0],ymm1[2,2],ymm15[6,4],ymm1[6,6]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm0 = xmm0[3,0],mem[1,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm15[3,4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm15[3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX1-ONLY-NEXT: vshufps $18, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm15 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm15 = ymm1[2,0],mem[1,0],ymm1[6,4],mem[5,4]
@@ -4087,7 +4087,7 @@ define void @load_i32_stride5_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm5[0,0],ymm13[3,0],ymm5[4,4],ymm13[7,4]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm15[2,0],ymm13[2,2],ymm15[6,4],ymm13[6,6]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm0 = xmm0[3,0],mem[1,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm15[3,4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm15[3,4,5,6,7]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm8[2,0],ymm3[1,0],ymm8[6,4],ymm3[5,4]
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} ymm15 = ymm15[0,1,2,0,4,5,6,4]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm15[6,7]
@@ -4099,7 +4099,7 @@ define void @load_i32_stride5_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm2[0,0],ymm11[3,0],ymm2[4,4],ymm11[7,4]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm15[2,0],ymm11[2,2],ymm15[6,4],ymm11[6,6]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm0 = xmm0[3,0],mem[1,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm15[3,4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm15[3,4,5,6,7]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm6[2,0],ymm12[1,0],ymm6[6,4],ymm12[5,4]
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} ymm15 = ymm15[0,1,2,0,4,5,6,4]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm15[6,7]
@@ -4111,7 +4111,7 @@ define void @load_i32_stride5_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm1[0,0],ymm7[3,0],ymm1[4,4],ymm7[7,4]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm15[2,0],ymm7[2,2],ymm15[6,4],ymm7[6,6]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm0 = xmm0[3,0],mem[1,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm15[3,4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm15[3,4,5,6,7]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm4[2,0],ymm10[1,0],ymm4[6,4],ymm10[5,4]
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} ymm15 = ymm15[0,1,2,0,4,5,6,4]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm15 = ymm0[0,1,2,3,4,5],ymm15[6,7]
@@ -4122,7 +4122,7 @@ define void @load_i32_stride5_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: # ymm7 = mem[0,1,2,3],ymm7[4,5],mem[6,7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm7, %xmm7
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm7 = xmm7[0],mem[1],xmm7[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm7[0,1,2],ymm0[3,4,5],ymm7[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm7[0,1,2],ymm0[3,4,5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm10[0,1,2,3,4,5,6],ymm4[7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm3[6,7]
; AVX1-ONLY-NEXT: vmovups (%rsp), %ymm1 # 32-byte Reload
@@ -4133,7 +4133,7 @@ define void @load_i32_stride5_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: # ymm4 = ymm4[0,1,2,3],mem[4,5],ymm4[6,7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm4, %xmm4
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm4 = xmm4[0],mem[1],xmm4[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3,4,5],ymm4[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX1-ONLY-NEXT: vblendps $127, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm4 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm4 = mem[0,1,2,3,4,5,6],ymm1[7]
@@ -4145,7 +4145,7 @@ define void @load_i32_stride5_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: # ymm7 = mem[0,1,2,3],ymm7[4,5],mem[6,7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm7, %xmm7
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm7 = xmm7[0],mem[1],xmm7[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm7[0,1,2],ymm4[3,4,5],ymm7[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm7[0,1,2],ymm4[3,4,5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm12[0,1,2,3,4,5,6],ymm6[7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm4[0,1,2,3,4,5],ymm1[6,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
@@ -4157,7 +4157,7 @@ define void @load_i32_stride5_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: # ymm6 = ymm6[0,1,2,3],mem[4,5],ymm6[6,7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm6, %xmm6
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm6 = xmm6[0],mem[1],xmm6[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm6[0,1,2],ymm4[3,4,5],ymm6[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm6[0,1,2],ymm4[3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX1-ONLY-NEXT: vblendps $127, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm6 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm6 = mem[0,1,2,3,4,5,6],ymm2[7]
@@ -4169,7 +4169,7 @@ define void @load_i32_stride5_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: # ymm7 = mem[0,1,2,3],ymm7[4,5],mem[6,7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm7, %xmm7
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm7 = xmm7[0],mem[1],xmm7[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm7[0,1,2],ymm6[3,4,5],ymm7[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm7[0,1,2],ymm6[3,4,5,6,7]
; AVX1-ONLY-NEXT: vblendps $127, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm2 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm2 = mem[0,1,2,3,4,5,6],ymm8[7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm6[0,1,2,3,4,5],ymm2[6,7]
@@ -4182,7 +4182,7 @@ define void @load_i32_stride5_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: # ymm6 = ymm6[0,1,2,3],mem[4,5],ymm6[6,7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm6, %xmm6
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm6 = xmm6[0],mem[1],xmm6[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm6[0,1,2],ymm5[3,4,5],ymm6[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm6[0,1,2],ymm5[3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
; AVX1-ONLY-NEXT: vblendps $128, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm6 = ymm6[0,1,2,3,4,5,6],mem[7]
@@ -4195,7 +4195,7 @@ define void @load_i32_stride5_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: # ymm7 = mem[0,1,2,3],ymm7[4,5],mem[6,7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm7, %xmm7
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm7 = xmm7[0],mem[1],xmm7[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm7[0,1,2],ymm6[3,4,5],ymm7[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm7[0,1,2],ymm6[3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
; AVX1-ONLY-NEXT: vblendps $127, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm7 = mem[0,1,2,3,4,5,6],ymm7[7]
@@ -4209,7 +4209,7 @@ define void @load_i32_stride5_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: # ymm8 = ymm8[0,1,2,3],mem[4,5],ymm8[6,7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm8, %xmm8
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm8 = xmm8[0],mem[1],xmm8[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm8[0,1,2],ymm7[3,4,5],ymm8[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm8[0,1,2],ymm7[3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
; AVX1-ONLY-NEXT: vblendps $128, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm8 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm8 = ymm8[0,1,2,3,4,5,6],mem[7]
@@ -4523,7 +4523,7 @@ define void @load_i32_stride5_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-ONLY-NEXT: vmovdqa %ymm10, %ymm4
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm10[0,1,2,3,4,5,6],ymm1[7]
; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm1 = ymm1[2,3,0,1,6,7,4,5]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm0[0,1,2],ymm1[3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm0[0,1,2],ymm1[3,4,5,6,7]
; AVX2-ONLY-NEXT: vmovdqa 288(%rdi), %ymm3
; AVX2-ONLY-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vpbroadcastq {{.*#+}} ymm0 = [21474836480,21474836480,21474836480,21474836480]
@@ -4537,7 +4537,7 @@ define void @load_i32_stride5_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-ONLY-NEXT: vinserti128 $1, 576(%rdi), %ymm0, %ymm3
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm3 = ymm14[0,1,2,3,4,5,6],ymm3[7]
; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm3 = ymm3[2,3,0,1,6,7,4,5]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm3[3,4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm3[3,4,5,6,7]
; AVX2-ONLY-NEXT: vmovdqa 608(%rdi), %ymm3
; AVX2-ONLY-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vpermd %ymm3, %ymm0, %ymm3
@@ -4551,7 +4551,7 @@ define void @load_i32_stride5_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm3 = ymm5[0,1,2,3,4,5,6],ymm3[7]
; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm3 = ymm3[2,3,0,1,6,7,4,5]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm3[3,4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm3[3,4,5,6,7]
; AVX2-ONLY-NEXT: vmovdqa 928(%rdi), %ymm3
; AVX2-ONLY-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vpermd %ymm3, %ymm0, %ymm3
@@ -4564,7 +4564,7 @@ define void @load_i32_stride5_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-ONLY-NEXT: vinserti128 $1, 1216(%rdi), %ymm0, %ymm3
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm3 = ymm9[0,1,2,3,4,5,6],ymm3[7]
; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm3 = ymm3[2,3,0,1,6,7,4,5]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm3[3,4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm3[3,4,5,6,7]
; AVX2-ONLY-NEXT: vmovdqa 1248(%rdi), %ymm3
; AVX2-ONLY-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vpermd %ymm3, %ymm0, %ymm3
@@ -4575,7 +4575,7 @@ define void @load_i32_stride5_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-ONLY-NEXT: vinserti128 $1, 1056(%rdi), %ymm0, %ymm3
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm3 = ymm15[0,1,2,3,4,5,6],ymm3[7]
; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm3 = ymm3[2,3,0,1,6,7,4,5]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm3[3,4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm3[3,4,5,6,7]
; AVX2-ONLY-NEXT: vmovdqa 1088(%rdi), %ymm8
; AVX2-ONLY-NEXT: vpermd %ymm8, %ymm0, %ymm3
; AVX2-ONLY-NEXT: vmovdqu %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -4589,7 +4589,7 @@ define void @load_i32_stride5_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-ONLY-NEXT: vpblendd $127, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
; AVX2-ONLY-NEXT: # ymm3 = mem[0,1,2,3,4,5,6],ymm3[7]
; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm3 = ymm3[2,3,0,1,6,7,4,5]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm3 = ymm1[0,1,2],ymm3[3,4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm3 = ymm1[0,1,2],ymm3[3,4,5,6,7]
; AVX2-ONLY-NEXT: vmovdqa 128(%rdi), %ymm1
; AVX2-ONLY-NEXT: vpermd %ymm1, %ymm0, %ymm15
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm15[6,7]
@@ -4601,7 +4601,7 @@ define void @load_i32_stride5_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-ONLY-NEXT: vpblendd $127, (%rsp), %ymm15, %ymm15 # 32-byte Folded Reload
; AVX2-ONLY-NEXT: # ymm15 = mem[0,1,2,3,4,5,6],ymm15[7]
; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm15 = ymm15[2,3,0,1,6,7,4,5]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm15 = ymm3[0,1,2],ymm15[3,4,5],ymm3[6,7]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm15 = ymm3[0,1,2],ymm15[3,4,5,6,7]
; AVX2-ONLY-NEXT: vmovdqa 448(%rdi), %ymm3
; AVX2-ONLY-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vpermd %ymm3, %ymm0, %ymm13
@@ -4615,7 +4615,7 @@ define void @load_i32_stride5_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm13 = ymm14[0,1,2,3,4,5,6],ymm13[7]
; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm13 = ymm13[2,3,0,1,6,7,4,5]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm13 = ymm2[0,1,2],ymm13[3,4,5],ymm2[6,7]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm13 = ymm2[0,1,2],ymm13[3,4,5,6,7]
; AVX2-ONLY-NEXT: vmovdqa 768(%rdi), %ymm2
; AVX2-ONLY-NEXT: vpermd %ymm2, %ymm0, %ymm0
; AVX2-ONLY-NEXT: vmovdqa %ymm2, %ymm3
@@ -4628,7 +4628,7 @@ define void @load_i32_stride5_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-ONLY-NEXT: vpalignr {{.*#+}} ymm13 = ymm12[12,13,14,15],ymm2[0,1,2,3,4,5,6,7,8,9,10,11],ymm12[28,29,30,31],ymm2[16,17,18,19,20,21,22,23,24,25,26,27]
; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm13 = ymm13[0,3,2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm13 = ymm13[0,1,2],ymm0[3,4,5],ymm13[6,7]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm13 = ymm13[0,1,2],ymm0[3,4,5,6,7]
; AVX2-ONLY-NEXT: vpbroadcastq {{.*#+}} ymm0 = [25769803777,25769803777,25769803777,25769803777]
; AVX2-ONLY-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm13[0,1,2,3,4,5],ymm15[6,7]
@@ -4640,7 +4640,7 @@ define void @load_i32_stride5_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-ONLY-NEXT: vpalignr {{.*#+}} ymm15 = ymm10[12,13,14,15],ymm2[0,1,2,3,4,5,6,7,8,9,10,11],ymm10[28,29,30,31],ymm2[16,17,18,19,20,21,22,23,24,25,26,27]
; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm15 = ymm15[0,3,2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm13 = ymm15[0,1,2],ymm13[3,4,5],ymm15[6,7]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm13 = ymm15[0,1,2],ymm13[3,4,5,6,7]
; AVX2-ONLY-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm13[0,1,2,3,4,5],ymm15[6,7]
; AVX2-ONLY-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -4651,7 +4651,7 @@ define void @load_i32_stride5_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-ONLY-NEXT: vpalignr $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm15 # 32-byte Folded Reload
; AVX2-ONLY-NEXT: # ymm15 = mem[12,13,14,15],ymm2[0,1,2,3,4,5,6,7,8,9,10,11],mem[28,29,30,31],ymm2[16,17,18,19,20,21,22,23,24,25,26,27]
; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm15 = ymm15[0,3,2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm13 = ymm15[0,1,2],ymm13[3,4,5],ymm15[6,7]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm13 = ymm15[0,1,2],ymm13[3,4,5,6,7]
; AVX2-ONLY-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm13[0,1,2,3,4,5],ymm15[6,7]
; AVX2-ONLY-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -4661,7 +4661,7 @@ define void @load_i32_stride5_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm13 = ymm13[3,0,2,2,7,4,6,6]
; AVX2-ONLY-NEXT: vpalignr {{.*#+}} ymm15 = ymm7[12,13,14,15],ymm6[0,1,2,3,4,5,6,7,8,9,10,11],ymm7[28,29,30,31],ymm6[16,17,18,19,20,21,22,23,24,25,26,27]
; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm15 = ymm15[0,3,2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm13 = ymm15[0,1,2],ymm13[3,4,5],ymm15[6,7]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm13 = ymm15[0,1,2],ymm13[3,4,5,6,7]
; AVX2-ONLY-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm13[0,1,2,3,4,5],ymm15[6,7]
; AVX2-ONLY-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -4673,7 +4673,7 @@ define void @load_i32_stride5_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-ONLY-NEXT: vpalignr $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm15 # 32-byte Folded Reload
; AVX2-ONLY-NEXT: # ymm15 = mem[12,13,14,15],ymm2[0,1,2,3,4,5,6,7,8,9,10,11],mem[28,29,30,31],ymm2[16,17,18,19,20,21,22,23,24,25,26,27]
; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm15 = ymm15[0,3,2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm13 = ymm15[0,1,2],ymm13[3,4,5],ymm15[6,7]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm13 = ymm15[0,1,2],ymm13[3,4,5,6,7]
; AVX2-ONLY-NEXT: vpermd %ymm8, %ymm0, %ymm15
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm13[0,1,2,3,4,5],ymm15[6,7]
; AVX2-ONLY-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -4684,7 +4684,7 @@ define void @load_i32_stride5_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-ONLY-NEXT: vmovdqa %ymm11, %ymm14
; AVX2-ONLY-NEXT: vmovdqa %ymm9, %ymm11
; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm15 = ymm15[0,3,2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm13 = ymm15[0,1,2],ymm13[3,4,5],ymm15[6,7]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm13 = ymm15[0,1,2],ymm13[3,4,5,6,7]
; AVX2-ONLY-NEXT: vpermd %ymm3, %ymm0, %ymm15
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm13[0,1,2,3,4,5],ymm15[6,7]
; AVX2-ONLY-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -4696,7 +4696,7 @@ define void @load_i32_stride5_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-ONLY-NEXT: vpalignr {{.*#+}} ymm15 = ymm3[12,13,14,15],ymm2[0,1,2,3,4,5,6,7,8,9,10,11],ymm3[28,29,30,31],ymm2[16,17,18,19,20,21,22,23,24,25,26,27]
; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm15 = ymm15[0,3,2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm13 = ymm15[0,1,2],ymm13[3,4,5],ymm15[6,7]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm13 = ymm15[0,1,2],ymm13[3,4,5,6,7]
; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
; AVX2-ONLY-NEXT: vpermd %ymm5, %ymm0, %ymm15
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm4 = ymm13[0,1,2,3,4,5],ymm15[6,7]
@@ -4709,7 +4709,7 @@ define void @load_i32_stride5_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
; AVX2-ONLY-NEXT: vpalignr {{.*#+}} ymm15 = ymm4[12,13,14,15],ymm6[0,1,2,3,4,5,6,7,8,9,10,11],ymm4[28,29,30,31],ymm6[16,17,18,19,20,21,22,23,24,25,26,27]
; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm15 = ymm15[0,3,2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm13 = ymm15[0,1,2],ymm13[3,4,5],ymm15[6,7]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm13 = ymm15[0,1,2],ymm13[3,4,5,6,7]
; AVX2-ONLY-NEXT: vpermd %ymm1, %ymm0, %ymm0
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm15 = ymm13[0,1,2,3,4,5],ymm0[6,7]
; AVX2-ONLY-NEXT: vperm2i128 {{.*#+}} ymm0 = ymm10[0,1],ymm7[0,1]
@@ -4717,7 +4717,7 @@ define void @load_i32_stride5_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm4 = ymm6[0,1,2,3],ymm4[4,5],ymm6[6,7]
; AVX2-ONLY-NEXT: vmovdqa {{.*#+}} xmm7 = <4,1,6,u>
; AVX2-ONLY-NEXT: vpermd %ymm4, %ymm7, %ymm4
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm4[0,1,2],ymm0[3,4,5],ymm4[6,7]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm4[0,1,2],ymm0[3,4,5,6,7]
; AVX2-ONLY-NEXT: vpbroadcastq {{.*#+}} ymm6 = [30064771074,30064771074,30064771074,30064771074]
; AVX2-ONLY-NEXT: vpermd %ymm1, %ymm6, %ymm1
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
@@ -4729,14 +4729,14 @@ define void @load_i32_stride5_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-ONLY-NEXT: # ymm4 = mem[0,1],ymm10[0,1]
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm4 = ymm4[0,1,2,3,4],ymm10[5],ymm4[6,7]
; AVX2-ONLY-NEXT: vpermd %ymm1, %ymm7, %ymm1
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm4[3,4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm4[3,4,5,6,7]
; AVX2-ONLY-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm4 # 32-byte Folded Reload
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm13 = ymm1[0,1,2,3,4,5],ymm4[6,7]
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm3[4,5],ymm2[6,7]
; AVX2-ONLY-NEXT: vperm2i128 {{.*#+}} ymm4 = ymm9[0,1],ymm8[0,1]
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm4 = ymm4[0,1,2,3,4],ymm8[5],ymm4[6,7]
; AVX2-ONLY-NEXT: vpermd %ymm1, %ymm7, %ymm1
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm4[3,4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm4[3,4,5,6,7]
; AVX2-ONLY-NEXT: vpermd %ymm5, %ymm6, %ymm3
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm3[6,7]
; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
@@ -4747,7 +4747,7 @@ define void @load_i32_stride5_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-ONLY-NEXT: # ymm4 = mem[0,1],ymm8[0,1]
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm4 = ymm4[0,1,2,3,4],ymm8[5],ymm4[6,7]
; AVX2-ONLY-NEXT: vpermd %ymm3, %ymm7, %ymm3
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2],ymm4[3,4,5],ymm3[6,7]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2],ymm4[3,4,5,6,7]
; AVX2-ONLY-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm4 # 32-byte Folded Reload
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm4 = ymm3[0,1,2,3,4,5],ymm4[6,7]
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm3 = ymm11[0,1,2,3],ymm14[4,5],ymm11[6,7]
@@ -4755,7 +4755,7 @@ define void @load_i32_stride5_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-ONLY-NEXT: # ymm8 = mem[0,1],ymm12[0,1]
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm8 = ymm8[0,1,2,3,4],ymm12[5],ymm8[6,7]
; AVX2-ONLY-NEXT: vpermd %ymm3, %ymm7, %ymm3
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2],ymm8[3,4,5],ymm3[6,7]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2],ymm8[3,4,5,6,7]
; AVX2-ONLY-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm2 # 32-byte Folded Reload
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm2[6,7]
; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
@@ -4766,7 +4766,7 @@ define void @load_i32_stride5_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-ONLY-NEXT: # ymm8 = mem[0,1],ymm10[0,1]
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm8 = ymm8[0,1,2,3,4],ymm10[5],ymm8[6,7]
; AVX2-ONLY-NEXT: vpermd %ymm3, %ymm7, %ymm3
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2],ymm8[3,4,5],ymm3[6,7]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2],ymm8[3,4,5,6,7]
; AVX2-ONLY-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm8 # 32-byte Folded Reload
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm8[6,7]
; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
@@ -4777,7 +4777,7 @@ define void @load_i32_stride5_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-ONLY-NEXT: # ymm10 = mem[0,1],ymm5[0,1]
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm10 = ymm10[0,1,2,3,4],ymm5[5],ymm10[6,7]
; AVX2-ONLY-NEXT: vpermd %ymm8, %ymm7, %ymm8
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm8 = ymm8[0,1,2],ymm10[3,4,5],ymm8[6,7]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm8 = ymm8[0,1,2],ymm10[3,4,5,6,7]
; AVX2-ONLY-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm9 # 32-byte Folded Reload
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5],ymm9[6,7]
; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
@@ -4788,7 +4788,7 @@ define void @load_i32_stride5_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-ONLY-NEXT: vperm2i128 $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm9 # 32-byte Folded Reload
; AVX2-ONLY-NEXT: # ymm9 = mem[0,1],ymm10[0,1]
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm9 = ymm9[0,1,2,3,4],ymm10[5],ymm9[6,7]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm7 = ymm7[0,1,2],ymm9[3,4,5],ymm7[6,7]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm7 = ymm7[0,1,2],ymm9[3,4,5,6,7]
; AVX2-ONLY-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm6 = ymm7[0,1,2,3,4,5],ymm6[6,7]
; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-6.ll b/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-6.ll
index 6daf83dbb7e91..8848f259f79a4 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-6.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-6.ll
@@ -87,7 +87,7 @@ define void @load_i32_stride6_vf2(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
; AVX2-ONLY-NEXT: vpermilps {{.*#+}} xmm1 = xmm1[3,3,3,3]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0],xmm3[1],xmm1[2,3]
; AVX2-ONLY-NEXT: vmovaps {{.*#+}} xmm3 = <4,2,u,u>
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],mem[4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],mem[4,5,6,7]
; AVX2-ONLY-NEXT: vpermps %ymm0, %ymm3, %ymm3
; AVX2-ONLY-NEXT: vmovaps {{.*#+}} xmm6 = <5,3,u,u>
; AVX2-ONLY-NEXT: vpermps %ymm0, %ymm6, %ymm0
@@ -116,7 +116,7 @@ define void @load_i32_stride6_vf2(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0],xmm2[1],xmm0[2,3]
; AVX512F-SLOW-NEXT: vmovaps {{.*#+}} xmm2 = <4,2,u,u>
; AVX512F-SLOW-NEXT: vmovaps 32(%rdi), %ymm5
-; AVX512F-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],mem[4,5],ymm5[6,7]
+; AVX512F-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],mem[4,5,6,7]
; AVX512F-SLOW-NEXT: vpermps %ymm5, %ymm2, %ymm2
; AVX512F-SLOW-NEXT: vmovaps {{.*#+}} xmm6 = <5,3,u,u>
; AVX512F-SLOW-NEXT: vpermps %ymm5, %ymm6, %ymm5
@@ -145,7 +145,7 @@ define void @load_i32_stride6_vf2(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
; AVX512F-FAST-NEXT: vpermi2d %xmm3, %xmm1, %xmm5
; AVX512F-FAST-NEXT: vmovaps {{.*#+}} xmm1 = <4,2,u,u>
; AVX512F-FAST-NEXT: vmovaps 32(%rdi), %ymm3
-; AVX512F-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],mem[4,5],ymm3[6,7]
+; AVX512F-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],mem[4,5,6,7]
; AVX512F-FAST-NEXT: vpermps %ymm3, %ymm1, %ymm1
; AVX512F-FAST-NEXT: vmovaps {{.*#+}} xmm6 = <5,3,u,u>
; AVX512F-FAST-NEXT: vpermps %ymm3, %ymm6, %ymm3
@@ -174,7 +174,7 @@ define void @load_i32_stride6_vf2(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
; AVX512BW-SLOW-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0],xmm2[1],xmm0[2,3]
; AVX512BW-SLOW-NEXT: vmovaps {{.*#+}} xmm2 = <4,2,u,u>
; AVX512BW-SLOW-NEXT: vmovaps 32(%rdi), %ymm5
-; AVX512BW-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],mem[4,5],ymm5[6,7]
+; AVX512BW-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],mem[4,5,6,7]
; AVX512BW-SLOW-NEXT: vpermps %ymm5, %ymm2, %ymm2
; AVX512BW-SLOW-NEXT: vmovaps {{.*#+}} xmm6 = <5,3,u,u>
; AVX512BW-SLOW-NEXT: vpermps %ymm5, %ymm6, %ymm5
@@ -203,7 +203,7 @@ define void @load_i32_stride6_vf2(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
; AVX512BW-FAST-NEXT: vpermi2d %xmm3, %xmm1, %xmm5
; AVX512BW-FAST-NEXT: vmovaps {{.*#+}} xmm1 = <4,2,u,u>
; AVX512BW-FAST-NEXT: vmovaps 32(%rdi), %ymm3
-; AVX512BW-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],mem[4,5],ymm3[6,7]
+; AVX512BW-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],mem[4,5,6,7]
; AVX512BW-FAST-NEXT: vpermps %ymm3, %ymm1, %ymm1
; AVX512BW-FAST-NEXT: vmovaps {{.*#+}} xmm6 = <5,3,u,u>
; AVX512BW-FAST-NEXT: vpermps %ymm3, %ymm6, %ymm3
@@ -358,7 +358,7 @@ define void @load_i32_stride6_vf4(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm8 = xmm4[0,1],xmm7[2,3]
; AVX2-ONLY-NEXT: vpshufd {{.*#+}} xmm8 = xmm8[0,1,0,2]
; AVX2-ONLY-NEXT: vmovdqa {{.*#+}} xmm9 = <4,2,u,u>
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5],ymm2[6,7]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT: vpermd %ymm1, %ymm9, %ymm2
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm2 = xmm2[0,1],xmm8[2,3]
; AVX2-ONLY-NEXT: vpmovzxdq {{.*#+}} xmm4 = xmm4[0],zero,xmm4[1],zero
@@ -542,8 +542,8 @@ define void @load_i32_stride6_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
; AVX1-ONLY-LABEL: load_i32_stride6_vf8:
; AVX1-ONLY: # %bb.0:
; AVX1-ONLY-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; AVX1-ONLY-NEXT: vmovaps 128(%rdi), %ymm3
-; AVX1-ONLY-NEXT: vmovaps 160(%rdi), %ymm4
+; AVX1-ONLY-NEXT: vmovapd 160(%rdi), %ymm3
+; AVX1-ONLY-NEXT: vmovapd 128(%rdi), %ymm4
; AVX1-ONLY-NEXT: vmovaps 32(%rdi), %ymm6
; AVX1-ONLY-NEXT: vmovaps (%rdi), %ymm7
; AVX1-ONLY-NEXT: vmovaps 96(%rdi), %ymm0
@@ -555,21 +555,20 @@ define void @load_i32_stride6_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm8, %xmm9
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm10 = xmm8[0,1],xmm9[2,3]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm10 = xmm10[0,2],xmm9[0,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm10[0,1,2],ymm2[3,4,5],ymm10[6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm10 = ymm4[0,1,2,3],ymm3[4,5],ymm4[6,7]
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm11 = ymm10[2,3,0,1]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm12 = ymm11[2,0],ymm10[0,0],ymm11[6,4],ymm10[4,4]
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} ymm12 = ymm12[0,1,2,0,4,5,6,4]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm12[6,7]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm12 = ymm0[3,0],ymm5[1,0],ymm0[7,4],ymm5[5,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm5 = ymm12[2,0],ymm5[2,3],ymm12[6,4],ymm5[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm10[0,1,2],ymm2[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm10 = ymm4[2,3],ymm3[0,1]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm11 = ymm10[0],ymm4[0],ymm10[3],ymm4[2]
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} ymm11 = ymm11[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm11[6,7]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm11 = ymm0[3,0],ymm5[1,0],ymm0[7,4],ymm5[5,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm5 = ymm11[2,0],ymm5[2,3],ymm11[6,4],ymm5[6,7]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm8 = xmm8[1,0],xmm9[3,0]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm8 = xmm8[0,2],xmm9[1,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm8[0,1,2],ymm5[3,4,5],ymm8[6,7]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm8 = ymm11[3,0],ymm10[1,0],ymm11[7,4],ymm10[5,4]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm8[0,1,2],ymm5[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm8 = ymm10[3,1],ymm4[1,3],ymm10[7,5],ymm4[5,7]
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} ymm8 = ymm8[0,1,2,0,4,5,6,4]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5],ymm8[6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm3[0,1,2,3],ymm4[4,5],ymm3[6,7]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm8 = ymm4[0,1],ymm3[2],ymm4[3]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm9 = ymm8[2,3,0,1]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm10 = ymm9[0,0],ymm8[2,0],ymm9[4,4],ymm8[6,4]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm6[0,1],ymm7[2,3],ymm6[4,5,6,7]
@@ -577,15 +576,15 @@ define void @load_i32_stride6_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm6 = xmm7[2,0],xmm11[2,3]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm12 = ymm0[2,1],ymm1[2,0],ymm0[6,5],ymm1[6,4]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm12 = ymm12[2,3,0,1]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2],ymm12[3,4],ymm6[5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2],ymm12[3,4,5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4],ymm10[5,6,7]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm8 = ymm9[0,1],ymm8[3,1],ymm9[4,5],ymm8[7,5]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm7 = xmm7[3,1],xmm11[3,3]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm9 = ymm0[3,1],ymm1[2,1],ymm0[7,5],ymm1[6,5]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm9 = ymm9[2,3,0,1]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2],ymm9[3,4],ymm7[5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2],ymm9[3,4,5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4],ymm8[5,6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm4[0,1],ymm3[2,3],ymm4[4,5,6,7]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm3 = ymm3[0],ymm4[1],ymm3[2,3]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm4 = ymm3[2,3,0,1]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm8 = ymm3[2,0],ymm4[0,0],ymm3[6,4],ymm4[4,4]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm8 = ymm4[0,2],ymm8[2,0],ymm4[4,6],ymm8[6,4]
@@ -630,39 +629,40 @@ define void @load_i32_stride6_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
; AVX2-SLOW-NEXT: vperm2f128 {{.*#+}} ymm8 = ymm7[0,1],ymm5[0,1]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5],ymm5[6,7]
; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm9 = ymm8[0,2,2,2,4,6,6,6]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm9[3,4,5],ymm2[6,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm9 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm9[3,4,5,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm9 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX2-SLOW-NEXT: vbroadcastsd {{.*#+}} ymm10 = [8589934596,8589934596,8589934596,8589934596]
; AVX2-SLOW-NEXT: vpermps %ymm9, %ymm10, %ymm10
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm10[6,7]
; AVX2-SLOW-NEXT: vmovaps {{.*#+}} xmm10 = <1,7,5,u>
; AVX2-SLOW-NEXT: vpermps %ymm6, %ymm10, %ymm6
; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm8 = ymm8[1,3,2,3,5,7,6,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2],ymm8[3,4,5],ymm6[6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2],ymm8[3,4,5,6,7]
; AVX2-SLOW-NEXT: vbroadcastsd {{.*#+}} ymm8 = [12884901893,12884901893,12884901893,12884901893]
; AVX2-SLOW-NEXT: vpermps %ymm9, %ymm8, %ymm8
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5],ymm8[6,7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm8 = ymm7[0,1],ymm5[2,3],ymm7[4,5],ymm5[6,7]
; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm8 = ymm8[2,0,2,3,6,4,6,7]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm8 = ymm8[0,2,0,3]
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm9 = ymm3[2,3,2,3,6,7,6,7]
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm10 = ymm4[0,0,2,3,4,4,6,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm9 = ymm9[0],ymm10[1,2,3],ymm9[4],ymm10[5,6,7]
-; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm9 = ymm9[0,3,2,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm8 = ymm9[0,1,2],ymm8[3,4],ymm9[5,6,7]
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm9 = ymm1[0,0,0,0,4,4,4,4]
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm10 = ymm0[0,0,2,3,4,4,6,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm9 = ymm10[0,1,2],ymm9[3],ymm10[4,5,6],ymm9[7]
-; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm9 = ymm9[0,1,0,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4],ymm9[5,6,7]
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm9 = ymm5[3,3,3,3,7,7,7,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm9 = ymm9[0],ymm7[1],ymm9[2,3,4],ymm7[5],ymm9[6,7]
-; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm9 = ymm9[0,2,0,3]
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm10 = ymm3[3,3,3,3,7,7,7,7]
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm11 = ymm4[0,1,3,3,4,5,7,7]
+; AVX2-SLOW-NEXT: vmovaps (%rdi), %xmm9
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} xmm10 = xmm9[2,3,2,3]
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm11 = ymm4[0,0,2,3,4,4,6,7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm10 = ymm10[0],ymm11[1,2,3],ymm10[4],ymm11[5,6,7]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm10 = ymm10[0,3,2,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm9 = ymm10[0,1,2],ymm9[3,4],ymm10[5,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm8 = ymm10[0,1,2],ymm8[3,4,5,6,7]
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm10 = ymm1[0,0,0,0,4,4,4,4]
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm11 = ymm0[0,0,2,3,4,4,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm10 = ymm11[0,1,2],ymm10[3],ymm11[4,5,6],ymm10[7]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm10 = ymm10[0,1,0,3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4],ymm10[5,6,7]
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm10 = ymm5[3,3,3,3,7,7,7,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm10 = ymm10[0],ymm7[1],ymm10[2,3,4],ymm7[5],ymm10[6,7]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm10 = ymm10[0,2,0,3]
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} xmm9 = xmm9[3,3,3,3]
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm11 = ymm4[0,1,3,3,4,5,7,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm9 = ymm9[0],ymm11[1,2,3],ymm9[4],ymm11[5,6,7]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm9 = ymm9[0,3,2,3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm9 = ymm9[0,1,2],ymm10[3,4,5,6,7]
; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm10 = ymm1[0,1,0,1,4,5,4,5]
; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm11 = ymm0[0,1,3,3,4,5,7,7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm10 = ymm11[0,1,2],ymm10[3],ymm11[4,5,6],ymm10[7]
@@ -673,7 +673,7 @@ define void @load_i32_stride6_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm7 = ymm5[0,1],ymm10[2,3],ymm5[4,5,6,7]
; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm7 = ymm7[0,1,0,2,4,5,4,6]
; AVX2-SLOW-NEXT: vmovaps {{.*#+}} xmm11 = <4,2,u,u>
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm4[0,1,2,3],ymm3[4,5],ymm4[6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm4[0,1,2,3],ymm3[4,5,6,7]
; AVX2-SLOW-NEXT: vpermps %ymm3, %ymm11, %ymm4
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm7[2,3,4,5,6,7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5,6,7]
@@ -714,39 +714,40 @@ define void @load_i32_stride6_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
; AVX2-FAST-NEXT: vperm2f128 {{.*#+}} ymm8 = ymm7[0,1],ymm5[0,1]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5],ymm5[6,7]
; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm9 = ymm8[0,2,2,2,4,6,6,6]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm9[3,4,5],ymm2[6,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm9 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm9[3,4,5,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm9 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX2-FAST-NEXT: vbroadcastsd {{.*#+}} ymm10 = [8589934596,8589934596,8589934596,8589934596]
; AVX2-FAST-NEXT: vpermps %ymm9, %ymm10, %ymm10
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm10[6,7]
; AVX2-FAST-NEXT: vmovaps {{.*#+}} xmm10 = <1,7,5,u>
; AVX2-FAST-NEXT: vpermps %ymm6, %ymm10, %ymm6
; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm8 = ymm8[1,3,2,3,5,7,6,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2],ymm8[3,4,5],ymm6[6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2],ymm8[3,4,5,6,7]
; AVX2-FAST-NEXT: vbroadcastsd {{.*#+}} ymm8 = [12884901893,12884901893,12884901893,12884901893]
; AVX2-FAST-NEXT: vpermps %ymm9, %ymm8, %ymm8
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5],ymm8[6,7]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm8 = ymm7[0,1],ymm5[2,3],ymm7[4,5],ymm5[6,7]
; AVX2-FAST-NEXT: vmovaps {{.*#+}} ymm9 = <u,u,u,4,2,u,u,u>
; AVX2-FAST-NEXT: vpermps %ymm8, %ymm9, %ymm8
-; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm9 = ymm3[2,3,2,3,6,7,6,7]
-; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm10 = ymm4[0,0,2,3,4,4,6,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm9 = ymm9[0],ymm10[1,2,3],ymm9[4],ymm10[5,6,7]
-; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm9 = ymm9[0,3,2,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm8 = ymm9[0,1,2],ymm8[3,4],ymm9[5,6,7]
-; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm9 = ymm1[0,0,0,0,4,4,4,4]
-; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm10 = ymm0[0,0,2,3,4,4,6,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm9 = ymm10[0,1,2],ymm9[3],ymm10[4,5,6],ymm9[7]
-; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm9 = ymm9[0,1,0,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4],ymm9[5,6,7]
-; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm9 = ymm5[3,3,3,3,7,7,7,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm9 = ymm9[0],ymm7[1],ymm9[2,3,4],ymm7[5],ymm9[6,7]
-; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm9 = ymm9[0,2,0,3]
-; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm10 = ymm3[3,3,3,3,7,7,7,7]
-; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm11 = ymm4[0,1,3,3,4,5,7,7]
+; AVX2-FAST-NEXT: vmovaps (%rdi), %xmm9
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} xmm10 = xmm9[2,3,2,3]
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm11 = ymm4[0,0,2,3,4,4,6,7]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm10 = ymm10[0],ymm11[1,2,3],ymm10[4],ymm11[5,6,7]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm10 = ymm10[0,3,2,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm9 = ymm10[0,1,2],ymm9[3,4],ymm10[5,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm8 = ymm10[0,1,2],ymm8[3,4,5,6,7]
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm10 = ymm1[0,0,0,0,4,4,4,4]
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm11 = ymm0[0,0,2,3,4,4,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm10 = ymm11[0,1,2],ymm10[3],ymm11[4,5,6],ymm10[7]
+; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm10 = ymm10[0,1,0,3]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4],ymm10[5,6,7]
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm10 = ymm5[3,3,3,3,7,7,7,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm10 = ymm10[0],ymm7[1],ymm10[2,3,4],ymm7[5],ymm10[6,7]
+; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm10 = ymm10[0,2,0,3]
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} xmm9 = xmm9[3,3,3,3]
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm11 = ymm4[0,1,3,3,4,5,7,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm9 = ymm9[0],ymm11[1,2,3],ymm9[4],ymm11[5,6,7]
+; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm9 = ymm9[0,3,2,3]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm9 = ymm9[0,1,2],ymm10[3,4,5,6,7]
; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm10 = ymm1[0,1,0,1,4,5,4,5]
; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm11 = ymm0[0,1,3,3,4,5,7,7]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm10 = ymm11[0,1,2],ymm10[3],ymm11[4,5,6],ymm10[7]
@@ -757,7 +758,7 @@ define void @load_i32_stride6_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm7 = ymm5[0,1],ymm10[2,3],ymm5[4,5,6,7]
; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm7 = ymm7[0,1,0,2,4,5,4,6]
; AVX2-FAST-NEXT: vmovaps {{.*#+}} xmm11 = <4,2,u,u>
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm4[0,1,2,3],ymm3[4,5],ymm4[6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm4[0,1,2,3],ymm3[4,5,6,7]
; AVX2-FAST-NEXT: vpermps %ymm3, %ymm11, %ymm4
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm7[2,3,4,5,6,7]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5,6,7]
@@ -798,39 +799,40 @@ define void @load_i32_stride6_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
; AVX2-FAST-PERLANE-NEXT: vperm2f128 {{.*#+}} ymm8 = ymm7[0,1],ymm5[0,1]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5],ymm5[6,7]
; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm9 = ymm8[0,2,2,2,4,6,6,6]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm9[3,4,5],ymm2[6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm9 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm9[3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm9 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vbroadcastsd {{.*#+}} ymm10 = [8589934596,8589934596,8589934596,8589934596]
; AVX2-FAST-PERLANE-NEXT: vpermps %ymm9, %ymm10, %ymm10
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm10[6,7]
; AVX2-FAST-PERLANE-NEXT: vmovaps {{.*#+}} xmm10 = <1,7,5,u>
; AVX2-FAST-PERLANE-NEXT: vpermps %ymm6, %ymm10, %ymm6
; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm8 = ymm8[1,3,2,3,5,7,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2],ymm8[3,4,5],ymm6[6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2],ymm8[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vbroadcastsd {{.*#+}} ymm8 = [12884901893,12884901893,12884901893,12884901893]
; AVX2-FAST-PERLANE-NEXT: vpermps %ymm9, %ymm8, %ymm8
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5],ymm8[6,7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm8 = ymm7[0,1],ymm5[2,3],ymm7[4,5],ymm5[6,7]
; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm8 = ymm8[2,0,2,3,6,4,6,7]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm8 = ymm8[0,2,0,3]
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm9 = ymm3[2,3,2,3,6,7,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm10 = ymm4[0,0,2,3,4,4,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm9 = ymm9[0],ymm10[1,2,3],ymm9[4],ymm10[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm9 = ymm9[0,3,2,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm8 = ymm9[0,1,2],ymm8[3,4],ymm9[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm9 = ymm1[0,0,0,0,4,4,4,4]
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm10 = ymm0[0,0,2,3,4,4,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm9 = ymm10[0,1,2],ymm9[3],ymm10[4,5,6],ymm9[7]
-; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm9 = ymm9[0,1,0,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4],ymm9[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm9 = ymm5[3,3,3,3,7,7,7,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm9 = ymm9[0],ymm7[1],ymm9[2,3,4],ymm7[5],ymm9[6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm9 = ymm9[0,2,0,3]
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm10 = ymm3[3,3,3,3,7,7,7,7]
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm11 = ymm4[0,1,3,3,4,5,7,7]
+; AVX2-FAST-PERLANE-NEXT: vmovaps (%rdi), %xmm9
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} xmm10 = xmm9[2,3,2,3]
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm11 = ymm4[0,0,2,3,4,4,6,7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm10 = ymm10[0],ymm11[1,2,3],ymm10[4],ymm11[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm10 = ymm10[0,3,2,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm9 = ymm10[0,1,2],ymm9[3,4],ymm10[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm8 = ymm10[0,1,2],ymm8[3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm10 = ymm1[0,0,0,0,4,4,4,4]
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm11 = ymm0[0,0,2,3,4,4,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm10 = ymm11[0,1,2],ymm10[3],ymm11[4,5,6],ymm10[7]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm10 = ymm10[0,1,0,3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4],ymm10[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm10 = ymm5[3,3,3,3,7,7,7,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm10 = ymm10[0],ymm7[1],ymm10[2,3,4],ymm7[5],ymm10[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm10 = ymm10[0,2,0,3]
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} xmm9 = xmm9[3,3,3,3]
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm11 = ymm4[0,1,3,3,4,5,7,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm9 = ymm9[0],ymm11[1,2,3],ymm9[4],ymm11[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm9 = ymm9[0,3,2,3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm9 = ymm9[0,1,2],ymm10[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm10 = ymm1[0,1,0,1,4,5,4,5]
; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm11 = ymm0[0,1,3,3,4,5,7,7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm10 = ymm11[0,1,2],ymm10[3],ymm11[4,5,6],ymm10[7]
@@ -841,7 +843,7 @@ define void @load_i32_stride6_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm7 = ymm5[0,1],ymm10[2,3],ymm5[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm7 = ymm7[0,1,0,2,4,5,4,6]
; AVX2-FAST-PERLANE-NEXT: vmovaps {{.*#+}} xmm11 = <4,2,u,u>
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm4[0,1,2,3],ymm3[4,5],ymm4[6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm4[0,1,2,3],ymm3[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpermps %ymm3, %ymm11, %ymm4
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm7[2,3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5,6,7]
@@ -874,7 +876,7 @@ define void @load_i32_stride6_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
; AVX512-NEXT: vmovdqa64 64(%rdi), %zmm3
; AVX512-NEXT: vmovdqa 128(%rdi), %ymm0
; AVX512-NEXT: vmovdqa 160(%rdi), %ymm1
-; AVX512-NEXT: vpblendd {{.*#+}} ymm4 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
+; AVX512-NEXT: vpblendd {{.*#+}} ymm4 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX512-NEXT: vmovdqa {{.*#+}} ymm5 = <0,6,12,18,24,30,u,u>
; AVX512-NEXT: vpermi2d %zmm3, %zmm2, %zmm5
; AVX512-NEXT: vmovdqa {{.*#+}} ymm6 = [0,1,2,3,4,5,12,10]
@@ -1223,200 +1225,198 @@ define void @load_i32_stride6_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
;
; AVX1-ONLY-LABEL: load_i32_stride6_vf16:
; AVX1-ONLY: # %bb.0:
-; AVX1-ONLY-NEXT: subq $360, %rsp # imm = 0x168
-; AVX1-ONLY-NEXT: vmovaps 288(%rdi), %ymm15
-; AVX1-ONLY-NEXT: vmovaps 256(%rdi), %ymm7
-; AVX1-ONLY-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 128(%rdi), %ymm1
-; AVX1-ONLY-NEXT: vmovaps 160(%rdi), %ymm3
-; AVX1-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 32(%rdi), %ymm5
+; AVX1-ONLY-NEXT: subq $296, %rsp # imm = 0x128
+; AVX1-ONLY-NEXT: vmovaps 224(%rdi), %ymm5
; AVX1-ONLY-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps (%rdi), %ymm8
-; AVX1-ONLY-NEXT: vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 288(%rdi), %ymm8
+; AVX1-ONLY-NEXT: vmovaps 256(%rdi), %ymm3
+; AVX1-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovapd 160(%rdi), %ymm0
+; AVX1-ONLY-NEXT: vmovapd 128(%rdi), %ymm14
+; AVX1-ONLY-NEXT: vmovaps 32(%rdi), %ymm7
+; AVX1-ONLY-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps (%rdi), %ymm9
+; AVX1-ONLY-NEXT: vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 96(%rdi), %ymm2
-; AVX1-ONLY-NEXT: vmovaps 64(%rdi), %ymm0
-; AVX1-ONLY-NEXT: vmovups %ymm0, (%rsp) # 32-byte Spill
-; AVX1-ONLY-NEXT: vinsertf128 $1, 96(%rdi), %ymm0, %ymm6
+; AVX1-ONLY-NEXT: vmovaps 64(%rdi), %ymm1
+; AVX1-ONLY-NEXT: vmovups %ymm1, (%rsp) # 32-byte Spill
+; AVX1-ONLY-NEXT: vinsertf128 $1, 96(%rdi), %ymm1, %ymm6
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm4 = ymm2[2,0],ymm6[0,0],ymm2[6,4],ymm6[4,4]
-; AVX1-ONLY-NEXT: vmovaps %ymm2, %ymm12
+; AVX1-ONLY-NEXT: vmovaps %ymm2, %ymm13
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm4 = ymm4[2,0],ymm6[2,2],ymm4[6,4],ymm6[6,6]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm9 = ymm8[0,1,2,3],ymm5[4,5],ymm8[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm9 = ymm9[0,1,2,3],ymm7[4,5],ymm9[6,7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm9, %xmm2
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm11 = xmm9[0,1],xmm2[2,3]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm11 = xmm11[0,2],xmm2[0,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm11[0,1,2],ymm4[3,4,5],ymm11[6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm3[0,1,2,3],ymm1[4,5],ymm3[6,7]
-; AVX1-ONLY-NEXT: vmovaps %ymm1, %ymm13
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm10 = ymm5[2,3,0,1]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm11 = ymm10[2,0],ymm5[0,0],ymm10[6,4],ymm5[4,4]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm11[0,1,2],ymm4[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm7 = ymm14[2,3],ymm0[0,1]
+; AVX1-ONLY-NEXT: vmovapd %ymm0, %ymm10
+; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm11 = ymm7[0],ymm14[0],ymm7[3],ymm14[2]
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} ymm11 = ymm11[0,1,2,0,4,5,6,4]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm4[0,1,2,3,4,5],ymm11[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vinsertf128 $1, 288(%rdi), %ymm7, %ymm1
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm4 = ymm15[2,0],ymm1[0,0],ymm15[6,4],ymm1[4,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm4[2,0],ymm1[2,2],ymm4[6,4],ymm1[6,6]
-; AVX1-ONLY-NEXT: vmovaps 224(%rdi), %ymm0
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 192(%rdi), %ymm14
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm14[0,1,2,3],ymm0[4,5],ymm14[6,7]
-; AVX1-ONLY-NEXT: vextractf128 $1, %ymm4, %xmm0
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm11 = xmm4[0,1],xmm0[2,3]
+; AVX1-ONLY-NEXT: vinsertf128 $1, 288(%rdi), %ymm3, %ymm1
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm4 = ymm8[2,0],ymm1[0,0],ymm8[6,4],ymm1[4,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm4 = ymm4[2,0],ymm1[2,2],ymm4[6,4],ymm1[6,6]
+; AVX1-ONLY-NEXT: vmovaps 192(%rdi), %ymm15
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm15[0,1,2,3],ymm5[4,5],ymm15[6,7]
+; AVX1-ONLY-NEXT: vextractf128 $1, %ymm3, %xmm0
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm11 = xmm3[0,1],xmm0[2,3]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm11 = xmm11[0,2],xmm0[0,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm11[0,1,2],ymm3[3,4,5],ymm11[6,7]
-; AVX1-ONLY-NEXT: vmovaps 320(%rdi), %ymm3
-; AVX1-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 352(%rdi), %ymm7
-; AVX1-ONLY-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm7[0,1,2,3],ymm3[4,5],ymm7[6,7]
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm7 = ymm3[2,3,0,1]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm11 = ymm7[2,0],ymm3[0,0],ymm7[6,4],ymm3[4,4]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm11[0,1,2],ymm4[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovapd 352(%rdi), %ymm5
+; AVX1-ONLY-NEXT: vmovupd %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovapd 320(%rdi), %ymm12
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm5 = ymm12[2,3],ymm5[0,1]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm11 = ymm5[0],ymm12[0],ymm5[3],ymm12[2]
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} ymm11 = ymm11[0,1,2,0,4,5,6,4]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5],ymm11[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm8 = ymm12[3,0],ymm6[1,0],ymm12[7,4],ymm6[5,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm6 = ymm8[2,0],ymm6[2,3],ymm8[6,4],ymm6[6,7]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm8 = xmm9[1,0],xmm2[3,0]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm2 = xmm8[0,2],xmm2[1,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm6[3,4,5],ymm2[6,7]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm5 = ymm10[3,0],ymm5[1,0],ymm10[7,4],ymm5[5,4]
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} ymm5 = ymm5[0,1,2,0,4,5,6,4]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm5[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5],ymm11[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm11 = ymm13[3,0],ymm6[1,0],ymm13[7,4],ymm6[5,4]
+; AVX1-ONLY-NEXT: vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm6 = ymm11[2,0],ymm6[2,3],ymm11[6,4],ymm6[6,7]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm9 = xmm9[1,0],xmm2[3,0]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm2 = xmm9[0,2],xmm2[1,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm6[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm6 = ymm7[3,1],ymm14[1,3],ymm7[7,5],ymm14[5,7]
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} ymm6 = ymm6[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm6[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm15[3,0],ymm1[1,0],ymm15[7,4],ymm1[5,4]
+; AVX1-ONLY-NEXT: vmovaps %ymm8, %ymm4
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm8[3,0],ymm1[1,0],ymm8[7,4],ymm1[5,4]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm2[2,0],ymm1[2,3],ymm2[6,4],ymm1[6,7]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm2 = xmm4[1,0],xmm0[3,0]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm2 = xmm3[1,0],xmm0[3,0]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm0 = xmm2[0,2],xmm0[1,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm7[3,0],ymm3[1,0],ymm7[7,4],ymm3[5,4]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm5[3,1],ymm12[1,3],ymm5[7,5],ymm12[5,7]
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm0 = ymm0[0,1],mem[2,3],ymm0[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups (%rsp), %ymm9 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm12[2,1],ymm9[2,0],ymm12[6,5],ymm9[6,4]
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm1[2,3,0,1]
-; AVX1-ONLY-NEXT: vextractf128 $1, %ymm0, %xmm3
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm2 = xmm0[2,0],xmm3[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4],ymm2[5,6,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm13[0,1,2,3],ymm12[4,5],ymm13[6,7]
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm5 = ymm4[2,3,0,1]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm5[0,0],ymm4[2,0],ymm5[4,4],ymm4[6,4]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm7 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm7 = mem[0,1],ymm14[2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm1 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm1 = ymm0[0,1],mem[2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups (%rsp), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm13[2,1],ymm8[2,0],ymm13[6,5],ymm8[6,4]
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,0,1]
+; AVX1-ONLY-NEXT: vextractf128 $1, %ymm1, %xmm3
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm2 = xmm1[2,0],xmm3[2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2],ymm0[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm5 = ymm14[0,1],ymm10[2],ymm14[3]
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm7 = ymm5[2,3,0,1]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm7[0,0],ymm5[2,0],ymm7[4,4],ymm5[6,4]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm11 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm11 = mem[0,1],ymm15[2,3],mem[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm15[2,1],ymm6[2,0],ymm15[6,5],ymm6[6,4]
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm2 = ymm2[2,3,0,1]
-; AVX1-ONLY-NEXT: vextractf128 $1, %ymm7, %xmm8
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm10 = xmm7[2,0],xmm8[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm10[0,1,2],ymm2[3,4],ymm10[5,6,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm10 = ymm1[0,1,2,3],ymm13[4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm11 = ymm10[2,3,0,1]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm14 = ymm11[0,0],ymm10[2,0],ymm11[4,4],ymm10[6,4]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm14[5,6,7]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm4[2,1],ymm6[2,0],ymm4[6,5],ymm6[6,4]
+; AVX1-ONLY-NEXT: vmovaps %ymm4, %ymm13
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,0,1]
+; AVX1-ONLY-NEXT: vextractf128 $1, %ymm11, %xmm15
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm4 = xmm11[2,0],xmm15[2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm4[0,1,2],ymm0[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm12[0,1,2,3],ymm10[4,5],ymm12[6,7]
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm4[2,3,0,1]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm9 = ymm0[0,0],ymm4[2,0],ymm0[4,4],ymm4[6,4]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm9[5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm4 = ymm5[0,1],ymm4[3,1],ymm5[4,5],ymm4[7,5]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm0 = xmm0[3,1],xmm3[3,3]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm2[3,1],ymm9[2,1],ymm2[7,5],ymm9[6,5]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm5 = ymm7[0,1],ymm5[3,1],ymm7[4,5],ymm5[7,5]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm1[3,1],xmm3[3,3]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm9[3,1],ymm8[2,1],ymm9[7,5],ymm8[6,5]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm3 = ymm3[2,3,0,1]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm3[3,4],ymm0[5,6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm4[5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm11[0,1],ymm10[3,1],ymm11[4,5],ymm10[7,5]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm3 = xmm7[3,1],xmm8[3,3]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm4 = ymm15[3,1],ymm6[2,1],ymm15[7,5],ymm6[6,5]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm3[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm5[5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[0,1],ymm4[3,1],ymm0[4,5],ymm4[7,5]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm11[3,1],xmm15[3,3]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm4 = ymm13[3,1],ymm6[2,1],ymm13[7,5],ymm6[6,5]
+; AVX1-ONLY-NEXT: vmovaps %ymm6, %ymm3
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm4 = ymm4[2,3,0,1]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2],ymm4[3,4],ymm3[5,6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3,4],ymm0[5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm4[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm3 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm3 = ymm12[0,1],mem[2,3],ymm12[4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm0 = mem[0,1],ymm14[2,3],mem[4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps 32(%rdi), %xmm4
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm0 = xmm4[2,2,3,3]
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm1 = xmm4[2,2,3,3]
; AVX1-ONLY-NEXT: vmovaps 16(%rdi), %xmm5
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm5[0],xmm0[1],xmm5[2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm5[0],xmm1[1],xmm5[2,3]
; AVX1-ONLY-NEXT: vmovapd 80(%rdi), %xmm7
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm8 = ymm7[1],ymm9[0],ymm7[2],ymm9[2]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm8 = ymm2[0,1],ymm8[2,0],ymm2[4,5],ymm8[6,4]
-; AVX1-ONLY-NEXT: vmovaps %ymm2, %ymm12
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm8[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm8 = ymm3[2,3,0,1]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm10 = ymm3[2,0],ymm8[0,0],ymm3[6,4],ymm8[4,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm10 = ymm8[0,2],ymm10[2,0],ymm8[4,6],ymm10[6,4]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm10[5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm10 = ymm13[0,1],ymm1[2,3],ymm13[4,5,6,7]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm8 = ymm7[1],ymm8[0],ymm7[2],ymm8[2]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm8 = ymm9[0,1],ymm8[2,0],ymm9[4,5],ymm8[6,4]
+; AVX1-ONLY-NEXT: vmovaps %ymm9, %ymm14
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm8[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm8 = ymm0[2,3,0,1]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm9 = ymm0[2,0],ymm8[0,0],ymm0[6,4],ymm8[4,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm9 = ymm8[0,2],ymm9[2,0],ymm8[4,6],ymm9[6,4]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm9[5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm9 = ymm10[0,1],ymm12[2,3],ymm10[4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps 224(%rdi), %xmm11
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm0 = xmm11[2,2,3,3]
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm12 = xmm11[2,2,3,3]
; AVX1-ONLY-NEXT: vmovaps 208(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm1[0],xmm0[1],xmm1[2,3]
-; AVX1-ONLY-NEXT: vmovapd 272(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm9 = ymm0[1],ymm6[0],ymm0[2],ymm6[2]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm9 = ymm15[0,1],ymm9[2,0],ymm15[4,5],ymm9[6,4]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm9[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm9 = ymm10[2,3,0,1]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm14 = ymm10[2,0],ymm9[0,0],ymm10[6,4],ymm9[4,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm14 = ymm9[0,2],ymm14[2,0],ymm9[4,6],ymm14[6,4]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm14[5,6,7]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm3[3,0],ymm8[1,0],ymm3[7,4],ymm8[5,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm8[0,3],ymm3[2,0],ymm8[4,7],ymm3[6,4]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm12 = xmm1[0],xmm12[1],xmm1[2,3]
+; AVX1-ONLY-NEXT: vmovapd 272(%rdi), %xmm2
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm6 = ymm2[1],ymm6[0],ymm2[2],ymm6[2]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm6 = ymm13[0,1],ymm6[2,0],ymm13[4,5],ymm6[6,4]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm12[0,1],ymm6[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm12 = ymm9[2,3,0,1]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm9[2,0],ymm12[0,0],ymm9[6,4],ymm12[4,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm12[0,2],ymm15[2,0],ymm12[4,6],ymm15[6,4]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4],ymm15[5,6,7]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[3,0],ymm8[1,0],ymm0[7,4],ymm8[5,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm8[0,3],ymm0[2,0],ymm8[4,7],ymm0[6,4]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm4 = xmm5[0,1],xmm4[2,3]
; AVX1-ONLY-NEXT: vshufps $215, (%rsp), %ymm7, %ymm5 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm5 = ymm7[3,1],mem[1,3],ymm7[7,5],mem[5,7]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm5 = ymm12[1,1],ymm5[2,0],ymm12[5,5],ymm5[6,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm5 = ymm14[1,1],ymm5[2,0],ymm14[5,5],ymm5[6,4]
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm4 = xmm4[1,3,2,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm5[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm4[0,1,2,3,4],ymm3[5,6,7]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm4 = ymm10[3,0],ymm9[1,0],ymm10[7,4],ymm9[5,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm4 = ymm9[0,3],ymm4[2,0],ymm9[4,7],ymm4[6,4]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm4[0,1,2,3,4],ymm0[5,6,7]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm4 = ymm9[3,0],ymm12[1,0],ymm9[7,4],ymm12[5,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm4 = ymm12[0,3],ymm4[2,0],ymm12[4,7],ymm4[6,4]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm11[2,3]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[3,1],ymm6[1,3],ymm0[7,5],ymm6[5,7]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm15[1,1],ymm0[2,0],ymm15[5,5],ymm0[6,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm2[3,1],ymm3[1,3],ymm2[7,5],ymm3[5,7]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm13[1,1],ymm2[2,0],ymm13[5,5],ymm2[6,4]
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm1 = xmm1[1,3,2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm4[5,6,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm1, 32(%rsi)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm1, (%rsi)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm1, 32(%rdx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm1, (%rdx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm1, 32(%rcx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm1, (%rcx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm1, 32(%r8)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm1, (%r8)
-; AVX1-ONLY-NEXT: vmovaps %ymm2, 32(%r9)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm1, (%r9)
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm4[5,6,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm2, 32(%rsi)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm2, (%rsi)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm2, 32(%rdx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm2, (%rdx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm2, 32(%rcx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm2, (%rcx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm2, 32(%r8)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm2, (%r8)
+; AVX1-ONLY-NEXT: vmovaps %ymm6, 32(%r9)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm2, (%r9)
; AVX1-ONLY-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; AVX1-ONLY-NEXT: vmovaps %ymm0, 32(%rax)
-; AVX1-ONLY-NEXT: vmovaps %ymm3, (%rax)
-; AVX1-ONLY-NEXT: addq $360, %rsp # imm = 0x168
+; AVX1-ONLY-NEXT: vmovaps %ymm1, 32(%rax)
+; AVX1-ONLY-NEXT: vmovaps %ymm0, (%rax)
+; AVX1-ONLY-NEXT: addq $296, %rsp # imm = 0x128
; AVX1-ONLY-NEXT: vzeroupper
; AVX1-ONLY-NEXT: retq
;
; AVX2-SLOW-LABEL: load_i32_stride6_vf16:
; AVX2-SLOW: # %bb.0:
-; AVX2-SLOW-NEXT: subq $264, %rsp # imm = 0x108
+; AVX2-SLOW-NEXT: subq $200, %rsp
; AVX2-SLOW-NEXT: vmovaps 288(%rdi), %ymm8
; AVX2-SLOW-NEXT: vmovaps 224(%rdi), %ymm10
-; AVX2-SLOW-NEXT: vmovaps 192(%rdi), %ymm11
+; AVX2-SLOW-NEXT: vmovaps 192(%rdi), %ymm5
+; AVX2-SLOW-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovaps 128(%rdi), %ymm0
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovaps 160(%rdi), %ymm2
@@ -1424,364 +1424,350 @@ define void @load_i32_stride6_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-SLOW-NEXT: vmovaps 96(%rdi), %ymm7
; AVX2-SLOW-NEXT: vmovaps (%rdi), %ymm3
; AVX2-SLOW-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovaps 32(%rdi), %ymm4
-; AVX2-SLOW-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovaps 64(%rdi), %ymm5
+; AVX2-SLOW-NEXT: vmovaps 32(%rdi), %ymm11
+; AVX2-SLOW-NEXT: vmovaps 64(%rdi), %ymm9
; AVX2-SLOW-NEXT: vmovaps {{.*#+}} xmm1 = <0,6,4,u>
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm6 = ymm3[0,1,2,3],ymm4[4,5],ymm3[6,7]
-; AVX2-SLOW-NEXT: vpermps %ymm6, %ymm1, %ymm9
-; AVX2-SLOW-NEXT: vperm2f128 {{.*#+}} ymm12 = ymm5[0,1],ymm7[0,1]
-; AVX2-SLOW-NEXT: vmovaps %ymm5, %ymm4
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm3[0,1,2,3],ymm11[4,5],ymm3[6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm4, %ymm1, %ymm6
+; AVX2-SLOW-NEXT: vperm2f128 {{.*#+}} ymm12 = ymm9[0,1],ymm7[0,1]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3,4,5],ymm7[6,7]
; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm13 = ymm12[0,2,2,2,4,6,6,6]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm9 = ymm9[0,1,2],ymm13[3,4,5],ymm9[6,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm13 = ymm2[0,1,2,3],ymm0[4,5],ymm2[6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2],ymm13[3,4,5,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm13 = ymm2[0,1,2,3],ymm0[4,5,6,7]
; AVX2-SLOW-NEXT: vbroadcastsd {{.*#+}} ymm0 = [8589934596,8589934596,8589934596,8589934596]
; AVX2-SLOW-NEXT: vpermps %ymm13, %ymm0, %ymm14
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm9[0,1,2,3,4,5],ymm14[6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm6[0,1,2,3,4,5],ymm14[6,7]
; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm9 = ymm11[0,1,2,3],ymm10[4,5],ymm11[6,7]
-; AVX2-SLOW-NEXT: vpermps %ymm9, %ymm1, %ymm14
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm6 = ymm5[0,1,2,3],ymm10[4,5],ymm5[6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm6, %ymm1, %ymm14
; AVX2-SLOW-NEXT: vmovaps 256(%rdi), %ymm5
; AVX2-SLOW-NEXT: vperm2f128 {{.*#+}} ymm15 = ymm5[0,1],ymm8[0,1]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm15[0,1,2,3,4,5],ymm8[6,7]
; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm15 = ymm2[0,2,2,2,4,6,6,6]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm14[0,1,2],ymm15[3,4,5],ymm14[6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm14[0,1,2],ymm15[3,4,5,6,7]
; AVX2-SLOW-NEXT: vmovaps 320(%rdi), %ymm14
; AVX2-SLOW-NEXT: vmovaps 352(%rdi), %ymm15
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm15[0,1,2,3],ymm14[4,5],ymm15[6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm15[0,1,2,3],ymm14[4,5,6,7]
; AVX2-SLOW-NEXT: vpermps %ymm1, %ymm0, %ymm0
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3,4,5],ymm0[6,7]
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovaps {{.*#+}} xmm0 = <1,7,5,u>
-; AVX2-SLOW-NEXT: vpermps %ymm6, %ymm0, %ymm3
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm6 = ymm12[1,3,2,3,5,7,6,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2],ymm6[3,4,5],ymm3[6,7]
-; AVX2-SLOW-NEXT: vbroadcastsd {{.*#+}} ymm6 = [12884901893,12884901893,12884901893,12884901893]
-; AVX2-SLOW-NEXT: vpermps %ymm13, %ymm6, %ymm12
+; AVX2-SLOW-NEXT: vpermps %ymm4, %ymm0, %ymm3
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm4 = ymm12[1,3,2,3,5,7,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2],ymm4[3,4,5,6,7]
+; AVX2-SLOW-NEXT: vbroadcastsd {{.*#+}} ymm4 = [12884901893,12884901893,12884901893,12884901893]
+; AVX2-SLOW-NEXT: vpermps %ymm13, %ymm4, %ymm12
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm12[6,7]
; AVX2-SLOW-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpermps %ymm9, %ymm0, %ymm0
+; AVX2-SLOW-NEXT: vpermps %ymm6, %ymm0, %ymm0
; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm2 = ymm2[1,3,2,3,5,7,6,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm2[3,4,5],ymm0[6,7]
-; AVX2-SLOW-NEXT: vpermps %ymm1, %ymm6, %ymm1
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm2[3,4,5,6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm1, %ymm4, %ymm1
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm0 = ymm9[2,3,2,3,6,7,6,7]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm1 = ymm13[0,0,2,3,4,4,6,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm4[0,1],ymm7[2,3],ymm4[4,5],ymm7[6,7]
-; AVX2-SLOW-NEXT: vmovaps %ymm4, %ymm12
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm1 = ymm1[2,0,2,3,6,4,6,7]
-; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
-; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm1 = ymm3[0,0,0,0,4,4,4,4]
+; AVX2-SLOW-NEXT: vmovaps (%rdi), %xmm0
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm2 = ymm11[0,0,2,3,4,4,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm2[1,2,3],ymm1[4],ymm2[5,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm9[0,1],ymm7[2,3],ymm9[4,5],ymm7[6,7]
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm2 = ymm2[2,0,2,3,6,4,6,7]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,2,0,3]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,3,2,3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm2 = ymm12[0,0,0,0,4,4,4,4]
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm2 = ymm6[0,0,2,3,4,4,6,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
-; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm0 = ymm11[2,3,2,3,6,7,6,7]
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm1 = ymm10[0,0,2,3,4,4,6,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm5[0,1],ymm8[2,3],ymm5[4,5],ymm8[6,7]
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm1 = ymm1[2,0,2,3,6,4,6,7]
-; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
-; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm1 = ymm15[0,0,0,0,4,4,4,4]
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm2 = ymm14[0,0,2,3,4,4,6,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
-; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm0 = ymm9[3,3,3,3,7,7,7,7]
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm1 = ymm13[0,1,3,3,4,5,7,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm1 = ymm7[3,3,3,3,7,7,7,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm4[1],ymm1[2,3,4],ymm4[5],ymm1[6,7]
-; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm3 = ymm6[0,0,2,3,4,4,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3],ymm3[4,5,6],ymm2[7]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,1,0,3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm1, (%rsp) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovaps 192(%rdi), %xmm1
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} xmm2 = xmm1[2,3,2,3]
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm3 = ymm10[0,0,2,3,4,4,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0],ymm3[1,2,3],ymm2[4],ymm3[5,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm5[0,1],ymm8[2,3],ymm5[4,5],ymm8[6,7]
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm3 = ymm3[2,0,2,3,6,4,6,7]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm3 = ymm3[0,2,0,3]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,3,2,3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm3[3,4,5,6,7]
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm3 = ymm15[0,0,0,0,4,4,4,4]
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm4 = ymm14[0,0,2,3,4,4,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3],ymm4[4,5,6],ymm3[7]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm3 = ymm3[0,1,0,3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm2 = ymm11[0,1,3,3,4,5,7,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm2[1,2,3],ymm0[4],ymm2[5,6,7]
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm2 = ymm7[3,3,3,3,7,7,7,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0],ymm9[1],ymm2[2,3,4],ymm9[5],ymm2[6,7]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,2,0,3]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm1 = ymm3[0,1,0,1,4,5,4,5]
-; AVX2-SLOW-NEXT: vmovaps %ymm3, %ymm9
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm2 = ymm6[0,1,3,3,4,5,7,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
-; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm0 = ymm11[3,3,3,3,7,7,7,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm2[3,4,5,6,7]
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm2 = ymm12[0,1,0,1,4,5,4,5]
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm3 = ymm6[0,1,3,3,4,5,7,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3],ymm3[4,5,6],ymm2[7]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,1,0,3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm13 = ymm0[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} xmm0 = xmm1[3,3,3,3]
; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm1 = ymm10[0,1,3,3,4,5,7,7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm1 = ymm8[3,3,3,3,7,7,7,7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm5[1],ymm1[2,3,4],ymm5[5],ymm1[6,7]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm1 = ymm15[0,1,0,1,4,5,4,5]
; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm2 = ymm14[0,1,3,3,4,5,7,7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm0, (%rsp) # 32-byte Spill
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm10[0,1,2,3],ymm11[4,5],ymm10[6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-SLOW-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm1 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm1 = ymm10[0,1,2,3],mem[4,5,6,7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm5[0,1,2,3],ymm8[4,5,6,7]
-; AVX2-SLOW-NEXT: vmovaps 272(%rdi), %xmm3
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm3[2,3],ymm2[4,5,6,7]
+; AVX2-SLOW-NEXT: vmovaps 272(%rdi), %xmm4
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm4[2,3],ymm2[4,5,6,7]
; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,1,0,2,4,5,4,6]
-; AVX2-SLOW-NEXT: vmovaps {{.*#+}} xmm4 = <4,2,u,u>
-; AVX2-SLOW-NEXT: vpermps %ymm1, %ymm4, %ymm8
+; AVX2-SLOW-NEXT: vmovaps {{.*#+}} xmm5 = <4,2,u,u>
+; AVX2-SLOW-NEXT: vpermps %ymm1, %ymm5, %ymm8
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm8[0,1],ymm0[2,3,4,5,6,7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm8 = ymm15[0,1],ymm14[2,3],ymm15[4,5,6,7]
; AVX2-SLOW-NEXT: vbroadcastf128 {{.*#+}} ymm10 = [0,2,0,6,0,2,0,6]
; AVX2-SLOW-NEXT: # ymm10 = mem[0,1,0,1]
-; AVX2-SLOW-NEXT: vpermps %ymm8, %ymm10, %ymm11
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm11[5,6,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm12[0,1,2,3],ymm7[4,5,6,7]
-; AVX2-SLOW-NEXT: vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm7 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm7 = ymm13[0,1,2,3],mem[4,5],ymm13[6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm8, %ymm10, %ymm14
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm14[5,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm7 = ymm9[0,1,2,3],ymm7[4,5,6,7]
+; AVX2-SLOW-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm9 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm9 = ymm11[0,1,2,3],mem[4,5,6,7]
; AVX2-SLOW-NEXT: vmovaps 80(%rdi), %xmm11
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm14 = ymm5[0,1],ymm11[2,3],ymm5[4,5,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm14 = ymm7[0,1],ymm11[2,3],ymm7[4,5,6,7]
; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm14 = ymm14[0,1,0,2,4,5,4,6]
-; AVX2-SLOW-NEXT: vpermps %ymm7, %ymm4, %ymm4
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm14[2,3,4,5,6,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm14 = ymm9[0,1],ymm6[2,3],ymm9[4,5,6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm9, %ymm5, %ymm5
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1],ymm14[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm14 = ymm12[0,1],ymm6[2,3],ymm12[4,5,6,7]
; AVX2-SLOW-NEXT: vpermps %ymm14, %ymm10, %ymm10
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4],ymm10[5,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4],ymm10[5,6,7]
; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm2 = ymm2[1,1,1,1,5,5,5,5]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm3[3],ymm2[4,5,6,7]
-; AVX2-SLOW-NEXT: vmovaps {{.*#+}} xmm3 = <5,3,u,u>
-; AVX2-SLOW-NEXT: vpermps %ymm1, %ymm3, %ymm1
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm4[3],ymm2[4,5,6,7]
+; AVX2-SLOW-NEXT: vmovaps {{.*#+}} xmm4 = <5,3,u,u>
+; AVX2-SLOW-NEXT: vpermps %ymm1, %ymm4, %ymm1
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2,3,4,5,6,7]
; AVX2-SLOW-NEXT: vbroadcastf128 {{.*#+}} ymm2 = [0,3,1,7,0,3,1,7]
; AVX2-SLOW-NEXT: # ymm2 = mem[0,1,0,1]
; AVX2-SLOW-NEXT: vpermps %ymm8, %ymm2, %ymm8
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm8[5,6,7]
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm5 = ymm5[1,1,1,1,5,5,5,5]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2],ymm11[3],ymm5[4,5,6,7]
-; AVX2-SLOW-NEXT: vpermps %ymm7, %ymm3, %ymm3
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1],ymm5[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm7 = ymm7[1,1,1,1,5,5,5,5]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2],ymm11[3],ymm7[4,5,6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm9, %ymm4, %ymm4
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm7[2,3,4,5,6,7]
; AVX2-SLOW-NEXT: vpermps %ymm14, %ymm2, %ymm2
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2,3,4],ymm2[5,6,7]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-SLOW-NEXT: vmovaps %ymm3, 32(%rsi)
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-SLOW-NEXT: vmovaps %ymm3, (%rsi)
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-SLOW-NEXT: vmovaps %ymm3, 32(%rdx)
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-SLOW-NEXT: vmovaps %ymm3, (%rdx)
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-SLOW-NEXT: vmovaps %ymm3, 32(%rcx)
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-SLOW-NEXT: vmovaps %ymm3, (%rcx)
-; AVX2-SLOW-NEXT: vmovups (%rsp), %ymm3 # 32-byte Reload
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm4[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovaps %ymm4, 32(%rsi)
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovaps %ymm4, (%rsi)
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovaps %ymm4, 32(%rdx)
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovaps %ymm4, (%rdx)
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovaps %ymm4, 32(%rcx)
+; AVX2-SLOW-NEXT: vmovups (%rsp), %ymm4 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovaps %ymm4, (%rcx)
; AVX2-SLOW-NEXT: vmovaps %ymm3, 32(%r8)
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-SLOW-NEXT: vmovaps %ymm3, (%r8)
+; AVX2-SLOW-NEXT: vmovaps %ymm13, (%r8)
; AVX2-SLOW-NEXT: vmovaps %ymm0, 32(%r9)
-; AVX2-SLOW-NEXT: vmovaps %ymm4, (%r9)
+; AVX2-SLOW-NEXT: vmovaps %ymm5, (%r9)
; AVX2-SLOW-NEXT: movq {{[0-9]+}}(%rsp), %rax
; AVX2-SLOW-NEXT: vmovaps %ymm1, 32(%rax)
; AVX2-SLOW-NEXT: vmovaps %ymm2, (%rax)
-; AVX2-SLOW-NEXT: addq $264, %rsp # imm = 0x108
+; AVX2-SLOW-NEXT: addq $200, %rsp
; AVX2-SLOW-NEXT: vzeroupper
; AVX2-SLOW-NEXT: retq
;
; AVX2-FAST-LABEL: load_i32_stride6_vf16:
; AVX2-FAST: # %bb.0:
-; AVX2-FAST-NEXT: subq $264, %rsp # imm = 0x108
+; AVX2-FAST-NEXT: subq $200, %rsp
; AVX2-FAST-NEXT: vmovaps 288(%rdi), %ymm8
; AVX2-FAST-NEXT: vmovaps 224(%rdi), %ymm10
-; AVX2-FAST-NEXT: vmovaps 192(%rdi), %ymm11
-; AVX2-FAST-NEXT: vmovaps 128(%rdi), %ymm7
-; AVX2-FAST-NEXT: vmovaps 160(%rdi), %ymm0
+; AVX2-FAST-NEXT: vmovaps 192(%rdi), %ymm5
+; AVX2-FAST-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovaps 128(%rdi), %ymm0
; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps 96(%rdi), %ymm5
-; AVX2-FAST-NEXT: vmovaps (%rdi), %ymm2
+; AVX2-FAST-NEXT: vmovaps 160(%rdi), %ymm2
; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps 32(%rdi), %ymm3
+; AVX2-FAST-NEXT: vmovaps 96(%rdi), %ymm7
+; AVX2-FAST-NEXT: vmovaps (%rdi), %ymm3
; AVX2-FAST-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps 64(%rdi), %ymm4
+; AVX2-FAST-NEXT: vmovaps 32(%rdi), %ymm11
+; AVX2-FAST-NEXT: vmovaps 64(%rdi), %ymm9
; AVX2-FAST-NEXT: vmovaps {{.*#+}} xmm1 = <0,6,4,u>
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm6 = ymm2[0,1,2,3],ymm3[4,5],ymm2[6,7]
-; AVX2-FAST-NEXT: vpermps %ymm6, %ymm1, %ymm9
-; AVX2-FAST-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vperm2f128 {{.*#+}} ymm12 = ymm4[0,1],ymm5[0,1]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3,4,5],ymm5[6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm3[0,1,2,3],ymm11[4,5],ymm3[6,7]
+; AVX2-FAST-NEXT: vpermps %ymm4, %ymm1, %ymm6
+; AVX2-FAST-NEXT: vperm2f128 {{.*#+}} ymm12 = ymm9[0,1],ymm7[0,1]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3,4,5],ymm7[6,7]
; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm13 = ymm12[0,2,2,2,4,6,6,6]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm9 = ymm9[0,1,2],ymm13[3,4,5],ymm9[6,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm13 = ymm0[0,1,2,3],ymm7[4,5],ymm0[6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2],ymm13[3,4,5,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm13 = ymm2[0,1,2,3],ymm0[4,5,6,7]
; AVX2-FAST-NEXT: vbroadcastsd {{.*#+}} ymm0 = [8589934596,8589934596,8589934596,8589934596]
; AVX2-FAST-NEXT: vpermps %ymm13, %ymm0, %ymm14
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm9[0,1,2,3,4,5],ymm14[6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm6[0,1,2,3,4,5],ymm14[6,7]
; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm9 = ymm11[0,1,2,3],ymm10[4,5],ymm11[6,7]
-; AVX2-FAST-NEXT: vpermps %ymm9, %ymm1, %ymm14
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm6 = ymm5[0,1,2,3],ymm10[4,5],ymm5[6,7]
+; AVX2-FAST-NEXT: vpermps %ymm6, %ymm1, %ymm14
; AVX2-FAST-NEXT: vmovaps 256(%rdi), %ymm5
; AVX2-FAST-NEXT: vperm2f128 {{.*#+}} ymm15 = ymm5[0,1],ymm8[0,1]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm15[0,1,2,3,4,5],ymm8[6,7]
; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm15 = ymm2[0,2,2,2,4,6,6,6]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm14[0,1,2],ymm15[3,4,5],ymm14[6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm14[0,1,2],ymm15[3,4,5,6,7]
; AVX2-FAST-NEXT: vmovaps 320(%rdi), %ymm14
; AVX2-FAST-NEXT: vmovaps 352(%rdi), %ymm15
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm15[0,1,2,3],ymm14[4,5],ymm15[6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm15[0,1,2,3],ymm14[4,5,6,7]
; AVX2-FAST-NEXT: vpermps %ymm1, %ymm0, %ymm0
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3,4,5],ymm0[6,7]
; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovaps {{.*#+}} xmm0 = <1,7,5,u>
-; AVX2-FAST-NEXT: vpermps %ymm6, %ymm0, %ymm3
-; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm6 = ymm12[1,3,2,3,5,7,6,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2],ymm6[3,4,5],ymm3[6,7]
-; AVX2-FAST-NEXT: vbroadcastsd {{.*#+}} ymm6 = [12884901893,12884901893,12884901893,12884901893]
-; AVX2-FAST-NEXT: vpermps %ymm13, %ymm6, %ymm12
+; AVX2-FAST-NEXT: vpermps %ymm4, %ymm0, %ymm3
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm4 = ymm12[1,3,2,3,5,7,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2],ymm4[3,4,5,6,7]
+; AVX2-FAST-NEXT: vbroadcastsd {{.*#+}} ymm4 = [12884901893,12884901893,12884901893,12884901893]
+; AVX2-FAST-NEXT: vpermps %ymm13, %ymm4, %ymm12
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm12[6,7]
; AVX2-FAST-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpermps %ymm9, %ymm0, %ymm0
+; AVX2-FAST-NEXT: vpermps %ymm6, %ymm0, %ymm0
; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm2 = ymm2[1,3,2,3,5,7,6,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm2[3,4,5],ymm0[6,7]
-; AVX2-FAST-NEXT: vpermps %ymm1, %ymm6, %ymm1
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm2[3,4,5,6,7]
+; AVX2-FAST-NEXT: vpermps %ymm1, %ymm4, %ymm1
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
-; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm0 = ymm12[2,3,2,3,6,7,6,7]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm1 = ymm9[0,0,2,3,4,4,6,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm4[0,1],ymm13[2,3],ymm4[4,5],ymm13[6,7]
-; AVX2-FAST-NEXT: vmovaps {{.*#+}} ymm2 = [2,0,6,4,2,0,6,7]
-; AVX2-FAST-NEXT: vpermps %ymm1, %ymm2, %ymm1
-; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm1 = ymm6[0,0,0,0,4,4,4,4]
-; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm3 = ymm7[0,0,2,3,4,4,6,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm3[0,1,2],ymm1[3],ymm3[4,5,6],ymm1[7]
-; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
-; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm5[0,1],ymm8[2,3],ymm5[4,5],ymm8[6,7]
-; AVX2-FAST-NEXT: vpermps %ymm0, %ymm2, %ymm0
-; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm1 = ymm11[2,3,2,3,6,7,6,7]
-; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm2 = ymm10[0,0,2,3,4,4,6,7]
+; AVX2-FAST-NEXT: vmovaps (%rdi), %xmm0
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm2 = ymm11[0,0,2,3,4,4,6,7]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm2[1,2,3],ymm1[4],ymm2[5,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm9[0,1],ymm7[2,3],ymm9[4,5],ymm7[6,7]
+; AVX2-FAST-NEXT: vmovaps {{.*#+}} ymm3 = [2,0,6,4,2,0,6,7]
+; AVX2-FAST-NEXT: vpermps %ymm2, %ymm3, %ymm2
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,3,2,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4],ymm1[5,6,7]
-; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm1 = ymm15[0,0,0,0,4,4,4,4]
-; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm2 = ymm14[0,0,2,3,4,4,6,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
-; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
-; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm0 = ymm12[3,3,3,3,7,7,7,7]
-; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm1 = ymm9[0,1,3,3,4,5,7,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm2 = ymm12[0,0,0,0,4,4,4,4]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm4 = ymm6[0,0,2,3,4,4,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm4[0,1,2],ymm2[3],ymm4[4,5,6],ymm2[7]
+; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,1,0,3]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-FAST-NEXT: vmovups %ymm1, (%rsp) # 32-byte Spill
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm5[0,1],ymm8[2,3],ymm5[4,5],ymm8[6,7]
+; AVX2-FAST-NEXT: vpermps %ymm1, %ymm3, %ymm1
+; AVX2-FAST-NEXT: vmovaps 192(%rdi), %xmm2
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} xmm3 = xmm2[2,3,2,3]
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm4 = ymm10[0,0,2,3,4,4,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0],ymm4[1,2,3],ymm3[4],ymm4[5,6,7]
+; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm3 = ymm3[0,3,2,3]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm3[0,1,2],ymm1[3,4,5,6,7]
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm3 = ymm15[0,0,0,0,4,4,4,4]
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm4 = ymm14[0,0,2,3,4,4,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3],ymm4[4,5,6],ymm3[7]
+; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm3 = ymm3[0,1,0,3]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm3[5,6,7]
+; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm1 = ymm11[0,1,3,3,4,5,7,7]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
-; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm1 = ymm13[3,3,3,3,7,7,7,7]
-; AVX2-FAST-NEXT: vmovaps %ymm13, %ymm12
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm4[1],ymm1[2,3,4],ymm4[5],ymm1[6,7]
-; AVX2-FAST-NEXT: vmovaps %ymm4, %ymm9
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm1 = ymm7[3,3,3,3,7,7,7,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm9[1],ymm1[2,3,4],ymm9[5],ymm1[6,7]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
-; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm1 = ymm6[0,1,0,1,4,5,4,5]
-; AVX2-FAST-NEXT: vmovaps %ymm6, %ymm13
-; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm2 = ymm7[0,1,3,3,4,5,7,7]
-; AVX2-FAST-NEXT: vmovaps %ymm7, %ymm6
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm1 = ymm12[0,1,0,1,4,5,4,5]
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm3 = ymm6[0,1,3,3,4,5,7,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm3[0,1,2],ymm1[3],ymm3[4,5,6],ymm1[7]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
-; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm0 = ymm11[3,3,3,3,7,7,7,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm13 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} xmm0 = xmm2[3,3,3,3]
; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm1 = ymm10[0,1,3,3,4,5,7,7]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm1 = ymm8[3,3,3,3,7,7,7,7]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm5[1],ymm1[2,3,4],ymm5[5],ymm1[6,7]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm1 = ymm15[0,1,0,1,4,5,4,5]
; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm2 = ymm14[0,1,3,3,4,5,7,7]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
-; AVX2-FAST-NEXT: vmovups %ymm0, (%rsp) # 32-byte Spill
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm10[0,1,2,3],ymm11[4,5],ymm10[6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-FAST-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm1 = ymm10[0,1,2,3],mem[4,5,6,7]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm5[0,1,2,3],ymm8[4,5,6,7]
-; AVX2-FAST-NEXT: vmovaps 272(%rdi), %xmm3
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm3[2,3],ymm2[4,5,6,7]
+; AVX2-FAST-NEXT: vmovaps 272(%rdi), %xmm4
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm4[2,3],ymm2[4,5,6,7]
; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,1,0,2,4,5,4,6]
-; AVX2-FAST-NEXT: vmovaps {{.*#+}} xmm4 = <4,2,u,u>
-; AVX2-FAST-NEXT: vpermps %ymm1, %ymm4, %ymm8
+; AVX2-FAST-NEXT: vmovaps {{.*#+}} xmm5 = <4,2,u,u>
+; AVX2-FAST-NEXT: vpermps %ymm1, %ymm5, %ymm8
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm8[0,1],ymm0[2,3,4,5,6,7]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm8 = ymm15[0,1],ymm14[2,3],ymm15[4,5,6,7]
; AVX2-FAST-NEXT: vbroadcastf128 {{.*#+}} ymm10 = [0,2,0,6,0,2,0,6]
; AVX2-FAST-NEXT: # ymm10 = mem[0,1,0,1]
-; AVX2-FAST-NEXT: vpermps %ymm8, %ymm10, %ymm11
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm11[5,6,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm5 = ymm9[0,1,2,3],ymm12[4,5,6,7]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
-; AVX2-FAST-NEXT: vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm7 = ymm7[0,1,2,3],mem[4,5],ymm7[6,7]
+; AVX2-FAST-NEXT: vpermps %ymm8, %ymm10, %ymm14
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm14[5,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm7 = ymm9[0,1,2,3],ymm7[4,5,6,7]
+; AVX2-FAST-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm9 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm9 = ymm11[0,1,2,3],mem[4,5,6,7]
; AVX2-FAST-NEXT: vmovaps 80(%rdi), %xmm11
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm14 = ymm5[0,1],ymm11[2,3],ymm5[4,5,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm14 = ymm7[0,1],ymm11[2,3],ymm7[4,5,6,7]
; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm14 = ymm14[0,1,0,2,4,5,4,6]
-; AVX2-FAST-NEXT: vpermps %ymm7, %ymm4, %ymm4
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm14[2,3,4,5,6,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm14 = ymm13[0,1],ymm6[2,3],ymm13[4,5,6,7]
+; AVX2-FAST-NEXT: vpermps %ymm9, %ymm5, %ymm5
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1],ymm14[2,3,4,5,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm14 = ymm12[0,1],ymm6[2,3],ymm12[4,5,6,7]
; AVX2-FAST-NEXT: vpermps %ymm14, %ymm10, %ymm10
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4],ymm10[5,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4],ymm10[5,6,7]
; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm2 = ymm2[1,1,1,1,5,5,5,5]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm3[3],ymm2[4,5,6,7]
-; AVX2-FAST-NEXT: vmovaps {{.*#+}} xmm3 = <5,3,u,u>
-; AVX2-FAST-NEXT: vpermps %ymm1, %ymm3, %ymm1
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm4[3],ymm2[4,5,6,7]
+; AVX2-FAST-NEXT: vmovaps {{.*#+}} xmm4 = <5,3,u,u>
+; AVX2-FAST-NEXT: vpermps %ymm1, %ymm4, %ymm1
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2,3,4,5,6,7]
; AVX2-FAST-NEXT: vbroadcastf128 {{.*#+}} ymm2 = [0,3,1,7,0,3,1,7]
; AVX2-FAST-NEXT: # ymm2 = mem[0,1,0,1]
; AVX2-FAST-NEXT: vpermps %ymm8, %ymm2, %ymm8
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm8[5,6,7]
-; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm5 = ymm5[1,1,1,1,5,5,5,5]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2],ymm11[3],ymm5[4,5,6,7]
-; AVX2-FAST-NEXT: vpermps %ymm7, %ymm3, %ymm3
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1],ymm5[2,3,4,5,6,7]
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm7 = ymm7[1,1,1,1,5,5,5,5]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2],ymm11[3],ymm7[4,5,6,7]
+; AVX2-FAST-NEXT: vpermps %ymm9, %ymm4, %ymm4
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm7[2,3,4,5,6,7]
; AVX2-FAST-NEXT: vpermps %ymm14, %ymm2, %ymm2
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2,3,4],ymm2[5,6,7]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovaps %ymm3, 32(%rsi)
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovaps %ymm3, (%rsi)
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovaps %ymm3, 32(%rdx)
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovaps %ymm3, (%rdx)
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovaps %ymm3, 32(%rcx)
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovaps %ymm3, (%rcx)
-; AVX2-FAST-NEXT: vmovups (%rsp), %ymm3 # 32-byte Reload
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm4[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovaps %ymm4, 32(%rsi)
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovaps %ymm4, (%rsi)
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovaps %ymm4, 32(%rdx)
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovaps %ymm4, (%rdx)
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovaps %ymm4, 32(%rcx)
+; AVX2-FAST-NEXT: vmovups (%rsp), %ymm4 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovaps %ymm4, (%rcx)
; AVX2-FAST-NEXT: vmovaps %ymm3, 32(%r8)
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovaps %ymm3, (%r8)
+; AVX2-FAST-NEXT: vmovaps %ymm13, (%r8)
; AVX2-FAST-NEXT: vmovaps %ymm0, 32(%r9)
-; AVX2-FAST-NEXT: vmovaps %ymm4, (%r9)
+; AVX2-FAST-NEXT: vmovaps %ymm5, (%r9)
; AVX2-FAST-NEXT: movq {{[0-9]+}}(%rsp), %rax
; AVX2-FAST-NEXT: vmovaps %ymm1, 32(%rax)
; AVX2-FAST-NEXT: vmovaps %ymm2, (%rax)
-; AVX2-FAST-NEXT: addq $264, %rsp # imm = 0x108
+; AVX2-FAST-NEXT: addq $200, %rsp
; AVX2-FAST-NEXT: vzeroupper
; AVX2-FAST-NEXT: retq
;
; AVX2-FAST-PERLANE-LABEL: load_i32_stride6_vf16:
; AVX2-FAST-PERLANE: # %bb.0:
-; AVX2-FAST-PERLANE-NEXT: subq $264, %rsp # imm = 0x108
+; AVX2-FAST-PERLANE-NEXT: subq $200, %rsp
; AVX2-FAST-PERLANE-NEXT: vmovaps 288(%rdi), %ymm8
; AVX2-FAST-PERLANE-NEXT: vmovaps 224(%rdi), %ymm10
-; AVX2-FAST-PERLANE-NEXT: vmovaps 192(%rdi), %ymm11
+; AVX2-FAST-PERLANE-NEXT: vmovaps 192(%rdi), %ymm5
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovaps 128(%rdi), %ymm0
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovaps 160(%rdi), %ymm2
@@ -1789,172 +1775,165 @@ define void @load_i32_stride6_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-PERLANE-NEXT: vmovaps 96(%rdi), %ymm7
; AVX2-FAST-PERLANE-NEXT: vmovaps (%rdi), %ymm3
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovaps 32(%rdi), %ymm4
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovaps 64(%rdi), %ymm5
+; AVX2-FAST-PERLANE-NEXT: vmovaps 32(%rdi), %ymm11
+; AVX2-FAST-PERLANE-NEXT: vmovaps 64(%rdi), %ymm9
; AVX2-FAST-PERLANE-NEXT: vmovaps {{.*#+}} xmm1 = <0,6,4,u>
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm6 = ymm3[0,1,2,3],ymm4[4,5],ymm3[6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm6, %ymm1, %ymm9
-; AVX2-FAST-PERLANE-NEXT: vperm2f128 {{.*#+}} ymm12 = ymm5[0,1],ymm7[0,1]
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm5, %ymm4
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm3[0,1,2,3],ymm11[4,5],ymm3[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm4, %ymm1, %ymm6
+; AVX2-FAST-PERLANE-NEXT: vperm2f128 {{.*#+}} ymm12 = ymm9[0,1],ymm7[0,1]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3,4,5],ymm7[6,7]
; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm13 = ymm12[0,2,2,2,4,6,6,6]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm9 = ymm9[0,1,2],ymm13[3,4,5],ymm9[6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm13 = ymm2[0,1,2,3],ymm0[4,5],ymm2[6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2],ymm13[3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm13 = ymm2[0,1,2,3],ymm0[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vbroadcastsd {{.*#+}} ymm0 = [8589934596,8589934596,8589934596,8589934596]
; AVX2-FAST-PERLANE-NEXT: vpermps %ymm13, %ymm0, %ymm14
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm9[0,1,2,3,4,5],ymm14[6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm6[0,1,2,3,4,5],ymm14[6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm9 = ymm11[0,1,2,3],ymm10[4,5],ymm11[6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm9, %ymm1, %ymm14
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm6 = ymm5[0,1,2,3],ymm10[4,5],ymm5[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm6, %ymm1, %ymm14
; AVX2-FAST-PERLANE-NEXT: vmovaps 256(%rdi), %ymm5
; AVX2-FAST-PERLANE-NEXT: vperm2f128 {{.*#+}} ymm15 = ymm5[0,1],ymm8[0,1]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm15[0,1,2,3,4,5],ymm8[6,7]
; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm15 = ymm2[0,2,2,2,4,6,6,6]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm14[0,1,2],ymm15[3,4,5],ymm14[6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm14[0,1,2],ymm15[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovaps 320(%rdi), %ymm14
; AVX2-FAST-PERLANE-NEXT: vmovaps 352(%rdi), %ymm15
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm15[0,1,2,3],ymm14[4,5],ymm15[6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm15[0,1,2,3],ymm14[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpermps %ymm1, %ymm0, %ymm0
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3,4,5],ymm0[6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovaps {{.*#+}} xmm0 = <1,7,5,u>
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm6, %ymm0, %ymm3
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm6 = ymm12[1,3,2,3,5,7,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2],ymm6[3,4,5],ymm3[6,7]
-; AVX2-FAST-PERLANE-NEXT: vbroadcastsd {{.*#+}} ymm6 = [12884901893,12884901893,12884901893,12884901893]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm13, %ymm6, %ymm12
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm4, %ymm0, %ymm3
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm4 = ymm12[1,3,2,3,5,7,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2],ymm4[3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vbroadcastsd {{.*#+}} ymm4 = [12884901893,12884901893,12884901893,12884901893]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm13, %ymm4, %ymm12
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm12[6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm9, %ymm0, %ymm0
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm6, %ymm0, %ymm0
; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm2 = ymm2[1,3,2,3,5,7,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm2[3,4,5],ymm0[6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm1, %ymm6, %ymm1
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm2[3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm1, %ymm4, %ymm1
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm0 = ymm9[2,3,2,3,6,7,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm1 = ymm13[0,0,2,3,4,4,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm4[0,1],ymm7[2,3],ymm4[4,5],ymm7[6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm4, %ymm12
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm1 = ymm1[2,0,2,3,6,4,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
-; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm1 = ymm3[0,0,0,0,4,4,4,4]
+; AVX2-FAST-PERLANE-NEXT: vmovaps (%rdi), %xmm0
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm2 = ymm11[0,0,2,3,4,4,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm2[1,2,3],ymm1[4],ymm2[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm9[0,1],ymm7[2,3],ymm9[4,5],ymm7[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm2 = ymm2[2,0,2,3,6,4,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,2,0,3]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,3,2,3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm2 = ymm12[0,0,0,0,4,4,4,4]
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm2 = ymm6[0,0,2,3,4,4,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
-; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm0 = ymm11[2,3,2,3,6,7,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm1 = ymm10[0,0,2,3,4,4,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm5[0,1],ymm8[2,3],ymm5[4,5],ymm8[6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm1 = ymm1[2,0,2,3,6,4,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
-; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm1 = ymm15[0,0,0,0,4,4,4,4]
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm2 = ymm14[0,0,2,3,4,4,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
-; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm0 = ymm9[3,3,3,3,7,7,7,7]
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm1 = ymm13[0,1,3,3,4,5,7,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm1 = ymm7[3,3,3,3,7,7,7,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm4[1],ymm1[2,3,4],ymm4[5],ymm1[6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm3 = ymm6[0,0,2,3,4,4,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3],ymm3[4,5,6],ymm2[7]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,1,0,3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, (%rsp) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovaps 192(%rdi), %xmm1
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} xmm2 = xmm1[2,3,2,3]
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm3 = ymm10[0,0,2,3,4,4,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0],ymm3[1,2,3],ymm2[4],ymm3[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm5[0,1],ymm8[2,3],ymm5[4,5],ymm8[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm3 = ymm3[2,0,2,3,6,4,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm3 = ymm3[0,2,0,3]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,3,2,3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm3[3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm3 = ymm15[0,0,0,0,4,4,4,4]
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm4 = ymm14[0,0,2,3,4,4,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3],ymm4[4,5,6],ymm3[7]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm3 = ymm3[0,1,0,3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm2 = ymm11[0,1,3,3,4,5,7,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm2[1,2,3],ymm0[4],ymm2[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm2 = ymm7[3,3,3,3,7,7,7,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0],ymm9[1],ymm2[2,3,4],ymm9[5],ymm2[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,2,0,3]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm1 = ymm3[0,1,0,1,4,5,4,5]
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm3, %ymm9
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm2 = ymm6[0,1,3,3,4,5,7,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
-; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm0 = ymm11[3,3,3,3,7,7,7,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm2[3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm2 = ymm12[0,1,0,1,4,5,4,5]
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm3 = ymm6[0,1,3,3,4,5,7,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3],ymm3[4,5,6],ymm2[7]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,1,0,3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm13 = ymm0[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} xmm0 = xmm1[3,3,3,3]
; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm1 = ymm10[0,1,3,3,4,5,7,7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm1 = ymm8[3,3,3,3,7,7,7,7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm5[1],ymm1[2,3,4],ymm5[5],ymm1[6,7]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm1 = ymm15[0,1,0,1,4,5,4,5]
; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm2 = ymm14[0,1,3,3,4,5,7,7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, (%rsp) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm10[0,1,2,3],ymm11[4,5],ymm10[6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm1 = ymm10[0,1,2,3],mem[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm5[0,1,2,3],ymm8[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovaps 272(%rdi), %xmm3
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm3[2,3],ymm2[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovaps 272(%rdi), %xmm4
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm4[2,3],ymm2[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,1,0,2,4,5,4,6]
-; AVX2-FAST-PERLANE-NEXT: vmovaps {{.*#+}} xmm4 = <4,2,u,u>
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm1, %ymm4, %ymm8
+; AVX2-FAST-PERLANE-NEXT: vmovaps {{.*#+}} xmm5 = <4,2,u,u>
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm1, %ymm5, %ymm8
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm8[0,1],ymm0[2,3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm8 = ymm15[0,1],ymm14[2,3],ymm15[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vbroadcastf128 {{.*#+}} ymm10 = [0,2,0,6,0,2,0,6]
; AVX2-FAST-PERLANE-NEXT: # ymm10 = mem[0,1,0,1]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm8, %ymm10, %ymm11
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm11[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm12[0,1,2,3],ymm7[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm7 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm7 = ymm13[0,1,2,3],mem[4,5],ymm13[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm8, %ymm10, %ymm14
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm14[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm7 = ymm9[0,1,2,3],ymm7[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm9 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm9 = ymm11[0,1,2,3],mem[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovaps 80(%rdi), %xmm11
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm14 = ymm5[0,1],ymm11[2,3],ymm5[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm14 = ymm7[0,1],ymm11[2,3],ymm7[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm14 = ymm14[0,1,0,2,4,5,4,6]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm7, %ymm4, %ymm4
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm14[2,3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm14 = ymm9[0,1],ymm6[2,3],ymm9[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm9, %ymm5, %ymm5
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1],ymm14[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm14 = ymm12[0,1],ymm6[2,3],ymm12[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpermps %ymm14, %ymm10, %ymm10
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4],ymm10[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4],ymm10[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm2 = ymm2[1,1,1,1,5,5,5,5]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm3[3],ymm2[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovaps {{.*#+}} xmm3 = <5,3,u,u>
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm1, %ymm3, %ymm1
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm4[3],ymm2[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovaps {{.*#+}} xmm4 = <5,3,u,u>
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm1, %ymm4, %ymm1
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2,3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vbroadcastf128 {{.*#+}} ymm2 = [0,3,1,7,0,3,1,7]
; AVX2-FAST-PERLANE-NEXT: # ymm2 = mem[0,1,0,1]
; AVX2-FAST-PERLANE-NEXT: vpermps %ymm8, %ymm2, %ymm8
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm8[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm5 = ymm5[1,1,1,1,5,5,5,5]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2],ymm11[3],ymm5[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm7, %ymm3, %ymm3
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1],ymm5[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm7 = ymm7[1,1,1,1,5,5,5,5]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2],ymm11[3],ymm7[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm9, %ymm4, %ymm4
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm7[2,3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpermps %ymm14, %ymm2, %ymm2
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2,3,4],ymm2[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm3, 32(%rsi)
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm3, (%rsi)
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm3, 32(%rdx)
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm3, (%rdx)
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm3, 32(%rcx)
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm3, (%rcx)
-; AVX2-FAST-PERLANE-NEXT: vmovups (%rsp), %ymm3 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm4[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm4, 32(%rsi)
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm4, (%rsi)
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm4, 32(%rdx)
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm4, (%rdx)
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm4, 32(%rcx)
+; AVX2-FAST-PERLANE-NEXT: vmovups (%rsp), %ymm4 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm4, (%rcx)
; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm3, 32(%r8)
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm3, (%r8)
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm13, (%r8)
; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm0, 32(%r9)
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm4, (%r9)
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm5, (%r9)
; AVX2-FAST-PERLANE-NEXT: movq {{[0-9]+}}(%rsp), %rax
; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm1, 32(%rax)
; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm2, (%rax)
-; AVX2-FAST-PERLANE-NEXT: addq $264, %rsp # imm = 0x108
+; AVX2-FAST-PERLANE-NEXT: addq $200, %rsp
; AVX2-FAST-PERLANE-NEXT: vzeroupper
; AVX2-FAST-PERLANE-NEXT: retq
;
@@ -2774,61 +2753,56 @@ define void @load_i32_stride6_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
;
; AVX1-ONLY-LABEL: load_i32_stride6_vf32:
; AVX1-ONLY: # %bb.0:
-; AVX1-ONLY-NEXT: subq $1064, %rsp # imm = 0x428
+; AVX1-ONLY-NEXT: subq $1016, %rsp # imm = 0x3F8
+; AVX1-ONLY-NEXT: vmovaps 416(%rdi), %ymm7
+; AVX1-ONLY-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 480(%rdi), %ymm2
; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 448(%rdi), %ymm3
; AVX1-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 128(%rdi), %ymm4
-; AVX1-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 160(%rdi), %ymm5
-; AVX1-ONLY-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovapd 160(%rdi), %ymm4
+; AVX1-ONLY-NEXT: vmovupd %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovapd 128(%rdi), %ymm5
+; AVX1-ONLY-NEXT: vmovupd %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 32(%rdi), %ymm6
; AVX1-ONLY-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps (%rdi), %ymm7
-; AVX1-ONLY-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps (%rdi), %ymm9
+; AVX1-ONLY-NEXT: vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 96(%rdi), %ymm1
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 64(%rdi), %ymm0
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vinsertf128 $1, 96(%rdi), %ymm0, %ymm10
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm1[2,0],ymm10[0,0],ymm1[6,4],ymm10[4,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm10[2,2],ymm0[6,4],ymm10[6,6]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm7[0,1,2,3],ymm6[4,5],ymm7[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vextractf128 $1, %ymm1, %xmm9
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm9[2,3]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm9[0,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm5[0,1,2,3],ymm4[4,5],ymm5[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm4[2,3,0,1]
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm1[2,0],ymm4[0,0],ymm1[6,4],ymm4[4,4]
+; AVX1-ONLY-NEXT: vinsertf128 $1, 96(%rdi), %ymm0, %ymm8
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm1[2,0],ymm8[0,0],ymm1[6,4],ymm8[4,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm8[2,2],ymm0[6,4],ymm8[6,6]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm11 = ymm9[0,1,2,3],ymm6[4,5],ymm9[6,7]
+; AVX1-ONLY-NEXT: vextractf128 $1, %ymm11, %xmm6
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm11[0,1],xmm6[2,3]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm6[0,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm12 = ymm5[2,3],ymm4[0,1]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm1 = ymm12[0],ymm5[0],ymm12[3],ymm5[2]
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vinsertf128 $1, 480(%rdi), %ymm3, %ymm7
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm2[2,0],ymm7[0,0],ymm2[6,4],ymm7[4,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm7[2,2],ymm0[6,4],ymm7[6,6]
-; AVX1-ONLY-NEXT: vmovaps 416(%rdi), %ymm1
-; AVX1-ONLY-NEXT: vmovups %ymm1, (%rsp) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 384(%rdi), %ymm2
-; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5],ymm2[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vextractf128 $1, %ymm1, %xmm5
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm5[2,3]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm5[0,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT: vmovaps 512(%rdi), %ymm1
+; AVX1-ONLY-NEXT: vinsertf128 $1, 480(%rdi), %ymm3, %ymm5
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm2[2,0],ymm5[0,0],ymm2[6,4],ymm5[4,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm5[2,2],ymm0[6,4],ymm5[6,6]
+; AVX1-ONLY-NEXT: vmovaps 384(%rdi), %ymm1
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 544(%rdi), %ymm2
-; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm14 = ymm2[0,1,2,3],ymm1[4,5],ymm2[6,7]
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm14[2,3,0,1]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm7[4,5],ymm1[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm1[2,0],ymm14[0,0],ymm1[6,4],ymm14[4,4]
+; AVX1-ONLY-NEXT: vextractf128 $1, %ymm1, %xmm4
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm4[2,3]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm4[0,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovapd 544(%rdi), %ymm1
+; AVX1-ONLY-NEXT: vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovapd 512(%rdi), %ymm2
+; AVX1-ONLY-NEXT: vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm2[2,3],ymm1[0,1]
+; AVX1-ONLY-NEXT: vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm1 = ymm1[0],ymm2[0],ymm1[3],ymm2[2]
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -2843,19 +2817,18 @@ define void @load_i32_stride6_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 192(%rdi), %ymm2
; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm13 = ymm2[0,1,2,3],ymm1[4,5],ymm2[6,7]
-; AVX1-ONLY-NEXT: vextractf128 $1, %ymm13, %xmm2
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm13[0,1],xmm2[2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm10 = ymm2[0,1,2,3],ymm1[4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT: vextractf128 $1, %ymm10, %xmm2
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm10[0,1],xmm2[2,3]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm2[0,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT: vmovaps 320(%rdi), %ymm1
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 352(%rdi), %ymm4
-; AVX1-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm12 = ymm4[0,1,2,3],ymm1[4,5],ymm4[6,7]
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm12[2,3,0,1]
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm1[2,0],ymm12[0,0],ymm1[6,4],ymm12[4,4]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovapd 352(%rdi), %ymm1
+; AVX1-ONLY-NEXT: vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovapd 320(%rdi), %ymm7
+; AVX1-ONLY-NEXT: vmovupd %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm7[2,3],ymm1[0,1]
+; AVX1-ONLY-NEXT: vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm1 = ymm1[0],ymm7[0],ymm1[3],ymm7[2]
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -2863,107 +2836,103 @@ define void @load_i32_stride6_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 640(%rdi), %ymm0
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vinsertf128 $1, 672(%rdi), %ymm0, %ymm6
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm1[2,0],ymm6[0,0],ymm1[6,4],ymm6[4,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm4 = ymm0[2,0],ymm6[2,2],ymm0[6,4],ymm6[6,6]
+; AVX1-ONLY-NEXT: vinsertf128 $1, 672(%rdi), %ymm0, %ymm9
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm1[2,0],ymm9[0,0],ymm1[6,4],ymm9[4,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm0[2,0],ymm9[2,2],ymm0[6,4],ymm9[6,6]
; AVX1-ONLY-NEXT: vmovaps 608(%rdi), %ymm0
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 576(%rdi), %ymm1
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm11 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT: vextractf128 $1, %ymm11, %xmm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm15 = xmm11[0,1],xmm1[2,3]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm15 = xmm15[0,2],xmm1[0,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm15[0,1,2],ymm4[3,4,5],ymm15[6,7]
-; AVX1-ONLY-NEXT: vmovaps 704(%rdi), %ymm8
-; AVX1-ONLY-NEXT: vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 736(%rdi), %ymm0
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm0[0,1,2,3],ymm8[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm8[2,3,0,1]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm0[2,0],ymm8[0,0],ymm0[6,4],ymm8[4,4]
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} ymm15 = ymm15[0,1,2,0,4,5,6,4]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5],ymm15[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm4 = ymm15[3,0],ymm10[1,0],ymm15[7,4],ymm10[5,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm4 = ymm4[2,0],ymm10[2,3],ymm4[6,4],ymm10[6,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm10 = xmm0[1,0],xmm9[3,0]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm9 = xmm10[0,2],xmm9[1,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm9[0,1,2],ymm4[3,4,5],ymm9[6,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps $19, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm9 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm9 = ymm0[3,0],mem[1,0],ymm0[7,4],mem[5,4]
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} ymm9 = ymm9[0,1,2,0,4,5,6,4]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5],ymm9[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm4 = ymm9[3,0],ymm7[1,0],ymm9[7,4],ymm7[5,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm4 = ymm4[2,0],ymm7[2,3],ymm4[6,4],ymm7[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT: vextractf128 $1, %ymm7, %xmm1
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm14 = xmm7[0,1],xmm1[2,3]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm14 = xmm14[0,2],xmm1[0,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1,2],ymm15[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovapd 736(%rdi), %ymm13
+; AVX1-ONLY-NEXT: vmovupd %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovapd 704(%rdi), %ymm0
+; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm15 = ymm0[2,3],ymm13[0,1]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm13 = ymm15[0],ymm0[0],ymm15[3],ymm0[2]
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} ymm13 = ymm13[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm13 = ymm14[0,1,2,3,4,5],ymm13[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm13 = ymm14[3,0],ymm8[1,0],ymm14[7,4],ymm8[5,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm8 = ymm13[2,0],ymm8[2,3],ymm13[6,4],ymm8[6,7]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm11 = xmm11[1,0],xmm6[3,0]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm6 = xmm11[0,2],xmm6[1,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2],ymm8[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm8 = ymm12[3,1],ymm11[1,3],ymm12[7,5],ymm11[5,7]
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} ymm8 = ymm8[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5],ymm8[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm6 = ymm12[3,0],ymm5[1,0],ymm12[7,4],ymm5[5,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm5 = ymm6[2,0],ymm5[2,3],ymm6[6,4],ymm5[6,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm7 = xmm0[1,0],xmm5[3,0]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm5 = xmm7[0,2],xmm5[1,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm5[0,1,2],ymm4[3,4,5],ymm5[6,7]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm6 = xmm0[1,0],xmm4[3,0]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm4 = xmm6[0,2],xmm4[1,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2],ymm5[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm5 = ymm0[3,0],ymm14[1,0],ymm0[7,4],ymm14[5,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm5 = ymm0[3,1],ymm8[1,3],ymm0[7,5],ymm8[5,7]
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} ymm5 = ymm5[0,1,2,0,4,5,6,4]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5],ymm5[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm4 = ymm14[3,0],ymm3[1,0],ymm14[7,4],ymm3[5,4]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm4 = ymm0[3,0],ymm3[1,0],ymm0[7,4],ymm3[5,4]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm4[2,0],ymm3[2,3],ymm4[6,4],ymm3[6,7]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm4 = xmm13[1,0],xmm2[3,0]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm4 = xmm10[1,0],xmm2[3,0]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm2 = xmm4[0,2],xmm2[1,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm3[3,4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm3[3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm0[3,0],ymm12[1,0],ymm0[7,4],ymm12[5,4]
+; AVX1-ONLY-NEXT: vshufps $215, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm3 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm3 = ymm0[3,1],mem[1,3],ymm0[7,5],mem[5,7]
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} ymm3 = ymm3[0,1,2,0,4,5,6,4]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm3[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm5[3,0],ymm6[1,0],ymm5[7,4],ymm6[5,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm2[2,0],ymm6[2,3],ymm2[6,4],ymm6[6,7]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm2 = xmm11[1,0],xmm1[3,0]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm5[3,0],ymm9[1,0],ymm5[7,4],ymm9[5,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm2[2,0],ymm9[2,3],ymm2[6,4],ymm9[6,7]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm2 = xmm7[1,0],xmm1[3,0]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm0 = xmm2[0,2],xmm1[1,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm3[3,4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm1[3,0],ymm8[1,0],ymm1[7,4],ymm8[5,4]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm3[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm15[3,1],ymm7[1,3],ymm15[7,5],ymm7[5,7]
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm3 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm3 = ymm0[0,1],mem[2,3],ymm0[4,5,6,7]
-; AVX1-ONLY-NEXT: vshufps $38, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm1 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm1 = ymm15[2,1],mem[2,0],ymm15[6,5],mem[6,4]
+; AVX1-ONLY-NEXT: vshufps $38, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm1 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm1 = ymm14[2,1],mem[2,0],ymm14[6,5],mem[6,4]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm1[2,3,0,1]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm3, %xmm0
; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm2 = xmm3[2,0],xmm0[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4],ymm2[5,6,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm4 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm4 = ymm0[0,1,2,3],mem[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm4 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm4 = ymm11[0,1,2,3],mem[4,5],ymm11[6,7]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm4[2,3,0,1]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm0[0,0],ymm4[2,0],ymm0[4,4],ymm4[6,4]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm2[5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups (%rsp), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm2 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm2 = ymm0[0,1],mem[2,3],ymm0[4,5,6,7]
-; AVX1-ONLY-NEXT: vshufps $38, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm1 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm1 = ymm9[2,1],mem[2,0],ymm9[6,5],mem[6,4]
+; AVX1-ONLY-NEXT: vshufps $38, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm1 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm1 = ymm12[2,1],mem[2,0],ymm12[6,5],mem[6,4]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm1[2,3,0,1]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm2, %xmm0
-; AVX1-ONLY-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm6 = xmm2[2,0],xmm0[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm6[0,1,2],ymm1[3,4],ymm6[5,6,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm10 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm10 = ymm0[0,1,2,3],mem[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm6[0,1,2],ymm1[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm10 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm10 = ymm8[0,1,2,3],mem[4,5],ymm8[6,7]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm10[2,3,0,1]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm6 = ymm0[0,0],ymm10[2,0],ymm0[4,4],ymm10[6,4]
@@ -2972,16 +2941,15 @@ define void @load_i32_stride6_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm11 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm11 = ymm0[0,1],mem[2,3],ymm0[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm6 = ymm5[2,1],ymm7[2,0],ymm5[6,5],ymm7[6,4]
+; AVX1-ONLY-NEXT: vshufps $38, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm6 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm6 = ymm5[2,1],mem[2,0],ymm5[6,5],mem[6,4]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm8 = ymm6[2,3,0,1]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm11, %xmm0
; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm9 = xmm11[2,0],xmm0[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm12 = ymm9[0,1,2],ymm8[3,4],ymm9[5,6,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm8 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm8 = ymm0[0,1,2,3],mem[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm12 = ymm9[0,1,2],ymm8[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm8 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm8 = ymm7[0,1,2,3],mem[4,5],ymm7[6,7]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm9 = ymm8[2,3,0,1]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm13 = ymm9[0,0],ymm8[2,0],ymm9[4,4],ymm8[6,4]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm12[0,1,2,3,4],ymm13[5,6,7]
@@ -2989,12 +2957,13 @@ define void @load_i32_stride6_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm12 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm12 = ymm0[0,1],mem[2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm13 = ymm14[2,1],ymm1[2,0],ymm14[6,5],ymm1[6,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm13 = ymm1[2,1],ymm6[2,0],ymm1[6,5],ymm6[6,4]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm13 = ymm13[2,3,0,1]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm12, %xmm14
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm15 = xmm12[2,0],xmm14[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm13 = ymm15[0,1,2],ymm13[3,4],ymm15[5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm13 = ymm15[0,1,2],ymm13[3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm15 = ymm0[0,1,2,3],mem[4,5],ymm0[6,7]
@@ -3007,44 +2976,45 @@ define void @load_i32_stride6_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: vshufps $247, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm3 = xmm3[3,1],mem[3,3]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm5 = ymm13[3,1],ymm6[2,1],ymm13[7,5],ymm6[6,5]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm5 = ymm13[3,1],ymm7[2,1],ymm13[7,5],ymm7[6,5]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm5 = ymm5[2,3,0,1]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2],ymm5[3,4],ymm3[5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2],ymm5[3,4,5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4],ymm4[5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm3[0,1],ymm10[3,1],ymm3[4,5],ymm10[7,5]
-; AVX1-ONLY-NEXT: vshufps $247, (%rsp), %xmm2, %xmm2 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: vshufps $247, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm2 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm2 = xmm2[3,1],mem[3,3]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm4 = ymm5[3,1],ymm10[2,1],ymm5[7,5],ymm10[6,5]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm4 = ymm4[2,3,0,1]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm4[3,4],ymm2[5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm4[3,4,5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm2, (%rsp) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[0,1],ymm15[3,1],ymm0[4,5],ymm15[7,5]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm2 = xmm12[3,1],xmm14[3,3]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm1, %ymm14
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm4[3,1],ymm1[2,1],ymm4[7,5],ymm1[6,5]
+; AVX1-ONLY-NEXT: vmovaps %ymm1, %ymm4
+; AVX1-ONLY-NEXT: vmovaps %ymm6, %ymm14
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm1[3,1],ymm6[2,1],ymm1[7,5],ymm6[6,5]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm3 = ymm3[2,3,0,1]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm3[3,4],ymm2[5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm3[3,4,5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3,4],ymm0[5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm9[0,1],ymm8[3,1],ymm9[4,5],ymm8[7,5]
; AVX1-ONLY-NEXT: vshufps $247, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11, %xmm1 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm1 = xmm11[3,1],mem[3,3]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm8[3,1],ymm7[2,1],ymm8[7,5],ymm7[6,5]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm2[3,1],ymm15[2,1],ymm2[7,5],ymm15[6,5]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm2 = ymm2[2,3,0,1]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4],ymm1[5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm7 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm7 = mem[0,1],ymm0[2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm8 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm8 = mem[0,1],ymm0[2,3],mem[4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps 32(%rdi), %xmm0
; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[2,2,3,3]
@@ -3053,13 +3023,13 @@ define void @load_i32_stride6_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm1[0],xmm0[1],xmm1[2,3]
; AVX1-ONLY-NEXT: vmovapd 80(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm1 = ymm1[1],ymm6[0],ymm1[2],ymm6[2]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm1 = ymm1[1],ymm7[0],ymm1[2],ymm7[2]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm13[0,1],ymm1[2,0],ymm13[4,5],ymm1[6,4]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm11 = ymm7[2,3,0,1]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm7[2,0],ymm11[0,0],ymm7[6,4],ymm11[4,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm11[0,2],ymm1[2,0],ymm11[4,6],ymm1[6,4]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm8[2,3,0,1]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm8[2,0],ymm1[0,0],ymm8[6,4],ymm1[4,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm1[0,2],ymm2[2,0],ymm1[4,6],ymm2[6,4]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm6 # 32-byte Folded Reload
@@ -3067,100 +3037,99 @@ define void @load_i32_stride6_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: vmovaps 416(%rdi), %xmm0
; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[2,2,3,3]
-; AVX1-ONLY-NEXT: vmovaps 400(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm1[0],xmm0[1],xmm1[2,3]
-; AVX1-ONLY-NEXT: vmovapd 464(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm1 = ymm1[1],ymm10[0],ymm1[2],ymm10[2]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm5[0,1],ymm1[2,0],ymm5[4,5],ymm1[6,4]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm6[2,3,0,1]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm6[2,0],ymm1[0,0],ymm6[6,4],ymm1[4,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm1[0,2],ymm2[2,0],ymm1[4,6],ymm2[6,4]
+; AVX1-ONLY-NEXT: vmovaps 400(%rdi), %xmm2
+; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm2[0],xmm0[1],xmm2[2,3]
+; AVX1-ONLY-NEXT: vmovapd 464(%rdi), %xmm2
+; AVX1-ONLY-NEXT: vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm2 = ymm2[1],ymm10[0],ymm2[2],ymm10[2]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm5[0,1],ymm2[2,0],ymm5[4,5],ymm2[6,4]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm2[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm3 = ymm6[2,3,0,1]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm6[2,0],ymm3[0,0],ymm6[6,4],ymm3[4,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm3[0,2],ymm2[2,0],ymm3[4,6],ymm2[6,4]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm12 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm12 = mem[0,1],ymm0[2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm9 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm9 = mem[0,1],ymm0[2,3],mem[4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps 224(%rdi), %xmm0
; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[2,2,3,3]
-; AVX1-ONLY-NEXT: vmovaps 208(%rdi), %xmm15
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm15[0],xmm0[1],xmm15[2,3]
-; AVX1-ONLY-NEXT: vmovapd 272(%rdi), %xmm2
-; AVX1-ONLY-NEXT: vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm2 = ymm2[1],ymm14[0],ymm2[2],ymm14[2]
+; AVX1-ONLY-NEXT: vmovaps 208(%rdi), %xmm13
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm13[0],xmm0[1],xmm13[2,3]
+; AVX1-ONLY-NEXT: vmovapd 272(%rdi), %xmm12
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm2 = ymm12[1],ymm14[0],ymm12[2],ymm14[2]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm4[0,1],ymm2[2,0],ymm4[4,5],ymm2[6,4]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm0[0,1],ymm2[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm12[2,3,0,1]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm12[2,0],ymm0[0,0],ymm12[6,4],ymm0[4,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm0[0,2],ymm3[2,0],ymm0[4,6],ymm3[6,4]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7]
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm9[2,3,0,1]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm5 = ymm9[2,0],ymm0[0,0],ymm9[6,4],ymm0[4,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm5 = ymm0[0,2],ymm5[2,0],ymm0[4,6],ymm5[6,4]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm5[5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX1-ONLY-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm4 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm4 = mem[0,1],ymm2[2,3],mem[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovaps 608(%rdi), %xmm13
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm3 = xmm13[2,2,3,3]
-; AVX1-ONLY-NEXT: vmovaps 592(%rdi), %xmm5
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm5[0],xmm3[1],xmm5[2,3]
-; AVX1-ONLY-NEXT: vmovapd 656(%rdi), %xmm9
-; AVX1-ONLY-NEXT: vmovupd {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm10 = ymm9[1],ymm3[0],ymm9[2],ymm3[2]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm10 = ymm8[0,1],ymm10[2,0],ymm8[4,5],ymm10[6,4]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm10[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm10 = ymm4[2,3,0,1]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm14 = ymm4[2,0],ymm10[0,0],ymm4[6,4],ymm10[4,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm14 = ymm10[0,2],ymm14[2,0],ymm10[4,6],ymm14[6,4]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm14[5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm7[3,0],ymm11[1,0],ymm7[7,4],ymm11[5,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm11[0,3],ymm2[2,0],ymm11[4,7],ymm2[6,4]
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; AVX1-ONLY-NEXT: vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7, %xmm7 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm7 = xmm7[0,1],mem[2,3]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps $215, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm11 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm11 = ymm11[3,1],mem[1,3],ymm11[7,5],mem[5,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm11 = ymm14[1,1],ymm11[2,0],ymm14[5,5],ymm11[6,4]
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm7 = xmm7[1,3,2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1],ymm11[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm7[0,1,2,3,4],ymm2[5,6,7]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm6 = ymm6[3,0],ymm1[1,0],ymm6[7,4],ymm1[5,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm1[0,3],ymm6[2,0],ymm1[4,7],ymm6[6,4]
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX1-ONLY-NEXT: vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6, %xmm6 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm6 = xmm6[0,1],mem[2,3]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps $215, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm7 = ymm7[3,1],mem[1,3],ymm7[7,5],mem[5,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm7 = ymm11[1,1],ymm7[2,0],ymm11[5,5],ymm7[6,4]
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm6 = xmm6[1,3,2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1],ymm7[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm6[0,1,2,3,4],ymm1[5,6,7]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm6 = ymm12[3,0],ymm0[1,0],ymm12[7,4],ymm0[5,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[0,3],ymm6[2,0],ymm0[4,7],ymm6[6,4]
-; AVX1-ONLY-NEXT: vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm6 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm6 = xmm15[0,1],mem[2,3]
+; AVX1-ONLY-NEXT: vmovaps 608(%rdi), %xmm11
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm5 = xmm11[2,2,3,3]
+; AVX1-ONLY-NEXT: vmovaps 592(%rdi), %xmm10
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm10[0],xmm5[1],xmm10[2,3]
+; AVX1-ONLY-NEXT: vmovapd 656(%rdi), %xmm5
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm14 = ymm5[1],ymm15[0],ymm5[2],ymm15[2]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps $215, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm7 = ymm7[3,1],mem[1,3],ymm7[7,5],mem[5,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm7 = ymm11[1,1],ymm7[2,0],ymm11[5,5],ymm7[6,4]
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm6 = xmm6[1,3,2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1],ymm7[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm6[0,1,2,3,4],ymm0[5,6,7]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm4 = ymm4[3,0],ymm10[1,0],ymm4[7,4],ymm10[5,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm4 = ymm10[0,3],ymm4[2,0],ymm10[4,7],ymm4[6,4]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm5 = xmm5[0,1],xmm13[2,3]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm9[3,1],ymm3[1,3],ymm9[7,5],ymm3[5,7]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm8[1,1],ymm3[2,0],ymm8[5,5],ymm3[6,4]
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm5 = xmm5[1,3,2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm5[0,1],ymm3[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4],ymm4[5,6,7]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm14 = ymm7[0,1],ymm14[2,0],ymm7[4,5],ymm14[6,4]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm14[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm14 = ymm4[2,3,0,1]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm4[2,0],ymm14[0,0],ymm4[6,4],ymm14[4,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm14[0,2],ymm15[2,0],ymm14[4,6],ymm15[6,4]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm15[5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm8[3,0],ymm1[1,0],ymm8[7,4],ymm1[5,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm1[0,3],ymm2[2,0],ymm1[4,7],ymm2[6,4]
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX1-ONLY-NEXT: vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm2 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm2 = xmm2[0,1],mem[2,3]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps $215, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm8 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm8 = ymm8[3,1],mem[1,3],ymm8[7,5],mem[5,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm8 = ymm15[1,1],ymm8[2,0],ymm15[5,5],ymm8[6,4]
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm2 = xmm2[1,3,2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm8[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm1[5,6,7]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm6[3,0],ymm3[1,0],ymm6[7,4],ymm3[5,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm3[0,3],ymm1[2,0],ymm3[4,7],ymm1[6,4]
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX1-ONLY-NEXT: vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm3 = xmm3[0,1],mem[2,3]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps $215, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm6 = ymm6[3,1],mem[1,3],ymm6[7,5],mem[5,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm6 = ymm8[1,1],ymm6[2,0],ymm8[5,5],ymm6[6,4]
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm3 = xmm3[1,3,2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1],ymm6[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm3[0,1,2,3,4],ymm1[5,6,7]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm9[3,0],ymm0[1,0],ymm9[7,4],ymm0[5,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[0,3],ymm3[2,0],ymm0[4,7],ymm3[6,4]
+; AVX1-ONLY-NEXT: vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13, %xmm3 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm3 = xmm13[0,1],mem[2,3]
+; AVX1-ONLY-NEXT: vshufps $215, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm6 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm6 = ymm12[3,1],mem[1,3],ymm12[7,5],mem[5,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm6 = ymm8[1,1],ymm6[2,0],ymm8[5,5],ymm6[6,4]
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm3 = xmm3[1,3,2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1],ymm6[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3,4],ymm0[5,6,7]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm4[3,0],ymm14[1,0],ymm4[7,4],ymm14[5,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm14[0,3],ymm3[2,0],ymm14[4,7],ymm3[6,4]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm4 = xmm10[0,1],xmm11[2,3]
+; AVX1-ONLY-NEXT: vshufps $215, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm5 = ymm5[3,1],mem[1,3],ymm5[7,5],mem[5,7]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm5 = ymm7[1,1],ymm5[2,0],ymm7[5,5],ymm5[6,4]
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm4 = xmm4[1,3,2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm5[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm4[0,1,2,3,4],ymm3[5,6,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm4, 96(%rsi)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
@@ -3189,7 +3158,7 @@ define void @load_i32_stride6_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: vmovaps %ymm4, 96(%r8)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm4, 32(%r8)
-; AVX1-ONLY-NEXT: vmovups (%rsp), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm4, 64(%r8)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm4, (%r8)
@@ -3206,371 +3175,371 @@ define void @load_i32_stride6_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: vmovaps %ymm0, 32(%rax)
; AVX1-ONLY-NEXT: vmovaps %ymm1, 64(%rax)
; AVX1-ONLY-NEXT: vmovaps %ymm2, (%rax)
-; AVX1-ONLY-NEXT: addq $1064, %rsp # imm = 0x428
+; AVX1-ONLY-NEXT: addq $1016, %rsp # imm = 0x3F8
; AVX1-ONLY-NEXT: vzeroupper
; AVX1-ONLY-NEXT: retq
;
; AVX2-SLOW-LABEL: load_i32_stride6_vf32:
; AVX2-SLOW: # %bb.0:
-; AVX2-SLOW-NEXT: subq $1192, %rsp # imm = 0x4A8
+; AVX2-SLOW-NEXT: subq $1160, %rsp # imm = 0x488
; AVX2-SLOW-NEXT: vmovaps 480(%rdi), %ymm6
; AVX2-SLOW-NEXT: vmovaps 448(%rdi), %ymm7
; AVX2-SLOW-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovaps 416(%rdi), %ymm8
; AVX2-SLOW-NEXT: vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovaps 128(%rdi), %ymm2
-; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovaps 128(%rdi), %ymm0
+; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovaps 160(%rdi), %ymm4
; AVX2-SLOW-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovaps 96(%rdi), %ymm3
; AVX2-SLOW-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovaps (%rdi), %ymm0
-; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovaps 32(%rdi), %ymm1
+; AVX2-SLOW-NEXT: vmovaps (%rdi), %ymm1
; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovaps 32(%rdi), %ymm2
+; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovaps 64(%rdi), %ymm5
; AVX2-SLOW-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovaps {{.*#+}} xmm13 = <0,6,4,u>
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm12 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
-; AVX2-SLOW-NEXT: vpermps %ymm12, %ymm13, %ymm1
-; AVX2-SLOW-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm5[0,1],ymm3[0,1]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm9 = ymm0[0,1,2,3,4,5],ymm3[6,7]
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm3 = ymm9[0,2,2,2,4,6,6,6]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm1[0,1,2],ymm3[3,4,5],ymm1[6,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm10 = ymm4[0,1,2,3],ymm2[4,5],ymm4[6,7]
+; AVX2-SLOW-NEXT: vmovaps {{.*#+}} xmm14 = <0,6,4,u>
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpermps %ymm1, %ymm14, %ymm2
+; AVX2-SLOW-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm5[0,1],ymm3[0,1]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm11 = ymm1[0,1,2,3,4,5],ymm3[6,7]
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm3 = ymm11[0,2,2,2,4,6,6,6]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm2[0,1,2],ymm3[3,4,5,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm10 = ymm4[0,1,2,3],ymm0[4,5,6,7]
; AVX2-SLOW-NEXT: vbroadcastsd {{.*#+}} ymm2 = [8589934596,8589934596,8589934596,8589934596]
; AVX2-SLOW-NEXT: vpermps %ymm10, %ymm2, %ymm5
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm3[0,1,2,3,4,5],ymm5[6,7]
; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovups %ymm6, (%rsp) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vperm2f128 {{.*#+}} ymm3 = ymm7[0,1],ymm6[0,1]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm7 = ymm3[0,1,2,3,4,5],ymm6[6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm12 = ymm3[0,1,2,3,4,5],ymm6[6,7]
; AVX2-SLOW-NEXT: vmovaps 384(%rdi), %ymm0
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm6 = ymm0[0,1,2,3],ymm8[4,5],ymm0[6,7]
-; AVX2-SLOW-NEXT: vpermps %ymm6, %ymm13, %ymm5
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm8 = ymm7[0,2,2,2,4,6,6,6]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2],ymm8[3,4,5],ymm5[6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm6, %ymm14, %ymm5
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm7 = ymm12[0,2,2,2,4,6,6,6]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2],ymm7[3,4,5,6,7]
; AVX2-SLOW-NEXT: vmovaps 512(%rdi), %ymm0
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovaps 544(%rdi), %ymm1
; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
-; AVX2-SLOW-NEXT: vpermps %ymm4, %ymm2, %ymm8
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm5[0,1,2,3,4,5],ymm8[6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm4, %ymm2, %ymm9
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm5[0,1,2,3,4,5],ymm9[6,7]
; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovaps 288(%rdi), %ymm1
; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovaps 256(%rdi), %ymm0
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vperm2f128 {{.*#+}} ymm8 = ymm0[0,1],ymm1[0,1]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm11 = ymm8[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-SLOW-NEXT: vperm2f128 {{.*#+}} ymm5 = ymm0[0,1],ymm1[0,1]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm9 = ymm5[0,1,2,3,4,5],ymm1[6,7]
; AVX2-SLOW-NEXT: vmovaps 224(%rdi), %ymm0
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovaps 192(%rdi), %ymm1
; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
-; AVX2-SLOW-NEXT: vpermps %ymm3, %ymm13, %ymm1
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm15 = ymm11[0,2,2,2,4,6,6,6]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm15[3,4,5],ymm1[6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm3, %ymm14, %ymm15
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm13 = ymm9[0,2,2,2,4,6,6,6]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm13 = ymm15[0,1,2],ymm13[3,4,5,6,7]
; AVX2-SLOW-NEXT: vmovaps 320(%rdi), %ymm0
-; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovaps 352(%rdi), %ymm5
-; AVX2-SLOW-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm15 = ymm5[0,1,2,3],ymm0[4,5],ymm5[6,7]
-; AVX2-SLOW-NEXT: vpermps %ymm15, %ymm2, %ymm8
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm8[6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm0, (%rsp) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovaps 352(%rdi), %ymm1
; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm8 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm8, %ymm2, %ymm15
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm13[0,1,2,3,4,5],ymm15[6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovaps 608(%rdi), %ymm0
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovaps 576(%rdi), %ymm1
; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
-; AVX2-SLOW-NEXT: vpermps %ymm5, %ymm13, %ymm8
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm13 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm13, %ymm14, %ymm14
; AVX2-SLOW-NEXT: vmovaps 672(%rdi), %ymm1
; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovaps 640(%rdi), %ymm0
-; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vperm2f128 {{.*#+}} ymm13 = ymm0[0,1],ymm1[0,1]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3,4,5],ymm1[6,7]
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm14 = ymm13[0,2,2,2,4,6,6,6]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2],ymm14[3,4,5],ymm8[6,7]
+; AVX2-SLOW-NEXT: vmovaps 640(%rdi), %ymm15
+; AVX2-SLOW-NEXT: vperm2f128 {{.*#+}} ymm7 = ymm15[0,1],ymm1[0,1]
+; AVX2-SLOW-NEXT: vmovups %ymm15, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm5 = ymm7[0,2,2,2,4,6,6,6]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm14[0,1,2],ymm5[3,4,5,6,7]
; AVX2-SLOW-NEXT: vmovaps 704(%rdi), %ymm1
; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovaps 736(%rdi), %ymm0
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm14 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm14 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-SLOW-NEXT: vpermps %ymm14, %ymm2, %ymm0
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm8[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm5[0,1,2,3,4,5],ymm0[6,7]
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovaps {{.*#+}} xmm0 = <1,7,5,u>
-; AVX2-SLOW-NEXT: vpermps %ymm12, %ymm0, %ymm8
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm9 = ymm9[1,3,2,3,5,7,6,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2],ymm9[3,4,5],ymm8[6,7]
-; AVX2-SLOW-NEXT: vbroadcastsd {{.*#+}} ymm9 = [12884901893,12884901893,12884901893,12884901893]
-; AVX2-SLOW-NEXT: vpermps %ymm10, %ymm9, %ymm10
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5],ymm10[6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpermps %ymm6, %ymm0, %ymm6
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm7 = ymm7[1,3,2,3,5,7,6,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2],ymm7[3,4,5],ymm6[6,7]
-; AVX2-SLOW-NEXT: vpermps %ymm4, %ymm9, %ymm4
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm6[0,1,2,3,4,5],ymm4[6,7]
+; AVX2-SLOW-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm5 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm11 = ymm11[1,3,2,3,5,7,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2],ymm11[3,4,5,6,7]
+; AVX2-SLOW-NEXT: vbroadcastsd {{.*#+}} ymm11 = [12884901893,12884901893,12884901893,12884901893]
+; AVX2-SLOW-NEXT: vpermps %ymm10, %ymm11, %ymm10
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5],ymm10[6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpermps %ymm6, %ymm0, %ymm5
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm6 = ymm12[1,3,2,3,5,7,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2],ymm6[3,4,5,6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm4, %ymm11, %ymm4
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm5[0,1,2,3,4,5],ymm4[6,7]
; AVX2-SLOW-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpermps %ymm3, %ymm0, %ymm2
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm3 = ymm11[1,3,2,3,5,7,6,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm3[3,4,5],ymm2[6,7]
-; AVX2-SLOW-NEXT: vpermps %ymm15, %ymm9, %ymm3
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm3[6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpermps %ymm5, %ymm0, %ymm0
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm1 = ymm13[1,3,2,3,5,7,6,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5],ymm0[6,7]
-; AVX2-SLOW-NEXT: vpermps %ymm14, %ymm9, %ymm1
+; AVX2-SLOW-NEXT: vpermps %ymm3, %ymm0, %ymm1
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm3 = ymm9[1,3,2,3,5,7,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm3[3,4,5,6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm8, %ymm11, %ymm2
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpermps %ymm13, %ymm0, %ymm0
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm1 = ymm7[1,3,2,3,5,7,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm14, %ymm11, %ymm1
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm0 = ymm10[2,3,2,3,6,7,6,7]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm1 = ymm11[0,0,2,3,4,4,6,7]
+; AVX2-SLOW-NEXT: vmovaps (%rdi), %xmm2
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} xmm0 = xmm2[2,3,2,3]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm1 = ymm13[0,0,2,3,4,4,6,7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm8[0,1],ymm9[2,3],ymm8[4,5],ymm9[6,7]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm11[0,1],ymm12[2,3],ymm11[4,5],ymm12[6,7]
; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm1 = ymm1[2,0,2,3,6,4,6,7]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm1 = ymm6[0,0,0,0,4,4,4,4]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm2 = ymm7[0,0,2,3,4,4,6,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm1 = ymm9[0,0,0,0,4,4,4,4]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm3 = ymm10[0,0,2,3,4,4,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm3[0,1,2],ymm1[3],ymm3[4,5,6],ymm1[7]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm0 = mem[2,3,2,3,6,7,6,7]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm1 = ymm14[0,0,2,3,4,4,6,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
-; AVX2-SLOW-NEXT: vmovups (%rsp), %ymm1 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm1 = mem[0,1],ymm1[2,3],mem[4,5],ymm1[6,7]
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm1 = ymm1[2,0,2,3,6,4,6,7]
-; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
+; AVX2-SLOW-NEXT: vmovaps 384(%rdi), %xmm1
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} xmm0 = xmm1[2,3,2,3]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm3 = ymm8[0,0,2,3,4,4,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm3[1,2,3],ymm0[4],ymm3[5,6,7]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-SLOW-NEXT: vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm3 = mem[0,1],ymm3[2,3],mem[4,5],ymm3[6,7]
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm3 = ymm3[2,0,2,3,6,4,6,7]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm3 = ymm3[0,2,0,3]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
-; AVX2-SLOW-NEXT: vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm1 = mem[0,0,0,0,4,4,4,4]
-; AVX2-SLOW-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm2 = mem[0,0,2,3,4,4,6,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
-; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm3[3,4,5,6,7]
+; AVX2-SLOW-NEXT: vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm3 = mem[0,0,0,0,4,4,4,4]
+; AVX2-SLOW-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm4 = mem[0,0,2,3,4,4,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3],ymm4[4,5,6],ymm3[7]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm3 = ymm3[0,1,0,3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm3[5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm0 = mem[2,3,2,3,6,7,6,7]
-; AVX2-SLOW-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm1 = mem[0,0,2,3,4,4,6,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm1 = mem[0,1],ymm1[2,3],mem[4,5],ymm1[6,7]
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm1 = ymm1[2,0,2,3,6,4,6,7]
-; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
+; AVX2-SLOW-NEXT: vmovaps 576(%rdi), %xmm0
+; AVX2-SLOW-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[2,3,2,3]
+; AVX2-SLOW-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm4 = mem[0,0,2,3,4,4,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm4[1,2,3],ymm0[4],ymm4[5,6,7]
+; AVX2-SLOW-NEXT: vblendps $204, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm4 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm4 = ymm15[0,1],mem[2,3],ymm15[4,5],mem[6,7]
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm4 = ymm4[2,0,2,3,6,4,6,7]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm4 = ymm4[0,2,0,3]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
-; AVX2-SLOW-NEXT: vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm1 = mem[0,0,0,0,4,4,4,4]
-; AVX2-SLOW-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm2 = mem[0,0,2,3,4,4,6,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
-; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm4[3,4,5,6,7]
+; AVX2-SLOW-NEXT: vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm4 = mem[0,0,0,0,4,4,4,4]
+; AVX2-SLOW-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm5 = mem[0,0,2,3,4,4,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm5[0,1,2],ymm4[3],ymm5[4,5,6],ymm4[7]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm4 = ymm4[0,1,0,3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm4[5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovaps 192(%rdi), %xmm0
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} xmm4 = xmm0[2,3,2,3]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm5 = ymm7[0,0,2,3,4,4,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0],ymm5[1,2,3],ymm4[4],ymm5[5,6,7]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm14[0,1],ymm15[2,3],ymm14[4,5],ymm15[6,7]
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm5 = ymm5[2,0,2,3,6,4,6,7]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm5 = ymm5[0,2,0,3]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm4 = ymm4[0,3,2,3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2],ymm5[3,4,5,6,7]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm5 = ymm3[0,0,0,0,4,4,4,4]
+; AVX2-SLOW-NEXT: vpermilps $224, (%rsp), %ymm6 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm6 = mem[0,0,2,3,4,4,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm6[0,1,2],ymm5[3],ymm6[4,5,6],ymm5[7]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm5 = ymm5[0,1,0,3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4],ymm5[5,6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} xmm2 = xmm2[3,3,3,3]
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm4 = ymm13[0,1,3,3,4,5,7,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0],ymm4[1,2,3],ymm2[4],ymm4[5,6,7]
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm4 = ymm12[3,3,3,3,7,7,7,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0],ymm11[1],ymm4[2,3,4],ymm11[5],ymm4[6,7]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm4 = ymm4[0,2,0,3]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,3,2,3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm4[3,4,5,6,7]
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm4 = ymm9[0,1,0,1,4,5,4,5]
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm5 = ymm10[0,1,3,3,4,5,7,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm5[0,1,2],ymm4[3],ymm5[4,5,6],ymm4[7]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm4 = ymm4[0,1,0,3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm4[5,6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} xmm1 = xmm1[3,3,3,3]
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm2 = ymm8[0,1,3,3,4,5,7,7]
+; AVX2-SLOW-NEXT: vmovaps %ymm8, %ymm13
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm2[1,2,3],ymm1[4],ymm2[5,6,7]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm2 = ymm9[3,3,3,3,7,7,7,7]
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm0 = ymm5[2,3,2,3,6,7,6,7]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm1 = ymm4[0,0,2,3,4,4,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0],ymm5[1],ymm2[2,3,4],ymm5[5],ymm2[6,7]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,2,0,3]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,3,2,3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm2 = ymm8[0,1,0,1,4,5,4,5]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm4 = ymm6[0,1,3,3,4,5,7,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm4[0,1,2],ymm2[3],ymm4[4,5,6],ymm2[7]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,1,0,3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm1 = ymm7[0,1,3,3,4,5,7,7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm12[0,1],ymm13[2,3],ymm12[4,5],ymm13[6,7]
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm1 = ymm1[2,0,2,3,6,4,6,7]
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm1 = ymm15[3,3,3,3,7,7,7,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm14[1],ymm1[2,3,4],ymm14[5],ymm1[6,7]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm1 = ymm3[0,0,0,0,4,4,4,4]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm2 = ymm15[0,0,2,3,4,4,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm1 = ymm3[0,1,0,1,4,5,4,5]
+; AVX2-SLOW-NEXT: vmovups (%rsp), %ymm10 # 32-byte Reload
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm2 = ymm10[0,1,3,3,4,5,7,7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm0 = ymm10[3,3,3,3,7,7,7,7]
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm1 = ymm11[0,1,3,3,4,5,7,7]
+; AVX2-SLOW-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; AVX2-SLOW-NEXT: # xmm0 = mem[3,3,3,3]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm1 = ymm3[0,1,3,3,4,5,7,7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm1 = ymm9[3,3,3,3,7,7,7,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm8[1],ymm1[2,3,4],ymm8[5],ymm1[6,7]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm1 = ymm12[3,3,3,3,7,7,7,7]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm11[1],ymm1[2,3,4],ymm11[5],ymm1[6,7]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm1 = ymm6[0,1,0,1,4,5,4,5]
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm2 = ymm7[0,1,3,3,4,5,7,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
+; AVX2-SLOW-NEXT: vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm1 = mem[0,1,0,1,4,5,4,5]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm2 = ymm15[0,1,3,3,4,5,7,7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm0 = ymm6[3,3,3,3,7,7,7,7]
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm1 = ymm14[0,1,3,3,4,5,7,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
-; AVX2-SLOW-NEXT: vmovups (%rsp), %ymm8 # 32-byte Reload
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm1 = ymm8[3,3,3,3,7,7,7,7]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm10[1],ymm1[2,3,4],ymm10[5],ymm1[6,7]
-; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
-; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm1 = ymm14[0,1,0,1,4,5,4,5]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm2 = ymm7[0,1,3,3,4,5,7,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
-; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-SLOW-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm0 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm0 = ymm13[0,1,2,3],mem[4,5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm0 = ymm5[3,3,3,3,7,7,7,7]
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm1 = ymm4[0,1,3,3,4,5,7,7]
-; AVX2-SLOW-NEXT: vmovaps %ymm4, %ymm5
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm1 = ymm13[3,3,3,3,7,7,7,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm12[1],ymm1[2,3,4],ymm12[5],ymm1[6,7]
-; AVX2-SLOW-NEXT: vmovaps %ymm12, %ymm13
-; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
-; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm1 = ymm3[0,1,0,1,4,5,4,5]
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm2 = ymm15[0,1,3,3,4,5,7,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
-; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm0 = ymm3[3,3,3,3,7,7,7,7]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm1 = ymm12[0,1,3,3,4,5,7,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm1 = ymm11[3,3,3,3,7,7,7,7]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm9[1],ymm1[2,3,4],ymm9[5],ymm1[6,7]
-; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
-; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
-; AVX2-SLOW-NEXT: vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm1 = mem[0,1,0,1,4,5,4,5]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm2 = ymm15[0,1,3,3,4,5,7,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
-; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vblendps $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm0 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm0 = mem[0,1,2,3],ymm6[4,5],mem[6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm10[0,1,2,3],ymm8[4,5,6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm14 = ymm5[0,1,2,3],ymm9[4,5,6,7]
; AVX2-SLOW-NEXT: vmovaps 464(%rdi), %xmm1
; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3],ymm2[4,5,6,7]
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
-; AVX2-SLOW-NEXT: vmovaps {{.*#+}} xmm2 = <4,2,u,u>
-; AVX2-SLOW-NEXT: vpermps %ymm0, %ymm2, %ymm4
-; AVX2-SLOW-NEXT: vmovaps %ymm2, %ymm6
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm4[0,1],ymm1[2,3,4,5,6,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm7 = ymm14[0,1],ymm7[2,3],ymm14[4,5,6,7]
-; AVX2-SLOW-NEXT: vbroadcastf128 {{.*#+}} ymm10 = [0,2,0,6,0,2,0,6]
-; AVX2-SLOW-NEXT: # ymm10 = mem[0,1,0,1]
-; AVX2-SLOW-NEXT: vpermps %ymm7, %ymm10, %ymm4
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm4[5,6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm0 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm0 = ymm5[0,1,2,3],mem[4,5],ymm5[6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm14[0,1],ymm1[2,3],ymm14[4,5,6,7]
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm2 = ymm2[0,1,0,2,4,5,4,6]
+; AVX2-SLOW-NEXT: vmovaps {{.*#+}} xmm1 = <4,2,u,u>
+; AVX2-SLOW-NEXT: vpermps %ymm0, %ymm1, %ymm4
+; AVX2-SLOW-NEXT: vmovaps %ymm1, %ymm9
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm2[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm7 = ymm8[0,1],ymm6[2,3],ymm8[4,5,6,7]
+; AVX2-SLOW-NEXT: vbroadcastf128 {{.*#+}} ymm2 = [0,2,0,6,0,2,0,6]
+; AVX2-SLOW-NEXT: # ymm2 = mem[0,1,0,1]
+; AVX2-SLOW-NEXT: vpermps %ymm7, %ymm2, %ymm5
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm4[0,1,2,3,4],ymm5[5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm2 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm2 = ymm13[0,1,2,3],mem[4,5,6,7]
-; AVX2-SLOW-NEXT: vmovaps 272(%rdi), %xmm13
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm2[0,1],ymm13[2,3],ymm2[4,5,6,7]
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm5 = ymm5[0,1,0,2,4,5,4,6]
-; AVX2-SLOW-NEXT: vmovaps %ymm6, %ymm1
-; AVX2-SLOW-NEXT: vpermps %ymm0, %ymm6, %ymm6
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1],ymm5[2,3,4,5,6,7]
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm5 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm5 = mem[0,1],ymm0[2,3],mem[4,5,6,7]
-; AVX2-SLOW-NEXT: vpermps %ymm5, %ymm10, %ymm8
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm6[0,1,2,3,4],ymm8[5,6,7]
+; AVX2-SLOW-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm0 = ymm0[0,1,2,3],mem[4,5,6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-SLOW-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm13 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm13 = ymm1[0,1,2,3],mem[4,5,6,7]
+; AVX2-SLOW-NEXT: vmovaps 272(%rdi), %xmm6
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm13[0,1],ymm6[2,3],ymm13[4,5,6,7]
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm4 = ymm4[0,1,0,2,4,5,4,6]
+; AVX2-SLOW-NEXT: vmovaps %ymm9, %ymm1
+; AVX2-SLOW-NEXT: vpermps %ymm0, %ymm9, %ymm8
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm8[0,1],ymm4[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm10 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm10 = mem[0,1],ymm10[2,3],mem[4,5,6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm10, %ymm2, %ymm8
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm4[0,1,2,3,4],ymm8[5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm0, (%rsp) # 32-byte Spill
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm8 = ymm12[0,1,2,3],ymm3[4,5],ymm12[6,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm12 = ymm9[0,1,2,3],ymm11[4,5,6,7]
-; AVX2-SLOW-NEXT: vmovaps 656(%rdi), %xmm6
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm9 = ymm12[0,1],ymm6[2,3],ymm12[4,5,6,7]
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm9 = ymm9[0,1,0,2,4,5,4,6]
-; AVX2-SLOW-NEXT: vpermps %ymm8, %ymm1, %ymm11
+; AVX2-SLOW-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm9 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm9 = ymm3[0,1,2,3],mem[4,5,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm12 = ymm11[0,1,2,3],ymm12[4,5,6,7]
+; AVX2-SLOW-NEXT: vmovaps 656(%rdi), %xmm4
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm8 = ymm12[0,1],ymm4[2,3],ymm12[4,5,6,7]
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm8 = ymm8[0,1,0,2,4,5,4,6]
+; AVX2-SLOW-NEXT: vpermps %ymm9, %ymm1, %ymm11
; AVX2-SLOW-NEXT: vmovaps %ymm1, %ymm3
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm9 = ymm11[0,1],ymm9[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm8 = ymm11[0,1],ymm8[2,3,4,5,6,7]
; AVX2-SLOW-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: # ymm15 = mem[0,1],ymm15[2,3],mem[4,5,6,7]
-; AVX2-SLOW-NEXT: vpermps %ymm15, %ymm10, %ymm11
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm9[0,1,2,3,4],ymm11[5,6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm15, %ymm2, %ymm11
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm8[0,1,2,3,4],ymm11[5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm9 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm9 = ymm0[0,1,2,3],mem[4,5,6,7]
+; AVX2-SLOW-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm8 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm8 = ymm0[0,1,2,3],mem[4,5,6,7]
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm1 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm1 = ymm0[0,1,2,3],mem[4,5],ymm0[6,7]
+; AVX2-SLOW-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm1 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm1 = ymm0[0,1,2,3],mem[4,5,6,7]
; AVX2-SLOW-NEXT: vmovaps 80(%rdi), %xmm0
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm11 = ymm9[0,1],ymm0[2,3],ymm9[4,5,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm11 = ymm8[0,1],ymm0[2,3],ymm8[4,5,6,7]
; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm11 = ymm11[0,1,0,2,4,5,4,6]
-; AVX2-SLOW-NEXT: vpermps %ymm1, %ymm3, %ymm4
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm11[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm1, %ymm3, %ymm5
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1],ymm11[2,3,4,5,6,7]
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX2-SLOW-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm11 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: # ymm11 = mem[0,1],ymm3[2,3],mem[4,5,6,7]
-; AVX2-SLOW-NEXT: vpermps %ymm11, %ymm10, %ymm10
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm10 = ymm4[0,1,2,3,4],ymm10[5,6,7]
-; AVX2-SLOW-NEXT: vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm3 = mem[1,1,1,1,5,5,5,5]
-; AVX2-SLOW-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm3 = ymm3[0,1,2],mem[3],ymm3[4,5,6,7]
-; AVX2-SLOW-NEXT: vmovaps {{.*#+}} xmm4 = <5,3,u,u>
-; AVX2-SLOW-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm14 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm14[0,1],ymm3[2,3,4,5,6,7]
-; AVX2-SLOW-NEXT: vbroadcastf128 {{.*#+}} ymm14 = [0,3,1,7,0,3,1,7]
-; AVX2-SLOW-NEXT: # ymm14 = mem[0,1,0,1]
-; AVX2-SLOW-NEXT: vpermps %ymm7, %ymm14, %ymm7
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4],ymm7[5,6,7]
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm2 = ymm2[1,1,1,1,5,5,5,5]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm13[3],ymm2[4,5,6,7]
-; AVX2-SLOW-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm7 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm7[0,1],ymm2[2,3,4,5,6,7]
-; AVX2-SLOW-NEXT: vpermps %ymm5, %ymm14, %ymm5
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm5[5,6,7]
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm5 = ymm12[1,1,1,1,5,5,5,5]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2],ymm6[3],ymm5[4,5,6,7]
-; AVX2-SLOW-NEXT: vpermps %ymm8, %ymm4, %ymm6
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm6[0,1],ymm5[2,3,4,5,6,7]
-; AVX2-SLOW-NEXT: vpermps %ymm15, %ymm14, %ymm6
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4],ymm6[5,6,7]
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm6 = ymm9[1,1,1,1,5,5,5,5]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm6[0,1,2],ymm0[3],ymm6[4,5,6,7]
-; AVX2-SLOW-NEXT: vpermps %ymm1, %ymm4, %ymm1
+; AVX2-SLOW-NEXT: vpermps %ymm11, %ymm2, %ymm2
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm2 = ymm14[1,1,1,1,5,5,5,5]
+; AVX2-SLOW-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm2 = ymm2[0,1,2],mem[3],ymm2[4,5,6,7]
+; AVX2-SLOW-NEXT: vmovaps {{.*#+}} xmm14 = <5,3,u,u>
+; AVX2-SLOW-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm3 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT: vbroadcastf128 {{.*#+}} ymm3 = [0,3,1,7,0,3,1,7]
+; AVX2-SLOW-NEXT: # ymm3 = mem[0,1,0,1]
+; AVX2-SLOW-NEXT: vpermps %ymm7, %ymm3, %ymm7
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm7[5,6,7]
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm7 = ymm13[1,1,1,1,5,5,5,5]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm6 = ymm7[0,1,2],ymm6[3],ymm7[4,5,6,7]
+; AVX2-SLOW-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm7 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm6 = ymm7[0,1],ymm6[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm10, %ymm3, %ymm7
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4],ymm7[5,6,7]
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm7 = ymm12[1,1,1,1,5,5,5,5]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm7[0,1,2],ymm4[3],ymm7[4,5,6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm9, %ymm14, %ymm7
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm7[0,1],ymm4[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm15, %ymm3, %ymm7
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4],ymm7[5,6,7]
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm7 = ymm8[1,1,1,1,5,5,5,5]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm7[0,1,2],ymm0[3],ymm7[4,5,6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm1, %ymm14, %ymm1
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
-; AVX2-SLOW-NEXT: vpermps %ymm11, %ymm14, %ymm1
+; AVX2-SLOW-NEXT: vpermps %ymm11, %ymm3, %ymm1
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-SLOW-NEXT: vmovaps %ymm1, 96(%rsi)
@@ -3610,13 +3579,13 @@ define void @load_i32_stride6_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-SLOW-NEXT: vmovaps %ymm1, 32(%r9)
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-SLOW-NEXT: vmovaps %ymm1, 64(%r9)
-; AVX2-SLOW-NEXT: vmovaps %ymm10, (%r9)
+; AVX2-SLOW-NEXT: vmovaps %ymm5, (%r9)
; AVX2-SLOW-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; AVX2-SLOW-NEXT: vmovaps %ymm5, 96(%rax)
-; AVX2-SLOW-NEXT: vmovaps %ymm2, 32(%rax)
-; AVX2-SLOW-NEXT: vmovaps %ymm3, 64(%rax)
+; AVX2-SLOW-NEXT: vmovaps %ymm4, 96(%rax)
+; AVX2-SLOW-NEXT: vmovaps %ymm6, 32(%rax)
+; AVX2-SLOW-NEXT: vmovaps %ymm2, 64(%rax)
; AVX2-SLOW-NEXT: vmovaps %ymm0, (%rax)
-; AVX2-SLOW-NEXT: addq $1192, %rsp # imm = 0x4A8
+; AVX2-SLOW-NEXT: addq $1160, %rsp # imm = 0x488
; AVX2-SLOW-NEXT: vzeroupper
; AVX2-SLOW-NEXT: retq
;
@@ -3628,352 +3597,354 @@ define void @load_i32_stride6_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-NEXT: vmovups %ymm7, (%rsp) # 32-byte Spill
; AVX2-FAST-NEXT: vmovaps 416(%rdi), %ymm8
; AVX2-FAST-NEXT: vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps 128(%rdi), %ymm2
-; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovaps 128(%rdi), %ymm0
+; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovaps 160(%rdi), %ymm4
; AVX2-FAST-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovaps 96(%rdi), %ymm3
; AVX2-FAST-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps (%rdi), %ymm0
-; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps 32(%rdi), %ymm1
+; AVX2-FAST-NEXT: vmovaps (%rdi), %ymm1
; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovaps 32(%rdi), %ymm2
+; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovaps 64(%rdi), %ymm5
; AVX2-FAST-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps {{.*#+}} xmm13 = <0,6,4,u>
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm12 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
-; AVX2-FAST-NEXT: vpermps %ymm12, %ymm13, %ymm1
-; AVX2-FAST-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm5[0,1],ymm3[0,1]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm9 = ymm0[0,1,2,3,4,5],ymm3[6,7]
-; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm3 = ymm9[0,2,2,2,4,6,6,6]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm1[0,1,2],ymm3[3,4,5],ymm1[6,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm10 = ymm4[0,1,2,3],ymm2[4,5],ymm4[6,7]
+; AVX2-FAST-NEXT: vmovaps {{.*#+}} xmm14 = <0,6,4,u>
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
+; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vpermps %ymm1, %ymm14, %ymm2
+; AVX2-FAST-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm5[0,1],ymm3[0,1]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm11 = ymm1[0,1,2,3,4,5],ymm3[6,7]
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm3 = ymm11[0,2,2,2,4,6,6,6]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm2[0,1,2],ymm3[3,4,5,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm10 = ymm4[0,1,2,3],ymm0[4,5,6,7]
; AVX2-FAST-NEXT: vbroadcastsd {{.*#+}} ymm2 = [8589934596,8589934596,8589934596,8589934596]
; AVX2-FAST-NEXT: vpermps %ymm10, %ymm2, %ymm5
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm3[0,1,2,3,4,5],ymm5[6,7]
; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vperm2f128 {{.*#+}} ymm3 = ymm7[0,1],ymm6[0,1]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm7 = ymm3[0,1,2,3,4,5],ymm6[6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm12 = ymm3[0,1,2,3,4,5],ymm6[6,7]
; AVX2-FAST-NEXT: vmovaps 384(%rdi), %ymm0
; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm6 = ymm0[0,1,2,3],ymm8[4,5],ymm0[6,7]
-; AVX2-FAST-NEXT: vpermps %ymm6, %ymm13, %ymm5
-; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm8 = ymm7[0,2,2,2,4,6,6,6]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2],ymm8[3,4,5],ymm5[6,7]
+; AVX2-FAST-NEXT: vpermps %ymm6, %ymm14, %ymm5
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm7 = ymm12[0,2,2,2,4,6,6,6]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2],ymm7[3,4,5,6,7]
; AVX2-FAST-NEXT: vmovaps 512(%rdi), %ymm0
; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovaps 544(%rdi), %ymm1
; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
-; AVX2-FAST-NEXT: vpermps %ymm4, %ymm2, %ymm8
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm5[0,1,2,3,4,5],ymm8[6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-FAST-NEXT: vpermps %ymm4, %ymm2, %ymm9
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm5[0,1,2,3,4,5],ymm9[6,7]
; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovaps 288(%rdi), %ymm1
; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovaps 256(%rdi), %ymm0
; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vperm2f128 {{.*#+}} ymm8 = ymm0[0,1],ymm1[0,1]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm11 = ymm8[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-FAST-NEXT: vperm2f128 {{.*#+}} ymm5 = ymm0[0,1],ymm1[0,1]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm9 = ymm5[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FAST-NEXT: vmovaps 224(%rdi), %ymm0
; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovaps 192(%rdi), %ymm1
; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
-; AVX2-FAST-NEXT: vpermps %ymm3, %ymm13, %ymm1
-; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm15 = ymm11[0,2,2,2,4,6,6,6]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm15[3,4,5],ymm1[6,7]
+; AVX2-FAST-NEXT: vpermps %ymm3, %ymm14, %ymm15
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm13 = ymm9[0,2,2,2,4,6,6,6]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm13 = ymm15[0,1,2],ymm13[3,4,5,6,7]
; AVX2-FAST-NEXT: vmovaps 320(%rdi), %ymm0
; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps 352(%rdi), %ymm5
-; AVX2-FAST-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm15 = ymm5[0,1,2,3],ymm0[4,5],ymm5[6,7]
-; AVX2-FAST-NEXT: vpermps %ymm15, %ymm2, %ymm8
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm8[6,7]
+; AVX2-FAST-NEXT: vmovaps 352(%rdi), %ymm1
; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm8 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-FAST-NEXT: vpermps %ymm8, %ymm2, %ymm15
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm5 = ymm13[0,1,2,3,4,5],ymm15[6,7]
+; AVX2-FAST-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovaps 608(%rdi), %ymm0
; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovaps 576(%rdi), %ymm1
; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm5 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
-; AVX2-FAST-NEXT: vpermps %ymm5, %ymm13, %ymm8
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm13 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
+; AVX2-FAST-NEXT: vpermps %ymm13, %ymm14, %ymm14
; AVX2-FAST-NEXT: vmovaps 672(%rdi), %ymm1
; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps 640(%rdi), %ymm0
-; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vperm2f128 {{.*#+}} ymm13 = ymm0[0,1],ymm1[0,1]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3,4,5],ymm1[6,7]
-; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm14 = ymm13[0,2,2,2,4,6,6,6]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2],ymm14[3,4,5],ymm8[6,7]
+; AVX2-FAST-NEXT: vmovaps 640(%rdi), %ymm15
+; AVX2-FAST-NEXT: vperm2f128 {{.*#+}} ymm7 = ymm15[0,1],ymm1[0,1]
+; AVX2-FAST-NEXT: vmovups %ymm15, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm5 = ymm7[0,2,2,2,4,6,6,6]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm5 = ymm14[0,1,2],ymm5[3,4,5,6,7]
; AVX2-FAST-NEXT: vmovaps 704(%rdi), %ymm1
; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovaps 736(%rdi), %ymm0
; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm14 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm14 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FAST-NEXT: vpermps %ymm14, %ymm2, %ymm0
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm8[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm5[0,1,2,3,4,5],ymm0[6,7]
; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovaps {{.*#+}} xmm0 = <1,7,5,u>
-; AVX2-FAST-NEXT: vpermps %ymm12, %ymm0, %ymm8
-; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm9 = ymm9[1,3,2,3,5,7,6,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2],ymm9[3,4,5],ymm8[6,7]
-; AVX2-FAST-NEXT: vbroadcastsd {{.*#+}} ymm9 = [12884901893,12884901893,12884901893,12884901893]
-; AVX2-FAST-NEXT: vpermps %ymm10, %ymm9, %ymm10
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5],ymm10[6,7]
-; AVX2-FAST-NEXT: vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpermps %ymm6, %ymm0, %ymm6
-; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm7 = ymm7[1,3,2,3,5,7,6,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2],ymm7[3,4,5],ymm6[6,7]
-; AVX2-FAST-NEXT: vpermps %ymm4, %ymm9, %ymm4
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm6[0,1,2,3,4,5],ymm4[6,7]
+; AVX2-FAST-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm5 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm11 = ymm11[1,3,2,3,5,7,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2],ymm11[3,4,5,6,7]
+; AVX2-FAST-NEXT: vbroadcastsd {{.*#+}} ymm11 = [12884901893,12884901893,12884901893,12884901893]
+; AVX2-FAST-NEXT: vpermps %ymm10, %ymm11, %ymm10
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5],ymm10[6,7]
+; AVX2-FAST-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vpermps %ymm6, %ymm0, %ymm5
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm6 = ymm12[1,3,2,3,5,7,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2],ymm6[3,4,5,6,7]
+; AVX2-FAST-NEXT: vpermps %ymm4, %ymm11, %ymm4
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm5[0,1,2,3,4,5],ymm4[6,7]
; AVX2-FAST-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpermps %ymm3, %ymm0, %ymm2
-; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm3 = ymm11[1,3,2,3,5,7,6,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm3[3,4,5],ymm2[6,7]
-; AVX2-FAST-NEXT: vpermps %ymm15, %ymm9, %ymm3
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm3[6,7]
-; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpermps %ymm5, %ymm0, %ymm0
-; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm1 = ymm13[1,3,2,3,5,7,6,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5],ymm0[6,7]
-; AVX2-FAST-NEXT: vpermps %ymm14, %ymm9, %ymm1
+; AVX2-FAST-NEXT: vpermps %ymm3, %ymm0, %ymm1
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm3 = ymm9[1,3,2,3,5,7,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm3[3,4,5,6,7]
+; AVX2-FAST-NEXT: vpermps %ymm8, %ymm11, %ymm2
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6,7]
+; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vpermps %ymm13, %ymm0, %ymm0
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm1 = ymm7[1,3,2,3,5,7,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
+; AVX2-FAST-NEXT: vpermps %ymm14, %ymm11, %ymm1
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm0 = ymm5[2,3,2,3,6,7,6,7]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
-; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm1 = ymm11[0,0,2,3,4,4,6,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
+; AVX2-FAST-NEXT: vmovaps (%rdi), %xmm2
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} xmm0 = xmm2[2,3,2,3]
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm9[0,1],ymm10[2,3],ymm9[4,5],ymm10[6,7]
-; AVX2-FAST-NEXT: vmovaps {{.*#+}} ymm2 = [2,0,6,4,2,0,6,7]
-; AVX2-FAST-NEXT: vpermps %ymm1, %ymm2, %ymm1
-; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
-; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm1 = ymm7[0,0,0,0,4,4,4,4]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm3 = ymm8[0,0,2,3,4,4,6,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm3[0,1,2],ymm1[3],ymm3[4,5,6],ymm1[7]
-; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
-; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
-; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm0 = ymm14[2,3,2,3,6,7,6,7]
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm1 = ymm9[0,0,2,3,4,4,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
-; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm1 = ymm12[0,0,2,3,4,4,6,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-FAST-NEXT: vblendps $51, (%rsp), %ymm1, %ymm1 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm1 = mem[0,1],ymm1[2,3],mem[4,5],ymm1[6,7]
-; AVX2-FAST-NEXT: vpermps %ymm1, %ymm2, %ymm1
-; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
-; AVX2-FAST-NEXT: vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm1 = mem[0,0,0,0,4,4,4,4]
-; AVX2-FAST-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm3 = mem[0,0,2,3,4,4,6,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm3[0,1,2],ymm1[3],ymm3[4,5,6],ymm1[7]
-; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
-; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm0 = mem[2,3,2,3,6,7,6,7]
-; AVX2-FAST-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm1 = mem[0,0,2,3,4,4,6,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-FAST-NEXT: vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm1 = mem[0,1],ymm1[2,3],mem[4,5],ymm1[6,7]
-; AVX2-FAST-NEXT: vpermps %ymm1, %ymm2, %ymm1
-; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
-; AVX2-FAST-NEXT: vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm1 = mem[0,0,0,0,4,4,4,4]
-; AVX2-FAST-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm3 = mem[0,0,2,3,4,4,6,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm3[0,1,2],ymm1[3],ymm3[4,5,6],ymm1[7]
-; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
-; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm6[0,1],ymm13[2,3],ymm6[4,5],ymm13[6,7]
-; AVX2-FAST-NEXT: vpermps %ymm0, %ymm2, %ymm0
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm1 = ymm3[2,3,2,3,6,7,6,7]
-; AVX2-FAST-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm2 = mem[0,0,2,3,4,4,6,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm2[1,2,3],ymm1[4],ymm2[5,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm11[0,1],ymm12[2,3],ymm11[4,5],ymm12[6,7]
+; AVX2-FAST-NEXT: vmovaps {{.*#+}} ymm7 = [2,0,6,4,2,0,6,7]
+; AVX2-FAST-NEXT: vpermps %ymm3, %ymm7, %ymm3
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,3,2,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4],ymm1[5,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm3[3,4,5,6,7]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm3 = ymm8[0,0,0,0,4,4,4,4]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm4 = ymm10[0,0,2,3,4,4,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3],ymm4[4,5,6],ymm3[7]
+; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm3 = ymm3[0,1,0,3]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm3[5,6,7]
+; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovaps 384(%rdi), %xmm1
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} xmm3 = xmm1[2,3,2,3]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm4 = ymm0[0,0,2,3,4,4,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0],ymm4[1,2,3],ymm3[4],ymm4[5,6,7]
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm1 = ymm4[0,0,0,0,4,4,4,4]
+; AVX2-FAST-NEXT: vblendps $51, (%rsp), %ymm4, %ymm4 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm4 = mem[0,1],ymm4[2,3],mem[4,5],ymm4[6,7]
+; AVX2-FAST-NEXT: vpermps %ymm4, %ymm7, %ymm4
+; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm3 = ymm3[0,3,2,3]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2],ymm4[3,4,5,6,7]
+; AVX2-FAST-NEXT: vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm4 = mem[0,0,0,0,4,4,4,4]
+; AVX2-FAST-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm5 = mem[0,0,2,3,4,4,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm5[0,1,2],ymm4[3],ymm5[4,5,6],ymm4[7]
+; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm4 = ymm4[0,1,0,3]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4],ymm4[5,6,7]
+; AVX2-FAST-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovaps 576(%rdi), %xmm3
+; AVX2-FAST-NEXT: vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} xmm4 = xmm3[2,3,2,3]
+; AVX2-FAST-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm5 = mem[0,0,2,3,4,4,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0],ymm5[1,2,3],ymm4[4],ymm5[5,6,7]
+; AVX2-FAST-NEXT: vblendps $204, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm5 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm5 = ymm15[0,1],mem[2,3],ymm15[4,5],mem[6,7]
+; AVX2-FAST-NEXT: vpermps %ymm5, %ymm7, %ymm5
+; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm4 = ymm4[0,3,2,3]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2],ymm5[3,4,5,6,7]
+; AVX2-FAST-NEXT: vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm5 = mem[0,0,0,0,4,4,4,4]
+; AVX2-FAST-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm6 = mem[0,0,2,3,4,4,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm5 = ymm6[0,1,2],ymm5[3],ymm6[4,5,6],ymm5[7]
+; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm5 = ymm5[0,1,0,3]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm4[0,1,2,3,4],ymm5[5,6,7]
+; AVX2-FAST-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
-; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm2 = ymm15[0,0,2,3,4,4,6,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
-; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
-; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm0 = ymm5[3,3,3,3,7,7,7,7]
-; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm1 = ymm11[0,1,3,3,4,5,7,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
-; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm1 = ymm10[3,3,3,3,7,7,7,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm9[1],ymm1[2,3,4],ymm9[5],ymm1[6,7]
-; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
-; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
-; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm1 = ymm7[0,1,0,1,4,5,4,5]
-; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm2 = ymm8[0,1,3,3,4,5,7,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
-; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
-; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm0 = ymm14[3,3,3,3,7,7,7,7]
-; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm1 = ymm12[0,1,3,3,4,5,7,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm14[0,1],ymm15[2,3],ymm14[4,5],ymm15[6,7]
+; AVX2-FAST-NEXT: vpermps %ymm4, %ymm7, %ymm4
+; AVX2-FAST-NEXT: vmovaps 192(%rdi), %xmm13
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} xmm5 = xmm13[2,3,2,3]
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
-; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm1 = ymm7[3,3,3,3,7,7,7,7]
-; AVX2-FAST-NEXT: vmovups (%rsp), %ymm11 # 32-byte Reload
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm11[1],ymm1[2,3,4],ymm11[5],ymm1[6,7]
-; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
-; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
-; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm1 = ymm10[0,1,0,1,4,5,4,5]
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm6 = ymm7[0,0,2,3,4,4,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0],ymm6[1,2,3],ymm5[4],ymm6[5,6,7]
+; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm5 = ymm5[0,3,2,3]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm5[0,1,2],ymm4[3,4,5,6,7]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm5 = ymm3[0,0,0,0,4,4,4,4]
+; AVX2-FAST-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm6 = mem[0,0,2,3,4,4,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm5 = ymm6[0,1,2],ymm5[3],ymm6[4,5,6],ymm5[7]
+; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm5 = ymm5[0,1,0,3]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4],ymm5[5,6,7]
+; AVX2-FAST-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} xmm2 = xmm2[3,3,3,3]
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm4 = ymm9[0,1,3,3,4,5,7,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0],ymm4[1,2,3],ymm2[4],ymm4[5,6,7]
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm4 = ymm12[3,3,3,3,7,7,7,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0],ymm11[1],ymm4[2,3,4],ymm11[5],ymm4[6,7]
+; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm4 = ymm4[0,2,0,3]
+; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,3,2,3]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm4[3,4,5,6,7]
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm4 = ymm8[0,1,0,1,4,5,4,5]
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm5 = ymm10[0,1,3,3,4,5,7,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm5[0,1,2],ymm4[3],ymm5[4,5,6],ymm4[7]
+; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm4 = ymm4[0,1,0,3]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm4[5,6,7]
+; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} xmm1 = xmm1[3,3,3,3]
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm2 = ymm0[0,1,3,3,4,5,7,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm2[1,2,3],ymm1[4],ymm2[5,6,7]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm2 = ymm9[3,3,3,3,7,7,7,7]
+; AVX2-FAST-NEXT: vmovups (%rsp), %ymm5 # 32-byte Reload
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0],ymm5[1],ymm2[2,3,4],ymm5[5],ymm2[6,7]
+; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,2,0,3]
+; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,3,2,3]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm2 = ymm8[0,1,3,3,4,5,7,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
-; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm2 = ymm8[0,1,0,1,4,5,4,5]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm4 = ymm6[0,1,3,3,4,5,7,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm4[0,1,2],ymm2[3],ymm4[4,5,6],ymm2[7]
+; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,1,0,3]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm2[5,6,7]
; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm0 = ymm3[3,3,3,3,7,7,7,7]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm1 = ymm5[0,1,3,3,4,5,7,7]
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} xmm0 = xmm13[3,3,3,3]
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm1 = ymm7[0,1,3,3,4,5,7,7]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
-; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm1 = ymm13[3,3,3,3,7,7,7,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm6[1],ymm1[2,3,4],ymm6[5],ymm1[6,7]
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm1 = ymm15[3,3,3,3,7,7,7,7]
+; AVX2-FAST-NEXT: vmovaps %ymm15, %ymm13
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm14[1],ymm1[2,3,4],ymm14[5],ymm1[6,7]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
-; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm1 = ymm4[0,1,0,1,4,5,4,5]
-; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm2 = ymm15[0,1,3,3,4,5,7,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm1 = ymm3[0,1,0,1,4,5,4,5]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm2 = ymm10[0,1,3,3,4,5,7,7]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; AVX2-FAST-NEXT: # xmm0 = mem[3,3,3,3]
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm0 = ymm3[3,3,3,3,7,7,7,7]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
-; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm1 = ymm15[0,1,3,3,4,5,7,7]
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm1 = ymm3[0,1,3,3,4,5,7,7]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm1 = ymm12[3,3,3,3,7,7,7,7]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm9[1],ymm1[2,3,4],ymm9[5],ymm1[6,7]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm11[1],ymm1[2,3,4],ymm11[5],ymm1[6,7]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
; AVX2-FAST-NEXT: vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
; AVX2-FAST-NEXT: # ymm1 = mem[0,1,0,1,4,5,4,5]
-; AVX2-FAST-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm2 = mem[0,1,3,3,4,5,7,7]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm2 = ymm15[0,1,3,3,4,5,7,7]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vblendps $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm0 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm0 = mem[0,1,2,3],ymm14[4,5],mem[6,7]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm0 = ymm0[0,1,2,3],mem[4,5,6,7]
; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm14 = ymm11[0,1,2,3],ymm7[4,5,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm14 = ymm5[0,1,2,3],ymm9[4,5,6,7]
; AVX2-FAST-NEXT: vmovaps 464(%rdi), %xmm1
; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm14[0,1],ymm1[2,3],ymm14[4,5,6,7]
-; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
-; AVX2-FAST-NEXT: vmovaps {{.*#+}} xmm2 = <4,2,u,u>
-; AVX2-FAST-NEXT: vpermps %ymm0, %ymm2, %ymm4
-; AVX2-FAST-NEXT: vmovaps %ymm2, %ymm11
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm4[0,1],ymm1[2,3,4,5,6,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm7 = ymm10[0,1],ymm8[2,3],ymm10[4,5,6,7]
-; AVX2-FAST-NEXT: vbroadcastf128 {{.*#+}} ymm10 = [0,2,0,6,0,2,0,6]
-; AVX2-FAST-NEXT: # ymm10 = mem[0,1,0,1]
-; AVX2-FAST-NEXT: vpermps %ymm7, %ymm10, %ymm4
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm4[5,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm14[0,1],ymm1[2,3],ymm14[4,5,6,7]
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm2 = ymm2[0,1,0,2,4,5,4,6]
+; AVX2-FAST-NEXT: vmovaps {{.*#+}} xmm1 = <4,2,u,u>
+; AVX2-FAST-NEXT: vpermps %ymm0, %ymm1, %ymm4
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm2[2,3,4,5,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm7 = ymm8[0,1],ymm6[2,3],ymm8[4,5,6,7]
+; AVX2-FAST-NEXT: vbroadcastf128 {{.*#+}} ymm2 = [0,2,0,6,0,2,0,6]
+; AVX2-FAST-NEXT: # ymm2 = mem[0,1,0,1]
+; AVX2-FAST-NEXT: vpermps %ymm7, %ymm2, %ymm5
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm4[0,1,2,3,4],ymm5[5,6,7]
; AVX2-FAST-NEXT: vmovups %ymm0, (%rsp) # 32-byte Spill
-; AVX2-FAST-NEXT: vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm0 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm0 = ymm5[0,1,2,3],mem[4,5],ymm5[6,7]
-; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm6[0,1,2,3],ymm13[4,5,6,7]
-; AVX2-FAST-NEXT: vmovaps 272(%rdi), %xmm13
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm5 = ymm2[0,1],ymm13[2,3],ymm2[4,5,6,7]
-; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm5 = ymm5[0,1,0,2,4,5,4,6]
-; AVX2-FAST-NEXT: vmovaps %ymm11, %ymm1
-; AVX2-FAST-NEXT: vpermps %ymm0, %ymm11, %ymm6
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1],ymm5[2,3,4,5,6,7]
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-FAST-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm5 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm5 = mem[0,1],ymm0[2,3],mem[4,5,6,7]
-; AVX2-FAST-NEXT: vpermps %ymm5, %ymm10, %ymm8
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm6[0,1,2,3,4],ymm8[5,6,7]
+; AVX2-FAST-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm0 = ymm0[0,1,2,3],mem[4,5,6,7]
+; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm13 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm13 = mem[0,1,2,3],ymm13[4,5,6,7]
+; AVX2-FAST-NEXT: vmovaps 272(%rdi), %xmm6
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm13[0,1],ymm6[2,3],ymm13[4,5,6,7]
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm4 = ymm4[0,1,0,2,4,5,4,6]
+; AVX2-FAST-NEXT: vpermps %ymm0, %ymm1, %ymm8
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm8[0,1],ymm4[2,3,4,5,6,7]
+; AVX2-FAST-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm10 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm10 = mem[0,1],ymm10[2,3],mem[4,5,6,7]
+; AVX2-FAST-NEXT: vpermps %ymm10, %ymm2, %ymm8
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm4[0,1,2,3,4],ymm8[5,6,7]
; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm8 = ymm15[0,1,2,3],ymm3[4,5],ymm15[6,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm12 = ymm9[0,1,2,3],ymm12[4,5,6,7]
-; AVX2-FAST-NEXT: vmovaps 656(%rdi), %xmm6
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm9 = ymm12[0,1],ymm6[2,3],ymm12[4,5,6,7]
-; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm9 = ymm9[0,1,0,2,4,5,4,6]
-; AVX2-FAST-NEXT: vpermps %ymm8, %ymm11, %ymm11
+; AVX2-FAST-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm9 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm9 = ymm3[0,1,2,3],mem[4,5,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm12 = ymm11[0,1,2,3],ymm12[4,5,6,7]
+; AVX2-FAST-NEXT: vmovaps 656(%rdi), %xmm4
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm8 = ymm12[0,1],ymm4[2,3],ymm12[4,5,6,7]
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm8 = ymm8[0,1,0,2,4,5,4,6]
+; AVX2-FAST-NEXT: vpermps %ymm9, %ymm1, %ymm11
; AVX2-FAST-NEXT: vmovaps %ymm1, %ymm3
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm9 = ymm11[0,1],ymm9[2,3,4,5,6,7]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-FAST-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm15 = mem[0,1],ymm0[2,3],mem[4,5,6,7]
-; AVX2-FAST-NEXT: vpermps %ymm15, %ymm10, %ymm11
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm9[0,1,2,3,4],ymm11[5,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm8 = ymm11[0,1],ymm8[2,3,4,5,6,7]
+; AVX2-FAST-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm15 = mem[0,1],ymm15[2,3],mem[4,5,6,7]
+; AVX2-FAST-NEXT: vpermps %ymm15, %ymm2, %ymm11
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm8[0,1,2,3,4],ymm11[5,6,7]
; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-FAST-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm9 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm9 = ymm0[0,1,2,3],mem[4,5,6,7]
+; AVX2-FAST-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm8 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm8 = ymm0[0,1,2,3],mem[4,5,6,7]
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-FAST-NEXT: vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm1 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm1 = ymm0[0,1,2,3],mem[4,5],ymm0[6,7]
+; AVX2-FAST-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm1 = ymm0[0,1,2,3],mem[4,5,6,7]
; AVX2-FAST-NEXT: vmovaps 80(%rdi), %xmm0
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm11 = ymm9[0,1],ymm0[2,3],ymm9[4,5,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm11 = ymm8[0,1],ymm0[2,3],ymm8[4,5,6,7]
; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm11 = ymm11[0,1,0,2,4,5,4,6]
-; AVX2-FAST-NEXT: vpermps %ymm1, %ymm3, %ymm4
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm11[2,3,4,5,6,7]
+; AVX2-FAST-NEXT: vpermps %ymm1, %ymm3, %ymm5
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1],ymm11[2,3,4,5,6,7]
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX2-FAST-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm11 # 32-byte Folded Reload
; AVX2-FAST-NEXT: # ymm11 = mem[0,1],ymm3[2,3],mem[4,5,6,7]
-; AVX2-FAST-NEXT: vpermps %ymm11, %ymm10, %ymm10
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm10 = ymm4[0,1,2,3,4],ymm10[5,6,7]
-; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm3 = ymm14[1,1,1,1,5,5,5,5]
-; AVX2-FAST-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm3 = ymm3[0,1,2],mem[3],ymm3[4,5,6,7]
-; AVX2-FAST-NEXT: vmovaps {{.*#+}} xmm4 = <5,3,u,u>
-; AVX2-FAST-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm14 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm14[0,1],ymm3[2,3,4,5,6,7]
-; AVX2-FAST-NEXT: vbroadcastf128 {{.*#+}} ymm14 = [0,3,1,7,0,3,1,7]
-; AVX2-FAST-NEXT: # ymm14 = mem[0,1,0,1]
-; AVX2-FAST-NEXT: vpermps %ymm7, %ymm14, %ymm7
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4],ymm7[5,6,7]
-; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm2 = ymm2[1,1,1,1,5,5,5,5]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm13[3],ymm2[4,5,6,7]
-; AVX2-FAST-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm7 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm7[0,1],ymm2[2,3,4,5,6,7]
-; AVX2-FAST-NEXT: vpermps %ymm5, %ymm14, %ymm5
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm5[5,6,7]
-; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm5 = ymm12[1,1,1,1,5,5,5,5]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2],ymm6[3],ymm5[4,5,6,7]
-; AVX2-FAST-NEXT: vpermps %ymm8, %ymm4, %ymm6
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm5 = ymm6[0,1],ymm5[2,3,4,5,6,7]
-; AVX2-FAST-NEXT: vpermps %ymm15, %ymm14, %ymm6
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4],ymm6[5,6,7]
-; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm6 = ymm9[1,1,1,1,5,5,5,5]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm6[0,1,2],ymm0[3],ymm6[4,5,6,7]
-; AVX2-FAST-NEXT: vpermps %ymm1, %ymm4, %ymm1
+; AVX2-FAST-NEXT: vpermps %ymm11, %ymm2, %ymm2
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm2 = ymm14[1,1,1,1,5,5,5,5]
+; AVX2-FAST-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm2 = ymm2[0,1,2],mem[3],ymm2[4,5,6,7]
+; AVX2-FAST-NEXT: vmovaps {{.*#+}} xmm14 = <5,3,u,u>
+; AVX2-FAST-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm3 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3,4,5,6,7]
+; AVX2-FAST-NEXT: vbroadcastf128 {{.*#+}} ymm3 = [0,3,1,7,0,3,1,7]
+; AVX2-FAST-NEXT: # ymm3 = mem[0,1,0,1]
+; AVX2-FAST-NEXT: vpermps %ymm7, %ymm3, %ymm7
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm7[5,6,7]
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm7 = ymm13[1,1,1,1,5,5,5,5]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm6 = ymm7[0,1,2],ymm6[3],ymm7[4,5,6,7]
+; AVX2-FAST-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm7 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm6 = ymm7[0,1],ymm6[2,3,4,5,6,7]
+; AVX2-FAST-NEXT: vpermps %ymm10, %ymm3, %ymm7
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4],ymm7[5,6,7]
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm7 = ymm12[1,1,1,1,5,5,5,5]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm7[0,1,2],ymm4[3],ymm7[4,5,6,7]
+; AVX2-FAST-NEXT: vpermps %ymm9, %ymm14, %ymm7
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm7[0,1],ymm4[2,3,4,5,6,7]
+; AVX2-FAST-NEXT: vpermps %ymm15, %ymm3, %ymm7
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4],ymm7[5,6,7]
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm7 = ymm8[1,1,1,1,5,5,5,5]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm7[0,1,2],ymm0[3],ymm7[4,5,6,7]
+; AVX2-FAST-NEXT: vpermps %ymm1, %ymm14, %ymm1
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
-; AVX2-FAST-NEXT: vpermps %ymm11, %ymm14, %ymm1
+; AVX2-FAST-NEXT: vpermps %ymm11, %ymm3, %ymm1
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FAST-NEXT: vmovaps %ymm1, 96(%rsi)
@@ -4013,11 +3984,11 @@ define void @load_i32_stride6_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-NEXT: vmovaps %ymm1, 32(%r9)
; AVX2-FAST-NEXT: vmovups (%rsp), %ymm1 # 32-byte Reload
; AVX2-FAST-NEXT: vmovaps %ymm1, 64(%r9)
-; AVX2-FAST-NEXT: vmovaps %ymm10, (%r9)
+; AVX2-FAST-NEXT: vmovaps %ymm5, (%r9)
; AVX2-FAST-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; AVX2-FAST-NEXT: vmovaps %ymm5, 96(%rax)
-; AVX2-FAST-NEXT: vmovaps %ymm2, 32(%rax)
-; AVX2-FAST-NEXT: vmovaps %ymm3, 64(%rax)
+; AVX2-FAST-NEXT: vmovaps %ymm4, 96(%rax)
+; AVX2-FAST-NEXT: vmovaps %ymm6, 32(%rax)
+; AVX2-FAST-NEXT: vmovaps %ymm2, 64(%rax)
; AVX2-FAST-NEXT: vmovaps %ymm0, (%rax)
; AVX2-FAST-NEXT: addq $1160, %rsp # imm = 0x488
; AVX2-FAST-NEXT: vzeroupper
@@ -4025,365 +3996,365 @@ define void @load_i32_stride6_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
;
; AVX2-FAST-PERLANE-LABEL: load_i32_stride6_vf32:
; AVX2-FAST-PERLANE: # %bb.0:
-; AVX2-FAST-PERLANE-NEXT: subq $1192, %rsp # imm = 0x4A8
+; AVX2-FAST-PERLANE-NEXT: subq $1160, %rsp # imm = 0x488
; AVX2-FAST-PERLANE-NEXT: vmovaps 480(%rdi), %ymm6
; AVX2-FAST-PERLANE-NEXT: vmovaps 448(%rdi), %ymm7
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovaps 416(%rdi), %ymm8
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovaps 128(%rdi), %ymm2
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovaps 128(%rdi), %ymm0
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovaps 160(%rdi), %ymm4
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovaps 96(%rdi), %ymm3
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovaps (%rdi), %ymm0
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovaps 32(%rdi), %ymm1
+; AVX2-FAST-PERLANE-NEXT: vmovaps (%rdi), %ymm1
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovaps 32(%rdi), %ymm2
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovaps 64(%rdi), %ymm5
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovaps {{.*#+}} xmm13 = <0,6,4,u>
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm12 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm12, %ymm13, %ymm1
-; AVX2-FAST-PERLANE-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm5[0,1],ymm3[0,1]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm9 = ymm0[0,1,2,3,4,5],ymm3[6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm3 = ymm9[0,2,2,2,4,6,6,6]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm1[0,1,2],ymm3[3,4,5],ymm1[6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm10 = ymm4[0,1,2,3],ymm2[4,5],ymm4[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovaps {{.*#+}} xmm14 = <0,6,4,u>
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm1, %ymm14, %ymm2
+; AVX2-FAST-PERLANE-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm5[0,1],ymm3[0,1]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm11 = ymm1[0,1,2,3,4,5],ymm3[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm3 = ymm11[0,2,2,2,4,6,6,6]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm2[0,1,2],ymm3[3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm10 = ymm4[0,1,2,3],ymm0[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vbroadcastsd {{.*#+}} ymm2 = [8589934596,8589934596,8589934596,8589934596]
; AVX2-FAST-PERLANE-NEXT: vpermps %ymm10, %ymm2, %ymm5
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm3[0,1,2,3,4,5],ymm5[6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm6, (%rsp) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vperm2f128 {{.*#+}} ymm3 = ymm7[0,1],ymm6[0,1]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm7 = ymm3[0,1,2,3,4,5],ymm6[6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm12 = ymm3[0,1,2,3,4,5],ymm6[6,7]
; AVX2-FAST-PERLANE-NEXT: vmovaps 384(%rdi), %ymm0
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm6 = ymm0[0,1,2,3],ymm8[4,5],ymm0[6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm6, %ymm13, %ymm5
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm8 = ymm7[0,2,2,2,4,6,6,6]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2],ymm8[3,4,5],ymm5[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm6, %ymm14, %ymm5
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm7 = ymm12[0,2,2,2,4,6,6,6]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2],ymm7[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovaps 512(%rdi), %ymm0
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovaps 544(%rdi), %ymm1
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm4, %ymm2, %ymm8
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm5[0,1,2,3,4,5],ymm8[6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm4, %ymm2, %ymm9
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm5[0,1,2,3,4,5],ymm9[6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovaps 288(%rdi), %ymm1
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovaps 256(%rdi), %ymm0
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vperm2f128 {{.*#+}} ymm8 = ymm0[0,1],ymm1[0,1]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm11 = ymm8[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-FAST-PERLANE-NEXT: vperm2f128 {{.*#+}} ymm5 = ymm0[0,1],ymm1[0,1]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm9 = ymm5[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FAST-PERLANE-NEXT: vmovaps 224(%rdi), %ymm0
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovaps 192(%rdi), %ymm1
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm3, %ymm13, %ymm1
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm15 = ymm11[0,2,2,2,4,6,6,6]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm15[3,4,5],ymm1[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm3, %ymm14, %ymm15
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm13 = ymm9[0,2,2,2,4,6,6,6]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm13 = ymm15[0,1,2],ymm13[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovaps 320(%rdi), %ymm0
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovaps 352(%rdi), %ymm5
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm15 = ymm5[0,1,2,3],ymm0[4,5],ymm5[6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm15, %ymm2, %ymm8
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm8[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, (%rsp) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovaps 352(%rdi), %ymm1
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm8 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm8, %ymm2, %ymm15
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm13[0,1,2,3,4,5],ymm15[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovaps 608(%rdi), %ymm0
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovaps 576(%rdi), %ymm1
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm5, %ymm13, %ymm8
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm13 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm13, %ymm14, %ymm14
; AVX2-FAST-PERLANE-NEXT: vmovaps 672(%rdi), %ymm1
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovaps 640(%rdi), %ymm0
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vperm2f128 {{.*#+}} ymm13 = ymm0[0,1],ymm1[0,1]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3,4,5],ymm1[6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm14 = ymm13[0,2,2,2,4,6,6,6]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2],ymm14[3,4,5],ymm8[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovaps 640(%rdi), %ymm15
+; AVX2-FAST-PERLANE-NEXT: vperm2f128 {{.*#+}} ymm7 = ymm15[0,1],ymm1[0,1]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm15, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm5 = ymm7[0,2,2,2,4,6,6,6]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm14[0,1,2],ymm5[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovaps 704(%rdi), %ymm1
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovaps 736(%rdi), %ymm0
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm14 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm14 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpermps %ymm14, %ymm2, %ymm0
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm8[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm5[0,1,2,3,4,5],ymm0[6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovaps {{.*#+}} xmm0 = <1,7,5,u>
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm12, %ymm0, %ymm8
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm9 = ymm9[1,3,2,3,5,7,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2],ymm9[3,4,5],ymm8[6,7]
-; AVX2-FAST-PERLANE-NEXT: vbroadcastsd {{.*#+}} ymm9 = [12884901893,12884901893,12884901893,12884901893]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm10, %ymm9, %ymm10
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5],ymm10[6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm6, %ymm0, %ymm6
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm7 = ymm7[1,3,2,3,5,7,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2],ymm7[3,4,5],ymm6[6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm4, %ymm9, %ymm4
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm6[0,1,2,3,4,5],ymm4[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm5 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm11 = ymm11[1,3,2,3,5,7,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2],ymm11[3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vbroadcastsd {{.*#+}} ymm11 = [12884901893,12884901893,12884901893,12884901893]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm10, %ymm11, %ymm10
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5],ymm10[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm6, %ymm0, %ymm5
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm6 = ymm12[1,3,2,3,5,7,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2],ymm6[3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm4, %ymm11, %ymm4
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm5[0,1,2,3,4,5],ymm4[6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm3, %ymm0, %ymm2
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm3 = ymm11[1,3,2,3,5,7,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm3[3,4,5],ymm2[6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm15, %ymm9, %ymm3
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm3[6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm5, %ymm0, %ymm0
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm1 = ymm13[1,3,2,3,5,7,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5],ymm0[6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm14, %ymm9, %ymm1
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm3, %ymm0, %ymm1
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm3 = ymm9[1,3,2,3,5,7,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm3[3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm8, %ymm11, %ymm2
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm13, %ymm0, %ymm0
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm1 = ymm7[1,3,2,3,5,7,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm14, %ymm11, %ymm1
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm0 = ymm10[2,3,2,3,6,7,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm1 = ymm11[0,0,2,3,4,4,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovaps (%rdi), %xmm2
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} xmm0 = xmm2[2,3,2,3]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm1 = ymm13[0,0,2,3,4,4,6,7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm8[0,1],ymm9[2,3],ymm8[4,5],ymm9[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm11[0,1],ymm12[2,3],ymm11[4,5],ymm12[6,7]
; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm1 = ymm1[2,0,2,3,6,4,6,7]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm1 = ymm6[0,0,0,0,4,4,4,4]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm2 = ymm7[0,0,2,3,4,4,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm1 = ymm9[0,0,0,0,4,4,4,4]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm3 = ymm10[0,0,2,3,4,4,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm3[0,1,2],ymm1[3],ymm3[4,5,6],ymm1[7]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm0 = mem[2,3,2,3,6,7,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm1 = ymm14[0,0,2,3,4,4,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups (%rsp), %ymm1 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm1 = mem[0,1],ymm1[2,3],mem[4,5],ymm1[6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm1 = ymm1[2,0,2,3,6,4,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
+; AVX2-FAST-PERLANE-NEXT: vmovaps 384(%rdi), %xmm1
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} xmm0 = xmm1[2,3,2,3]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm3 = ymm8[0,0,2,3,4,4,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm3[1,2,3],ymm0[4],ymm3[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm3 = mem[0,1],ymm3[2,3],mem[4,5],ymm3[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm3 = ymm3[2,0,2,3,6,4,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm3 = ymm3[0,2,0,3]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm1 = mem[0,0,0,0,4,4,4,4]
-; AVX2-FAST-PERLANE-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm2 = mem[0,0,2,3,4,4,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
-; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm3[3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm3 = mem[0,0,0,0,4,4,4,4]
+; AVX2-FAST-PERLANE-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm4 = mem[0,0,2,3,4,4,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3],ymm4[4,5,6],ymm3[7]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm3 = ymm3[0,1,0,3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm3[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm0 = mem[2,3,2,3,6,7,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm1 = mem[0,0,2,3,4,4,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovaps 576(%rdi), %xmm0
+; AVX2-FAST-PERLANE-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[2,3,2,3]
+; AVX2-FAST-PERLANE-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm4 = mem[0,0,2,3,4,4,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm4[1,2,3],ymm0[4],ymm4[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps $204, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm4 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm4 = ymm15[0,1],mem[2,3],ymm15[4,5],mem[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm4 = ymm4[2,0,2,3,6,4,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm4 = ymm4[0,2,0,3]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm4[3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm4 = mem[0,0,0,0,4,4,4,4]
+; AVX2-FAST-PERLANE-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm5 = mem[0,0,2,3,4,4,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm5[0,1,2],ymm4[3],ymm5[4,5,6],ymm4[7]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm4 = ymm4[0,1,0,3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm4[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovaps 192(%rdi), %xmm0
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} xmm4 = xmm0[2,3,2,3]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm5 = ymm7[0,0,2,3,4,4,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0],ymm5[1,2,3],ymm4[4],ymm5[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm14[0,1],ymm15[2,3],ymm14[4,5],ymm15[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm5 = ymm5[2,0,2,3,6,4,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm5 = ymm5[0,2,0,3]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm4 = ymm4[0,3,2,3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2],ymm5[3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm5 = ymm3[0,0,0,0,4,4,4,4]
+; AVX2-FAST-PERLANE-NEXT: vpermilps $224, (%rsp), %ymm6 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm6 = mem[0,0,2,3,4,4,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm6[0,1,2],ymm5[3],ymm6[4,5,6],ymm5[7]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm5 = ymm5[0,1,0,3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4],ymm5[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} xmm2 = xmm2[3,3,3,3]
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm4 = ymm13[0,1,3,3,4,5,7,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0],ymm4[1,2,3],ymm2[4],ymm4[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm4 = ymm12[3,3,3,3,7,7,7,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0],ymm11[1],ymm4[2,3,4],ymm11[5],ymm4[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm4 = ymm4[0,2,0,3]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,3,2,3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm4[3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm4 = ymm9[0,1,0,1,4,5,4,5]
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm5 = ymm10[0,1,3,3,4,5,7,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm5[0,1,2],ymm4[3],ymm5[4,5,6],ymm4[7]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm4 = ymm4[0,1,0,3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm4[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} xmm1 = xmm1[3,3,3,3]
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm2 = ymm8[0,1,3,3,4,5,7,7]
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm8, %ymm13
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm2[1,2,3],ymm1[4],ymm2[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm2 = ymm9[3,3,3,3,7,7,7,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0],ymm5[1],ymm2[2,3,4],ymm5[5],ymm2[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,2,0,3]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,3,2,3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm2 = ymm8[0,1,0,1,4,5,4,5]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm4 = ymm6[0,1,3,3,4,5,7,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm4[0,1,2],ymm2[3],ymm4[4,5,6],ymm2[7]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,1,0,3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm1 = ymm7[0,1,3,3,4,5,7,7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm1 = mem[0,1],ymm1[2,3],mem[4,5],ymm1[6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm1 = ymm1[2,0,2,3,6,4,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm1 = ymm15[3,3,3,3,7,7,7,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm14[1],ymm1[2,3,4],ymm14[5],ymm1[6,7]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm1 = mem[0,0,0,0,4,4,4,4]
-; AVX2-FAST-PERLANE-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm2 = mem[0,0,2,3,4,4,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm1 = ymm3[0,1,0,1,4,5,4,5]
+; AVX2-FAST-PERLANE-NEXT: vmovups (%rsp), %ymm10 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm2 = ymm10[0,1,3,3,4,5,7,7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm0 = ymm5[2,3,2,3,6,7,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm1 = ymm4[0,0,2,3,4,4,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # xmm0 = mem[3,3,3,3]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm1 = ymm3[0,1,3,3,4,5,7,7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm12[0,1],ymm13[2,3],ymm12[4,5],ymm13[6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm1 = ymm1[2,0,2,3,6,4,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm1 = ymm12[3,3,3,3,7,7,7,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm11[1],ymm1[2,3,4],ymm11[5],ymm1[6,7]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm1 = ymm3[0,0,0,0,4,4,4,4]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm1 = mem[0,1,0,1,4,5,4,5]
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm2 = ymm15[0,0,2,3,4,4,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm2 = ymm15[0,1,3,3,4,5,7,7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm0 = ymm10[3,3,3,3,7,7,7,7]
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm1 = ymm11[0,1,3,3,4,5,7,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm1 = ymm9[3,3,3,3,7,7,7,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm8[1],ymm1[2,3,4],ymm8[5],ymm1[6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
-; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm1 = ymm6[0,1,0,1,4,5,4,5]
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm2 = ymm7[0,1,3,3,4,5,7,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
-; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm0 = ymm6[3,3,3,3,7,7,7,7]
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm1 = ymm14[0,1,3,3,4,5,7,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups (%rsp), %ymm8 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm1 = ymm8[3,3,3,3,7,7,7,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm10[1],ymm1[2,3,4],ymm10[5],ymm1[6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
-; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm1 = ymm14[0,1,0,1,4,5,4,5]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm2 = ymm7[0,1,3,3,4,5,7,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
-; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm0 = ymm5[3,3,3,3,7,7,7,7]
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm1 = ymm4[0,1,3,3,4,5,7,7]
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm4, %ymm5
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm1 = ymm13[3,3,3,3,7,7,7,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm12[1],ymm1[2,3,4],ymm12[5],ymm1[6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm12, %ymm13
-; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
-; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm1 = ymm3[0,1,0,1,4,5,4,5]
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm2 = ymm15[0,1,3,3,4,5,7,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
-; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm0 = ymm3[3,3,3,3,7,7,7,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm1 = ymm12[0,1,3,3,4,5,7,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm1 = ymm11[3,3,3,3,7,7,7,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm9[1],ymm1[2,3,4],ymm9[5],ymm1[6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
-; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm1 = mem[0,1,0,1,4,5,4,5]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm2 = ymm15[0,1,3,3,4,5,7,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
-; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm0 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm0 = ymm13[0,1,2,3],mem[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vblendps $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm0 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm0 = mem[0,1,2,3],ymm6[4,5],mem[6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm10[0,1,2,3],ymm8[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm14 = ymm5[0,1,2,3],ymm9[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovaps 464(%rdi), %xmm1
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3],ymm2[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
-; AVX2-FAST-PERLANE-NEXT: vmovaps {{.*#+}} xmm2 = <4,2,u,u>
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm0, %ymm2, %ymm4
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm2, %ymm6
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm4[0,1],ymm1[2,3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm7 = ymm14[0,1],ymm7[2,3],ymm14[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vbroadcastf128 {{.*#+}} ymm10 = [0,2,0,6,0,2,0,6]
-; AVX2-FAST-PERLANE-NEXT: # ymm10 = mem[0,1,0,1]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm7, %ymm10, %ymm4
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm4[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm0 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm0 = ymm5[0,1,2,3],mem[4,5],ymm5[6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm14[0,1],ymm1[2,3],ymm14[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm2 = ymm2[0,1,0,2,4,5,4,6]
+; AVX2-FAST-PERLANE-NEXT: vmovaps {{.*#+}} xmm1 = <4,2,u,u>
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm0, %ymm1, %ymm4
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm1, %ymm9
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm2[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm7 = ymm8[0,1],ymm6[2,3],ymm8[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vbroadcastf128 {{.*#+}} ymm2 = [0,2,0,6,0,2,0,6]
+; AVX2-FAST-PERLANE-NEXT: # ymm2 = mem[0,1,0,1]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm7, %ymm2, %ymm5
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm4[0,1,2,3,4],ymm5[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm2 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm2 = ymm13[0,1,2,3],mem[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovaps 272(%rdi), %xmm13
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm2[0,1],ymm13[2,3],ymm2[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm5 = ymm5[0,1,0,2,4,5,4,6]
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm6, %ymm1
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm0, %ymm6, %ymm6
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1],ymm5[2,3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm5 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm5 = mem[0,1],ymm0[2,3],mem[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm5, %ymm10, %ymm8
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm6[0,1,2,3,4],ymm8[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm0 = ymm0[0,1,2,3],mem[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm13 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm13 = ymm1[0,1,2,3],mem[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovaps 272(%rdi), %xmm6
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm13[0,1],ymm6[2,3],ymm13[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm4 = ymm4[0,1,0,2,4,5,4,6]
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm9, %ymm1
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm0, %ymm9, %ymm8
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm8[0,1],ymm4[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm10 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm10 = mem[0,1],ymm10[2,3],mem[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm10, %ymm2, %ymm8
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm4[0,1,2,3,4],ymm8[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, (%rsp) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm8 = ymm12[0,1,2,3],ymm3[4,5],ymm12[6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm12 = ymm9[0,1,2,3],ymm11[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovaps 656(%rdi), %xmm6
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm9 = ymm12[0,1],ymm6[2,3],ymm12[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm9 = ymm9[0,1,0,2,4,5,4,6]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm8, %ymm1, %ymm11
+; AVX2-FAST-PERLANE-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm9 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm9 = ymm3[0,1,2,3],mem[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm12 = ymm11[0,1,2,3],ymm12[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovaps 656(%rdi), %xmm4
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm8 = ymm12[0,1],ymm4[2,3],ymm12[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm8 = ymm8[0,1,0,2,4,5,4,6]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm9, %ymm1, %ymm11
; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm1, %ymm3
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm9 = ymm11[0,1],ymm9[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm8 = ymm11[0,1],ymm8[2,3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # ymm15 = mem[0,1],ymm15[2,3],mem[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm15, %ymm10, %ymm11
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm9[0,1,2,3,4],ymm11[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm15, %ymm2, %ymm11
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm8[0,1,2,3,4],ymm11[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm9 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm9 = ymm0[0,1,2,3],mem[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm8 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm8 = ymm0[0,1,2,3],mem[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm1 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm1 = ymm0[0,1,2,3],mem[4,5],ymm0[6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm1 = ymm0[0,1,2,3],mem[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovaps 80(%rdi), %xmm0
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm11 = ymm9[0,1],ymm0[2,3],ymm9[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm11 = ymm8[0,1],ymm0[2,3],ymm8[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm11 = ymm11[0,1,0,2,4,5,4,6]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm1, %ymm3, %ymm4
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm11[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm1, %ymm3, %ymm5
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1],ymm11[2,3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm11 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # ymm11 = mem[0,1],ymm3[2,3],mem[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm11, %ymm10, %ymm10
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm10 = ymm4[0,1,2,3,4],ymm10[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm3 = mem[1,1,1,1,5,5,5,5]
-; AVX2-FAST-PERLANE-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm3 = ymm3[0,1,2],mem[3],ymm3[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovaps {{.*#+}} xmm4 = <5,3,u,u>
-; AVX2-FAST-PERLANE-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm14 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm14[0,1],ymm3[2,3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vbroadcastf128 {{.*#+}} ymm14 = [0,3,1,7,0,3,1,7]
-; AVX2-FAST-PERLANE-NEXT: # ymm14 = mem[0,1,0,1]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm7, %ymm14, %ymm7
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4],ymm7[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm2 = ymm2[1,1,1,1,5,5,5,5]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm13[3],ymm2[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm7 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm7[0,1],ymm2[2,3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm5, %ymm14, %ymm5
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm5[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm5 = ymm12[1,1,1,1,5,5,5,5]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2],ymm6[3],ymm5[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm8, %ymm4, %ymm6
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm6[0,1],ymm5[2,3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm15, %ymm14, %ymm6
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4],ymm6[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm6 = ymm9[1,1,1,1,5,5,5,5]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm6[0,1,2],ymm0[3],ymm6[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm1, %ymm4, %ymm1
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm11, %ymm2, %ymm2
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm2 = ymm14[1,1,1,1,5,5,5,5]
+; AVX2-FAST-PERLANE-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm2 = ymm2[0,1,2],mem[3],ymm2[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovaps {{.*#+}} xmm14 = <5,3,u,u>
+; AVX2-FAST-PERLANE-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm3 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vbroadcastf128 {{.*#+}} ymm3 = [0,3,1,7,0,3,1,7]
+; AVX2-FAST-PERLANE-NEXT: # ymm3 = mem[0,1,0,1]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm7, %ymm3, %ymm7
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm7[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm7 = ymm13[1,1,1,1,5,5,5,5]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm6 = ymm7[0,1,2],ymm6[3],ymm7[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm7 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm6 = ymm7[0,1],ymm6[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm10, %ymm3, %ymm7
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4],ymm7[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm7 = ymm12[1,1,1,1,5,5,5,5]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm7[0,1,2],ymm4[3],ymm7[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm9, %ymm14, %ymm7
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm7[0,1],ymm4[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm15, %ymm3, %ymm7
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4],ymm7[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm7 = ymm8[1,1,1,1,5,5,5,5]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm7[0,1,2],ymm0[3],ymm7[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm1, %ymm14, %ymm1
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm11, %ymm14, %ymm1
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm11, %ymm3, %ymm1
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm1, 96(%rsi)
@@ -4423,13 +4394,13 @@ define void @load_i32_stride6_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm1, 32(%r9)
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm1, 64(%r9)
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm10, (%r9)
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm5, (%r9)
; AVX2-FAST-PERLANE-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm5, 96(%rax)
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm2, 32(%rax)
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm3, 64(%rax)
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm4, 96(%rax)
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm6, 32(%rax)
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm2, 64(%rax)
; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm0, (%rax)
-; AVX2-FAST-PERLANE-NEXT: addq $1192, %rsp # imm = 0x4A8
+; AVX2-FAST-PERLANE-NEXT: addq $1160, %rsp # imm = 0x488
; AVX2-FAST-PERLANE-NEXT: vzeroupper
; AVX2-FAST-PERLANE-NEXT: retq
;
@@ -6044,39 +6015,39 @@ define void @load_i32_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
;
; AVX1-ONLY-LABEL: load_i32_stride6_vf64:
; AVX1-ONLY: # %bb.0:
-; AVX1-ONLY-NEXT: subq $2568, %rsp # imm = 0xA08
+; AVX1-ONLY-NEXT: subq $2488, %rsp # imm = 0x9B8
+; AVX1-ONLY-NEXT: vmovaps 608(%rdi), %ymm6
+; AVX1-ONLY-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 672(%rdi), %ymm2
; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 640(%rdi), %ymm3
; AVX1-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 320(%rdi), %ymm4
-; AVX1-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 352(%rdi), %ymm5
-; AVX1-ONLY-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 224(%rdi), %ymm6
-; AVX1-ONLY-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 192(%rdi), %ymm7
+; AVX1-ONLY-NEXT: vmovapd 352(%rdi), %ymm4
+; AVX1-ONLY-NEXT: vmovupd %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovapd 320(%rdi), %ymm5
+; AVX1-ONLY-NEXT: vmovupd %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 224(%rdi), %ymm7
; AVX1-ONLY-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 192(%rdi), %ymm8
+; AVX1-ONLY-NEXT: vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 288(%rdi), %ymm1
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 256(%rdi), %ymm0
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vinsertf128 $1, 288(%rdi), %ymm0, %ymm8
-; AVX1-ONLY-NEXT: vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm1[2,0],ymm8[0,0],ymm1[6,4],ymm8[4,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm8[2,2],ymm0[6,4],ymm8[6,6]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm7[0,1,2,3],ymm6[4,5],ymm7[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vextractf128 $1, %ymm1, %xmm6
-; AVX1-ONLY-NEXT: vmovaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm6[2,3]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm6[0,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm5[0,1,2,3],ymm4[4,5],ymm5[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm4[2,3,0,1]
+; AVX1-ONLY-NEXT: vinsertf128 $1, 288(%rdi), %ymm0, %ymm9
+; AVX1-ONLY-NEXT: vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm1[2,0],ymm9[0,0],ymm1[6,4],ymm9[4,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm9[2,2],ymm0[6,4],ymm9[6,6]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm8[0,1,2,3],ymm7[4,5],ymm8[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm1[2,0],ymm4[0,0],ymm1[6,4],ymm4[4,4]
+; AVX1-ONLY-NEXT: vextractf128 $1, %ymm1, %xmm7
+; AVX1-ONLY-NEXT: vmovaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm7[2,3]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm7[0,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm5[2,3],ymm4[0,1]
+; AVX1-ONLY-NEXT: vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm1 = ymm1[0],ymm5[0],ymm1[3],ymm5[2]
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -6084,26 +6055,21 @@ define void @load_i32_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm2[2,0],ymm1[0,0],ymm2[6,4],ymm1[4,4]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm1[2,2],ymm0[6,4],ymm1[6,6]
-; AVX1-ONLY-NEXT: vmovaps 608(%rdi), %ymm1
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 576(%rdi), %ymm2
-; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5],ymm2[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vextractf128 $1, %ymm1, %xmm2
-; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm2[0,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT: vmovaps 704(%rdi), %ymm1
+; AVX1-ONLY-NEXT: vmovaps 576(%rdi), %ymm1
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 736(%rdi), %ymm2
-; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm6[4,5],ymm1[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm2 = ymm1[2,3,0,1]
-; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm2[2,0],ymm1[0,0],ymm2[6,4],ymm1[4,4]
+; AVX1-ONLY-NEXT: vextractf128 $1, %ymm1, %xmm14
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm14[2,3]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm14[0,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovapd 736(%rdi), %ymm1
+; AVX1-ONLY-NEXT: vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovapd 704(%rdi), %ymm2
+; AVX1-ONLY-NEXT: vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm2[2,3],ymm1[0,1]
+; AVX1-ONLY-NEXT: vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm1 = ymm1[0],ymm2[0],ymm1[3],ymm2[2]
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -6111,29 +6077,26 @@ define void @load_i32_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 1024(%rdi), %ymm0
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vinsertf128 $1, 1056(%rdi), %ymm0, %ymm2
-; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm1[2,0],ymm2[0,0],ymm1[6,4],ymm2[4,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm2[2,2],ymm0[6,4],ymm2[6,6]
+; AVX1-ONLY-NEXT: vinsertf128 $1, 1056(%rdi), %ymm0, %ymm13
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm1[2,0],ymm13[0,0],ymm1[6,4],ymm13[4,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm13[2,2],ymm0[6,4],ymm13[6,6]
; AVX1-ONLY-NEXT: vmovaps 992(%rdi), %ymm1
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 960(%rdi), %ymm2
; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5],ymm2[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vextractf128 $1, %ymm1, %xmm13
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm13[2,3]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm13[0,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT: vmovaps 1088(%rdi), %ymm1
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 1120(%rdi), %ymm2
-; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm1[4,5],ymm2[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm2[2,3,0,1]
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm1[2,0],ymm2[0,0],ymm1[6,4],ymm2[4,4]
+; AVX1-ONLY-NEXT: vextractf128 $1, %ymm1, %xmm11
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm11[2,3]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm11[0,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovapd 1120(%rdi), %ymm1
+; AVX1-ONLY-NEXT: vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovapd 1088(%rdi), %ymm2
+; AVX1-ONLY-NEXT: vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm2[2,3],ymm1[0,1]
+; AVX1-ONLY-NEXT: vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm1 = ymm1[0],ymm2[0],ymm1[3],ymm2[2]
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -6141,9 +6104,9 @@ define void @load_i32_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 1408(%rdi), %ymm0
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vinsertf128 $1, 1440(%rdi), %ymm0, %ymm11
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm1[2,0],ymm11[0,0],ymm1[6,4],ymm11[4,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm11[2,2],ymm0[6,4],ymm11[6,6]
+; AVX1-ONLY-NEXT: vinsertf128 $1, 1440(%rdi), %ymm0, %ymm10
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm1[2,0],ymm10[0,0],ymm1[6,4],ymm10[4,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm10[2,2],ymm0[6,4],ymm10[6,6]
; AVX1-ONLY-NEXT: vmovaps 1376(%rdi), %ymm1
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 1344(%rdi), %ymm2
@@ -6153,16 +6116,14 @@ define void @load_i32_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm1, %xmm9
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm9[2,3]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm9[0,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT: vmovaps 1472(%rdi), %ymm1
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 1504(%rdi), %ymm2
-; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm1[4,5],ymm2[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm2[2,3,0,1]
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm1[2,0],ymm2[0,0],ymm1[6,4],ymm2[4,4]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovapd 1504(%rdi), %ymm1
+; AVX1-ONLY-NEXT: vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovapd 1472(%rdi), %ymm2
+; AVX1-ONLY-NEXT: vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm2[2,3],ymm1[0,1]
+; AVX1-ONLY-NEXT: vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm1 = ymm1[0],ymm2[0],ymm1[3],ymm2[2]
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -6170,28 +6131,26 @@ define void @load_i32_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 64(%rdi), %ymm0
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vinsertf128 $1, 96(%rdi), %ymm0, %ymm7
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm1[2,0],ymm7[0,0],ymm1[6,4],ymm7[4,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm7[2,2],ymm0[6,4],ymm7[6,6]
+; AVX1-ONLY-NEXT: vinsertf128 $1, 96(%rdi), %ymm0, %ymm8
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm1[2,0],ymm8[0,0],ymm1[6,4],ymm8[4,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm8[2,2],ymm0[6,4],ymm8[6,6]
; AVX1-ONLY-NEXT: vmovaps 32(%rdi), %ymm1
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps (%rdi), %ymm2
; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5],ymm2[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vextractf128 $1, %ymm1, %xmm6
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm6[2,3]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm6[0,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT: vmovaps 128(%rdi), %ymm1
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 160(%rdi), %ymm2
-; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm1[4,5],ymm2[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm2[2,3,0,1]
-; AVX1-ONLY-NEXT: vmovups %ymm1, (%rsp) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm1[2,0],ymm2[0,0],ymm1[6,4],ymm2[4,4]
+; AVX1-ONLY-NEXT: vextractf128 $1, %ymm1, %xmm7
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm7[2,3]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm7[0,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovapd 160(%rdi), %ymm1
+; AVX1-ONLY-NEXT: vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovapd 128(%rdi), %ymm2
+; AVX1-ONLY-NEXT: vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm2[2,3],ymm1[0,1]
+; AVX1-ONLY-NEXT: vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm1 = ymm1[0],ymm2[0],ymm1[3],ymm2[2]
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -6199,28 +6158,26 @@ define void @load_i32_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 448(%rdi), %ymm0
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vinsertf128 $1, 480(%rdi), %ymm0, %ymm5
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm1[2,0],ymm5[0,0],ymm1[6,4],ymm5[4,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm5[2,2],ymm0[6,4],ymm5[6,6]
+; AVX1-ONLY-NEXT: vinsertf128 $1, 480(%rdi), %ymm0, %ymm6
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm1[2,0],ymm6[0,0],ymm1[6,4],ymm6[4,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm6[2,2],ymm0[6,4],ymm6[6,6]
; AVX1-ONLY-NEXT: vmovaps 416(%rdi), %ymm1
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 384(%rdi), %ymm2
; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5],ymm2[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vextractf128 $1, %ymm1, %xmm4
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm4[2,3]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm4[0,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT: vmovaps 512(%rdi), %ymm1
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 544(%rdi), %ymm2
-; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm1[4,5],ymm2[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm2[2,3,0,1]
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm1[2,0],ymm2[0,0],ymm1[6,4],ymm2[4,4]
+; AVX1-ONLY-NEXT: vextractf128 $1, %ymm1, %xmm5
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm5[2,3]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm5[0,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovapd 544(%rdi), %ymm1
+; AVX1-ONLY-NEXT: vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovapd 512(%rdi), %ymm2
+; AVX1-ONLY-NEXT: vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm2[2,3],ymm1[0,1]
+; AVX1-ONLY-NEXT: vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm1 = ymm1[0],ymm2[0],ymm1[3],ymm2[2]
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -6228,28 +6185,26 @@ define void @load_i32_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 832(%rdi), %ymm0
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vinsertf128 $1, 864(%rdi), %ymm0, %ymm3
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm1[2,0],ymm3[0,0],ymm1[6,4],ymm3[4,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm3[2,2],ymm0[6,4],ymm3[6,6]
+; AVX1-ONLY-NEXT: vinsertf128 $1, 864(%rdi), %ymm0, %ymm4
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm1[2,0],ymm4[0,0],ymm1[6,4],ymm4[4,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm4[2,2],ymm0[6,4],ymm4[6,6]
; AVX1-ONLY-NEXT: vmovaps 800(%rdi), %ymm1
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 768(%rdi), %ymm2
; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5],ymm2[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vextractf128 $1, %ymm1, %xmm2
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm2[0,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT: vmovaps 896(%rdi), %ymm1
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 928(%rdi), %ymm8
-; AVX1-ONLY-NEXT: vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3],ymm1[4,5],ymm8[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm8[2,3,0,1]
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm1[2,0],ymm8[0,0],ymm1[6,4],ymm8[4,4]
+; AVX1-ONLY-NEXT: vextractf128 $1, %ymm1, %xmm3
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm3[2,3]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm3[0,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovapd 928(%rdi), %ymm1
+; AVX1-ONLY-NEXT: vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovapd 896(%rdi), %ymm2
+; AVX1-ONLY-NEXT: vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm2[2,3],ymm1[0,1]
+; AVX1-ONLY-NEXT: vmovupd %ymm1, (%rsp) # 32-byte Spill
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm1 = ymm1[0],ymm2[0],ymm1[3],ymm2[2]
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -6257,150 +6212,149 @@ define void @load_i32_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 1216(%rdi), %ymm0
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vinsertf128 $1, 1248(%rdi), %ymm0, %ymm12
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm1[2,0],ymm12[0,0],ymm1[6,4],ymm12[4,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm8 = ymm0[2,0],ymm12[2,2],ymm0[6,4],ymm12[6,6]
-; AVX1-ONLY-NEXT: vmovaps 1184(%rdi), %ymm0
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 1152(%rdi), %ymm10
-; AVX1-ONLY-NEXT: vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm10[0,1,2,3],ymm0[4,5],ymm10[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vextractf128 $1, %ymm0, %xmm10
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm15 = xmm0[0,1],xmm10[2,3]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm15 = xmm15[0,2],xmm10[0,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm15 = ymm15[0,1,2],ymm8[3,4,5],ymm15[6,7]
-; AVX1-ONLY-NEXT: vmovaps 1280(%rdi), %ymm8
-; AVX1-ONLY-NEXT: vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 1312(%rdi), %ymm0
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm0[0,1,2,3],ymm8[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm8[2,3,0,1]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm14 = ymm0[2,0],ymm8[0,0],ymm0[6,4],ymm8[4,4]
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} ymm14 = ymm14[0,1,2,0,4,5,6,4]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm14 = ymm15[0,1,2,3,4,5],ymm14[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm14 = ymm14[3,0],ymm0[1,0],ymm14[7,4],ymm0[5,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm14 = ymm14[2,0],ymm0[2,3],ymm14[6,4],ymm0[6,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm15 = xmm0[1,0],xmm1[3,0]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm15 = xmm15[0,2],xmm1[1,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm14 = ymm15[0,1,2],ymm14[3,4,5],ymm15[6,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps $19, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm15 = ymm0[3,0],mem[1,0],ymm0[7,4],mem[5,4]
+; AVX1-ONLY-NEXT: vinsertf128 $1, 1248(%rdi), %ymm0, %ymm2
+; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm1[2,0],ymm2[0,0],ymm1[6,4],ymm2[4,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm2[2,2],ymm0[6,4],ymm2[6,6]
+; AVX1-ONLY-NEXT: vmovaps 1184(%rdi), %ymm1
+; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 1152(%rdi), %ymm12
+; AVX1-ONLY-NEXT: vmovups %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm12[0,1,2,3],ymm1[4,5],ymm12[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vextractf128 $1, %ymm1, %xmm12
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm15 = xmm1[0,1],xmm12[2,3]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm15 = xmm15[0,2],xmm12[0,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm15[0,1,2],ymm0[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovapd 1312(%rdi), %ymm15
+; AVX1-ONLY-NEXT: vmovupd %ymm15, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovapd 1280(%rdi), %ymm0
+; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm2 = ymm0[2,3],ymm15[0,1]
+; AVX1-ONLY-NEXT: vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm15 = ymm2[0],ymm0[0],ymm2[3],ymm0[2]
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} ymm15 = ymm15[0,1,2,0,4,5,6,4]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4,5],ymm15[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm15[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm14 = ymm0[3,0],ymm1[1,0],ymm0[7,4],ymm1[5,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm14 = ymm14[2,0],ymm1[2,3],ymm14[6,4],ymm1[6,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm15 = xmm0[1,0],xmm1[3,0]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm15 = xmm15[0,2],xmm1[1,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm14 = ymm15[0,1,2],ymm14[3,4,5],ymm15[6,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps $19, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm15 = ymm0[3,0],mem[1,0],ymm0[7,4],mem[5,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[3,0],ymm1[1,0],ymm0[7,4],ymm1[5,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm0[2,0],ymm1[2,3],ymm0[6,4],ymm1[6,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm15 = xmm1[1,0],xmm0[3,0]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm15 = xmm15[0,2],xmm0[1,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm15[0,1,2],ymm2[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps $215, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm15 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm15 = ymm1[3,1],mem[1,3],ymm1[7,5],mem[5,7]
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} ymm15 = ymm15[0,1,2,0,4,5,6,4]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4,5],ymm15[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm15[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm14 = ymm0[3,0],ymm1[1,0],ymm0[7,4],ymm1[5,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm14 = ymm14[2,0],ymm1[2,3],ymm14[6,4],ymm1[6,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm15 = xmm0[1,0],xmm13[3,0]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm13 = xmm15[0,2],xmm13[1,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0,1,2],ymm14[3,4,5],ymm13[6,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps $19, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm14 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm14 = ymm0[3,0],mem[1,0],ymm0[7,4],mem[5,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[3,0],ymm1[1,0],ymm0[7,4],ymm1[5,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm1[2,3],ymm0[6,4],ymm1[6,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm15 = xmm1[1,0],xmm14[3,0]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm14 = xmm15[0,2],xmm14[1,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm14[0,1,2],ymm0[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm14 = ymm1[3,1],ymm15[1,3],ymm1[7,5],ymm15[5,7]
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} ymm14 = ymm14[0,1,2,0,4,5,6,4]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3,4,5],ymm14[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm13 = ymm14[3,0],ymm11[1,0],ymm14[7,4],ymm11[5,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm11 = ymm13[2,0],ymm11[2,3],ymm13[6,4],ymm11[6,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm13 = xmm0[1,0],xmm9[3,0]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm9 = xmm13[0,2],xmm9[1,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm9 = ymm9[0,1,2],ymm11[3,4,5],ymm9[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm14[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps $19, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm11 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm11 = ymm0[3,0],mem[1,0],ymm0[7,4],mem[5,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[3,0],ymm13[1,0],ymm0[7,4],ymm13[5,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm13[2,3],ymm0[6,4],ymm13[6,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm13 = xmm1[1,0],xmm11[3,0]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm11 = xmm13[0,2],xmm11[1,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm11[0,1,2],ymm0[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm11 = ymm1[3,1],ymm13[1,3],ymm1[7,5],ymm13[5,7]
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} ymm11 = ymm11[0,1,2,0,4,5,6,4]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm9 = ymm9[0,1,2,3,4,5],ymm11[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm11[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm9 = ymm0[3,0],ymm7[1,0],ymm0[7,4],ymm7[5,4]
-; AVX1-ONLY-NEXT: vmovaps %ymm0, %ymm13
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm7 = ymm9[2,0],ymm7[2,3],ymm9[6,4],ymm7[6,7]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[3,0],ymm10[1,0],ymm0[7,4],ymm10[5,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm10[2,3],ymm0[6,4],ymm10[6,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm10 = xmm1[1,0],xmm9[3,0]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm9 = xmm10[0,2],xmm9[1,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm9[0,1,2],ymm0[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm9 = ymm1[3,1],ymm14[1,3],ymm1[7,5],ymm14[5,7]
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} ymm9 = ymm9[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm9[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm9 = xmm0[1,0],xmm6[3,0]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm6 = xmm9[0,2],xmm6[1,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2],ymm7[3,4,5],ymm6[6,7]
-; AVX1-ONLY-NEXT: vmovups (%rsp), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps $19, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm7 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm7 = ymm0[3,0],mem[1,0],ymm0[7,4],mem[5,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[3,0],ymm8[1,0],ymm0[7,4],ymm8[5,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm8[2,3],ymm0[6,4],ymm8[6,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm8 = xmm1[1,0],xmm7[3,0]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm7 = xmm8[0,2],xmm7[1,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm7[0,1,2],ymm0[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps $215, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm7 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm7 = ymm1[3,1],mem[1,3],ymm1[7,5],mem[5,7]
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} ymm7 = ymm7[0,1,2,0,4,5,6,4]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5],ymm7[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm6 = ymm0[3,0],ymm5[1,0],ymm0[7,4],ymm5[5,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm5 = ymm6[2,0],ymm5[2,3],ymm6[6,4],ymm5[6,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm6 = xmm0[1,0],xmm4[3,0]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm4 = xmm6[0,2],xmm4[1,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2],ymm5[3,4,5],ymm4[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm7[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps $19, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm5 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm5 = ymm0[3,0],mem[1,0],ymm0[7,4],mem[5,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[3,0],ymm6[1,0],ymm0[7,4],ymm6[5,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm6[2,3],ymm0[6,4],ymm6[6,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm6 = xmm1[1,0],xmm5[3,0]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm5 = xmm6[0,2],xmm5[1,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm5[0,1,2],ymm0[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps $215, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm5 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm5 = ymm1[3,1],mem[1,3],ymm1[7,5],mem[5,7]
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} ymm5 = ymm5[0,1,2,0,4,5,6,4]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5],ymm5[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm4 = ymm0[3,0],ymm3[1,0],ymm0[7,4],ymm3[5,4]
-; AVX1-ONLY-NEXT: vmovaps %ymm0, %ymm6
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm4[2,0],ymm3[2,3],ymm4[6,4],ymm3[6,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm4 = xmm0[1,0],xmm2[3,0]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm2 = xmm4[0,2],xmm2[1,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm3[3,4,5],ymm2[6,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps $19, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm3 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm3 = ymm0[3,0],mem[1,0],ymm0[7,4],mem[5,4]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm5[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm6[3,0],ymm4[1,0],ymm6[7,4],ymm4[5,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm4[2,3],ymm0[6,4],ymm4[6,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm4 = xmm1[1,0],xmm3[3,0]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm3 = xmm4[0,2],xmm3[1,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm3[0,1,2],ymm0[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups (%rsp), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps $215, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm3 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm3 = ymm1[3,1],mem[1,3],ymm1[7,5],mem[5,7]
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} ymm3 = ymm3[0,1,2,0,4,5,6,4]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm3[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm3[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm3[3,0],ymm12[1,0],ymm3[7,4],ymm12[5,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm2[2,0],ymm12[2,3],ymm2[6,4],ymm12[6,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm2 = xmm0[1,0],xmm10[3,0]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm0 = xmm2[0,2],xmm10[1,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5],ymm0[6,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm1[3,0],ymm8[1,0],ymm1[7,4],ymm8[5,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm3[3,0],ymm1[1,0],ymm3[7,4],ymm1[5,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm1[2,3],ymm0[6,4],ymm1[6,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm2 = xmm1[1,0],xmm12[3,0]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm2[0,2],xmm12[1,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm1[3,1],ymm2[1,3],ymm1[7,5],ymm2[5,7]
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm2 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm2 = ymm0[0,1],mem[2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm12 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm12 = ymm0[0,1],mem[2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vshufps $38, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm0 = ymm0[2,1],mem[2,0],ymm0[6,5],mem[6,4]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,0,1]
-; AVX1-ONLY-NEXT: vextractf128 $1, %ymm2, %xmm12
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm2[2,0],xmm12[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4],ymm1[5,6,7]
+; AVX1-ONLY-NEXT: vextractf128 $1, %ymm12, %xmm1
+; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm12[2,0],xmm1[2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX1-ONLY-NEXT: vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm1 = ymm1[0,1,2,3],mem[4,5],ymm1[6,7]
@@ -6421,10 +6375,9 @@ define void @load_i32_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm1, %xmm4
; AVX1-ONLY-NEXT: vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm1[2,0],xmm4[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4],ymm1[5,6,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm11 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm11 = ymm1[0,1,2,3],mem[4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm11 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm11 = ymm15[0,1,2,3],mem[4,5],ymm15[6,7]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm11[2,3,0,1]
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm1[0,0],ymm11[2,0],ymm1[4,4],ymm11[6,4]
@@ -6441,12 +6394,11 @@ define void @load_i32_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm1, %xmm4
; AVX1-ONLY-NEXT: vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm1[2,0],xmm4[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4],ymm1[5,6,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm10 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm10 = ymm1[0,1,2,3],mem[4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm10 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm10 = ymm13[0,1,2,3],mem[4,5],ymm13[6,7]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm10[2,3,0,1]
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovups %ymm1, (%rsp) # 32-byte Spill
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm1[0,0],ymm10[2,0],ymm1[4,4],ymm10[6,4]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -6454,18 +6406,18 @@ define void @load_i32_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm1 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm1 = ymm0[0,1],mem[2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufps $38, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm0 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm0 = ymm14[2,1],mem[2,0],ymm14[6,5],mem[6,4]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps $38, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm0 = ymm0[2,1],mem[2,0],ymm0[6,5],mem[6,4]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,0,1]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm1, %xmm4
; AVX1-ONLY-NEXT: vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm1[2,0],xmm4[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4],ymm1[5,6,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm8 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm8 = ymm1[0,1,2,3],mem[4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm8 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm8 = ymm14[0,1,2,3],mem[4,5],ymm14[6,7]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm8[2,3,0,1]
-; AVX1-ONLY-NEXT: vmovups %ymm1, (%rsp) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm1[0,0],ymm8[2,0],ymm1[4,4],ymm8[6,4]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -6479,10 +6431,9 @@ define void @load_i32_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm1, %xmm3
; AVX1-ONLY-NEXT: vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm1[2,0],xmm3[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4],ymm1[5,6,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm7 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm7 = ymm1[0,1,2,3],mem[4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm7 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm7 = ymm2[0,1,2,3],mem[4,5],ymm2[6,7]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm7[2,3,0,1]
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm1[0,0],ymm7[2,0],ymm1[4,4],ymm7[6,4]
@@ -6492,20 +6443,21 @@ define void @load_i32_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm1 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm1 = ymm0[0,1],mem[2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufps $38, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm0 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm0 = ymm13[2,1],mem[2,0],ymm13[6,5],mem[6,4]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps $38, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm0 = ymm0[2,1],mem[2,0],ymm0[6,5],mem[6,4]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,0,1]
-; AVX1-ONLY-NEXT: vextractf128 $1, %ymm1, %xmm3
-; AVX1-ONLY-NEXT: vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm1[2,0],xmm3[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4],ymm1[5,6,7]
+; AVX1-ONLY-NEXT: vextractf128 $1, %ymm1, %xmm2
+; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm1[2,0],xmm2[2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX1-ONLY-NEXT: vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm1 = ymm1[0,1,2,3],mem[4,5],ymm1[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm3 = ymm1[2,3,0,1]
-; AVX1-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm3[0,0],ymm1[2,0],ymm3[4,4],ymm1[6,4]
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm2 = ymm1[2,3,0,1]
+; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm2[0,0],ymm1[2,0],ymm2[4,4],ymm1[6,4]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
@@ -6518,7 +6470,7 @@ define void @load_i32_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm13, %xmm1
; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm13[2,0],xmm1[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4],ymm1[5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX1-ONLY-NEXT: vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm5 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm5 = ymm1[0,1,2,3],mem[4,5],ymm1[6,7]
@@ -6535,7 +6487,7 @@ define void @load_i32_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,0,1]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm4, %xmm6
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm3 = xmm4[2,0],xmm6[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm3[0,1,2],ymm0[3,4],ymm3[5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm3[0,1,2],ymm0[3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX1-ONLY-NEXT: vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm1 = ymm1[0,1,2,3],mem[4,5],ymm1[6,7]
@@ -6546,12 +6498,13 @@ define void @load_i32_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vshufps $116, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm0 = ymm0[0,1],mem[3,1],ymm0[4,5],mem[7,5]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm14 = xmm2[3,1],xmm12[3,3]
+; AVX1-ONLY-NEXT: vshufps $247, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12, %xmm14 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm14 = xmm12[3,1],mem[3,3]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
; AVX1-ONLY-NEXT: vshufps $103, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm12 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm12 = ymm12[3,1],mem[2,1],ymm12[7,5],mem[6,5]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm12 = ymm12[2,3,0,1]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm12 = ymm14[0,1,2],ymm12[3,4],ymm14[5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm12 = ymm14[0,1,2],ymm12[3,4,5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm12[0,1,2,3,4],ymm0[5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
@@ -6563,10 +6516,10 @@ define void @load_i32_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: vshufps $103, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm12 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm12 = ymm14[3,1],mem[2,1],ymm14[7,5],mem[6,5]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm12 = ymm12[2,3,0,1]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm11 = ymm11[0,1,2],ymm12[3,4],ymm11[5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm11 = ymm11[0,1,2],ymm12[3,4,5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm11[0,1,2,3,4],ymm0[5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovups (%rsp), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm0[0,1],ymm10[3,1],ymm0[4,5],ymm10[7,5]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vshufps $247, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm11 # 16-byte Folded Reload
@@ -6575,10 +6528,10 @@ define void @load_i32_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: vshufps $103, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm12 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm12 = ymm10[3,1],mem[2,1],ymm10[7,5],mem[6,5]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm12 = ymm12[2,3,0,1]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm11 = ymm11[0,1,2],ymm12[3,4],ymm11[5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm11 = ymm11[0,1,2],ymm12[3,4,5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm11[0,1,2,3,4],ymm2[5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups (%rsp), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[0,1],ymm8[3,1],ymm0[4,5],ymm8[7,5]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX1-ONLY-NEXT: vshufps $247, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm8 # 16-byte Folded Reload
@@ -6587,7 +6540,7 @@ define void @load_i32_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: vshufps $103, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm10 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm10 = ymm10[3,1],mem[2,1],ymm10[7,5],mem[6,5]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm10 = ymm10[2,3,0,1]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2],ymm10[3,4],ymm8[5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2],ymm10[3,4,5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm8[0,1,2,3,4],ymm0[5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
@@ -6598,16 +6551,15 @@ define void @load_i32_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm8 = ymm2[3,1],ymm15[2,1],ymm2[7,5],ymm15[6,5]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm8 = ymm8[2,3,0,1]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2],ymm8[3,4],ymm7[5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2],ymm8[3,4,5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm7[0,1,2,3,4],ymm0[5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm3[0,1],ymm1[3,1],ymm3[4,5],ymm1[7,5]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm4[3,1],xmm6[3,3]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm9, %ymm7
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm3[3,1],ymm9[2,1],ymm3[7,5],ymm9[6,5]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm2 = ymm2[2,3,0,1]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4],ymm1[5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
@@ -6618,7 +6570,7 @@ define void @load_i32_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm4[3,1],ymm5[2,1],ymm4[7,5],ymm5[6,5]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm2 = ymm2[2,3,0,1]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4],ymm1[5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
@@ -6631,7 +6583,7 @@ define void @load_i32_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm6[3,1],ymm8[2,1],ymm6[7,5],ymm8[6,5]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm2 = ymm2[2,3,0,1]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4],ymm1[5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
@@ -6649,9 +6601,9 @@ define void @load_i32_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm6[0,1],ymm1[2,0],ymm6[4,5],ymm1[6,4]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm9 = ymm2[2,3,0,1]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm2[2,0],ymm9[0,0],ymm2[6,4],ymm9[4,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm9[0,2],ymm1[2,0],ymm9[4,6],ymm1[6,4]
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm8 = ymm2[2,3,0,1]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm2[2,0],ymm8[0,0],ymm2[6,4],ymm8[4,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm8[0,2],ymm1[2,0],ymm8[4,6],ymm1[6,4]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
@@ -6671,7 +6623,7 @@ define void @load_i32_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,0],ymm2[4,5],ymm1[6,4]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps %ymm6, %ymm1
-; AVX1-ONLY-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovups %ymm6, (%rsp) # 32-byte Spill
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm6 = ymm6[2,3,0,1]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm1[2,0],ymm6[0,0],ymm1[6,4],ymm6[4,4]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm6[0,2],ymm1[2,0],ymm6[4,6],ymm1[6,4]
@@ -6707,7 +6659,7 @@ define void @load_i32_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm1[0],xmm0[1],xmm1[2,3]
; AVX1-ONLY-NEXT: vmovapd 656(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vmovupd %ymm1, (%rsp) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vshufpd $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm1 = ymm1[1],mem[0],ymm1[2],mem[2]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm14[0,1],ymm1[2,0],ymm14[4,5],ymm1[6,4]
@@ -6728,7 +6680,7 @@ define void @load_i32_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm1[0],xmm0[1],xmm1[2,3]
; AVX1-ONLY-NEXT: vmovapd 848(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm1 = ymm1[1],ymm7[0],ymm1[2],ymm7[2]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm1 = ymm1[1],ymm9[0],ymm1[2],ymm9[2]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm3[0,1],ymm1[2,0],ymm3[4,5],ymm1[6,4]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4,5,6,7]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm3 = ymm11[2,3,0,1]
@@ -6758,8 +6710,8 @@ define void @load_i32_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm8 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm8 = mem[0,1],ymm0[2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm9 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm9 = mem[0,1],ymm0[2,3],mem[4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps 1184(%rdi), %xmm0
; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[2,2,3,3]
@@ -6772,8 +6724,8 @@ define void @load_i32_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm7[0,1],ymm1[2,0],ymm7[4,5],ymm1[6,4]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm8[2,3,0,1]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm7 = ymm8[2,0],ymm1[0,0],ymm8[6,4],ymm1[4,4]
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm9[2,3,0,1]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm7 = ymm9[2,0],ymm1[0,0],ymm9[6,4],ymm1[4,4]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm7 = ymm1[0,2],ymm7[2,0],ymm1[4,6],ymm7[6,4]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm7[5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -6799,8 +6751,8 @@ define void @load_i32_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm13 = ymm15[0,1,2,3,4],ymm13[5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm13 = ymm13[3,0],ymm9[1,0],ymm13[7,4],ymm9[5,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm9 = ymm9[0,3],ymm13[2,0],ymm9[4,7],ymm13[6,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm13 = ymm13[3,0],ymm8[1,0],ymm13[7,4],ymm8[5,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm8 = ymm8[0,3],ymm13[2,0],ymm8[4,7],ymm13[6,4]
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
; AVX1-ONLY-NEXT: vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13, %xmm13 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm13 = xmm13[0,1],mem[2,3]
@@ -6811,99 +6763,99 @@ define void @load_i32_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm14[1,1],ymm15[2,0],ymm14[5,5],ymm15[6,4]
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm13 = xmm13[1,3,2,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0,1],ymm15[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm14 = ymm13[0,1,2,3,4],ymm9[5,6,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm13 = ymm9[3,0],ymm6[1,0],ymm9[7,4],ymm6[5,4]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm14 = ymm13[0,1,2,3,4],ymm8[5,6,7]
+; AVX1-ONLY-NEXT: vmovups (%rsp), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm13 = ymm8[3,0],ymm6[1,0],ymm8[7,4],ymm6[5,4]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm6 = ymm6[0,3],ymm13[2,0],ymm6[4,7],ymm13[6,4]
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; AVX1-ONLY-NEXT: vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9, %xmm13 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm13 = xmm9[0,1],mem[2,3]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps $215, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm15 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm15 = ymm9[3,1],mem[1,3],ymm9[7,5],mem[5,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm9[1,1],ymm15[2,0],ymm9[5,5],ymm15[6,4]
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT: vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8, %xmm13 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm13 = xmm8[0,1],mem[2,3]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps $215, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm15 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm15 = ymm8[3,1],mem[1,3],ymm8[7,5],mem[5,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm8[1,1],ymm15[2,0],ymm8[5,5],ymm15[6,4]
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm13 = xmm13[1,3,2,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0,1],ymm15[2,3,4,5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm13[0,1,2,3,4],ymm6[5,6,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm13 = ymm9[3,0],ymm5[1,0],ymm9[7,4],ymm5[5,4]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm13 = ymm8[3,0],ymm5[1,0],ymm8[7,4],ymm5[5,4]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm5 = ymm5[0,3],ymm13[2,0],ymm5[4,7],ymm13[6,4]
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; AVX1-ONLY-NEXT: vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9, %xmm13 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm13 = xmm9[0,1],mem[2,3]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps $215, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm15 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm15 = ymm9[3,1],mem[1,3],ymm9[7,5],mem[5,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm9[1,1],ymm15[2,0],ymm9[5,5],ymm15[6,4]
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT: vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8, %xmm13 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm13 = xmm8[0,1],mem[2,3]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps $215, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm15 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm15 = ymm8[3,1],mem[1,3],ymm8[7,5],mem[5,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm8[1,1],ymm15[2,0],ymm8[5,5],ymm15[6,4]
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm13 = xmm13[1,3,2,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0,1],ymm15[2,3,4,5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm13[0,1,2,3,4],ymm5[5,6,7]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm12 = ymm12[3,0],ymm4[1,0],ymm12[7,4],ymm4[5,4]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm4 = ymm4[0,3],ymm12[2,0],ymm4[4,7],ymm12[6,4]
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; AVX1-ONLY-NEXT: vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9, %xmm12 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm12 = xmm9[0,1],mem[2,3]
-; AVX1-ONLY-NEXT: vmovups (%rsp), %ymm9 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps $215, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm13 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm13 = ymm9[3,1],mem[1,3],ymm9[7,5],mem[5,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm13 = ymm9[1,1],ymm13[2,0],ymm9[5,5],ymm13[6,4]
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT: vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8, %xmm12 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm12 = xmm8[0,1],mem[2,3]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps $215, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm13 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm13 = ymm8[3,1],mem[1,3],ymm8[7,5],mem[5,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm13 = ymm8[1,1],ymm13[2,0],ymm8[5,5],ymm13[6,4]
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm12 = xmm12[1,3,2,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm12 = ymm12[0,1],ymm13[2,3,4,5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm12[0,1,2,3,4],ymm4[5,6,7]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm11 = ymm11[3,0],ymm3[1,0],ymm11[7,4],ymm3[5,4]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm3[0,3],ymm11[2,0],ymm3[4,7],ymm11[6,4]
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; AVX1-ONLY-NEXT: vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9, %xmm11 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm11 = xmm9[0,1],mem[2,3]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps $215, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm12 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm12 = ymm9[3,1],mem[1,3],ymm9[7,5],mem[5,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm12 = ymm9[1,1],ymm12[2,0],ymm9[5,5],ymm12[6,4]
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT: vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8, %xmm11 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm11 = xmm8[0,1],mem[2,3]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps $215, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm12 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm12 = ymm8[3,1],mem[1,3],ymm8[7,5],mem[5,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm12 = ymm8[1,1],ymm12[2,0],ymm8[5,5],ymm12[6,4]
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm11 = xmm11[1,3,2,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm11 = ymm11[0,1],ymm12[2,3,4,5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm11[0,1,2,3,4],ymm3[5,6,7]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm10 = ymm10[3,0],ymm2[1,0],ymm10[7,4],ymm2[5,4]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm2[0,3],ymm10[2,0],ymm2[4,7],ymm10[6,4]
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; AVX1-ONLY-NEXT: vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9, %xmm10 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm10 = xmm9[0,1],mem[2,3]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps $215, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm11 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm11 = ymm9[3,1],mem[1,3],ymm9[7,5],mem[5,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm11 = ymm9[1,1],ymm11[2,0],ymm9[5,5],ymm11[6,4]
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT: vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8, %xmm10 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm10 = xmm8[0,1],mem[2,3]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps $215, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm11 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm11 = ymm8[3,1],mem[1,3],ymm8[7,5],mem[5,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm11 = ymm8[1,1],ymm11[2,0],ymm8[5,5],ymm11[6,4]
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm10 = xmm10[1,3,2,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm10 = ymm10[0,1],ymm11[2,3,4,5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm10[0,1,2,3,4],ymm2[5,6,7]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm8 = ymm8[3,0],ymm1[1,0],ymm8[7,4],ymm1[5,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm1[0,3],ymm8[2,0],ymm1[4,7],ymm8[6,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm9 = ymm9[3,0],ymm1[1,0],ymm9[7,4],ymm1[5,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm1[0,3],ymm9[2,0],ymm1[4,7],ymm9[6,4]
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; AVX1-ONLY-NEXT: vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8, %xmm8 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm8 = xmm8[0,1],mem[2,3]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps $215, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm10 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm10 = ymm9[3,1],mem[1,3],ymm9[7,5],mem[5,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm10 = ymm9[1,1],ymm10[2,0],ymm9[5,5],ymm10[6,4]
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm8 = xmm8[1,3,2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1],ymm10[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm8[0,1,2,3,4],ymm1[5,6,7]
+; AVX1-ONLY-NEXT: vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8, %xmm9 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm9 = xmm8[0,1],mem[2,3]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps $215, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm10 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm10 = ymm8[3,1],mem[1,3],ymm8[7,5],mem[5,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm10 = ymm8[1,1],ymm10[2,0],ymm8[5,5],ymm10[6,4]
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm9 = xmm9[1,3,2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm9 = ymm9[0,1],ymm10[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm9[0,1,2,3,4],ymm1[5,6,7]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm7 = ymm7[3,0],ymm0[1,0],ymm7[7,4],ymm0[5,4]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[0,3],ymm7[2,0],ymm0[4,7],ymm7[6,4]
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
; AVX1-ONLY-NEXT: vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7, %xmm7 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm7 = xmm7[0,1],mem[2,3]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps $215, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm8 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm8 = ymm8[3,1],mem[1,3],ymm8[7,5],mem[5,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm8 = ymm9[1,1],ymm8[2,0],ymm9[5,5],ymm8[6,4]
+; AVX1-ONLY-NEXT: vshufps $215, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm9 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm9 = ymm8[3,1],mem[1,3],ymm8[7,5],mem[5,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm9 = ymm8[1,1],ymm9[2,0],ymm8[5,5],ymm9[6,4]
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm7 = xmm7[1,3,2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1],ymm8[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1],ymm9[2,3,4,5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm7[0,1,2,3,4],ymm0[5,6,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm7, 192(%rsi)
@@ -6994,42 +6946,42 @@ define void @load_i32_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: vmovaps %ymm5, 64(%rax)
; AVX1-ONLY-NEXT: vmovaps %ymm6, 32(%rax)
; AVX1-ONLY-NEXT: vmovaps %ymm14, (%rax)
-; AVX1-ONLY-NEXT: addq $2568, %rsp # imm = 0xA08
+; AVX1-ONLY-NEXT: addq $2488, %rsp # imm = 0x9B8
; AVX1-ONLY-NEXT: vzeroupper
; AVX1-ONLY-NEXT: retq
;
; AVX2-SLOW-LABEL: load_i32_stride6_vf64:
; AVX2-SLOW: # %bb.0:
-; AVX2-SLOW-NEXT: subq $2472, %rsp # imm = 0x9A8
+; AVX2-SLOW-NEXT: subq $2504, %rsp # imm = 0x9C8
; AVX2-SLOW-NEXT: vmovaps 672(%rdi), %ymm2
; AVX2-SLOW-NEXT: vmovaps 640(%rdi), %ymm3
; AVX2-SLOW-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovaps 608(%rdi), %ymm4
-; AVX2-SLOW-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovaps 320(%rdi), %ymm5
+; AVX2-SLOW-NEXT: vmovaps 608(%rdi), %ymm5
; AVX2-SLOW-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovaps 352(%rdi), %ymm6
+; AVX2-SLOW-NEXT: vmovaps 320(%rdi), %ymm6
; AVX2-SLOW-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovaps 288(%rdi), %ymm7
+; AVX2-SLOW-NEXT: vmovaps 352(%rdi), %ymm7
; AVX2-SLOW-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovaps 256(%rdi), %ymm8
+; AVX2-SLOW-NEXT: vmovaps 288(%rdi), %ymm8
; AVX2-SLOW-NEXT: vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovaps 256(%rdi), %ymm9
+; AVX2-SLOW-NEXT: vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovaps 224(%rdi), %ymm0
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovaps 192(%rdi), %ymm1
; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovaps {{.*#+}} xmm9 = <0,6,4,u>
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm14 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
-; AVX2-SLOW-NEXT: vpermps %ymm14, %ymm9, %ymm0
-; AVX2-SLOW-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm8[0,1],ymm7[0,1]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm7[6,7]
+; AVX2-SLOW-NEXT: vmovaps {{.*#+}} xmm4 = <0,6,4,u>
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm15 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm15, %ymm4, %ymm0
+; AVX2-SLOW-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm9[0,1],ymm8[0,1]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm8[6,7]
; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm1 = ymm1[0,2,2,2,4,6,6,6]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5],ymm0[6,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm6[0,1,2,3],ymm5[4,5],ymm6[6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm7[0,1,2,3],ymm6[4,5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vbroadcastsd {{.*#+}} ymm15 = [8589934596,8589934596,8589934596,8589934596]
-; AVX2-SLOW-NEXT: vpermps %ymm1, %ymm15, %ymm1
+; AVX2-SLOW-NEXT: vbroadcastsd {{.*#+}} ymm9 = [8589934596,8589934596,8589934596,8589934596]
+; AVX2-SLOW-NEXT: vpermps %ymm1, %ymm9, %ymm1
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -7038,18 +6990,18 @@ define void @load_i32_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovaps 576(%rdi), %ymm0
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm4[4,5],ymm0[6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm5[4,5],ymm0[6,7]
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpermps %ymm0, %ymm9, %ymm0
+; AVX2-SLOW-NEXT: vpermps %ymm0, %ymm4, %ymm0
; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm1 = ymm1[0,2,2,2,4,6,6,6]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5],ymm0[6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
; AVX2-SLOW-NEXT: vmovaps 704(%rdi), %ymm1
; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovaps 736(%rdi), %ymm2
; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5],ymm2[6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpermps %ymm1, %ymm15, %ymm1
+; AVX2-SLOW-NEXT: vpermps %ymm1, %ymm9, %ymm1
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovaps 1056(%rdi), %ymm1
@@ -7065,16 +7017,16 @@ define void @load_i32_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpermps %ymm0, %ymm9, %ymm0
+; AVX2-SLOW-NEXT: vpermps %ymm0, %ymm4, %ymm0
; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm1 = ymm2[0,2,2,2,4,6,6,6]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5],ymm0[6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
; AVX2-SLOW-NEXT: vmovaps 1088(%rdi), %ymm1
-; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovups %ymm1, (%rsp) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovaps 1120(%rdi), %ymm2
; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5],ymm2[6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpermps %ymm1, %ymm15, %ymm1
+; AVX2-SLOW-NEXT: vpermps %ymm1, %ymm9, %ymm1
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovaps 1440(%rdi), %ymm1
@@ -7090,16 +7042,16 @@ define void @load_i32_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpermps %ymm0, %ymm9, %ymm0
+; AVX2-SLOW-NEXT: vpermps %ymm0, %ymm4, %ymm0
; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm1 = ymm2[0,2,2,2,4,6,6,6]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5],ymm0[6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
; AVX2-SLOW-NEXT: vmovaps 1472(%rdi), %ymm1
; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovaps 1504(%rdi), %ymm2
; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5],ymm2[6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpermps %ymm1, %ymm15, %ymm1
+; AVX2-SLOW-NEXT: vpermps %ymm1, %ymm9, %ymm1
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovaps 96(%rdi), %ymm1
@@ -7107,21 +7059,21 @@ define void @load_i32_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-SLOW-NEXT: vmovaps 64(%rdi), %ymm0
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[0,1],ymm1[0,1]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm12 = ymm0[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm13 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-SLOW-NEXT: vmovaps (%rdi), %ymm0
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovaps 32(%rdi), %ymm1
; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm11 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
-; AVX2-SLOW-NEXT: vpermps %ymm11, %ymm9, %ymm0
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm1 = ymm12[0,2,2,2,4,6,6,6]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5],ymm0[6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm12 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm12, %ymm4, %ymm0
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm1 = ymm13[0,2,2,2,4,6,6,6]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
; AVX2-SLOW-NEXT: vmovaps 128(%rdi), %ymm1
; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovaps 160(%rdi), %ymm2
; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm10 = ymm2[0,1,2,3],ymm1[4,5],ymm2[6,7]
-; AVX2-SLOW-NEXT: vpermps %ymm10, %ymm15, %ymm1
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm11 = ymm2[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm11, %ymm9, %ymm1
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovaps 480(%rdi), %ymm1
@@ -7129,21 +7081,21 @@ define void @load_i32_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-SLOW-NEXT: vmovaps 448(%rdi), %ymm0
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[0,1],ymm1[0,1]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm8 = ymm0[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm10 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-SLOW-NEXT: vmovaps 416(%rdi), %ymm0
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovaps 384(%rdi), %ymm1
; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm6 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
-; AVX2-SLOW-NEXT: vpermps %ymm6, %ymm9, %ymm0
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm1 = ymm8[0,2,2,2,4,6,6,6]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5],ymm0[6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm7 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm7, %ymm4, %ymm0
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm1 = ymm10[0,2,2,2,4,6,6,6]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
; AVX2-SLOW-NEXT: vmovaps 512(%rdi), %ymm1
; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovaps 544(%rdi), %ymm2
; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm7 = ymm2[0,1,2,3],ymm1[4,5],ymm2[6,7]
-; AVX2-SLOW-NEXT: vpermps %ymm7, %ymm15, %ymm1
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm8 = ymm2[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm8, %ymm9, %ymm1
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovaps 864(%rdi), %ymm1
@@ -7151,101 +7103,101 @@ define void @load_i32_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-SLOW-NEXT: vmovaps 832(%rdi), %ymm0
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[0,1],ymm1[0,1]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm0[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm6 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-SLOW-NEXT: vmovaps 800(%rdi), %ymm0
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovaps 768(%rdi), %ymm1
; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
-; AVX2-SLOW-NEXT: vpermps %ymm3, %ymm9, %ymm0
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm1 = ymm5[0,2,2,2,4,6,6,6]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5],ymm0[6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm3, %ymm4, %ymm0
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm1 = ymm6[0,2,2,2,4,6,6,6]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
; AVX2-SLOW-NEXT: vmovaps 896(%rdi), %ymm1
; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovaps 928(%rdi), %ymm2
; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm2[0,1,2,3],ymm1[4,5],ymm2[6,7]
-; AVX2-SLOW-NEXT: vpermps %ymm4, %ymm15, %ymm1
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm2[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm5, %ymm9, %ymm1
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovaps 1184(%rdi), %ymm0
-; AVX2-SLOW-NEXT: vmovups %ymm0, (%rsp) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovaps 1152(%rdi), %ymm1
; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
-; AVX2-SLOW-NEXT: vpermps %ymm2, %ymm9, %ymm0
-; AVX2-SLOW-NEXT: vmovaps 1248(%rdi), %ymm9
-; AVX2-SLOW-NEXT: vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpermps %ymm2, %ymm4, %ymm0
+; AVX2-SLOW-NEXT: vmovaps 1248(%rdi), %ymm4
+; AVX2-SLOW-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovaps 1216(%rdi), %ymm1
; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm1[0,1],ymm9[0,1]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm9 = ymm1[0,1,2,3,4,5],ymm9[6,7]
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm1 = ymm9[0,2,2,2,4,6,6,6]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5],ymm0[6,7]
+; AVX2-SLOW-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm1[0,1],ymm4[0,1]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm1[0,1,2,3,4,5],ymm4[6,7]
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm1 = ymm4[0,2,2,2,4,6,6,6]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
; AVX2-SLOW-NEXT: vmovaps 1280(%rdi), %ymm1
; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovaps 1312(%rdi), %ymm13
-; AVX2-SLOW-NEXT: vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm13[0,1,2,3],ymm1[4,5],ymm13[6,7]
-; AVX2-SLOW-NEXT: vpermps %ymm1, %ymm15, %ymm15
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm15[6,7]
+; AVX2-SLOW-NEXT: vmovaps 1312(%rdi), %ymm14
+; AVX2-SLOW-NEXT: vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm14[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm1, %ymm9, %ymm9
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm9[6,7]
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovaps {{.*#+}} xmm15 = <1,7,5,u>
-; AVX2-SLOW-NEXT: vpermps %ymm14, %ymm15, %ymm0
-; AVX2-SLOW-NEXT: vpermilps $237, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm13 = mem[1,3,2,3,5,7,6,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm13 = ymm0[0,1,2],ymm13[3,4,5],ymm0[6,7]
-; AVX2-SLOW-NEXT: vbroadcastsd {{.*#+}} ymm0 = [12884901893,12884901893,12884901893,12884901893]
-; AVX2-SLOW-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm14 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3,4,5],ymm14[6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm13 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: vpermilps $237, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm14 = mem[1,3,2,3,5,7,6,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0,1,2],ymm14[3,4,5],ymm13[6,7]
-; AVX2-SLOW-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm14 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3,4,5],ymm14[6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm13 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: vpermilps $237, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm14 = mem[1,3,2,3,5,7,6,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0,1,2],ymm14[3,4,5],ymm13[6,7]
-; AVX2-SLOW-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm14 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3,4,5],ymm14[6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm13 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: vmovaps {{.*#+}} xmm9 = <1,7,5,u>
+; AVX2-SLOW-NEXT: vpermps %ymm15, %ymm9, %ymm0
; AVX2-SLOW-NEXT: vpermilps $237, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: # ymm14 = mem[1,3,2,3,5,7,6,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0,1,2],ymm14[3,4,5],ymm13[6,7]
-; AVX2-SLOW-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm14 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3,4,5],ymm14[6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpermps %ymm11, %ymm15, %ymm11
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm12 = ymm12[1,3,2,3,5,7,6,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm11 = ymm11[0,1,2],ymm12[3,4,5],ymm11[6,7]
-; AVX2-SLOW-NEXT: vpermps %ymm10, %ymm0, %ymm10
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm10 = ymm11[0,1,2,3,4,5],ymm10[6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpermps %ymm6, %ymm15, %ymm6
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm8 = ymm8[1,3,2,3,5,7,6,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2],ymm8[3,4,5],ymm6[6,7]
-; AVX2-SLOW-NEXT: vpermps %ymm7, %ymm0, %ymm7
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5],ymm7[6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpermps %ymm3, %ymm15, %ymm3
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm5 = ymm5[1,3,2,3,5,7,6,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2],ymm5[3,4,5],ymm3[6,7]
-; AVX2-SLOW-NEXT: vpermps %ymm4, %ymm0, %ymm4
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm4[6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm14 = ymm0[0,1,2],ymm14[3,4,5,6,7]
+; AVX2-SLOW-NEXT: vbroadcastsd {{.*#+}} ymm0 = [12884901893,12884901893,12884901893,12884901893]
+; AVX2-SLOW-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4,5],ymm15[6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm14 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: vpermilps $237, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm15 = mem[1,3,2,3,5,7,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1,2],ymm15[3,4,5,6,7]
+; AVX2-SLOW-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4,5],ymm15[6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm14 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: vpermilps $237, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm15 = mem[1,3,2,3,5,7,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1,2],ymm15[3,4,5,6,7]
+; AVX2-SLOW-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4,5],ymm15[6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm14 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: vpermilps $237, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm15 = mem[1,3,2,3,5,7,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1,2],ymm15[3,4,5,6,7]
+; AVX2-SLOW-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4,5],ymm15[6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpermps %ymm12, %ymm9, %ymm12
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm13 = ymm13[1,3,2,3,5,7,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm12 = ymm12[0,1,2],ymm13[3,4,5,6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm11, %ymm0, %ymm11
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm11 = ymm12[0,1,2,3,4,5],ymm11[6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpermps %ymm7, %ymm9, %ymm7
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm10 = ymm10[1,3,2,3,5,7,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2],ymm10[3,4,5,6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm8, %ymm0, %ymm8
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5],ymm8[6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpermps %ymm3, %ymm9, %ymm3
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm6 = ymm6[1,3,2,3,5,7,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2],ymm6[3,4,5,6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm5, %ymm0, %ymm5
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm5[6,7]
; AVX2-SLOW-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpermps %ymm2, %ymm15, %ymm2
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm3 = ymm9[1,3,2,3,5,7,6,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm3[3,4,5],ymm2[6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm2, %ymm9, %ymm2
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm3 = ymm4[1,3,2,3,5,7,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm3[3,4,5,6,7]
; AVX2-SLOW-NEXT: vpermps %ymm1, %ymm0, %ymm0
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3,4,5],ymm0[6,7]
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm0 = mem[2,3,2,3,6,7,6,7]
+; AVX2-SLOW-NEXT: vmovaps 192(%rdi), %xmm6
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} xmm0 = xmm6[2,3,2,3]
; AVX2-SLOW-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: # ymm1 = mem[0,0,2,3,4,4,6,7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
@@ -7255,27 +7207,28 @@ define void @load_i32_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm1 = ymm1[2,0,2,3,6,4,6,7]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
-; AVX2-SLOW-NEXT: vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm1 = mem[0,0,0,0,4,4,4,4]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm1 = ymm5[0,0,0,0,4,4,4,4]
; AVX2-SLOW-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: # ymm2 = mem[0,0,2,3,4,4,6,7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm0 = mem[2,3,2,3,6,7,6,7]
+; AVX2-SLOW-NEXT: vmovaps 576(%rdi), %xmm0
+; AVX2-SLOW-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[2,3,2,3]
; AVX2-SLOW-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: # ymm1 = mem[0,0,2,3,4,4,6,7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps $204, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm1 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm1 = ymm9[0,1],mem[2,3],ymm9[4,5],mem[6,7]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm13[0,1],ymm4[2,3],ymm13[4,5],ymm4[6,7]
; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm1 = ymm1[2,0,2,3,6,4,6,7]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
; AVX2-SLOW-NEXT: vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: # ymm1 = mem[0,0,0,0,4,4,4,4]
; AVX2-SLOW-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
@@ -7284,88 +7237,50 @@ define void @load_i32_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm0 = ymm8[2,3,2,3,6,7,6,7]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm1 = ymm6[0,0,2,3,4,4,6,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm14[0,1],ymm15[2,3],ymm14[4,5],ymm15[6,7]
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm1 = ymm1[2,0,2,3,6,4,6,7]
-; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
-; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm1 = ymm4[0,0,0,0,4,4,4,4]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm2 = ymm5[0,0,2,3,4,4,6,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
-; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm0 = ymm13[2,3,2,3,6,7,6,7]
+; AVX2-SLOW-NEXT: vmovaps 960(%rdi), %xmm0
+; AVX2-SLOW-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[2,3,2,3]
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm1 = ymm12[0,0,2,3,4,4,6,7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm1 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm1 = mem[0,1],ymm7[2,3],mem[4,5],ymm7[6,7]
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm1 = ymm1[2,0,2,3,6,4,6,7]
-; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
-; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm1 = ymm11[0,0,0,0,4,4,4,4]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm2 = ymm10[0,0,2,3,4,4,6,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
-; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm0 = mem[2,3,2,3,6,7,6,7]
-; AVX2-SLOW-NEXT: vpermilps $224, (%rsp), %ymm1 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm1 = mem[0,0,2,3,4,4,6,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-SLOW-NEXT: vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: # ymm1 = mem[0,1],ymm1[2,3],mem[4,5],ymm1[6,7]
; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm1 = ymm1[2,0,2,3,6,4,6,7]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
-; AVX2-SLOW-NEXT: vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm1 = mem[0,0,0,0,4,4,4,4]
-; AVX2-SLOW-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm1 = ymm3[0,0,0,0,4,4,4,4]
+; AVX2-SLOW-NEXT: vpermilps $224, (%rsp), %ymm2 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: # ymm2 = mem[0,0,2,3,4,4,6,7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm0 = mem[2,3,2,3,6,7,6,7]
+; AVX2-SLOW-NEXT: vmovaps 1344(%rdi), %xmm0
+; AVX2-SLOW-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[2,3,2,3]
; AVX2-SLOW-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: # ymm1 = mem[0,0,2,3,4,4,6,7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm1 = mem[0,1],ymm1[2,3],mem[4,5],ymm1[6,7]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm11[0,1],ymm14[2,3],ymm11[4,5],ymm14[6,7]
; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm1 = ymm1[2,0,2,3,6,4,6,7]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
; AVX2-SLOW-NEXT: vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: # ymm1 = mem[0,0,0,0,4,4,4,4]
-; AVX2-SLOW-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm2 = mem[0,0,2,3,4,4,6,7]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm2 = ymm15[0,0,2,3,4,4,6,7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm0 = mem[2,3,2,3,6,7,6,7]
+; AVX2-SLOW-NEXT: vmovaps 1152(%rdi), %xmm2
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} xmm0 = xmm2[2,3,2,3]
; AVX2-SLOW-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: # ymm1 = mem[0,0,2,3,4,4,6,7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
@@ -7375,17 +7290,18 @@ define void @load_i32_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm1 = ymm1[2,0,2,3,6,4,6,7]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
; AVX2-SLOW-NEXT: vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: # ymm1 = mem[0,0,0,0,4,4,4,4]
-; AVX2-SLOW-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm2 = mem[0,0,2,3,4,4,6,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
+; AVX2-SLOW-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm7 = mem[0,0,2,3,4,4,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm7[0,1,2],ymm1[3],ymm7[4,5,6],ymm1[7]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm0 = mem[2,3,2,3,6,7,6,7]
+; AVX2-SLOW-NEXT: vmovaps (%rdi), %xmm0
+; AVX2-SLOW-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[2,3,2,3]
; AVX2-SLOW-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: # ymm1 = mem[0,0,2,3,4,4,6,7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
@@ -7395,155 +7311,195 @@ define void @load_i32_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm1 = ymm1[2,0,2,3,6,4,6,7]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
; AVX2-SLOW-NEXT: vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: # ymm1 = mem[0,0,0,0,4,4,4,4]
-; AVX2-SLOW-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm2 = mem[0,0,2,3,4,4,6,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
-; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm0 = mem[3,3,3,3,7,7,7,7]
-; AVX2-SLOW-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm1 = mem[0,1,3,3,4,5,7,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
-; AVX2-SLOW-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm1 = mem[3,3,3,3,7,7,7,7]
-; AVX2-SLOW-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm1 = ymm1[0],mem[1],ymm1[2,3,4],mem[5],ymm1[6,7]
-; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
-; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
-; AVX2-SLOW-NEXT: vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm1 = mem[0,1,0,1,4,5,4,5]
-; AVX2-SLOW-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm2 = mem[0,1,3,3,4,5,7,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
-; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm0 = mem[3,3,3,3,7,7,7,7]
-; AVX2-SLOW-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm1 = mem[0,1,3,3,4,5,7,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
-; AVX2-SLOW-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm1 = mem[3,3,3,3,7,7,7,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm9[1],ymm1[2,3,4],ymm9[5],ymm1[6,7]
-; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
-; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
-; AVX2-SLOW-NEXT: vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm1 = mem[0,1,0,1,4,5,4,5]
-; AVX2-SLOW-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm2 = mem[0,1,3,3,4,5,7,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
-; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm0 = ymm8[3,3,3,3,7,7,7,7]
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm1 = ymm6[0,1,3,3,4,5,7,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm1 = ymm15[3,3,3,3,7,7,7,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm14[1],ymm1[2,3,4],ymm14[5],ymm1[6,7]
-; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
-; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm1 = ymm4[0,1,0,1,4,5,4,5]
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm2 = ymm5[0,1,3,3,4,5,7,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
-; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm0 = ymm13[3,3,3,3,7,7,7,7]
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm1 = ymm12[0,1,3,3,4,5,7,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm1 = ymm7[3,3,3,3,7,7,7,7]
-; AVX2-SLOW-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm1 = ymm1[0],mem[1],ymm1[2,3,4],mem[5],ymm1[6,7]
-; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
-; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm1 = ymm11[0,1,0,1,4,5,4,5]
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm2 = ymm10[0,1,3,3,4,5,7,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
+; AVX2-SLOW-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm8 = mem[0,0,2,3,4,4,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm8[0,1,2],ymm1[3],ymm8[4,5,6],ymm1[7]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm0 = mem[3,3,3,3,7,7,7,7]
-; AVX2-SLOW-NEXT: vpermilps $244, (%rsp), %ymm1 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm1 = mem[0,1,3,3,4,5,7,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
-; AVX2-SLOW-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm1 = mem[3,3,3,3,7,7,7,7]
-; AVX2-SLOW-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm1 = ymm1[0],mem[1],ymm1[2,3,4],mem[5],ymm1[6,7]
-; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
+; AVX2-SLOW-NEXT: vmovaps 384(%rdi), %xmm0
+; AVX2-SLOW-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[2,3,2,3]
+; AVX2-SLOW-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm8 = mem[0,0,2,3,4,4,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm8[1,2,3],ymm0[4],ymm8[5,6,7]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-SLOW-NEXT: vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm8 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm8 = mem[0,1],ymm1[2,3],mem[4,5],ymm1[6,7]
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm8 = ymm8[2,0,2,3,6,4,6,7]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm8 = ymm8[0,2,0,3]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
-; AVX2-SLOW-NEXT: vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm1 = mem[0,1,0,1,4,5,4,5]
-; AVX2-SLOW-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm2 = mem[0,1,3,3,4,5,7,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
-; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm8[3,4,5,6,7]
+; AVX2-SLOW-NEXT: vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm8 = mem[0,0,0,0,4,4,4,4]
+; AVX2-SLOW-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm9 = mem[0,0,2,3,4,4,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm8 = ymm9[0,1,2],ymm8[3],ymm9[4,5,6],ymm8[7]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm8 = ymm8[0,1,0,3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm8[5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm0 = ymm12[3,3,3,3,7,7,7,7]
+; AVX2-SLOW-NEXT: vmovaps 768(%rdi), %xmm0
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} xmm8 = xmm0[2,3,2,3]
+; AVX2-SLOW-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm9 = mem[0,0,2,3,4,4,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0],ymm9[1,2,3],ymm8[4],ymm9[5,6,7]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-SLOW-NEXT: vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm9 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm9 = mem[0,1],ymm1[2,3],mem[4,5],ymm1[6,7]
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm9 = ymm9[2,0,2,3,6,4,6,7]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm9 = ymm9[0,2,0,3]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm8 = ymm8[0,3,2,3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2],ymm9[3,4,5,6,7]
+; AVX2-SLOW-NEXT: vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm9 = mem[0,0,0,0,4,4,4,4]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm10 = ymm7[0,0,2,3,4,4,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm9 = ymm10[0,1,2],ymm9[3],ymm10[4,5,6],ymm9[7]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm9 = ymm9[0,1,0,3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4],ymm9[5,6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} xmm6 = xmm6[3,3,3,3]
+; AVX2-SLOW-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm8 = mem[0,1,3,3,4,5,7,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0],ymm8[1,2,3],ymm6[4],ymm8[5,6,7]
+; AVX2-SLOW-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm8 = mem[3,3,3,3,7,7,7,7]
+; AVX2-SLOW-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm8 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm8 = ymm8[0],mem[1],ymm8[2,3,4],mem[5],ymm8[6,7]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm8 = ymm8[0,2,0,3]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm6 = ymm6[0,3,2,3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2],ymm8[3,4,5,6,7]
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm8 = ymm5[0,1,0,1,4,5,4,5]
+; AVX2-SLOW-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm9 = mem[0,1,3,3,4,5,7,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm8 = ymm9[0,1,2],ymm8[3],ymm9[4,5,6],ymm8[7]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm8 = ymm8[0,1,0,3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm6[0,1,2,3,4],ymm8[5,6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
+; AVX2-SLOW-NEXT: # xmm5 = mem[3,3,3,3]
+; AVX2-SLOW-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm6 = mem[0,1,3,3,4,5,7,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0],ymm6[1,2,3],ymm5[4],ymm6[5,6,7]
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm6 = ymm4[3,3,3,3,7,7,7,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0],ymm13[1],ymm6[2,3,4],ymm13[5],ymm6[6,7]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm6 = ymm6[0,2,0,3]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm5 = ymm5[0,3,2,3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2],ymm6[3,4,5,6,7]
+; AVX2-SLOW-NEXT: vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm6 = mem[0,1,0,1,4,5,4,5]
+; AVX2-SLOW-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm8 = mem[0,1,3,3,4,5,7,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm6 = ymm8[0,1,2],ymm6[3],ymm8[4,5,6],ymm6[7]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm6 = ymm6[0,1,0,3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm5[0,1,2,3,4],ymm6[5,6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
+; AVX2-SLOW-NEXT: # xmm4 = mem[3,3,3,3]
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm5 = ymm12[0,1,3,3,4,5,7,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0],ymm5[1,2,3],ymm4[4],ymm5[5,6,7]
+; AVX2-SLOW-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm5 = mem[3,3,3,3,7,7,7,7]
+; AVX2-SLOW-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm5 = ymm5[0],mem[1],ymm5[2,3,4],mem[5],ymm5[6,7]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm5 = ymm5[0,2,0,3]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm4 = ymm4[0,3,2,3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2],ymm5[3,4,5,6,7]
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm5 = ymm3[0,1,0,1,4,5,4,5]
+; AVX2-SLOW-NEXT: vpermilps $244, (%rsp), %ymm6 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm6 = mem[0,1,3,3,4,5,7,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm6[0,1,2],ymm5[3],ymm6[4,5,6],ymm5[7]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm5 = ymm5[0,1,0,3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm4[0,1,2,3,4],ymm5[5,6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; AVX2-SLOW-NEXT: # xmm3 = mem[3,3,3,3]
+; AVX2-SLOW-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm4 = mem[0,1,3,3,4,5,7,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0],ymm4[1,2,3],ymm3[4],ymm4[5,6,7]
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm4 = ymm14[3,3,3,3,7,7,7,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0],ymm11[1],ymm4[2,3,4],ymm11[5],ymm4[6,7]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm4 = ymm4[0,2,0,3]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm3 = ymm3[0,3,2,3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2],ymm4[3,4,5,6,7]
+; AVX2-SLOW-NEXT: vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm4 = mem[0,1,0,1,4,5,4,5]
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm5 = ymm15[0,1,3,3,4,5,7,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm5[0,1,2],ymm4[3],ymm5[4,5,6],ymm4[7]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm4 = ymm4[0,1,0,3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4],ymm4[5,6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} xmm2 = xmm2[3,3,3,3]
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm1 = ymm11[0,1,3,3,4,5,7,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
-; AVX2-SLOW-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm1 = mem[3,3,3,3,7,7,7,7]
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm3 = ymm11[0,1,3,3,4,5,7,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0],ymm3[1,2,3],ymm2[4],ymm3[5,6,7]
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm13[1],ymm1[2,3,4],ymm13[5],ymm1[6,7]
-; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm3 = ymm13[3,3,3,3,7,7,7,7]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0],ymm12[1],ymm3[2,3,4],ymm12[5],ymm3[6,7]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm3 = ymm3[0,2,0,3]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,3,2,3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm3[3,4,5,6,7]
+; AVX2-SLOW-NEXT: vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm3 = mem[0,1,0,1,4,5,4,5]
+; AVX2-SLOW-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm4 = mem[0,1,3,3,4,5,7,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3],ymm4[4,5,6],ymm3[7]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm3 = ymm3[0,1,0,3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm2 = ymm9[0,1,3,3,4,5,7,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm2[1,2,3],ymm0[4],ymm2[5,6,7]
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm2 = ymm1[3,3,3,3,7,7,7,7]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0],ymm14[1],ymm2[2,3,4],ymm14[5],ymm2[6,7]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,2,0,3]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
-; AVX2-SLOW-NEXT: vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm1 = mem[0,1,0,1,4,5,4,5]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm2 = ymm15[0,1,3,3,4,5,7,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
-; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm2[3,4,5,6,7]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm2 = ymm10[0,1,0,1,4,5,4,5]
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm3 = ymm7[0,1,3,3,4,5,7,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3],ymm3[4,5,6],ymm2[7]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,1,0,3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm0 = ymm6[3,3,3,3,7,7,7,7]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm1 = ymm5[0,1,3,3,4,5,7,7]
+; AVX2-SLOW-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; AVX2-SLOW-NEXT: # xmm0 = mem[3,3,3,3]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm1 = ymm3[0,1,3,3,4,5,7,7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm1 = ymm8[3,3,3,3,7,7,7,7]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm7[1],ymm1[2,3,4],ymm7[5],ymm1[6,7]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm1 = ymm5[3,3,3,3,7,7,7,7]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm4[1],ymm1[2,3,4],ymm4[5],ymm1[6,7]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm1 = ymm10[0,1,0,1,4,5,4,5]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm2 = ymm9[0,1,3,3,4,5,7,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm1 = ymm7[0,1,0,1,4,5,4,5]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm2 = ymm6[0,1,3,3,4,5,7,7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm0 = mem[3,3,3,3,7,7,7,7]
+; AVX2-SLOW-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; AVX2-SLOW-NEXT: # xmm0 = mem[3,3,3,3]
; AVX2-SLOW-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: # ymm1 = mem[0,1,3,3,4,5,7,7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
; AVX2-SLOW-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: # ymm1 = mem[3,3,3,3,7,7,7,7]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm4[1],ymm1[2,3,4],ymm4[5],ymm1[6,7]
+; AVX2-SLOW-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm1 = ymm1[0],mem[1],ymm1[2,3,4],mem[5],ymm1[6,7]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
; AVX2-SLOW-NEXT: vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: # ymm1 = mem[0,1,0,1,4,5,4,5]
; AVX2-SLOW-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
@@ -7553,146 +7509,147 @@ define void @load_i32_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm2 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm2 = ymm0[0,1,2,3],mem[4,5],ymm0[6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm1 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm1 = mem[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-SLOW-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm1 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm1 = ymm0[0,1,2,3],mem[4,5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-SLOW-NEXT: vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm2 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm2 = mem[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovaps 272(%rdi), %xmm0
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3],ymm2[4,5,6,7]
; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,1,0,2,4,5,4,6]
-; AVX2-SLOW-NEXT: vmovaps {{.*#+}} xmm1 = <4,2,u,u>
-; AVX2-SLOW-NEXT: vpermps %ymm2, %ymm1, %ymm2
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm0[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT: vmovaps {{.*#+}} xmm8 = <4,2,u,u>
+; AVX2-SLOW-NEXT: vpermps %ymm1, %ymm8, %ymm1
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm0[2,3,4,5,6,7]
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm14 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm14 = mem[0,1],ymm0[2,3],mem[4,5,6,7]
+; AVX2-SLOW-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm15 = mem[0,1],ymm0[2,3],mem[4,5,6,7]
; AVX2-SLOW-NEXT: vbroadcastf128 {{.*#+}} ymm0 = [0,2,0,6,0,2,0,6]
; AVX2-SLOW-NEXT: # ymm0 = mem[0,1,0,1]
-; AVX2-SLOW-NEXT: vpermps %ymm14, %ymm0, %ymm3
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm15, %ymm0, %ymm2
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm2 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm2 = ymm3[0,1,2,3],mem[4,5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm5[0,1,2,3],ymm6[4,5],ymm5[6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm4[0,1,2,3],ymm5[4,5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm7[0,1,2,3],ymm8[4,5,6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovaps 464(%rdi), %xmm2
+; AVX2-SLOW-NEXT: vmovaps 464(%rdi), %xmm1
+; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm3[0,1],ymm1[2,3],ymm3[4,5,6,7]
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
+; AVX2-SLOW-NEXT: vpermps %ymm2, %ymm8, %ymm2
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm7[0,1],ymm6[2,3],ymm7[4,5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm5[0,1],ymm2[2,3],ymm5[4,5,6,7]
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm2 = ymm2[0,1,0,2,4,5,4,6]
-; AVX2-SLOW-NEXT: vpermps %ymm3, %ymm1, %ymm3
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3,4,5,6,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm10[0,1],ymm9[2,3],ymm10[4,5,6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpermps %ymm3, %ymm0, %ymm3
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm2, %ymm0, %ymm2
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-SLOW-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm2 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm2 = ymm1[0,1,2,3],mem[4,5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm3 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm3 = mem[0,1,2,3],ymm2[4,5],mem[6,7]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-SLOW-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm3 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm3 = ymm1[0,1,2,3],mem[4,5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovaps 656(%rdi), %xmm1
+; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm3[0,1],ymm1[2,3],ymm3[4,5,6,7]
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
+; AVX2-SLOW-NEXT: vpermps %ymm2, %ymm8, %ymm2
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5,6,7]
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm5 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm5 = ymm2[0,1,2,3],mem[4,5,6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovaps 656(%rdi), %xmm2
+; AVX2-SLOW-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm2 = mem[0,1],ymm2[2,3],mem[4,5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm5[0,1],ymm2[2,3],ymm5[4,5,6,7]
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm2 = ymm2[0,1,0,2,4,5,4,6]
-; AVX2-SLOW-NEXT: vpermps %ymm3, %ymm1, %ymm3
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3,4,5,6,7]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm3 = mem[0,1],ymm3[2,3],mem[4,5,6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpermps %ymm3, %ymm0, %ymm3
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm2, %ymm0, %ymm2
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm2 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm2 = ymm9[0,1,2,3],mem[4,5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm11[0,1,2,3],ymm12[4,5],ymm11[6,7]
+; AVX2-SLOW-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm3 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm3 = ymm14[0,1,2,3],mem[4,5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm5 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm5 = ymm13[0,1,2,3],mem[4,5,6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovaps 848(%rdi), %xmm2
+; AVX2-SLOW-NEXT: vmovaps 848(%rdi), %xmm1
+; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm3[0,1],ymm1[2,3],ymm3[4,5,6,7]
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
+; AVX2-SLOW-NEXT: vpermps %ymm2, %ymm8, %ymm2
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT: vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm2 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm2 = ymm10[0,1],mem[2,3],ymm10[4,5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm5[0,1],ymm2[2,3],ymm5[4,5,6,7]
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm2 = ymm2[0,1,0,2,4,5,4,6]
-; AVX2-SLOW-NEXT: vpermps %ymm3, %ymm1, %ymm3
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3,4,5,6,7]
-; AVX2-SLOW-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm3 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm3 = mem[0,1],ymm15[2,3],mem[4,5,6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpermps %ymm3, %ymm0, %ymm3
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm2, %ymm0, %ymm2
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-SLOW-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm2 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm2 = ymm1[0,1,2,3],mem[4,5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm3 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm3 = ymm2[0,1,2,3],mem[4,5],ymm2[6,7]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-SLOW-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm3 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm3 = ymm1[0,1,2,3],mem[4,5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm5 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm5 = ymm2[0,1,2,3],mem[4,5,6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovaps 1040(%rdi), %xmm2
-; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm5[0,1],ymm2[2,3],ymm5[4,5,6,7]
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm2 = ymm2[0,1,0,2,4,5,4,6]
-; AVX2-SLOW-NEXT: vpermps %ymm3, %ymm1, %ymm3
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3,4,5,6,7]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm13 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm13 = mem[0,1],ymm3[2,3],mem[4,5,6,7]
-; AVX2-SLOW-NEXT: vpermps %ymm13, %ymm0, %ymm3
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovaps 1040(%rdi), %xmm1
+; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm3[0,1],ymm1[2,3],ymm3[4,5,6,7]
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
+; AVX2-SLOW-NEXT: vpermps %ymm2, %ymm8, %ymm2
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5,6,7]
; AVX2-SLOW-NEXT: vmovups (%rsp), %ymm2 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm3 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm3 = ymm2[0,1,2,3],mem[4,5],ymm2[6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm3, (%rsp) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm10 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm10 = ymm2[0,1,2,3],mem[4,5,6,7]
-; AVX2-SLOW-NEXT: vmovaps 1232(%rdi), %xmm12
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm10[0,1],ymm12[2,3],ymm10[4,5,6,7]
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm2 = ymm2[0,1,0,2,4,5,4,6]
-; AVX2-SLOW-NEXT: vpermps %ymm3, %ymm1, %ymm3
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3,4,5,6,7]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm9 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm9 = mem[0,1],ymm3[2,3],mem[4,5,6,7]
-; AVX2-SLOW-NEXT: vpermps %ymm9, %ymm0, %ymm3
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7]
+; AVX2-SLOW-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm14 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm14 = mem[0,1],ymm2[2,3],mem[4,5,6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm14, %ymm0, %ymm2
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm2 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm2 = ymm11[0,1,2,3],mem[4,5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm11 = ymm12[0,1,2,3],ymm13[4,5,6,7]
+; AVX2-SLOW-NEXT: vmovaps 1232(%rdi), %xmm1
+; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm11[0,1],ymm1[2,3],ymm11[4,5,6,7]
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
+; AVX2-SLOW-NEXT: vpermps %ymm2, %ymm8, %ymm2
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5,6,7]
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm11 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm11 = ymm2[0,1,2,3],mem[4,5],ymm2[6,7]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm7 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm7 = ymm2[0,1,2,3],mem[4,5,6,7]
-; AVX2-SLOW-NEXT: vmovaps 1424(%rdi), %xmm8
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm7[0,1],ymm8[2,3],ymm7[4,5,6,7]
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm2 = ymm2[0,1,0,2,4,5,4,6]
-; AVX2-SLOW-NEXT: vpermps %ymm11, %ymm1, %ymm3
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3,4,5,6,7]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm6 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm6 = mem[0,1],ymm3[2,3],mem[4,5,6,7]
-; AVX2-SLOW-NEXT: vpermps %ymm6, %ymm0, %ymm3
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm5 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm5 = ymm4[0,1,2,3],mem[4,5,6,7]
+; AVX2-SLOW-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm10 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm10 = mem[0,1],ymm2[2,3],mem[4,5,6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm10, %ymm0, %ymm2
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-SLOW-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm13 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm13 = ymm1[0,1,2,3],mem[4,5,6,7]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-SLOW-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm9 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm9 = ymm1[0,1,2,3],mem[4,5,6,7]
+; AVX2-SLOW-NEXT: vmovaps 1424(%rdi), %xmm12
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm9[0,1],ymm12[2,3],ymm9[4,5,6,7]
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
+; AVX2-SLOW-NEXT: vpermps %ymm13, %ymm8, %ymm2
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5,6,7]
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm4 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm4 = ymm2[0,1,2,3],mem[4,5],ymm2[6,7]
+; AVX2-SLOW-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm6 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm6 = mem[0,1],ymm2[2,3],mem[4,5,6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm6, %ymm0, %ymm2
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-SLOW-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm5 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm5 = ymm1[0,1,2,3],mem[4,5,6,7]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-SLOW-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm4 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm4 = ymm1[0,1,2,3],mem[4,5,6,7]
; AVX2-SLOW-NEXT: vmovaps 80(%rdi), %xmm2
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm5[0,1],ymm2[2,3],ymm5[4,5,6,7]
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm3 = ymm3[0,1,0,2,4,5,4,6]
-; AVX2-SLOW-NEXT: vpermps %ymm4, %ymm1, %ymm1
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm3[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm5[0,1],ymm2[2,3],ymm5[4,5,6,7]
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
+; AVX2-SLOW-NEXT: vpermps %ymm4, %ymm8, %ymm3
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm3[0,1],ymm1[2,3,4,5,6,7]
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX2-SLOW-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: # ymm3 = mem[0,1],ymm3[2,3],mem[4,5,6,7]
@@ -7704,57 +7661,58 @@ define void @load_i32_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-SLOW-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: # ymm0 = ymm0[0,1,2],mem[3],ymm0[4,5,6,7]
; AVX2-SLOW-NEXT: vmovaps {{.*#+}} xmm1 = <5,3,u,u>
-; AVX2-SLOW-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm15 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm15 = ymm15[0,1],ymm0[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm7 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1],ymm0[2,3,4,5,6,7]
; AVX2-SLOW-NEXT: vbroadcastf128 {{.*#+}} ymm0 = [0,3,1,7,0,3,1,7]
; AVX2-SLOW-NEXT: # ymm0 = mem[0,1,0,1]
-; AVX2-SLOW-NEXT: vpermps %ymm14, %ymm0, %ymm14
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm14 = ymm15[0,1,2,3,4],ymm14[5,6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm14 = mem[1,1,1,1,5,5,5,5]
-; AVX2-SLOW-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm14 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm14 = ymm14[0,1,2],mem[3],ymm14[4,5,6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm15, %ymm0, %ymm15
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4],ymm15[5,6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm7 = mem[1,1,1,1,5,5,5,5]
+; AVX2-SLOW-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm7 = ymm7[0,1,2],mem[3],ymm7[4,5,6,7]
; AVX2-SLOW-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm15 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm14 = ymm15[0,1],ymm14[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm7 = ymm15[0,1],ymm7[2,3,4,5,6,7]
; AVX2-SLOW-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4],ymm15[5,6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm14 = mem[1,1,1,1,5,5,5,5]
-; AVX2-SLOW-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm14 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm14 = ymm14[0,1,2],mem[3],ymm14[4,5,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4],ymm15[5,6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm7 = mem[1,1,1,1,5,5,5,5]
+; AVX2-SLOW-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm7 = ymm7[0,1,2],mem[3],ymm7[4,5,6,7]
; AVX2-SLOW-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm15 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm14 = ymm15[0,1],ymm14[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm7 = ymm15[0,1],ymm7[2,3,4,5,6,7]
; AVX2-SLOW-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4],ymm15[5,6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm14 = mem[1,1,1,1,5,5,5,5]
-; AVX2-SLOW-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm14 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm14 = ymm14[0,1,2],mem[3],ymm14[4,5,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4],ymm15[5,6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm7 = mem[1,1,1,1,5,5,5,5]
+; AVX2-SLOW-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm7 = ymm7[0,1,2],mem[3],ymm7[4,5,6,7]
; AVX2-SLOW-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm15 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm14 = ymm15[0,1],ymm14[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm7 = ymm15[0,1],ymm7[2,3,4,5,6,7]
; AVX2-SLOW-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4],ymm15[5,6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm14 = mem[1,1,1,1,5,5,5,5]
-; AVX2-SLOW-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm14 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm14 = ymm14[0,1,2],mem[3],ymm14[4,5,6,7]
-; AVX2-SLOW-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm15 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm14 = ymm15[0,1],ymm14[2,3,4,5,6,7]
-; AVX2-SLOW-NEXT: vpermps %ymm13, %ymm0, %ymm13
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm13 = ymm14[0,1,2,3,4],ymm13[5,6,7]
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm10 = ymm10[1,1,1,1,5,5,5,5]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm10 = ymm10[0,1,2],ymm12[3],ymm10[4,5,6,7]
-; AVX2-SLOW-NEXT: vpermps (%rsp), %ymm1, %ymm12 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm10 = ymm12[0,1],ymm10[2,3,4,5,6,7]
-; AVX2-SLOW-NEXT: vpermps %ymm9, %ymm0, %ymm9
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm9 = ymm10[0,1,2,3,4],ymm9[5,6,7]
-; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm7 = ymm7[1,1,1,1,5,5,5,5]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2],ymm8[3],ymm7[4,5,6,7]
-; AVX2-SLOW-NEXT: vpermps %ymm11, %ymm1, %ymm8
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4],ymm15[5,6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm7, (%rsp) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm15 = mem[1,1,1,1,5,5,5,5]
+; AVX2-SLOW-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm15 = ymm15[0,1,2],mem[3],ymm15[4,5,6,7]
+; AVX2-SLOW-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm7 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1],ymm15[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm14, %ymm0, %ymm14
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm14 = ymm7[0,1,2,3,4],ymm14[5,6,7]
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm7 = ymm11[1,1,1,1,5,5,5,5]
+; AVX2-SLOW-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm7 = ymm7[0,1,2],mem[3],ymm7[4,5,6,7]
+; AVX2-SLOW-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm11 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm7 = ymm11[0,1],ymm7[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm10, %ymm0, %ymm10
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm10 = ymm7[0,1,2,3,4],ymm10[5,6,7]
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm7 = ymm9[1,1,1,1,5,5,5,5]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2],ymm12[3],ymm7[4,5,6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm13, %ymm1, %ymm8
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm7 = ymm8[0,1],ymm7[2,3,4,5,6,7]
; AVX2-SLOW-NEXT: vpermps %ymm6, %ymm0, %ymm6
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm6 = ymm7[0,1,2,3,4],ymm6[5,6,7]
@@ -7846,9 +7804,9 @@ define void @load_i32_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-SLOW-NEXT: vmovaps %ymm0, (%r9)
; AVX2-SLOW-NEXT: movq {{[0-9]+}}(%rsp), %rax
; AVX2-SLOW-NEXT: vmovaps %ymm6, 224(%rax)
-; AVX2-SLOW-NEXT: vmovaps %ymm9, 192(%rax)
-; AVX2-SLOW-NEXT: vmovaps %ymm13, 160(%rax)
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovaps %ymm10, 192(%rax)
+; AVX2-SLOW-NEXT: vmovaps %ymm14, 160(%rax)
+; AVX2-SLOW-NEXT: vmovups (%rsp), %ymm0 # 32-byte Reload
; AVX2-SLOW-NEXT: vmovaps %ymm0, 128(%rax)
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-SLOW-NEXT: vmovaps %ymm0, 96(%rax)
@@ -7857,268 +7815,268 @@ define void @load_i32_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-SLOW-NEXT: vmovaps %ymm0, 32(%rax)
; AVX2-SLOW-NEXT: vmovaps %ymm1, (%rax)
-; AVX2-SLOW-NEXT: addq $2472, %rsp # imm = 0x9A8
+; AVX2-SLOW-NEXT: addq $2504, %rsp # imm = 0x9C8
; AVX2-SLOW-NEXT: vzeroupper
; AVX2-SLOW-NEXT: retq
;
; AVX2-FAST-LABEL: load_i32_stride6_vf64:
; AVX2-FAST: # %bb.0:
-; AVX2-FAST-NEXT: subq $2472, %rsp # imm = 0x9A8
-; AVX2-FAST-NEXT: vmovaps 672(%rdi), %ymm3
-; AVX2-FAST-NEXT: vmovaps 640(%rdi), %ymm4
-; AVX2-FAST-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: subq $2504, %rsp # imm = 0x9C8
+; AVX2-FAST-NEXT: vmovaps 672(%rdi), %ymm2
+; AVX2-FAST-NEXT: vmovaps 640(%rdi), %ymm3
+; AVX2-FAST-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovaps 608(%rdi), %ymm5
; AVX2-FAST-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps 320(%rdi), %ymm2
-; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps 352(%rdi), %ymm6
+; AVX2-FAST-NEXT: vmovaps 320(%rdi), %ymm6
; AVX2-FAST-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps 288(%rdi), %ymm7
+; AVX2-FAST-NEXT: vmovaps 352(%rdi), %ymm7
; AVX2-FAST-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps 256(%rdi), %ymm8
+; AVX2-FAST-NEXT: vmovaps 288(%rdi), %ymm8
; AVX2-FAST-NEXT: vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovaps 256(%rdi), %ymm9
+; AVX2-FAST-NEXT: vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovaps 224(%rdi), %ymm0
; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovaps 192(%rdi), %ymm1
; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps {{.*#+}} xmm10 = <0,6,4,u>
+; AVX2-FAST-NEXT: vmovaps {{.*#+}} xmm4 = <0,6,4,u>
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm15 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
-; AVX2-FAST-NEXT: vpermps %ymm15, %ymm10, %ymm0
-; AVX2-FAST-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm8[0,1],ymm7[0,1]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm7[6,7]
+; AVX2-FAST-NEXT: vpermps %ymm15, %ymm4, %ymm0
+; AVX2-FAST-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm9[0,1],ymm8[0,1]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm8[6,7]
; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm1 = ymm1[0,2,2,2,4,6,6,6]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm0[0,1,2],ymm1[3,4,5],ymm0[6,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm6[0,1,2,3],ymm2[4,5],ymm6[6,7]
-; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vbroadcastsd {{.*#+}} ymm0 = [8589934596,8589934596,8589934596,8589934596]
-; AVX2-FAST-NEXT: vpermps %ymm2, %ymm0, %ymm2
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm7[0,1,2,3],ymm6[4,5,6,7]
; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm4[0,1],ymm3[0,1]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm1[0,1,2,3,4,5],ymm3[6,7]
-; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps 576(%rdi), %ymm1
+; AVX2-FAST-NEXT: vbroadcastsd {{.*#+}} ymm10 = [8589934596,8589934596,8589934596,8589934596]
+; AVX2-FAST-NEXT: vpermps %ymm1, %ymm10, %ymm1
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovups %ymm2, (%rsp) # 32-byte Spill
+; AVX2-FAST-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm3[0,1],ymm2[0,1]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm0[0,1,2,3,4,5],ymm2[6,7]
+; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovaps 576(%rdi), %ymm0
+; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm5[4,5],ymm0[6,7]
+; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vpermps %ymm0, %ymm4, %ymm0
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm1 = ymm1[0,2,2,2,4,6,6,6]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
+; AVX2-FAST-NEXT: vmovaps 704(%rdi), %ymm1
; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm5[4,5],ymm1[6,7]
+; AVX2-FAST-NEXT: vmovaps 736(%rdi), %ymm2
+; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vpermps %ymm1, %ymm10, %ymm1
-; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm2 = ymm2[0,2,2,2,4,6,6,6]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5],ymm1[6,7]
-; AVX2-FAST-NEXT: vmovaps 704(%rdi), %ymm2
-; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps 736(%rdi), %ymm3
-; AVX2-FAST-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2,3],ymm2[4,5],ymm3[6,7]
-; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpermps %ymm2, %ymm0, %ymm2
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovaps 1056(%rdi), %ymm1
; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps 1056(%rdi), %ymm2
+; AVX2-FAST-NEXT: vmovaps 1024(%rdi), %ymm0
+; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[0,1],ymm1[0,1]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps 1024(%rdi), %ymm1
+; AVX2-FAST-NEXT: vmovaps 992(%rdi), %ymm0
+; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovaps 960(%rdi), %ymm1
; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm1[0,1],ymm2[0,1]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm1[0,1,2,3,4,5],ymm2[6,7]
-; AVX2-FAST-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps 992(%rdi), %ymm1
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
+; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vpermps %ymm0, %ymm4, %ymm0
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm1 = ymm2[0,2,2,2,4,6,6,6]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
+; AVX2-FAST-NEXT: vmovaps 1088(%rdi), %ymm1
; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps 960(%rdi), %ymm2
+; AVX2-FAST-NEXT: vmovaps 1120(%rdi), %ymm2
; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5],ymm2[6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vpermps %ymm1, %ymm10, %ymm1
-; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm2 = ymm3[0,2,2,2,4,6,6,6]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5],ymm1[6,7]
-; AVX2-FAST-NEXT: vmovaps 1088(%rdi), %ymm2
-; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps 1120(%rdi), %ymm3
-; AVX2-FAST-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2,3],ymm2[4,5],ymm3[6,7]
-; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpermps %ymm2, %ymm0, %ymm2
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovaps 1440(%rdi), %ymm1
; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps 1440(%rdi), %ymm2
+; AVX2-FAST-NEXT: vmovaps 1408(%rdi), %ymm0
+; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[0,1],ymm1[0,1]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps 1408(%rdi), %ymm1
+; AVX2-FAST-NEXT: vmovaps 1376(%rdi), %ymm0
+; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovaps 1344(%rdi), %ymm1
; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm1[0,1],ymm2[0,1]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm1[0,1,2,3,4,5],ymm2[6,7]
-; AVX2-FAST-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps 1376(%rdi), %ymm1
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
+; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vpermps %ymm0, %ymm4, %ymm0
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm1 = ymm2[0,2,2,2,4,6,6,6]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
+; AVX2-FAST-NEXT: vmovaps 1472(%rdi), %ymm1
; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps 1344(%rdi), %ymm2
+; AVX2-FAST-NEXT: vmovaps 1504(%rdi), %ymm2
; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5],ymm2[6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vpermps %ymm1, %ymm10, %ymm1
-; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm2 = ymm3[0,2,2,2,4,6,6,6]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5],ymm1[6,7]
-; AVX2-FAST-NEXT: vmovaps 1472(%rdi), %ymm2
-; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps 1504(%rdi), %ymm3
-; AVX2-FAST-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2,3],ymm2[4,5],ymm3[6,7]
-; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpermps %ymm2, %ymm0, %ymm2
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovaps 96(%rdi), %ymm1
; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps 96(%rdi), %ymm2
-; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps 64(%rdi), %ymm1
+; AVX2-FAST-NEXT: vmovaps 64(%rdi), %ymm0
+; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[0,1],ymm1[0,1]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm13 = ymm0[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-FAST-NEXT: vmovaps (%rdi), %ymm0
+; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovaps 32(%rdi), %ymm1
; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm1[0,1],ymm2[0,1]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm13 = ymm1[0,1,2,3,4,5],ymm2[6,7]
-; AVX2-FAST-NEXT: vmovaps (%rdi), %ymm1
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm12 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-FAST-NEXT: vpermps %ymm12, %ymm4, %ymm0
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm1 = ymm13[0,2,2,2,4,6,6,6]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
+; AVX2-FAST-NEXT: vmovaps 128(%rdi), %ymm1
; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps 32(%rdi), %ymm2
+; AVX2-FAST-NEXT: vmovaps 160(%rdi), %ymm2
; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm12 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
-; AVX2-FAST-NEXT: vpermps %ymm12, %ymm10, %ymm1
-; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm2 = ymm13[0,2,2,2,4,6,6,6]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5],ymm1[6,7]
-; AVX2-FAST-NEXT: vmovaps 128(%rdi), %ymm2
-; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps 160(%rdi), %ymm3
-; AVX2-FAST-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm11 = ymm3[0,1,2,3],ymm2[4,5],ymm3[6,7]
-; AVX2-FAST-NEXT: vpermps %ymm11, %ymm0, %ymm2
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm11 = ymm2[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-FAST-NEXT: vpermps %ymm11, %ymm10, %ymm1
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovaps 480(%rdi), %ymm1
; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps 480(%rdi), %ymm2
-; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps 448(%rdi), %ymm1
+; AVX2-FAST-NEXT: vmovaps 448(%rdi), %ymm0
+; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[0,1],ymm1[0,1]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm9 = ymm0[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-FAST-NEXT: vmovaps 416(%rdi), %ymm0
+; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovaps 384(%rdi), %ymm1
; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm1[0,1],ymm2[0,1]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm9 = ymm1[0,1,2,3,4,5],ymm2[6,7]
-; AVX2-FAST-NEXT: vmovaps 416(%rdi), %ymm1
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm7 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
+; AVX2-FAST-NEXT: vpermps %ymm7, %ymm4, %ymm0
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm1 = ymm9[0,2,2,2,4,6,6,6]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
+; AVX2-FAST-NEXT: vmovaps 512(%rdi), %ymm1
; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps 384(%rdi), %ymm2
+; AVX2-FAST-NEXT: vmovaps 544(%rdi), %ymm2
; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm7 = ymm2[0,1,2,3],ymm1[4,5],ymm2[6,7]
-; AVX2-FAST-NEXT: vpermps %ymm7, %ymm10, %ymm1
-; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm2 = ymm9[0,2,2,2,4,6,6,6]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5],ymm1[6,7]
-; AVX2-FAST-NEXT: vmovaps 512(%rdi), %ymm2
-; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps 544(%rdi), %ymm3
-; AVX2-FAST-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm8 = ymm3[0,1,2,3],ymm2[4,5],ymm3[6,7]
-; AVX2-FAST-NEXT: vpermps %ymm8, %ymm0, %ymm2
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm8 = ymm2[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-FAST-NEXT: vpermps %ymm8, %ymm10, %ymm1
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovaps 864(%rdi), %ymm1
; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps 864(%rdi), %ymm2
-; AVX2-FAST-NEXT: vmovups %ymm2, (%rsp) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps 832(%rdi), %ymm1
+; AVX2-FAST-NEXT: vmovaps 832(%rdi), %ymm0
+; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[0,1],ymm1[0,1]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm6 = ymm0[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-FAST-NEXT: vmovaps 800(%rdi), %ymm0
+; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovaps 768(%rdi), %ymm1
; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm1[0,1],ymm2[0,1]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm6 = ymm1[0,1,2,3,4,5],ymm2[6,7]
-; AVX2-FAST-NEXT: vmovaps 800(%rdi), %ymm1
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
+; AVX2-FAST-NEXT: vpermps %ymm3, %ymm4, %ymm0
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm1 = ymm6[0,2,2,2,4,6,6,6]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
+; AVX2-FAST-NEXT: vmovaps 896(%rdi), %ymm1
; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps 768(%rdi), %ymm2
-; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm2[0,1,2,3],ymm1[4,5],ymm2[6,7]
-; AVX2-FAST-NEXT: vpermps %ymm4, %ymm10, %ymm1
-; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm2 = ymm6[0,2,2,2,4,6,6,6]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5],ymm1[6,7]
-; AVX2-FAST-NEXT: vmovaps 896(%rdi), %ymm2
+; AVX2-FAST-NEXT: vmovaps 928(%rdi), %ymm2
; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps 928(%rdi), %ymm3
-; AVX2-FAST-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm5 = ymm3[0,1,2,3],ymm2[4,5],ymm3[6,7]
-; AVX2-FAST-NEXT: vpermps %ymm5, %ymm0, %ymm2
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm5 = ymm2[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-FAST-NEXT: vpermps %ymm5, %ymm10, %ymm1
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovaps 1184(%rdi), %ymm0
+; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovaps 1152(%rdi), %ymm1
; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps 1184(%rdi), %ymm1
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
+; AVX2-FAST-NEXT: vpermps %ymm2, %ymm4, %ymm0
+; AVX2-FAST-NEXT: vmovaps 1248(%rdi), %ymm4
+; AVX2-FAST-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovaps 1216(%rdi), %ymm1
+; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm1[0,1],ymm4[0,1]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm1[0,1,2,3,4,5],ymm4[6,7]
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm1 = ymm4[0,2,2,2,4,6,6,6]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
+; AVX2-FAST-NEXT: vmovaps 1280(%rdi), %ymm1
; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps 1152(%rdi), %ymm2
-; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm2[0,1,2,3],ymm1[4,5],ymm2[6,7]
-; AVX2-FAST-NEXT: vpermps %ymm3, %ymm10, %ymm1
-; AVX2-FAST-NEXT: vmovaps 1248(%rdi), %ymm10
-; AVX2-FAST-NEXT: vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps 1216(%rdi), %ymm2
-; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vperm2f128 {{.*#+}} ymm2 = ymm2[0,1],ymm10[0,1]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm10 = ymm2[0,1,2,3,4,5],ymm10[6,7]
-; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm2 = ymm10[0,2,2,2,4,6,6,6]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5],ymm1[6,7]
-; AVX2-FAST-NEXT: vmovaps 1280(%rdi), %ymm2
-; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovaps 1312(%rdi), %ymm14
; AVX2-FAST-NEXT: vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm14[0,1,2,3],ymm2[4,5],ymm14[6,7]
-; AVX2-FAST-NEXT: vpermps %ymm2, %ymm0, %ymm0
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm14[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-FAST-NEXT: vpermps %ymm1, %ymm10, %ymm10
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm10[6,7]
; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps {{.*#+}} xmm1 = <1,7,5,u>
-; AVX2-FAST-NEXT: vpermps %ymm15, %ymm1, %ymm0
+; AVX2-FAST-NEXT: vmovaps {{.*#+}} xmm10 = <1,7,5,u>
+; AVX2-FAST-NEXT: vpermps %ymm15, %ymm10, %ymm0
; AVX2-FAST-NEXT: vpermilps $237, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
; AVX2-FAST-NEXT: # ymm14 = mem[1,3,2,3,5,7,6,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm14 = ymm0[0,1,2],ymm14[3,4,5],ymm0[6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm14 = ymm0[0,1,2],ymm14[3,4,5,6,7]
; AVX2-FAST-NEXT: vbroadcastsd {{.*#+}} ymm0 = [12884901893,12884901893,12884901893,12884901893]
; AVX2-FAST-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4,5],ymm15[6,7]
; AVX2-FAST-NEXT: vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm14 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm14 # 32-byte Folded Reload
; AVX2-FAST-NEXT: vpermilps $237, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
; AVX2-FAST-NEXT: # ymm15 = mem[1,3,2,3,5,7,6,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1,2],ymm15[3,4,5],ymm14[6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1,2],ymm15[3,4,5,6,7]
; AVX2-FAST-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4,5],ymm15[6,7]
; AVX2-FAST-NEXT: vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm14 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm14 # 32-byte Folded Reload
; AVX2-FAST-NEXT: vpermilps $237, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
; AVX2-FAST-NEXT: # ymm15 = mem[1,3,2,3,5,7,6,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1,2],ymm15[3,4,5],ymm14[6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1,2],ymm15[3,4,5,6,7]
; AVX2-FAST-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4,5],ymm15[6,7]
; AVX2-FAST-NEXT: vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm14 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm14 # 32-byte Folded Reload
; AVX2-FAST-NEXT: vpermilps $237, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
; AVX2-FAST-NEXT: # ymm15 = mem[1,3,2,3,5,7,6,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1,2],ymm15[3,4,5],ymm14[6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1,2],ymm15[3,4,5,6,7]
; AVX2-FAST-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4,5],ymm15[6,7]
; AVX2-FAST-NEXT: vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpermps %ymm12, %ymm1, %ymm12
+; AVX2-FAST-NEXT: vpermps %ymm12, %ymm10, %ymm12
; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm13 = ymm13[1,3,2,3,5,7,6,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm12 = ymm12[0,1,2],ymm13[3,4,5],ymm12[6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm12 = ymm12[0,1,2],ymm13[3,4,5,6,7]
; AVX2-FAST-NEXT: vpermps %ymm11, %ymm0, %ymm11
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm11 = ymm12[0,1,2,3,4,5],ymm11[6,7]
; AVX2-FAST-NEXT: vmovups %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpermps %ymm7, %ymm1, %ymm7
+; AVX2-FAST-NEXT: vpermps %ymm7, %ymm10, %ymm7
; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm9 = ymm9[1,3,2,3,5,7,6,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2],ymm9[3,4,5],ymm7[6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2],ymm9[3,4,5,6,7]
; AVX2-FAST-NEXT: vpermps %ymm8, %ymm0, %ymm8
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5],ymm8[6,7]
; AVX2-FAST-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpermps %ymm4, %ymm1, %ymm4
+; AVX2-FAST-NEXT: vpermps %ymm3, %ymm10, %ymm3
; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm6 = ymm6[1,3,2,3,5,7,6,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2],ymm6[3,4,5],ymm4[6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2],ymm6[3,4,5,6,7]
; AVX2-FAST-NEXT: vpermps %ymm5, %ymm0, %ymm5
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5],ymm5[6,7]
-; AVX2-FAST-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpermps %ymm3, %ymm1, %ymm1
-; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm3 = ymm10[1,3,2,3,5,7,6,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm3[3,4,5],ymm1[6,7]
-; AVX2-FAST-NEXT: vpermps %ymm2, %ymm0, %ymm0
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm5[6,7]
+; AVX2-FAST-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vpermps %ymm2, %ymm10, %ymm2
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm3 = ymm4[1,3,2,3,5,7,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm3[3,4,5,6,7]
+; AVX2-FAST-NEXT: vpermps %ymm1, %ymm0, %ymm0
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3,4,5],ymm0[6,7]
; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm0 = mem[2,3,2,3,6,7,6,7]
+; AVX2-FAST-NEXT: vmovaps 192(%rdi), %xmm6
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} xmm0 = xmm6[2,3,2,3]
; AVX2-FAST-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
; AVX2-FAST-NEXT: # ymm1 = mem[0,0,2,3,4,4,6,7]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FAST-NEXT: vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm2 # 32-byte Folded Reload
; AVX2-FAST-NEXT: # ymm2 = mem[0,1],ymm0[2,3],mem[4,5],ymm0[6,7]
-; AVX2-FAST-NEXT: vmovaps {{.*#+}} ymm4 = [2,0,6,4,2,0,6,7]
-; AVX2-FAST-NEXT: vpermps %ymm2, %ymm4, %ymm2
+; AVX2-FAST-NEXT: vmovaps {{.*#+}} ymm0 = [2,0,6,4,2,0,6,7]
+; AVX2-FAST-NEXT: vpermps %ymm2, %ymm0, %ymm2
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,3,2,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4],ymm1[5,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
; AVX2-FAST-NEXT: vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
; AVX2-FAST-NEXT: # ymm2 = mem[0,0,0,0,4,4,4,4]
; AVX2-FAST-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
@@ -8127,279 +8085,283 @@ define void @load_i32_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,1,0,3]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm1 = mem[2,3,2,3,6,7,6,7]
-; AVX2-FAST-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm2 = mem[0,0,2,3,4,4,6,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm2[1,2,3],ymm1[4],ymm2[5,6,7]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-FAST-NEXT: vblendps $204, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm2 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm2 = ymm5[0,1],mem[2,3],ymm5[4,5],mem[6,7]
-; AVX2-FAST-NEXT: vpermps %ymm2, %ymm4, %ymm2
-; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,3,2,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4],ymm1[5,6,7]
-; AVX2-FAST-NEXT: vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm2 = mem[0,0,0,0,4,4,4,4]
+; AVX2-FAST-NEXT: vmovaps 576(%rdi), %xmm1
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} xmm2 = xmm1[2,3,2,3]
; AVX2-FAST-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
; AVX2-FAST-NEXT: # ymm3 = mem[0,0,2,3,4,4,6,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3],ymm3[4,5,6],ymm2[7]
-; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,1,0,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
-; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm1 = ymm9[2,3,2,3,6,7,6,7]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
-; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm2 = ymm10[0,0,2,3,4,4,6,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm2[1,2,3],ymm1[4],ymm2[5,6,7]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm8[0,1],ymm6[2,3],ymm8[4,5],ymm6[6,7]
-; AVX2-FAST-NEXT: vpermps %ymm2, %ymm4, %ymm2
-; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,3,2,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4],ymm1[5,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0],ymm3[1,2,3],ymm2[4],ymm3[5,6,7]
+; AVX2-FAST-NEXT: vmovups (%rsp), %ymm3 # 32-byte Reload
+; AVX2-FAST-NEXT: vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm3 = mem[0,1],ymm3[2,3],mem[4,5],ymm3[6,7]
+; AVX2-FAST-NEXT: vpermps %ymm3, %ymm0, %ymm3
+; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,3,2,3]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm3[3,4,5,6,7]
+; AVX2-FAST-NEXT: vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm3 = mem[0,0,0,0,4,4,4,4]
+; AVX2-FAST-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm4 = mem[0,0,2,3,4,4,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3],ymm4[4,5,6],ymm3[7]
+; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm3 = ymm3[0,1,0,3]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7]
+; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovaps 960(%rdi), %xmm2
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} xmm3 = xmm2[2,3,2,3]
+; AVX2-FAST-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm4 = mem[0,0,2,3,4,4,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0],ymm4[1,2,3],ymm3[4],ymm4[5,6,7]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-FAST-NEXT: vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm4 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm4 = mem[0,1],ymm4[2,3],mem[4,5],ymm4[6,7]
+; AVX2-FAST-NEXT: vpermps %ymm4, %ymm0, %ymm4
+; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm3 = ymm3[0,3,2,3]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2],ymm4[3,4,5,6,7]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm4 = ymm13[0,0,0,0,4,4,4,4]
+; AVX2-FAST-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm5 = mem[0,0,2,3,4,4,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm5[0,1,2],ymm4[3],ymm5[4,5,6],ymm4[7]
+; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm4 = ymm4[0,1,0,3]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4],ymm4[5,6,7]
+; AVX2-FAST-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovaps 1344(%rdi), %xmm3
+; AVX2-FAST-NEXT: vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} xmm4 = xmm3[2,3,2,3]
+; AVX2-FAST-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm5 = mem[0,0,2,3,4,4,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0],ymm5[1,2,3],ymm4[4],ymm5[5,6,7]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm5 = ymm11[0,1],ymm14[2,3],ymm11[4,5],ymm14[6,7]
+; AVX2-FAST-NEXT: vpermps %ymm5, %ymm0, %ymm5
+; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm4 = ymm4[0,3,2,3]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2],ymm5[3,4,5,6,7]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm5 = ymm3[0,0,0,0,4,4,4,4]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm7 = ymm15[0,0,2,3,4,4,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm5 = ymm7[0,1,2],ymm5[3],ymm7[4,5,6],ymm5[7]
+; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm5 = ymm5[0,1,0,3]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4],ymm5[5,6,7]
+; AVX2-FAST-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovaps 1152(%rdi), %xmm4
+; AVX2-FAST-NEXT: vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} xmm5 = xmm4[2,3,2,3]
+; AVX2-FAST-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm7 = mem[0,0,2,3,4,4,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0],ymm7[1,2,3],ymm5[4],ymm7[5,6,7]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-FAST-NEXT: vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm7 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm7 = mem[0,1],ymm4[2,3],mem[4,5],ymm4[6,7]
+; AVX2-FAST-NEXT: vpermps %ymm7, %ymm0, %ymm7
+; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm5 = ymm5[0,3,2,3]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2],ymm7[3,4,5,6,7]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm7 = ymm12[0,0,0,0,4,4,4,4]
+; AVX2-FAST-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm8 = mem[0,0,2,3,4,4,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm7 = ymm8[0,1,2],ymm7[3],ymm8[4,5,6],ymm7[7]
+; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm7 = ymm7[0,1,0,3]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4],ymm7[5,6,7]
+; AVX2-FAST-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovaps (%rdi), %xmm4
+; AVX2-FAST-NEXT: vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} xmm5 = xmm4[2,3,2,3]
+; AVX2-FAST-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm8 = mem[0,0,2,3,4,4,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0],ymm8[1,2,3],ymm5[4],ymm8[5,6,7]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-FAST-NEXT: vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm8 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm8 = mem[0,1],ymm4[2,3],mem[4,5],ymm4[6,7]
+; AVX2-FAST-NEXT: vpermps %ymm8, %ymm0, %ymm8
+; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm5 = ymm5[0,3,2,3]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2],ymm8[3,4,5,6,7]
+; AVX2-FAST-NEXT: vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm8 = mem[0,0,0,0,4,4,4,4]
+; AVX2-FAST-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm9 = mem[0,0,2,3,4,4,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm8 = ymm9[0,1,2],ymm8[3],ymm9[4,5,6],ymm8[7]
+; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm8 = ymm8[0,1,0,3]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm5[0,1,2,3,4],ymm8[5,6,7]
+; AVX2-FAST-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovaps 384(%rdi), %xmm4
+; AVX2-FAST-NEXT: vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} xmm8 = xmm4[2,3,2,3]
+; AVX2-FAST-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm9 = mem[0,0,2,3,4,4,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0],ymm9[1,2,3],ymm8[4],ymm9[5,6,7]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-FAST-NEXT: vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm9 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm9 = mem[0,1],ymm4[2,3],mem[4,5],ymm4[6,7]
+; AVX2-FAST-NEXT: vpermps %ymm9, %ymm0, %ymm9
+; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm8 = ymm8[0,3,2,3]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2],ymm9[3,4,5,6,7]
+; AVX2-FAST-NEXT: vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm9 = mem[0,0,0,0,4,4,4,4]
+; AVX2-FAST-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm10 = mem[0,0,2,3,4,4,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm9 = ymm10[0,1,2],ymm9[3],ymm10[4,5,6],ymm9[7]
+; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm9 = ymm9[0,1,0,3]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm8[0,1,2,3,4],ymm9[5,6,7]
+; AVX2-FAST-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-FAST-NEXT: vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm8 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm8 = mem[0,1],ymm4[2,3],mem[4,5],ymm4[6,7]
+; AVX2-FAST-NEXT: vpermps %ymm8, %ymm0, %ymm8
+; AVX2-FAST-NEXT: vmovaps 768(%rdi), %xmm0
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} xmm9 = xmm0[2,3,2,3]
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
-; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm2 = ymm7[0,0,0,0,4,4,4,4]
-; AVX2-FAST-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm3 = mem[0,0,2,3,4,4,6,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3],ymm3[4,5,6],ymm2[7]
-; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,1,0,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm10 = ymm7[0,0,2,3,4,4,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm9 = ymm9[0],ymm10[1,2,3],ymm9[4],ymm10[5,6,7]
+; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm9 = ymm9[0,3,2,3]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm8 = ymm9[0,1,2],ymm8[3,4,5,6,7]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm9 = ymm5[0,0,0,0,4,4,4,4]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm10 = ymm4[0,0,2,3,4,4,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm9 = ymm10[0,1,2],ymm9[3],ymm10[4,5,6],ymm9[7]
+; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm9 = ymm9[0,1,0,3]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4],ymm9[5,6,7]
+; AVX2-FAST-NEXT: vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} xmm6 = xmm6[3,3,3,3]
+; AVX2-FAST-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm8 = mem[0,1,3,3,4,5,7,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0],ymm8[1,2,3],ymm6[4],ymm8[5,6,7]
+; AVX2-FAST-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm8 = mem[3,3,3,3,7,7,7,7]
+; AVX2-FAST-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm8 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm8 = ymm8[0],mem[1],ymm8[2,3,4],mem[5],ymm8[6,7]
+; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm8 = ymm8[0,2,0,3]
+; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm6 = ymm6[0,3,2,3]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2],ymm8[3,4,5,6,7]
+; AVX2-FAST-NEXT: vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm8 = mem[0,1,0,1,4,5,4,5]
+; AVX2-FAST-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm9 = mem[0,1,3,3,4,5,7,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm8 = ymm9[0,1,2],ymm8[3],ymm9[4,5,6],ymm8[7]
+; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm8 = ymm8[0,1,0,3]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4],ymm8[5,6,7]
+; AVX2-FAST-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} xmm1 = xmm1[3,3,3,3]
+; AVX2-FAST-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm6 = mem[0,1,3,3,4,5,7,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm6[1,2,3],ymm1[4],ymm6[5,6,7]
+; AVX2-FAST-NEXT: vpermilps $255, (%rsp), %ymm6 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm6 = mem[3,3,3,3,7,7,7,7]
+; AVX2-FAST-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm6 = ymm6[0],mem[1],ymm6[2,3,4],mem[5],ymm6[6,7]
+; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm6 = ymm6[0,2,0,3]
+; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,3,2,3]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm6[3,4,5,6,7]
+; AVX2-FAST-NEXT: vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm6 = mem[0,1,0,1,4,5,4,5]
+; AVX2-FAST-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm8 = mem[0,1,3,3,4,5,7,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm6 = ymm8[0,1,2],ymm6[3],ymm8[4,5,6],ymm6[7]
+; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm6 = ymm6[0,1,0,3]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm6[5,6,7]
; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
-; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm1 = ymm15[2,3,2,3,6,7,6,7]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
-; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm2 = ymm14[0,0,2,3,4,4,6,7]
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} xmm1 = xmm2[3,3,3,3]
+; AVX2-FAST-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm2 = mem[0,1,3,3,4,5,7,7]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm2[1,2,3],ymm1[4],ymm2[5,6,7]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm12[0,1],ymm11[2,3],ymm12[4,5],ymm11[6,7]
-; AVX2-FAST-NEXT: vpermps %ymm2, %ymm4, %ymm2
+; AVX2-FAST-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm2 = mem[3,3,3,3,7,7,7,7]
+; AVX2-FAST-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm2 = ymm2[0],mem[1],ymm2[2,3,4],mem[5],ymm2[6,7]
+; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,2,0,3]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,3,2,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4],ymm1[5,6,7]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
-; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm2 = ymm13[0,0,0,0,4,4,4,4]
-; AVX2-FAST-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm3 = mem[0,0,2,3,4,4,6,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3],ymm3[4,5,6],ymm2[7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm2 = ymm13[0,1,0,1,4,5,4,5]
+; AVX2-FAST-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm6 = mem[0,1,3,3,4,5,7,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm6[0,1,2],ymm2[3],ymm6[4,5,6],ymm2[7]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,1,0,3]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm1 = mem[2,3,2,3,6,7,6,7]
-; AVX2-FAST-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm2 = mem[0,0,2,3,4,4,6,7]
+; AVX2-FAST-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; AVX2-FAST-NEXT: # xmm1 = mem[3,3,3,3]
+; AVX2-FAST-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm2 = mem[0,1,3,3,4,5,7,7]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm2[1,2,3],ymm1[4],ymm2[5,6,7]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-NEXT: vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm2 = mem[0,1],ymm2[2,3],mem[4,5],ymm2[6,7]
-; AVX2-FAST-NEXT: vpermps %ymm2, %ymm4, %ymm2
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm2 = ymm14[3,3,3,3,7,7,7,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0],ymm11[1],ymm2[2,3,4],ymm11[5],ymm2[6,7]
+; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,2,0,3]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,3,2,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4],ymm1[5,6,7]
-; AVX2-FAST-NEXT: vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm2 = mem[0,0,0,0,4,4,4,4]
-; AVX2-FAST-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm3 = mem[0,0,2,3,4,4,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm2 = ymm3[0,1,0,1,4,5,4,5]
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm3 = ymm15[0,1,3,3,4,5,7,7]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3],ymm3[4,5,6],ymm2[7]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,1,0,3]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm1 = mem[2,3,2,3,6,7,6,7]
-; AVX2-FAST-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm2 = mem[0,0,2,3,4,4,6,7]
+; AVX2-FAST-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; AVX2-FAST-NEXT: # xmm1 = mem[3,3,3,3]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm2 = ymm11[0,1,3,3,4,5,7,7]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm2[1,2,3],ymm1[4],ymm2[5,6,7]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-FAST-NEXT: vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm2 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm2 = mem[0,1],ymm0[2,3],mem[4,5],ymm0[6,7]
-; AVX2-FAST-NEXT: vpermps %ymm2, %ymm4, %ymm2
+; AVX2-FAST-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm2 = mem[3,3,3,3,7,7,7,7]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0],ymm13[1],ymm2[2,3,4],ymm13[5],ymm2[6,7]
+; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,2,0,3]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,3,2,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4],ymm1[5,6,7]
-; AVX2-FAST-NEXT: vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm2 = mem[0,0,0,0,4,4,4,4]
-; AVX2-FAST-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm3 = mem[0,0,2,3,4,4,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm2 = ymm12[0,1,0,1,4,5,4,5]
+; AVX2-FAST-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm3 = mem[0,1,3,3,4,5,7,7]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3],ymm3[4,5,6],ymm2[7]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,1,0,3]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm1 = mem[2,3,2,3,6,7,6,7]
-; AVX2-FAST-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm2 = mem[0,0,2,3,4,4,6,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm2[1,2,3],ymm1[4],ymm2[5,6,7]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-NEXT: vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm2 = mem[0,1],ymm2[2,3],mem[4,5],ymm2[6,7]
-; AVX2-FAST-NEXT: vpermps %ymm2, %ymm4, %ymm2
-; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,3,2,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4],ymm1[5,6,7]
-; AVX2-FAST-NEXT: vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm2 = mem[0,0,0,0,4,4,4,4]
-; AVX2-FAST-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm3 = mem[0,0,2,3,4,4,6,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3],ymm3[4,5,6],ymm2[7]
-; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,1,0,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
-; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovups (%rsp), %ymm1 # 32-byte Reload
-; AVX2-FAST-NEXT: vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm1 = mem[0,1],ymm1[2,3],mem[4,5],ymm1[6,7]
-; AVX2-FAST-NEXT: vpermps %ymm1, %ymm4, %ymm0
-; AVX2-FAST-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm1 = mem[2,3,2,3,6,7,6,7]
-; AVX2-FAST-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm2 = mem[0,0,2,3,4,4,6,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm2[1,2,3],ymm1[4],ymm2[5,6,7]
-; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,3,2,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4],ymm1[5,6,7]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm1 = ymm3[0,0,0,0,4,4,4,4]
-; AVX2-FAST-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm2 = mem[0,0,2,3,4,4,6,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
-; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
-; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm0 = mem[3,3,3,3,7,7,7,7]
-; AVX2-FAST-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm1 = mem[0,1,3,3,4,5,7,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
-; AVX2-FAST-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm1 = mem[3,3,3,3,7,7,7,7]
-; AVX2-FAST-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm1 = ymm1[0],mem[1],ymm1[2,3,4],mem[5],ymm1[6,7]
-; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
-; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
-; AVX2-FAST-NEXT: vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm1 = mem[0,1,0,1,4,5,4,5]
-; AVX2-FAST-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm2 = mem[0,1,3,3,4,5,7,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
-; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
-; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm0 = mem[3,3,3,3,7,7,7,7]
-; AVX2-FAST-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm1 = mem[0,1,3,3,4,5,7,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
-; AVX2-FAST-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm1 = mem[3,3,3,3,7,7,7,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm5[1],ymm1[2,3,4],ymm5[5],ymm1[6,7]
-; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
-; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
-; AVX2-FAST-NEXT: vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm1 = mem[0,1,0,1,4,5,4,5]
-; AVX2-FAST-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm2 = mem[0,1,3,3,4,5,7,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
-; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
-; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm0 = ymm9[3,3,3,3,7,7,7,7]
-; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm1 = ymm10[0,1,3,3,4,5,7,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
-; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm1 = ymm6[3,3,3,3,7,7,7,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm8[1],ymm1[2,3,4],ymm8[5],ymm1[6,7]
-; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
-; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
-; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm1 = ymm7[0,1,0,1,4,5,4,5]
-; AVX2-FAST-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm2 = mem[0,1,3,3,4,5,7,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
-; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
-; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm0 = ymm15[3,3,3,3,7,7,7,7]
-; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm1 = ymm14[0,1,3,3,4,5,7,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
-; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm1 = ymm11[3,3,3,3,7,7,7,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm12[1],ymm1[2,3,4],ymm12[5],ymm1[6,7]
-; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
-; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
-; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm1 = ymm13[0,1,0,1,4,5,4,5]
-; AVX2-FAST-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm2 = mem[0,1,3,3,4,5,7,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
-; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
-; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm0 = mem[3,3,3,3,7,7,7,7]
-; AVX2-FAST-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm1 = mem[0,1,3,3,4,5,7,7]
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm1 = ymm7[0,1,3,3,4,5,7,7]
+; AVX2-FAST-NEXT: vmovaps %ymm7, %ymm12
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
-; AVX2-FAST-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm1 = mem[3,3,3,3,7,7,7,7]
-; AVX2-FAST-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm1 = ymm1[0],mem[1],ymm1[2,3,4],mem[5],ymm1[6,7]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm1 = ymm10[3,3,3,3,7,7,7,7]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm14[1],ymm1[2,3,4],ymm14[5],ymm1[6,7]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
-; AVX2-FAST-NEXT: vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm1 = mem[0,1,0,1,4,5,4,5]
-; AVX2-FAST-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm2 = mem[0,1,3,3,4,5,7,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm1 = ymm5[0,1,0,1,4,5,4,5]
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm2 = ymm4[0,1,3,3,4,5,7,7]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
-; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm0 = ymm12[3,3,3,3,7,7,7,7]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
-; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm1 = ymm11[0,1,3,3,4,5,7,7]
+; AVX2-FAST-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; AVX2-FAST-NEXT: # xmm0 = mem[3,3,3,3]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm1 = ymm3[0,1,3,3,4,5,7,7]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
-; AVX2-FAST-NEXT: vpermilps $255, (%rsp), %ymm1 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm1 = mem[3,3,3,3,7,7,7,7]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm13[1],ymm1[2,3,4],ymm13[5],ymm1[6,7]
-; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
-; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
-; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm1 = ymm3[0,1,0,1,4,5,4,5]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
-; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm2 = ymm15[0,1,3,3,4,5,7,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
-; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
-; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm0 = ymm6[3,3,3,3,7,7,7,7]
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm1 = ymm6[3,3,3,3,7,7,7,7]
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm1 = ymm5[0,1,3,3,4,5,7,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm1 = ymm8[3,3,3,3,7,7,7,7]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm7[1],ymm1[2,3,4],ymm7[5],ymm1[6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm5[1],ymm1[2,3,4],ymm5[5],ymm1[6,7]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
-; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm1 = ymm10[0,1,0,1,4,5,4,5]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm2 = ymm9[0,1,3,3,4,5,7,7]
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm1 = ymm9[0,1,0,1,4,5,4,5]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm2 = ymm7[0,1,3,3,4,5,7,7]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm0 = mem[3,3,3,3,7,7,7,7]
+; AVX2-FAST-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; AVX2-FAST-NEXT: # xmm0 = mem[3,3,3,3]
; AVX2-FAST-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
; AVX2-FAST-NEXT: # ymm1 = mem[0,1,3,3,4,5,7,7]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
; AVX2-FAST-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
; AVX2-FAST-NEXT: # ymm1 = mem[3,3,3,3,7,7,7,7]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm4[1],ymm1[2,3,4],ymm4[5],ymm1[6,7]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm15[1],ymm1[2,3,4],ymm15[5],ymm1[6,7]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
; AVX2-FAST-NEXT: vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
; AVX2-FAST-NEXT: # ymm1 = mem[0,1,0,1,4,5,4,5]
; AVX2-FAST-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
@@ -8409,150 +8371,153 @@ define void @load_i32_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-FAST-NEXT: vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm2 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm2 = ymm0[0,1,2,3],mem[4,5],ymm0[6,7]
-; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-FAST-NEXT: vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm1 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm1 = mem[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-FAST-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm1 = ymm0[0,1,2,3],mem[4,5,6,7]
; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-NEXT: vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm2 = mem[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovaps 272(%rdi), %xmm0
; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3],ymm2[4,5,6,7]
; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,1,0,2,4,5,4,6]
-; AVX2-FAST-NEXT: vmovaps {{.*#+}} xmm1 = <4,2,u,u>
-; AVX2-FAST-NEXT: vpermps %ymm2, %ymm1, %ymm2
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm0[2,3,4,5,6,7]
+; AVX2-FAST-NEXT: vmovaps {{.*#+}} xmm8 = <4,2,u,u>
+; AVX2-FAST-NEXT: vpermps %ymm1, %ymm8, %ymm1
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm0[2,3,4,5,6,7]
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-FAST-NEXT: vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm14 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm14 = ymm0[0,1],mem[2,3],ymm0[4,5,6,7]
-; AVX2-FAST-NEXT: vbroadcastf128 {{.*#+}} ymm0 = [0,2,0,6,0,2,0,6]
-; AVX2-FAST-NEXT: # ymm0 = mem[0,1,0,1]
-; AVX2-FAST-NEXT: vpermps %ymm14, %ymm0, %ymm3
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7]
-; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm5[0,1,2,3],ymm6[4,5],ymm5[6,7]
-; AVX2-FAST-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm5 = ymm7[0,1,2,3],ymm8[4,5,6,7]
-; AVX2-FAST-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps 464(%rdi), %xmm2
-; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm5[0,1],ymm2[2,3],ymm5[4,5,6,7]
-; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm2 = ymm2[0,1,0,2,4,5,4,6]
-; AVX2-FAST-NEXT: vpermps %ymm3, %ymm1, %ymm3
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3,4,5,6,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm10[0,1],ymm9[2,3],ymm10[4,5,6,7]
-; AVX2-FAST-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpermps %ymm3, %ymm0, %ymm3
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7]
-; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-NEXT: vblendps $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm3 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm3 = mem[0,1,2,3],ymm2[4,5],mem[6,7]
-; AVX2-FAST-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm5 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm5 = ymm2[0,1,2,3],mem[4,5,6,7]
-; AVX2-FAST-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps 656(%rdi), %xmm2
-; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm5[0,1],ymm2[2,3],ymm5[4,5,6,7]
-; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm2 = ymm2[0,1,0,2,4,5,4,6]
-; AVX2-FAST-NEXT: vpermps %ymm3, %ymm1, %ymm3
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3,4,5,6,7]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-FAST-NEXT: vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm3 = ymm3[0,1],mem[2,3],ymm3[4,5,6,7]
-; AVX2-FAST-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpermps %ymm3, %ymm0, %ymm3
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7]
-; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm11[0,1,2,3],ymm12[4,5],ymm11[6,7]
+; AVX2-FAST-NEXT: vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm0 = ymm0[0,1],mem[2,3],ymm0[4,5,6,7]
+; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vbroadcastf128 {{.*#+}} ymm4 = [0,2,0,6,0,2,0,6]
+; AVX2-FAST-NEXT: # ymm4 = mem[0,1,0,1]
+; AVX2-FAST-NEXT: vpermps %ymm0, %ymm4, %ymm2
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm3 = ymm3[0,1,2,3],mem[4,5,6,7]
; AVX2-FAST-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vblendps $240, (%rsp), %ymm13, %ymm5 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm5 = ymm13[0,1,2,3],mem[4,5,6,7]
-; AVX2-FAST-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps 848(%rdi), %xmm2
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm5[0,1,2,3],ymm6[4,5,6,7]
+; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovaps 464(%rdi), %xmm0
+; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm0[2,3],ymm1[4,5,6,7]
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
+; AVX2-FAST-NEXT: vpermps %ymm3, %ymm8, %ymm2
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm9[0,1],ymm7[2,3],ymm9[4,5,6,7]
+; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vpermps %ymm0, %ymm4, %ymm2
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-NEXT: vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm0 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm0 = mem[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-NEXT: vblendps $240, (%rsp), %ymm1, %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm2 = ymm1[0,1,2,3],mem[4,5,6,7]
; AVX2-FAST-NEXT: vmovups %ymm2, (%rsp) # 32-byte Spill
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm5[0,1],ymm2[2,3],ymm5[4,5,6,7]
-; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm2 = ymm2[0,1,0,2,4,5,4,6]
-; AVX2-FAST-NEXT: vpermps %ymm3, %ymm1, %ymm3
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3,4,5,6,7]
-; AVX2-FAST-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm3 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm3 = mem[0,1],ymm15[2,3],mem[4,5,6,7]
-; AVX2-FAST-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpermps %ymm3, %ymm0, %ymm3
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7]
-; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-NEXT: vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm3 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm3 = ymm2[0,1,2,3],mem[4,5],ymm2[6,7]
-; AVX2-FAST-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovaps 656(%rdi), %xmm1
+; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3],ymm2[4,5,6,7]
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
+; AVX2-FAST-NEXT: vpermps %ymm0, %ymm8, %ymm2
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5,6,7]
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm5 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm5 = ymm2[0,1,2,3],mem[4,5,6,7]
-; AVX2-FAST-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps 1040(%rdi), %xmm2
-; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm5[0,1],ymm2[2,3],ymm5[4,5,6,7]
-; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm2 = ymm2[0,1,0,2,4,5,4,6]
-; AVX2-FAST-NEXT: vpermps %ymm3, %ymm1, %ymm3
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3,4,5,6,7]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-FAST-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm13 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm13 = mem[0,1],ymm3[2,3],mem[4,5,6,7]
-; AVX2-FAST-NEXT: vpermps %ymm13, %ymm0, %ymm3
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7]
+; AVX2-FAST-NEXT: vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm0 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm0 = ymm2[0,1],mem[2,3],ymm2[4,5,6,7]
+; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vpermps %ymm0, %ymm4, %ymm2
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm0 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm0 = ymm12[0,1,2,3],mem[4,5,6,7]
+; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm14[0,1,2,3],ymm10[4,5,6,7]
; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-NEXT: vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm3 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm3 = ymm2[0,1,2,3],mem[4,5],ymm2[6,7]
-; AVX2-FAST-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm10 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm10 = ymm2[0,1,2,3],mem[4,5,6,7]
-; AVX2-FAST-NEXT: vmovaps 1232(%rdi), %xmm12
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm10[0,1],ymm12[2,3],ymm10[4,5,6,7]
-; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm2 = ymm2[0,1,0,2,4,5,4,6]
-; AVX2-FAST-NEXT: vpermps %ymm3, %ymm1, %ymm3
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3,4,5,6,7]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-FAST-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm9 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm9 = mem[0,1],ymm3[2,3],mem[4,5,6,7]
-; AVX2-FAST-NEXT: vpermps %ymm9, %ymm0, %ymm3
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7]
+; AVX2-FAST-NEXT: vmovaps 848(%rdi), %xmm1
+; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3],ymm2[4,5,6,7]
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
+; AVX2-FAST-NEXT: vpermps %ymm0, %ymm8, %ymm2
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5,6,7]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm0 = mem[0,1],ymm0[2,3],mem[4,5,6,7]
+; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vpermps %ymm0, %ymm4, %ymm2
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm0 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm0 = ymm1[0,1,2,3],mem[4,5,6,7]
+; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm2 = ymm1[0,1,2,3],mem[4,5,6,7]
; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovaps 1040(%rdi), %xmm1
+; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3],ymm2[4,5,6,7]
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
+; AVX2-FAST-NEXT: vpermps %ymm0, %ymm8, %ymm2
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5,6,7]
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-NEXT: vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm11 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm11 = ymm2[0,1,2,3],mem[4,5],ymm2[6,7]
+; AVX2-FAST-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm14 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm14 = mem[0,1],ymm2[2,3],mem[4,5,6,7]
+; AVX2-FAST-NEXT: vpermps %ymm14, %ymm4, %ymm2
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm0 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm0 = ymm11[0,1,2,3],mem[4,5,6,7]
+; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm11 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm11 = ymm13[0,1,2,3],mem[4,5,6,7]
+; AVX2-FAST-NEXT: vmovaps 1232(%rdi), %xmm1
+; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm11[0,1],ymm1[2,3],ymm11[4,5,6,7]
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
+; AVX2-FAST-NEXT: vpermps %ymm0, %ymm8, %ymm2
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5,6,7]
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm7 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm7 = ymm2[0,1,2,3],mem[4,5,6,7]
-; AVX2-FAST-NEXT: vmovaps 1424(%rdi), %xmm8
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm7[0,1],ymm8[2,3],ymm7[4,5,6,7]
-; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm2 = ymm2[0,1,0,2,4,5,4,6]
-; AVX2-FAST-NEXT: vpermps %ymm11, %ymm1, %ymm3
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3,4,5,6,7]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-FAST-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm6 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm6 = mem[0,1],ymm3[2,3],mem[4,5,6,7]
-; AVX2-FAST-NEXT: vpermps %ymm6, %ymm0, %ymm3
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7]
-; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm5 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm5 = ymm4[0,1,2,3],mem[4,5,6,7]
+; AVX2-FAST-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm10 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm10 = mem[0,1],ymm2[2,3],mem[4,5,6,7]
+; AVX2-FAST-NEXT: vpermps %ymm10, %ymm4, %ymm2
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm0 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm0 = ymm1[0,1,2,3],mem[4,5,6,7]
+; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm12 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm12 = ymm1[0,1,2,3],mem[4,5,6,7]
+; AVX2-FAST-NEXT: vmovaps 1424(%rdi), %xmm13
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm12[0,1],ymm13[2,3],ymm12[4,5,6,7]
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
+; AVX2-FAST-NEXT: vpermps %ymm0, %ymm8, %ymm2
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5,6,7]
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-NEXT: vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm4 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm4 = ymm2[0,1,2,3],mem[4,5],ymm2[6,7]
+; AVX2-FAST-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm6 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm6 = mem[0,1],ymm2[2,3],mem[4,5,6,7]
+; AVX2-FAST-NEXT: vpermps %ymm6, %ymm4, %ymm2
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm5 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm5 = ymm15[0,1,2,3],mem[4,5,6,7]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm9 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm9 = ymm1[0,1,2,3],mem[4,5,6,7]
; AVX2-FAST-NEXT: vmovaps 80(%rdi), %xmm2
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm5[0,1],ymm2[2,3],ymm5[4,5,6,7]
-; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm3 = ymm3[0,1,0,2,4,5,4,6]
-; AVX2-FAST-NEXT: vpermps %ymm4, %ymm1, %ymm1
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm3[2,3,4,5,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm5[0,1],ymm2[2,3],ymm5[4,5,6,7]
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
+; AVX2-FAST-NEXT: vpermps %ymm9, %ymm8, %ymm3
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm3[0,1],ymm1[2,3,4,5,6,7]
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX2-FAST-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
; AVX2-FAST-NEXT: # ymm3 = mem[0,1],ymm3[2,3],mem[4,5,6,7]
-; AVX2-FAST-NEXT: vpermps %ymm3, %ymm0, %ymm0
+; AVX2-FAST-NEXT: vpermps %ymm3, %ymm4, %ymm0
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7]
; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
@@ -8560,63 +8525,63 @@ define void @load_i32_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-FAST-NEXT: # ymm0 = ymm0[0,1,2],mem[3],ymm0[4,5,6,7]
; AVX2-FAST-NEXT: vmovaps {{.*#+}} xmm1 = <5,3,u,u>
-; AVX2-FAST-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm15 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm15 = ymm15[0,1],ymm0[2,3,4,5,6,7]
+; AVX2-FAST-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm7 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1],ymm0[2,3,4,5,6,7]
; AVX2-FAST-NEXT: vbroadcastf128 {{.*#+}} ymm0 = [0,3,1,7,0,3,1,7]
; AVX2-FAST-NEXT: # ymm0 = mem[0,1,0,1]
-; AVX2-FAST-NEXT: vpermps %ymm14, %ymm0, %ymm14
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm14 = ymm15[0,1,2,3,4],ymm14[5,6,7]
-; AVX2-FAST-NEXT: vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm14 = mem[1,1,1,1,5,5,5,5]
-; AVX2-FAST-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm14 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm14 = ymm14[0,1,2],mem[3],ymm14[4,5,6,7]
-; AVX2-FAST-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm15 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm14 = ymm15[0,1],ymm14[2,3,4,5,6,7]
; AVX2-FAST-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4],ymm15[5,6,7]
-; AVX2-FAST-NEXT: vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm14 = mem[1,1,1,1,5,5,5,5]
-; AVX2-FAST-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm14 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm14 = ymm14[0,1,2],mem[3],ymm14[4,5,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4],ymm15[5,6,7]
+; AVX2-FAST-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm7 = mem[1,1,1,1,5,5,5,5]
+; AVX2-FAST-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm7 = ymm7[0,1,2],mem[3],ymm7[4,5,6,7]
; AVX2-FAST-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm15 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm14 = ymm15[0,1],ymm14[2,3,4,5,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm7 = ymm15[0,1],ymm7[2,3,4,5,6,7]
; AVX2-FAST-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4],ymm15[5,6,7]
-; AVX2-FAST-NEXT: vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm14 = mem[1,1,1,1,5,5,5,5]
-; AVX2-FAST-NEXT: vblendps $8, (%rsp), %ymm14, %ymm14 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm14 = ymm14[0,1,2],mem[3],ymm14[4,5,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm7[0,1,2,3,4],ymm15[5,6,7]
+; AVX2-FAST-NEXT: vpermilps $85, (%rsp), %ymm7 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm7 = mem[1,1,1,1,5,5,5,5]
+; AVX2-FAST-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm7 = ymm7[0,1,2],mem[3],ymm7[4,5,6,7]
; AVX2-FAST-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm15 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm14 = ymm15[0,1],ymm14[2,3,4,5,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm7 = ymm15[0,1],ymm7[2,3,4,5,6,7]
; AVX2-FAST-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4],ymm15[5,6,7]
-; AVX2-FAST-NEXT: vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm14 = mem[1,1,1,1,5,5,5,5]
-; AVX2-FAST-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm14 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm14 = ymm14[0,1,2],mem[3],ymm14[4,5,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4],ymm15[5,6,7]
+; AVX2-FAST-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm7 = mem[1,1,1,1,5,5,5,5]
+; AVX2-FAST-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm7 = ymm7[0,1,2],mem[3],ymm7[4,5,6,7]
; AVX2-FAST-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm15 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm14 = ymm15[0,1],ymm14[2,3,4,5,6,7]
-; AVX2-FAST-NEXT: vpermps %ymm13, %ymm0, %ymm13
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm13 = ymm14[0,1,2,3,4],ymm13[5,6,7]
-; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm10 = ymm10[1,1,1,1,5,5,5,5]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm10 = ymm10[0,1,2],ymm12[3],ymm10[4,5,6,7]
-; AVX2-FAST-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm12 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm10 = ymm12[0,1],ymm10[2,3,4,5,6,7]
-; AVX2-FAST-NEXT: vpermps %ymm9, %ymm0, %ymm9
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm9 = ymm10[0,1,2,3,4],ymm9[5,6,7]
-; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm7 = ymm7[1,1,1,1,5,5,5,5]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2],ymm8[3],ymm7[4,5,6,7]
-; AVX2-FAST-NEXT: vpermps %ymm11, %ymm1, %ymm8
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm7 = ymm15[0,1],ymm7[2,3,4,5,6,7]
+; AVX2-FAST-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4],ymm15[5,6,7]
+; AVX2-FAST-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm15 = mem[1,1,1,1,5,5,5,5]
+; AVX2-FAST-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm15 = ymm15[0,1,2],mem[3],ymm15[4,5,6,7]
+; AVX2-FAST-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm7 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1],ymm15[2,3,4,5,6,7]
+; AVX2-FAST-NEXT: vpermps %ymm14, %ymm0, %ymm14
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm14 = ymm7[0,1,2,3,4],ymm14[5,6,7]
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm7 = ymm11[1,1,1,1,5,5,5,5]
+; AVX2-FAST-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm7 = ymm7[0,1,2],mem[3],ymm7[4,5,6,7]
+; AVX2-FAST-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm11 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm7 = ymm11[0,1],ymm7[2,3,4,5,6,7]
+; AVX2-FAST-NEXT: vpermps %ymm10, %ymm0, %ymm10
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm10 = ymm7[0,1,2,3,4],ymm10[5,6,7]
+; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm7 = ymm12[1,1,1,1,5,5,5,5]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2],ymm13[3],ymm7[4,5,6,7]
+; AVX2-FAST-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm8 # 32-byte Folded Reload
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm7 = ymm8[0,1],ymm7[2,3,4,5,6,7]
; AVX2-FAST-NEXT: vpermps %ymm6, %ymm0, %ymm6
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm6 = ymm7[0,1,2,3,4],ymm6[5,6,7]
; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm5 = ymm5[1,1,1,1,5,5,5,5]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm5[0,1,2],ymm2[3],ymm5[4,5,6,7]
-; AVX2-FAST-NEXT: vpermps %ymm4, %ymm1, %ymm1
+; AVX2-FAST-NEXT: vpermps %ymm9, %ymm1, %ymm1
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2,3,4,5,6,7]
; AVX2-FAST-NEXT: vpermps %ymm3, %ymm0, %ymm0
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm0[5,6,7]
@@ -8702,53 +8667,52 @@ define void @load_i32_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-NEXT: vmovaps %ymm0, (%r9)
; AVX2-FAST-NEXT: movq {{[0-9]+}}(%rsp), %rax
; AVX2-FAST-NEXT: vmovaps %ymm6, 224(%rax)
-; AVX2-FAST-NEXT: vmovaps %ymm9, 192(%rax)
-; AVX2-FAST-NEXT: vmovaps %ymm13, 160(%rax)
+; AVX2-FAST-NEXT: vmovaps %ymm10, 192(%rax)
+; AVX2-FAST-NEXT: vmovaps %ymm14, 160(%rax)
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FAST-NEXT: vmovaps %ymm0, 128(%rax)
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FAST-NEXT: vmovaps %ymm0, 96(%rax)
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovaps %ymm0, 64(%rax)
+; AVX2-FAST-NEXT: vmovaps %ymm4, 64(%rax)
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FAST-NEXT: vmovaps %ymm0, 32(%rax)
; AVX2-FAST-NEXT: vmovaps %ymm1, (%rax)
-; AVX2-FAST-NEXT: addq $2472, %rsp # imm = 0x9A8
+; AVX2-FAST-NEXT: addq $2504, %rsp # imm = 0x9C8
; AVX2-FAST-NEXT: vzeroupper
; AVX2-FAST-NEXT: retq
;
; AVX2-FAST-PERLANE-LABEL: load_i32_stride6_vf64:
; AVX2-FAST-PERLANE: # %bb.0:
-; AVX2-FAST-PERLANE-NEXT: subq $2472, %rsp # imm = 0x9A8
+; AVX2-FAST-PERLANE-NEXT: subq $2504, %rsp # imm = 0x9C8
; AVX2-FAST-PERLANE-NEXT: vmovaps 672(%rdi), %ymm2
; AVX2-FAST-PERLANE-NEXT: vmovaps 640(%rdi), %ymm3
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovaps 608(%rdi), %ymm4
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovaps 320(%rdi), %ymm5
+; AVX2-FAST-PERLANE-NEXT: vmovaps 608(%rdi), %ymm5
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovaps 352(%rdi), %ymm6
+; AVX2-FAST-PERLANE-NEXT: vmovaps 320(%rdi), %ymm6
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovaps 288(%rdi), %ymm7
+; AVX2-FAST-PERLANE-NEXT: vmovaps 352(%rdi), %ymm7
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovaps 256(%rdi), %ymm8
+; AVX2-FAST-PERLANE-NEXT: vmovaps 288(%rdi), %ymm8
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovaps 256(%rdi), %ymm9
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovaps 224(%rdi), %ymm0
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovaps 192(%rdi), %ymm1
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovaps {{.*#+}} xmm9 = <0,6,4,u>
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm14 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm14, %ymm9, %ymm0
-; AVX2-FAST-PERLANE-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm8[0,1],ymm7[0,1]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm7[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovaps {{.*#+}} xmm4 = <0,6,4,u>
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm15 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm15, %ymm4, %ymm0
+; AVX2-FAST-PERLANE-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm9[0,1],ymm8[0,1]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm8[6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm1 = ymm1[0,2,2,2,4,6,6,6]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5],ymm0[6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm6[0,1,2,3],ymm5[4,5],ymm6[6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm7[0,1,2,3],ymm6[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vbroadcastsd {{.*#+}} ymm15 = [8589934596,8589934596,8589934596,8589934596]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm1, %ymm15, %ymm1
+; AVX2-FAST-PERLANE-NEXT: vbroadcastsd {{.*#+}} ymm9 = [8589934596,8589934596,8589934596,8589934596]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm1, %ymm9, %ymm1
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -8757,18 +8721,18 @@ define void @load_i32_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovaps 576(%rdi), %ymm0
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm4[4,5],ymm0[6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm5[4,5],ymm0[6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm0, %ymm9, %ymm0
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm0, %ymm4, %ymm0
; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm1 = ymm1[0,2,2,2,4,6,6,6]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5],ymm0[6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovaps 704(%rdi), %ymm1
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovaps 736(%rdi), %ymm2
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5],ymm2[6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm1, %ymm15, %ymm1
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm1, %ymm9, %ymm1
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovaps 1056(%rdi), %ymm1
@@ -8784,16 +8748,16 @@ define void @load_i32_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm0, %ymm9, %ymm0
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm0, %ymm4, %ymm0
; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm1 = ymm2[0,2,2,2,4,6,6,6]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5],ymm0[6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovaps 1088(%rdi), %ymm1
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, (%rsp) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovaps 1120(%rdi), %ymm2
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5],ymm2[6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm1, %ymm15, %ymm1
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm1, %ymm9, %ymm1
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovaps 1440(%rdi), %ymm1
@@ -8809,16 +8773,16 @@ define void @load_i32_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm0, %ymm9, %ymm0
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm0, %ymm4, %ymm0
; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm1 = ymm2[0,2,2,2,4,6,6,6]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5],ymm0[6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovaps 1472(%rdi), %ymm1
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovaps 1504(%rdi), %ymm2
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5],ymm2[6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm1, %ymm15, %ymm1
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm1, %ymm9, %ymm1
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovaps 96(%rdi), %ymm1
@@ -8826,21 +8790,21 @@ define void @load_i32_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-PERLANE-NEXT: vmovaps 64(%rdi), %ymm0
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[0,1],ymm1[0,1]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm12 = ymm0[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm13 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FAST-PERLANE-NEXT: vmovaps (%rdi), %ymm0
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovaps 32(%rdi), %ymm1
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm11 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm11, %ymm9, %ymm0
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm1 = ymm12[0,2,2,2,4,6,6,6]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5],ymm0[6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm12 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm12, %ymm4, %ymm0
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm1 = ymm13[0,2,2,2,4,6,6,6]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovaps 128(%rdi), %ymm1
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovaps 160(%rdi), %ymm2
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm10 = ymm2[0,1,2,3],ymm1[4,5],ymm2[6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm10, %ymm15, %ymm1
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm11 = ymm2[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm11, %ymm9, %ymm1
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovaps 480(%rdi), %ymm1
@@ -8848,21 +8812,21 @@ define void @load_i32_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-PERLANE-NEXT: vmovaps 448(%rdi), %ymm0
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[0,1],ymm1[0,1]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm8 = ymm0[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm10 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FAST-PERLANE-NEXT: vmovaps 416(%rdi), %ymm0
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovaps 384(%rdi), %ymm1
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm6 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm6, %ymm9, %ymm0
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm1 = ymm8[0,2,2,2,4,6,6,6]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5],ymm0[6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm7 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm7, %ymm4, %ymm0
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm1 = ymm10[0,2,2,2,4,6,6,6]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovaps 512(%rdi), %ymm1
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovaps 544(%rdi), %ymm2
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm7 = ymm2[0,1,2,3],ymm1[4,5],ymm2[6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm7, %ymm15, %ymm1
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm8 = ymm2[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm8, %ymm9, %ymm1
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovaps 864(%rdi), %ymm1
@@ -8870,101 +8834,101 @@ define void @load_i32_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-PERLANE-NEXT: vmovaps 832(%rdi), %ymm0
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[0,1],ymm1[0,1]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm0[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm6 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FAST-PERLANE-NEXT: vmovaps 800(%rdi), %ymm0
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovaps 768(%rdi), %ymm1
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm3, %ymm9, %ymm0
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm1 = ymm5[0,2,2,2,4,6,6,6]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5],ymm0[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm3, %ymm4, %ymm0
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm1 = ymm6[0,2,2,2,4,6,6,6]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovaps 896(%rdi), %ymm1
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovaps 928(%rdi), %ymm2
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm2[0,1,2,3],ymm1[4,5],ymm2[6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm4, %ymm15, %ymm1
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm2[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm5, %ymm9, %ymm1
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovaps 1184(%rdi), %ymm0
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, (%rsp) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovaps 1152(%rdi), %ymm1
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm2, %ymm9, %ymm0
-; AVX2-FAST-PERLANE-NEXT: vmovaps 1248(%rdi), %ymm9
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm2, %ymm4, %ymm0
+; AVX2-FAST-PERLANE-NEXT: vmovaps 1248(%rdi), %ymm4
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovaps 1216(%rdi), %ymm1
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm1[0,1],ymm9[0,1]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm9 = ymm1[0,1,2,3,4,5],ymm9[6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm1 = ymm9[0,2,2,2,4,6,6,6]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5],ymm0[6,7]
+; AVX2-FAST-PERLANE-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm1[0,1],ymm4[0,1]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm1[0,1,2,3,4,5],ymm4[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm1 = ymm4[0,2,2,2,4,6,6,6]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovaps 1280(%rdi), %ymm1
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovaps 1312(%rdi), %ymm13
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm13[0,1,2,3],ymm1[4,5],ymm13[6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm1, %ymm15, %ymm15
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm15[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovaps 1312(%rdi), %ymm14
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm14[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm1, %ymm9, %ymm9
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm9[6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovaps {{.*#+}} xmm15 = <1,7,5,u>
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm14, %ymm15, %ymm0
-; AVX2-FAST-PERLANE-NEXT: vpermilps $237, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm13 = mem[1,3,2,3,5,7,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm13 = ymm0[0,1,2],ymm13[3,4,5],ymm0[6,7]
-; AVX2-FAST-PERLANE-NEXT: vbroadcastsd {{.*#+}} ymm0 = [12884901893,12884901893,12884901893,12884901893]
-; AVX2-FAST-PERLANE-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm14 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3,4,5],ymm14[6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm13 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: vpermilps $237, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm14 = mem[1,3,2,3,5,7,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0,1,2],ymm14[3,4,5],ymm13[6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm14 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3,4,5],ymm14[6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm13 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: vpermilps $237, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm14 = mem[1,3,2,3,5,7,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0,1,2],ymm14[3,4,5],ymm13[6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm14 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3,4,5],ymm14[6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm13 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: vmovaps {{.*#+}} xmm9 = <1,7,5,u>
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm15, %ymm9, %ymm0
; AVX2-FAST-PERLANE-NEXT: vpermilps $237, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # ymm14 = mem[1,3,2,3,5,7,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0,1,2],ymm14[3,4,5],ymm13[6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm14 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3,4,5],ymm14[6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm11, %ymm15, %ymm11
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm12 = ymm12[1,3,2,3,5,7,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm11 = ymm11[0,1,2],ymm12[3,4,5],ymm11[6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm10, %ymm0, %ymm10
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm10 = ymm11[0,1,2,3,4,5],ymm10[6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm6, %ymm15, %ymm6
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm8 = ymm8[1,3,2,3,5,7,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2],ymm8[3,4,5],ymm6[6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm7, %ymm0, %ymm7
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5],ymm7[6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm3, %ymm15, %ymm3
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm5 = ymm5[1,3,2,3,5,7,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2],ymm5[3,4,5],ymm3[6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm4, %ymm0, %ymm4
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm4[6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm14 = ymm0[0,1,2],ymm14[3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vbroadcastsd {{.*#+}} ymm0 = [12884901893,12884901893,12884901893,12884901893]
+; AVX2-FAST-PERLANE-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4,5],ymm15[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm14 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: vpermilps $237, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm15 = mem[1,3,2,3,5,7,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1,2],ymm15[3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4,5],ymm15[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm14 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: vpermilps $237, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm15 = mem[1,3,2,3,5,7,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1,2],ymm15[3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4,5],ymm15[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm14 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: vpermilps $237, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm15 = mem[1,3,2,3,5,7,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1,2],ymm15[3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4,5],ymm15[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm12, %ymm9, %ymm12
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm13 = ymm13[1,3,2,3,5,7,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm12 = ymm12[0,1,2],ymm13[3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm11, %ymm0, %ymm11
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm11 = ymm12[0,1,2,3,4,5],ymm11[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm7, %ymm9, %ymm7
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm10 = ymm10[1,3,2,3,5,7,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2],ymm10[3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm8, %ymm0, %ymm8
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5],ymm8[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm3, %ymm9, %ymm3
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm6 = ymm6[1,3,2,3,5,7,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2],ymm6[3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm5, %ymm0, %ymm5
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm5[6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm2, %ymm15, %ymm2
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm3 = ymm9[1,3,2,3,5,7,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm3[3,4,5],ymm2[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm2, %ymm9, %ymm2
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm3 = ymm4[1,3,2,3,5,7,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm3[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpermps %ymm1, %ymm0, %ymm0
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3,4,5],ymm0[6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm0 = mem[2,3,2,3,6,7,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovaps 192(%rdi), %xmm6
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} xmm0 = xmm6[2,3,2,3]
; AVX2-FAST-PERLANE-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # ymm1 = mem[0,0,2,3,4,4,6,7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
@@ -8974,27 +8938,28 @@ define void @load_i32_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm1 = ymm1[2,0,2,3,6,4,6,7]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm1 = mem[0,0,0,0,4,4,4,4]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm1 = ymm5[0,0,0,0,4,4,4,4]
; AVX2-FAST-PERLANE-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # ymm2 = mem[0,0,2,3,4,4,6,7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm0 = mem[2,3,2,3,6,7,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovaps 576(%rdi), %xmm0
+; AVX2-FAST-PERLANE-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[2,3,2,3]
; AVX2-FAST-PERLANE-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # ymm1 = mem[0,0,2,3,4,4,6,7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps $204, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm1 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm1 = ymm9[0,1],mem[2,3],ymm9[4,5],mem[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm13[0,1],ymm4[2,3],ymm13[4,5],ymm4[6,7]
; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm1 = ymm1[2,0,2,3,6,4,6,7]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # ymm1 = mem[0,0,0,0,4,4,4,4]
; AVX2-FAST-PERLANE-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
@@ -9003,88 +8968,50 @@ define void @load_i32_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm0 = ymm8[2,3,2,3,6,7,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm1 = ymm6[0,0,2,3,4,4,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm14[0,1],ymm15[2,3],ymm14[4,5],ymm15[6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm1 = ymm1[2,0,2,3,6,4,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
-; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm1 = ymm4[0,0,0,0,4,4,4,4]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm2 = ymm5[0,0,2,3,4,4,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
-; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm0 = ymm13[2,3,2,3,6,7,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovaps 960(%rdi), %xmm0
+; AVX2-FAST-PERLANE-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[2,3,2,3]
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm1 = ymm12[0,0,2,3,4,4,6,7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm1 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm1 = mem[0,1],ymm7[2,3],mem[4,5],ymm7[6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm1 = ymm1[2,0,2,3,6,4,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
-; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm1 = ymm11[0,0,0,0,4,4,4,4]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm2 = ymm10[0,0,2,3,4,4,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
-; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm0 = mem[2,3,2,3,6,7,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermilps $224, (%rsp), %ymm1 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm1 = mem[0,0,2,3,4,4,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # ymm1 = mem[0,1],ymm1[2,3],mem[4,5],ymm1[6,7]
; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm1 = ymm1[2,0,2,3,6,4,6,7]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm1 = mem[0,0,0,0,4,4,4,4]
-; AVX2-FAST-PERLANE-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm1 = ymm3[0,0,0,0,4,4,4,4]
+; AVX2-FAST-PERLANE-NEXT: vpermilps $224, (%rsp), %ymm2 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # ymm2 = mem[0,0,2,3,4,4,6,7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm0 = mem[2,3,2,3,6,7,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovaps 1344(%rdi), %xmm0
+; AVX2-FAST-PERLANE-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[2,3,2,3]
; AVX2-FAST-PERLANE-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # ymm1 = mem[0,0,2,3,4,4,6,7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm1 = mem[0,1],ymm1[2,3],mem[4,5],ymm1[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm11[0,1],ymm14[2,3],ymm11[4,5],ymm14[6,7]
; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm1 = ymm1[2,0,2,3,6,4,6,7]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # ymm1 = mem[0,0,0,0,4,4,4,4]
-; AVX2-FAST-PERLANE-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm2 = mem[0,0,2,3,4,4,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm2 = ymm15[0,0,2,3,4,4,6,7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm0 = mem[2,3,2,3,6,7,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovaps 1152(%rdi), %xmm2
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} xmm0 = xmm2[2,3,2,3]
; AVX2-FAST-PERLANE-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # ymm1 = mem[0,0,2,3,4,4,6,7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
@@ -9094,17 +9021,18 @@ define void @load_i32_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm1 = ymm1[2,0,2,3,6,4,6,7]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # ymm1 = mem[0,0,0,0,4,4,4,4]
-; AVX2-FAST-PERLANE-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm2 = mem[0,0,2,3,4,4,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
+; AVX2-FAST-PERLANE-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm7 = mem[0,0,2,3,4,4,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm7[0,1,2],ymm1[3],ymm7[4,5,6],ymm1[7]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm0 = mem[2,3,2,3,6,7,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovaps (%rdi), %xmm0
+; AVX2-FAST-PERLANE-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[2,3,2,3]
; AVX2-FAST-PERLANE-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # ymm1 = mem[0,0,2,3,4,4,6,7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
@@ -9114,155 +9042,195 @@ define void @load_i32_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm1 = ymm1[2,0,2,3,6,4,6,7]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # ymm1 = mem[0,0,0,0,4,4,4,4]
-; AVX2-FAST-PERLANE-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm2 = mem[0,0,2,3,4,4,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
+; AVX2-FAST-PERLANE-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm8 = mem[0,0,2,3,4,4,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm8[0,1,2],ymm1[3],ymm8[4,5,6],ymm1[7]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm0 = mem[3,3,3,3,7,7,7,7]
-; AVX2-FAST-PERLANE-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm1 = mem[0,1,3,3,4,5,7,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm1 = mem[3,3,3,3,7,7,7,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm1 = ymm1[0],mem[1],ymm1[2,3,4],mem[5],ymm1[6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
-; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm1 = mem[0,1,0,1,4,5,4,5]
-; AVX2-FAST-PERLANE-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm2 = mem[0,1,3,3,4,5,7,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
-; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm0 = mem[3,3,3,3,7,7,7,7]
-; AVX2-FAST-PERLANE-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm1 = mem[0,1,3,3,4,5,7,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm1 = mem[3,3,3,3,7,7,7,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm9[1],ymm1[2,3,4],ymm9[5],ymm1[6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
-; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm1 = mem[0,1,0,1,4,5,4,5]
-; AVX2-FAST-PERLANE-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm2 = mem[0,1,3,3,4,5,7,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
-; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm0 = ymm8[3,3,3,3,7,7,7,7]
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm1 = ymm6[0,1,3,3,4,5,7,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm1 = ymm15[3,3,3,3,7,7,7,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm14[1],ymm1[2,3,4],ymm14[5],ymm1[6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
-; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm1 = ymm4[0,1,0,1,4,5,4,5]
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm2 = ymm5[0,1,3,3,4,5,7,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
-; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm0 = ymm13[3,3,3,3,7,7,7,7]
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm1 = ymm12[0,1,3,3,4,5,7,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm1 = ymm7[3,3,3,3,7,7,7,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm1 = ymm1[0],mem[1],ymm1[2,3,4],mem[5],ymm1[6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
-; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm1 = ymm11[0,1,0,1,4,5,4,5]
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm2 = ymm10[0,1,3,3,4,5,7,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
-; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm0 = mem[3,3,3,3,7,7,7,7]
-; AVX2-FAST-PERLANE-NEXT: vpermilps $244, (%rsp), %ymm1 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm1 = mem[0,1,3,3,4,5,7,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm1 = mem[3,3,3,3,7,7,7,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm1 = ymm1[0],mem[1],ymm1[2,3,4],mem[5],ymm1[6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
+; AVX2-FAST-PERLANE-NEXT: vmovaps 384(%rdi), %xmm0
+; AVX2-FAST-PERLANE-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[2,3,2,3]
+; AVX2-FAST-PERLANE-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm8 = mem[0,0,2,3,4,4,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm8[1,2,3],ymm0[4],ymm8[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm8 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm8 = mem[0,1],ymm1[2,3],mem[4,5],ymm1[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm8 = ymm8[2,0,2,3,6,4,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm8 = ymm8[0,2,0,3]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm1 = mem[0,1,0,1,4,5,4,5]
-; AVX2-FAST-PERLANE-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm2 = mem[0,1,3,3,4,5,7,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
-; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm8[3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm8 = mem[0,0,0,0,4,4,4,4]
+; AVX2-FAST-PERLANE-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm9 = mem[0,0,2,3,4,4,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm8 = ymm9[0,1,2],ymm8[3],ymm9[4,5,6],ymm8[7]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm8 = ymm8[0,1,0,3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm8[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm0 = ymm12[3,3,3,3,7,7,7,7]
+; AVX2-FAST-PERLANE-NEXT: vmovaps 768(%rdi), %xmm0
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} xmm8 = xmm0[2,3,2,3]
+; AVX2-FAST-PERLANE-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm9 = mem[0,0,2,3,4,4,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0],ymm9[1,2,3],ymm8[4],ymm9[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm9 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm9 = mem[0,1],ymm1[2,3],mem[4,5],ymm1[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm9 = ymm9[2,0,2,3,6,4,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm9 = ymm9[0,2,0,3]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm8 = ymm8[0,3,2,3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2],ymm9[3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm9 = mem[0,0,0,0,4,4,4,4]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm10 = ymm7[0,0,2,3,4,4,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm9 = ymm10[0,1,2],ymm9[3],ymm10[4,5,6],ymm9[7]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm9 = ymm9[0,1,0,3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4],ymm9[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} xmm6 = xmm6[3,3,3,3]
+; AVX2-FAST-PERLANE-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm8 = mem[0,1,3,3,4,5,7,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0],ymm8[1,2,3],ymm6[4],ymm8[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm8 = mem[3,3,3,3,7,7,7,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm8 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm8 = ymm8[0],mem[1],ymm8[2,3,4],mem[5],ymm8[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm8 = ymm8[0,2,0,3]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm6 = ymm6[0,3,2,3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2],ymm8[3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm8 = ymm5[0,1,0,1,4,5,4,5]
+; AVX2-FAST-PERLANE-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm9 = mem[0,1,3,3,4,5,7,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm8 = ymm9[0,1,2],ymm8[3],ymm9[4,5,6],ymm8[7]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm8 = ymm8[0,1,0,3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm6[0,1,2,3,4],ymm8[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # xmm5 = mem[3,3,3,3]
+; AVX2-FAST-PERLANE-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm6 = mem[0,1,3,3,4,5,7,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0],ymm6[1,2,3],ymm5[4],ymm6[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm6 = ymm4[3,3,3,3,7,7,7,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0],ymm13[1],ymm6[2,3,4],ymm13[5],ymm6[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm6 = ymm6[0,2,0,3]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm5 = ymm5[0,3,2,3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2],ymm6[3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm6 = mem[0,1,0,1,4,5,4,5]
+; AVX2-FAST-PERLANE-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm8 = mem[0,1,3,3,4,5,7,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm6 = ymm8[0,1,2],ymm6[3],ymm8[4,5,6],ymm6[7]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm6 = ymm6[0,1,0,3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm5[0,1,2,3,4],ymm6[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # xmm4 = mem[3,3,3,3]
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm5 = ymm12[0,1,3,3,4,5,7,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0],ymm5[1,2,3],ymm4[4],ymm5[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm5 = mem[3,3,3,3,7,7,7,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm5 = ymm5[0],mem[1],ymm5[2,3,4],mem[5],ymm5[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm5 = ymm5[0,2,0,3]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm4 = ymm4[0,3,2,3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2],ymm5[3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm5 = ymm3[0,1,0,1,4,5,4,5]
+; AVX2-FAST-PERLANE-NEXT: vpermilps $244, (%rsp), %ymm6 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm6 = mem[0,1,3,3,4,5,7,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm6[0,1,2],ymm5[3],ymm6[4,5,6],ymm5[7]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm5 = ymm5[0,1,0,3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm4[0,1,2,3,4],ymm5[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # xmm3 = mem[3,3,3,3]
+; AVX2-FAST-PERLANE-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm4 = mem[0,1,3,3,4,5,7,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0],ymm4[1,2,3],ymm3[4],ymm4[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm4 = ymm14[3,3,3,3,7,7,7,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0],ymm11[1],ymm4[2,3,4],ymm11[5],ymm4[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm4 = ymm4[0,2,0,3]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm3 = ymm3[0,3,2,3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2],ymm4[3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm4 = mem[0,1,0,1,4,5,4,5]
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm5 = ymm15[0,1,3,3,4,5,7,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm5[0,1,2],ymm4[3],ymm5[4,5,6],ymm4[7]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm4 = ymm4[0,1,0,3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4],ymm4[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} xmm2 = xmm2[3,3,3,3]
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm1 = ymm11[0,1,3,3,4,5,7,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm1 = mem[3,3,3,3,7,7,7,7]
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm3 = ymm11[0,1,3,3,4,5,7,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0],ymm3[1,2,3],ymm2[4],ymm3[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm13[1],ymm1[2,3,4],ymm13[5],ymm1[6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm3 = ymm13[3,3,3,3,7,7,7,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0],ymm12[1],ymm3[2,3,4],ymm12[5],ymm3[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm3 = ymm3[0,2,0,3]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,3,2,3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm3[3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm3 = mem[0,1,0,1,4,5,4,5]
+; AVX2-FAST-PERLANE-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm4 = mem[0,1,3,3,4,5,7,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3],ymm4[4,5,6],ymm3[7]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm3 = ymm3[0,1,0,3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm2 = ymm9[0,1,3,3,4,5,7,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm2[1,2,3],ymm0[4],ymm2[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm2 = ymm1[3,3,3,3,7,7,7,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0],ymm14[1],ymm2[2,3,4],ymm14[5],ymm2[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,2,0,3]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm1 = mem[0,1,0,1,4,5,4,5]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm2 = ymm15[0,1,3,3,4,5,7,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
-; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm2[3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm2 = ymm10[0,1,0,1,4,5,4,5]
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm3 = ymm7[0,1,3,3,4,5,7,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3],ymm3[4,5,6],ymm2[7]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,1,0,3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm0 = ymm6[3,3,3,3,7,7,7,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm1 = ymm5[0,1,3,3,4,5,7,7]
+; AVX2-FAST-PERLANE-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # xmm0 = mem[3,3,3,3]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm1 = ymm3[0,1,3,3,4,5,7,7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm1 = ymm8[3,3,3,3,7,7,7,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm7[1],ymm1[2,3,4],ymm7[5],ymm1[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm1 = ymm5[3,3,3,3,7,7,7,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm4[1],ymm1[2,3,4],ymm4[5],ymm1[6,7]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm1 = ymm10[0,1,0,1,4,5,4,5]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm2 = ymm9[0,1,3,3,4,5,7,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm1 = ymm7[0,1,0,1,4,5,4,5]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm2 = ymm6[0,1,3,3,4,5,7,7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm0 = mem[3,3,3,3,7,7,7,7]
+; AVX2-FAST-PERLANE-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # xmm0 = mem[3,3,3,3]
; AVX2-FAST-PERLANE-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # ymm1 = mem[0,1,3,3,4,5,7,7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # ymm1 = mem[3,3,3,3,7,7,7,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm4[1],ymm1[2,3,4],ymm4[5],ymm1[6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm1 = ymm1[0],mem[1],ymm1[2,3,4],mem[5],ymm1[6,7]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # ymm1 = mem[0,1,0,1,4,5,4,5]
; AVX2-FAST-PERLANE-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
@@ -9272,146 +9240,147 @@ define void @load_i32_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm2 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm2 = ymm0[0,1,2,3],mem[4,5],ymm0[6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm1 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm1 = mem[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm1 = ymm0[0,1,2,3],mem[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm2 = mem[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovaps 272(%rdi), %xmm0
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3],ymm2[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,1,0,2,4,5,4,6]
-; AVX2-FAST-PERLANE-NEXT: vmovaps {{.*#+}} xmm1 = <4,2,u,u>
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm2, %ymm1, %ymm2
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm0[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovaps {{.*#+}} xmm8 = <4,2,u,u>
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm1, %ymm8, %ymm1
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm0[2,3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm14 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm14 = mem[0,1],ymm0[2,3],mem[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm15 = mem[0,1],ymm0[2,3],mem[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vbroadcastf128 {{.*#+}} ymm0 = [0,2,0,6,0,2,0,6]
; AVX2-FAST-PERLANE-NEXT: # ymm0 = mem[0,1,0,1]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm14, %ymm0, %ymm3
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm15, %ymm0, %ymm2
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm2 = ymm3[0,1,2,3],mem[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm5[0,1,2,3],ymm6[4,5],ymm5[6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm4[0,1,2,3],ymm5[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm7[0,1,2,3],ymm8[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovaps 464(%rdi), %xmm2
+; AVX2-FAST-PERLANE-NEXT: vmovaps 464(%rdi), %xmm1
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm3[0,1],ymm1[2,3],ymm3[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm2, %ymm8, %ymm2
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm7[0,1],ymm6[2,3],ymm7[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm5[0,1],ymm2[2,3],ymm5[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm2 = ymm2[0,1,0,2,4,5,4,6]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm3, %ymm1, %ymm3
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm10[0,1],ymm9[2,3],ymm10[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm3, %ymm0, %ymm3
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm2, %ymm0, %ymm2
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm2 = ymm1[0,1,2,3],mem[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm3 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm3 = mem[0,1,2,3],ymm2[4,5],mem[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm3 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm3 = ymm1[0,1,2,3],mem[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovaps 656(%rdi), %xmm1
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm3[0,1],ymm1[2,3],ymm3[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm2, %ymm8, %ymm2
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm5 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm5 = ymm2[0,1,2,3],mem[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovaps 656(%rdi), %xmm2
+; AVX2-FAST-PERLANE-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm2 = mem[0,1],ymm2[2,3],mem[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm5[0,1],ymm2[2,3],ymm5[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm2 = ymm2[0,1,0,2,4,5,4,6]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm3, %ymm1, %ymm3
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm3 = mem[0,1],ymm3[2,3],mem[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm3, %ymm0, %ymm3
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm2, %ymm0, %ymm2
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm2 = ymm9[0,1,2,3],mem[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm11[0,1,2,3],ymm12[4,5],ymm11[6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm3 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm3 = ymm14[0,1,2,3],mem[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm5 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm5 = ymm13[0,1,2,3],mem[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovaps 848(%rdi), %xmm2
+; AVX2-FAST-PERLANE-NEXT: vmovaps 848(%rdi), %xmm1
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm3[0,1],ymm1[2,3],ymm3[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm2, %ymm8, %ymm2
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm2 = ymm10[0,1],mem[2,3],ymm10[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm5[0,1],ymm2[2,3],ymm5[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm2 = ymm2[0,1,0,2,4,5,4,6]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm3, %ymm1, %ymm3
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm3 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm3 = mem[0,1],ymm15[2,3],mem[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm3, %ymm0, %ymm3
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm2, %ymm0, %ymm2
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm2 = ymm1[0,1,2,3],mem[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm3 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm3 = ymm2[0,1,2,3],mem[4,5],ymm2[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm3 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm3 = ymm1[0,1,2,3],mem[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm5 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm5 = ymm2[0,1,2,3],mem[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovaps 1040(%rdi), %xmm2
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm5[0,1],ymm2[2,3],ymm5[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm2 = ymm2[0,1,0,2,4,5,4,6]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm3, %ymm1, %ymm3
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm13 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm13 = mem[0,1],ymm3[2,3],mem[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm13, %ymm0, %ymm3
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovaps 1040(%rdi), %xmm1
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm3[0,1],ymm1[2,3],ymm3[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm2, %ymm8, %ymm2
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups (%rsp), %ymm2 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm3 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm3 = ymm2[0,1,2,3],mem[4,5],ymm2[6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm3, (%rsp) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm10 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm10 = ymm2[0,1,2,3],mem[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovaps 1232(%rdi), %xmm12
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm10[0,1],ymm12[2,3],ymm10[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm2 = ymm2[0,1,0,2,4,5,4,6]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm3, %ymm1, %ymm3
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm9 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm9 = mem[0,1],ymm3[2,3],mem[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm9, %ymm0, %ymm3
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm14 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm14 = mem[0,1],ymm2[2,3],mem[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm14, %ymm0, %ymm2
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm2 = ymm11[0,1,2,3],mem[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm11 = ymm12[0,1,2,3],ymm13[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovaps 1232(%rdi), %xmm1
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm11[0,1],ymm1[2,3],ymm11[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm2, %ymm8, %ymm2
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm11 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm11 = ymm2[0,1,2,3],mem[4,5],ymm2[6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm7 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm7 = ymm2[0,1,2,3],mem[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovaps 1424(%rdi), %xmm8
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm7[0,1],ymm8[2,3],ymm7[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm2 = ymm2[0,1,0,2,4,5,4,6]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm11, %ymm1, %ymm3
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm6 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm6 = mem[0,1],ymm3[2,3],mem[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm6, %ymm0, %ymm3
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm5 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm5 = ymm4[0,1,2,3],mem[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm10 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm10 = mem[0,1],ymm2[2,3],mem[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm10, %ymm0, %ymm2
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm13 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm13 = ymm1[0,1,2,3],mem[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm9 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm9 = ymm1[0,1,2,3],mem[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovaps 1424(%rdi), %xmm12
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm9[0,1],ymm12[2,3],ymm9[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm13, %ymm8, %ymm2
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm4 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm4 = ymm2[0,1,2,3],mem[4,5],ymm2[6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm6 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm6 = mem[0,1],ymm2[2,3],mem[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm6, %ymm0, %ymm2
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm5 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm5 = ymm1[0,1,2,3],mem[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm4 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm4 = ymm1[0,1,2,3],mem[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovaps 80(%rdi), %xmm2
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm5[0,1],ymm2[2,3],ymm5[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm3 = ymm3[0,1,0,2,4,5,4,6]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm4, %ymm1, %ymm1
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm3[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm5[0,1],ymm2[2,3],ymm5[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm4, %ymm8, %ymm3
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm3[0,1],ymm1[2,3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # ymm3 = mem[0,1],ymm3[2,3],mem[4,5,6,7]
@@ -9423,57 +9392,58 @@ define void @load_i32_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-PERLANE-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # ymm0 = ymm0[0,1,2],mem[3],ymm0[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovaps {{.*#+}} xmm1 = <5,3,u,u>
-; AVX2-FAST-PERLANE-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm15 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm15 = ymm15[0,1],ymm0[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm7 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1],ymm0[2,3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vbroadcastf128 {{.*#+}} ymm0 = [0,3,1,7,0,3,1,7]
; AVX2-FAST-PERLANE-NEXT: # ymm0 = mem[0,1,0,1]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm14, %ymm0, %ymm14
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm14 = ymm15[0,1,2,3,4],ymm14[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm14 = mem[1,1,1,1,5,5,5,5]
-; AVX2-FAST-PERLANE-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm14 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm14 = ymm14[0,1,2],mem[3],ymm14[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm15, %ymm0, %ymm15
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4],ymm15[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm7 = mem[1,1,1,1,5,5,5,5]
+; AVX2-FAST-PERLANE-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm7 = ymm7[0,1,2],mem[3],ymm7[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm15 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm14 = ymm15[0,1],ymm14[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm7 = ymm15[0,1],ymm7[2,3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4],ymm15[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm14 = mem[1,1,1,1,5,5,5,5]
-; AVX2-FAST-PERLANE-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm14 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm14 = ymm14[0,1,2],mem[3],ymm14[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4],ymm15[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm7 = mem[1,1,1,1,5,5,5,5]
+; AVX2-FAST-PERLANE-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm7 = ymm7[0,1,2],mem[3],ymm7[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm15 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm14 = ymm15[0,1],ymm14[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm7 = ymm15[0,1],ymm7[2,3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4],ymm15[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm14 = mem[1,1,1,1,5,5,5,5]
-; AVX2-FAST-PERLANE-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm14 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm14 = ymm14[0,1,2],mem[3],ymm14[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4],ymm15[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm7 = mem[1,1,1,1,5,5,5,5]
+; AVX2-FAST-PERLANE-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm7 = ymm7[0,1,2],mem[3],ymm7[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm15 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm14 = ymm15[0,1],ymm14[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm7 = ymm15[0,1],ymm7[2,3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4],ymm15[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm14 = mem[1,1,1,1,5,5,5,5]
-; AVX2-FAST-PERLANE-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm14 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm14 = ymm14[0,1,2],mem[3],ymm14[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm15 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm14 = ymm15[0,1],ymm14[2,3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm13, %ymm0, %ymm13
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm13 = ymm14[0,1,2,3,4],ymm13[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm10 = ymm10[1,1,1,1,5,5,5,5]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm10 = ymm10[0,1,2],ymm12[3],ymm10[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps (%rsp), %ymm1, %ymm12 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm10 = ymm12[0,1],ymm10[2,3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm9, %ymm0, %ymm9
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm9 = ymm10[0,1,2,3,4],ymm9[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm7 = ymm7[1,1,1,1,5,5,5,5]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2],ymm8[3],ymm7[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm11, %ymm1, %ymm8
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4],ymm15[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm7, (%rsp) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm15 = mem[1,1,1,1,5,5,5,5]
+; AVX2-FAST-PERLANE-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm15 = ymm15[0,1,2],mem[3],ymm15[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm7 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1],ymm15[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm14, %ymm0, %ymm14
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm14 = ymm7[0,1,2,3,4],ymm14[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm7 = ymm11[1,1,1,1,5,5,5,5]
+; AVX2-FAST-PERLANE-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm7 = ymm7[0,1,2],mem[3],ymm7[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm11 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm7 = ymm11[0,1],ymm7[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm10, %ymm0, %ymm10
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm10 = ymm7[0,1,2,3,4],ymm10[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm7 = ymm9[1,1,1,1,5,5,5,5]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2],ymm12[3],ymm7[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm13, %ymm1, %ymm8
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm7 = ymm8[0,1],ymm7[2,3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpermps %ymm6, %ymm0, %ymm6
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm6 = ymm7[0,1,2,3,4],ymm6[5,6,7]
@@ -9565,9 +9535,9 @@ define void @load_i32_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm0, (%r9)
; AVX2-FAST-PERLANE-NEXT: movq {{[0-9]+}}(%rsp), %rax
; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm6, 224(%rax)
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm9, 192(%rax)
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm13, 160(%rax)
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm10, 192(%rax)
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm14, 160(%rax)
+; AVX2-FAST-PERLANE-NEXT: vmovups (%rsp), %ymm0 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm0, 128(%rax)
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm0, 96(%rax)
@@ -9576,7 +9546,7 @@ define void @load_i32_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm0, 32(%rax)
; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm1, (%rax)
-; AVX2-FAST-PERLANE-NEXT: addq $2472, %rsp # imm = 0x9A8
+; AVX2-FAST-PERLANE-NEXT: addq $2504, %rsp # imm = 0x9C8
; AVX2-FAST-PERLANE-NEXT: vzeroupper
; AVX2-FAST-PERLANE-NEXT: retq
;
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-7.ll b/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-7.ll
index f4cf3b2188993..fb83aeabccd13 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-7.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-7.ll
@@ -101,7 +101,7 @@ define void @load_i32_stride7_vf2(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm3 = xmm4[0,1,2],xmm3[3]
; AVX2-ONLY-NEXT: vpermilps {{.*#+}} xmm3 = xmm3[3,2,2,3]
; AVX2-ONLY-NEXT: vmovaps {{.*#+}} xmm4 = <4,3,u,u>
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX2-ONLY-NEXT: vpermps %ymm7, %ymm4, %ymm4
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm1[0],ymm0[1],ymm1[2,3,4],ymm0[5],ymm1[6,7]
; AVX2-ONLY-NEXT: vpermilps {{.*#+}} ymm7 = ymm7[1,0,2,3,5,4,6,7]
@@ -422,7 +422,7 @@ define void @load_i32_stride7_vf4(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
; AVX2-SLOW-NEXT: vbroadcastss 100(%rdi), %xmm9
; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm9 = xmm8[0,1,2],xmm9[3]
; AVX2-SLOW-NEXT: vmovaps {{.*#+}} xmm10 = <4,3,u,u>
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm11 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm11 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX2-SLOW-NEXT: vpermps %ymm11, %ymm10, %ymm10
; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm9 = xmm10[0,1],xmm9[2,3]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm8 = xmm6[0,1,2],xmm8[3]
@@ -479,7 +479,7 @@ define void @load_i32_stride7_vf4(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
; AVX2-FAST-NEXT: vbroadcastss 100(%rdi), %xmm9
; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm9 = xmm8[0,1,2],xmm9[3]
; AVX2-FAST-NEXT: vmovaps {{.*#+}} xmm10 = <4,3,u,u>
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm11 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm11 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX2-FAST-NEXT: vpermps %ymm11, %ymm10, %ymm10
; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm9 = xmm10[0,1],xmm9[2,3]
; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm8 = xmm6[0,1,2],xmm8[3]
@@ -536,7 +536,7 @@ define void @load_i32_stride7_vf4(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
; AVX2-FAST-PERLANE-NEXT: vbroadcastss 100(%rdi), %xmm9
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm9 = xmm8[0,1,2],xmm9[3]
; AVX2-FAST-PERLANE-NEXT: vmovaps {{.*#+}} xmm10 = <4,3,u,u>
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm11 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm11 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpermps %ymm11, %ymm10, %ymm10
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm9 = xmm10[0,1],xmm9[2,3]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm8 = xmm6[0,1,2],xmm8[3]
@@ -748,8 +748,8 @@ define void @load_i32_stride7_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
;
; AVX1-ONLY-LABEL: load_i32_stride7_vf8:
; AVX1-ONLY: # %bb.0:
-; AVX1-ONLY-NEXT: vmovaps 128(%rdi), %ymm7
; AVX1-ONLY-NEXT: vmovaps 160(%rdi), %ymm4
+; AVX1-ONLY-NEXT: vmovaps 128(%rdi), %ymm8
; AVX1-ONLY-NEXT: vmovaps 64(%rdi), %ymm11
; AVX1-ONLY-NEXT: vmovaps 32(%rdi), %ymm0
; AVX1-ONLY-NEXT: vmovaps (%rdi), %ymm1
@@ -762,31 +762,31 @@ define void @load_i32_stride7_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
; AVX1-ONLY-NEXT: vmovaps 32(%rdi), %xmm9
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm5 = xmm13[0,1],xmm5[2,3]
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm5 = xmm5[0,3,2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm5[0,1,2],ymm3[3,4],ymm5[5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm5[0,1,2],ymm3[3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps 160(%rdi), %xmm5
; AVX1-ONLY-NEXT: vmovaps 128(%rdi), %xmm6
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm8 = xmm6[1],xmm5[1]
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm7 = xmm6[1],xmm5[1]
; AVX1-ONLY-NEXT: vmovaps 192(%rdi), %xmm10
-; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm8 = zero,xmm8[1,2],xmm10[1]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm8, %ymm0, %ymm8
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4],ymm8[5,6,7]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm8 = ymm12[1,1],ymm11[2,2],ymm12[5,5],ymm11[6,6]
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm8 = ymm8[2,3,2,3]
+; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm7 = zero,xmm7[1,2],xmm10[1]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm7, %ymm0, %ymm7
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4],ymm7[5,6,7]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm7 = ymm12[1,1],ymm11[2,2],ymm12[5,5],ymm11[6,6]
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm7 = ymm7[2,3,2,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm14 = xmm9[0],xmm13[1],xmm9[2,3]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm14 = xmm14[1,0],mem[3,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm14[0,1,2],ymm8[3,4],ymm14[5,6,7]
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm14 = ymm4[2,3,0,1]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm14 = ymm7[0,0],ymm14[3,3],ymm7[4,4],ymm14[7,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm14[0,1,2],ymm7[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm14 = ymm8[2,3],ymm4[0,1]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm14 = ymm8[0,0],ymm14[3,3],ymm8[4,4],ymm14[7,7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm14, %xmm14
; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm14 = zero,xmm14[1,2],xmm10[2]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm14, %ymm0, %ymm14
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4],ymm14[5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4],ymm14[5,6,7]
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm14 = xmm13[2,3,2,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm14 = xmm14[0],xmm9[1],xmm14[2,3]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm2[3,1],ymm11[0,3],ymm2[7,5],ymm11[4,7]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm12[2,1],ymm15[2,0],ymm12[6,5],ymm15[6,4]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm15 = ymm14[0,1],ymm15[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm14 = ymm7[0],ymm4[0],ymm7[2],ymm4[2]
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm14 = ymm8[0],ymm4[0],ymm8[2],ymm4[2]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm14, %xmm14
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm10 = xmm14[0,1,2],xmm10[3]
; AVX1-ONLY-NEXT: vmovaps 192(%rdi), %ymm14
@@ -799,19 +799,19 @@ define void @load_i32_stride7_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm12 = xmm12[3,2,2,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm11 = ymm12[0,1],ymm11[2,3,4,5,6,7]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm12 = ymm14[0,1],ymm4[1,3],ymm14[4,5],ymm4[5,7]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm12 = ymm7[0,2],ymm12[2,0],ymm7[4,6],ymm12[6,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm12 = ymm8[0,2],ymm12[2,0],ymm8[4,6],ymm12[6,4]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm11 = ymm11[0,1,2,3,4],ymm12[5,6,7]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm12 = ymm14[1,0],ymm4[2,0],ymm14[5,4],ymm4[6,4]
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm13 = ymm7[2,3,0,1]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm7 = ymm7[3,0],ymm13[0,0],ymm7[7,4],ymm13[4,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm7 = ymm7[2,0],ymm12[2,0],ymm7[6,4],ymm12[6,4]
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm13 = ymm8[2,3,0,1]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm8 = ymm8[3,0],ymm13[0,0],ymm8[7,4],ymm13[4,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm8 = ymm8[2,0],ymm12[2,0],ymm8[6,4],ymm12[6,4]
; AVX1-ONLY-NEXT: vmovaps 96(%rdi), %xmm12
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm13 = xmm12[0,1,0,1]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm13 = xmm15[0,1,2],xmm13[3]
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm9 = xmm9[2,3,2,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm9 = mem[0],xmm9[1],mem[2,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm9 = xmm9[0,1],xmm13[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm9[0,1,2,3],ymm7[4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm9[0,1,2,3],ymm8[4,5,6,7]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm4 = ymm14[2,1],ymm4[3,3],ymm14[6,5],ymm4[7,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm9 = xmm5[0],xmm6[1],xmm5[2,3]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm9, %ymm0, %ymm9
@@ -836,10 +836,10 @@ define void @load_i32_stride7_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm2[2,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm5[4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps %ymm3, (%rsi)
-; AVX1-ONLY-NEXT: vmovaps %ymm8, (%rdx)
+; AVX1-ONLY-NEXT: vmovaps %ymm7, (%rdx)
; AVX1-ONLY-NEXT: vmovaps %ymm10, (%rcx)
; AVX1-ONLY-NEXT: vmovaps %ymm11, (%r8)
-; AVX1-ONLY-NEXT: vmovaps %ymm7, (%r9)
+; AVX1-ONLY-NEXT: vmovaps %ymm8, (%r9)
; AVX1-ONLY-NEXT: movq {{[0-9]+}}(%rsp), %rax
; AVX1-ONLY-NEXT: vmovaps %ymm4, (%rax)
; AVX1-ONLY-NEXT: movq {{[0-9]+}}(%rsp), %rax
@@ -858,11 +858,11 @@ define void @load_i32_stride7_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
; AVX2-SLOW-NEXT: vmovdqa 32(%rdi), %ymm1
; AVX2-SLOW-NEXT: vmovdqa 96(%rdi), %ymm10
; AVX2-SLOW-NEXT: vpbroadcastq 80(%rdi), %ymm2
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm10[4,5],ymm2[6,7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm10[4,5,6,7]
; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} xmm3 = <0,7,6,u>
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm6 = ymm0[0,1,2,3,4,5],ymm1[6],ymm0[7]
; AVX2-SLOW-NEXT: vpermd %ymm6, %ymm3, %ymm3
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3,4],ymm3[5,6,7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3,4,5,6,7]
; AVX2-SLOW-NEXT: vmovdqa 128(%rdi), %xmm6
; AVX2-SLOW-NEXT: vmovdqa 160(%rdi), %xmm3
; AVX2-SLOW-NEXT: vpunpckhqdq {{.*#+}} xmm6 = xmm6[1],xmm3[1]
@@ -881,7 +881,7 @@ define void @load_i32_stride7_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm8 = ymm1[0],ymm0[1],ymm1[2,3,4],ymm0[5],ymm1[6,7]
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm11 = ymm8[1,0,3,3,5,4,7,7]
; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm11 = ymm11[0,3,2,3]
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm7 = ymm11[0,1,2],ymm7[3,4],ymm11[5,6,7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm7 = ymm11[0,1,2],ymm7[3,4,5,6,7]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm6 = ymm7[0,1,2,3,4],ymm6[5,6,7]
; AVX2-SLOW-NEXT: vmovdqa 80(%rdi), %xmm7
; AVX2-SLOW-NEXT: vpalignr {{.*#+}} ymm11 = ymm10[8,9,10,11,12,13,14,15],ymm9[0,1,2,3,4,5,6,7],ymm10[24,25,26,27,28,29,30,31],ymm9[16,17,18,19,20,21,22,23]
@@ -907,7 +907,7 @@ define void @load_i32_stride7_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
; AVX2-SLOW-NEXT: vmovdqa 64(%rdi), %xmm11
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} xmm10 = xmm11[0,1,2],xmm10[3]
; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} xmm12 = <4,3,u,u>
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm13 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm13 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX2-SLOW-NEXT: vpermd %ymm13, %ymm12, %ymm12
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} xmm10 = xmm12[0,1],xmm10[2,3]
; AVX2-SLOW-NEXT: vpbroadcastq {{.*#+}} ymm12 = [30064771072,30064771072,30064771072,30064771072]
@@ -961,11 +961,11 @@ define void @load_i32_stride7_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
; AVX2-FAST-NEXT: vmovdqa 32(%rdi), %ymm1
; AVX2-FAST-NEXT: vmovdqa 96(%rdi), %ymm10
; AVX2-FAST-NEXT: vpbroadcastq 80(%rdi), %ymm2
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm10[4,5],ymm2[6,7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm10[4,5,6,7]
; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm3 = <0,7,6,u>
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm6 = ymm0[0,1,2,3,4,5],ymm1[6],ymm0[7]
; AVX2-FAST-NEXT: vpermd %ymm6, %ymm3, %ymm3
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3,4],ymm3[5,6,7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3,4,5,6,7]
; AVX2-FAST-NEXT: vmovdqa 128(%rdi), %xmm6
; AVX2-FAST-NEXT: vmovdqa 160(%rdi), %xmm3
; AVX2-FAST-NEXT: vpunpckhqdq {{.*#+}} xmm6 = xmm6[1],xmm3[1]
@@ -984,7 +984,7 @@ define void @load_i32_stride7_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm11 = <1,0,7,u,u,u,u,u>
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm8 = ymm1[0],ymm0[1],ymm1[2,3,4],ymm0[5],ymm1[6,7]
; AVX2-FAST-NEXT: vpermd %ymm8, %ymm11, %ymm11
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm7 = ymm11[0,1,2],ymm7[3,4],ymm11[5,6,7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm7 = ymm11[0,1,2],ymm7[3,4,5,6,7]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm6 = ymm7[0,1,2,3,4],ymm6[5,6,7]
; AVX2-FAST-NEXT: vmovdqa 80(%rdi), %xmm7
; AVX2-FAST-NEXT: vpalignr {{.*#+}} ymm11 = ymm10[8,9,10,11,12,13,14,15],ymm9[0,1,2,3,4,5,6,7],ymm10[24,25,26,27,28,29,30,31],ymm9[16,17,18,19,20,21,22,23]
@@ -1010,7 +1010,7 @@ define void @load_i32_stride7_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
; AVX2-FAST-NEXT: vmovdqa 64(%rdi), %xmm11
; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm10 = xmm11[0,1,2],xmm10[3]
; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm12 = <4,3,u,u>
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm13 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm13 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX2-FAST-NEXT: vpermd %ymm13, %ymm12, %ymm12
; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm10 = xmm12[0,1],xmm10[2,3]
; AVX2-FAST-NEXT: vpbroadcastq {{.*#+}} ymm12 = [30064771072,30064771072,30064771072,30064771072]
@@ -1064,11 +1064,11 @@ define void @load_i32_stride7_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
; AVX2-FAST-PERLANE-NEXT: vmovdqa 32(%rdi), %ymm1
; AVX2-FAST-PERLANE-NEXT: vmovdqa 96(%rdi), %ymm10
; AVX2-FAST-PERLANE-NEXT: vpbroadcastq 80(%rdi), %ymm2
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm10[4,5],ymm2[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm10[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm3 = <0,7,6,u>
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm6 = ymm0[0,1,2,3,4,5],ymm1[6],ymm0[7]
; AVX2-FAST-PERLANE-NEXT: vpermd %ymm6, %ymm3, %ymm3
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3,4],ymm3[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqa 128(%rdi), %xmm6
; AVX2-FAST-PERLANE-NEXT: vmovdqa 160(%rdi), %xmm3
; AVX2-FAST-PERLANE-NEXT: vpunpckhqdq {{.*#+}} xmm6 = xmm6[1],xmm3[1]
@@ -1087,7 +1087,7 @@ define void @load_i32_stride7_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm8 = ymm1[0],ymm0[1],ymm1[2,3,4],ymm0[5],ymm1[6,7]
; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} ymm11 = ymm8[1,0,3,3,5,4,7,7]
; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm11 = ymm11[0,3,2,3]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm7 = ymm11[0,1,2],ymm7[3,4],ymm11[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm7 = ymm11[0,1,2],ymm7[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm6 = ymm7[0,1,2,3,4],ymm6[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqa 80(%rdi), %xmm7
; AVX2-FAST-PERLANE-NEXT: vpalignr {{.*#+}} ymm11 = ymm10[8,9,10,11,12,13,14,15],ymm9[0,1,2,3,4,5,6,7],ymm10[24,25,26,27,28,29,30,31],ymm9[16,17,18,19,20,21,22,23]
@@ -1113,7 +1113,7 @@ define void @load_i32_stride7_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
; AVX2-FAST-PERLANE-NEXT: vmovdqa 64(%rdi), %xmm11
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm10 = xmm11[0,1,2],xmm10[3]
; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm12 = <4,3,u,u>
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm13 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm13 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpermd %ymm13, %ymm12, %ymm12
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm10 = xmm12[0,1],xmm10[2,3]
; AVX2-FAST-PERLANE-NEXT: vpbroadcastq {{.*#+}} ymm12 = [30064771072,30064771072,30064771072,30064771072]
@@ -1568,16 +1568,16 @@ define void @load_i32_stride7_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps (%rdi), %ymm1
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 96(%rdi), %ymm8
+; AVX1-ONLY-NEXT: vmovaps 96(%rdi), %ymm7
; AVX1-ONLY-NEXT: vmovaps 80(%rdi), %xmm0
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm8[0],ymm0[0],ymm8[2],ymm0[2]
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm7[0],ymm0[0],ymm7[2],ymm0[2]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6],ymm1[7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm1, %xmm1
; AVX1-ONLY-NEXT: vmovaps (%rdi), %xmm15
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm15[0,1],xmm1[2,3]
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm1 = xmm1[0,3,2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4],ymm1[5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps 160(%rdi), %xmm2
; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vmovaps 128(%rdi), %xmm1
@@ -1596,201 +1596,201 @@ define void @load_i32_stride7_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: vmovaps 224(%rdi), %xmm9
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm9[0,1],xmm1[2,3]
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm1 = xmm1[0,3,2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4],ymm1[5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps 384(%rdi), %xmm2
; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vmovaps 352(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm2[1]
-; AVX1-ONLY-NEXT: vmovaps 416(%rdi), %xmm7
-; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm1 = zero,xmm1[1,2],xmm7[1]
+; AVX1-ONLY-NEXT: vmovaps 416(%rdi), %xmm6
+; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm1 = zero,xmm1[1,2],xmm6[1]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 64(%rdi), %ymm5
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm8[1,1],ymm5[2,2],ymm8[5,5],ymm5[6,6]
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm1[2,3,2,3]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm7[1,1],ymm5[2,2],ymm7[5,5],ymm5[6,6]
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,2,3]
; AVX1-ONLY-NEXT: vmovaps 32(%rdi), %xmm10
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm10[0],xmm15[1],xmm10[2,3]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm2 = xmm2[1,0],mem[3,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4],ymm2[5,6,7]
-; AVX1-ONLY-NEXT: vmovaps 128(%rdi), %ymm2
-; AVX1-ONLY-NEXT: vmovaps 160(%rdi), %ymm6
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm3 = ymm6[2,3,0,1]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm2[0,0],ymm3[3,3],ymm2[4,4],ymm3[7,7]
-; AVX1-ONLY-NEXT: vextractf128 $1, %ymm3, %xmm3
-; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm3 = zero,xmm3[1,2],xmm12[2]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm3
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm3[5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm10[0],xmm15[1],xmm10[2,3]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm1[1,0],mem[3,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps 160(%rdi), %ymm8
+; AVX1-ONLY-NEXT: vmovaps 128(%rdi), %ymm1
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm2 = ymm1[2,3],ymm8[0,1]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm1[0,0],ymm2[3,3],ymm1[4,4],ymm2[7,7]
+; AVX1-ONLY-NEXT: vextractf128 $1, %ymm2, %xmm2
+; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm2 = zero,xmm2[1,2],xmm12[2]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm2
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 288(%rdi), %ymm3
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm13[1,1],ymm3[2,2],ymm13[5,5],ymm3[6,6]
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm1[2,3,2,3]
+; AVX1-ONLY-NEXT: vmovaps 288(%rdi), %ymm4
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm13[1,1],ymm4[2,2],ymm13[5,5],ymm4[6,6]
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,2,3]
; AVX1-ONLY-NEXT: vmovaps 256(%rdi), %xmm11
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm4 = xmm11[0],xmm9[1],xmm11[2,3]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm4 = xmm4[1,0],mem[3,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm4[0,1,2],ymm1[3,4],ymm4[5,6,7]
-; AVX1-ONLY-NEXT: vmovaps 352(%rdi), %ymm1
-; AVX1-ONLY-NEXT: vmovaps 384(%rdi), %ymm4
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm14 = ymm4[2,3,0,1]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm14 = ymm1[0,0],ymm14[3,3],ymm1[4,4],ymm14[7,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm3 = xmm11[0],xmm9[1],xmm11[2,3]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm3 = xmm3[1,0],mem[3,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2],ymm0[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps 384(%rdi), %ymm3
+; AVX1-ONLY-NEXT: vmovaps 352(%rdi), %ymm0
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm14 = ymm0[2,3],ymm3[0,1]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm14 = ymm0[0,0],ymm14[3,3],ymm0[4,4],ymm14[7,7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm14, %xmm14
-; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm14 = zero,xmm14[1,2],xmm7[2]
+; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm14 = zero,xmm14[1,2],xmm6[2]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm14, %ymm0, %ymm14
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm14[5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm0 = xmm15[2,3,2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm0[0],xmm10[1],xmm0[2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm14[5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm2 = xmm15[2,3,2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm2[0],xmm10[1],xmm2[2,3]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm14 = ymm14[3,1],ymm5[0,3],ymm14[7,5],ymm5[4,7]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm14 = ymm8[2,1],ymm14[2,0],ymm8[6,5],ymm14[6,4]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm14[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm14 = ymm2[0],ymm6[0],ymm2[2],ymm6[2]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm14 = ymm7[2,1],ymm14[2,0],ymm7[6,5],ymm14[6,4]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm14[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm14 = ymm1[0],ymm8[0],ymm1[2],ymm8[2]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm14, %xmm14
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm12 = xmm14[0,1,2],xmm12[3]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm12, %ymm0, %ymm12
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm12[5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm0 = xmm9[2,3,2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm0[0],xmm11[1],xmm0[2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm12[5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm2 = xmm9[2,3,2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm2[0],xmm11[1],xmm2[2,3]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm12 = ymm12[3,1],ymm3[0,3],ymm12[7,5],ymm3[4,7]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm12 = ymm12[3,1],ymm4[0,3],ymm12[7,5],ymm4[4,7]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm12 = ymm13[2,1],ymm12[2,0],ymm13[6,5],ymm12[6,4]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm12[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm12 = ymm1[0],ymm4[0],ymm1[2],ymm4[2]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm12[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm12 = ymm0[0],ymm3[0],ymm0[2],ymm3[2]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm12, %xmm12
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm7 = xmm12[0,1,2],xmm7[3]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm7, %ymm0, %ymm7
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm7[5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm5[1,0],ymm8[0,0],ymm5[5,4],ymm8[4,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm8[3,1],ymm0[0,2],ymm8[7,5],ymm0[4,6]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm6 = xmm12[0,1,2],xmm6[3]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm6, %ymm0, %ymm6
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm6[5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm5[1,0],ymm7[0,0],ymm5[5,4],ymm7[4,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm7[3,1],ymm2[0,2],ymm7[7,5],ymm2[4,6]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm5 = xmm10[0,1,2],xmm15[3]
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm5 = xmm5[3,2,2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1],ymm0[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovaps 192(%rdi), %ymm0
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm7 = ymm0[0,1],ymm6[1,3],ymm0[4,5],ymm6[5,7]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm7 = ymm2[0,2],ymm7[2,0],ymm2[4,6],ymm7[6,4]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4],ymm7[5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm3[1,0],ymm13[0,0],ymm3[5,4],ymm13[4,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm13[3,1],ymm3[0,2],ymm13[7,5],ymm3[4,6]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm5 = xmm11[0,1,2],xmm9[3]
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm5 = xmm5[3,2,2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm5[0,1],ymm3[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm5[0,1],ymm2[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps 192(%rdi), %ymm7
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm5 = ymm7[0,1],ymm8[1,3],ymm7[4,5],ymm8[5,7]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm5 = ymm1[0,2],ymm5[2,0],ymm1[4,6],ymm5[6,4]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm5[5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm4[1,0],ymm13[0,0],ymm4[5,4],ymm13[4,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm13[3,1],ymm2[0,2],ymm13[7,5],ymm2[4,6]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm4 = xmm11[0,1,2],xmm9[3]
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm4 = xmm4[3,2,2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm4[0,1],ymm2[2,3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps 416(%rdi), %ymm5
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm7 = ymm5[0,1],ymm4[1,3],ymm5[4,5],ymm4[5,7]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm7 = ymm1[0,2],ymm7[2,0],ymm1[4,6],ymm7[6,4]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4],ymm7[5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm3 = ymm2[2,3,0,1]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm2[3,0],ymm3[0,0],ymm2[7,4],ymm3[4,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm0[1,0],ymm6[2,0],ymm0[5,4],ymm6[6,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm2[2,0],ymm3[2,0],ymm2[6,4],ymm3[6,4]
-; AVX1-ONLY-NEXT: vmovaps 64(%rdi), %xmm8
-; AVX1-ONLY-NEXT: vmovaps 96(%rdi), %xmm9
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm3 = xmm9[0,1,0,1]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm3 = xmm8[0,1,2],xmm3[3]
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm10 = xmm10[2,3,2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm10 = mem[0],xmm10[1],mem[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm3 = xmm10[0,1],xmm3[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2,3],ymm2[4,5,6,7]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm4 = ymm5[0,1],ymm3[1,3],ymm5[4,5],ymm3[5,7]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm4 = ymm0[0,2],ymm4[2,0],ymm0[4,6],ymm4[6,4]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm4[5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm3 = ymm1[2,3,0,1]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm1[3,0],ymm3[0,0],ymm1[7,4],ymm3[4,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm5[1,0],ymm4[2,0],ymm5[5,4],ymm4[6,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm1[2,0],ymm3[2,0],ymm1[6,4],ymm3[6,4]
-; AVX1-ONLY-NEXT: vmovaps 320(%rdi), %xmm7
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm10 = xmm7[0,1,0,1]
-; AVX1-ONLY-NEXT: vmovaps 288(%rdi), %xmm12
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm10 = xmm12[0,1,2],xmm10[3]
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm2 = ymm1[2,3,0,1]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm1[3,0],ymm2[0,0],ymm1[7,4],ymm2[4,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm7[1,0],ymm8[2,0],ymm7[5,4],ymm8[6,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm1[2,0],ymm2[2,0],ymm1[6,4],ymm2[6,4]
+; AVX1-ONLY-NEXT: vmovaps 64(%rdi), %xmm2
+; AVX1-ONLY-NEXT: vmovaps 96(%rdi), %xmm9
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm4 = xmm9[0,1,0,1]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm4 = xmm2[0,1,2],xmm4[3]
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm6 = xmm10[2,3,2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm6 = mem[0],xmm6[1],mem[2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm4 = xmm6[0,1],xmm4[2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm4[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm0[2,3,0,1]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[3,0],ymm1[0,0],ymm0[7,4],ymm1[4,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm5[1,0],ymm3[2,0],ymm5[5,4],ymm3[6,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm1[2,0],ymm0[6,4],ymm1[6,4]
+; AVX1-ONLY-NEXT: vmovaps 320(%rdi), %xmm14
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm6 = xmm14[0,1,0,1]
+; AVX1-ONLY-NEXT: vmovaps 288(%rdi), %xmm10
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm6 = xmm10[0,1,2],xmm6[3]
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm11 = xmm11[2,3,2,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm11 = mem[0],xmm11[1],mem[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm10 = xmm11[0,1],xmm10[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm10[0,1,2,3],ymm1[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm1, (%rsp) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm6 = ymm0[2,1],ymm6[3,3],ymm0[6,5],ymm6[7,7]
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm6 = xmm11[0,1],xmm6[2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm6[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, (%rsp) # 32-byte Spill
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm6 = ymm7[2,1],ymm8[3,3],ymm7[6,5],ymm8[7,7]
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm10 = xmm2[0],xmm1[1],xmm2[2,3]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm10, %ymm0, %ymm10
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm6 = ymm10[1,0],ymm6[2,0],ymm10[5,4],ymm6[6,4]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm8 = xmm9[0,1,2],xmm8[3]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm10 = ymm14[0,0],ymm13[1,0],ymm14[4,4],ymm13[5,4]
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} ymm10 = ymm10[2,0,2,3,6,4,6,7]
-; AVX1-ONLY-NEXT: vextractf128 $1, %ymm10, %xmm10
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm8 = xmm10[0,1],xmm8[3,2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm15 = ymm8[0,1,2,3],ymm6[4,5,6,7]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm4 = ymm5[2,1],ymm4[3,3],ymm5[6,5],ymm4[7,7]
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm8 = xmm6[0],xmm3[1],xmm6[2,3]
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm8 = xmm1[0],xmm0[1],xmm1[2,3]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm8, %ymm0, %ymm8
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm4 = ymm8[1,0],ymm4[2,0],ymm8[5,4],ymm4[6,4]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm8 = xmm7[0,1,2],xmm12[3]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm6 = ymm8[1,0],ymm6[2,0],ymm8[5,4],ymm6[6,4]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm9[0,1,2],xmm2[3]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm10 = ymm11[0,0],ymm12[1,0],ymm11[4,4],ymm12[5,4]
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} ymm10 = ymm10[2,0,2,3,6,4,6,7]
-; AVX1-ONLY-NEXT: vextractf128 $1, %ymm10, %xmm10
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm8 = xmm10[0,1],xmm8[3,2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm8[0,1,2,3],ymm4[4,5,6,7]
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm8 = ymm0[2,3,0,1]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[3,0],ymm8[0,0],ymm0[7,4],ymm8[4,4]
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm8 = xmm1[2,3,2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm8 = xmm8[0],xmm2[1],xmm8[2,3]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm8, %ymm0, %ymm8
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm8[0,1],ymm0[2,0],ymm8[4,5],ymm0[6,4]
-; AVX1-ONLY-NEXT: vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm8 = mem[0,1,0,1]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm8 = xmm8[0,1,2],xmm9[3]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm9 = ymm14[1,0],ymm13[2,0],ymm14[5,4],ymm13[6,4]
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} ymm9 = ymm9[2,0,2,3,6,4,6,7]
-; AVX1-ONLY-NEXT: vextractf128 $1, %ymm9, %xmm9
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm8 = xmm9[0,1],xmm8[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm8[0,1,2,3],ymm0[4,5,6,7]
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm8 = ymm5[2,3,0,1]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm5 = ymm5[3,0],ymm8[0,0],ymm5[7,4],ymm8[4,4]
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm8 = xmm3[2,3,2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm8 = xmm8[0],xmm6[1],xmm8[2,3]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm8, %ymm0, %ymm8
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm5 = ymm8[0,1],ymm5[2,0],ymm8[4,5],ymm5[6,4]
-; AVX1-ONLY-NEXT: vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm8 = mem[0,1,0,1]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm3 = xmm8[0,1,2],xmm7[3]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm8 = ymm11[1,0],ymm12[2,0],ymm11[5,4],ymm12[6,4]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm8 = ymm13[0,0],ymm12[1,0],ymm13[4,4],ymm12[5,4]
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} ymm8 = ymm8[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT: vextractf128 $1, %ymm8, %xmm8
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm2 = xmm8[0,1],xmm2[3,2]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm15 = ymm2[0,1,2,3],ymm6[4,5,6,7]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm5[2,1],ymm3[3,3],ymm5[6,5],ymm3[7,7]
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm6 = xmm4[0],xmm2[1],xmm4[2,3]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm6, %ymm0, %ymm6
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm6[1,0],ymm3[2,0],ymm6[5,4],ymm3[6,4]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm6 = xmm14[0,1,2],xmm10[3]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm8 = ymm10[0,0],ymm11[1,0],ymm10[4,4],ymm11[5,4]
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} ymm8 = ymm8[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT: vextractf128 $1, %ymm8, %xmm8
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm6 = xmm8[0,1],xmm6[3,2]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm6[0,1,2,3],ymm3[4,5,6,7]
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm6 = ymm7[2,3,0,1]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm6 = ymm7[3,0],ymm6[0,0],ymm7[7,4],ymm6[4,4]
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm7 = xmm0[2,3,2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm7 = xmm7[0],xmm1[1],xmm7[2,3]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm7, %ymm0, %ymm7
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm6 = ymm7[0,1],ymm6[2,0],ymm7[4,5],ymm6[6,4]
+; AVX1-ONLY-NEXT: vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm7 = mem[0,1,0,1]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm7 = xmm7[0,1,2],xmm9[3]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm8 = ymm13[1,0],ymm12[2,0],ymm13[5,4],ymm12[6,4]
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} ymm8 = ymm8[2,0,2,3,6,4,6,7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm8, %xmm8
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm3 = xmm8[0,1],xmm3[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm5[4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm7 = xmm8[0,1],xmm7[2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm7[0,1,2,3],ymm6[4,5,6,7]
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm7 = ymm5[2,3,0,1]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm5 = ymm5[3,0],ymm7[0,0],ymm5[7,4],ymm7[4,4]
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm7 = xmm2[2,3,2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm7 = xmm7[0],xmm4[1],xmm7[2,3]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm7, %ymm0, %ymm7
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm5 = ymm7[0,1],ymm5[2,0],ymm7[4,5],ymm5[6,4]
+; AVX1-ONLY-NEXT: vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm7 = mem[0,1,0,1]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm7[0,1,2],xmm14[3]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm7 = ymm10[1,0],ymm11[2,0],ymm10[5,4],ymm11[6,4]
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} ymm7 = ymm7[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT: vextractf128 $1, %ymm7, %xmm7
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm7[0,1],xmm1[2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm5[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm5, 32(%rsi)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm5, (%rsi)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm1, 32(%rdx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm1, (%rdx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm1, 32(%rcx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm1, (%rcx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm1, 32(%r8)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm1, (%r8)
-; AVX1-ONLY-NEXT: vmovups (%rsp), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm1, 32(%r9)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm1, (%r9)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm0, 32(%rdx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm0, (%rdx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm0, 32(%rcx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm0, (%rcx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm0, 32(%r8)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm0, (%r8)
+; AVX1-ONLY-NEXT: vmovups (%rsp), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm0, 32(%r9)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm0, (%r9)
; AVX1-ONLY-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; AVX1-ONLY-NEXT: vmovaps %ymm4, 32(%rax)
+; AVX1-ONLY-NEXT: vmovaps %ymm3, 32(%rax)
; AVX1-ONLY-NEXT: vmovaps %ymm15, (%rax)
; AVX1-ONLY-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; AVX1-ONLY-NEXT: vmovaps %ymm3, 32(%rax)
-; AVX1-ONLY-NEXT: vmovaps %ymm0, (%rax)
+; AVX1-ONLY-NEXT: vmovaps %ymm1, 32(%rax)
+; AVX1-ONLY-NEXT: vmovaps %ymm6, (%rax)
; AVX1-ONLY-NEXT: addq $456, %rsp # imm = 0x1C8
; AVX1-ONLY-NEXT: vzeroupper
; AVX1-ONLY-NEXT: retq
@@ -1808,12 +1808,12 @@ define void @load_i32_stride7_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-SLOW-NEXT: vmovdqa 32(%rdi), %ymm4
; AVX2-SLOW-NEXT: vmovdqa 96(%rdi), %ymm12
; AVX2-SLOW-NEXT: vpbroadcastq 80(%rdi), %ymm1
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm12[4,5],ymm1[6,7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm12[4,5,6,7]
; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} xmm2 = <0,7,6,u>
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm9[0,1,2,3,4,5],ymm4[6],ymm9[7]
; AVX2-SLOW-NEXT: vmovdqa %ymm4, %ymm7
; AVX2-SLOW-NEXT: vpermd %ymm3, %ymm2, %ymm3
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm3[0,1,2],ymm1[3,4],ymm3[5,6,7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm3[0,1,2],ymm1[3,4,5,6,7]
; AVX2-SLOW-NEXT: vmovdqa 128(%rdi), %xmm3
; AVX2-SLOW-NEXT: vmovdqa 160(%rdi), %xmm4
; AVX2-SLOW-NEXT: vmovdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -1828,8 +1828,8 @@ define void @load_i32_stride7_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm10[0,1,2,3,4,5],ymm0[6],ymm10[7]
; AVX2-SLOW-NEXT: vpermd %ymm1, %ymm2, %ymm1
; AVX2-SLOW-NEXT: vpbroadcastq 304(%rdi), %ymm2
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm5[4,5],ymm2[6,7]
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4],ymm1[5,6,7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm5[4,5,6,7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
; AVX2-SLOW-NEXT: vmovdqa 352(%rdi), %xmm2
; AVX2-SLOW-NEXT: vmovdqa 384(%rdi), %xmm3
; AVX2-SLOW-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -1851,7 +1851,7 @@ define void @load_i32_stride7_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-SLOW-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm3 = ymm3[1,0,3,3,5,4,7,7]
; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm3 = ymm3[0,3,2,3]
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3,4],ymm3[5,6,7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3,4,5,6,7]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm2[0,1,2,3,4],ymm1[5,6,7]
; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm1 = mem[2,2,2,2]
@@ -1868,7 +1868,7 @@ define void @load_i32_stride7_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm7[0],ymm9[1],ymm7[2,3,4],ymm9[5],ymm7[6,7]
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm14 = ymm4[1,0,3,3,5,4,7,7]
; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm14 = ymm14[0,3,2,3]
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm14[0,1,2],ymm0[3,4],ymm14[5,6,7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm14[0,1,2],ymm0[3,4,5,6,7]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm3[5,6,7]
; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovdqa 80(%rdi), %xmm0
@@ -1919,7 +1919,7 @@ define void @load_i32_stride7_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-SLOW-NEXT: vmovdqa 64(%rdi), %xmm3
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} xmm0 = xmm3[0,1,2],xmm0[3]
; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} xmm5 = <4,3,u,u>
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm10 = ymm7[0,1,2,3],ymm9[4,5],ymm7[6,7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm10 = ymm7[0,1,2,3],ymm9[4,5,6,7]
; AVX2-SLOW-NEXT: vmovdqa %ymm7, %ymm12
; AVX2-SLOW-NEXT: vpermd %ymm10, %ymm5, %ymm10
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} xmm10 = xmm10[0,1],xmm0[2,3]
@@ -1931,7 +1931,7 @@ define void @load_i32_stride7_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm15 = ymm10[0,1,2,3],ymm11[4,5,6,7]
; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm10 = ymm14[0,1,2,3],ymm7[4,5],ymm14[6,7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm10 = ymm14[0,1,2,3],ymm7[4,5,6,7]
; AVX2-SLOW-NEXT: vpermd %ymm10, %ymm5, %ymm5
; AVX2-SLOW-NEXT: vpbroadcastd 324(%rdi), %xmm10
; AVX2-SLOW-NEXT: vmovdqa 288(%rdi), %xmm11
@@ -2034,12 +2034,12 @@ define void @load_i32_stride7_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-NEXT: vmovdqa 64(%rdi), %ymm11
; AVX2-FAST-NEXT: vmovdqa 96(%rdi), %ymm3
; AVX2-FAST-NEXT: vpbroadcastq 80(%rdi), %ymm2
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5],ymm2[6,7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7]
; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm7 = <0,7,6,u>
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm8 = ymm12[0,1,2,3,4,5],ymm5[6],ymm12[7]
; AVX2-FAST-NEXT: vmovdqa %ymm5, %ymm6
; AVX2-FAST-NEXT: vpermd %ymm8, %ymm7, %ymm8
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm8[0,1,2],ymm2[3,4],ymm8[5,6,7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm8[0,1,2],ymm2[3,4,5,6,7]
; AVX2-FAST-NEXT: vmovdqa 128(%rdi), %xmm8
; AVX2-FAST-NEXT: vmovdqa 160(%rdi), %xmm5
; AVX2-FAST-NEXT: vmovdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -2056,8 +2056,8 @@ define void @load_i32_stride7_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vpermd %ymm2, %ymm7, %ymm2
; AVX2-FAST-NEXT: vpbroadcastq 304(%rdi), %ymm7
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm7 = ymm7[0,1,2,3],ymm4[4,5],ymm7[6,7]
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2],ymm7[3,4],ymm2[5,6,7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm7 = ymm7[0,1,2,3],ymm4[4,5,6,7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2],ymm7[3,4,5,6,7]
; AVX2-FAST-NEXT: vmovdqa 352(%rdi), %xmm7
; AVX2-FAST-NEXT: vmovdqa 384(%rdi), %xmm0
; AVX2-FAST-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -2079,7 +2079,7 @@ define void @load_i32_stride7_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm6[0],ymm12[1],ymm6[2,3,4],ymm12[5],ymm6[6,7]
; AVX2-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vpermd %ymm1, %ymm7, %ymm14
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm8 = ymm14[0,1,2],ymm8[3,4],ymm14[5,6,7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm8 = ymm14[0,1,2],ymm8[3,4,5,6,7]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm8[0,1,2,3,4],ymm2[5,6,7]
; AVX2-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vpshufd {{.*#+}} xmm2 = mem[2,2,2,2]
@@ -2094,7 +2094,7 @@ define void @load_i32_stride7_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-NEXT: vpermd %ymm2, %ymm0, %ymm0
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm13[0],ymm5[1],ymm13[2,3,4],ymm5[5],ymm13[6,7]
; AVX2-FAST-NEXT: vpermd %ymm2, %ymm7, %ymm5
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm5[0,1,2],ymm0[3,4],ymm5[5,6,7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm5[0,1,2],ymm0[3,4,5,6,7]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovdqa 80(%rdi), %xmm0
@@ -2145,7 +2145,7 @@ define void @load_i32_stride7_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-NEXT: vmovdqa 64(%rdi), %xmm1
; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm0 = xmm1[0,1,2],xmm0[3]
; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm3 = <4,3,u,u>
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm4 = ymm6[0,1,2,3],ymm12[4,5],ymm6[6,7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm4 = ymm6[0,1,2,3],ymm12[4,5,6,7]
; AVX2-FAST-NEXT: vmovdqa %ymm6, %ymm8
; AVX2-FAST-NEXT: vpermd %ymm4, %ymm3, %ymm4
; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm4 = xmm4[0,1],xmm0[2,3]
@@ -2158,7 +2158,7 @@ define void @load_i32_stride7_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm4 = ymm6[0,1,2,3],ymm13[4,5],ymm6[6,7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm4 = ymm6[0,1,2,3],ymm13[4,5,6,7]
; AVX2-FAST-NEXT: vpermd %ymm4, %ymm3, %ymm3
; AVX2-FAST-NEXT: vpbroadcastd 324(%rdi), %xmm4
; AVX2-FAST-NEXT: vmovdqa 288(%rdi), %xmm5
@@ -2261,12 +2261,12 @@ define void @load_i32_stride7_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-PERLANE-NEXT: vmovdqa 32(%rdi), %ymm4
; AVX2-FAST-PERLANE-NEXT: vmovdqa 96(%rdi), %ymm12
; AVX2-FAST-PERLANE-NEXT: vpbroadcastq 80(%rdi), %ymm1
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm12[4,5],ymm1[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm12[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm2 = <0,7,6,u>
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm3 = ymm9[0,1,2,3,4,5],ymm4[6],ymm9[7]
; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm4, %ymm7
; AVX2-FAST-PERLANE-NEXT: vpermd %ymm3, %ymm2, %ymm3
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm3[0,1,2],ymm1[3,4],ymm3[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm3[0,1,2],ymm1[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqa 128(%rdi), %xmm3
; AVX2-FAST-PERLANE-NEXT: vmovdqa 160(%rdi), %xmm4
; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -2281,8 +2281,8 @@ define void @load_i32_stride7_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm10[0,1,2,3,4,5],ymm0[6],ymm10[7]
; AVX2-FAST-PERLANE-NEXT: vpermd %ymm1, %ymm2, %ymm1
; AVX2-FAST-PERLANE-NEXT: vpbroadcastq 304(%rdi), %ymm2
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm5[4,5],ymm2[6,7]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4],ymm1[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm5[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqa 352(%rdi), %xmm2
; AVX2-FAST-PERLANE-NEXT: vmovdqa 384(%rdi), %xmm3
; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -2304,7 +2304,7 @@ define void @load_i32_stride7_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} ymm3 = ymm3[1,0,3,3,5,4,7,7]
; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm3 = ymm3[0,3,2,3]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3,4],ymm3[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm2[0,1,2,3,4],ymm1[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} xmm1 = mem[2,2,2,2]
@@ -2321,7 +2321,7 @@ define void @load_i32_stride7_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm4 = ymm7[0],ymm9[1],ymm7[2,3,4],ymm9[5],ymm7[6,7]
; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} ymm14 = ymm4[1,0,3,3,5,4,7,7]
; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm14 = ymm14[0,3,2,3]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm14[0,1,2],ymm0[3,4],ymm14[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm14[0,1,2],ymm0[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm3[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovdqa 80(%rdi), %xmm0
@@ -2372,7 +2372,7 @@ define void @load_i32_stride7_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-PERLANE-NEXT: vmovdqa 64(%rdi), %xmm3
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm0 = xmm3[0,1,2],xmm0[3]
; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm5 = <4,3,u,u>
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm10 = ymm7[0,1,2,3],ymm9[4,5],ymm7[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm10 = ymm7[0,1,2,3],ymm9[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm7, %ymm12
; AVX2-FAST-PERLANE-NEXT: vpermd %ymm10, %ymm5, %ymm10
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm10 = xmm10[0,1],xmm0[2,3]
@@ -2384,7 +2384,7 @@ define void @load_i32_stride7_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm15 = ymm10[0,1,2,3],ymm11[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm10 = ymm14[0,1,2,3],ymm7[4,5],ymm14[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm10 = ymm14[0,1,2,3],ymm7[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpermd %ymm10, %ymm5, %ymm5
; AVX2-FAST-PERLANE-NEXT: vpbroadcastd 324(%rdi), %xmm10
; AVX2-FAST-PERLANE-NEXT: vmovdqa 288(%rdi), %xmm11
@@ -3441,7 +3441,7 @@ define void @load_i32_stride7_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
;
; AVX1-ONLY-LABEL: load_i32_stride7_vf32:
; AVX1-ONLY: # %bb.0:
-; AVX1-ONLY-NEXT: subq $1496, %rsp # imm = 0x5D8
+; AVX1-ONLY-NEXT: subq $1528, %rsp # imm = 0x5F8
; AVX1-ONLY-NEXT: vmovaps 480(%rdi), %ymm4
; AVX1-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 448(%rdi), %ymm3
@@ -3452,27 +3452,24 @@ define void @load_i32_stride7_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: vmovaps (%rdi), %ymm1
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 96(%rdi), %ymm6
+; AVX1-ONLY-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 80(%rdi), %xmm0
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm6[0],ymm0[0],ymm6[2],ymm0[2]
-; AVX1-ONLY-NEXT: vmovaps %ymm6, %ymm8
-; AVX1-ONLY-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6],ymm1[7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm1, %xmm1
-; AVX1-ONLY-NEXT: vmovaps (%rdi), %xmm2
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3]
-; AVX1-ONLY-NEXT: vmovaps %xmm2, %xmm12
-; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps (%rdi), %xmm10
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm10[0,1],xmm1[2,3]
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm1 = xmm1[0,3,2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4],ymm1[5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps 160(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vmovaps %xmm1, (%rsp) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vmovaps 128(%rdi), %xmm2
; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm2[1],xmm1[1]
-; AVX1-ONLY-NEXT: vmovaps 192(%rdi), %xmm2
-; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm1 = zero,xmm1[1,2],xmm2[1]
+; AVX1-ONLY-NEXT: vmovaps 192(%rdi), %xmm6
+; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm1 = zero,xmm1[1,2],xmm6[1]
+; AVX1-ONLY-NEXT: vmovaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -3482,17 +3479,18 @@ define void @load_i32_stride7_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm3[0,1,2,3,4,5],ymm4[6],ymm3[7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm1, %xmm1
-; AVX1-ONLY-NEXT: vmovaps 448(%rdi), %xmm7
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm7[0,1],xmm1[2,3]
+; AVX1-ONLY-NEXT: vmovaps 448(%rdi), %xmm8
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm8[0,1],xmm1[2,3]
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm1 = xmm1[0,3,2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4],ymm1[5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps 608(%rdi), %xmm2
; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vmovaps 576(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm2[1]
-; AVX1-ONLY-NEXT: vmovaps 640(%rdi), %xmm13
-; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm1 = zero,xmm1[1,2],xmm13[1]
+; AVX1-ONLY-NEXT: vmovaps 640(%rdi), %xmm11
+; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm1 = zero,xmm1[1,2],xmm11[1]
+; AVX1-ONLY-NEXT: vmovaps %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -3504,449 +3502,441 @@ define void @load_i32_stride7_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm0, %xmm0
; AVX1-ONLY-NEXT: vmovaps 224(%rdi), %xmm1
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm1[0,1],xmm0[2,3]
-; AVX1-ONLY-NEXT: vmovaps %xmm1, %xmm9
+; AVX1-ONLY-NEXT: vmovaps %xmm1, %xmm14
; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm1 = xmm0[0,3,2,3]
-; AVX1-ONLY-NEXT: vmovaps 320(%rdi), %ymm0
-; AVX1-ONLY-NEXT: vmovaps 304(%rdi), %xmm2
-; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm2 = ymm0[0],ymm2[0],ymm0[2],ymm2[2]
-; AVX1-ONLY-NEXT: vmovaps %ymm0, %ymm4
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4],ymm1[5,6,7]
-; AVX1-ONLY-NEXT: vmovaps 384(%rdi), %xmm3
-; AVX1-ONLY-NEXT: vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 352(%rdi), %xmm2
-; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm2 = xmm2[1],xmm3[1]
-; AVX1-ONLY-NEXT: vmovaps 416(%rdi), %xmm6
-; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm2 = zero,xmm2[1,2],xmm6[1]
-; AVX1-ONLY-NEXT: vmovaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm2
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[0,3,2,3]
+; AVX1-ONLY-NEXT: vmovaps 320(%rdi), %ymm12
+; AVX1-ONLY-NEXT: vmovaps 304(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 704(%rdi), %ymm2
-; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 672(%rdi), %ymm1
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm12[0],ymm1[0],ymm12[2],ymm1[2]
+; AVX1-ONLY-NEXT: vmovups %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps 384(%rdi), %xmm2
+; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 352(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm2[1]
+; AVX1-ONLY-NEXT: vmovaps 416(%rdi), %xmm2
+; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm1 = zero,xmm1[1,2],xmm2[1]
+; AVX1-ONLY-NEXT: vmovaps %xmm2, %xmm15
+; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 704(%rdi), %ymm1
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6],ymm1[7]
-; AVX1-ONLY-NEXT: vextractf128 $1, %ymm1, %xmm1
-; AVX1-ONLY-NEXT: vmovaps 672(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm0[0,1],xmm1[2,3]
-; AVX1-ONLY-NEXT: vmovaps %xmm0, %xmm15
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm1 = xmm1[0,3,2,3]
-; AVX1-ONLY-NEXT: vmovaps 768(%rdi), %ymm0
-; AVX1-ONLY-NEXT: vmovaps 752(%rdi), %xmm2
-; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm2 = ymm0[0],ymm2[0],ymm0[2],ymm2[2]
-; AVX1-ONLY-NEXT: vmovaps %ymm0, %ymm11
+; AVX1-ONLY-NEXT: vmovaps 672(%rdi), %ymm0
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4],ymm1[5,6,7]
-; AVX1-ONLY-NEXT: vmovaps 832(%rdi), %xmm3
-; AVX1-ONLY-NEXT: vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 800(%rdi), %xmm2
-; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm2 = xmm2[1],xmm3[1]
-; AVX1-ONLY-NEXT: vmovaps 864(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm2 = zero,xmm2[1,2],xmm0[1]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm2
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6],ymm0[7]
+; AVX1-ONLY-NEXT: vextractf128 $1, %ymm0, %xmm0
+; AVX1-ONLY-NEXT: vmovaps 672(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm1[0,1],xmm0[2,3]
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[0,3,2,3]
+; AVX1-ONLY-NEXT: vmovaps 768(%rdi), %ymm2
+; AVX1-ONLY-NEXT: vmovaps 752(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 64(%rdi), %ymm10
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm8[1,1],ymm10[2,2],ymm8[5,5],ymm10[6,6]
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm1[2,3,2,3]
-; AVX1-ONLY-NEXT: vmovaps 32(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm0[0],xmm12[1],xmm0[2,3]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm2 = xmm2[1,0],mem[3,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4],ymm2[5,6,7]
-; AVX1-ONLY-NEXT: vmovaps 128(%rdi), %ymm0
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
+; AVX1-ONLY-NEXT: vmovaps %ymm2, %ymm13
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps 832(%rdi), %xmm2
+; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 800(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm2[1]
+; AVX1-ONLY-NEXT: vmovaps 864(%rdi), %xmm2
+; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm1 = zero,xmm1[1,2],xmm2[1]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 160(%rdi), %ymm2
-; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm2 = ymm2[2,3,0,1]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm0[0,0],ymm2[3,3],ymm0[4,4],ymm2[7,7]
-; AVX1-ONLY-NEXT: vextractf128 $1, %ymm2, %xmm2
-; AVX1-ONLY-NEXT: vinsertps $49, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm2 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm2 = zero,xmm2[1,2],mem[0]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm2
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
+; AVX1-ONLY-NEXT: vmovaps 64(%rdi), %ymm1
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[1,1],ymm1[2,2],ymm0[5,5],ymm1[6,6]
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,2,3]
+; AVX1-ONLY-NEXT: vmovaps 32(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0],xmm10[1],xmm1[2,3]
+; AVX1-ONLY-NEXT: vmovaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm1[1,0],mem[3,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps 160(%rdi), %ymm9
+; AVX1-ONLY-NEXT: vmovaps 128(%rdi), %ymm4
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm4[2,3],ymm9[0,1]
+; AVX1-ONLY-NEXT: vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm4[0,0],ymm1[3,3],ymm4[4,4],ymm1[7,7]
+; AVX1-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vextractf128 $1, %ymm1, %xmm1
+; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm1 = zero,xmm1[1,2],xmm6[2]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 512(%rdi), %ymm0
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm5[1,1],ymm0[2,2],ymm5[5,5],ymm0[6,6]
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm1[2,3,2,3]
-; AVX1-ONLY-NEXT: vmovaps 480(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm0[0],xmm7[1],xmm0[2,3]
-; AVX1-ONLY-NEXT: vmovaps %xmm7, %xmm12
-; AVX1-ONLY-NEXT: vmovaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm2 = xmm2[1,0],mem[3,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4],ymm2[5,6,7]
-; AVX1-ONLY-NEXT: vmovaps 576(%rdi), %ymm0
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm5[1,1],ymm0[2,2],ymm5[5,5],ymm0[6,6]
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,2,3]
+; AVX1-ONLY-NEXT: vmovaps 480(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0],xmm8[1],xmm1[2,3]
+; AVX1-ONLY-NEXT: vmovaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm1[1,0],mem[3,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps 608(%rdi), %ymm7
+; AVX1-ONLY-NEXT: vmovaps 576(%rdi), %ymm2
+; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm2[2,3],ymm7[0,1]
+; AVX1-ONLY-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm2[0,0],ymm1[3,3],ymm2[4,4],ymm1[7,7]
+; AVX1-ONLY-NEXT: vextractf128 $1, %ymm1, %xmm1
+; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm1 = zero,xmm1[1,2],xmm11[2]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 288(%rdi), %ymm0
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 608(%rdi), %ymm2
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm12[1,1],ymm0[2,2],ymm12[5,5],ymm0[6,6]
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,2,3]
+; AVX1-ONLY-NEXT: vmovaps 256(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0],xmm14[1],xmm1[2,3]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm1[1,0],mem[3,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps 384(%rdi), %ymm2
; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm5 = ymm2[2,3,0,1]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm5 = ymm0[0,0],ymm5[3,3],ymm0[4,4],ymm5[7,7]
+; AVX1-ONLY-NEXT: vmovaps 352(%rdi), %ymm1
+; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm5 = ymm1[2,3],ymm2[0,1]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm5 = ymm1[0,0],ymm5[3,3],ymm1[4,4],ymm5[7,7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm5, %xmm5
-; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm5 = zero,xmm5[1,2],xmm13[2]
+; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm5 = zero,xmm5[1,2],xmm15[2]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm5, %ymm0, %ymm5
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm5[5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 288(%rdi), %ymm0
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm4[1,1],ymm0[2,2],ymm4[5,5],ymm0[6,6]
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm1[2,3,2,3]
-; AVX1-ONLY-NEXT: vmovaps 256(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm5 = xmm0[0],xmm9[1],xmm0[2,3]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm5 = xmm5[1,0],mem[3,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2],ymm1[3,4],ymm5[5,6,7]
-; AVX1-ONLY-NEXT: vmovaps 352(%rdi), %ymm0
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm5[5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 384(%rdi), %ymm1
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm7 = ymm1[2,3,0,1]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm7 = ymm0[0,0],ymm7[3,3],ymm0[4,4],ymm7[7,7]
-; AVX1-ONLY-NEXT: vextractf128 $1, %ymm7, %xmm7
-; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm7 = zero,xmm7[1,2],xmm6[2]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm7, %ymm0, %ymm7
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4],ymm7[5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 736(%rdi), %ymm8
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm7 = ymm11[1,1],ymm8[2,2],ymm11[5,5],ymm8[6,6]
-; AVX1-ONLY-NEXT: vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm7 = ymm7[2,3,2,3]
-; AVX1-ONLY-NEXT: vmovaps 704(%rdi), %xmm3
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm14 = xmm3[0],xmm15[1],xmm3[2,3]
-; AVX1-ONLY-NEXT: vmovaps %xmm15, %xmm1
-; AVX1-ONLY-NEXT: vmovaps %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm14 = xmm14[1,0],mem[3,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1,2],ymm7[3,4],ymm14[5,6,7]
-; AVX1-ONLY-NEXT: vmovaps 800(%rdi), %ymm2
-; AVX1-ONLY-NEXT: vmovaps 832(%rdi), %ymm6
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm15 = ymm6[2,3,0,1]
+; AVX1-ONLY-NEXT: vmovaps 736(%rdi), %ymm6
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm13[1,1],ymm6[2,2],ymm13[5,5],ymm6[6,6]
; AVX1-ONLY-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm2[0,0],ymm15[3,3],ymm2[4,4],ymm15[7,7]
-; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,2,3]
+; AVX1-ONLY-NEXT: vmovaps 704(%rdi), %xmm12
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm14 = xmm12[0],xmm2[1],xmm12[2,3]
+; AVX1-ONLY-NEXT: vmovaps %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm14 = xmm14[1,0],mem[3,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1,2],ymm0[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps 832(%rdi), %ymm3
+; AVX1-ONLY-NEXT: vmovaps 800(%rdi), %ymm1
+; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm15 = ymm1[2,3],ymm3[0,1]
+; AVX1-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm1[0,0],ymm15[3,3],ymm1[4,4],ymm15[7,7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm15, %xmm15
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm15 = zero,xmm15[1,2],xmm4[2]
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm15 = zero,xmm15[1,2],xmm0[2]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm15, %ymm0, %ymm15
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4],ymm15[5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm14 = xmm11[2,3,2,3]
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm14 = xmm14[0],xmm0[1],xmm14[2,3]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm15[3,1],ymm10[0,3],ymm15[7,5],ymm10[4,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm9[2,1],ymm15[2,0],ymm9[6,5],ymm15[6,4]
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm14 = xmm10[2,3,2,3]
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm14 = xmm14[0],xmm5[1],xmm14[2,3]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm15[3,1],ymm11[0,3],ymm15[7,5],ymm11[4,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm10[2,1],ymm15[2,0],ymm10[6,5],ymm15[6,4]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1],ymm15[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
-; AVX1-ONLY-NEXT: vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm15 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm15 = ymm7[0],mem[0],ymm7[2],mem[2]
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm15 = ymm4[0],ymm9[0],ymm4[2],ymm9[2]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm15, %xmm15
; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm15 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm15 = xmm15[0,1,2],mem[3]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm15, %ymm0, %ymm15
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4],ymm15[5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm14 = xmm12[2,3,2,3]
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm14 = xmm14[0],xmm5[1],xmm14[2,3]
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm14 = xmm8[2,3,2,3]
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm14 = xmm14[0],xmm1[1],xmm14[2,3]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps $199, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm15 = ymm15[3,1],mem[0,3],ymm15[7,5],mem[4,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm12[2,1],ymm15[2,0],ymm12[6,5],ymm15[6,4]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm15[3,1],ymm9[0,3],ymm15[7,5],ymm9[4,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm8[2,1],ymm15[2,0],ymm8[6,5],ymm15[6,4]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1],ymm15[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
-; AVX1-ONLY-NEXT: vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm15 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm15 = ymm12[0],mem[0],ymm12[2],mem[2]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm15 = ymm4[0],ymm7[0],ymm4[2],ymm7[2]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm15, %xmm15
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm15 = xmm15[0,1,2],xmm13[3]
+; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm15 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm15 = xmm15[0,1,2],mem[3]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm15, %ymm0, %ymm15
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4],ymm15[5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm14 = xmm1[2,3,2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm14 = xmm14[0],xmm3[1],xmm14[2,3]
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm14 = xmm2[2,3,2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm14 = xmm14[0],xmm12[1],xmm14[2,3]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm15[3,1],ymm8[0,3],ymm15[7,5],ymm8[4,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm1[2,1],ymm15[2,0],ymm1[6,5],ymm15[6,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm15[3,1],ymm6[0,3],ymm15[7,5],ymm6[4,7]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm13[2,1],ymm15[2,0],ymm13[6,5],ymm15[6,4]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1],ymm15[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm15 = ymm2[0],ymm6[0],ymm2[2],ymm6[2]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm15 = ymm2[0],ymm3[0],ymm2[2],ymm3[2]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm15, %xmm15
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm6 = xmm15[0,1,2],xmm4[3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm6 = xmm15[0,1,2],xmm0[3]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm6, %ymm0, %ymm6
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm14[0,1,2,3,4],ymm6[5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm6 = xmm13[2,3,2,3]
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm6 = xmm6[0],xmm1[1],xmm6[2,3]
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm6 = xmm6[0],xmm0[1],xmm6[2,3]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm14 = ymm14[3,1],ymm4[0,3],ymm14[7,5],ymm4[4,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm14 = ymm3[2,1],ymm14[2,0],ymm3[6,5],ymm14[6,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm14 = ymm14[3,1],ymm3[0,3],ymm14[7,5],ymm3[4,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm14 = ymm7[2,1],ymm14[2,0],ymm7[6,5],ymm14[6,4]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1],ymm14[2,3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm14 = ymm2[0],ymm15[0],ymm2[2],ymm15[2]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm14 = ymm12[0],ymm15[0],ymm12[2],ymm15[2]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm14, %xmm14
; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14, %xmm14 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm14 = xmm14[0,1,2],mem[3]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm14, %ymm0, %ymm14
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4],ymm14[5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm6 = ymm10[1,0],ymm9[0,0],ymm10[5,4],ymm9[4,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm6 = ymm9[3,1],ymm6[0,2],ymm9[7,5],ymm6[4,6]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm10 = xmm0[0,1,2],xmm11[3]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm6 = ymm11[1,0],ymm10[0,0],ymm11[5,4],ymm10[4,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm6 = ymm10[3,1],ymm6[0,2],ymm10[7,5],ymm6[4,6]
+; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm10 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm10 = xmm5[0,1,2],mem[3]
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm10 = xmm10[3,2,2,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm10[0,1],ymm6[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovaps 192(%rdi), %ymm0
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 192(%rdi), %ymm2
+; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm10 = ymm0[0,1],ymm11[1,3],ymm0[4,5],ymm11[5,7]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm10 = ymm7[0,2],ymm10[2,0],ymm7[4,6],ymm10[6,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm10 = ymm2[0,1],ymm11[1,3],ymm2[4,5],ymm11[5,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm10 = ymm2[0,2],ymm10[2,0],ymm2[4,6],ymm10[6,4]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4],ymm10[5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm6 = ymm0[1,0],ymm10[0,0],ymm0[5,4],ymm10[4,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm6 = ymm10[3,1],ymm6[0,2],ymm10[7,5],ymm6[4,6]
-; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm9 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm9 = xmm5[0,1,2],mem[3]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm6 = ymm9[1,0],ymm8[0,0],ymm9[5,4],ymm8[4,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm6 = ymm8[3,1],ymm6[0,2],ymm8[7,5],ymm6[4,6]
+; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm9 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm9 = xmm1[0,1,2],mem[3]
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm9 = xmm9[3,2,2,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm9 = ymm9[0,1],ymm6[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovaps 640(%rdi), %ymm0
+; AVX1-ONLY-NEXT: vmovaps 640(%rdi), %ymm6
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm10 = ymm0[0,1],ymm14[1,3],ymm0[4,5],ymm14[5,7]
-; AVX1-ONLY-NEXT: vmovaps %ymm0, %ymm7
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm10 = ymm12[0,2],ymm10[2,0],ymm12[4,6],ymm10[6,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm10 = ymm6[0,1],ymm14[1,3],ymm6[4,5],ymm14[5,7]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm10 = ymm4[0,2],ymm10[2,0],ymm4[4,6],ymm10[6,4]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm9 = ymm9[0,1,2,3,4],ymm10[5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm8 = ymm4[1,0],ymm3[0,0],ymm4[5,4],ymm3[4,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm3[3,1],ymm8[0,2],ymm3[7,5],ymm8[4,6]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm8 = xmm1[0,1,2],xmm13[3]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm8 = ymm3[1,0],ymm7[0,0],ymm3[5,4],ymm7[4,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm7 = ymm7[3,1],ymm8[0,2],ymm7[7,5],ymm8[4,6]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm8 = xmm0[0,1,2],xmm13[3]
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm8 = xmm8[3,2,2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm8[0,1],ymm0[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovaps 416(%rdi), %ymm1
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm8 = ymm1[0,1],ymm15[1,3],ymm1[4,5],ymm15[5,7]
-; AVX1-ONLY-NEXT: vmovaps %ymm15, %ymm9
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm8 = ymm2[0,2],ymm8[2,0],ymm2[4,6],ymm8[6,4]
-; AVX1-ONLY-NEXT: vmovaps %ymm2, %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm8[5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm8[0,1],ymm7[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps 416(%rdi), %ymm9
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm8 = ymm9[0,1],ymm15[1,3],ymm9[4,5],ymm15[5,7]
+; AVX1-ONLY-NEXT: vmovaps %ymm12, %ymm1
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm8 = ymm12[0,2],ymm8[2,0],ymm12[4,6],ymm8[6,4]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4],ymm8[5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[1,0],ymm2[0,0],ymm0[5,4],ymm2[4,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm2[3,1],ymm0[0,2],ymm2[7,5],ymm0[4,6]
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX1-ONLY-NEXT: vblendps $7, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm4 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm4 = mem[0,1,2],xmm2[3]
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm4 = xmm4[3,2,2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm4[0,1],ymm0[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovaps 864(%rdi), %ymm5
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm5 = ymm3[1,0],ymm0[0,0],ymm3[5,4],ymm0[4,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm4 = ymm0[3,1],ymm5[0,2],ymm0[7,5],ymm5[4,6]
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT: vblendps $7, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm5 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm5 = mem[0,1,2],xmm0[3]
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm5 = xmm5[3,2,2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm5[0,1],ymm4[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps 864(%rdi), %ymm15
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm4 = ymm5[0,1],ymm13[1,3],ymm5[4,5],ymm13[5,7]
-; AVX1-ONLY-NEXT: vmovaps %ymm5, %ymm15
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm5 = ymm15[0,1],ymm13[1,3],ymm15[4,5],ymm13[5,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm4 = ymm8[0,2],ymm4[2,0],ymm8[4,6],ymm4[6,4]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm4[5,6,7]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm5 = ymm8[0,2],ymm5[2,0],ymm8[4,6],ymm5[6,4]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm4[0,1,2,3,4],ymm5[5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm2[2,3,0,1]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm2[3,0],ymm0[0,0],ymm2[7,4],ymm0[4,4]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm6[1,0],ymm11[2,0],ymm6[5,4],ymm11[6,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm3[2,0],ymm0[6,4],ymm3[6,4]
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm4 = ymm2[2,3,0,1]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm2[3,0],ymm4[0,0],ymm2[7,4],ymm4[4,4]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm4 = ymm5[1,0],ymm11[2,0],ymm5[5,4],ymm11[6,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm3[2,0],ymm4[2,0],ymm3[6,4],ymm4[6,4]
; AVX1-ONLY-NEXT: vmovaps 64(%rdi), %xmm10
-; AVX1-ONLY-NEXT: vmovaps 96(%rdi), %xmm11
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm3 = xmm11[0,1,0,1]
-; AVX1-ONLY-NEXT: vmovaps %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm3 = xmm10[0,1,2],xmm3[3]
-; AVX1-ONLY-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm4 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm4 = mem[0],xmm4[1],mem[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm3 = xmm4[0,1],xmm3[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps 96(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm4 = xmm0[0,1,0,1]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm4 = xmm10[0,1,2],xmm4[3]
+; AVX1-ONLY-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm7 = mem[2,3,2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm7 = mem[0],xmm7[1],mem[2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm4 = xmm7[0,1],xmm4[2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm4[0,1,2,3],ymm3[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm12[2,3,0,1]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm12[3,0],ymm0[0,0],ymm12[7,4],ymm0[4,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm7[1,0],ymm14[2,0],ymm7[5,4],ymm14[6,4]
-; AVX1-ONLY-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm2[2,0],ymm0[6,4],ymm2[6,4]
-; AVX1-ONLY-NEXT: vmovaps 544(%rdi), %xmm2
-; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm3 = xmm2[0,1,0,1]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm3 = ymm0[2,3,0,1]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm0[3,0],ymm3[0,0],ymm0[7,4],ymm3[4,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm6[1,0],ymm14[2,0],ymm6[5,4],ymm14[6,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm2[2,0],ymm3[2,0],ymm2[6,4],ymm3[6,4]
+; AVX1-ONLY-NEXT: vmovaps 544(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm4 = xmm0[0,1,0,1]
; AVX1-ONLY-NEXT: vmovaps 512(%rdi), %xmm12
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm3 = xmm12[0,1,2],xmm3[3]
-; AVX1-ONLY-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm4 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm4 = mem[0],xmm4[1],mem[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm3 = xmm4[0,1],xmm3[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm4 = xmm12[0,1,2],xmm4[3]
+; AVX1-ONLY-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm7 = mem[2,3,2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm7 = mem[0],xmm7[1],mem[2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm4 = xmm7[0,1],xmm4[2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm4[0,1,2,3],ymm3[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm1[2,3,0,1]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm1[3,0],ymm0[0,0],ymm1[7,4],ymm0[4,4]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm5[1,0],ymm9[2,0],ymm5[5,4],ymm9[6,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm0[2,0],ymm1[2,0],ymm0[6,4],ymm1[6,4]
-; AVX1-ONLY-NEXT: vmovaps 320(%rdi), %xmm9
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm3 = xmm9[0,1,0,1]
-; AVX1-ONLY-NEXT: vmovaps %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 288(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm3 = xmm0[0,1,2],xmm3[3]
-; AVX1-ONLY-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm4 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm4 = mem[0],xmm4[1],mem[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm3 = xmm4[0,1],xmm3[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm3[0,1,2,3],ymm1[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm8[2,3,0,1]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm8[3,0],ymm1[0,0],ymm8[7,4],ymm1[4,4]
-; AVX1-ONLY-NEXT: vmovaps %ymm15, %ymm14
-; AVX1-ONLY-NEXT: vmovups %ymm15, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm3 = ymm1[2,3,0,1]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm1[3,0],ymm3[0,0],ymm1[7,4],ymm3[4,4]
+; AVX1-ONLY-NEXT: vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm9[1,0],ymm14[2,0],ymm9[5,4],ymm14[6,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm1[2,0],ymm3[2,0],ymm1[6,4],ymm3[6,4]
+; AVX1-ONLY-NEXT: vmovaps 320(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm4 = xmm0[0,1,0,1]
+; AVX1-ONLY-NEXT: vmovaps 288(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm4 = xmm1[0,1,2],xmm4[3]
+; AVX1-ONLY-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm7 = mem[2,3,2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm7 = mem[0],xmm7[1],mem[2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm4 = xmm7[0,1],xmm4[2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm4[0,1,2,3],ymm3[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm3 = ymm8[2,3,0,1]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm8[3,0],ymm3[0,0],ymm8[7,4],ymm3[4,4]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm15[1,0],ymm13[2,0],ymm15[5,4],ymm13[6,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm1[2,0],ymm3[2,0],ymm1[6,4],ymm3[6,4]
-; AVX1-ONLY-NEXT: vmovaps 768(%rdi), %xmm8
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm3 = xmm8[0,1,0,1]
-; AVX1-ONLY-NEXT: vmovaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 736(%rdi), %xmm15
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm3 = xmm15[0,1,2],xmm3[3]
+; AVX1-ONLY-NEXT: vmovups %ymm15, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm3[2,0],ymm0[6,4],ymm3[6,4]
+; AVX1-ONLY-NEXT: vmovaps 768(%rdi), %xmm11
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm3 = xmm11[0,1,0,1]
+; AVX1-ONLY-NEXT: vmovaps %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 736(%rdi), %xmm8
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm3 = xmm8[0,1,2],xmm3[3]
; AVX1-ONLY-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm13 = mem[2,3,2,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm13 = mem[0],xmm13[1],mem[2,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm3 = xmm13[0,1],xmm3[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm3[0,1,2,3],ymm1[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufps $246, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm1 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm1 = ymm6[2,1],mem[3,3],ymm6[6,5],mem[7,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vshufps $246, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm0 = ymm5[2,1],mem[3,3],ymm5[6,5],mem[7,7]
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX1-ONLY-NEXT: vblendps $13, (%rsp), %xmm2, %xmm13 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: vblendps $13, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm13 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm13 = mem[0],xmm2[1],mem[2,3]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm13, %ymm0, %ymm13
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm13[1,0],ymm1[2,0],ymm13[5,4],ymm1[6,4]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm10 = xmm11[0,1,2],xmm10[3]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm13[1,0],ymm0[2,0],ymm13[5,4],ymm0[6,4]
+; AVX1-ONLY-NEXT: vblendps $7, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10, %xmm10 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm10 = mem[0,1,2],xmm10[3]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX1-ONLY-NEXT: vshufps $16, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm13 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm13 = ymm3[0,0],mem[1,0],ymm3[4,4],mem[5,4]
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} ymm13 = ymm13[2,0,2,3,6,4,6,7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm13, %xmm13
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm10 = xmm13[0,1],xmm10[3,2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm10[0,1,2,3],ymm1[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufps $246, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm10 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm10 = ymm7[2,1],mem[3,3],ymm7[6,5],mem[7,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm10[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vshufps $246, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm10 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm10 = ymm6[2,1],mem[3,3],ymm6[6,5],mem[7,7]
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm13 = xmm2[0],xmm1[1],xmm2[2,3]
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm13 = xmm2[0],xmm0[1],xmm2[2,3]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm13, %ymm0, %ymm13
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm10 = ymm13[1,0],ymm10[2,0],ymm13[5,4],ymm10[6,4]
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm12 = xmm7[0,1,2],xmm12[3]
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm12 = xmm5[0,1,2],xmm12[3]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm13 = ymm3[0,0],ymm4[1,0],ymm3[4,4],ymm4[5,4]
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} ymm13 = ymm13[2,0,2,3,6,4,6,7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm13, %xmm13
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm12 = xmm13[0,1],xmm12[3,2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm12[0,1,2,3],ymm10[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufps $246, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm12 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm12 = ymm5[2,1],mem[3,3],ymm5[6,5],mem[7,7]
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm13 = xmm10[0],xmm11[1],xmm10[2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm12[0,1,2,3],ymm10[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm12 = ymm9[2,1],ymm14[3,3],ymm9[6,5],ymm14[7,7]
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm13 = xmm9[0],xmm7[1],xmm9[2,3]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm13, %ymm0, %ymm13
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm12 = ymm13[1,0],ymm12[2,0],ymm13[5,4],ymm12[6,4]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm9[0,1,2],xmm0[3]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps $16, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm13 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm13 = ymm9[0,0],mem[1,0],ymm9[4,4],mem[5,4]
+; AVX1-ONLY-NEXT: vblendps $7, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm1 = mem[0,1,2],xmm1[3]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm13 = ymm10[0,0],ymm14[1,0],ymm10[4,4],ymm14[5,4]
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} ymm13 = ymm13[2,0,2,3,6,4,6,7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm13, %xmm13
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm0 = xmm13[0,1],xmm0[3,2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm12[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufps $246, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm0 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm0 = ymm14[2,1],mem[3,3],ymm14[6,5],mem[7,7]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm13[0,1],xmm1[3,2]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm12[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vshufps $246, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm1 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm1 = ymm15[2,1],mem[3,3],ymm15[6,5],mem[7,7]
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
; AVX1-ONLY-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12, %xmm13 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm13 = xmm12[0],mem[1],xmm12[2,3]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm13, %ymm0, %ymm13
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm13[1,0],ymm0[2,0],ymm13[5,4],ymm0[6,4]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm13 = xmm8[0,1,2],xmm15[3]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps $16, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm15 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm15 = ymm14[0,0],mem[1,0],ymm14[4,4],mem[5,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm13[1,0],ymm1[2,0],ymm13[5,4],ymm1[6,4]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm13 = xmm11[0,1,2],xmm8[3]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps $16, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm15 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm15 = ymm11[0,0],mem[1,0],ymm11[4,4],mem[5,4]
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} ymm15 = ymm15[2,0,2,3,6,4,6,7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm15, %xmm15
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm13 = xmm15[0,1],xmm13[3,2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm15 = ymm13[0,1,2,3],ymm0[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm5[2,3,0,1]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm5[3,0],ymm0[0,0],ymm5[7,4],ymm0[4,4]
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm6 = xmm1[2,3,2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm15 = ymm13[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm6[2,3,0,1]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm6[3,0],ymm1[0,0],ymm6[7,4],ymm1[4,4]
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm6 = xmm0[2,3,2,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm6 = xmm6[0],xmm2[1],xmm6[2,3]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm6, %ymm0, %ymm6
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm6[0,1],ymm0[2,0],ymm6[4,5],ymm0[6,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm6[0,1],ymm1[2,0],ymm6[4,5],ymm1[6,4]
; AVX1-ONLY-NEXT: vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm6 = mem[0,1,0,1]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm6[0,1,2],xmm7[3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm6[0,1,2],xmm5[3]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm6 = ymm3[1,0],ymm4[2,0],ymm3[5,4],ymm4[6,4]
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} ymm6 = ymm6[2,0,2,3,6,4,6,7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm6, %xmm6
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm6[0,1],xmm2[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm0[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm1[2,3,0,1]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm1[3,0],ymm0[0,0],ymm1[7,4],ymm0[4,4]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm0[2,3,0,1]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm0[3,0],ymm1[0,0],ymm0[7,4],ymm1[4,4]
; AVX1-ONLY-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm6 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT: vblendps $2, (%rsp), %xmm6, %xmm6 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6, %xmm6 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm6 = xmm6[0],mem[1],xmm6[2,3]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm6, %ymm0, %ymm6
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm6[0,1],ymm0[2,0],ymm6[4,5],ymm0[6,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm6[0,1],ymm1[2,0],ymm6[4,5],ymm1[6,4]
; AVX1-ONLY-NEXT: vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm6 = mem[0,1,0,1]
; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6, %xmm5 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm5 = xmm6[0,1,2],mem[3]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps $33, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm6 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm6 = ymm1[1,0],mem[2,0],ymm1[5,4],mem[6,4]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps $33, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm6 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm6 = ymm0[1,0],mem[2,0],ymm0[5,4],mem[6,4]
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} ymm6 = ymm6[2,0,2,3,6,4,6,7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm6, %xmm6
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm5 = xmm6[0,1],xmm5[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm0[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm1[2,3,0,1]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm1[3,0],ymm0[0,0],ymm1[7,4],ymm0[4,4]
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm6 = xmm11[2,3,2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm6 = xmm6[0],xmm10[1],xmm6[2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm0[2,3,0,1]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm0[3,0],ymm1[0,0],ymm0[7,4],ymm1[4,4]
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm6 = xmm7[2,3,2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm6 = xmm6[0],xmm9[1],xmm6[2,3]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm6, %ymm0, %ymm6
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm6[0,1],ymm0[2,0],ymm6[4,5],ymm0[6,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm6[0,1],ymm1[2,0],ymm6[4,5],ymm1[6,4]
; AVX1-ONLY-NEXT: vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm6 = mem[0,1,0,1]
; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6, %xmm6 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm6 = xmm6[0,1,2],mem[3]
-; AVX1-ONLY-NEXT: vshufps $33, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm9 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm9 = ymm9[1,0],mem[2,0],ymm9[5,4],mem[6,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm9 = ymm10[1,0],ymm14[2,0],ymm10[5,4],ymm14[6,4]
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} ymm9 = ymm9[2,0,2,3,6,4,6,7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm9, %xmm9
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm6 = xmm9[0,1],xmm6[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm6[0,1,2,3],ymm0[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm6 = ymm1[2,3,0,1]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm6 = ymm1[3,0],ymm6[0,0],ymm1[7,4],ymm6[4,4]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm6[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm6 = ymm0[2,3,0,1]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm6 = ymm0[3,0],ymm6[0,0],ymm0[7,4],ymm6[4,4]
; AVX1-ONLY-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm8 = mem[2,3,2,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm8 = xmm8[0],xmm12[1],xmm8[2,3]
@@ -3956,66 +3946,66 @@ define void @load_i32_stride7_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: # xmm8 = mem[0,1,0,1]
; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8, %xmm7 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm7 = xmm8[0,1,2],mem[3]
-; AVX1-ONLY-NEXT: vshufps $33, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm8 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm8 = ymm14[1,0],mem[2,0],ymm14[5,4],mem[6,4]
+; AVX1-ONLY-NEXT: vshufps $33, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm8 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm8 = ymm11[1,0],mem[2,0],ymm11[5,4],mem[6,4]
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} ymm8 = ymm8[2,0,2,3,6,4,6,7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm8, %xmm8
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm7 = xmm8[0,1],xmm7[2,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm7[0,1,2,3],ymm6[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm1, 96(%rsi)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm0, 96(%rsi)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm3, 32(%rsi)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm7, 64(%rsi)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm7, (%rsi)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm1, 96(%rdx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm1, 32(%rdx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm1, 64(%rdx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm1, (%rdx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm1, 32(%rcx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm1, 96(%rcx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm1, 64(%rcx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm1, (%rcx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm1, 96(%r8)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm1, 32(%r8)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm1, 64(%r8)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm1, (%r8)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm1, 96(%r9)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm1, 32(%r9)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm1, 64(%r9)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm1, (%r9)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm0, 96(%rdx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm0, 32(%rdx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm0, 64(%rdx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm0, (%rdx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm0, 32(%rcx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm0, 96(%rcx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm0, 64(%rcx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm0, (%rcx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm0, 96(%r8)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm0, 32(%r8)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm0, 64(%r8)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm0, (%r8)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm0, 96(%r9)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm0, 32(%r9)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm0, 64(%r9)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm0, (%r9)
; AVX1-ONLY-NEXT: movq {{[0-9]+}}(%rsp), %rax
; AVX1-ONLY-NEXT: vmovaps %ymm15, 96(%rax)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm1, 32(%rax)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm1, 64(%rax)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm1, (%rax)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm0, 32(%rax)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm0, 64(%rax)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm0, (%rax)
; AVX1-ONLY-NEXT: movq {{[0-9]+}}(%rsp), %rax
; AVX1-ONLY-NEXT: vmovaps %ymm6, 96(%rax)
-; AVX1-ONLY-NEXT: vmovaps %ymm0, 32(%rax)
+; AVX1-ONLY-NEXT: vmovaps %ymm1, 32(%rax)
; AVX1-ONLY-NEXT: vmovaps %ymm5, (%rax)
; AVX1-ONLY-NEXT: vmovaps %ymm2, 64(%rax)
-; AVX1-ONLY-NEXT: addq $1496, %rsp # imm = 0x5D8
+; AVX1-ONLY-NEXT: addq $1528, %rsp # imm = 0x5F8
; AVX1-ONLY-NEXT: vzeroupper
; AVX1-ONLY-NEXT: retq
;
@@ -4032,7 +4022,7 @@ define void @load_i32_stride7_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-SLOW-NEXT: vmovdqa 32(%rdi), %ymm6
; AVX2-SLOW-NEXT: vmovdqa 96(%rdi), %ymm10
; AVX2-SLOW-NEXT: vpbroadcastq 80(%rdi), %ymm0
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm0[0,1,2,3],ymm10[4,5],ymm0[6,7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm0[0,1,2,3],ymm10[4,5,6,7]
; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} xmm0 = <0,7,6,u>
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm6[6],ymm3[7]
; AVX2-SLOW-NEXT: vmovdqa %ymm6, %ymm12
@@ -4040,7 +4030,7 @@ define void @load_i32_stride7_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-SLOW-NEXT: vmovdqa %ymm3, %ymm13
; AVX2-SLOW-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vpermd %ymm2, %ymm0, %ymm2
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4],ymm2[5,6,7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4,5,6,7]
; AVX2-SLOW-NEXT: vmovdqa 128(%rdi), %xmm2
; AVX2-SLOW-NEXT: vmovdqa 160(%rdi), %xmm3
; AVX2-SLOW-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -4056,8 +4046,8 @@ define void @load_i32_stride7_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-SLOW-NEXT: vmovdqu %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vpermd %ymm1, %ymm0, %ymm1
; AVX2-SLOW-NEXT: vpbroadcastq 528(%rdi), %ymm2
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm11[4,5],ymm2[6,7]
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4],ymm1[5,6,7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm11[4,5,6,7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
; AVX2-SLOW-NEXT: vmovdqa 576(%rdi), %xmm2
; AVX2-SLOW-NEXT: vmovdqa 608(%rdi), %xmm3
; AVX2-SLOW-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -4073,10 +4063,10 @@ define void @load_i32_stride7_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-SLOW-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vpermd %ymm1, %ymm0, %ymm1
; AVX2-SLOW-NEXT: vpbroadcastq 304(%rdi), %ymm2
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm9[4,5],ymm2[6,7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm9[4,5,6,7]
; AVX2-SLOW-NEXT: vmovdqa %ymm9, %ymm15
; AVX2-SLOW-NEXT: vmovdqu %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4],ymm1[5,6,7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
; AVX2-SLOW-NEXT: vmovdqa 352(%rdi), %xmm2
; AVX2-SLOW-NEXT: vmovdqa 384(%rdi), %xmm3
; AVX2-SLOW-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -4096,8 +4086,8 @@ define void @load_i32_stride7_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-SLOW-NEXT: vpermd %ymm1, %ymm0, %ymm0
; AVX2-SLOW-NEXT: vmovdqa 768(%rdi), %ymm9
; AVX2-SLOW-NEXT: vpbroadcastq 752(%rdi), %ymm1
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm9[4,5],ymm1[6,7]
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm9[4,5,6,7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
; AVX2-SLOW-NEXT: vmovdqa 800(%rdi), %xmm1
; AVX2-SLOW-NEXT: vmovdqa 832(%rdi), %xmm2
; AVX2-SLOW-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -4126,7 +4116,7 @@ define void @load_i32_stride7_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-SLOW-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm2 = ymm2[1,0,3,3,5,4,7,7]
; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm2 = ymm2[0,3,2,3]
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4],ymm2[5,6,7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4,5,6,7]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7]
; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm0 = mem[2,2,2,2]
@@ -4147,7 +4137,7 @@ define void @load_i32_stride7_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-SLOW-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm2 = ymm2[1,0,3,3,5,4,7,7]
; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm2 = ymm2[0,3,2,3]
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4],ymm2[5,6,7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4,5,6,7]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7]
; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm0 = mem[2,2,2,2]
@@ -4171,7 +4161,7 @@ define void @load_i32_stride7_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-SLOW-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm2 = ymm2[1,0,3,3,5,4,7,7]
; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm2 = ymm2[0,3,2,3]
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4],ymm2[5,6,7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4,5,6,7]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7]
; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm0 = mem[2,2,2,2]
@@ -4193,7 +4183,7 @@ define void @load_i32_stride7_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-SLOW-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm3 = ymm3[1,0,3,3,5,4,7,7]
; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm3 = ymm3[0,3,2,3]
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3,4],ymm3[5,6,7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3,4,5,6,7]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3,4],ymm1[5,6,7]
; AVX2-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovdqa 80(%rdi), %xmm1
@@ -4310,8 +4300,8 @@ define void @load_i32_stride7_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm1 = xmm0[0,1,2],xmm1[3]
; AVX2-SLOW-NEXT: vmovaps {{.*#+}} xmm3 = <4,3,u,u>
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm2 = mem[0,1,2,3],ymm2[4,5],mem[6,7]
+; AVX2-SLOW-NEXT: vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm2 = mem[0,1,2,3],ymm2[4,5,6,7]
; AVX2-SLOW-NEXT: vpermps %ymm2, %ymm3, %ymm2
; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3]
; AVX2-SLOW-NEXT: vbroadcastsd {{.*#+}} ymm4 = [30064771072,30064771072,30064771072,30064771072]
@@ -4324,8 +4314,8 @@ define void @load_i32_stride7_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm1 = mem[0,1,2,3],ymm1[4,5],mem[6,7]
+; AVX2-SLOW-NEXT: vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm1 = mem[0,1,2,3],ymm1[4,5,6,7]
; AVX2-SLOW-NEXT: vpermps %ymm1, %ymm3, %ymm2
; AVX2-SLOW-NEXT: vbroadcastss 548(%rdi), %xmm7
; AVX2-SLOW-NEXT: vmovaps 512(%rdi), %xmm1
@@ -4339,8 +4329,8 @@ define void @load_i32_stride7_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm7[4,5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm2 = mem[0,1,2,3],ymm2[4,5],mem[6,7]
+; AVX2-SLOW-NEXT: vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm2 = mem[0,1,2,3],ymm2[4,5,6,7]
; AVX2-SLOW-NEXT: vpermps %ymm2, %ymm3, %ymm7
; AVX2-SLOW-NEXT: vbroadcastss 324(%rdi), %xmm8
; AVX2-SLOW-NEXT: vmovaps 288(%rdi), %xmm2
@@ -4353,8 +4343,8 @@ define void @load_i32_stride7_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3],ymm8[4,5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm7 = mem[0,1,2,3],ymm7[4,5],mem[6,7]
+; AVX2-SLOW-NEXT: vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm7 = mem[0,1,2,3],ymm7[4,5,6,7]
; AVX2-SLOW-NEXT: vpermps %ymm7, %ymm3, %ymm3
; AVX2-SLOW-NEXT: vbroadcastss 772(%rdi), %xmm8
; AVX2-SLOW-NEXT: vmovaps 736(%rdi), %xmm7
@@ -4551,7 +4541,7 @@ define void @load_i32_stride7_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-NEXT: vmovdqa 32(%rdi), %ymm6
; AVX2-FAST-NEXT: vmovdqa 96(%rdi), %ymm10
; AVX2-FAST-NEXT: vpbroadcastq 80(%rdi), %ymm0
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm0[0,1,2,3],ymm10[4,5],ymm0[6,7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm0[0,1,2,3],ymm10[4,5,6,7]
; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm0 = <0,7,6,u>
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm6[6],ymm3[7]
; AVX2-FAST-NEXT: vmovdqa %ymm6, %ymm11
@@ -4559,7 +4549,7 @@ define void @load_i32_stride7_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-NEXT: vmovdqa %ymm3, %ymm12
; AVX2-FAST-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vpermd %ymm2, %ymm0, %ymm2
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4],ymm2[5,6,7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4,5,6,7]
; AVX2-FAST-NEXT: vmovdqa 128(%rdi), %xmm2
; AVX2-FAST-NEXT: vmovdqa 160(%rdi), %xmm3
; AVX2-FAST-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -4573,8 +4563,8 @@ define void @load_i32_stride7_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-NEXT: vmovdqu %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vpermd %ymm1, %ymm0, %ymm1
; AVX2-FAST-NEXT: vpbroadcastq 528(%rdi), %ymm2
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm13[4,5],ymm2[6,7]
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4],ymm1[5,6,7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm13[4,5,6,7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
; AVX2-FAST-NEXT: vmovdqa 576(%rdi), %xmm2
; AVX2-FAST-NEXT: vmovdqa 608(%rdi), %xmm3
; AVX2-FAST-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -4592,8 +4582,8 @@ define void @load_i32_stride7_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-NEXT: vpermd %ymm1, %ymm0, %ymm1
; AVX2-FAST-NEXT: vpbroadcastq 304(%rdi), %ymm2
; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm7[4,5],ymm2[6,7]
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4],ymm1[5,6,7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm7[4,5,6,7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
; AVX2-FAST-NEXT: vmovdqa 352(%rdi), %xmm2
; AVX2-FAST-NEXT: vmovdqa 384(%rdi), %xmm3
; AVX2-FAST-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -4613,8 +4603,8 @@ define void @load_i32_stride7_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-NEXT: vmovdqa 768(%rdi), %ymm2
; AVX2-FAST-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vpbroadcastq 752(%rdi), %ymm1
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
; AVX2-FAST-NEXT: vmovdqa 800(%rdi), %xmm1
; AVX2-FAST-NEXT: vmovdqa 832(%rdi), %xmm2
; AVX2-FAST-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -4641,7 +4631,7 @@ define void @load_i32_stride7_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm4 = ymm11[0],ymm12[1],ymm11[2,3,4],ymm12[5],ymm11[6,7]
; AVX2-FAST-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vpermd %ymm4, %ymm1, %ymm4
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3,4],ymm4[5,6,7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3,4,5,6,7]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4],ymm2[5,6,7]
; AVX2-FAST-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vpshufd {{.*#+}} xmm2 = mem[2,2,2,2]
@@ -4663,7 +4653,7 @@ define void @load_i32_stride7_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-NEXT: # ymm4 = ymm9[0],mem[1],ymm9[2,3,4],mem[5],ymm9[6,7]
; AVX2-FAST-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vpermd %ymm4, %ymm1, %ymm4
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3,4],ymm4[5,6,7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3,4,5,6,7]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4],ymm2[5,6,7]
; AVX2-FAST-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vpshufd {{.*#+}} xmm2 = mem[2,2,2,2]
@@ -4682,7 +4672,7 @@ define void @load_i32_stride7_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm7 = ymm14[0],ymm15[1],ymm14[2,3,4],ymm15[5],ymm14[6,7]
; AVX2-FAST-NEXT: vmovdqu %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vpermd %ymm7, %ymm1, %ymm15
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm15[0,1,2],ymm3[3,4],ymm15[5,6,7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm15[0,1,2],ymm3[3,4,5,6,7]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4],ymm2[5,6,7]
; AVX2-FAST-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vpshufd {{.*#+}} xmm2 = mem[2,2,2,2]
@@ -4702,7 +4692,7 @@ define void @load_i32_stride7_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm13[0],ymm8[1],ymm13[2,3,4],ymm8[5],ymm13[6,7]
; AVX2-FAST-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vpermd %ymm3, %ymm1, %ymm1
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4],ymm1[5,6,7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7]
; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovdqa 80(%rdi), %xmm0
@@ -4821,8 +4811,8 @@ define void @load_i32_stride7_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm1 = xmm2[0,1,2],xmm0[3]
; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm0 = <4,3,u,u>
; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-FAST-NEXT: vpblendd $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm3 = mem[0,1,2,3],ymm3[4,5],mem[6,7]
+; AVX2-FAST-NEXT: vpblendd $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm3 = mem[0,1,2,3],ymm3[4,5,6,7]
; AVX2-FAST-NEXT: vpermd %ymm3, %ymm0, %ymm3
; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm1 = xmm3[0,1],xmm1[2,3]
; AVX2-FAST-NEXT: vpbroadcastq {{.*#+}} ymm11 = [30064771072,30064771072,30064771072,30064771072]
@@ -4834,8 +4824,8 @@ define void @load_i32_stride7_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5,6,7]
; AVX2-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-FAST-NEXT: vpblendd $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm1 = ymm1[0,1,2,3],mem[4,5],ymm1[6,7]
+; AVX2-FAST-NEXT: vpblendd $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm1 = ymm1[0,1,2,3],mem[4,5,6,7]
; AVX2-FAST-NEXT: vpermd %ymm1, %ymm0, %ymm1
; AVX2-FAST-NEXT: vpbroadcastd 548(%rdi), %xmm4
; AVX2-FAST-NEXT: vmovdqa 512(%rdi), %xmm3
@@ -4848,8 +4838,8 @@ define void @load_i32_stride7_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm4[4,5,6,7]
; AVX2-FAST-NEXT: vmovdqu %ymm1, (%rsp) # 32-byte Spill
; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-FAST-NEXT: vpblendd $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm1 = ymm1[0,1,2,3],mem[4,5],ymm1[6,7]
+; AVX2-FAST-NEXT: vpblendd $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm1 = ymm1[0,1,2,3],mem[4,5,6,7]
; AVX2-FAST-NEXT: vpermd %ymm1, %ymm0, %ymm1
; AVX2-FAST-NEXT: vpbroadcastd 324(%rdi), %xmm6
; AVX2-FAST-NEXT: vmovdqa 288(%rdi), %xmm5
@@ -4863,8 +4853,8 @@ define void @load_i32_stride7_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm6[4,5,6,7]
; AVX2-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-FAST-NEXT: vpblendd $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm6 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm6 = mem[0,1,2,3],ymm1[4,5],mem[6,7]
+; AVX2-FAST-NEXT: vpblendd $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm6 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm6 = mem[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FAST-NEXT: vpermd %ymm6, %ymm0, %ymm0
; AVX2-FAST-NEXT: vpbroadcastd 772(%rdi), %xmm7
; AVX2-FAST-NEXT: vmovdqa 736(%rdi), %xmm6
@@ -5055,7 +5045,7 @@ define void @load_i32_stride7_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-PERLANE-NEXT: vmovdqa 32(%rdi), %ymm6
; AVX2-FAST-PERLANE-NEXT: vmovdqa 96(%rdi), %ymm10
; AVX2-FAST-PERLANE-NEXT: vpbroadcastq 80(%rdi), %ymm0
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm0[0,1,2,3],ymm10[4,5],ymm0[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm0[0,1,2,3],ymm10[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm0 = <0,7,6,u>
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm6[6],ymm3[7]
; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm6, %ymm12
@@ -5063,7 +5053,7 @@ define void @load_i32_stride7_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm3, %ymm13
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vpermd %ymm2, %ymm0, %ymm2
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4],ymm2[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqa 128(%rdi), %xmm2
; AVX2-FAST-PERLANE-NEXT: vmovdqa 160(%rdi), %xmm3
; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -5079,8 +5069,8 @@ define void @load_i32_stride7_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vpermd %ymm1, %ymm0, %ymm1
; AVX2-FAST-PERLANE-NEXT: vpbroadcastq 528(%rdi), %ymm2
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm11[4,5],ymm2[6,7]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4],ymm1[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm11[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqa 576(%rdi), %xmm2
; AVX2-FAST-PERLANE-NEXT: vmovdqa 608(%rdi), %xmm3
; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -5096,10 +5086,10 @@ define void @load_i32_stride7_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vpermd %ymm1, %ymm0, %ymm1
; AVX2-FAST-PERLANE-NEXT: vpbroadcastq 304(%rdi), %ymm2
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm9[4,5],ymm2[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm9[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm9, %ymm15
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4],ymm1[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqa 352(%rdi), %xmm2
; AVX2-FAST-PERLANE-NEXT: vmovdqa 384(%rdi), %xmm3
; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -5119,8 +5109,8 @@ define void @load_i32_stride7_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-PERLANE-NEXT: vpermd %ymm1, %ymm0, %ymm0
; AVX2-FAST-PERLANE-NEXT: vmovdqa 768(%rdi), %ymm9
; AVX2-FAST-PERLANE-NEXT: vpbroadcastq 752(%rdi), %ymm1
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm9[4,5],ymm1[6,7]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm9[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqa 800(%rdi), %xmm1
; AVX2-FAST-PERLANE-NEXT: vmovdqa 832(%rdi), %xmm2
; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -5149,7 +5139,7 @@ define void @load_i32_stride7_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} ymm2 = ymm2[1,0,3,3,5,4,7,7]
; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm2 = ymm2[0,3,2,3]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4],ymm2[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} xmm0 = mem[2,2,2,2]
@@ -5170,7 +5160,7 @@ define void @load_i32_stride7_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} ymm2 = ymm2[1,0,3,3,5,4,7,7]
; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm2 = ymm2[0,3,2,3]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4],ymm2[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} xmm0 = mem[2,2,2,2]
@@ -5194,7 +5184,7 @@ define void @load_i32_stride7_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} ymm2 = ymm2[1,0,3,3,5,4,7,7]
; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm2 = ymm2[0,3,2,3]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4],ymm2[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} xmm0 = mem[2,2,2,2]
@@ -5216,7 +5206,7 @@ define void @load_i32_stride7_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} ymm3 = ymm3[1,0,3,3,5,4,7,7]
; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm3 = ymm3[0,3,2,3]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3,4],ymm3[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3,4],ymm1[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovdqa 80(%rdi), %xmm1
@@ -5333,8 +5323,8 @@ define void @load_i32_stride7_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm1 = xmm0[0,1,2],xmm1[3]
; AVX2-FAST-PERLANE-NEXT: vmovaps {{.*#+}} xmm3 = <4,3,u,u>
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm2 = mem[0,1,2,3],ymm2[4,5],mem[6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm2 = mem[0,1,2,3],ymm2[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpermps %ymm2, %ymm3, %ymm2
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3]
; AVX2-FAST-PERLANE-NEXT: vbroadcastsd {{.*#+}} ymm4 = [30064771072,30064771072,30064771072,30064771072]
@@ -5347,8 +5337,8 @@ define void @load_i32_stride7_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm1 = mem[0,1,2,3],ymm1[4,5],mem[6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm1 = mem[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpermps %ymm1, %ymm3, %ymm2
; AVX2-FAST-PERLANE-NEXT: vbroadcastss 548(%rdi), %xmm7
; AVX2-FAST-PERLANE-NEXT: vmovaps 512(%rdi), %xmm1
@@ -5362,8 +5352,8 @@ define void @load_i32_stride7_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm7[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm2 = mem[0,1,2,3],ymm2[4,5],mem[6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm2 = mem[0,1,2,3],ymm2[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpermps %ymm2, %ymm3, %ymm7
; AVX2-FAST-PERLANE-NEXT: vbroadcastss 324(%rdi), %xmm8
; AVX2-FAST-PERLANE-NEXT: vmovaps 288(%rdi), %xmm2
@@ -5376,8 +5366,8 @@ define void @load_i32_stride7_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3],ymm8[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm7 = mem[0,1,2,3],ymm7[4,5],mem[6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm7 = mem[0,1,2,3],ymm7[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpermps %ymm7, %ymm3, %ymm3
; AVX2-FAST-PERLANE-NEXT: vbroadcastss 772(%rdi), %xmm8
; AVX2-FAST-PERLANE-NEXT: vmovaps 736(%rdi), %xmm7
@@ -7419,13 +7409,12 @@ define void @load_i32_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
;
; AVX1-ONLY-LABEL: load_i32_stride7_vf64:
; AVX1-ONLY: # %bb.0:
-; AVX1-ONLY-NEXT: subq $3224, %rsp # imm = 0xC98
+; AVX1-ONLY-NEXT: subq $3192, %rsp # imm = 0xC78
; AVX1-ONLY-NEXT: vmovaps 704(%rdi), %ymm2
; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 672(%rdi), %ymm3
; AVX1-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 768(%rdi), %ymm5
-; AVX1-ONLY-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 256(%rdi), %ymm4
; AVX1-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 224(%rdi), %ymm1
@@ -7434,39 +7423,42 @@ define void @load_i32_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: vmovaps 304(%rdi), %xmm0
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm6[0],ymm0[0],ymm6[2],ymm0[2]
+; AVX1-ONLY-NEXT: vmovaps %ymm6, %ymm10
; AVX1-ONLY-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm4[6],ymm1[7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm1, %xmm1
-; AVX1-ONLY-NEXT: vmovaps 224(%rdi), %xmm9
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm9[0,1],xmm1[2,3]
+; AVX1-ONLY-NEXT: vmovaps 224(%rdi), %xmm12
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm12[0,1],xmm1[2,3]
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm1 = xmm1[0,3,2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4],ymm1[5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps 384(%rdi), %xmm4
; AVX1-ONLY-NEXT: vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vmovaps 352(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm4[1]
-; AVX1-ONLY-NEXT: vmovaps 416(%rdi), %xmm7
-; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm1 = zero,xmm1[1,2],xmm7[1]
+; AVX1-ONLY-NEXT: vmovaps 416(%rdi), %xmm4
+; AVX1-ONLY-NEXT: vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm1 = zero,xmm1[1,2],xmm4[1]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 752(%rdi), %xmm0
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm5[0],ymm0[0],ymm5[2],ymm0[2]
+; AVX1-ONLY-NEXT: vmovaps %ymm5, %ymm7
+; AVX1-ONLY-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm3[0,1,2,3,4,5],ymm2[6],ymm3[7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm1, %xmm1
; AVX1-ONLY-NEXT: vmovaps 672(%rdi), %xmm2
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3]
-; AVX1-ONLY-NEXT: vmovaps %xmm2, %xmm5
-; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps %xmm2, %xmm4
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm1 = xmm1[0,3,2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4],ymm1[5,6,7]
-; AVX1-ONLY-NEXT: vmovaps 832(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 800(%rdi), %xmm2
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps 832(%rdi), %xmm2
; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm2[1],xmm1[1]
+; AVX1-ONLY-NEXT: vmovaps 800(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm2[1]
; AVX1-ONLY-NEXT: vmovaps 864(%rdi), %xmm2
; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm1 = zero,xmm1[1,2],xmm2[1]
@@ -7483,12 +7475,12 @@ define void @load_i32_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm1[0,1],xmm0[2,3]
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[0,3,2,3]
-; AVX1-ONLY-NEXT: vmovaps 1216(%rdi), %ymm4
+; AVX1-ONLY-NEXT: vmovaps 1216(%rdi), %ymm5
; AVX1-ONLY-NEXT: vmovaps 1200(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm4[0],ymm1[0],ymm4[2],ymm1[2]
-; AVX1-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm5[0],ymm1[0],ymm5[2],ymm1[2]
+; AVX1-ONLY-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps 1280(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vmovaps 1248(%rdi), %xmm2
@@ -7510,12 +7502,12 @@ define void @load_i32_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm1[0,1],xmm0[2,3]
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[0,3,2,3]
-; AVX1-ONLY-NEXT: vmovaps 1664(%rdi), %ymm2
-; AVX1-ONLY-NEXT: vmovups %ymm2, (%rsp) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 1664(%rdi), %ymm8
; AVX1-ONLY-NEXT: vmovaps 1648(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm8[0],ymm1[0],ymm8[2],ymm1[2]
+; AVX1-ONLY-NEXT: vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps 1728(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vmovaps 1696(%rdi), %xmm2
@@ -7533,25 +7525,24 @@ define void @load_i32_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6],ymm0[7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm0, %xmm0
-; AVX1-ONLY-NEXT: vmovaps (%rdi), %xmm13
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm13[0,1],xmm0[2,3]
-; AVX1-ONLY-NEXT: vmovaps %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps (%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm1[0,1],xmm0[2,3]
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[0,3,2,3]
-; AVX1-ONLY-NEXT: vmovaps 96(%rdi), %ymm2
+; AVX1-ONLY-NEXT: vmovaps 96(%rdi), %ymm3
; AVX1-ONLY-NEXT: vmovaps 80(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
-; AVX1-ONLY-NEXT: vmovaps %ymm2, %ymm10
-; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
-; AVX1-ONLY-NEXT: vmovaps 160(%rdi), %xmm2
-; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 128(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm3[0],ymm1[0],ymm3[2],ymm1[2]
+; AVX1-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps 160(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm2[1]
-; AVX1-ONLY-NEXT: vmovaps 192(%rdi), %xmm3
-; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm1 = zero,xmm1[1,2],xmm3[1]
-; AVX1-ONLY-NEXT: vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 128(%rdi), %xmm2
+; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm2[1],xmm1[1]
+; AVX1-ONLY-NEXT: vmovaps 192(%rdi), %xmm2
+; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm1 = zero,xmm1[1,2],xmm2[1]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -7562,24 +7553,24 @@ define void @load_i32_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6],ymm0[7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm0, %xmm0
; AVX1-ONLY-NEXT: vmovaps 448(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm1[0,1],xmm0[2,3]
-; AVX1-ONLY-NEXT: vmovaps %xmm1, %xmm15
; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm1[0,1],xmm0[2,3]
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[0,3,2,3]
; AVX1-ONLY-NEXT: vmovaps 544(%rdi), %ymm2
-; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 528(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
+; AVX1-ONLY-NEXT: vmovaps %ymm2, %ymm14
+; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps 608(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vmovaps 576(%rdi), %xmm2
; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm2[1],xmm1[1]
-; AVX1-ONLY-NEXT: vmovaps 640(%rdi), %xmm12
-; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm1 = zero,xmm1[1,2],xmm12[1]
-; AVX1-ONLY-NEXT: vmovaps %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 640(%rdi), %xmm2
+; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm1 = zero,xmm1[1,2],xmm2[1]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -7589,25 +7580,25 @@ define void @load_i32_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6],ymm0[7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm0, %xmm0
-; AVX1-ONLY-NEXT: vmovaps 896(%rdi), %xmm11
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm11[0,1],xmm0[2,3]
-; AVX1-ONLY-NEXT: vmovaps %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 896(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm1[0,1],xmm0[2,3]
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[0,3,2,3]
-; AVX1-ONLY-NEXT: vmovaps 992(%rdi), %ymm2
-; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 992(%rdi), %ymm6
; AVX1-ONLY-NEXT: vmovaps 976(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm6[0],ymm1[0],ymm6[2],ymm1[2]
+; AVX1-ONLY-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps 1056(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vmovaps 1024(%rdi), %xmm2
; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm2[1],xmm1[1]
-; AVX1-ONLY-NEXT: vmovaps 1088(%rdi), %xmm8
-; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm1 = zero,xmm1[1,2],xmm8[1]
-; AVX1-ONLY-NEXT: vmovaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX1-ONLY-NEXT: vmovaps 1088(%rdi), %xmm13
+; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm1 = zero,xmm1[1,2],xmm13[1]
+; AVX1-ONLY-NEXT: vmovaps %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 1376(%rdi), %ymm1
@@ -7625,7 +7616,7 @@ define void @load_i32_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: vmovaps 1424(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps 1504(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vmovaps 1472(%rdi), %xmm2
@@ -7637,43 +7628,45 @@ define void @load_i32_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 288(%rdi), %ymm14
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm6[1,1],ymm14[2,2],ymm6[5,5],ymm14[6,6]
-; AVX1-ONLY-NEXT: vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 288(%rdi), %ymm0
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm10[1,1],ymm0[2,2],ymm10[5,5],ymm0[6,6]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,2,3]
-; AVX1-ONLY-NEXT: vmovaps 256(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0],xmm9[1],xmm1[2,3]
-; AVX1-ONLY-NEXT: vmovaps %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 256(%rdi), %xmm15
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm15[0],xmm12[1],xmm15[2,3]
+; AVX1-ONLY-NEXT: vmovaps %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm1[1,0],mem[3,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4],ymm1[5,6,7]
-; AVX1-ONLY-NEXT: vmovaps 352(%rdi), %ymm2
-; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps 384(%rdi), %ymm1
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm1[2,3,0,1]
+; AVX1-ONLY-NEXT: vmovaps 352(%rdi), %ymm2
+; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm2[2,3],ymm1[0,1]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm2[0,0],ymm1[3,3],ymm2[4,4],ymm1[7,7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm1, %xmm1
-; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm1 = zero,xmm1[1,2],xmm7[2]
+; AVX1-ONLY-NEXT: vinsertps $49, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm1 = zero,xmm1[1,2],mem[0]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 736(%rdi), %ymm0
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm1[1,1],ymm0[2,2],ymm1[5,5],ymm0[6,6]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm7[1,1],ymm0[2,2],ymm7[5,5],ymm0[6,6]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,2,3]
-; AVX1-ONLY-NEXT: vmovaps 704(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0],xmm5[1],xmm1[2,3]
+; AVX1-ONLY-NEXT: vmovaps 704(%rdi), %xmm11
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm11[0],xmm4[1],xmm11[2,3]
+; AVX1-ONLY-NEXT: vmovaps %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps %xmm4, %xmm10
+; AVX1-ONLY-NEXT: vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm1[1,0],mem[3,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4],ymm1[5,6,7]
-; AVX1-ONLY-NEXT: vmovaps 800(%rdi), %ymm5
-; AVX1-ONLY-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps 832(%rdi), %ymm1
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm1[2,3,0,1]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm5[0,0],ymm1[3,3],ymm5[4,4],ymm1[7,7]
+; AVX1-ONLY-NEXT: vmovaps 800(%rdi), %ymm2
+; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm2[2,3],ymm1[0,1]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm2[0,0],ymm1[3,3],ymm2[4,4],ymm1[7,7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm1, %xmm1
; AVX1-ONLY-NEXT: vinsertps $49, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm1 = zero,xmm1[1,2],mem[0]
@@ -7682,42 +7675,41 @@ define void @load_i32_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 1184(%rdi), %ymm0
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm4[1,1],ymm0[2,2],ymm4[5,5],ymm0[6,6]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm5[1,1],ymm0[2,2],ymm5[5,5],ymm0[6,6]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,2,3]
; AVX1-ONLY-NEXT: vmovaps 1152(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm1 = xmm1[0],mem[1],xmm1[2,3]
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0],xmm7[1],xmm1[2,3]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm1[1,0],mem[3,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4],ymm1[5,6,7]
-; AVX1-ONLY-NEXT: vmovaps 1248(%rdi), %ymm4
-; AVX1-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps 1280(%rdi), %ymm1
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm1[2,3,0,1]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm4[0,0],ymm1[3,3],ymm4[4,4],ymm1[7,7]
+; AVX1-ONLY-NEXT: vmovaps 1248(%rdi), %ymm2
+; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm2[2,3],ymm1[0,1]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm2[0,0],ymm1[3,3],ymm2[4,4],ymm1[7,7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm1, %xmm1
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm1 = zero,xmm1[1,2],xmm4[2]
+; AVX1-ONLY-NEXT: vinsertps $49, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm1 = zero,xmm1[1,2],mem[0]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 1632(%rdi), %ymm0
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups (%rsp), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm1[1,1],ymm0[2,2],ymm1[5,5],ymm0[6,6]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm8[1,1],ymm0[2,2],ymm8[5,5],ymm0[6,6]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,2,3]
; AVX1-ONLY-NEXT: vmovaps 1600(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm1 = xmm1[0],mem[1],xmm1[2,3]
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0],xmm9[1],xmm1[2,3]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm1[1,0],mem[3,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4],ymm1[5,6,7]
-; AVX1-ONLY-NEXT: vmovaps 1696(%rdi), %ymm2
-; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps 1728(%rdi), %ymm1
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm1[2,3,0,1]
+; AVX1-ONLY-NEXT: vmovaps 1696(%rdi), %ymm2
+; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm2[2,3],ymm1[0,1]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm2[0,0],ymm1[3,3],ymm2[4,4],ymm1[7,7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm1, %xmm1
; AVX1-ONLY-NEXT: vinsertps $49, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
@@ -7727,599 +7719,603 @@ define void @load_i32_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 64(%rdi), %ymm0
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm10[1,1],ymm0[2,2],ymm10[5,5],ymm0[6,6]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm3[1,1],ymm0[2,2],ymm3[5,5],ymm0[6,6]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,2,3]
; AVX1-ONLY-NEXT: vmovaps 32(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0],xmm13[1],xmm1[2,3]
+; AVX1-ONLY-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm1 = xmm1[0],mem[1],xmm1[2,3]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm1[1,0],mem[3,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4],ymm1[5,6,7]
-; AVX1-ONLY-NEXT: vmovaps 128(%rdi), %ymm2
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm0[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps 160(%rdi), %ymm2
; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 160(%rdi), %ymm1
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm1[2,3,0,1]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm2[0,0],ymm1[3,3],ymm2[4,4],ymm1[7,7]
-; AVX1-ONLY-NEXT: vextractf128 $1, %ymm1, %xmm1
-; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm1 = zero,xmm1[1,2],xmm3[2]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 512(%rdi), %ymm0
+; AVX1-ONLY-NEXT: vmovaps 128(%rdi), %ymm0
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm10[1,1],ymm0[2,2],ymm10[5,5],ymm0[6,6]
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,2,3]
-; AVX1-ONLY-NEXT: vmovaps 480(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0],xmm15[1],xmm1[2,3]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm1[1,0],mem[3,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4],ymm1[5,6,7]
-; AVX1-ONLY-NEXT: vmovaps 576(%rdi), %ymm3
-; AVX1-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 608(%rdi), %ymm1
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm2 = ymm1[2,3,0,1]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm3[0,0],ymm2[3,3],ymm3[4,4],ymm2[7,7]
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm2 = ymm0[2,3],ymm2[0,1]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm0[0,0],ymm2[3,3],ymm0[4,4],ymm2[7,7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm2, %xmm2
-; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm2 = zero,xmm2[1,2],xmm12[2]
+; AVX1-ONLY-NEXT: vinsertps $49, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm2 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm2 = zero,xmm2[1,2],mem[0]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm2
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 960(%rdi), %ymm0
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 512(%rdi), %ymm0
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm1[1,1],ymm0[2,2],ymm1[5,5],ymm0[6,6]
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,2,3]
-; AVX1-ONLY-NEXT: vmovaps 928(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm1[0],xmm11[1],xmm1[2,3]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm14[1,1],ymm0[2,2],ymm14[5,5],ymm0[6,6]
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm1[2,3,2,3]
+; AVX1-ONLY-NEXT: vmovaps 480(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm2 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm2 = xmm0[0],mem[1],xmm0[2,3]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm2 = xmm2[1,0],mem[3,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2],ymm0[3,4],ymm2[5,6,7]
-; AVX1-ONLY-NEXT: vmovaps 1024(%rdi), %ymm2
-; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 1056(%rdi), %ymm1
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm1[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps 608(%rdi), %ymm0
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 576(%rdi), %ymm1
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm6 = ymm1[2,3,0,1]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm6 = ymm2[0,0],ymm6[3,3],ymm2[4,4],ymm6[7,7]
-; AVX1-ONLY-NEXT: vextractf128 $1, %ymm6, %xmm6
-; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm6 = zero,xmm6[1,2],xmm8[2]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm6, %ymm0, %ymm6
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm6[5,6,7]
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm3 = ymm1[2,3],ymm0[0,1]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm1[0,0],ymm3[3,3],ymm1[4,4],ymm3[7,7]
+; AVX1-ONLY-NEXT: vextractf128 $1, %ymm3, %xmm3
+; AVX1-ONLY-NEXT: vinsertps $49, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm3 = zero,xmm3[1,2],mem[0]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm3
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 960(%rdi), %ymm0
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 1408(%rdi), %ymm8
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm6 = ymm12[1,1],ymm8[2,2],ymm12[5,5],ymm8[6,6]
-; AVX1-ONLY-NEXT: vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm6 = ymm6[2,3,2,3]
-; AVX1-ONLY-NEXT: vmovaps 1376(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm11 = xmm0[0],xmm5[1],xmm0[2,3]
-; AVX1-ONLY-NEXT: vmovaps %xmm0, %xmm1
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm6[1,1],ymm0[2,2],ymm6[5,5],ymm0[6,6]
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm2 = ymm2[2,3,2,3]
+; AVX1-ONLY-NEXT: vmovaps 928(%rdi), %xmm0
; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm11 = xmm11[1,0],mem[3,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm11 = ymm11[0,1,2],ymm6[3,4],ymm11[5,6,7]
-; AVX1-ONLY-NEXT: vmovaps 1472(%rdi), %ymm2
-; AVX1-ONLY-NEXT: vmovaps 1504(%rdi), %ymm3
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm15 = ymm3[2,3,0,1]
-; AVX1-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm2[0,0],ymm15[3,3],ymm2[4,4],ymm15[7,7]
-; AVX1-ONLY-NEXT: vmovaps %ymm2, %ymm13
+; AVX1-ONLY-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm3 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm3 = xmm0[0],mem[1],xmm0[2,3]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm3 = xmm3[1,0],mem[3,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps 1056(%rdi), %ymm0
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 1024(%rdi), %ymm1
+; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm3 = ymm1[2,3],ymm0[0,1]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm1[0,0],ymm3[3,3],ymm1[4,4],ymm3[7,7]
+; AVX1-ONLY-NEXT: vextractf128 $1, %ymm3, %xmm3
+; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm3 = zero,xmm3[1,2],xmm13[2]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm3
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vextractf128 $1, %ymm15, %xmm15
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm15 = zero,xmm15[1,2],xmm6[2]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm15, %ymm0, %ymm15
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm11 = ymm11[0,1,2,3,4],ymm15[5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm11 = xmm9[2,3,2,3]
-; AVX1-ONLY-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11, %xmm11 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm11 = xmm11[0],mem[1],xmm11[2,3]
+; AVX1-ONLY-NEXT: vmovaps 1408(%rdi), %ymm6
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm8[1,1],ymm6[2,2],ymm8[5,5],ymm6[6,6]
+; AVX1-ONLY-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm3 = ymm3[2,3,2,3]
+; AVX1-ONLY-NEXT: vmovaps 1376(%rdi), %xmm4
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm13 = xmm4[0],xmm1[1],xmm4[2,3]
+; AVX1-ONLY-NEXT: vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm13 = xmm13[1,0],mem[3,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0,1,2],ymm3[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps 1504(%rdi), %ymm5
+; AVX1-ONLY-NEXT: vmovaps 1472(%rdi), %ymm0
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm14 = ymm0[2,3],ymm5[0,1]
+; AVX1-ONLY-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm14 = ymm0[0,0],ymm14[3,3],ymm0[4,4],ymm14[7,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vextractf128 $1, %ymm14, %xmm14
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm14 = zero,xmm14[1,2],xmm3[2]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm14, %ymm0, %ymm14
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3,4],ymm14[5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm13 = xmm12[2,3,2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm13 = xmm13[0],xmm15[1],xmm13[2,3]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps $199, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm14 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm14 = ymm14[3,1],mem[0,3],ymm14[7,5],mem[4,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm14 = ymm2[2,1],ymm14[2,0],ymm2[6,5],ymm14[6,4]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0,1],ymm14[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
+; AVX1-ONLY-NEXT: vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm14 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm14 = ymm12[0],mem[0],ymm12[2],mem[2]
+; AVX1-ONLY-NEXT: vextractf128 $1, %ymm14, %xmm14
+; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14, %xmm14 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm14 = xmm14[0,1,2],mem[3]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm14, %ymm0, %ymm14
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3,4],ymm14[5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm13 = xmm10[2,3,2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm13 = xmm13[0],xmm11[1],xmm13[2,3]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps $199, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm14 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm14 = ymm14[3,1],mem[0,3],ymm14[7,5],mem[4,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm14 = ymm2[2,1],ymm14[2,0],ymm2[6,5],ymm14[6,4]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0,1],ymm14[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm15[3,1],ymm14[0,3],ymm15[7,5],ymm14[4,7]
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm14 = ymm15[0],ymm12[0],ymm15[2],ymm12[2]
+; AVX1-ONLY-NEXT: vextractf128 $1, %ymm14, %xmm14
+; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14, %xmm14 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm14 = xmm14[0,1,2],mem[3]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm14, %ymm0, %ymm14
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3,4],ymm14[5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm13 = xmm7[2,3,2,3]
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm13 = xmm13[0],xmm11[1],xmm13[2,3]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps $199, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm14 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm14 = ymm14[3,1],mem[0,3],ymm14[7,5],mem[4,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm14 = ymm2[2,1],ymm14[2,0],ymm2[6,5],ymm14[6,4]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0,1],ymm14[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm14 = ymm10[0],ymm7[0],ymm10[2],ymm7[2]
+; AVX1-ONLY-NEXT: vextractf128 $1, %ymm14, %xmm14
+; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14, %xmm14 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm14 = xmm14[0,1,2],mem[3]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm14, %ymm0, %ymm14
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3,4],ymm14[5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm13 = xmm9[2,3,2,3]
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm13 = xmm13[0],xmm9[1],xmm13[2,3]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps $199, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm14 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm14 = ymm14[3,1],mem[0,3],ymm14[7,5],mem[4,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm14 = ymm2[2,1],ymm14[2,0],ymm2[6,5],ymm14[6,4]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0,1],ymm14[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT: vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm14 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm14 = ymm2[0],mem[0],ymm2[2],mem[2]
+; AVX1-ONLY-NEXT: vextractf128 $1, %ymm14, %xmm14
+; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14, %xmm14 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm14 = xmm14[0,1,2],mem[3]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm14, %ymm0, %ymm14
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3,4],ymm14[5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm13 = xmm1[2,3,2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm13 = xmm13[0],xmm4[1],xmm13[2,3]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm14 = ymm14[3,1],ymm6[0,3],ymm14[7,5],ymm6[4,7]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm14 = ymm8[2,1],ymm14[2,0],ymm8[6,5],ymm14[6,4]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0,1],ymm14[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm14 = ymm0[0],ymm5[0],ymm0[2],ymm5[2]
+; AVX1-ONLY-NEXT: vextractf128 $1, %ymm14, %xmm14
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm8 = xmm14[0,1,2],xmm3[3]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm8, %ymm0, %ymm8
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm13[0,1,2,3,4],ymm8[5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm8 = mem[2,3,2,3]
+; AVX1-ONLY-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8, %xmm8 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm8 = xmm8[0],mem[1],xmm8[2,3]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps $199, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm13 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm13 = ymm13[3,1],mem[0,3],ymm13[7,5],mem[4,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm0[2,1],ymm15[2,0],ymm0[6,5],ymm15[6,4]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm11 = ymm11[0,1],ymm15[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm13 = ymm0[2,1],ymm13[2,0],ymm0[6,5],ymm13[6,4]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1],ymm13[2,3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm15 = ymm0[0],mem[0],ymm0[2],mem[2]
-; AVX1-ONLY-NEXT: vextractf128 $1, %ymm15, %xmm15
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm15 = xmm15[0,1,2],xmm7[3]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm15, %ymm0, %ymm15
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm11 = ymm11[0,1,2,3,4],ymm15[5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm11 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm11 = xmm11[0],xmm7[1],xmm11[2,3]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps $199, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm15 = ymm15[3,1],mem[0,3],ymm15[7,5],mem[4,7]
+; AVX1-ONLY-NEXT: vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm13 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm13 = ymm0[0],mem[0],ymm0[2],mem[2]
+; AVX1-ONLY-NEXT: vextractf128 $1, %ymm13, %xmm13
+; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13, %xmm13 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm13 = xmm13[0,1,2],mem[3]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm13, %ymm0, %ymm13
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4],ymm13[5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm8 = mem[2,3,2,3]
+; AVX1-ONLY-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8, %xmm8 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm8 = xmm8[0],mem[1],xmm8[2,3]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps $199, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm13 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm13 = ymm13[3,1],mem[0,3],ymm13[7,5],mem[4,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm0[2,1],ymm15[2,0],ymm0[6,5],ymm15[6,4]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm11 = ymm11[0,1],ymm15[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm13 = ymm0[2,1],ymm13[2,0],ymm0[6,5],ymm13[6,4]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1],ymm13[2,3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm15 = ymm0[0],mem[0],ymm0[2],mem[2]
-; AVX1-ONLY-NEXT: vextractf128 $1, %ymm15, %xmm15
-; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm15 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm15 = xmm15[0,1,2],mem[3]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm15, %ymm0, %ymm15
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm11 = ymm11[0,1,2,3,4],ymm15[5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm11 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11, %xmm11 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm11 = xmm11[0],mem[1],xmm11[2,3]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps $199, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm15 = ymm15[3,1],mem[0,3],ymm15[7,5],mem[4,7]
+; AVX1-ONLY-NEXT: vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm13 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm13 = ymm0[0],mem[0],ymm0[2],mem[2]
+; AVX1-ONLY-NEXT: vextractf128 $1, %ymm13, %xmm13
+; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13, %xmm13 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm13 = xmm13[0,1,2],mem[3]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm13, %ymm0, %ymm13
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4],ymm13[5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm8 = mem[2,3,2,3]
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm8 = xmm8[0],xmm1[1],xmm8[2,3]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm13 = ymm13[3,1],ymm4[0,3],ymm13[7,5],ymm4[4,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm13 = ymm3[2,1],ymm13[2,0],ymm3[6,5],ymm13[6,4]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1],ymm13[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm0[2,1],ymm15[2,0],ymm0[6,5],ymm15[6,4]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm11 = ymm11[0,1],ymm15[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm13 = ymm0[0],ymm6[0],ymm0[2],ymm6[2]
+; AVX1-ONLY-NEXT: vextractf128 $1, %ymm13, %xmm13
+; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13, %xmm13 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm13 = xmm13[0,1,2],mem[3]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm13, %ymm0, %ymm13
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4],ymm13[5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm15 = ymm2[0],ymm0[0],ymm2[2],ymm0[2]
-; AVX1-ONLY-NEXT: vextractf128 $1, %ymm15, %xmm15
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm15 = xmm15[0,1,2],xmm4[3]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm15, %ymm0, %ymm15
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm11 = ymm11[0,1,2,3,4],ymm15[5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm11 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm11 = xmm11[0],xmm2[1],xmm11[2,3]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps $199, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm15 = ymm15[3,1],mem[0,3],ymm15[7,5],mem[4,7]
-; AVX1-ONLY-NEXT: vmovups (%rsp), %ymm4 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm4[2,1],ymm15[2,0],ymm4[6,5],ymm15[6,4]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm11 = ymm11[0,1],ymm15[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX1-ONLY-NEXT: vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm15 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm15 = ymm9[0],mem[0],ymm9[2],mem[2]
-; AVX1-ONLY-NEXT: vextractf128 $1, %ymm15, %xmm15
-; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm15 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm15 = xmm15[0,1,2],mem[3]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm15, %ymm0, %ymm15
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm11 = ymm11[0,1,2,3,4],ymm15[5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm11 = xmm5[2,3,2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm11 = xmm11[0],xmm1[1],xmm11[2,3]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm15[3,1],ymm8[0,3],ymm15[7,5],ymm8[4,7]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm12[2,1],ymm15[2,0],ymm12[6,5],ymm15[6,4]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm11 = ymm11[0,1],ymm15[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm15 = ymm13[0],ymm3[0],ymm13[2],ymm3[2]
-; AVX1-ONLY-NEXT: vextractf128 $1, %ymm15, %xmm15
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm4 = xmm15[0,1,2],xmm6[3]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm4, %ymm0, %ymm4
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm11[0,1,2,3,4],ymm4[5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm4 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4, %xmm4 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm4 = xmm4[0],mem[1],xmm4[2,3]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps $199, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm11 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm11 = ymm11[3,1],mem[0,3],ymm11[7,5],mem[4,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm11 = ymm1[2,1],ymm11[2,0],ymm1[6,5],ymm11[6,4]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm11[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm11 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm11 = ymm1[0],mem[0],ymm1[2],mem[2]
-; AVX1-ONLY-NEXT: vextractf128 $1, %ymm11, %xmm11
-; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11, %xmm11 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm11 = xmm11[0,1,2],mem[3]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm11, %ymm0, %ymm11
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4],ymm11[5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm4 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4, %xmm4 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm4 = xmm4[0],mem[1],xmm4[2,3]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm8 = ymm8[1,0],ymm0[0,0],ymm8[5,4],ymm0[4,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm8 = ymm0[3,1],ymm8[0,2],ymm0[7,5],ymm8[4,6]
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT: vblendps $7, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm13 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm13 = mem[0,1,2],xmm0[3]
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm13 = xmm13[3,2,2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm13[0,1],ymm8[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps 416(%rdi), %ymm0
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vshufps $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm14 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm14 = ymm0[0,1],mem[1,3],ymm0[4,5],mem[5,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm14 = ymm0[0,2],ymm14[2,0],ymm0[4,6],ymm14[6,4]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4],ymm14[5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm11 = ymm11[3,1],ymm13[0,3],ymm11[7,5],ymm13[4,7]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm11 = ymm10[2,1],ymm11[2,0],ymm10[6,5],ymm11[6,4]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm11[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm11 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm11 = ymm1[0],mem[0],ymm1[2],mem[2]
-; AVX1-ONLY-NEXT: vextractf128 $1, %ymm11, %xmm11
-; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11, %xmm11 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm11 = xmm11[0,1,2],mem[3]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm11, %ymm0, %ymm11
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4],ymm11[5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm4 = xmm10[2,3,2,3]
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm4 = xmm4[0],xmm5[1],xmm4[2,3]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm11 = ymm11[3,1],ymm6[0,3],ymm11[7,5],ymm6[4,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm11 = ymm12[2,1],ymm11[2,0],ymm12[6,5],ymm11[6,4]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm11[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm11 = ymm14[0],ymm1[0],ymm14[2],ymm1[2]
-; AVX1-ONLY-NEXT: vextractf128 $1, %ymm11, %xmm11
-; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11, %xmm11 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm11 = xmm11[0,1,2],mem[3]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm11, %ymm0, %ymm11
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4],ymm11[5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm4 = ymm3[1,0],ymm11[0,0],ymm3[5,4],ymm11[4,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm4 = ymm11[3,1],ymm4[0,2],ymm11[7,5],ymm4[4,6]
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm11 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm11 = xmm3[0,1,2],mem[3]
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm11 = xmm11[3,2,2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm11[0,1],ymm4[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovaps 416(%rdi), %ymm8
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm8 = ymm0[1,0],ymm13[0,0],ymm0[5,4],ymm13[4,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm8 = ymm13[3,1],ymm8[0,2],ymm13[7,5],ymm8[4,6]
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT: vblendps $7, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm14 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm14 = mem[0,1,2],xmm0[3]
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm14 = xmm14[3,2,2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm14[0,1],ymm8[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps 864(%rdi), %ymm0
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm14 = ymm0[0,1],ymm12[1,3],ymm0[4,5],ymm12[5,7]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm14 = ymm15[0,2],ymm14[2,0],ymm15[4,6],ymm14[6,4]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4],ymm14[5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufps $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm15 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm15 = ymm8[0,1],mem[1,3],ymm8[4,5],mem[5,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm3[0,2],ymm15[2,0],ymm3[4,6],ymm15[6,4]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4],ymm15[5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm4 = ymm4[1,0],ymm11[0,0],ymm4[5,4],ymm11[4,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm4 = ymm11[3,1],ymm4[0,2],ymm11[7,5],ymm4[4,6]
-; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7, %xmm15 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm15 = xmm7[0,1,2],mem[3]
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm15 = xmm15[3,2,2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm15[0,1],ymm4[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovaps 864(%rdi), %ymm3
-; AVX1-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufps $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm15 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm15 = ymm3[0,1],mem[1,3],ymm3[4,5],mem[5,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm3[0,2],ymm15[2,0],ymm3[4,6],ymm15[6,4]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4],ymm15[5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm4 = ymm4[1,0],ymm3[0,0],ymm4[5,4],ymm3[4,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm4 = ymm3[3,1],ymm4[0,2],ymm3[7,5],ymm4[4,6]
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm15 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm15 = xmm3[0,1,2],mem[3]
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm15 = xmm15[3,2,2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm15[0,1],ymm4[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovaps 1312(%rdi), %ymm3
-; AVX1-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm3[0,1],ymm0[1,3],ymm3[4,5],ymm0[5,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm0[0,2],ymm15[2,0],ymm0[4,6],ymm15[6,4]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm4[0,1,2,3,4],ymm15[5,6,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm8 = ymm8[1,0],ymm0[0,0],ymm8[5,4],ymm0[4,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm8 = ymm0[3,1],ymm8[0,2],ymm0[7,5],ymm8[4,6]
+; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11, %xmm14 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm14 = xmm11[0,1,2],mem[3]
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm14 = xmm14[3,2,2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm14[0,1],ymm8[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps 1312(%rdi), %ymm0
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups (%rsp), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm4 = ymm3[1,0],ymm0[0,0],ymm3[5,4],ymm0[4,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm4 = ymm0[3,1],ymm4[0,2],ymm0[7,5],ymm4[4,6]
-; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm8 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm8 = xmm2[0,1,2],mem[3]
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm8 = xmm8[3,2,2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm8[0,1],ymm4[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm14 = ymm0[0,1],ymm7[1,3],ymm0[4,5],ymm7[5,7]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm14 = ymm10[0,2],ymm14[2,0],ymm10[4,6],ymm14[6,4]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm8[0,1,2,3,4],ymm14[5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm8 = ymm5[1,0],ymm0[0,0],ymm5[5,4],ymm0[4,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm8 = ymm0[3,1],ymm8[0,2],ymm0[7,5],ymm8[4,6]
+; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9, %xmm14 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm14 = xmm9[0,1,2],mem[3]
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm14 = xmm14[3,2,2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm14[0,1],ymm8[2,3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps 1760(%rdi), %ymm0
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufps $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm8 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm8 = ymm0[0,1],mem[1,3],ymm0[4,5],mem[5,7]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm8 = ymm9[0,2],ymm8[2,0],ymm9[4,6],ymm8[6,4]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm4[0,1,2,3,4],ymm8[5,6,7]
+; AVX1-ONLY-NEXT: vshufps $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm14 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm14 = ymm0[0,1],mem[1,3],ymm0[4,5],mem[5,7]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm14 = ymm2[0,2],ymm14[2,0],ymm2[4,6],ymm14[6,4]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm8[0,1,2,3,4],ymm14[5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm2[1,0],ymm0[0,0],ymm2[5,4],ymm0[4,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm0[3,1],ymm2[0,2],ymm0[7,5],ymm2[4,6]
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT: vblendps $7, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm5 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm5 = mem[0,1,2],xmm0[3]
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm5 = xmm5[3,2,2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm5[0,1],ymm2[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps 1536(%rdi), %ymm0
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm5 = ymm0[0,1],ymm11[1,3],ymm0[4,5],ymm11[5,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[1,0],ymm2[0,0],ymm0[5,4],ymm2[4,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm2[3,1],ymm0[0,2],ymm2[7,5],ymm0[4,6]
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX1-ONLY-NEXT: vblendps $7, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm3 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm3 = mem[0,1,2],xmm2[3]
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm3 = xmm3[3,2,2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm3[0,1],ymm0[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovaps 1536(%rdi), %ymm2
-; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm5 = ymm0[0,2],ymm5[2,0],ymm0[4,6],ymm5[6,4]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3,4],ymm5[5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm4[1,0],ymm3[0,0],ymm4[5,4],ymm3[4,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm3[3,1],ymm2[0,2],ymm3[7,5],ymm2[4,6]
+; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm5 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm5 = xmm1[0,1,2],mem[3]
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm5 = xmm5[3,2,2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm5[0,1],ymm2[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps 1088(%rdi), %ymm0
+; AVX1-ONLY-NEXT: vmovaps %ymm6, %ymm14
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm5 = ymm0[0,1],ymm6[1,3],ymm0[4,5],ymm6[5,7]
+; AVX1-ONLY-NEXT: vmovaps %ymm0, %ymm10
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm2[0,1],ymm15[1,3],ymm2[4,5],ymm15[5,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm2[0,2],ymm3[2,0],ymm2[4,6],ymm3[6,4]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm3[5,6,7]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm5 = ymm15[0,2],ymm5[2,0],ymm15[4,6],ymm5[6,4]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3,4],ymm5[5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm6[1,0],ymm12[0,0],ymm6[5,4],ymm12[4,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm12[3,1],ymm0[0,2],ymm12[7,5],ymm0[4,6]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm3 = xmm5[0,1,2],xmm10[3]
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm3 = xmm3[3,2,2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm3[0,1],ymm0[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovaps 1088(%rdi), %ymm11
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm11[0,1],ymm1[1,3],ymm11[4,5],ymm1[5,7]
-; AVX1-ONLY-NEXT: vmovups %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps %ymm1, %ymm10
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm14[0,2],ymm3[2,0],ymm14[4,6],ymm3[6,4]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm3[5,6,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm1[1,0],ymm0[0,0],ymm1[5,4],ymm0[4,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm0[3,1],ymm2[0,2],ymm0[7,5],ymm2[4,6]
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9, %xmm5 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm5 = xmm9[0,1,2],mem[3]
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm5 = xmm5[3,2,2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm5[0,1],ymm2[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps 640(%rdi), %ymm0
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm5 = ymm0[0,1],ymm3[1,3],ymm0[4,5],ymm3[5,7]
+; AVX1-ONLY-NEXT: vmovaps %ymm0, %ymm12
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm5 = ymm6[0,2],ymm5[2,0],ymm6[4,6],ymm5[6,4]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3,4],ymm5[5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm13[1,0],ymm1[0,0],ymm13[5,4],ymm1[4,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm1[3,1],ymm0[0,2],ymm1[7,5],ymm0[4,6]
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8, %xmm3 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm3 = xmm8[0,1,2],mem[3]
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm3 = xmm3[3,2,2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm3[0,1],ymm0[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovaps 640(%rdi), %ymm7
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm1[1,0],ymm0[0,0],ymm1[5,4],ymm0[4,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm0[3,1],ymm2[0,2],ymm0[7,5],ymm2[4,6]
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4, %xmm5 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm5 = xmm4[0,1,2],mem[3]
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm5 = xmm5[3,2,2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1],ymm2[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps 192(%rdi), %ymm13
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm8 = ymm13[0,1],ymm1[1,3],ymm13[4,5],ymm1[5,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm7[0,1],ymm2[1,3],ymm7[4,5],ymm2[5,7]
-; AVX1-ONLY-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm6[0,2],ymm3[2,0],ymm6[4,6],ymm3[6,4]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm3[5,6,7]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm8 = ymm2[0,2],ymm8[2,0],ymm2[4,6],ymm8[6,4]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm5[0,1,2,3,4],ymm8[5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm5 = ymm2[2,3,0,1]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm2[3,0],ymm5[0,0],ymm2[7,4],ymm5[4,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm5 = ymm13[1,0],ymm1[2,0],ymm13[5,4],ymm1[6,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm5[2,0],ymm0[6,4],ymm5[6,4]
+; AVX1-ONLY-NEXT: vmovaps 64(%rdi), %xmm2
+; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 96(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm5 = xmm1[0,1,0,1]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm5 = xmm2[0,1,2],xmm5[3]
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm7 = xmm4[2,3,2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm7 = mem[0],xmm7[1],mem[2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm5 = xmm7[0,1],xmm5[2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm5[0,1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[1,0],ymm1[0,0],ymm0[5,4],ymm1[4,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm1[3,1],ymm0[0,2],ymm1[7,5],ymm0[4,6]
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm3 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm3 = xmm5[0,1,2],mem[3]
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm3 = xmm3[3,2,2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1],ymm0[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovaps 192(%rdi), %ymm9
-; AVX1-ONLY-NEXT: vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm4 = ymm9[0,1],ymm12[1,3],ymm9[4,5],ymm12[5,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm4 = ymm0[0,2],ymm4[2,0],ymm0[4,6],ymm4[6,4]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm3[0,1,2,3,4],ymm4[5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm3 = ymm0[2,3,0,1]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm0[3,0],ymm3[0,0],ymm0[7,4],ymm3[4,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm4 = ymm9[1,0],ymm12[2,0],ymm9[5,4],ymm12[6,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm3[2,0],ymm4[2,0],ymm3[6,4],ymm4[6,4]
-; AVX1-ONLY-NEXT: vmovaps 64(%rdi), %xmm12
-; AVX1-ONLY-NEXT: vmovaps 96(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm4 = xmm0[0,1,0,1]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm4 = xmm12[0,1,2],xmm4[3]
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm5 = xmm5[2,3,2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm5 = mem[0],xmm5[1],mem[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm4 = xmm5[0,1],xmm4[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm4[0,1,2,3],ymm3[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm3 = ymm0[2,3,0,1]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm0[3,0],ymm3[0,0],ymm0[7,4],ymm3[4,4]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps $33, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm4 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm4 = ymm9[1,0],mem[2,0],ymm9[5,4],mem[6,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm3[2,0],ymm4[2,0],ymm3[6,4],ymm4[6,4]
-; AVX1-ONLY-NEXT: vmovaps 320(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm4 = xmm0[0,1,0,1]
-; AVX1-ONLY-NEXT: vmovaps 288(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm4 = xmm0[0,1,2],xmm4[3]
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm1[2,3,0,1]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm1[3,0],ymm0[0,0],ymm1[7,4],ymm0[4,4]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm4 = ymm7[1,0],ymm8[2,0],ymm7[5,4],ymm8[6,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm4[2,0],ymm0[6,4],ymm4[6,4]
+; AVX1-ONLY-NEXT: vmovaps 320(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm4 = xmm1[0,1,0,1]
+; AVX1-ONLY-NEXT: vmovaps 288(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm4 = xmm1[0,1,2],xmm4[3]
; AVX1-ONLY-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm5 = mem[2,3,2,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm5 = mem[0],xmm5[1],mem[2,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm4 = xmm5[0,1],xmm4[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm4[0,1,2,3],ymm3[4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm4[0,1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm3 = ymm6[2,3,0,1]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm6[3,0],ymm3[0,0],ymm6[7,4],ymm3[4,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm7[1,0],ymm2[2,0],ymm7[5,4],ymm2[6,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm1[2,0],ymm3[2,0],ymm1[6,4],ymm3[6,4]
-; AVX1-ONLY-NEXT: vmovaps 544(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm3 = xmm0[0,1,0,1]
-; AVX1-ONLY-NEXT: vmovaps 512(%rdi), %xmm5
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm3 = xmm5[0,1,2],xmm3[3]
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm4 = xmm8[2,3,2,3]
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm6[2,3,0,1]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm6[3,0],ymm0[0,0],ymm6[7,4],ymm0[4,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm12[1,0],ymm3[2,0],ymm12[5,4],ymm3[6,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm1[2,0],ymm0[6,4],ymm1[6,4]
+; AVX1-ONLY-NEXT: vmovaps 544(%rdi), %xmm2
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm1 = xmm2[0,1,0,1]
+; AVX1-ONLY-NEXT: vmovaps %xmm2, %xmm6
+; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 512(%rdi), %xmm2
+; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm2[0,1,2],xmm1[3]
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm4 = xmm9[2,3,2,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm4 = mem[0],xmm4[1],mem[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm3 = xmm4[0,1],xmm3[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm4[0,1],xmm1[2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm0[2,3,0,1]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm0[3,0],ymm1[0,0],ymm0[7,4],ymm1[4,4]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps $33, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm3 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm3 = ymm7[1,0],mem[2,0],ymm7[5,4],mem[6,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm1[2,0],ymm3[2,0],ymm1[6,4],ymm3[6,4]
-; AVX1-ONLY-NEXT: vmovaps 768(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm3 = xmm0[0,1,0,1]
-; AVX1-ONLY-NEXT: vmovaps 736(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm3 = xmm0[0,1,2],xmm3[3]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm1[2,3,0,1]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm1[3,0],ymm0[0,0],ymm1[7,4],ymm0[4,4]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps $33, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm1 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm1 = ymm5[1,0],mem[2,0],ymm5[5,4],mem[6,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm1[2,0],ymm0[6,4],ymm1[6,4]
+; AVX1-ONLY-NEXT: vmovaps 768(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm1 = xmm1[0,1,0,1]
+; AVX1-ONLY-NEXT: vmovaps 736(%rdi), %xmm2
+; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm2[0,1,2],xmm1[3]
; AVX1-ONLY-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm4 = mem[2,3,2,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm4 = mem[0],xmm4[1],mem[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm3 = xmm4[0,1],xmm3[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3],ymm1[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm14[2,3,0,1]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm14[3,0],ymm1[0,0],ymm14[7,4],ymm1[4,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm11[1,0],ymm10[2,0],ymm11[5,4],ymm10[6,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm1[2,0],ymm2[2,0],ymm1[6,4],ymm2[6,4]
-; AVX1-ONLY-NEXT: vmovaps 992(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm2 = xmm0[0,1,0,1]
-; AVX1-ONLY-NEXT: vmovaps 960(%rdi), %xmm4
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm4[0,1,2],xmm2[3]
-; AVX1-ONLY-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm3 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm3 = mem[0],xmm3[1],mem[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm3[0,1],xmm2[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm1[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm0[2,3,0,1]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm0[3,0],ymm1[0,0],ymm0[7,4],ymm1[4,4]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps $33, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm2 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm2 = ymm3[1,0],mem[2,0],ymm3[5,4],mem[6,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm1[2,0],ymm2[2,0],ymm1[6,4],ymm2[6,4]
-; AVX1-ONLY-NEXT: vmovaps 1216(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm2 = xmm0[0,1,0,1]
-; AVX1-ONLY-NEXT: vmovaps 1184(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm0[0,1,2],xmm2[3]
-; AVX1-ONLY-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm13 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm13 = mem[0],xmm13[1],mem[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm13[0,1],xmm2[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm4[0,1],xmm1[2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm0[2,3,0,1]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm0[3,0],ymm1[0,0],ymm0[7,4],ymm1[4,4]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm0[1,0],ymm15[2,0],ymm0[5,4],ymm15[6,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm1[2,0],ymm2[2,0],ymm1[6,4],ymm2[6,4]
-; AVX1-ONLY-NEXT: vmovaps 1440(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm13 = xmm0[0,1,0,1]
-; AVX1-ONLY-NEXT: vmovaps 1408(%rdi), %xmm2
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm13 = xmm2[0,1,2],xmm13[3]
-; AVX1-ONLY-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm6 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm6 = mem[0],xmm6[1],mem[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm6 = xmm6[0,1],xmm13[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm6[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm15[2,3,0,1]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm15[3,0],ymm0[0,0],ymm15[7,4],ymm0[4,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm10[1,0],ymm14[2,0],ymm10[5,4],ymm14[6,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm1[2,0],ymm0[6,4],ymm1[6,4]
+; AVX1-ONLY-NEXT: vmovaps 992(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm1 = xmm1[0,1,0,1]
+; AVX1-ONLY-NEXT: vmovaps 960(%rdi), %xmm2
+; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm2[0,1,2],xmm1[3]
+; AVX1-ONLY-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm4 = mem[2,3,2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm4 = mem[0],xmm4[1],mem[2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm4[0,1],xmm1[2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm0[2,3,0,1]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm0[3,0],ymm1[0,0],ymm0[7,4],ymm1[4,4]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps $33, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm6 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm6 = ymm0[1,0],mem[2,0],ymm0[5,4],mem[6,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm6 = ymm1[2,0],ymm6[2,0],ymm1[6,4],ymm6[6,4]
-; AVX1-ONLY-NEXT: vmovaps 1664(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm1[2,3,0,1]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm1[3,0],ymm0[0,0],ymm1[7,4],ymm0[4,4]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps $33, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm1 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm1 = ymm4[1,0],mem[2,0],ymm4[5,4],mem[6,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm1[2,0],ymm0[6,4],ymm1[6,4]
+; AVX1-ONLY-NEXT: vmovaps 1216(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm13 = xmm1[0,1,0,1]
-; AVX1-ONLY-NEXT: vmovaps 1632(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm13 = xmm1[0,1,2],xmm13[3]
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm1 = xmm1[0,1,0,1]
+; AVX1-ONLY-NEXT: vmovaps 1184(%rdi), %xmm2
+; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm2[0,1,2],xmm1[3]
; AVX1-ONLY-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm10 = mem[2,3,2,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm10 = mem[0],xmm10[1],mem[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm10 = xmm10[0,1],xmm13[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm10[0,1,2,3],ymm6[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps $246, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm6 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm6 = ymm15[2,1],mem[3,3],ymm15[6,5],mem[7,7]
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm10 = xmm14[0],xmm8[1],xmm14[2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm10[0,1],xmm1[2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm1[2,3,0,1]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm1[3,0],ymm0[0,0],ymm1[7,4],ymm0[4,4]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm1[1,0],ymm11[2,0],ymm1[5,4],ymm11[6,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm1[2,0],ymm0[6,4],ymm1[6,4]
+; AVX1-ONLY-NEXT: vmovaps 1440(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm10 = xmm1[0,1,0,1]
+; AVX1-ONLY-NEXT: vmovaps 1408(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm10 = xmm1[0,1,2],xmm10[3]
+; AVX1-ONLY-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm3 = mem[2,3,2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm3 = mem[0],xmm3[1],mem[2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm3 = xmm3[0,1],xmm10[2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm2[2,3,0,1]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm2[3,0],ymm0[0,0],ymm2[7,4],ymm0[4,4]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps $33, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm3 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm3 = ymm2[1,0],mem[2,0],ymm2[5,4],mem[6,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm0[2,0],ymm3[2,0],ymm0[6,4],ymm3[6,4]
+; AVX1-ONLY-NEXT: vmovaps 1664(%rdi), %xmm9
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm10 = xmm9[0,1,0,1]
+; AVX1-ONLY-NEXT: vmovaps 1632(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm10 = xmm0[0,1,2],xmm10[3]
+; AVX1-ONLY-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm12 = mem[2,3,2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm12 = mem[0],xmm12[1],mem[2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm10 = xmm12[0,1],xmm10[2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm10[0,1,2,3],ymm3[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vshufps $246, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm3 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm3 = ymm13[2,1],mem[3,3],ymm13[6,5],mem[7,7]
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; AVX1-ONLY-NEXT: vblendps $13, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12, %xmm10 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm10 = mem[0],xmm12[1],mem[2,3]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm10, %ymm0, %ymm10
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm6 = ymm10[1,0],ymm6[2,0],ymm10[5,4],ymm6[6,4]
-; AVX1-ONLY-NEXT: vblendps $7, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12, %xmm10 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm10 = mem[0,1,2],xmm12[3]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm13 = ymm12[0,0],ymm11[1,0],ymm12[4,4],ymm11[5,4]
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} ymm13 = ymm13[2,0,2,3,6,4,6,7]
-; AVX1-ONLY-NEXT: vextractf128 $1, %ymm13, %xmm13
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm10 = xmm13[0,1],xmm10[3,2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm10[0,1,2,3],ymm6[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufps $246, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm6 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm6 = ymm9[2,1],mem[3,3],ymm9[6,5],mem[7,7]
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
-; AVX1-ONLY-NEXT: vblendps $13, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13, %xmm10 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm10 = mem[0],xmm13[1],mem[2,3]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm10[1,0],ymm3[2,0],ymm10[5,4],ymm3[6,4]
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; AVX1-ONLY-NEXT: vblendps $7, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10, %xmm10 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm10 = mem[0,1,2],xmm10[3]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm11 = ymm14[0,0],ymm15[1,0],ymm14[4,4],ymm15[5,4]
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} ymm11 = ymm11[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT: vextractf128 $1, %ymm11, %xmm11
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm10 = xmm11[0,1],xmm10[3,2]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm10[0,1,2,3],ymm3[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm7[2,1],ymm8[3,3],ymm7[6,5],ymm8[7,7]
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
+; AVX1-ONLY-NEXT: vblendps $13, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11, %xmm10 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm10 = mem[0],xmm11[1],mem[2,3]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm10, %ymm0, %ymm10
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm6 = ymm10[1,0],ymm6[2,0],ymm10[5,4],ymm6[6,4]
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; AVX1-ONLY-NEXT: vblendps $7, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9, %xmm9 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm9 = mem[0,1,2],xmm9[3]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm10[1,0],ymm3[2,0],ymm10[5,4],ymm3[6,4]
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; AVX1-ONLY-NEXT: vblendps $7, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7, %xmm8 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm8 = mem[0,1,2],xmm7[3]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
; AVX1-ONLY-NEXT: vshufps $16, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm10 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm10 = ymm10[0,0],mem[1,0],ymm10[4,4],mem[5,4]
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} ymm10 = ymm10[2,0,2,3,6,4,6,7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm10, %xmm10
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm9 = xmm10[0,1],xmm9[3,2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm9[0,1,2,3],ymm6[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm8 = xmm10[0,1],xmm8[3,2]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm8[0,1,2,3],ymm3[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps $246, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm6 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm6 = ymm10[2,1],mem[3,3],ymm10[6,5],mem[7,7]
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; AVX1-ONLY-NEXT: vblendps $13, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9, %xmm9 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm9 = mem[0],xmm9[1],mem[2,3]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm9, %ymm0, %ymm9
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm6 = ymm9[1,0],ymm6[2,0],ymm9[5,4],ymm6[6,4]
-; AVX1-ONLY-NEXT: vblendps $7, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm5 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm5 = mem[0,1,2],xmm5[3]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps $16, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm9 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm9 = ymm9[0,0],mem[1,0],ymm9[4,4],mem[5,4]
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} ymm9 = ymm9[2,0,2,3,6,4,6,7]
-; AVX1-ONLY-NEXT: vextractf128 $1, %ymm9, %xmm9
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm5 = xmm9[0,1],xmm5[3,2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm6[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufps $246, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm5 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm5 = ymm7[2,1],mem[3,3],ymm7[6,5],mem[7,7]
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; AVX1-ONLY-NEXT: vblendps $13, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7, %xmm6 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm6 = mem[0],xmm7[1],mem[2,3]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm6, %ymm0, %ymm6
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm5 = ymm6[1,0],ymm5[2,0],ymm6[5,4],ymm5[6,4]
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX1-ONLY-NEXT: vblendps $7, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6, %xmm6 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm6 = mem[0,1,2],xmm6[3]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps $16, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm9 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm9 = ymm9[0,0],mem[1,0],ymm9[4,4],mem[5,4]
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} ymm9 = ymm9[2,0,2,3,6,4,6,7]
-; AVX1-ONLY-NEXT: vextractf128 $1, %ymm9, %xmm9
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm6 = xmm9[0,1],xmm6[3,2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm6[0,1,2,3],ymm5[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps $246, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm5 = ymm5[2,1],mem[3,3],ymm5[6,5],mem[7,7]
+; AVX1-ONLY-NEXT: vshufps $246, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm3 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm3 = ymm10[2,1],mem[3,3],ymm10[6,5],mem[7,7]
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT: vblendps $13, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8, %xmm8 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm8 = mem[0],xmm8[1],mem[2,3]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm8, %ymm0, %ymm8
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm8[1,0],ymm3[2,0],ymm8[5,4],ymm3[6,4]
+; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6, %xmm7 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm7 = xmm6[0,1,2],mem[3]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps $16, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm8 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm8 = ymm8[0,0],mem[1,0],ymm8[4,4],mem[5,4]
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} ymm8 = ymm8[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT: vextractf128 $1, %ymm8, %xmm8
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm7 = xmm8[0,1],xmm7[3,2]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm7[0,1,2,3],ymm3[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vshufps $246, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm3 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm3 = ymm5[2,1],mem[3,3],ymm5[6,5],mem[7,7]
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT: vblendps $13, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8, %xmm7 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm7 = mem[0],xmm8[1],mem[2,3]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm7, %ymm0, %ymm7
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm7[1,0],ymm3[2,0],ymm7[5,4],ymm3[6,4]
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT: vblendps $7, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm6 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm6 = mem[0,1,2],xmm5[3]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps $16, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm7 = ymm7[0,0],mem[1,0],ymm7[4,4],mem[5,4]
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} ymm7 = ymm7[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT: vextractf128 $1, %ymm7, %xmm7
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm6 = xmm7[0,1],xmm6[3,2]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm6[0,1,2,3],ymm3[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps $246, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm3 = ymm3[2,1],mem[3,3],ymm3[6,5],mem[7,7]
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
; AVX1-ONLY-NEXT: vblendps $13, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6, %xmm6 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm6 = mem[0],xmm6[1],mem[2,3]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm6, %ymm0, %ymm6
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm5 = ymm6[1,0],ymm5[2,0],ymm6[5,4],ymm5[6,4]
-; AVX1-ONLY-NEXT: vblendps $7, (%rsp), %xmm4, %xmm4 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm4 = mem[0,1,2],xmm4[3]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm6[1,0],ymm3[2,0],ymm6[5,4],ymm3[6,4]
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT: vblendps $7, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm5 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm5 = mem[0,1,2],xmm5[3]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
; AVX1-ONLY-NEXT: vshufps $16, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm6 = ymm6[0,0],mem[1,0],ymm6[4,4],mem[5,4]
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} ymm6 = ymm6[2,0,2,3,6,4,6,7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm6, %xmm6
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm4 = xmm6[0,1],xmm4[3,2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm5[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufps $246, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm4 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm4 = ymm3[2,1],mem[3,3],ymm3[6,5],mem[7,7]
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; AVX1-ONLY-NEXT: vblendps $13, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9, %xmm5 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm5 = mem[0],xmm9[1],mem[2,3]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm5 = xmm6[0,1],xmm5[3,2]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm5[0,1,2,3],ymm3[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vshufps $246, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm3 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm3 = ymm4[2,1],mem[3,3],ymm4[6,5],mem[7,7]
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; AVX1-ONLY-NEXT: vblendps $13, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7, %xmm5 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm5 = mem[0],xmm7[1],mem[2,3]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm5, %ymm0, %ymm5
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm4 = ymm5[1,0],ymm4[2,0],ymm5[5,4],ymm4[6,4]
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; AVX1-ONLY-NEXT: vblendps $7, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm3 = mem[0,1,2],xmm3[3]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm5[1,0],ymm3[2,0],ymm5[5,4],ymm3[6,4]
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX1-ONLY-NEXT: vblendps $7, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4, %xmm4 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm4 = mem[0,1,2],xmm4[3]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
; AVX1-ONLY-NEXT: vshufps $16, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm5 = ymm5[0,0],mem[1,0],ymm5[4,4],mem[5,4]
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} ymm5 = ymm5[2,0,2,3,6,4,6,7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm5, %xmm5
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm3 = xmm5[0,1],xmm3[3,2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm4[4,5,6,7]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm4 = xmm5[0,1],xmm4[3,2]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm4[0,1,2,3],ymm3[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX1-ONLY-NEXT: vshufps $246, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
@@ -8329,44 +8325,44 @@ define void @load_i32_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: # xmm4 = mem[0],xmm4[1],mem[2,3]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm4, %ymm0, %ymm4
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm4[1,0],ymm3[2,0],ymm4[5,4],ymm3[6,4]
-; AVX1-ONLY-NEXT: vblendps $7, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm2 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm2 = mem[0,1,2],xmm2[3]
+; AVX1-ONLY-NEXT: vblendps $7, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm1 = mem[0,1,2],xmm1[3]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX1-ONLY-NEXT: vshufps $16, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm4 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm4 = ymm4[0,0],mem[1,0],ymm4[4,4],mem[5,4]
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} ymm4 = ymm4[2,0,2,3,6,4,6,7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm4, %xmm4
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm2 = xmm4[0,1],xmm2[3,2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufps $246, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm2 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm2 = ymm0[2,1],mem[3,3],ymm0[6,5],mem[7,7]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm4[0,1],xmm1[3,2]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vshufps $246, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm1 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm1 = ymm2[2,1],mem[3,3],ymm2[6,5],mem[7,7]
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
; AVX1-ONLY-NEXT: vblendps $13, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm3 = mem[0],xmm3[1],mem[2,3]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm3
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm3[1,0],ymm2[2,0],ymm3[5,4],ymm2[6,4]
-; AVX1-ONLY-NEXT: vblendps $7, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm1 = mem[0,1,2],xmm1[3]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm3[1,0],ymm1[2,0],ymm3[5,4],ymm1[6,4]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm9[0,1,2],xmm0[3]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX1-ONLY-NEXT: vshufps $16, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm3 = ymm3[0,0],mem[1,0],ymm3[4,4],mem[5,4]
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} ymm3 = ymm3[2,0,2,3,6,4,6,7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm3, %xmm3
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm3[0,1],xmm1[3,2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm2[4,5,6,7]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm0 = xmm3[0,1],xmm0[3,2]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm15[2,3,0,1]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm15[3,0],ymm1[0,0],ymm15[7,4],ymm1[4,4]
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm1 = xmm8[2,3,2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0],xmm14[1],xmm1[2,3]
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm13[2,3,0,1]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm13[3,0],ymm0[0,0],ymm13[7,4],ymm0[4,4]
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm1 = xmm12[2,3,2,3]
+; AVX1-ONLY-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm1 = xmm1[0],mem[1],xmm1[2,3]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,0],ymm1[4,5],ymm0[6,4]
; AVX1-ONLY-NEXT: vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm1 = mem[0,1,0,1]
; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm1 = xmm1[0,1,2],mem[3]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm12[1,0],ymm11[2,0],ymm12[5,4],ymm11[6,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm14[1,0],ymm15[2,0],ymm14[5,4],ymm15[6,4]
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} ymm2 = ymm2[2,0,2,3,6,4,6,7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm2, %xmm2
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3]
@@ -8374,7 +8370,7 @@ define void @load_i32_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm2[2,3,0,1]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm2[3,0],ymm1[0,0],ymm2[7,4],ymm1[4,4]
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm2 = xmm13[2,3,2,3]
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm2 = xmm11[2,3,2,3]
; AVX1-ONLY-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm2 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm2 = xmm2[0],mem[1],xmm2[2,3]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm2
@@ -8412,7 +8408,7 @@ define void @load_i32_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm3 = ymm4[2,3,0,1]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm4[3,0],ymm3[0,0],ymm4[7,4],ymm3[4,4]
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm4 = xmm7[2,3,2,3]
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm4 = xmm8[2,3,2,3]
; AVX1-ONLY-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4, %xmm4 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm4 = xmm4[0],mem[1],xmm4[2,3]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm4, %ymm0, %ymm4
@@ -8439,7 +8435,7 @@ define void @load_i32_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm5[0,1],ymm3[2,0],ymm5[4,5],ymm3[6,4]
; AVX1-ONLY-NEXT: vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm5 = mem[0,1,0,1]
-; AVX1-ONLY-NEXT: vblendps $8, (%rsp), %xmm5, %xmm5 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm5 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm5 = xmm5[0,1,2],mem[3]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
; AVX1-ONLY-NEXT: vshufps $33, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
@@ -8447,15 +8443,15 @@ define void @load_i32_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} ymm6 = ymm6[2,0,2,3,6,4,6,7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm6, %xmm6
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm5 = xmm6[0,1],xmm5[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm5[0,1,2,3],ymm3[4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm3[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm5 = ymm6[2,3,0,1]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm5 = ymm6[3,0],ymm5[0,0],ymm6[7,4],ymm5[4,4]
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm6 = xmm9[2,3,2,3]
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm3 = ymm6[2,3,0,1]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm6[3,0],ymm3[0,0],ymm6[7,4],ymm3[4,4]
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm6 = xmm7[2,3,2,3]
; AVX1-ONLY-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6, %xmm6 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm6 = xmm6[0],mem[1],xmm6[2,3]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm6, %ymm0, %ymm6
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm5 = ymm6[0,1],ymm5[2,0],ymm6[4,5],ymm5[6,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm6[0,1],ymm3[2,0],ymm6[4,5],ymm3[6,4]
; AVX1-ONLY-NEXT: vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm6 = mem[0,1,0,1]
; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6, %xmm6 # 16-byte Folded Reload
@@ -8466,16 +8462,16 @@ define void @load_i32_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} ymm7 = ymm7[2,0,2,3,6,4,6,7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm7, %xmm7
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm6 = xmm7[0,1],xmm6[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm6[0,1,2,3],ymm5[4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm3[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm6 = ymm7[2,3,0,1]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm6 = ymm7[3,0],ymm6[0,0],ymm7[7,4],ymm6[4,4]
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm3 = ymm7[2,3,0,1]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm7[3,0],ymm3[0,0],ymm7[7,4],ymm3[4,4]
; AVX1-ONLY-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm7 = mem[2,3,2,3]
; AVX1-ONLY-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7, %xmm7 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm7 = xmm7[0],mem[1],xmm7[2,3]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm7, %ymm0, %ymm7
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm6 = ymm7[0,1],ymm6[2,0],ymm7[4,5],ymm6[6,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm7[0,1],ymm3[2,0],ymm7[4,5],ymm3[6,4]
; AVX1-ONLY-NEXT: vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm7 = mem[0,1,0,1]
; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7, %xmm7 # 16-byte Folded Reload
@@ -8486,7 +8482,7 @@ define void @load_i32_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} ymm8 = ymm8[2,0,2,3,6,4,6,7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm8, %xmm8
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm7 = xmm8[0,1],xmm7[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm7[0,1,2,3],ymm6[4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm7[0,1,2,3],ymm3[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm7 = ymm8[2,3,0,1]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm7 = ymm8[3,0],ymm7[0,0],ymm8[7,4],ymm7[4,4]
@@ -8498,14 +8494,13 @@ define void @load_i32_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm7 = ymm8[0,1],ymm7[2,0],ymm8[4,5],ymm7[6,4]
; AVX1-ONLY-NEXT: vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm8 = mem[0,1,0,1]
-; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8, %xmm8 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm8 = xmm8[0,1,2],mem[3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm8 = xmm8[0,1,2],xmm9[3]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps $33, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm10 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm10 = ymm9[1,0],mem[2,0],ymm9[5,4],mem[6,4]
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} ymm10 = ymm10[2,0,2,3,6,4,6,7]
-; AVX1-ONLY-NEXT: vextractf128 $1, %ymm10, %xmm10
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm8 = xmm10[0,1],xmm8[2,3]
+; AVX1-ONLY-NEXT: vshufps $33, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm9 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm9 = ymm9[1,0],mem[2,0],ymm9[5,4],mem[6,4]
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} ymm9 = ymm9[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT: vextractf128 $1, %ymm9, %xmm9
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm8 = xmm9[0,1],xmm8[2,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm8[0,1,2,3],ymm7[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm8, 192(%rsi)
@@ -8606,14 +8601,14 @@ define void @load_i32_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: vmovaps %ymm8, (%rax)
; AVX1-ONLY-NEXT: movq {{[0-9]+}}(%rsp), %rax
; AVX1-ONLY-NEXT: vmovaps %ymm7, 224(%rax)
-; AVX1-ONLY-NEXT: vmovaps %ymm6, 192(%rax)
-; AVX1-ONLY-NEXT: vmovaps %ymm5, 160(%rax)
-; AVX1-ONLY-NEXT: vmovaps %ymm3, 128(%rax)
+; AVX1-ONLY-NEXT: vmovaps %ymm3, 192(%rax)
+; AVX1-ONLY-NEXT: vmovaps %ymm6, 160(%rax)
+; AVX1-ONLY-NEXT: vmovaps %ymm5, 128(%rax)
; AVX1-ONLY-NEXT: vmovaps %ymm4, 96(%rax)
; AVX1-ONLY-NEXT: vmovaps %ymm1, 64(%rax)
; AVX1-ONLY-NEXT: vmovaps %ymm2, 32(%rax)
; AVX1-ONLY-NEXT: vmovaps %ymm0, (%rax)
-; AVX1-ONLY-NEXT: addq $3224, %rsp # imm = 0xC98
+; AVX1-ONLY-NEXT: addq $3192, %rsp # imm = 0xC78
; AVX1-ONLY-NEXT: vzeroupper
; AVX1-ONLY-NEXT: retq
;
@@ -8641,8 +8636,8 @@ define void @load_i32_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-SLOW-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vpermd %ymm1, %ymm0, %ymm1
; AVX2-SLOW-NEXT: vpbroadcastq 304(%rdi), %ymm2
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm8[4,5],ymm2[6,7]
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4],ymm1[5,6,7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm8[4,5,6,7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
; AVX2-SLOW-NEXT: vmovdqa 352(%rdi), %xmm2
; AVX2-SLOW-NEXT: vmovdqa 384(%rdi), %xmm3
; AVX2-SLOW-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -8659,8 +8654,8 @@ define void @load_i32_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-SLOW-NEXT: vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vpermd %ymm1, %ymm0, %ymm1
; AVX2-SLOW-NEXT: vpbroadcastq 752(%rdi), %ymm2
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm13[4,5],ymm2[6,7]
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4],ymm1[5,6,7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm13[4,5,6,7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
; AVX2-SLOW-NEXT: vmovdqa 800(%rdi), %xmm2
; AVX2-SLOW-NEXT: vmovdqa 832(%rdi), %xmm3
; AVX2-SLOW-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -8674,8 +8669,8 @@ define void @load_i32_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-SLOW-NEXT: vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vpermd %ymm1, %ymm0, %ymm1
; AVX2-SLOW-NEXT: vpbroadcastq 1200(%rdi), %ymm2
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm12[4,5],ymm2[6,7]
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4],ymm1[5,6,7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm12[4,5,6,7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
; AVX2-SLOW-NEXT: vmovdqa 1248(%rdi), %xmm2
; AVX2-SLOW-NEXT: vmovdqa 1280(%rdi), %xmm3
; AVX2-SLOW-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -8694,8 +8689,8 @@ define void @load_i32_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-SLOW-NEXT: vmovdqa 1664(%rdi), %ymm3
; AVX2-SLOW-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vpbroadcastq 1648(%rdi), %ymm2
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5],ymm2[6,7]
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4],ymm1[5,6,7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
; AVX2-SLOW-NEXT: vmovdqa 1696(%rdi), %xmm2
; AVX2-SLOW-NEXT: vmovdqa 1728(%rdi), %xmm3
; AVX2-SLOW-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -8708,14 +8703,14 @@ define void @load_i32_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-SLOW-NEXT: vmovdqa 96(%rdi), %ymm2
; AVX2-SLOW-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vpbroadcastq 80(%rdi), %ymm1
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX2-SLOW-NEXT: vmovdqa (%rdi), %ymm3
; AVX2-SLOW-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovdqa 32(%rdi), %ymm2
; AVX2-SLOW-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm2[6],ymm3[7]
; AVX2-SLOW-NEXT: vpermd %ymm2, %ymm0, %ymm2
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4],ymm2[5,6,7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4,5,6,7]
; AVX2-SLOW-NEXT: vmovdqa 128(%rdi), %xmm2
; AVX2-SLOW-NEXT: vmovdqa 160(%rdi), %xmm3
; AVX2-SLOW-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -8734,9 +8729,9 @@ define void @load_i32_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-SLOW-NEXT: vpermd %ymm1, %ymm0, %ymm1
; AVX2-SLOW-NEXT: vmovdqa 544(%rdi), %ymm6
; AVX2-SLOW-NEXT: vpbroadcastq 528(%rdi), %ymm2
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm6[4,5],ymm2[6,7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm6[4,5,6,7]
; AVX2-SLOW-NEXT: vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4],ymm1[5,6,7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
; AVX2-SLOW-NEXT: vmovdqa 576(%rdi), %xmm2
; AVX2-SLOW-NEXT: vmovdqa 608(%rdi), %xmm3
; AVX2-SLOW-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -8754,9 +8749,9 @@ define void @load_i32_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-SLOW-NEXT: vpermd %ymm1, %ymm0, %ymm1
; AVX2-SLOW-NEXT: vmovdqa 992(%rdi), %ymm7
; AVX2-SLOW-NEXT: vpbroadcastq 976(%rdi), %ymm2
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm7[4,5],ymm2[6,7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm7[4,5,6,7]
; AVX2-SLOW-NEXT: vmovdqu %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4],ymm1[5,6,7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
; AVX2-SLOW-NEXT: vmovdqa 1024(%rdi), %xmm2
; AVX2-SLOW-NEXT: vmovdqa 1056(%rdi), %xmm3
; AVX2-SLOW-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -8775,8 +8770,8 @@ define void @load_i32_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-SLOW-NEXT: vmovdqa 1440(%rdi), %ymm2
; AVX2-SLOW-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vpbroadcastq 1424(%rdi), %ymm1
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
; AVX2-SLOW-NEXT: vmovdqa 1472(%rdi), %xmm1
; AVX2-SLOW-NEXT: vmovdqa 1504(%rdi), %xmm2
; AVX2-SLOW-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -8805,7 +8800,7 @@ define void @load_i32_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-SLOW-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm2 = ymm2[1,0,3,3,5,4,7,7]
; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm2 = ymm2[0,3,2,3]
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4],ymm2[5,6,7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4,5,6,7]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7]
; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm0 = mem[2,2,2,2]
@@ -8827,7 +8822,7 @@ define void @load_i32_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-SLOW-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm2 = ymm2[1,0,3,3,5,4,7,7]
; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm2 = ymm2[0,3,2,3]
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4],ymm2[5,6,7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4,5,6,7]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7]
; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm0 = mem[2,2,2,2]
@@ -8850,7 +8845,7 @@ define void @load_i32_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-SLOW-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm2 = ymm2[1,0,3,3,5,4,7,7]
; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm2 = ymm2[0,3,2,3]
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4],ymm2[5,6,7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4,5,6,7]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7]
; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm0 = mem[2,2,2,2]
@@ -8873,7 +8868,7 @@ define void @load_i32_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-SLOW-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm2 = ymm2[1,0,3,3,5,4,7,7]
; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm2 = ymm2[0,3,2,3]
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4],ymm2[5,6,7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4,5,6,7]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7]
; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm0 = mem[2,2,2,2]
@@ -8894,7 +8889,7 @@ define void @load_i32_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-SLOW-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm2 = ymm2[1,0,3,3,5,4,7,7]
; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm2 = ymm2[0,3,2,3]
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4],ymm2[5,6,7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4,5,6,7]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7]
; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm0 = mem[2,2,2,2]
@@ -8916,7 +8911,7 @@ define void @load_i32_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-SLOW-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm2 = ymm2[1,0,3,3,5,4,7,7]
; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm2 = ymm2[0,3,2,3]
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4],ymm2[5,6,7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4,5,6,7]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7]
; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm0 = mem[2,2,2,2]
@@ -8941,7 +8936,7 @@ define void @load_i32_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-SLOW-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm2 = ymm2[1,0,3,3,5,4,7,7]
; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm2 = ymm2[0,3,2,3]
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4],ymm2[5,6,7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4,5,6,7]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7]
; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm0 = mem[2,2,2,2]
@@ -8966,7 +8961,7 @@ define void @load_i32_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-SLOW-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm2 = ymm2[1,0,3,3,5,4,7,7]
; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm2 = ymm2[0,3,2,3]
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4],ymm2[5,6,7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4,5,6,7]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7]
; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovdqa 304(%rdi), %xmm0
@@ -9209,8 +9204,8 @@ define void @load_i32_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm1 = xmm0[0,1,2],xmm1[3]
; AVX2-SLOW-NEXT: vmovaps {{.*#+}} xmm5 = <4,3,u,u>
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm2 = mem[0,1,2,3],ymm2[4,5],mem[6,7]
+; AVX2-SLOW-NEXT: vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm2 = mem[0,1,2,3],ymm2[4,5,6,7]
; AVX2-SLOW-NEXT: vpermps %ymm2, %ymm5, %ymm2
; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3]
; AVX2-SLOW-NEXT: vbroadcastsd {{.*#+}} ymm9 = [30064771072,30064771072,30064771072,30064771072]
@@ -9221,8 +9216,8 @@ define void @load_i32_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm1 = mem[0,1,2,3],ymm1[4,5],mem[6,7]
+; AVX2-SLOW-NEXT: vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm1 = mem[0,1,2,3],ymm1[4,5,6,7]
; AVX2-SLOW-NEXT: vpermps %ymm1, %ymm5, %ymm2
; AVX2-SLOW-NEXT: vbroadcastss 324(%rdi), %xmm3
; AVX2-SLOW-NEXT: vmovaps 288(%rdi), %xmm1
@@ -9236,8 +9231,8 @@ define void @load_i32_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm2 = mem[0,1,2,3],ymm2[4,5],mem[6,7]
+; AVX2-SLOW-NEXT: vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm2 = mem[0,1,2,3],ymm2[4,5,6,7]
; AVX2-SLOW-NEXT: vpermps %ymm2, %ymm5, %ymm3
; AVX2-SLOW-NEXT: vbroadcastss 548(%rdi), %xmm4
; AVX2-SLOW-NEXT: vmovaps 512(%rdi), %xmm2
@@ -9250,8 +9245,8 @@ define void @load_i32_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm4[4,5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm3, (%rsp) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm3 = mem[0,1,2,3],ymm3[4,5],mem[6,7]
+; AVX2-SLOW-NEXT: vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm3 = mem[0,1,2,3],ymm3[4,5,6,7]
; AVX2-SLOW-NEXT: vpermps %ymm3, %ymm5, %ymm4
; AVX2-SLOW-NEXT: vbroadcastss 772(%rdi), %xmm6
; AVX2-SLOW-NEXT: vmovaps 736(%rdi), %xmm3
@@ -9265,8 +9260,8 @@ define void @load_i32_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm6[4,5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm4 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm4 = mem[0,1,2,3],ymm4[4,5],mem[6,7]
+; AVX2-SLOW-NEXT: vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm4 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm4 = mem[0,1,2,3],ymm4[4,5,6,7]
; AVX2-SLOW-NEXT: vpermps %ymm4, %ymm5, %ymm6
; AVX2-SLOW-NEXT: vbroadcastss 996(%rdi), %xmm7
; AVX2-SLOW-NEXT: vmovaps 960(%rdi), %xmm4
@@ -9279,8 +9274,8 @@ define void @load_i32_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm7[4,5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm6 = mem[0,1,2,3],ymm6[4,5],mem[6,7]
+; AVX2-SLOW-NEXT: vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm6 = mem[0,1,2,3],ymm6[4,5,6,7]
; AVX2-SLOW-NEXT: vpermps %ymm6, %ymm5, %ymm6
; AVX2-SLOW-NEXT: vbroadcastss 1220(%rdi), %xmm8
; AVX2-SLOW-NEXT: vmovaps 1184(%rdi), %xmm7
@@ -9293,8 +9288,8 @@ define void @load_i32_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm8[4,5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm6 = mem[0,1,2,3],ymm6[4,5],mem[6,7]
+; AVX2-SLOW-NEXT: vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm6 = mem[0,1,2,3],ymm6[4,5,6,7]
; AVX2-SLOW-NEXT: vpermps %ymm6, %ymm5, %ymm6
; AVX2-SLOW-NEXT: vbroadcastss 1444(%rdi), %xmm8
; AVX2-SLOW-NEXT: vmovaps 1408(%rdi), %xmm13
@@ -9309,8 +9304,8 @@ define void @load_i32_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm8[4,5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm6 = ymm6[0,1,2,3],mem[4,5],ymm6[6,7]
+; AVX2-SLOW-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm6 = ymm6[0,1,2,3],mem[4,5,6,7]
; AVX2-SLOW-NEXT: vpermps %ymm6, %ymm5, %ymm5
; AVX2-SLOW-NEXT: vbroadcastss 1668(%rdi), %xmm6
; AVX2-SLOW-NEXT: vmovaps 1632(%rdi), %xmm15
@@ -9694,8 +9689,8 @@ define void @load_i32_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vpermd %ymm1, %ymm0, %ymm1
; AVX2-FAST-NEXT: vpbroadcastq 304(%rdi), %ymm2
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm10[4,5],ymm2[6,7]
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4],ymm1[5,6,7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm10[4,5,6,7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
; AVX2-FAST-NEXT: vmovdqa 352(%rdi), %xmm2
; AVX2-FAST-NEXT: vmovdqa 384(%rdi), %xmm3
; AVX2-FAST-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -9711,8 +9706,8 @@ define void @load_i32_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-NEXT: vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vpermd %ymm1, %ymm0, %ymm1
; AVX2-FAST-NEXT: vpbroadcastq 752(%rdi), %ymm2
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm7[4,5],ymm2[6,7]
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4],ymm1[5,6,7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm7[4,5,6,7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
; AVX2-FAST-NEXT: vmovdqa 800(%rdi), %xmm2
; AVX2-FAST-NEXT: vmovdqa 832(%rdi), %xmm3
; AVX2-FAST-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -9726,8 +9721,8 @@ define void @load_i32_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-NEXT: vmovdqu %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vpermd %ymm1, %ymm0, %ymm1
; AVX2-FAST-NEXT: vpbroadcastq 1200(%rdi), %ymm2
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm8[4,5],ymm2[6,7]
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4],ymm1[5,6,7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm8[4,5,6,7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
; AVX2-FAST-NEXT: vmovdqa 1248(%rdi), %xmm2
; AVX2-FAST-NEXT: vmovdqa 1280(%rdi), %xmm3
; AVX2-FAST-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -9745,9 +9740,9 @@ define void @load_i32_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-NEXT: vpermd %ymm1, %ymm0, %ymm1
; AVX2-FAST-NEXT: vmovdqa 1664(%rdi), %ymm3
; AVX2-FAST-NEXT: vpbroadcastq 1648(%rdi), %ymm2
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5],ymm2[6,7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7]
; AVX2-FAST-NEXT: vmovdqa %ymm3, %ymm6
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4],ymm1[5,6,7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
; AVX2-FAST-NEXT: vmovdqa 1696(%rdi), %xmm2
; AVX2-FAST-NEXT: vmovdqa 1728(%rdi), %xmm3
; AVX2-FAST-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -9760,14 +9755,14 @@ define void @load_i32_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-NEXT: vmovdqa 96(%rdi), %ymm2
; AVX2-FAST-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vpbroadcastq 80(%rdi), %ymm1
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX2-FAST-NEXT: vmovdqa (%rdi), %ymm3
; AVX2-FAST-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovdqa 32(%rdi), %ymm2
; AVX2-FAST-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm2[6],ymm3[7]
; AVX2-FAST-NEXT: vpermd %ymm2, %ymm0, %ymm2
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4],ymm2[5,6,7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4,5,6,7]
; AVX2-FAST-NEXT: vmovdqa 128(%rdi), %xmm2
; AVX2-FAST-NEXT: vmovdqa 160(%rdi), %xmm3
; AVX2-FAST-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -9786,8 +9781,8 @@ define void @load_i32_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-NEXT: vmovdqa 544(%rdi), %ymm3
; AVX2-FAST-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vpbroadcastq 528(%rdi), %ymm2
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5],ymm2[6,7]
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4],ymm1[5,6,7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
; AVX2-FAST-NEXT: vmovdqa 576(%rdi), %xmm2
; AVX2-FAST-NEXT: vmovdqa 608(%rdi), %xmm3
; AVX2-FAST-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -9807,8 +9802,8 @@ define void @load_i32_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-NEXT: vmovdqa 992(%rdi), %ymm3
; AVX2-FAST-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vpbroadcastq 976(%rdi), %ymm2
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5],ymm2[6,7]
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4],ymm1[5,6,7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
; AVX2-FAST-NEXT: vmovdqa 1024(%rdi), %xmm2
; AVX2-FAST-NEXT: vmovdqa 1056(%rdi), %xmm3
; AVX2-FAST-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -9827,8 +9822,8 @@ define void @load_i32_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-NEXT: vmovdqa 1440(%rdi), %ymm2
; AVX2-FAST-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vpbroadcastq 1424(%rdi), %ymm1
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
; AVX2-FAST-NEXT: vmovdqa 1472(%rdi), %xmm1
; AVX2-FAST-NEXT: vmovdqa 1504(%rdi), %xmm2
; AVX2-FAST-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -9857,7 +9852,7 @@ define void @load_i32_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm12[0],ymm13[1],ymm12[2,3,4],ymm13[5],ymm12[6,7]
; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vpermd %ymm0, %ymm5, %ymm4
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3,4],ymm4[5,6,7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3,4,5,6,7]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4],ymm2[5,6,7]
; AVX2-FAST-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vpshufd {{.*#+}} xmm2 = mem[2,2,2,2]
@@ -9877,7 +9872,7 @@ define void @load_i32_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm10[0],ymm14[1],ymm10[2,3,4],ymm14[5],ymm10[6,7]
; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vpermd %ymm0, %ymm5, %ymm4
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3,4],ymm4[5,6,7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3,4,5,6,7]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4],ymm2[5,6,7]
; AVX2-FAST-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vpshufd {{.*#+}} xmm2 = mem[2,2,2,2]
@@ -9898,7 +9893,7 @@ define void @load_i32_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-NEXT: # ymm0 = ymm11[0],mem[1],ymm11[2,3,4],mem[5],ymm11[6,7]
; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vpermd %ymm0, %ymm5, %ymm4
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3,4],ymm4[5,6,7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3,4,5,6,7]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4],ymm2[5,6,7]
; AVX2-FAST-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vpshufd {{.*#+}} xmm2 = mem[2,2,2,2]
@@ -9921,7 +9916,7 @@ define void @load_i32_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-NEXT: # ymm4 = ymm0[0],mem[1],ymm0[2,3,4],mem[5],ymm0[6,7]
; AVX2-FAST-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vpermd %ymm4, %ymm5, %ymm4
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3,4],ymm4[5,6,7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3,4,5,6,7]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4],ymm2[5,6,7]
; AVX2-FAST-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vpshufd {{.*#+}} xmm2 = mem[2,2,2,2]
@@ -9943,7 +9938,7 @@ define void @load_i32_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-NEXT: # ymm4 = mem[0],ymm0[1],mem[2,3,4],ymm0[5],mem[6,7]
; AVX2-FAST-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vpermd %ymm4, %ymm5, %ymm4
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3,4],ymm4[5,6,7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3,4,5,6,7]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4],ymm2[5,6,7]
; AVX2-FAST-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vpshufd {{.*#+}} xmm2 = mem[2,2,2,2]
@@ -9965,7 +9960,7 @@ define void @load_i32_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-NEXT: # ymm0 = ymm0[0],mem[1],ymm0[2,3,4],mem[5],ymm0[6,7]
; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vpermd %ymm0, %ymm5, %ymm4
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3,4],ymm4[5,6,7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3,4,5,6,7]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4],ymm2[5,6,7]
; AVX2-FAST-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vpshufd {{.*#+}} xmm2 = mem[2,2,2,2]
@@ -9986,7 +9981,7 @@ define void @load_i32_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-NEXT: # ymm0 = ymm15[0],mem[1],ymm15[2,3,4],mem[5],ymm15[6,7]
; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vpermd %ymm0, %ymm5, %ymm4
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3,4],ymm4[5,6,7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3,4,5,6,7]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4],ymm2[5,6,7]
; AVX2-FAST-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vpshufd {{.*#+}} xmm2 = mem[2,2,2,2]
@@ -10009,7 +10004,7 @@ define void @load_i32_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-NEXT: # ymm1 = ymm1[0],mem[1],ymm1[2,3,4],mem[5],ymm1[6,7]
; AVX2-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vpermd %ymm1, %ymm5, %ymm1
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4],ymm1[5,6,7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7]
; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovdqa 304(%rdi), %xmm0
@@ -10251,8 +10246,8 @@ define void @load_i32_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm1 = xmm0[0,1,2],xmm1[3]
; AVX2-FAST-NEXT: vmovaps {{.*#+}} xmm5 = <4,3,u,u>
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-NEXT: vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm2 = ymm2[0,1,2,3],mem[4,5],ymm2[6,7]
+; AVX2-FAST-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm2 = ymm2[0,1,2,3],mem[4,5,6,7]
; AVX2-FAST-NEXT: vpermps %ymm2, %ymm5, %ymm2
; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3]
; AVX2-FAST-NEXT: vbroadcastsd {{.*#+}} ymm10 = [30064771072,30064771072,30064771072,30064771072]
@@ -10263,8 +10258,8 @@ define void @load_i32_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-FAST-NEXT: vblendps $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm1 = mem[0,1,2,3],ymm1[4,5],mem[6,7]
+; AVX2-FAST-NEXT: vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm1 = mem[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FAST-NEXT: vpermps %ymm1, %ymm5, %ymm2
; AVX2-FAST-NEXT: vbroadcastss 324(%rdi), %xmm3
; AVX2-FAST-NEXT: vmovaps 288(%rdi), %xmm1
@@ -10278,8 +10273,8 @@ define void @load_i32_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7]
; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-NEXT: vblendps $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm2 = mem[0,1,2,3],ymm2[4,5],mem[6,7]
+; AVX2-FAST-NEXT: vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm2 = mem[0,1,2,3],ymm2[4,5,6,7]
; AVX2-FAST-NEXT: vpermps %ymm2, %ymm5, %ymm3
; AVX2-FAST-NEXT: vbroadcastss 548(%rdi), %xmm4
; AVX2-FAST-NEXT: vmovaps 512(%rdi), %xmm2
@@ -10292,8 +10287,8 @@ define void @load_i32_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm4[4,5,6,7]
; AVX2-FAST-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-FAST-NEXT: vblendps $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm3 = mem[0,1,2,3],ymm3[4,5],mem[6,7]
+; AVX2-FAST-NEXT: vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm3 = mem[0,1,2,3],ymm3[4,5,6,7]
; AVX2-FAST-NEXT: vpermps %ymm3, %ymm5, %ymm4
; AVX2-FAST-NEXT: vbroadcastss 772(%rdi), %xmm6
; AVX2-FAST-NEXT: vmovaps 736(%rdi), %xmm3
@@ -10307,8 +10302,8 @@ define void @load_i32_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm6[4,5,6,7]
; AVX2-FAST-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX2-FAST-NEXT: vblendps $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm4 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm4 = mem[0,1,2,3],ymm4[4,5],mem[6,7]
+; AVX2-FAST-NEXT: vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm4 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm4 = mem[0,1,2,3],ymm4[4,5,6,7]
; AVX2-FAST-NEXT: vpermps %ymm4, %ymm5, %ymm6
; AVX2-FAST-NEXT: vbroadcastss 996(%rdi), %xmm7
; AVX2-FAST-NEXT: vmovaps 960(%rdi), %xmm4
@@ -10321,8 +10316,8 @@ define void @load_i32_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm7[4,5,6,7]
; AVX2-FAST-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-FAST-NEXT: vblendps $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm6 = mem[0,1,2,3],ymm6[4,5],mem[6,7]
+; AVX2-FAST-NEXT: vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm6 = mem[0,1,2,3],ymm6[4,5,6,7]
; AVX2-FAST-NEXT: vpermps %ymm6, %ymm5, %ymm6
; AVX2-FAST-NEXT: vbroadcastss 1220(%rdi), %xmm8
; AVX2-FAST-NEXT: vmovaps 1184(%rdi), %xmm7
@@ -10336,8 +10331,8 @@ define void @load_i32_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm8[4,5,6,7]
; AVX2-FAST-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-FAST-NEXT: vblendps $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm6 = mem[0,1,2,3],ymm6[4,5],mem[6,7]
+; AVX2-FAST-NEXT: vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm6 = mem[0,1,2,3],ymm6[4,5,6,7]
; AVX2-FAST-NEXT: vpermps %ymm6, %ymm5, %ymm6
; AVX2-FAST-NEXT: vbroadcastss 1444(%rdi), %xmm9
; AVX2-FAST-NEXT: vmovaps 1408(%rdi), %xmm8
@@ -10351,8 +10346,8 @@ define void @load_i32_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm9[4,5,6,7]
; AVX2-FAST-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-FAST-NEXT: vblendps $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm6 = mem[0,1,2,3],ymm6[4,5],mem[6,7]
+; AVX2-FAST-NEXT: vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm6 = mem[0,1,2,3],ymm6[4,5,6,7]
; AVX2-FAST-NEXT: vpermps %ymm6, %ymm5, %ymm5
; AVX2-FAST-NEXT: vbroadcastss 1668(%rdi), %xmm6
; AVX2-FAST-NEXT: vmovaps 1632(%rdi), %xmm11
@@ -10732,8 +10727,8 @@ define void @load_i32_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vpermd %ymm1, %ymm0, %ymm1
; AVX2-FAST-PERLANE-NEXT: vpbroadcastq 304(%rdi), %ymm2
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm8[4,5],ymm2[6,7]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4],ymm1[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm8[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqa 352(%rdi), %xmm2
; AVX2-FAST-PERLANE-NEXT: vmovdqa 384(%rdi), %xmm3
; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -10750,8 +10745,8 @@ define void @load_i32_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vpermd %ymm1, %ymm0, %ymm1
; AVX2-FAST-PERLANE-NEXT: vpbroadcastq 752(%rdi), %ymm2
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm13[4,5],ymm2[6,7]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4],ymm1[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm13[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqa 800(%rdi), %xmm2
; AVX2-FAST-PERLANE-NEXT: vmovdqa 832(%rdi), %xmm3
; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -10765,8 +10760,8 @@ define void @load_i32_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vpermd %ymm1, %ymm0, %ymm1
; AVX2-FAST-PERLANE-NEXT: vpbroadcastq 1200(%rdi), %ymm2
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm12[4,5],ymm2[6,7]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4],ymm1[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm12[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqa 1248(%rdi), %xmm2
; AVX2-FAST-PERLANE-NEXT: vmovdqa 1280(%rdi), %xmm3
; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -10785,8 +10780,8 @@ define void @load_i32_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-PERLANE-NEXT: vmovdqa 1664(%rdi), %ymm3
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vpbroadcastq 1648(%rdi), %ymm2
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5],ymm2[6,7]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4],ymm1[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqa 1696(%rdi), %xmm2
; AVX2-FAST-PERLANE-NEXT: vmovdqa 1728(%rdi), %xmm3
; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -10799,14 +10794,14 @@ define void @load_i32_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-PERLANE-NEXT: vmovdqa 96(%rdi), %ymm2
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vpbroadcastq 80(%rdi), %ymm1
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqa (%rdi), %ymm3
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovdqa 32(%rdi), %ymm2
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm2[6],ymm3[7]
; AVX2-FAST-PERLANE-NEXT: vpermd %ymm2, %ymm0, %ymm2
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4],ymm2[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqa 128(%rdi), %xmm2
; AVX2-FAST-PERLANE-NEXT: vmovdqa 160(%rdi), %xmm3
; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -10825,9 +10820,9 @@ define void @load_i32_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-PERLANE-NEXT: vpermd %ymm1, %ymm0, %ymm1
; AVX2-FAST-PERLANE-NEXT: vmovdqa 544(%rdi), %ymm6
; AVX2-FAST-PERLANE-NEXT: vpbroadcastq 528(%rdi), %ymm2
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm6[4,5],ymm2[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm6[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4],ymm1[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqa 576(%rdi), %xmm2
; AVX2-FAST-PERLANE-NEXT: vmovdqa 608(%rdi), %xmm3
; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -10845,9 +10840,9 @@ define void @load_i32_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-PERLANE-NEXT: vpermd %ymm1, %ymm0, %ymm1
; AVX2-FAST-PERLANE-NEXT: vmovdqa 992(%rdi), %ymm7
; AVX2-FAST-PERLANE-NEXT: vpbroadcastq 976(%rdi), %ymm2
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm7[4,5],ymm2[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm7[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4],ymm1[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqa 1024(%rdi), %xmm2
; AVX2-FAST-PERLANE-NEXT: vmovdqa 1056(%rdi), %xmm3
; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -10866,8 +10861,8 @@ define void @load_i32_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-PERLANE-NEXT: vmovdqa 1440(%rdi), %ymm2
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vpbroadcastq 1424(%rdi), %ymm1
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4],ymm0[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqa 1472(%rdi), %xmm1
; AVX2-FAST-PERLANE-NEXT: vmovdqa 1504(%rdi), %xmm2
; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -10896,7 +10891,7 @@ define void @load_i32_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} ymm2 = ymm2[1,0,3,3,5,4,7,7]
; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm2 = ymm2[0,3,2,3]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4],ymm2[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} xmm0 = mem[2,2,2,2]
@@ -10918,7 +10913,7 @@ define void @load_i32_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} ymm2 = ymm2[1,0,3,3,5,4,7,7]
; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm2 = ymm2[0,3,2,3]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4],ymm2[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} xmm0 = mem[2,2,2,2]
@@ -10941,7 +10936,7 @@ define void @load_i32_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} ymm2 = ymm2[1,0,3,3,5,4,7,7]
; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm2 = ymm2[0,3,2,3]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4],ymm2[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} xmm0 = mem[2,2,2,2]
@@ -10964,7 +10959,7 @@ define void @load_i32_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} ymm2 = ymm2[1,0,3,3,5,4,7,7]
; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm2 = ymm2[0,3,2,3]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4],ymm2[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} xmm0 = mem[2,2,2,2]
@@ -10985,7 +10980,7 @@ define void @load_i32_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} ymm2 = ymm2[1,0,3,3,5,4,7,7]
; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm2 = ymm2[0,3,2,3]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4],ymm2[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} xmm0 = mem[2,2,2,2]
@@ -11007,7 +11002,7 @@ define void @load_i32_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} ymm2 = ymm2[1,0,3,3,5,4,7,7]
; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm2 = ymm2[0,3,2,3]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4],ymm2[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} xmm0 = mem[2,2,2,2]
@@ -11032,7 +11027,7 @@ define void @load_i32_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} ymm2 = ymm2[1,0,3,3,5,4,7,7]
; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm2 = ymm2[0,3,2,3]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4],ymm2[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} xmm0 = mem[2,2,2,2]
@@ -11057,7 +11052,7 @@ define void @load_i32_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} ymm2 = ymm2[1,0,3,3,5,4,7,7]
; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm2 = ymm2[0,3,2,3]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4],ymm2[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovdqa 304(%rdi), %xmm0
@@ -11300,8 +11295,8 @@ define void @load_i32_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm1 = xmm0[0,1,2],xmm1[3]
; AVX2-FAST-PERLANE-NEXT: vmovaps {{.*#+}} xmm5 = <4,3,u,u>
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm2 = mem[0,1,2,3],ymm2[4,5],mem[6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm2 = mem[0,1,2,3],ymm2[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpermps %ymm2, %ymm5, %ymm2
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3]
; AVX2-FAST-PERLANE-NEXT: vbroadcastsd {{.*#+}} ymm9 = [30064771072,30064771072,30064771072,30064771072]
@@ -11312,8 +11307,8 @@ define void @load_i32_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm1 = mem[0,1,2,3],ymm1[4,5],mem[6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm1 = mem[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpermps %ymm1, %ymm5, %ymm2
; AVX2-FAST-PERLANE-NEXT: vbroadcastss 324(%rdi), %xmm3
; AVX2-FAST-PERLANE-NEXT: vmovaps 288(%rdi), %xmm1
@@ -11327,8 +11322,8 @@ define void @load_i32_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm2 = mem[0,1,2,3],ymm2[4,5],mem[6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm2 = mem[0,1,2,3],ymm2[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpermps %ymm2, %ymm5, %ymm3
; AVX2-FAST-PERLANE-NEXT: vbroadcastss 548(%rdi), %xmm4
; AVX2-FAST-PERLANE-NEXT: vmovaps 512(%rdi), %xmm2
@@ -11341,8 +11336,8 @@ define void @load_i32_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm4[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm3, (%rsp) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm3 = mem[0,1,2,3],ymm3[4,5],mem[6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm3 = mem[0,1,2,3],ymm3[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpermps %ymm3, %ymm5, %ymm4
; AVX2-FAST-PERLANE-NEXT: vbroadcastss 772(%rdi), %xmm6
; AVX2-FAST-PERLANE-NEXT: vmovaps 736(%rdi), %xmm3
@@ -11356,8 +11351,8 @@ define void @load_i32_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm6[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm4 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm4 = mem[0,1,2,3],ymm4[4,5],mem[6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm4 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm4 = mem[0,1,2,3],ymm4[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpermps %ymm4, %ymm5, %ymm6
; AVX2-FAST-PERLANE-NEXT: vbroadcastss 996(%rdi), %xmm7
; AVX2-FAST-PERLANE-NEXT: vmovaps 960(%rdi), %xmm4
@@ -11370,8 +11365,8 @@ define void @load_i32_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm7[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm6 = mem[0,1,2,3],ymm6[4,5],mem[6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm6 = mem[0,1,2,3],ymm6[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpermps %ymm6, %ymm5, %ymm6
; AVX2-FAST-PERLANE-NEXT: vbroadcastss 1220(%rdi), %xmm8
; AVX2-FAST-PERLANE-NEXT: vmovaps 1184(%rdi), %xmm7
@@ -11384,8 +11379,8 @@ define void @load_i32_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm8[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm6 = mem[0,1,2,3],ymm6[4,5],mem[6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm6 = mem[0,1,2,3],ymm6[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpermps %ymm6, %ymm5, %ymm6
; AVX2-FAST-PERLANE-NEXT: vbroadcastss 1444(%rdi), %xmm8
; AVX2-FAST-PERLANE-NEXT: vmovaps 1408(%rdi), %xmm13
@@ -11400,8 +11395,8 @@ define void @load_i32_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm8[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm6 = ymm6[0,1,2,3],mem[4,5],ymm6[6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm6 = ymm6[0,1,2,3],mem[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpermps %ymm6, %ymm5, %ymm5
; AVX2-FAST-PERLANE-NEXT: vbroadcastss 1668(%rdi), %xmm6
; AVX2-FAST-PERLANE-NEXT: vmovaps 1632(%rdi), %xmm15
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-load-i64-stride-3.ll b/llvm/test/CodeGen/X86/vector-interleaved-load-i64-stride-3.ll
index 76d48345849d7..5ae18683f155e 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-load-i64-stride-3.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-load-i64-stride-3.ll
@@ -145,19 +145,17 @@ define void @load_i64_stride3_vf4(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
; AVX1-ONLY-LABEL: load_i64_stride3_vf4:
; AVX1-ONLY: # %bb.0:
; AVX1-ONLY-NEXT: vmovapd 32(%rdi), %ymm0
-; AVX1-ONLY-NEXT: vmovapd (%rdi), %ymm1
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm1[0,1],ymm0[2],ymm1[3]
-; AVX1-ONLY-NEXT: vmovaps 16(%rdi), %xmm3
-; AVX1-ONLY-NEXT: vinsertf128 $1, 64(%rdi), %ymm3, %ymm3
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm2[0],ymm3[1],ymm2[2],ymm3[3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm1[0,1],ymm0[2,3]
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm1 = ymm1[1],ymm0[0],ymm1[3],ymm0[2]
+; AVX1-ONLY-NEXT: vmovaps 16(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = mem[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT: vinsertf128 $1, 64(%rdi), %ymm1, %ymm1
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm3 = ymm2[0],ymm1[1],ymm2[2],ymm1[3]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm2 = ymm2[1],ymm0[0],ymm2[3],ymm0[2]
; AVX1-ONLY-NEXT: vbroadcastsd 80(%rdi), %ymm4
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm1[0,1,2],ymm4[3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm2[0,1,2],ymm4[3]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],mem[2,3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm3[0],ymm0[1],ymm3[2],ymm0[3]
-; AVX1-ONLY-NEXT: vmovapd %ymm2, (%rsi)
-; AVX1-ONLY-NEXT: vmovapd %ymm1, (%rdx)
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[2],ymm0[3]
+; AVX1-ONLY-NEXT: vmovapd %ymm3, (%rsi)
+; AVX1-ONLY-NEXT: vmovapd %ymm2, (%rdx)
; AVX1-ONLY-NEXT: vmovapd %ymm0, (%rcx)
; AVX1-ONLY-NEXT: vzeroupper
; AVX1-ONLY-NEXT: retq
@@ -168,7 +166,7 @@ define void @load_i64_stride3_vf4(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
; AVX2-ONLY-NEXT: vmovaps (%rdi), %ymm1
; AVX2-ONLY-NEXT: vinsertf128 $1, 64(%rdi), %ymm0, %ymm2
; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm3 = ymm1[0,3,2,3]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm0[4,5],ymm3[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm0[4,5,6,7]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm2[6,7]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5,6,7]
; AVX2-ONLY-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[2,3,0,1,6,7,4,5]
@@ -257,35 +255,31 @@ define void @load_i64_stride3_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
; AVX1-ONLY-LABEL: load_i64_stride3_vf8:
; AVX1-ONLY: # %bb.0:
; AVX1-ONLY-NEXT: vmovapd 128(%rdi), %ymm0
-; AVX1-ONLY-NEXT: vmovapd 96(%rdi), %ymm1
-; AVX1-ONLY-NEXT: vmovapd 32(%rdi), %ymm2
-; AVX1-ONLY-NEXT: vmovapd (%rdi), %ymm3
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm4 = ymm3[0,1],ymm2[2],ymm3[3]
-; AVX1-ONLY-NEXT: vmovaps 16(%rdi), %xmm5
-; AVX1-ONLY-NEXT: vinsertf128 $1, 64(%rdi), %ymm5, %ymm5
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm4 = ymm4[0],ymm5[1],ymm4[2],ymm5[3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm6 = ymm1[0,1],ymm0[2],ymm1[3]
-; AVX1-ONLY-NEXT: vmovaps 112(%rdi), %xmm7
-; AVX1-ONLY-NEXT: vinsertf128 $1, 160(%rdi), %ymm7, %ymm7
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm6 = ymm6[0],ymm7[1],ymm6[2],ymm7[3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm3 = ymm3[0,1],ymm2[2,3]
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm3 = ymm3[1],ymm2[0],ymm3[3],ymm2[2]
+; AVX1-ONLY-NEXT: vmovapd 32(%rdi), %ymm1
+; AVX1-ONLY-NEXT: vmovaps 16(%rdi), %xmm2
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm3 = mem[0,1],ymm1[2,3]
+; AVX1-ONLY-NEXT: vinsertf128 $1, 64(%rdi), %ymm2, %ymm2
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm4 = ymm3[0],ymm2[1],ymm3[2],ymm2[3]
+; AVX1-ONLY-NEXT: vmovaps 112(%rdi), %xmm5
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm6 = mem[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT: vinsertf128 $1, 160(%rdi), %ymm5, %ymm5
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm7 = ymm6[0],ymm5[1],ymm6[2],ymm5[3]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm3 = ymm3[1],ymm1[0],ymm3[3],ymm1[2]
; AVX1-ONLY-NEXT: vbroadcastsd 80(%rdi), %ymm8
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm3 = ymm3[0,1,2],ymm8[3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm1[0,1],ymm0[2,3]
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm1 = ymm1[1],ymm0[0],ymm1[3],ymm0[2]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm6 = ymm6[1],ymm0[0],ymm6[3],ymm0[2]
; AVX1-ONLY-NEXT: vbroadcastsd 176(%rdi), %ymm8
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm1[0,1,2],ymm8[3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm2[0,1],mem[2,3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm5[0],ymm2[1],ymm5[2],ymm2[3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm6 = ymm6[0,1,2],ymm8[3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm1[0,1],mem[2,3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm2[0],ymm1[1],ymm2[2],ymm1[3]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],mem[2,3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm7[0],ymm0[1],ymm7[2],ymm0[3]
-; AVX1-ONLY-NEXT: vmovapd %ymm6, 32(%rsi)
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm5[0],ymm0[1],ymm5[2],ymm0[3]
+; AVX1-ONLY-NEXT: vmovapd %ymm7, 32(%rsi)
; AVX1-ONLY-NEXT: vmovapd %ymm4, (%rsi)
-; AVX1-ONLY-NEXT: vmovapd %ymm1, 32(%rdx)
+; AVX1-ONLY-NEXT: vmovapd %ymm6, 32(%rdx)
; AVX1-ONLY-NEXT: vmovapd %ymm3, (%rdx)
; AVX1-ONLY-NEXT: vmovapd %ymm0, 32(%rcx)
-; AVX1-ONLY-NEXT: vmovapd %ymm2, (%rcx)
+; AVX1-ONLY-NEXT: vmovapd %ymm1, (%rcx)
; AVX1-ONLY-NEXT: vzeroupper
; AVX1-ONLY-NEXT: retq
;
@@ -297,11 +291,11 @@ define void @load_i64_stride3_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
; AVX2-ONLY-NEXT: vmovaps (%rdi), %ymm3
; AVX2-ONLY-NEXT: vinsertf128 $1, 64(%rdi), %ymm0, %ymm4
; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm5 = ymm3[0,3,2,3]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm2[4,5],ymm5[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm2[4,5,6,7]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm5[0,1,2,3,4,5],ymm4[6,7]
; AVX2-ONLY-NEXT: vinsertf128 $1, 160(%rdi), %ymm0, %ymm5
; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm6 = ymm1[0,3,2,3]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm0[4,5],ymm6[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm0[4,5,6,7]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm6[0,1,2,3,4,5],ymm5[6,7]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm3[2,3],ymm2[4,5,6,7]
; AVX2-ONLY-NEXT: vpermilps {{.*#+}} ymm2 = ymm2[2,3,0,1,6,7,4,5]
@@ -468,65 +462,57 @@ define void @load_i64_stride3_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-LABEL: load_i64_stride3_vf16:
; AVX1-ONLY: # %bb.0:
; AVX1-ONLY-NEXT: vmovapd 32(%rdi), %ymm1
-; AVX1-ONLY-NEXT: vmovapd (%rdi), %ymm2
; AVX1-ONLY-NEXT: vmovapd 224(%rdi), %ymm0
-; AVX1-ONLY-NEXT: vmovapd 192(%rdi), %ymm3
-; AVX1-ONLY-NEXT: vmovapd 320(%rdi), %ymm4
-; AVX1-ONLY-NEXT: vmovapd 288(%rdi), %ymm5
-; AVX1-ONLY-NEXT: vmovapd 128(%rdi), %ymm6
-; AVX1-ONLY-NEXT: vmovapd 96(%rdi), %ymm8
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm7 = ymm8[0,1],ymm6[2],ymm8[3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm8 = ymm8[0,1],ymm6[2,3]
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm8 = ymm8[1],ymm6[0],ymm8[3],ymm6[2]
+; AVX1-ONLY-NEXT: vmovapd 320(%rdi), %ymm2
+; AVX1-ONLY-NEXT: vmovapd 128(%rdi), %ymm3
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm4 = mem[0,1],ymm3[2,3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm5 = mem[0,1],ymm2[2,3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm6 = mem[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm7 = mem[0,1],ymm1[2,3]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm8 = ymm4[1],ymm3[0],ymm4[3],ymm3[2]
; AVX1-ONLY-NEXT: vbroadcastsd 176(%rdi), %ymm9
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm8 = ymm8[0,1,2],ymm9[3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm9 = ymm5[0,1],ymm4[2,3]
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm9 = ymm9[1],ymm4[0],ymm9[3],ymm4[2]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm9 = ymm5[1],ymm2[0],ymm5[3],ymm2[2]
; AVX1-ONLY-NEXT: vbroadcastsd 368(%rdi), %ymm10
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm9 = ymm9[0,1,2],ymm10[3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm10 = ymm2[0,1],ymm1[2,3]
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm10 = ymm10[1],ymm1[0],ymm10[3],ymm1[2]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm10 = ymm7[1],ymm1[0],ymm7[3],ymm1[2]
; AVX1-ONLY-NEXT: vbroadcastsd 80(%rdi), %ymm11
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm10 = ymm10[0,1,2],ymm11[3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm11 = ymm3[0,1],ymm0[2,3]
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm11 = ymm11[1],ymm0[0],ymm11[3],ymm0[2]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm11 = ymm6[1],ymm0[0],ymm6[3],ymm0[2]
; AVX1-ONLY-NEXT: vbroadcastsd 272(%rdi), %ymm12
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm11 = ymm11[0,1,2],ymm12[3]
; AVX1-ONLY-NEXT: vmovaps 112(%rdi), %xmm12
; AVX1-ONLY-NEXT: vinsertf128 $1, 160(%rdi), %ymm12, %ymm12
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm7 = ymm7[0],ymm12[1],ymm7[2],ymm12[3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm5 = ymm5[0,1],ymm4[2],ymm5[3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm6 = ymm6[0,1],mem[2,3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm6 = ymm12[0],ymm6[1],ymm12[2],ymm6[3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm4 = ymm4[0],ymm12[1],ymm4[2],ymm12[3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm3 = ymm3[0,1],mem[2,3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm3 = ymm12[0],ymm3[1],ymm12[2],ymm3[3]
; AVX1-ONLY-NEXT: vmovaps 304(%rdi), %xmm12
; AVX1-ONLY-NEXT: vinsertf128 $1, 352(%rdi), %ymm12, %ymm12
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm5 = ymm5[0],ymm12[1],ymm5[2],ymm12[3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm3 = ymm3[0,1],ymm0[2],ymm3[3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm2[0,1],ymm1[2],ymm2[3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm4 = ymm4[0,1],mem[2,3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm4 = ymm12[0],ymm4[1],ymm12[2],ymm4[3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm2[0,1],mem[2,3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm12[0],ymm2[1],ymm12[2],ymm2[3]
; AVX1-ONLY-NEXT: vmovaps 16(%rdi), %xmm12
; AVX1-ONLY-NEXT: vinsertf128 $1, 64(%rdi), %ymm12, %ymm12
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm2[0],ymm12[1],ymm2[2],ymm12[3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm7 = ymm7[0],ymm12[1],ymm7[2],ymm12[3]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm1[0,1],mem[2,3]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm12[0],ymm1[1],ymm12[2],ymm1[3]
; AVX1-ONLY-NEXT: vmovaps 208(%rdi), %xmm12
; AVX1-ONLY-NEXT: vinsertf128 $1, 256(%rdi), %ymm12, %ymm12
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm3 = ymm3[0],ymm12[1],ymm3[2],ymm12[3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm6 = ymm6[0],ymm12[1],ymm6[2],ymm12[3]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],mem[2,3]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm12[0],ymm0[1],ymm12[2],ymm0[3]
-; AVX1-ONLY-NEXT: vmovapd %ymm2, (%rsi)
-; AVX1-ONLY-NEXT: vmovapd %ymm3, 64(%rsi)
+; AVX1-ONLY-NEXT: vmovapd %ymm7, (%rsi)
+; AVX1-ONLY-NEXT: vmovapd %ymm6, 64(%rsi)
; AVX1-ONLY-NEXT: vmovapd %ymm5, 96(%rsi)
-; AVX1-ONLY-NEXT: vmovapd %ymm7, 32(%rsi)
+; AVX1-ONLY-NEXT: vmovapd %ymm4, 32(%rsi)
; AVX1-ONLY-NEXT: vmovapd %ymm11, 64(%rdx)
; AVX1-ONLY-NEXT: vmovapd %ymm10, (%rdx)
; AVX1-ONLY-NEXT: vmovapd %ymm9, 96(%rdx)
; AVX1-ONLY-NEXT: vmovapd %ymm8, 32(%rdx)
; AVX1-ONLY-NEXT: vmovapd %ymm0, 64(%rcx)
; AVX1-ONLY-NEXT: vmovapd %ymm1, (%rcx)
-; AVX1-ONLY-NEXT: vmovapd %ymm4, 96(%rcx)
-; AVX1-ONLY-NEXT: vmovapd %ymm6, 32(%rcx)
+; AVX1-ONLY-NEXT: vmovapd %ymm2, 96(%rcx)
+; AVX1-ONLY-NEXT: vmovapd %ymm3, 32(%rcx)
; AVX1-ONLY-NEXT: vzeroupper
; AVX1-ONLY-NEXT: retq
;
@@ -542,19 +528,19 @@ define void @load_i64_stride3_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-ONLY-NEXT: vmovaps 96(%rdi), %ymm11
; AVX2-ONLY-NEXT: vinsertf128 $1, 160(%rdi), %ymm0, %ymm0
; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm1 = ymm11[0,3,2,3]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5,6,7]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX2-ONLY-NEXT: vinsertf128 $1, 352(%rdi), %ymm0, %ymm1
; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm2 = ymm10[0,3,2,3]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm5[4,5],ymm2[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm5[4,5,6,7]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
; AVX2-ONLY-NEXT: vinsertf128 $1, 256(%rdi), %ymm0, %ymm2
; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm4 = ymm7[0,3,2,3]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm6[4,5],ymm4[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm6[4,5,6,7]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm4[0,1,2,3,4,5],ymm2[6,7]
; AVX2-ONLY-NEXT: vinsertf128 $1, 64(%rdi), %ymm0, %ymm4
; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm12 = ymm9[0,3,2,3]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3],ymm8[4,5],ymm12[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3],ymm8[4,5,6,7]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm12[0,1,2,3,4,5],ymm4[6,7]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1],ymm11[2,3],ymm3[4,5,6,7]
; AVX2-ONLY-NEXT: vpermilps {{.*#+}} ymm3 = ymm3[2,3,0,1,6,7,4,5]
@@ -885,152 +871,136 @@ define void @load_i64_stride3_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-LABEL: load_i64_stride3_vf32:
; AVX1-ONLY: # %bb.0:
; AVX1-ONLY-NEXT: subq $296, %rsp # imm = 0x128
-; AVX1-ONLY-NEXT: vmovapd 32(%rdi), %ymm15
-; AVX1-ONLY-NEXT: vmovapd (%rdi), %ymm2
-; AVX1-ONLY-NEXT: vmovapd 704(%rdi), %ymm3
-; AVX1-ONLY-NEXT: vmovapd 672(%rdi), %ymm0
-; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovapd 512(%rdi), %ymm4
-; AVX1-ONLY-NEXT: vmovapd 480(%rdi), %ymm1
-; AVX1-ONLY-NEXT: vmovupd %ymm1, (%rsp) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovapd 320(%rdi), %ymm6
-; AVX1-ONLY-NEXT: vmovapd 288(%rdi), %ymm5
-; AVX1-ONLY-NEXT: vmovupd %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovapd 128(%rdi), %ymm8
-; AVX1-ONLY-NEXT: vmovapd 96(%rdi), %ymm9
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm10 = ymm9[0,1],ymm8[2,3]
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm10 = ymm10[1],ymm8[0],ymm10[3],ymm8[2]
-; AVX1-ONLY-NEXT: vbroadcastsd 176(%rdi), %ymm11
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm7 = ymm10[0,1,2],ymm11[3]
-; AVX1-ONLY-NEXT: vmovupd %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm11 = ymm5[0,1],ymm6[2,3]
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm11 = ymm11[1],ymm6[0],ymm11[3],ymm6[2]
+; AVX1-ONLY-NEXT: vmovapd 32(%rdi), %ymm2
+; AVX1-ONLY-NEXT: vmovapd 704(%rdi), %ymm0
+; AVX1-ONLY-NEXT: vmovapd 512(%rdi), %ymm1
+; AVX1-ONLY-NEXT: vmovapd 320(%rdi), %ymm8
+; AVX1-ONLY-NEXT: vmovapd 128(%rdi), %ymm6
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm7 = mem[0,1],ymm6[2,3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm3 = mem[0,1],ymm8[2,3]
+; AVX1-ONLY-NEXT: vmovupd %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm4 = mem[0,1],ymm1[2,3]
+; AVX1-ONLY-NEXT: vmovupd %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovapd %ymm1, %ymm5
+; AVX1-ONLY-NEXT: vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = mem[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT: vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovupd %ymm0, (%rsp) # 32-byte Spill
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm11 = mem[0,1],ymm2[2,3]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm9 = ymm7[1],ymm6[0],ymm7[3],ymm6[2]
+; AVX1-ONLY-NEXT: vbroadcastsd 176(%rdi), %ymm10
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm9 = ymm9[0,1,2],ymm10[3]
+; AVX1-ONLY-NEXT: vmovupd %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm10 = ymm3[1],ymm8[0],ymm3[3],ymm8[2]
; AVX1-ONLY-NEXT: vbroadcastsd 368(%rdi), %ymm12
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm5 = ymm11[0,1,2],ymm12[3]
-; AVX1-ONLY-NEXT: vmovupd %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovapd %ymm4, %ymm5
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm12 = ymm1[0,1],ymm4[2,3]
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm12 = ymm12[1],ymm4[0],ymm12[3],ymm4[2]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm3 = ymm10[0,1,2],ymm12[3]
+; AVX1-ONLY-NEXT: vmovupd %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm12 = ymm4[1],ymm5[0],ymm4[3],ymm5[2]
; AVX1-ONLY-NEXT: vbroadcastsd 560(%rdi), %ymm13
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm12[0,1,2],ymm13[3]
-; AVX1-ONLY-NEXT: vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovapd %ymm3, %ymm7
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm13 = ymm0[0,1],ymm3[2,3]
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm13 = ymm13[1],ymm3[0],ymm13[3],ymm3[2]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm3 = ymm12[0,1,2],ymm13[3]
+; AVX1-ONLY-NEXT: vmovupd %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm13 = ymm1[1],ymm0[0],ymm1[3],ymm0[2]
; AVX1-ONLY-NEXT: vbroadcastsd 752(%rdi), %ymm14
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm13[0,1,2],ymm14[3]
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovapd %ymm15, %ymm1
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm14 = ymm2[0,1],ymm15[2,3]
-; AVX1-ONLY-NEXT: vmovapd %ymm2, %ymm4
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm14 = ymm14[1],ymm15[0],ymm14[3],ymm15[2]
+; AVX1-ONLY-NEXT: vmovapd %ymm2, %ymm13
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm14 = ymm11[1],ymm2[0],ymm11[3],ymm2[2]
; AVX1-ONLY-NEXT: vbroadcastsd 80(%rdi), %ymm15
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm14[0,1,2],ymm15[3]
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovapd 224(%rdi), %ymm0
-; AVX1-ONLY-NEXT: vmovapd 192(%rdi), %ymm2
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm15 = ymm2[0,1],ymm0[2,3]
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm15 = ymm15[1],ymm0[0],ymm15[3],ymm0[2]
-; AVX1-ONLY-NEXT: vbroadcastsd 272(%rdi), %ymm10
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm3 = ymm15[0,1,2],ymm10[3]
-; AVX1-ONLY-NEXT: vmovupd %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovapd 416(%rdi), %ymm10
-; AVX1-ONLY-NEXT: vmovapd 384(%rdi), %ymm3
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm11 = ymm3[0,1],ymm10[2,3]
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm11 = ymm11[1],ymm10[0],ymm11[3],ymm10[2]
-; AVX1-ONLY-NEXT: vbroadcastsd 464(%rdi), %ymm12
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm11 = ymm11[0,1,2],ymm12[3]
-; AVX1-ONLY-NEXT: vmovupd %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovapd 608(%rdi), %ymm12
-; AVX1-ONLY-NEXT: vmovapd 576(%rdi), %ymm11
-; AVX1-ONLY-NEXT: vmovupd %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm13 = ymm11[0,1],ymm12[2,3]
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm13 = ymm13[1],ymm12[0],ymm13[3],ymm12[2]
-; AVX1-ONLY-NEXT: vbroadcastsd 656(%rdi), %ymm14
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm13 = ymm13[0,1,2],ymm14[3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm9 = ymm9[0,1],ymm8[2],ymm9[3]
-; AVX1-ONLY-NEXT: vmovaps 112(%rdi), %xmm14
-; AVX1-ONLY-NEXT: vinsertf128 $1, 160(%rdi), %ymm14, %ymm14
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm9 = ymm9[0],ymm14[1],ymm9[2],ymm14[3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm8 = ymm8[0,1],mem[2,3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm8 = ymm14[0],ymm8[1],ymm14[2],ymm8[3]
-; AVX1-ONLY-NEXT: vmovupd %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm14 = ymm4[0,1],ymm1[2],ymm4[3]
-; AVX1-ONLY-NEXT: vmovaps 16(%rdi), %xmm8
-; AVX1-ONLY-NEXT: vinsertf128 $1, 64(%rdi), %ymm8, %ymm8
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm14 = ymm14[0],ymm8[1],ymm14[2],ymm8[3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm4 = ymm1[0,1],mem[2,3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm8[0],ymm4[1],ymm8[2],ymm4[3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm3 = mem[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm15 = ymm3[1],ymm0[0],ymm3[3],ymm0[2]
+; AVX1-ONLY-NEXT: vbroadcastsd 272(%rdi), %ymm9
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm15[0,1,2],ymm9[3]
; AVX1-ONLY-NEXT: vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendpd $11, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm8 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm8 = mem[0,1],ymm6[2],mem[3]
-; AVX1-ONLY-NEXT: vmovaps 304(%rdi), %xmm15
-; AVX1-ONLY-NEXT: vinsertf128 $1, 352(%rdi), %ymm15, %ymm15
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm8 = ymm8[0],ymm15[1],ymm8[2],ymm15[3]
+; AVX1-ONLY-NEXT: vmovapd 416(%rdi), %ymm9
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm4 = mem[0,1],ymm9[2,3]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm1 = ymm4[1],ymm9[0],ymm4[3],ymm9[2]
+; AVX1-ONLY-NEXT: vbroadcastsd 464(%rdi), %ymm10
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm10 = ymm1[0,1,2],ymm10[3]
+; AVX1-ONLY-NEXT: vmovapd 608(%rdi), %ymm1
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm5 = mem[0,1],ymm1[2,3]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm2 = ymm5[1],ymm1[0],ymm5[3],ymm1[2]
+; AVX1-ONLY-NEXT: vbroadcastsd 656(%rdi), %ymm12
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm12 = ymm2[0,1,2],ymm12[3]
+; AVX1-ONLY-NEXT: vmovaps 112(%rdi), %xmm2
+; AVX1-ONLY-NEXT: vinsertf128 $1, 160(%rdi), %ymm2, %ymm2
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm7 = ymm7[0],ymm2[1],ymm7[2],ymm2[3]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm6 = ymm6[0,1],mem[2,3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm6 = ymm15[0],ymm6[1],ymm15[2],ymm6[3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm2[0,1],ymm0[2],ymm2[3]
-; AVX1-ONLY-NEXT: vmovaps 208(%rdi), %xmm15
-; AVX1-ONLY-NEXT: vinsertf128 $1, 256(%rdi), %ymm15, %ymm15
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm1[0],ymm15[1],ymm1[2],ymm15[3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm2[0],ymm6[1],ymm2[2],ymm6[3]
+; AVX1-ONLY-NEXT: vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 16(%rdi), %xmm6
+; AVX1-ONLY-NEXT: vinsertf128 $1, 64(%rdi), %ymm6, %ymm6
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm11 = ymm11[0],ymm6[1],ymm11[2],ymm6[3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm13 = ymm13[0,1],mem[2,3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm6[0],ymm13[1],ymm6[2],ymm13[3]
+; AVX1-ONLY-NEXT: vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 304(%rdi), %xmm13
+; AVX1-ONLY-NEXT: vinsertf128 $1, 352(%rdi), %ymm13, %ymm13
+; AVX1-ONLY-NEXT: vblendpd $5, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm6 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm6 = mem[0],ymm13[1],mem[2],ymm13[3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm8 = ymm8[0,1],mem[2,3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm13[0],ymm8[1],ymm13[2],ymm8[3]
+; AVX1-ONLY-NEXT: vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 208(%rdi), %xmm13
+; AVX1-ONLY-NEXT: vinsertf128 $1, 256(%rdi), %ymm13, %ymm13
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm3 = ymm3[0],ymm13[1],ymm3[2],ymm13[3]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],mem[2,3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm15[0],ymm0[1],ymm15[2],ymm0[3]
-; AVX1-ONLY-NEXT: vblendpd $11, (%rsp), %ymm5, %ymm15 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm15 = mem[0,1],ymm5[2],mem[3]
-; AVX1-ONLY-NEXT: vmovaps 496(%rdi), %xmm4
-; AVX1-ONLY-NEXT: vinsertf128 $1, 544(%rdi), %ymm4, %ymm4
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm15 = ymm15[0],ymm4[1],ymm15[2],ymm4[3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm5 = ymm5[0,1],mem[2,3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm4 = ymm4[0],ymm5[1],ymm4[2],ymm5[3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm3[0,1],ymm10[2],ymm3[3]
-; AVX1-ONLY-NEXT: vmovaps 400(%rdi), %xmm5
-; AVX1-ONLY-NEXT: vinsertf128 $1, 448(%rdi), %ymm5, %ymm5
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm2[0],ymm5[1],ymm2[2],ymm5[3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm10 = ymm10[0,1],mem[2,3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm5 = ymm5[0],ymm10[1],ymm5[2],ymm10[3]
-; AVX1-ONLY-NEXT: vblendpd $11, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm10 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm10 = mem[0,1],ymm7[2],mem[3]
-; AVX1-ONLY-NEXT: vmovaps 688(%rdi), %xmm11
-; AVX1-ONLY-NEXT: vinsertf128 $1, 736(%rdi), %ymm11, %ymm11
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm10 = ymm10[0],ymm11[1],ymm10[2],ymm11[3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm7 = ymm7[0,1],mem[2,3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm7 = ymm11[0],ymm7[1],ymm11[2],ymm7[3]
-; AVX1-ONLY-NEXT: vblendpd $11, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm3 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm3 = mem[0,1],ymm12[2],mem[3]
-; AVX1-ONLY-NEXT: vmovaps 592(%rdi), %xmm11
-; AVX1-ONLY-NEXT: vinsertf128 $1, 640(%rdi), %ymm11, %ymm11
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm3 = ymm3[0],ymm11[1],ymm3[2],ymm11[3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm12 = ymm12[0,1],mem[2,3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm11 = ymm11[0],ymm12[1],ymm11[2],ymm12[3]
-; AVX1-ONLY-NEXT: vmovapd %ymm3, 192(%rsi)
-; AVX1-ONLY-NEXT: vmovapd %ymm2, 128(%rsi)
-; AVX1-ONLY-NEXT: vmovapd %ymm1, 64(%rsi)
-; AVX1-ONLY-NEXT: vmovapd %ymm14, (%rsi)
-; AVX1-ONLY-NEXT: vmovapd %ymm10, 224(%rsi)
-; AVX1-ONLY-NEXT: vmovapd %ymm15, 160(%rsi)
-; AVX1-ONLY-NEXT: vmovapd %ymm8, 96(%rsi)
-; AVX1-ONLY-NEXT: vmovapd %ymm9, 32(%rsi)
-; AVX1-ONLY-NEXT: vmovapd %ymm13, 192(%rdx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm1, 128(%rdx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm1, 64(%rdx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm1, (%rdx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm1, 224(%rdx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm1, 160(%rdx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm1, 96(%rdx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm1, 32(%rdx)
-; AVX1-ONLY-NEXT: vmovapd %ymm11, 192(%rcx)
-; AVX1-ONLY-NEXT: vmovapd %ymm7, 224(%rcx)
-; AVX1-ONLY-NEXT: vmovapd %ymm5, 128(%rcx)
-; AVX1-ONLY-NEXT: vmovapd %ymm4, 160(%rcx)
-; AVX1-ONLY-NEXT: vmovapd %ymm0, 64(%rcx)
-; AVX1-ONLY-NEXT: vmovapd %ymm6, 96(%rcx)
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm8 = ymm13[0],ymm0[1],ymm13[2],ymm0[3]
+; AVX1-ONLY-NEXT: vmovaps 496(%rdi), %xmm13
+; AVX1-ONLY-NEXT: vinsertf128 $1, 544(%rdi), %ymm13, %ymm13
+; AVX1-ONLY-NEXT: vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm2 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm2 = mem[0,1],ymm13[2,3],mem[4,5],ymm13[6,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm14 = ymm0[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0,1],ymm14[2,3],ymm13[4,5],ymm14[6,7]
+; AVX1-ONLY-NEXT: vmovaps 400(%rdi), %xmm14
+; AVX1-ONLY-NEXT: vinsertf128 $1, 448(%rdi), %ymm14, %ymm14
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm4 = ymm4[0],ymm14[1],ymm4[2],ymm14[3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm9 = ymm9[0,1],mem[2,3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm9 = ymm14[0],ymm9[1],ymm14[2],ymm9[3]
+; AVX1-ONLY-NEXT: vmovaps 688(%rdi), %xmm14
+; AVX1-ONLY-NEXT: vinsertf128 $1, 736(%rdi), %ymm14, %ymm14
+; AVX1-ONLY-NEXT: vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm0 = mem[0,1],ymm14[2,3],mem[4,5],ymm14[6,7]
+; AVX1-ONLY-NEXT: vmovups (%rsp), %ymm15 # 32-byte Reload
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm15 = ymm15[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1],ymm15[2,3],ymm14[4,5],ymm15[6,7]
+; AVX1-ONLY-NEXT: vmovaps 592(%rdi), %xmm15
+; AVX1-ONLY-NEXT: vinsertf128 $1, 640(%rdi), %ymm15, %ymm15
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm5 = ymm5[0],ymm15[1],ymm5[2],ymm15[3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm1[0,1],mem[2,3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm15[0],ymm1[1],ymm15[2],ymm1[3]
+; AVX1-ONLY-NEXT: vmovapd %ymm5, 192(%rsi)
+; AVX1-ONLY-NEXT: vmovapd %ymm4, 128(%rsi)
+; AVX1-ONLY-NEXT: vmovapd %ymm3, 64(%rsi)
+; AVX1-ONLY-NEXT: vmovapd %ymm11, (%rsi)
+; AVX1-ONLY-NEXT: vmovaps %ymm0, 224(%rsi)
+; AVX1-ONLY-NEXT: vmovaps %ymm2, 160(%rsi)
+; AVX1-ONLY-NEXT: vmovapd %ymm6, 96(%rsi)
+; AVX1-ONLY-NEXT: vmovapd %ymm7, 32(%rsi)
+; AVX1-ONLY-NEXT: vmovapd %ymm12, 192(%rdx)
+; AVX1-ONLY-NEXT: vmovapd %ymm10, 128(%rdx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm0, 64(%rdx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm0, (%rdx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm0, 224(%rdx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm0, 160(%rdx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm0, 96(%rdx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm0, 32(%rdx)
+; AVX1-ONLY-NEXT: vmovapd %ymm1, 192(%rcx)
+; AVX1-ONLY-NEXT: vmovaps %ymm14, 224(%rcx)
+; AVX1-ONLY-NEXT: vmovapd %ymm9, 128(%rcx)
+; AVX1-ONLY-NEXT: vmovaps %ymm13, 160(%rcx)
+; AVX1-ONLY-NEXT: vmovapd %ymm8, 64(%rcx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm0, 96(%rcx)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm0, (%rcx)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
@@ -1056,47 +1026,47 @@ define void @load_i64_stride3_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-ONLY-NEXT: vmovaps 96(%rdi), %ymm5
; AVX2-ONLY-NEXT: vinsertf128 $1, 160(%rdi), %ymm0, %ymm6
; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm8 = ymm5[0,3,2,3]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3],ymm4[4,5],ymm8[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3],ymm4[4,5,6,7]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm8[0,1,2,3,4,5],ymm6[6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vinsertf128 $1, 352(%rdi), %ymm0, %ymm6
; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm8 = ymm7[0,3,2,3]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3],ymm9[4,5],ymm8[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3],ymm9[4,5,6,7]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm8[0,1,2,3,4,5],ymm6[6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vinsertf128 $1, 544(%rdi), %ymm0, %ymm6
; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm8 = ymm10[0,3,2,3]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3],ymm11[4,5],ymm8[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3],ymm11[4,5,6,7]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm8[0,1,2,3,4,5],ymm6[6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vinsertf128 $1, 736(%rdi), %ymm0, %ymm6
; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm8 = ymm12[0,3,2,3]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3],ymm13[4,5],ymm8[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3],ymm13[4,5,6,7]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm8[0,1,2,3,4,5],ymm6[6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vinsertf128 $1, 64(%rdi), %ymm0, %ymm6
; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm8 = ymm2[0,3,2,3]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3],ymm1[4,5],ymm8[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm8[0,1,2,3,4,5],ymm6[6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovaps 224(%rdi), %ymm14
; AVX2-ONLY-NEXT: vmovaps 192(%rdi), %ymm15
; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm6 = ymm15[0,3,2,3]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm14[4,5],ymm6[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm14[4,5,6,7]
; AVX2-ONLY-NEXT: vinsertf128 $1, 256(%rdi), %ymm0, %ymm8
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm6[0,1,2,3,4,5],ymm8[6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, (%rsp) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovaps 416(%rdi), %ymm3
; AVX2-ONLY-NEXT: vmovaps 384(%rdi), %ymm0
; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm6 = ymm0[0,3,2,3]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm3[4,5],ymm6[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm3[4,5,6,7]
; AVX2-ONLY-NEXT: vinsertf128 $1, 448(%rdi), %ymm0, %ymm8
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm6[0,1,2,3,4,5],ymm8[6,7]
; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovaps 608(%rdi), %ymm2
; AVX2-ONLY-NEXT: vmovaps 576(%rdi), %ymm1
; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm6 = ymm1[0,3,2,3]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm2[4,5],ymm6[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm2[4,5,6,7]
; AVX2-ONLY-NEXT: vinsertf128 $1, 640(%rdi), %ymm0, %ymm8
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5],ymm8[6,7]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm5[2,3],ymm4[4,5,6,7]
@@ -1770,282 +1740,249 @@ define void @load_i64_stride3_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
;
; AVX1-ONLY-LABEL: load_i64_stride3_vf64:
; AVX1-ONLY: # %bb.0:
-; AVX1-ONLY-NEXT: subq $1160, %rsp # imm = 0x488
-; AVX1-ONLY-NEXT: vmovapd 896(%rdi), %ymm2
-; AVX1-ONLY-NEXT: vmovapd 864(%rdi), %ymm3
-; AVX1-ONLY-NEXT: vmovupd %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovapd 704(%rdi), %ymm4
-; AVX1-ONLY-NEXT: vmovapd 672(%rdi), %ymm5
+; AVX1-ONLY-NEXT: subq $1128, %rsp # imm = 0x468
+; AVX1-ONLY-NEXT: vmovapd 896(%rdi), %ymm0
+; AVX1-ONLY-NEXT: vmovapd 704(%rdi), %ymm1
+; AVX1-ONLY-NEXT: vmovapd 512(%rdi), %ymm2
+; AVX1-ONLY-NEXT: vmovapd 320(%rdi), %ymm3
+; AVX1-ONLY-NEXT: vmovapd 128(%rdi), %ymm4
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm5 = mem[0,1],ymm4[2,3]
; AVX1-ONLY-NEXT: vmovupd %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovapd 512(%rdi), %ymm6
-; AVX1-ONLY-NEXT: vmovapd 480(%rdi), %ymm7
+; AVX1-ONLY-NEXT: vmovupd %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm6 = mem[0,1],ymm3[2,3]
+; AVX1-ONLY-NEXT: vmovupd %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovupd %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm7 = mem[0,1],ymm2[2,3]
; AVX1-ONLY-NEXT: vmovupd %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovapd 320(%rdi), %ymm8
-; AVX1-ONLY-NEXT: vmovapd 288(%rdi), %ymm9
-; AVX1-ONLY-NEXT: vmovupd %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovapd 128(%rdi), %ymm1
+; AVX1-ONLY-NEXT: vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm8 = mem[0,1],ymm1[2,3]
+; AVX1-ONLY-NEXT: vmovupd %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovapd %ymm1, %ymm9
; AVX1-ONLY-NEXT: vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovapd 96(%rdi), %ymm0
-; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3]
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[1],ymm1[0],ymm0[3],ymm1[2]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm10 = mem[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT: vmovupd %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovapd %ymm0, %ymm11
+; AVX1-ONLY-NEXT: vmovupd %ymm0, (%rsp) # 32-byte Spill
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm0 = ymm5[1],ymm4[0],ymm5[3],ymm4[2]
; AVX1-ONLY-NEXT: vbroadcastsd 176(%rdi), %ymm1
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3]
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovupd %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm9[0,1],ymm8[2,3]
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[1],ymm8[0],ymm0[3],ymm8[2]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm0 = ymm6[1],ymm3[0],ymm6[3],ymm3[2]
; AVX1-ONLY-NEXT: vbroadcastsd 368(%rdi), %ymm1
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3]
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovupd %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm7[0,1],ymm6[2,3]
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[1],ymm6[0],ymm0[3],ymm6[2]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm0 = ymm7[1],ymm2[0],ymm7[3],ymm2[2]
; AVX1-ONLY-NEXT: vbroadcastsd 560(%rdi), %ymm1
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3]
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovupd %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm5[0,1],ymm4[2,3]
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[1],ymm4[0],ymm0[3],ymm4[2]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm0 = ymm8[1],ymm9[0],ymm8[3],ymm9[2]
; AVX1-ONLY-NEXT: vbroadcastsd 752(%rdi), %ymm1
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3]
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovupd %ymm2, (%rsp) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm3[0,1],ymm2[2,3]
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[1],ymm2[0],ymm0[3],ymm2[2]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm0 = ymm10[1],ymm11[0],ymm10[3],ymm11[2]
; AVX1-ONLY-NEXT: vbroadcastsd 944(%rdi), %ymm1
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3]
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovapd 1088(%rdi), %ymm6
-; AVX1-ONLY-NEXT: vmovapd 1056(%rdi), %ymm0
+; AVX1-ONLY-NEXT: vmovapd 1088(%rdi), %ymm11
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = mem[0,1],ymm11[2,3]
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm6[2,3]
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[1],ymm6[0],ymm0[3],ymm6[2]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[1],ymm11[0],ymm0[3],ymm11[2]
; AVX1-ONLY-NEXT: vbroadcastsd 1136(%rdi), %ymm1
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3]
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovapd 1280(%rdi), %ymm5
-; AVX1-ONLY-NEXT: vmovapd 1248(%rdi), %ymm0
+; AVX1-ONLY-NEXT: vmovapd 1280(%rdi), %ymm8
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = mem[0,1],ymm8[2,3]
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm5[2,3]
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[1],ymm5[0],ymm0[3],ymm5[2]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[1],ymm8[0],ymm0[3],ymm8[2]
; AVX1-ONLY-NEXT: vbroadcastsd 1328(%rdi), %ymm1
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3]
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovapd 1472(%rdi), %ymm4
-; AVX1-ONLY-NEXT: vmovapd 1440(%rdi), %ymm0
+; AVX1-ONLY-NEXT: vmovapd 1472(%rdi), %ymm5
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = mem[0,1],ymm5[2,3]
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm4[2,3]
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[1],ymm4[0],ymm0[3],ymm4[2]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[1],ymm5[0],ymm0[3],ymm5[2]
; AVX1-ONLY-NEXT: vbroadcastsd 1520(%rdi), %ymm1
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3]
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovapd 32(%rdi), %ymm1
-; AVX1-ONLY-NEXT: vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovapd (%rdi), %ymm0
+; AVX1-ONLY-NEXT: vmovapd 32(%rdi), %ymm0
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3]
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[1],ymm1[0],ymm0[3],ymm1[2]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = mem[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT: vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm0 = ymm1[1],ymm0[0],ymm1[3],ymm0[2]
; AVX1-ONLY-NEXT: vbroadcastsd 80(%rdi), %ymm1
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3]
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovapd 224(%rdi), %ymm11
-; AVX1-ONLY-NEXT: vmovapd 192(%rdi), %ymm0
+; AVX1-ONLY-NEXT: vmovapd 224(%rdi), %ymm10
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = mem[0,1],ymm10[2,3]
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm11[2,3]
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[1],ymm11[0],ymm0[3],ymm11[2]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[1],ymm10[0],ymm0[3],ymm10[2]
; AVX1-ONLY-NEXT: vbroadcastsd 272(%rdi), %ymm1
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3]
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovapd 416(%rdi), %ymm10
-; AVX1-ONLY-NEXT: vmovapd 384(%rdi), %ymm0
+; AVX1-ONLY-NEXT: vmovapd 416(%rdi), %ymm9
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = mem[0,1],ymm9[2,3]
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm10[2,3]
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[1],ymm10[0],ymm0[3],ymm10[2]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[1],ymm9[0],ymm0[3],ymm9[2]
; AVX1-ONLY-NEXT: vbroadcastsd 464(%rdi), %ymm1
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3]
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovapd 608(%rdi), %ymm8
-; AVX1-ONLY-NEXT: vmovapd 576(%rdi), %ymm0
+; AVX1-ONLY-NEXT: vmovapd 608(%rdi), %ymm7
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = mem[0,1],ymm7[2,3]
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm8[2,3]
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[1],ymm8[0],ymm0[3],ymm8[2]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[1],ymm7[0],ymm0[3],ymm7[2]
; AVX1-ONLY-NEXT: vbroadcastsd 656(%rdi), %ymm1
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3]
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovapd 800(%rdi), %ymm3
-; AVX1-ONLY-NEXT: vmovapd 768(%rdi), %ymm0
+; AVX1-ONLY-NEXT: vmovapd 800(%rdi), %ymm4
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm12 = mem[0,1],ymm4[2,3]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm0 = ymm12[1],ymm4[0],ymm12[3],ymm4[2]
+; AVX1-ONLY-NEXT: vbroadcastsd 848(%rdi), %ymm6
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1,2],ymm6[3]
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm3[2,3]
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[1],ymm3[0],ymm0[3],ymm3[2]
-; AVX1-ONLY-NEXT: vbroadcastsd 848(%rdi), %ymm1
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3]
-; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovapd 992(%rdi), %ymm2
-; AVX1-ONLY-NEXT: vmovapd 960(%rdi), %ymm12
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm12[0,1],ymm2[2,3]
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[1],ymm2[0],ymm0[3],ymm2[2]
-; AVX1-ONLY-NEXT: vbroadcastsd 1040(%rdi), %ymm9
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1,2],ymm9[3]
+; AVX1-ONLY-NEXT: vmovapd 992(%rdi), %ymm3
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm6 = mem[0,1],ymm3[2,3]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm0 = ymm6[1],ymm3[0],ymm6[3],ymm3[2]
+; AVX1-ONLY-NEXT: vbroadcastsd 1040(%rdi), %ymm15
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1,2],ymm15[3]
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovapd 1184(%rdi), %ymm1
-; AVX1-ONLY-NEXT: vmovapd 1152(%rdi), %ymm9
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm15 = ymm9[0,1],ymm1[2,3]
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm15 = ymm15[1],ymm1[0],ymm15[3],ymm1[2]
+; AVX1-ONLY-NEXT: vmovapd 1184(%rdi), %ymm2
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm15 = mem[0,1],ymm2[2,3]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm0 = ymm15[1],ymm2[0],ymm15[3],ymm2[2]
; AVX1-ONLY-NEXT: vbroadcastsd 1232(%rdi), %ymm14
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm15[0,1,2],ymm14[3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1,2],ymm14[3]
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovapd 1376(%rdi), %ymm14
-; AVX1-ONLY-NEXT: vmovapd 1344(%rdi), %ymm7
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm15 = ymm7[0,1],ymm14[2,3]
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm15 = ymm15[1],ymm14[0],ymm15[3],ymm14[2]
+; AVX1-ONLY-NEXT: vmovapd 1376(%rdi), %ymm0
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm14 = mem[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm1 = ymm14[1],ymm0[0],ymm14[3],ymm0[2]
; AVX1-ONLY-NEXT: vbroadcastsd 1424(%rdi), %ymm13
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm15[0,1,2],ymm13[3]
-; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vblendps $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm13 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm13 = mem[0,1,2,3],ymm0[4,5],mem[6,7]
-; AVX1-ONLY-NEXT: vmovaps 112(%rdi), %xmm15
-; AVX1-ONLY-NEXT: vinsertf128 $1, 160(%rdi), %ymm15, %ymm15
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0,1],ymm15[2,3],ymm13[4,5],ymm15[6,7]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm1[0,1,2],ymm13[3]
+; AVX1-ONLY-NEXT: vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 112(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vinsertf128 $1, 160(%rdi), %ymm1, %ymm1
+; AVX1-ONLY-NEXT: vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm13 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm13 = mem[0,1],ymm1[2,3],mem[4,5],ymm1[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm13 = ymm0[0,1,2,3],mem[4,5,6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm15[0,1],ymm13[2,3],ymm15[4,5],ymm13[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vblendps $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm13 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm13 = mem[0,1,2,3],ymm0[4,5],mem[6,7]
-; AVX1-ONLY-NEXT: vmovaps 304(%rdi), %xmm15
-; AVX1-ONLY-NEXT: vinsertf128 $1, 352(%rdi), %ymm15, %ymm15
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0,1],ymm15[2,3],ymm13[4,5],ymm15[6,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm13[2,3],ymm1[4,5],ymm13[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 304(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vinsertf128 $1, 352(%rdi), %ymm1, %ymm1
+; AVX1-ONLY-NEXT: vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm13 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm13 = mem[0,1],ymm1[2,3],mem[4,5],ymm1[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm13 = ymm0[0,1,2,3],mem[4,5,6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm15[0,1],ymm13[2,3],ymm15[4,5],ymm13[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vblendps $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm13 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm13 = mem[0,1,2,3],ymm0[4,5],mem[6,7]
-; AVX1-ONLY-NEXT: vmovaps 496(%rdi), %xmm15
-; AVX1-ONLY-NEXT: vinsertf128 $1, 544(%rdi), %ymm15, %ymm15
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0,1],ymm15[2,3],ymm13[4,5],ymm15[6,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm13[2,3],ymm1[4,5],ymm13[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 496(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vinsertf128 $1, 544(%rdi), %ymm1, %ymm1
+; AVX1-ONLY-NEXT: vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm13 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm13 = mem[0,1],ymm1[2,3],mem[4,5],ymm1[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm13 = ymm0[0,1,2,3],mem[4,5,6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm15[0,1],ymm13[2,3],ymm15[4,5],ymm13[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vblendps $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm13 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm13 = mem[0,1,2,3],ymm0[4,5],mem[6,7]
-; AVX1-ONLY-NEXT: vmovaps 688(%rdi), %xmm15
-; AVX1-ONLY-NEXT: vinsertf128 $1, 736(%rdi), %ymm15, %ymm15
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0,1],ymm15[2,3],ymm13[4,5],ymm15[6,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm13[2,3],ymm1[4,5],ymm13[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 688(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vinsertf128 $1, 736(%rdi), %ymm1, %ymm1
+; AVX1-ONLY-NEXT: vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm13 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm13 = mem[0,1],ymm1[2,3],mem[4,5],ymm1[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm13 = ymm0[0,1,2,3],mem[4,5,6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm15[0,1],ymm13[2,3],ymm15[4,5],ymm13[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups (%rsp), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vblendps $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm13 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm13 = mem[0,1,2,3],ymm0[4,5],mem[6,7]
-; AVX1-ONLY-NEXT: vmovaps 880(%rdi), %xmm15
-; AVX1-ONLY-NEXT: vinsertf128 $1, 928(%rdi), %ymm15, %ymm15
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0,1],ymm15[2,3],ymm13[4,5],ymm15[6,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm13[2,3],ymm1[4,5],ymm13[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 880(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vinsertf128 $1, 928(%rdi), %ymm1, %ymm1
+; AVX1-ONLY-NEXT: vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm13 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm13 = mem[0,1],ymm1[2,3],mem[4,5],ymm1[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm13 = ymm0[0,1,2,3],mem[4,5,6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm15[0,1],ymm13[2,3],ymm15[4,5],ymm13[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendpd $11, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm13 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm13 = mem[0,1],ymm6[2],mem[3]
-; AVX1-ONLY-NEXT: vmovaps 1072(%rdi), %xmm15
-; AVX1-ONLY-NEXT: vinsertf128 $1, 1120(%rdi), %ymm15, %ymm15
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm13[0],ymm15[1],ymm13[2],ymm15[3]
+; AVX1-ONLY-NEXT: vmovups (%rsp), %ymm13 # 32-byte Reload
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm13[2,3],ymm1[4,5],ymm13[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 1072(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vinsertf128 $1, 1120(%rdi), %ymm1, %ymm1
+; AVX1-ONLY-NEXT: vblendpd $5, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm13 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm13 = mem[0],ymm1[1],mem[2],ymm1[3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm11 = ymm11[0,1],mem[2,3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm1[0],ymm11[1],ymm1[2],ymm11[3]
+; AVX1-ONLY-NEXT: vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 1264(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vinsertf128 $1, 1312(%rdi), %ymm1, %ymm1
+; AVX1-ONLY-NEXT: vblendpd $5, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm11 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm11 = mem[0],ymm1[1],mem[2],ymm1[3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm8 = ymm8[0,1],mem[2,3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm1[0],ymm8[1],ymm1[2],ymm8[3]
+; AVX1-ONLY-NEXT: vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 1456(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vinsertf128 $1, 1504(%rdi), %ymm1, %ymm1
+; AVX1-ONLY-NEXT: vblendpd $5, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm8 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm8 = mem[0],ymm1[1],mem[2],ymm1[3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm5 = ymm5[0,1],mem[2,3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm1[0],ymm5[1],ymm1[2],ymm5[3]
+; AVX1-ONLY-NEXT: vmovupd %ymm1, (%rsp) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 1360(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vinsertf128 $1, 1408(%rdi), %ymm1, %ymm1
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm5 = ymm14[0],ymm1[1],ymm14[2],ymm1[3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],mem[2,3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[2],ymm0[3]
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm6 = ymm6[0,1],mem[2,3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm15[0],ymm6[1],ymm15[2],ymm6[3]
+; AVX1-ONLY-NEXT: vmovaps 1168(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vinsertf128 $1, 1216(%rdi), %ymm0, %ymm0
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm15 = ymm15[0],ymm0[1],ymm15[2],ymm0[3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm2[0,1],mem[2,3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3]
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendpd $11, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm13 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm13 = mem[0,1],ymm5[2],mem[3]
-; AVX1-ONLY-NEXT: vmovaps 1264(%rdi), %xmm15
-; AVX1-ONLY-NEXT: vinsertf128 $1, 1312(%rdi), %ymm15, %ymm15
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm13 = ymm13[0],ymm15[1],ymm13[2],ymm15[3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm5 = ymm5[0,1],mem[2,3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm15[0],ymm5[1],ymm15[2],ymm5[3]
+; AVX1-ONLY-NEXT: vmovaps 976(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vinsertf128 $1, 1024(%rdi), %ymm0, %ymm0
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm6 = ymm6[0],ymm0[1],ymm6[2],ymm0[3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm3[0,1],mem[2,3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3]
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendpd $11, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm15 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm15 = mem[0,1],ymm4[2],mem[3]
-; AVX1-ONLY-NEXT: vmovaps 1456(%rdi), %xmm6
-; AVX1-ONLY-NEXT: vinsertf128 $1, 1504(%rdi), %ymm6, %ymm6
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm15 = ymm15[0],ymm6[1],ymm15[2],ymm6[3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm4 = ymm4[0,1],mem[2,3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm6[0],ymm4[1],ymm6[2],ymm4[3]
-; AVX1-ONLY-NEXT: vmovupd %ymm0, (%rsp) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm6 = ymm7[0,1],ymm14[2],ymm7[3]
-; AVX1-ONLY-NEXT: vmovaps 1360(%rdi), %xmm7
-; AVX1-ONLY-NEXT: vinsertf128 $1, 1408(%rdi), %ymm7, %ymm7
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm6 = ymm6[0],ymm7[1],ymm6[2],ymm7[3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm14 = ymm14[0,1],mem[2,3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm7[0],ymm14[1],ymm7[2],ymm14[3]
+; AVX1-ONLY-NEXT: vmovaps 784(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vinsertf128 $1, 832(%rdi), %ymm0, %ymm0
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm12 = ymm12[0],ymm0[1],ymm12[2],ymm0[3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm4[0,1],mem[2,3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3]
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm7 = ymm9[0,1],ymm1[2],ymm9[3]
-; AVX1-ONLY-NEXT: vmovaps 1168(%rdi), %xmm9
-; AVX1-ONLY-NEXT: vinsertf128 $1, 1216(%rdi), %ymm9, %ymm9
+; AVX1-ONLY-NEXT: vmovaps 592(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vinsertf128 $1, 640(%rdi), %ymm1, %ymm1
+; AVX1-ONLY-NEXT: vblendpd $5, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm4 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm4 = mem[0],ymm1[1],mem[2],ymm1[3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm7 = ymm7[0,1],mem[2,3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm3 = ymm1[0],ymm7[1],ymm1[2],ymm7[3]
+; AVX1-ONLY-NEXT: vmovaps 400(%rdi), %xmm7
+; AVX1-ONLY-NEXT: vinsertf128 $1, 448(%rdi), %ymm7, %ymm7
+; AVX1-ONLY-NEXT: vblendpd $5, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm2 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm2 = mem[0],ymm7[1],mem[2],ymm7[3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm9 = ymm9[0,1],mem[2,3]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm7 = ymm7[0],ymm9[1],ymm7[2],ymm9[3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm1[0,1],mem[2,3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm9[0],ymm0[1],ymm9[2],ymm0[3]
-; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm12[0,1],ymm2[2],ymm12[3]
-; AVX1-ONLY-NEXT: vmovaps 976(%rdi), %xmm12
-; AVX1-ONLY-NEXT: vinsertf128 $1, 1024(%rdi), %ymm12, %ymm12
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0],ymm12[1],ymm0[2],ymm12[3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm2[0,1],mem[2,3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm12[0],ymm1[1],ymm12[2],ymm1[3]
-; AVX1-ONLY-NEXT: vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendpd $11, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm1 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm1 = mem[0,1],ymm3[2],mem[3]
-; AVX1-ONLY-NEXT: vmovaps 784(%rdi), %xmm5
-; AVX1-ONLY-NEXT: vinsertf128 $1, 832(%rdi), %ymm5, %ymm5
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm1[0],ymm5[1],ymm1[2],ymm5[3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm3[0,1],mem[2,3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm5[0],ymm2[1],ymm5[2],ymm2[3]
-; AVX1-ONLY-NEXT: vblendpd $11, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm5 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm5 = mem[0,1],ymm8[2],mem[3]
-; AVX1-ONLY-NEXT: vmovaps 592(%rdi), %xmm4
-; AVX1-ONLY-NEXT: vinsertf128 $1, 640(%rdi), %ymm4, %ymm4
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm5 = ymm5[0],ymm4[1],ymm5[2],ymm4[3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm3 = ymm8[0,1],mem[2,3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm3 = ymm4[0],ymm3[1],ymm4[2],ymm3[3]
-; AVX1-ONLY-NEXT: vblendpd $11, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm4 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm4 = mem[0,1],ymm10[2],mem[3]
-; AVX1-ONLY-NEXT: vmovaps 400(%rdi), %xmm14
-; AVX1-ONLY-NEXT: vinsertf128 $1, 448(%rdi), %ymm14, %ymm14
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm4 = ymm4[0],ymm14[1],ymm4[2],ymm14[3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm8 = ymm10[0,1],mem[2,3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm8 = ymm14[0],ymm8[1],ymm14[2],ymm8[3]
-; AVX1-ONLY-NEXT: vblendpd $11, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm14 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm14 = mem[0,1],ymm11[2],mem[3]
; AVX1-ONLY-NEXT: vmovaps 208(%rdi), %xmm9
; AVX1-ONLY-NEXT: vinsertf128 $1, 256(%rdi), %ymm9, %ymm9
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm14 = ymm14[0],ymm9[1],ymm14[2],ymm9[3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm10 = ymm11[0,1],mem[2,3]
+; AVX1-ONLY-NEXT: vblendpd $5, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm1 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm1 = mem[0],ymm9[1],mem[2],ymm9[3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm10 = ymm10[0,1],mem[2,3]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm9 = ymm9[0],ymm10[1],ymm9[2],ymm10[3]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
-; AVX1-ONLY-NEXT: vblendps $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm10 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm10 = mem[0,1,2,3],ymm11[4,5],mem[6,7]
-; AVX1-ONLY-NEXT: vmovaps 16(%rdi), %xmm12
-; AVX1-ONLY-NEXT: vinsertf128 $1, 64(%rdi), %ymm12, %ymm12
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm10 = ymm10[0,1],ymm12[2,3],ymm10[4,5],ymm12[6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm11 = ymm11[0,1,2,3],mem[4,5,6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm11 = ymm12[0,1],ymm11[2,3],ymm12[4,5],ymm11[6,7]
-; AVX1-ONLY-NEXT: vmovapd %ymm6, 448(%rsi)
-; AVX1-ONLY-NEXT: vmovapd %ymm7, 384(%rsi)
-; AVX1-ONLY-NEXT: vmovapd %ymm0, 320(%rsi)
-; AVX1-ONLY-NEXT: vmovapd %ymm1, 256(%rsi)
-; AVX1-ONLY-NEXT: vmovapd %ymm5, 192(%rsi)
-; AVX1-ONLY-NEXT: vmovapd %ymm4, 128(%rsi)
-; AVX1-ONLY-NEXT: vmovapd %ymm14, 64(%rsi)
-; AVX1-ONLY-NEXT: vmovaps %ymm10, (%rsi)
-; AVX1-ONLY-NEXT: vmovapd %ymm15, 480(%rsi)
-; AVX1-ONLY-NEXT: vmovapd %ymm13, 416(%rsi)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm0, 352(%rsi)
+; AVX1-ONLY-NEXT: vmovaps 16(%rdi), %xmm10
+; AVX1-ONLY-NEXT: vinsertf128 $1, 64(%rdi), %ymm10, %ymm10
+; AVX1-ONLY-NEXT: vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm0 = mem[0,1],ymm10[2,3],mem[4,5],ymm10[6,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm10 = ymm10[0,1],ymm14[2,3],ymm10[4,5],ymm14[6,7]
+; AVX1-ONLY-NEXT: vmovapd %ymm5, 448(%rsi)
+; AVX1-ONLY-NEXT: vmovapd %ymm15, 384(%rsi)
+; AVX1-ONLY-NEXT: vmovapd %ymm6, 320(%rsi)
+; AVX1-ONLY-NEXT: vmovapd %ymm12, 256(%rsi)
+; AVX1-ONLY-NEXT: vmovapd %ymm4, 192(%rsi)
+; AVX1-ONLY-NEXT: vmovapd %ymm2, 128(%rsi)
+; AVX1-ONLY-NEXT: vmovapd %ymm1, 64(%rsi)
+; AVX1-ONLY-NEXT: vmovaps %ymm0, (%rsi)
+; AVX1-ONLY-NEXT: vmovapd %ymm8, 480(%rsi)
+; AVX1-ONLY-NEXT: vmovapd %ymm11, 416(%rsi)
+; AVX1-ONLY-NEXT: vmovapd %ymm13, 352(%rsi)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm0, 288(%rsi)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
@@ -2088,11 +2025,12 @@ define void @load_i64_stride3_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: vmovaps %ymm0, 96(%rdx)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm0, 32(%rdx)
-; AVX1-ONLY-NEXT: vmovaps %ymm11, (%rcx)
+; AVX1-ONLY-NEXT: vmovaps %ymm10, (%rcx)
; AVX1-ONLY-NEXT: vmovapd %ymm9, 64(%rcx)
-; AVX1-ONLY-NEXT: vmovapd %ymm8, 128(%rcx)
+; AVX1-ONLY-NEXT: vmovapd %ymm7, 128(%rcx)
; AVX1-ONLY-NEXT: vmovapd %ymm3, 192(%rcx)
-; AVX1-ONLY-NEXT: vmovapd %ymm2, 256(%rcx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm0, 256(%rcx)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm0, 320(%rcx)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
@@ -2115,7 +2053,7 @@ define void @load_i64_stride3_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX1-ONLY-NEXT: vmovaps %ymm0, 96(%rcx)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm0, 32(%rcx)
-; AVX1-ONLY-NEXT: addq $1160, %rsp # imm = 0x488
+; AVX1-ONLY-NEXT: addq $1128, %rsp # imm = 0x468
; AVX1-ONLY-NEXT: vzeroupper
; AVX1-ONLY-NEXT: retq
;
@@ -2144,34 +2082,34 @@ define void @load_i64_stride3_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vinsertf128 $1, 160(%rdi), %ymm0, %ymm0
; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,3,2,3]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm10[4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm10[4,5,6,7]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vinsertf128 $1, 352(%rdi), %ymm0, %ymm0
; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm1 = ymm9[0,3,2,3]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm8[4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm8[4,5,6,7]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vinsertf128 $1, 544(%rdi), %ymm0, %ymm0
; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm1 = ymm7[0,3,2,3]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm6[4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm6[4,5,6,7]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vinsertf128 $1, 736(%rdi), %ymm0, %ymm0
; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm1 = ymm5[0,3,2,3]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm4[4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm4[4,5,6,7]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vinsertf128 $1, 928(%rdi), %ymm0, %ymm0
; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm1 = ymm3[0,3,2,3]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovaps 1088(%rdi), %ymm1
; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovaps 1056(%rdi), %ymm10
; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm0 = ymm10[0,3,2,3]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT: vinsertf128 $1, 1120(%rdi), %ymm0, %ymm1
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -2179,7 +2117,7 @@ define void @load_i64_stride3_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovaps 1248(%rdi), %ymm9
; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm0 = ymm9[0,3,2,3]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT: vinsertf128 $1, 1312(%rdi), %ymm0, %ymm1
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -2187,7 +2125,7 @@ define void @load_i64_stride3_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovaps 1440(%rdi), %ymm8
; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm0 = ymm8[0,3,2,3]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT: vinsertf128 $1, 1504(%rdi), %ymm0, %ymm1
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -2195,7 +2133,7 @@ define void @load_i64_stride3_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovaps (%rdi), %ymm7
; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm0 = ymm7[0,3,2,3]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT: vinsertf128 $1, 64(%rdi), %ymm0, %ymm1
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -2203,14 +2141,14 @@ define void @load_i64_stride3_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovaps 192(%rdi), %ymm6
; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm0 = ymm6[0,3,2,3]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT: vinsertf128 $1, 256(%rdi), %ymm0, %ymm1
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovaps 416(%rdi), %ymm14
; AVX2-ONLY-NEXT: vmovaps 384(%rdi), %ymm5
; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm0 = ymm5[0,3,2,3]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm14[4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm14[4,5,6,7]
; AVX2-ONLY-NEXT: vinsertf128 $1, 448(%rdi), %ymm0, %ymm1
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -2218,7 +2156,7 @@ define void @load_i64_stride3_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-ONLY-NEXT: vmovups %ymm1, (%rsp) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovaps 576(%rdi), %ymm4
; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm0 = ymm4[0,3,2,3]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT: vinsertf128 $1, 640(%rdi), %ymm0, %ymm1
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -2226,21 +2164,21 @@ define void @load_i64_stride3_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovaps 768(%rdi), %ymm3
; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm0 = ymm3[0,3,2,3]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT: vinsertf128 $1, 832(%rdi), %ymm0, %ymm1
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovaps 992(%rdi), %ymm15
; AVX2-ONLY-NEXT: vmovaps 960(%rdi), %ymm2
; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm0 = ymm2[0,3,2,3]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm15[4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm15[4,5,6,7]
; AVX2-ONLY-NEXT: vinsertf128 $1, 1024(%rdi), %ymm0, %ymm1
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovaps 1184(%rdi), %ymm13
; AVX2-ONLY-NEXT: vmovaps 1152(%rdi), %ymm1
; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm0 = ymm1[0,3,2,3]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm13[4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm13[4,5,6,7]
; AVX2-ONLY-NEXT: vinsertf128 $1, 1216(%rdi), %ymm0, %ymm11
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm11[6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -2248,7 +2186,7 @@ define void @load_i64_stride3_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-ONLY-NEXT: vmovups %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovaps 1344(%rdi), %ymm0
; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm11 = ymm0[0,3,2,3]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm11 = ymm11[0,1,2,3],ymm12[4,5],ymm11[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm11 = ymm11[0,1,2,3],ymm12[4,5,6,7]
; AVX2-ONLY-NEXT: vinsertf128 $1, 1408(%rdi), %ymm0, %ymm12
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm11 = ymm11[0,1,2,3,4,5],ymm12[6,7]
; AVX2-ONLY-NEXT: vmovups %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-store-i32-stride-7.ll b/llvm/test/CodeGen/X86/vector-interleaved-store-i32-stride-7.ll
index 910c59fc09099..dc57a11304d0a 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-store-i32-stride-7.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-store-i32-stride-7.ll
@@ -812,7 +812,7 @@ define void @store_i32_stride7_vf8(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm9 = ymm9[2,3,2,3]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm12 = ymm0[3,3],ymm1[3,3],ymm0[7,7],ymm1[7,7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm12, %xmm12
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm9 = ymm12[0,1,2],ymm9[3,4],ymm12[5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm9 = ymm12[0,1,2],ymm9[3,4,5,6,7]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm7 = ymm7[3,3],ymm8[3,3],ymm7[7,7],ymm8[7,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm9[0,1,2,3,4],ymm7[5,6],ymm9[7]
; AVX1-ONLY-NEXT: vbroadcastsd 24(%rax), %ymm8
@@ -825,7 +825,7 @@ define void @store_i32_stride7_vf8(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm9 = xmm6[2,2,2,2]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm9 = xmm5[0,1,2],xmm9[3]
; AVX1-ONLY-NEXT: vbroadcastsd 8(%rax), %ymm11
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm9 = ymm9[0,1,2,3],ymm11[4,5],ymm9[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm9 = ymm9[0,1,2,3],ymm11[4,5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1],ymm9[2,3,4],ymm8[5,6,7]
; AVX1-ONLY-NEXT: vunpcklps {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm2[0],ymm3[0],ymm2[2],ymm3[2]
@@ -879,7 +879,7 @@ define void @store_i32_stride7_vf8(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
; AVX2-SLOW-NEXT: vunpckhps {{.*#+}} ymm10 = ymm8[2],ymm3[2],ymm8[3],ymm3[3],ymm8[6],ymm3[6],ymm8[7],ymm3[7]
; AVX2-SLOW-NEXT: vunpckhps {{.*#+}} ymm11 = ymm0[2],ymm2[2],ymm0[3],ymm2[3],ymm0[6],ymm2[6],ymm0[7],ymm2[7]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm11 = ymm11[2,2,2,2]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm10 = ymm11[0,1,2,3],ymm10[4,5],ymm11[6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm10 = ymm11[0,1,2,3],ymm10[4,5,6,7]
; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm11 = ymm7[1,1,2,2,5,5,6,6]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm11 = ymm11[0,1],ymm6[2],ymm11[3,4,5],ymm6[6],ymm11[7]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm11 = ymm11[2,1,2,3]
@@ -896,7 +896,7 @@ define void @store_i32_stride7_vf8(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
; AVX2-SLOW-NEXT: vpermilps {{.*#+}} xmm11 = xmm5[2,2,2,2]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm11 = xmm4[0,1,2],xmm11[3]
; AVX2-SLOW-NEXT: vbroadcastsd 8(%rax), %ymm10
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm10 = ymm11[0,1,2,3],ymm10[4,5],ymm11[6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm10 = ymm11[0,1,2,3],ymm10[4,5,6,7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm11 = ymm9[0,1],ymm10[2,3,4],ymm9[5,6,7]
; AVX2-SLOW-NEXT: vbroadcastss %xmm12, %xmm9
; AVX2-SLOW-NEXT: vbroadcastss %xmm13, %xmm10
@@ -926,7 +926,7 @@ define void @store_i32_stride7_vf8(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm8 = ymm8[3,3,3,3]
; AVX2-SLOW-NEXT: vunpckhps {{.*#+}} ymm10 = ymm2[2],ymm0[2],ymm2[3],ymm0[3],ymm2[6],ymm0[6],ymm2[7],ymm0[7]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm10 = ymm10[3,3,3,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm8 = ymm10[0,1,2],ymm8[3,4],ymm10[5,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm8 = ymm10[0,1,2],ymm8[3,4,5,6,7]
; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm6 = ymm6[3,3],ymm7[3,3],ymm6[7,7],ymm7[7,7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm6 = ymm8[0,1,2,3,4],ymm6[5,6],ymm8[7]
; AVX2-SLOW-NEXT: vbroadcastsd 24(%rax), %ymm7
@@ -984,7 +984,7 @@ define void @store_i32_stride7_vf8(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
; AVX2-FAST-NEXT: vunpckhps {{.*#+}} ymm6 = ymm10[2],ymm3[2],ymm10[3],ymm3[3],ymm10[6],ymm3[6],ymm10[7],ymm3[7]
; AVX2-FAST-NEXT: vunpckhps {{.*#+}} ymm9 = ymm1[2],ymm2[2],ymm1[3],ymm2[3],ymm1[6],ymm2[6],ymm1[7],ymm2[7]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm9 = ymm9[2,2,2,2]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm6 = ymm9[0,1,2,3],ymm6[4,5],ymm9[6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm6 = ymm9[0,1,2,3],ymm6[4,5,6,7]
; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm9 = ymm8[1,1,2,2,5,5,6,6]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm9 = ymm9[0,1],ymm7[2],ymm9[3,4,5],ymm7[6],ymm9[7]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm9 = ymm9[2,1,2,3]
@@ -1015,13 +1015,13 @@ define void @store_i32_stride7_vf8(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
; AVX2-FAST-NEXT: vpermilps {{.*#+}} xmm9 = xmm5[2,2,2,2]
; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm9 = xmm4[0,1,2],xmm9[3]
; AVX2-FAST-NEXT: vbroadcastsd 8(%rax), %ymm15
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm9 = ymm9[0,1,2,3],ymm15[4,5],ymm9[6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm9 = ymm9[0,1,2,3],ymm15[4,5,6,7]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1],ymm9[2,3,4],ymm6[5,6,7]
; AVX2-FAST-NEXT: vunpckhps {{.*#+}} ymm9 = ymm3[2],ymm10[2],ymm3[3],ymm10[3],ymm3[6],ymm10[6],ymm3[7],ymm10[7]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm9 = ymm9[3,3,3,3]
; AVX2-FAST-NEXT: vunpckhps {{.*#+}} ymm10 = ymm2[2],ymm1[2],ymm2[3],ymm1[3],ymm2[6],ymm1[6],ymm2[7],ymm1[7]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm10 = ymm10[3,3,3,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm9 = ymm10[0,1,2],ymm9[3,4],ymm10[5,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm9 = ymm10[0,1,2],ymm9[3,4,5,6,7]
; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm7 = ymm7[3,3],ymm8[3,3],ymm7[7,7],ymm8[7,7]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm7 = ymm9[0,1,2,3,4],ymm7[5,6],ymm9[7]
; AVX2-FAST-NEXT: vbroadcastsd 24(%rax), %ymm8
@@ -1091,7 +1091,7 @@ define void @store_i32_stride7_vf8(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
; AVX2-FAST-PERLANE-NEXT: vunpckhps {{.*#+}} ymm10 = ymm8[2],ymm3[2],ymm8[3],ymm3[3],ymm8[6],ymm3[6],ymm8[7],ymm3[7]
; AVX2-FAST-PERLANE-NEXT: vunpckhps {{.*#+}} ymm11 = ymm0[2],ymm2[2],ymm0[3],ymm2[3],ymm0[6],ymm2[6],ymm0[7],ymm2[7]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm11 = ymm11[2,2,2,2]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm10 = ymm11[0,1,2,3],ymm10[4,5],ymm11[6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm10 = ymm11[0,1,2,3],ymm10[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm11 = ymm7[1,1,2,2,5,5,6,6]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm11 = ymm11[0,1],ymm6[2],ymm11[3,4,5],ymm6[6],ymm11[7]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm11 = ymm11[2,1,2,3]
@@ -1108,7 +1108,7 @@ define void @store_i32_stride7_vf8(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} xmm11 = xmm5[2,2,2,2]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm11 = xmm4[0,1,2],xmm11[3]
; AVX2-FAST-PERLANE-NEXT: vbroadcastsd 8(%rax), %ymm10
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm10 = ymm11[0,1,2,3],ymm10[4,5],ymm11[6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm10 = ymm11[0,1,2,3],ymm10[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm11 = ymm9[0,1],ymm10[2,3,4],ymm9[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vbroadcastss %xmm12, %xmm9
; AVX2-FAST-PERLANE-NEXT: vbroadcastss %xmm13, %xmm10
@@ -1138,7 +1138,7 @@ define void @store_i32_stride7_vf8(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm8 = ymm8[3,3,3,3]
; AVX2-FAST-PERLANE-NEXT: vunpckhps {{.*#+}} ymm10 = ymm2[2],ymm0[2],ymm2[3],ymm0[3],ymm2[6],ymm0[6],ymm2[7],ymm0[7]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm10 = ymm10[3,3,3,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm8 = ymm10[0,1,2],ymm8[3,4],ymm10[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm8 = ymm10[0,1,2],ymm8[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm6 = ymm6[3,3],ymm7[3,3],ymm6[7,7],ymm7[7,7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm6 = ymm8[0,1,2,3,4],ymm6[5,6],ymm8[7]
; AVX2-FAST-PERLANE-NEXT: vbroadcastsd 24(%rax), %ymm7
@@ -1710,7 +1710,7 @@ define void @store_i32_stride7_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vshufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm3 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm3 = ymm13[3,3],mem[3,3],ymm13[7,7],mem[7,7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm3, %xmm3
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm3[0,1,2],ymm0[3,4],ymm3[5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm3[0,1,2],ymm0[3,4,5,6,7]
; AVX1-ONLY-NEXT: vbroadcastss 60(%r8), %ymm3
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm3[5],ymm0[6,7]
; AVX1-ONLY-NEXT: vbroadcastss 60(%r9), %ymm3
@@ -1722,7 +1722,7 @@ define void @store_i32_stride7_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,2,3]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm10[3,3],ymm8[3,3],ymm10[7,7],ymm8[7,7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm3, %xmm3
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm3[0,1,2],ymm0[3,4],ymm3[5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm3[0,1,2],ymm0[3,4,5,6,7]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm2[3,3],ymm1[3,3],ymm2[7,7],ymm1[7,7]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm4[2,3],ymm1[1,2],ymm4[6,7],ymm1[5,6]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm1[2,3,2,3]
@@ -1743,7 +1743,7 @@ define void @store_i32_stride7_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vmovaps (%rsp), %xmm0 # 16-byte Reload
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm0[0,1,2],xmm1[3]
; AVX1-ONLY-NEXT: vbroadcastsd 40(%rax), %ymm2
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm3[0,1],ymm1[2,3,4],ymm3[5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm12[3,3],xmm11[3,3]
@@ -1755,7 +1755,7 @@ define void @store_i32_stride7_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm2 = xmm15[2,2,2,2]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm14[0,1,2],xmm2[3]
; AVX1-ONLY-NEXT: vbroadcastsd 8(%rax), %ymm3
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2,3,4],ymm1[5,6,7]
; AVX1-ONLY-NEXT: vunpcklps {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm2 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm2 = ymm13[0],mem[0],ymm13[1],mem[1],ymm13[4],mem[4],ymm13[5],mem[5]
@@ -1866,7 +1866,7 @@ define void @store_i32_stride7_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovaps (%rcx), %ymm13
; AVX2-SLOW-NEXT: vunpckhps {{.*#+}} ymm6 = ymm0[2],ymm13[2],ymm0[3],ymm13[3],ymm0[6],ymm13[6],ymm0[7],ymm13[7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm6[4,5],ymm4[6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm6[4,5,6,7]
; AVX2-SLOW-NEXT: vmovaps (%r8), %ymm0
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovaps (%r9), %ymm1
@@ -1885,7 +1885,7 @@ define void @store_i32_stride7_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-SLOW-NEXT: vmovaps 32(%rdx), %ymm4
; AVX2-SLOW-NEXT: vmovaps 32(%rcx), %ymm6
; AVX2-SLOW-NEXT: vunpckhps {{.*#+}} ymm0 = ymm4[2],ymm6[2],ymm4[3],ymm6[3],ymm4[6],ymm6[6],ymm4[7],ymm6[7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX2-SLOW-NEXT: vmovaps 32(%r8), %ymm2
; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6,7]
@@ -1901,7 +1901,7 @@ define void @store_i32_stride7_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[3,3,3,3]
; AVX2-SLOW-NEXT: vunpckhps {{.*#+}} ymm1 = ymm15[2],ymm14[2],ymm15[3],ymm14[3],ymm15[6],ymm14[6],ymm15[7],ymm14[7]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[3,3,3,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4],ymm1[5,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
; AVX2-SLOW-NEXT: vbroadcastss 60(%r8), %ymm1
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5],ymm0[6,7]
; AVX2-SLOW-NEXT: vbroadcastss 60(%r9), %ymm1
@@ -1920,7 +1920,7 @@ define void @store_i32_stride7_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-SLOW-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm1 = xmm0[0,1,2],xmm1[3]
; AVX2-SLOW-NEXT: vbroadcastsd 40(%rax), %ymm10
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm10[4,5],ymm1[6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm10[4,5,6,7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm3[0,1],ymm1[2,3,4],ymm3[5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vbroadcastss %xmm11, %xmm3
@@ -1950,7 +1950,7 @@ define void @store_i32_stride7_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-SLOW-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm1 = xmm8[0,1,2],xmm1[3]
; AVX2-SLOW-NEXT: vbroadcastsd 8(%rax), %ymm2
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4],ymm0[5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vbroadcastss %xmm9, %xmm0
@@ -2005,7 +2005,7 @@ define void @store_i32_stride7_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-SLOW-NEXT: vmovaps %ymm12, %ymm2
; AVX2-SLOW-NEXT: vmovaps %ymm5, %ymm12
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm4 = ymm4[3,3,3,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3,4],ymm4[5,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3,4,5,6,7]
; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm4 = ymm0[3,3],ymm1[3,3],ymm0[7,7],ymm1[7,7]
; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm5 = mem[2,3,2,3,6,7,6,7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm5[0],ymm4[1,2],ymm5[3,4],ymm4[5,6],ymm5[7]
@@ -2121,7 +2121,7 @@ define void @store_i32_stride7_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-FAST-NEXT: vmovaps (%rcx), %ymm2
; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vunpckhps {{.*#+}} ymm1 = ymm1[2],ymm2[2],ymm1[3],ymm2[3],ymm1[6],ymm2[6],ymm1[7],ymm2[7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FAST-NEXT: vmovaps (%r8), %ymm2
; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovaps (%r9), %ymm1
@@ -2140,7 +2140,7 @@ define void @store_i32_stride7_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-FAST-NEXT: vmovaps 32(%rdx), %ymm2
; AVX2-FAST-NEXT: vmovaps 32(%rcx), %ymm11
; AVX2-FAST-NEXT: vunpckhps {{.*#+}} ymm0 = ymm2[2],ymm11[2],ymm2[3],ymm11[3],ymm2[6],ymm11[6],ymm2[7],ymm11[7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm0[4,5],ymm4[6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm0[4,5,6,7]
; AVX2-FAST-NEXT: vmovaps 32(%r8), %ymm10
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm4[0,1,2,3,4,5],ymm10[6,7]
; AVX2-FAST-NEXT: vmovaps 32(%r9), %ymm4
@@ -2154,7 +2154,7 @@ define void @store_i32_stride7_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[3,3,3,3]
; AVX2-FAST-NEXT: vunpckhps {{.*#+}} ymm1 = ymm12[2],ymm13[2],ymm12[3],ymm13[3],ymm12[6],ymm13[6],ymm12[7],ymm13[7]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[3,3,3,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4],ymm1[5,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
; AVX2-FAST-NEXT: vbroadcastss 60(%r8), %ymm1
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5],ymm0[6,7]
; AVX2-FAST-NEXT: vbroadcastss 60(%r9), %ymm1
@@ -2175,7 +2175,7 @@ define void @store_i32_stride7_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-FAST-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm15 = xmm8[0,1,2],xmm15[3]
; AVX2-FAST-NEXT: vbroadcastsd 40(%rax), %ymm14
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm14 = ymm15[0,1,2,3],ymm14[4,5],ymm15[6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm14 = ymm15[0,1,2,3],ymm14[4,5,6,7]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm14[2,3,4],ymm1[5,6,7]
; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vbroadcastss %xmm9, %xmm1
@@ -2218,7 +2218,7 @@ define void @store_i32_stride7_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-FAST-NEXT: vmovaps (%rsp), %xmm14 # 16-byte Reload
; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm3 = xmm14[0,1,2],xmm3[3]
; AVX2-FAST-NEXT: vbroadcastsd 8(%rax), %ymm4
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm4[4,5],ymm3[6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm4[4,5,6,7]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm3[2,3,4],ymm1[5,6,7]
; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vbroadcastss %xmm0, %xmm3
@@ -2259,7 +2259,7 @@ define void @store_i32_stride7_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-FAST-NEXT: vmovaps %ymm0, %ymm1
; AVX2-FAST-NEXT: vmovaps %ymm6, %ymm0
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm5 = ymm5[3,3,3,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm5[0,1,2],ymm4[3,4],ymm5[5,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm5[0,1,2],ymm4[3,4,5,6,7]
; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm5 = ymm2[3,3],ymm7[3,3],ymm2[7,7],ymm7[7,7]
; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm6 = mem[2,3,2,3,6,7,6,7]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm5 = ymm6[0],ymm5[1,2],ymm6[3,4],ymm5[5,6],ymm6[7]
@@ -2372,7 +2372,7 @@ define void @store_i32_stride7_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovaps (%rcx), %ymm13
; AVX2-FAST-PERLANE-NEXT: vunpckhps {{.*#+}} ymm6 = ymm0[2],ymm13[2],ymm0[3],ymm13[3],ymm0[6],ymm13[6],ymm0[7],ymm13[7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm6[4,5],ymm4[6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm6[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovaps (%r8), %ymm0
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovaps (%r9), %ymm1
@@ -2391,7 +2391,7 @@ define void @store_i32_stride7_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-FAST-PERLANE-NEXT: vmovaps 32(%rdx), %ymm4
; AVX2-FAST-PERLANE-NEXT: vmovaps 32(%rcx), %ymm6
; AVX2-FAST-PERLANE-NEXT: vunpckhps {{.*#+}} ymm0 = ymm4[2],ymm6[2],ymm4[3],ymm6[3],ymm4[6],ymm6[6],ymm4[7],ymm6[7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovaps 32(%r8), %ymm2
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6,7]
@@ -2407,7 +2407,7 @@ define void @store_i32_stride7_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[3,3,3,3]
; AVX2-FAST-PERLANE-NEXT: vunpckhps {{.*#+}} ymm1 = ymm15[2],ymm14[2],ymm15[3],ymm14[3],ymm15[6],ymm14[6],ymm15[7],ymm14[7]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[3,3,3,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4],ymm1[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vbroadcastss 60(%r8), %ymm1
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5],ymm0[6,7]
; AVX2-FAST-PERLANE-NEXT: vbroadcastss 60(%r9), %ymm1
@@ -2426,7 +2426,7 @@ define void @store_i32_stride7_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-FAST-PERLANE-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm1 = xmm0[0,1,2],xmm1[3]
; AVX2-FAST-PERLANE-NEXT: vbroadcastsd 40(%rax), %ymm10
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm10[4,5],ymm1[6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm10[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm3[0,1],ymm1[2,3,4],ymm3[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vbroadcastss %xmm11, %xmm3
@@ -2456,7 +2456,7 @@ define void @store_i32_stride7_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-FAST-PERLANE-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm1 = xmm8[0,1,2],xmm1[3]
; AVX2-FAST-PERLANE-NEXT: vbroadcastsd 8(%rax), %ymm2
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4],ymm0[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vbroadcastss %xmm9, %xmm0
@@ -2511,7 +2511,7 @@ define void @store_i32_stride7_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm12, %ymm2
; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm5, %ymm12
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm4 = ymm4[3,3,3,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3,4],ymm4[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm4 = ymm0[3,3],ymm1[3,3],ymm0[7,7],ymm1[7,7]
; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm5 = mem[2,3,2,3,6,7,6,7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm5[0],ymm4[1,2],ymm5[3,4],ymm4[5,6],ymm5[7]
@@ -3727,7 +3727,7 @@ define void @store_i32_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm6[3,3],ymm14[3,3],ymm6[7,7],ymm14[7,7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm1, %xmm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4],ymm1[5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
; AVX1-ONLY-NEXT: vbroadcastss 124(%r8), %ymm1
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5],ymm0[6,7]
; AVX1-ONLY-NEXT: vbroadcastss 124(%r9), %ymm1
@@ -3771,7 +3771,7 @@ define void @store_i32_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vblendps $7, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm1 = mem[0,1,2],xmm1[3]
; AVX1-ONLY-NEXT: vbroadcastsd 8(%rax), %ymm14
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm14[4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm14[4,5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4],ymm0[5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm3[3,3],ymm4[3,3],ymm3[7,7],ymm4[7,7]
@@ -3780,7 +3780,7 @@ define void @store_i32_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vshufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm1 = ymm1[3,3],mem[3,3],ymm1[7,7],mem[7,7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm1, %xmm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4],ymm1[5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX1-ONLY-NEXT: vshufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm1 = ymm1[3,3],mem[3,3],ymm1[7,7],mem[7,7]
@@ -3804,7 +3804,7 @@ define void @store_i32_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm14[0,1,2],xmm1[3]
; AVX1-ONLY-NEXT: vbroadcastsd 40(%rax), %ymm13
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm13[4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm13[4,5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4],ymm0[5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm5[3,3],ymm8[3,3],ymm5[7,7],ymm8[7,7]
@@ -3813,7 +3813,7 @@ define void @store_i32_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm13 = ymm7[3,3],ymm0[3,3],ymm7[7,7],ymm0[7,7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm13, %xmm13
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm13[0,1,2],ymm1[3,4],ymm13[5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm13[0,1,2],ymm1[3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm11 = ymm2[3,3],ymm11[3,3],ymm2[7,7],ymm11[7,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
@@ -3836,7 +3836,7 @@ define void @store_i32_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm9 = xmm13[0,1,2],xmm9[3]
; AVX1-ONLY-NEXT: vbroadcastsd 72(%rax), %ymm10
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm9 = ymm9[0,1,2,3],ymm10[4,5],ymm9[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm9 = ymm9[0,1,2,3],ymm10[4,5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm8[0,1],ymm9[2,3,4],ymm8[5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
@@ -3847,7 +3847,7 @@ define void @store_i32_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm10 = ymm6[3,3],ymm1[3,3],ymm6[7,7],ymm1[7,7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm10, %xmm10
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm9 = ymm10[0,1,2],ymm9[3,4],ymm10[5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm9 = ymm10[0,1,2],ymm9[3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX1-ONLY-NEXT: vshufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm4 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm4 = ymm2[3,3],mem[3,3],ymm2[7,7],mem[7,7]
@@ -3870,7 +3870,7 @@ define void @store_i32_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vblendps $7, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm3 = mem[0,1,2],xmm3[3]
; AVX1-ONLY-NEXT: vbroadcastsd 104(%rax), %ymm5
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm5[4,5],ymm3[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm5[4,5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm3[2,3,4],ymm2[5,6,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX1-ONLY-NEXT: vunpcklps {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
@@ -4069,7 +4069,7 @@ define void @store_i32_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-SLOW-NEXT: vmovaps (%rcx), %ymm1
; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vunpckhps {{.*#+}} ymm1 = ymm2[2],ymm1[2],ymm2[3],ymm1[3],ymm2[6],ymm1[6],ymm2[7],ymm1[7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-SLOW-NEXT: vmovaps (%r8), %ymm2
; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovaps (%r9), %ymm1
@@ -4092,7 +4092,7 @@ define void @store_i32_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-SLOW-NEXT: vmovaps 32(%rcx), %ymm2
; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vunpckhps {{.*#+}} ymm1 = ymm1[2],ymm2[2],ymm1[3],ymm2[3],ymm1[6],ymm2[6],ymm1[7],ymm2[7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-SLOW-NEXT: vmovaps 32(%r8), %ymm2
; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovaps 32(%r9), %ymm1
@@ -4115,7 +4115,7 @@ define void @store_i32_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-SLOW-NEXT: vmovaps 64(%rcx), %ymm2
; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vunpckhps {{.*#+}} ymm1 = ymm1[2],ymm2[2],ymm1[3],ymm2[3],ymm1[6],ymm2[6],ymm1[7],ymm2[7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-SLOW-NEXT: vmovaps 64(%r8), %ymm2
; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovaps 64(%r9), %ymm1
@@ -4169,7 +4169,7 @@ define void @store_i32_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-SLOW-NEXT: vpermilps {{.*#+}} xmm4 = xmm8[2,2,2,2]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm4 = xmm9[0,1,2],xmm4[3]
; AVX2-SLOW-NEXT: vbroadcastsd 8(%rax), %ymm5
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm5[4,5],ymm4[6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm5[4,5,6,7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm4[2,3,4],ymm0[5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
@@ -4199,7 +4199,7 @@ define void @store_i32_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-SLOW-NEXT: vpermilps {{.*#+}} xmm4 = xmm9[2,2,2,2]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm4 = xmm7[0,1,2],xmm4[3]
; AVX2-SLOW-NEXT: vbroadcastsd 40(%rax), %ymm5
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm5[4,5],ymm4[6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm5[4,5,6,7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm4[2,3,4],ymm0[5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vbroadcastss %xmm13, %xmm0
@@ -4228,7 +4228,7 @@ define void @store_i32_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-SLOW-NEXT: vpermilps {{.*#+}} xmm4 = xmm6[2,2,2,2]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm4 = xmm7[0,1,2],xmm4[3]
; AVX2-SLOW-NEXT: vbroadcastsd 72(%rax), %ymm5
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm5[4,5],ymm4[6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm5[4,5,6,7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm4[2,3,4],ymm0[5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovaps (%rsp), %xmm9 # 16-byte Reload
@@ -4258,7 +4258,7 @@ define void @store_i32_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-SLOW-NEXT: vpermilps {{.*#+}} xmm4 = xmm11[2,2,2,2]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm4 = xmm6[0,1,2],xmm4[3]
; AVX2-SLOW-NEXT: vbroadcastsd 104(%rax), %ymm5
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm5[4,5],ymm4[6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm5[4,5,6,7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm4[2,3,4],ymm0[5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vbroadcastss 112(%rdx), %ymm0
@@ -4274,7 +4274,7 @@ define void @store_i32_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm5 = ymm5[2,2,2,2]
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX2-SLOW-NEXT: vunpckhps {{.*#+}} ymm6 = ymm4[2],ymm1[2],ymm4[3],ymm1[3],ymm4[6],ymm1[6],ymm4[7],ymm1[7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm6[4,5],ymm5[6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm6[4,5,6,7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5],mem[6,7]
; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm7 = mem[1,2,2,3,5,6,6,7]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm7 = ymm7[2,2,2,2]
@@ -4289,7 +4289,7 @@ define void @store_i32_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-SLOW-NEXT: vunpckhps {{.*#+}} ymm1 = ymm2[2],ymm3[2],ymm2[3],ymm3[3],ymm2[6],ymm3[6],ymm2[7],ymm3[7]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[3,3,3,3]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[3,3,3,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4],ymm1[5,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
; AVX2-SLOW-NEXT: vbroadcastss 124(%r8), %ymm1
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5],ymm0[6,7]
; AVX2-SLOW-NEXT: vbroadcastss 124(%r9), %ymm1
@@ -4369,7 +4369,7 @@ define void @store_i32_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-SLOW-NEXT: vunpckhps {{.*#+}} ymm8 = ymm8[2],ymm5[2],ymm8[3],ymm5[3],ymm8[6],ymm5[6],ymm8[7],ymm5[7]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm3 = ymm3[3,3,3,3]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm8 = ymm8[3,3,3,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm8[0,1,2],ymm3[3,4],ymm8[5,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm8[0,1,2],ymm3[3,4,5,6,7]
; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm8 = ymm11[3,3],ymm13[3,3],ymm11[7,7],ymm13[7,7]
; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm9 = mem[2,3,2,3,6,7,6,7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm8 = ymm9[0],ymm8[1,2],ymm9[3,4],ymm8[5,6],ymm9[7]
@@ -4391,7 +4391,7 @@ define void @store_i32_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-SLOW-NEXT: vunpckhps {{.*#+}} ymm11 = ymm14[2],ymm12[2],ymm14[3],ymm12[3],ymm14[6],ymm12[6],ymm14[7],ymm12[7]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm3 = ymm3[3,3,3,3]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm11 = ymm11[3,3,3,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm11[0,1,2],ymm3[3,4],ymm11[5,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm11[0,1,2],ymm3[3,4,5,6,7]
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-SLOW-NEXT: vshufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm11 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: # ymm11 = ymm0[3,3],mem[3,3],ymm0[7,7],mem[7,7]
@@ -4418,7 +4418,7 @@ define void @store_i32_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-SLOW-NEXT: vunpckhps {{.*#+}} ymm13 = ymm0[2],ymm15[2],ymm0[3],ymm15[3],ymm0[6],ymm15[6],ymm0[7],ymm15[7]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm12 = ymm12[3,3,3,3]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm13 = ymm13[3,3,3,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm12 = ymm13[0,1,2],ymm12[3,4],ymm13[5,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm12 = ymm13[0,1,2],ymm12[3,4,5,6,7]
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-SLOW-NEXT: vshufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm13 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: # ymm13 = ymm0[3,3],mem[3,3],ymm0[7,7],mem[7,7]
@@ -4599,7 +4599,7 @@ define void @store_i32_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-FAST-NEXT: vunpckhps {{.*#+}} ymm1 = ymm9[2],ymm8[2],ymm9[3],ymm8[3],ymm9[6],ymm8[6],ymm9[7],ymm8[7]
; AVX2-FAST-NEXT: vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FAST-NEXT: vmovaps (%r8), %ymm6
; AVX2-FAST-NEXT: vmovaps (%r9), %ymm5
; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm1 = ymm5[1,1,2,2,5,5,6,6]
@@ -4622,7 +4622,7 @@ define void @store_i32_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-FAST-NEXT: vmovaps 32(%rcx), %ymm2
; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vunpckhps {{.*#+}} ymm1 = ymm1[2],ymm2[2],ymm1[3],ymm2[3],ymm1[6],ymm2[6],ymm1[7],ymm2[7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FAST-NEXT: vmovaps 32(%r8), %ymm2
; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovaps 32(%r9), %ymm1
@@ -4645,7 +4645,7 @@ define void @store_i32_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-FAST-NEXT: vmovaps 64(%rcx), %ymm2
; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vunpckhps {{.*#+}} ymm1 = ymm1[2],ymm2[2],ymm1[3],ymm2[3],ymm1[6],ymm2[6],ymm1[7],ymm2[7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FAST-NEXT: vmovaps 64(%r8), %ymm2
; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovaps 64(%r9), %ymm1
@@ -4685,7 +4685,7 @@ define void @store_i32_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-FAST-NEXT: vunpckhps {{.*#+}} ymm15 = ymm3[2],ymm2[2],ymm3[3],ymm2[3],ymm3[6],ymm2[6],ymm3[7],ymm2[7]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm15 = ymm15[2,2,2,2]
; AVX2-FAST-NEXT: vunpckhps {{.*#+}} ymm0 = ymm4[2],ymm1[2],ymm4[3],ymm1[3],ymm4[6],ymm1[6],ymm4[7],ymm1[7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm15 = ymm15[0,1,2,3],ymm0[4,5],ymm15[6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm15 = ymm15[0,1,2,3],ymm0[4,5,6,7]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm15 = ymm15[0,1,2,3,4,5],mem[6,7]
; AVX2-FAST-NEXT: vbroadcastsd {{.*#+}} ymm11 = [25769803781,25769803781,25769803781,25769803781]
; AVX2-FAST-NEXT: vpermps 96(%r9), %ymm11, %ymm11
@@ -4700,7 +4700,7 @@ define void @store_i32_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-FAST-NEXT: vunpckhps {{.*#+}} ymm1 = ymm2[2],ymm3[2],ymm2[3],ymm3[3],ymm2[6],ymm3[6],ymm2[7],ymm3[7]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[3,3,3,3]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[3,3,3,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4],ymm1[5,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
; AVX2-FAST-NEXT: vbroadcastss 124(%r8), %ymm1
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5],ymm0[6,7]
; AVX2-FAST-NEXT: vbroadcastss 124(%r9), %ymm1
@@ -4735,7 +4735,7 @@ define void @store_i32_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-FAST-NEXT: vpermilps {{.*#+}} xmm2 = xmm11[2,2,2,2]
; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm2 = xmm4[0,1,2],xmm2[3]
; AVX2-FAST-NEXT: vbroadcastsd 8(%rax), %ymm3
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5],ymm2[6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm2[2,3,4],ymm1[5,6,7]
; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm1 = ymm10[1,1,1,1,5,5,5,5]
@@ -4778,7 +4778,7 @@ define void @store_i32_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-FAST-NEXT: vpermilps {{.*#+}} xmm2 = xmm8[2,2,2,2]
; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm2 = xmm9[0,1,2],xmm2[3]
; AVX2-FAST-NEXT: vbroadcastsd 40(%rax), %ymm3
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5],ymm2[6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm2[2,3,4],ymm1[5,6,7]
; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
@@ -4826,7 +4826,7 @@ define void @store_i32_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-FAST-NEXT: vpermilps {{.*#+}} xmm5 = xmm1[2,2,2,2]
; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm5 = xmm6[0,1,2],xmm5[3]
; AVX2-FAST-NEXT: vbroadcastsd 72(%rax), %ymm6
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm6[4,5],ymm5[6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm6[4,5,6,7]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm4[0,1],ymm5[2,3,4],ymm4[5,6,7]
; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
@@ -4873,7 +4873,7 @@ define void @store_i32_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-FAST-NEXT: vpermilps {{.*#+}} xmm6 = xmm5[2,2,2,2]
; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm6 = xmm1[0,1,2],xmm6[3]
; AVX2-FAST-NEXT: vbroadcastsd 104(%rax), %ymm9
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm9[4,5],ymm6[6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm9[4,5,6,7]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm5 = ymm0[0,1],ymm6[2,3,4],ymm0[5,6,7]
; AVX2-FAST-NEXT: vbroadcastss 16(%rdx), %ymm6
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
@@ -4893,7 +4893,7 @@ define void @store_i32_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-FAST-NEXT: vunpckhps {{.*#+}} ymm7 = ymm1[2],ymm10[2],ymm1[3],ymm10[3],ymm1[6],ymm10[6],ymm1[7],ymm10[7]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm6 = ymm6[3,3,3,3]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm7 = ymm7[3,3,3,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm6 = ymm7[0,1,2],ymm6[3,4],ymm7[5,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm6 = ymm7[0,1,2],ymm6[3,4,5,6,7]
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FAST-NEXT: vshufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm7 # 32-byte Folded Reload
; AVX2-FAST-NEXT: # ymm7 = ymm0[3,3],mem[3,3],ymm0[7,7],mem[7,7]
@@ -4918,7 +4918,7 @@ define void @store_i32_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-FAST-NEXT: vunpckhps {{.*#+}} ymm11 = ymm0[2],ymm12[2],ymm0[3],ymm12[3],ymm0[6],ymm12[6],ymm0[7],ymm12[7]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm10 = ymm10[3,3,3,3]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm11 = ymm11[3,3,3,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm10 = ymm11[0,1,2],ymm10[3,4],ymm11[5,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm10 = ymm11[0,1,2],ymm10[3,4,5,6,7]
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FAST-NEXT: vshufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm11 # 32-byte Folded Reload
; AVX2-FAST-NEXT: # ymm11 = ymm0[3,3],mem[3,3],ymm0[7,7],mem[7,7]
@@ -4942,7 +4942,7 @@ define void @store_i32_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-FAST-NEXT: vunpckhps {{.*#+}} ymm13 = ymm15[2],ymm14[2],ymm15[3],ymm14[3],ymm15[6],ymm14[6],ymm15[7],ymm14[7]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm12 = ymm12[3,3,3,3]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm13 = ymm13[3,3,3,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm12 = ymm13[0,1,2],ymm12[3,4],ymm13[5,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm12 = ymm13[0,1,2],ymm12[3,4,5,6,7]
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FAST-NEXT: vshufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm13 # 32-byte Folded Reload
; AVX2-FAST-NEXT: # ymm13 = ymm0[3,3],mem[3,3],ymm0[7,7],mem[7,7]
@@ -5112,7 +5112,7 @@ define void @store_i32_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-FAST-PERLANE-NEXT: vmovaps (%rcx), %ymm1
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vunpckhps {{.*#+}} ymm1 = ymm2[2],ymm1[2],ymm2[3],ymm1[3],ymm2[6],ymm1[6],ymm2[7],ymm1[7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovaps (%r8), %ymm2
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovaps (%r9), %ymm1
@@ -5135,7 +5135,7 @@ define void @store_i32_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-FAST-PERLANE-NEXT: vmovaps 32(%rcx), %ymm2
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vunpckhps {{.*#+}} ymm1 = ymm1[2],ymm2[2],ymm1[3],ymm2[3],ymm1[6],ymm2[6],ymm1[7],ymm2[7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovaps 32(%r8), %ymm2
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovaps 32(%r9), %ymm1
@@ -5158,7 +5158,7 @@ define void @store_i32_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-FAST-PERLANE-NEXT: vmovaps 64(%rcx), %ymm2
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vunpckhps {{.*#+}} ymm1 = ymm1[2],ymm2[2],ymm1[3],ymm2[3],ymm1[6],ymm2[6],ymm1[7],ymm2[7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovaps 64(%r8), %ymm2
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovaps 64(%r9), %ymm1
@@ -5212,7 +5212,7 @@ define void @store_i32_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} xmm4 = xmm8[2,2,2,2]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm4 = xmm9[0,1,2],xmm4[3]
; AVX2-FAST-PERLANE-NEXT: vbroadcastsd 8(%rax), %ymm5
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm5[4,5],ymm4[6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm5[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm4[2,3,4],ymm0[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
@@ -5242,7 +5242,7 @@ define void @store_i32_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} xmm4 = xmm9[2,2,2,2]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm4 = xmm7[0,1,2],xmm4[3]
; AVX2-FAST-PERLANE-NEXT: vbroadcastsd 40(%rax), %ymm5
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm5[4,5],ymm4[6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm5[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm4[2,3,4],ymm0[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vbroadcastss %xmm13, %xmm0
@@ -5271,7 +5271,7 @@ define void @store_i32_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} xmm4 = xmm6[2,2,2,2]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm4 = xmm7[0,1,2],xmm4[3]
; AVX2-FAST-PERLANE-NEXT: vbroadcastsd 72(%rax), %ymm5
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm5[4,5],ymm4[6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm5[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm4[2,3,4],ymm0[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovaps (%rsp), %xmm9 # 16-byte Reload
@@ -5301,7 +5301,7 @@ define void @store_i32_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} xmm4 = xmm11[2,2,2,2]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm4 = xmm6[0,1,2],xmm4[3]
; AVX2-FAST-PERLANE-NEXT: vbroadcastsd 104(%rax), %ymm5
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm5[4,5],ymm4[6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm5[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm4[2,3,4],ymm0[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vbroadcastss 112(%rdx), %ymm0
@@ -5317,7 +5317,7 @@ define void @store_i32_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm5 = ymm5[2,2,2,2]
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vunpckhps {{.*#+}} ymm6 = ymm4[2],ymm1[2],ymm4[3],ymm1[3],ymm4[6],ymm1[6],ymm4[7],ymm1[7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm6[4,5],ymm5[6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm6[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5],mem[6,7]
; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm7 = mem[1,2,2,3,5,6,6,7]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm7 = ymm7[2,2,2,2]
@@ -5332,7 +5332,7 @@ define void @store_i32_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-FAST-PERLANE-NEXT: vunpckhps {{.*#+}} ymm1 = ymm2[2],ymm3[2],ymm2[3],ymm3[3],ymm2[6],ymm3[6],ymm2[7],ymm3[7]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[3,3,3,3]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[3,3,3,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4],ymm1[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vbroadcastss 124(%r8), %ymm1
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5],ymm0[6,7]
; AVX2-FAST-PERLANE-NEXT: vbroadcastss 124(%r9), %ymm1
@@ -5412,7 +5412,7 @@ define void @store_i32_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-FAST-PERLANE-NEXT: vunpckhps {{.*#+}} ymm8 = ymm8[2],ymm5[2],ymm8[3],ymm5[3],ymm8[6],ymm5[6],ymm8[7],ymm5[7]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm3 = ymm3[3,3,3,3]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm8 = ymm8[3,3,3,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm8[0,1,2],ymm3[3,4],ymm8[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm8[0,1,2],ymm3[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm8 = ymm11[3,3],ymm13[3,3],ymm11[7,7],ymm13[7,7]
; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm9 = mem[2,3,2,3,6,7,6,7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm8 = ymm9[0],ymm8[1,2],ymm9[3,4],ymm8[5,6],ymm9[7]
@@ -5434,7 +5434,7 @@ define void @store_i32_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-FAST-PERLANE-NEXT: vunpckhps {{.*#+}} ymm11 = ymm14[2],ymm12[2],ymm14[3],ymm12[3],ymm14[6],ymm12[6],ymm14[7],ymm12[7]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm3 = ymm3[3,3,3,3]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm11 = ymm11[3,3,3,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm11[0,1,2],ymm3[3,4],ymm11[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm11[0,1,2],ymm3[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vshufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm11 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # ymm11 = ymm0[3,3],mem[3,3],ymm0[7,7],mem[7,7]
@@ -5461,7 +5461,7 @@ define void @store_i32_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-FAST-PERLANE-NEXT: vunpckhps {{.*#+}} ymm13 = ymm0[2],ymm15[2],ymm0[3],ymm15[3],ymm0[6],ymm15[6],ymm0[7],ymm15[7]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm12 = ymm12[3,3,3,3]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm13 = ymm13[3,3,3,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm12 = ymm13[0,1,2],ymm12[3,4],ymm13[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm12 = ymm13[0,1,2],ymm12[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vshufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm13 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # ymm13 = ymm0[3,3],mem[3,3],ymm0[7,7],mem[7,7]
@@ -7819,7 +7819,7 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm3 = ymm3[2,3,2,3]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm13 = ymm8[3,3],ymm9[3,3],ymm8[7,7],ymm9[7,7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm13, %xmm13
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm13[0,1,2],ymm3[3,4],ymm13[5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm13[0,1,2],ymm3[3,4,5,6,7]
; AVX1-ONLY-NEXT: vbroadcastss 220(%r8), %ymm13
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4],ymm13[5],ymm3[6,7]
; AVX1-ONLY-NEXT: vbroadcastss 220(%r9), %ymm13
@@ -7848,7 +7848,7 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm15[3,3],ymm2[3,3],ymm15[7,7],ymm2[7,7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm1, %xmm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4],ymm1[5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
; AVX1-ONLY-NEXT: vbroadcastss 252(%r8), %ymm1
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5],ymm0[6,7]
; AVX1-ONLY-NEXT: vbroadcastss 252(%r9), %ymm1
@@ -7892,7 +7892,7 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vblendps $7, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm1 = mem[0,1,2],xmm1[3]
; AVX1-ONLY-NEXT: vbroadcastsd 8(%rax), %ymm2
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4],ymm0[5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
@@ -7902,7 +7902,7 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm4[3,3],ymm10[3,3],ymm4[7,7],ymm10[7,7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm1, %xmm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4],ymm1[5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX1-ONLY-NEXT: vshufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm1 = ymm1[3,3],mem[3,3],ymm1[7,7],mem[7,7]
@@ -7926,7 +7926,7 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vblendps $7, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm1 = mem[0,1,2],xmm1[3]
; AVX1-ONLY-NEXT: vbroadcastsd 40(%rax), %ymm2
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4],ymm0[5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
@@ -7937,7 +7937,7 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vshufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm1 = ymm1[3,3],mem[3,3],ymm1[7,7],mem[7,7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm1, %xmm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4],ymm1[5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX1-ONLY-NEXT: vshufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm1 = ymm1[3,3],mem[3,3],ymm1[7,7],mem[7,7]
@@ -7961,7 +7961,7 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vblendps $7, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm1 = mem[0,1,2],xmm1[3]
; AVX1-ONLY-NEXT: vbroadcastsd 72(%rax), %ymm2
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4],ymm0[5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
@@ -7971,7 +7971,7 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm1[3,3],ymm6[3,3],ymm1[7,7],ymm6[7,7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm1, %xmm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4],ymm1[5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX1-ONLY-NEXT: vshufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm1 = ymm1[3,3],mem[3,3],ymm1[7,7],mem[7,7]
@@ -7995,7 +7995,7 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vblendps $7, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm1 = mem[0,1,2],xmm1[3]
; AVX1-ONLY-NEXT: vbroadcastsd 104(%rax), %ymm2
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4],ymm0[5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
@@ -8005,7 +8005,7 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm1[3,3],ymm4[3,3],ymm1[7,7],ymm4[7,7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm1, %xmm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4],ymm1[5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX1-ONLY-NEXT: vshufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm1 = ymm1[3,3],mem[3,3],ymm1[7,7],mem[7,7]
@@ -8029,7 +8029,7 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vblendps $7, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm1 = mem[0,1,2],xmm1[3]
; AVX1-ONLY-NEXT: vbroadcastsd 136(%rax), %ymm2
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4],ymm0[5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
@@ -8040,7 +8040,7 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm12[3,3],ymm1[3,3],ymm12[7,7],ymm1[7,7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm2, %xmm2
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2],ymm0[3,4],ymm2[5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2],ymm0[3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX1-ONLY-NEXT: vshufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm2 = ymm2[3,3],mem[3,3],ymm2[7,7],mem[7,7]
@@ -8064,7 +8064,7 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm15[0,1,2],xmm2[3]
; AVX1-ONLY-NEXT: vbroadcastsd 168(%rax), %ymm3
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm2[2,3,4],ymm0[5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
@@ -8075,7 +8075,7 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm9[3,3],ymm7[3,3],ymm9[7,7],ymm7[7,7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm2, %xmm2
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2],ymm0[3,4],ymm2[5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2],ymm0[3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX1-ONLY-NEXT: vshufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm2 = ymm2[3,3],mem[3,3],ymm2[7,7],mem[7,7]
@@ -8099,7 +8099,7 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm5[0,1,2],xmm2[3]
; AVX1-ONLY-NEXT: vbroadcastsd 200(%rax), %ymm3
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm2[2,3,4],ymm0[5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
@@ -8488,7 +8488,7 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-SLOW-NEXT: vmovaps (%rcx), %ymm1
; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vunpckhps {{.*#+}} ymm1 = ymm2[2],ymm1[2],ymm2[3],ymm1[3],ymm2[6],ymm1[6],ymm2[7],ymm1[7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-SLOW-NEXT: vmovaps (%r8), %ymm2
; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovaps (%r9), %ymm1
@@ -8511,7 +8511,7 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-SLOW-NEXT: vmovaps 32(%rcx), %ymm2
; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vunpckhps {{.*#+}} ymm1 = ymm1[2],ymm2[2],ymm1[3],ymm2[3],ymm1[6],ymm2[6],ymm1[7],ymm2[7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-SLOW-NEXT: vmovaps 32(%r8), %ymm2
; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovaps 32(%r9), %ymm1
@@ -8534,7 +8534,7 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-SLOW-NEXT: vmovaps 64(%rcx), %ymm2
; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vunpckhps {{.*#+}} ymm1 = ymm1[2],ymm2[2],ymm1[3],ymm2[3],ymm1[6],ymm2[6],ymm1[7],ymm2[7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-SLOW-NEXT: vmovaps 64(%r8), %ymm2
; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovaps 64(%r9), %ymm1
@@ -8557,7 +8557,7 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-SLOW-NEXT: vmovaps 96(%rcx), %ymm2
; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vunpckhps {{.*#+}} ymm1 = ymm1[2],ymm2[2],ymm1[3],ymm2[3],ymm1[6],ymm2[6],ymm1[7],ymm2[7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-SLOW-NEXT: vmovaps 96(%r8), %ymm2
; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovaps 96(%r9), %ymm1
@@ -8580,7 +8580,7 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-SLOW-NEXT: vmovaps 128(%rcx), %ymm2
; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vunpckhps {{.*#+}} ymm1 = ymm1[2],ymm2[2],ymm1[3],ymm2[3],ymm1[6],ymm2[6],ymm1[7],ymm2[7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-SLOW-NEXT: vmovaps 128(%r8), %ymm2
; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovaps 128(%r9), %ymm1
@@ -8603,7 +8603,7 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-SLOW-NEXT: vmovaps 160(%rcx), %ymm2
; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vunpckhps {{.*#+}} ymm1 = ymm1[2],ymm2[2],ymm1[3],ymm2[3],ymm1[6],ymm2[6],ymm1[7],ymm2[7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-SLOW-NEXT: vmovaps 160(%r8), %ymm2
; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovaps 160(%r9), %ymm1
@@ -8626,7 +8626,7 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-SLOW-NEXT: vmovaps 192(%rcx), %ymm2
; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vunpckhps {{.*#+}} ymm1 = ymm1[2],ymm2[2],ymm1[3],ymm2[3],ymm1[6],ymm2[6],ymm1[7],ymm2[7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-SLOW-NEXT: vmovaps 192(%r8), %ymm2
; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovaps 192(%r9), %ymm1
@@ -8725,7 +8725,7 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-SLOW-NEXT: vpermilps {{.*#+}} xmm6 = xmm0[2,2,2,2]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm6 = xmm1[0,1,2],xmm6[3]
; AVX2-SLOW-NEXT: vbroadcastsd 8(%rax), %ymm8
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm8[4,5],ymm6[6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm8[4,5,6,7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm5[0,1],ymm6[2,3,4],ymm5[5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
@@ -8755,7 +8755,7 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-SLOW-NEXT: vpermilps {{.*#+}} xmm6 = xmm13[2,2,2,2]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm6 = xmm9[0,1,2],xmm6[3]
; AVX2-SLOW-NEXT: vbroadcastsd 40(%rax), %ymm7
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm7[4,5],ymm6[6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm7[4,5,6,7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm5[0,1],ymm6[2,3,4],ymm5[5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
@@ -8785,7 +8785,7 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-SLOW-NEXT: vpermilps {{.*#+}} xmm6 = xmm0[2,2,2,2]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm6 = xmm1[0,1,2],xmm6[3]
; AVX2-SLOW-NEXT: vbroadcastsd 72(%rax), %ymm7
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm7[4,5],ymm6[6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm7[4,5,6,7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm5[0,1],ymm6[2,3,4],ymm5[5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
@@ -8816,7 +8816,7 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-SLOW-NEXT: vpermilps {{.*#+}} xmm6 = xmm11[2,2,2,2]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm6 = xmm8[0,1,2],xmm6[3]
; AVX2-SLOW-NEXT: vbroadcastsd 104(%rax), %ymm7
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm7[4,5],ymm6[6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm7[4,5,6,7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm5[0,1],ymm6[2,3,4],ymm5[5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
@@ -8847,7 +8847,7 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-SLOW-NEXT: vpermilps {{.*#+}} xmm6 = xmm8[2,2,2,2]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm6 = xmm0[0,1,2],xmm6[3]
; AVX2-SLOW-NEXT: vbroadcastsd 136(%rax), %ymm7
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm7[4,5],ymm6[6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm7[4,5,6,7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm5[0,1],ymm6[2,3,4],ymm5[5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
@@ -8878,7 +8878,7 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-SLOW-NEXT: vpermilps {{.*#+}} xmm6 = xmm11[2,2,2,2]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm6 = xmm0[0,1,2],xmm6[3]
; AVX2-SLOW-NEXT: vbroadcastsd 168(%rax), %ymm7
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm7[4,5],ymm6[6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm7[4,5,6,7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm5[0,1],ymm6[2,3,4],ymm5[5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
@@ -8909,7 +8909,7 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-SLOW-NEXT: vpermilps {{.*#+}} xmm6 = xmm9[2,2,2,2]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm6 = xmm0[0,1,2],xmm6[3]
; AVX2-SLOW-NEXT: vbroadcastsd 200(%rax), %ymm7
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm7[4,5],ymm6[6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm7[4,5,6,7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm5[0,1],ymm6[2,3,4],ymm5[5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
@@ -8920,7 +8920,7 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
; AVX2-SLOW-NEXT: vunpckhps {{.*#+}} ymm6 = ymm9[2],ymm14[2],ymm9[3],ymm14[3],ymm9[6],ymm14[6],ymm9[7],ymm14[7]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm6 = ymm6[3,3,3,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm6[0,1,2],ymm5[3,4],ymm6[5,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm6[0,1,2],ymm5[3,4,5,6,7]
; AVX2-SLOW-NEXT: vbroadcastss 220(%r8), %ymm6
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4],ymm6[5],ymm5[6,7]
; AVX2-SLOW-NEXT: vbroadcastss 220(%r9), %ymm6
@@ -8940,7 +8940,7 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-SLOW-NEXT: vunpckhps {{.*#+}} ymm5 = ymm10[2],ymm4[2],ymm10[3],ymm4[3],ymm10[6],ymm4[6],ymm10[7],ymm4[7]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm5 = ymm5[2,2,2,2]
; AVX2-SLOW-NEXT: vunpckhps {{.*#+}} ymm6 = ymm12[2],ymm2[2],ymm12[3],ymm2[3],ymm12[6],ymm2[6],ymm12[7],ymm2[7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm6[4,5],ymm5[6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm6[4,5,6,7]
; AVX2-SLOW-NEXT: vblendps $192, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: # ymm5 = ymm5[0,1,2,3,4,5],mem[6,7]
; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm7 = mem[1,2,2,3,5,6,6,7]
@@ -8956,7 +8956,7 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-SLOW-NEXT: vunpckhps {{.*#+}} ymm1 = ymm4[2],ymm10[2],ymm4[3],ymm10[3],ymm4[6],ymm10[6],ymm4[7],ymm10[7]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[3,3,3,3]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[3,3,3,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4],ymm1[5,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
; AVX2-SLOW-NEXT: vbroadcastss 252(%r8), %ymm1
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5],ymm0[6,7]
; AVX2-SLOW-NEXT: vbroadcastss 252(%r9), %ymm1
@@ -9110,7 +9110,7 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-SLOW-NEXT: vunpckhps {{.*#+}} ymm1 = ymm5[2],ymm4[2],ymm5[3],ymm4[3],ymm5[6],ymm4[6],ymm5[7],ymm4[7]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[3,3,3,3]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[3,3,3,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4],ymm1[5,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm8[3,3],ymm11[3,3],ymm8[7,7],ymm11[7,7]
; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm2 = mem[2,3,2,3,6,7,6,7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0],ymm1[1,2],ymm2[3,4],ymm1[5,6],ymm2[7]
@@ -9133,7 +9133,7 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-SLOW-NEXT: vunpckhps {{.*#+}} ymm1 = ymm12[2],ymm2[2],ymm12[3],ymm2[3],ymm12[6],ymm2[6],ymm12[7],ymm2[7]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[3,3,3,3]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[3,3,3,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4],ymm1[5,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-SLOW-NEXT: vshufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: # ymm1 = ymm1[3,3],mem[3,3],ymm1[7,7],mem[7,7]
@@ -9160,7 +9160,7 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-SLOW-NEXT: vunpckhps {{.*#+}} ymm1 = ymm9[2],ymm2[2],ymm9[3],ymm2[3],ymm9[6],ymm2[6],ymm9[7],ymm2[7]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[3,3,3,3]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[3,3,3,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4],ymm1[5,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-SLOW-NEXT: vshufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: # ymm1 = ymm1[3,3],mem[3,3],ymm1[7,7],mem[7,7]
@@ -9187,7 +9187,7 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-SLOW-NEXT: vunpckhps {{.*#+}} ymm1 = ymm3[2],ymm2[2],ymm3[3],ymm2[3],ymm3[6],ymm2[6],ymm3[7],ymm2[7]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[3,3,3,3]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[3,3,3,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4],ymm1[5,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-SLOW-NEXT: vshufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: # ymm1 = ymm1[3,3],mem[3,3],ymm1[7,7],mem[7,7]
@@ -9214,7 +9214,7 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-SLOW-NEXT: vunpckhps {{.*#+}} ymm0 = ymm2[2],ymm1[2],ymm2[3],ymm1[3],ymm2[6],ymm1[6],ymm2[7],ymm1[7]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm9 = ymm9[3,3,3,3]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[3,3,3,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm9[3,4],ymm0[5,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm9[3,4,5,6,7]
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-SLOW-NEXT: vshufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm9 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: # ymm9 = ymm1[3,3],mem[3,3],ymm1[7,7],mem[7,7]
@@ -9241,7 +9241,7 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-SLOW-NEXT: vunpckhps {{.*#+}} ymm0 = ymm12[2],ymm1[2],ymm12[3],ymm1[3],ymm12[6],ymm1[6],ymm12[7],ymm1[7]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm9 = ymm9[3,3,3,3]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[3,3,3,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm9[3,4],ymm0[5,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm9[3,4,5,6,7]
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-SLOW-NEXT: vshufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm9 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: # ymm9 = ymm1[3,3],mem[3,3],ymm1[7,7],mem[7,7]
@@ -9566,7 +9566,7 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vunpckhps {{.*#+}} ymm1 = ymm14[2],ymm1[2],ymm14[3],ymm1[3],ymm14[6],ymm1[6],ymm14[7],ymm1[7]
; AVX2-FAST-NEXT: vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FAST-NEXT: vmovaps (%r8), %ymm2
; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovaps (%r9), %ymm12
@@ -9589,7 +9589,7 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-FAST-NEXT: vmovaps 32(%rcx), %ymm2
; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vunpckhps {{.*#+}} ymm1 = ymm1[2],ymm2[2],ymm1[3],ymm2[3],ymm1[6],ymm2[6],ymm1[7],ymm2[7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FAST-NEXT: vmovaps 32(%r8), %ymm2
; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovaps 32(%r9), %ymm1
@@ -9612,7 +9612,7 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-FAST-NEXT: vmovaps 64(%rcx), %ymm2
; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vunpckhps {{.*#+}} ymm1 = ymm1[2],ymm2[2],ymm1[3],ymm2[3],ymm1[6],ymm2[6],ymm1[7],ymm2[7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FAST-NEXT: vmovaps 64(%r8), %ymm2
; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovaps 64(%r9), %ymm1
@@ -9635,7 +9635,7 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-FAST-NEXT: vmovaps 96(%rcx), %ymm2
; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vunpckhps {{.*#+}} ymm1 = ymm1[2],ymm2[2],ymm1[3],ymm2[3],ymm1[6],ymm2[6],ymm1[7],ymm2[7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FAST-NEXT: vmovaps 96(%r8), %ymm2
; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovaps 96(%r9), %ymm1
@@ -9658,7 +9658,7 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-FAST-NEXT: vmovaps 128(%rcx), %ymm2
; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vunpckhps {{.*#+}} ymm1 = ymm1[2],ymm2[2],ymm1[3],ymm2[3],ymm1[6],ymm2[6],ymm1[7],ymm2[7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FAST-NEXT: vmovaps 128(%r8), %ymm2
; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovaps 128(%r9), %ymm1
@@ -9681,7 +9681,7 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-FAST-NEXT: vmovaps 160(%rcx), %ymm2
; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vunpckhps {{.*#+}} ymm1 = ymm1[2],ymm2[2],ymm1[3],ymm2[3],ymm1[6],ymm2[6],ymm1[7],ymm2[7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FAST-NEXT: vmovaps 160(%r8), %ymm2
; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovaps 160(%r9), %ymm1
@@ -9705,7 +9705,7 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-FAST-NEXT: vunpckhps {{.*#+}} ymm1 = ymm8[2],ymm9[2],ymm8[3],ymm9[3],ymm8[6],ymm9[6],ymm8[7],ymm9[7]
; AVX2-FAST-NEXT: vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FAST-NEXT: vmovaps 192(%r8), %ymm2
; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovaps 192(%r9), %ymm1
@@ -9783,7 +9783,7 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm7 = ymm7[3,3,3,3]
; AVX2-FAST-NEXT: vunpckhps {{.*#+}} ymm15 = ymm11[2],ymm13[2],ymm11[3],ymm13[3],ymm11[6],ymm13[6],ymm11[7],ymm13[7]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm15 = ymm15[3,3,3,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm7 = ymm15[0,1,2],ymm7[3,4],ymm15[5,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm7 = ymm15[0,1,2],ymm7[3,4,5,6,7]
; AVX2-FAST-NEXT: vbroadcastss 220(%r8), %ymm15
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4],ymm15[5],ymm7[6,7]
; AVX2-FAST-NEXT: vbroadcastss 220(%r9), %ymm15
@@ -9803,7 +9803,7 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-FAST-NEXT: vunpckhps {{.*#+}} ymm7 = ymm2[2],ymm1[2],ymm2[3],ymm1[3],ymm2[6],ymm1[6],ymm2[7],ymm1[7]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm7 = ymm7[2,2,2,2]
; AVX2-FAST-NEXT: vunpckhps {{.*#+}} ymm15 = ymm3[2],ymm0[2],ymm3[3],ymm0[3],ymm3[6],ymm0[6],ymm3[7],ymm0[7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3],ymm15[4,5],ymm7[6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3],ymm15[4,5,6,7]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm6 = ymm7[0,1,2,3,4,5],ymm6[6,7]
; AVX2-FAST-NEXT: vbroadcastsd {{.*#+}} ymm7 = [25769803781,25769803781,25769803781,25769803781]
; AVX2-FAST-NEXT: vpermps 224(%r9), %ymm7, %ymm7
@@ -9818,7 +9818,7 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-FAST-NEXT: vunpckhps {{.*#+}} ymm1 = ymm1[2],ymm2[2],ymm1[3],ymm2[3],ymm1[6],ymm2[6],ymm1[7],ymm2[7]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[3,3,3,3]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[3,3,3,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4],ymm1[5,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
; AVX2-FAST-NEXT: vbroadcastss 252(%r8), %ymm1
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5],ymm0[6,7]
; AVX2-FAST-NEXT: vbroadcastss 252(%r9), %ymm1
@@ -9852,7 +9852,7 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-FAST-NEXT: vpermilps {{.*#+}} xmm1 = xmm6[2,2,2,2]
; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm1 = xmm3[0,1,2],xmm1[3]
; AVX2-FAST-NEXT: vbroadcastsd 8(%rax), %ymm2
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4],ymm0[5,6,7]
; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
@@ -9898,7 +9898,7 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-FAST-NEXT: vpermilps {{.*#+}} xmm1 = xmm7[2,2,2,2]
; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm1 = xmm3[0,1,2],xmm1[3]
; AVX2-FAST-NEXT: vbroadcastsd 40(%rax), %ymm2
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4],ymm0[5,6,7]
; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
@@ -9946,7 +9946,7 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-FAST-NEXT: vpermilps {{.*#+}} xmm1 = xmm3[2,2,2,2]
; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm1 = xmm8[0,1,2],xmm1[3]
; AVX2-FAST-NEXT: vbroadcastsd 72(%rax), %ymm2
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4],ymm0[5,6,7]
; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
@@ -9994,7 +9994,7 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-FAST-NEXT: vpermilps {{.*#+}} xmm1 = xmm7[2,2,2,2]
; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm1 = xmm8[0,1,2],xmm1[3]
; AVX2-FAST-NEXT: vbroadcastsd 104(%rax), %ymm2
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4],ymm0[5,6,7]
; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
@@ -10042,7 +10042,7 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-FAST-NEXT: vpermilps {{.*#+}} xmm1 = xmm7[2,2,2,2]
; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm1 = xmm8[0,1,2],xmm1[3]
; AVX2-FAST-NEXT: vbroadcastsd 136(%rax), %ymm2
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4],ymm0[5,6,7]
; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
@@ -10090,7 +10090,7 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-FAST-NEXT: vpermilps {{.*#+}} xmm1 = xmm7[2,2,2,2]
; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm1 = xmm8[0,1,2],xmm1[3]
; AVX2-FAST-NEXT: vbroadcastsd 168(%rax), %ymm2
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4],ymm0[5,6,7]
; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
@@ -10138,7 +10138,7 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-FAST-NEXT: vpermilps {{.*#+}} xmm1 = xmm6[2,2,2,2]
; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm1 = xmm3[0,1,2],xmm1[3]
; AVX2-FAST-NEXT: vbroadcastsd 200(%rax), %ymm2
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4],ymm0[5,6,7]
; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
@@ -10180,7 +10180,7 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-FAST-NEXT: vunpckhps {{.*#+}} ymm1 = ymm3[2],ymm2[2],ymm3[3],ymm2[3],ymm3[6],ymm2[6],ymm3[7],ymm2[7]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[3,3,3,3]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[3,3,3,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4],ymm1[5,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FAST-NEXT: vshufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
; AVX2-FAST-NEXT: # ymm1 = ymm1[3,3],mem[3,3],ymm1[7,7],mem[7,7]
@@ -10209,7 +10209,7 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-FAST-NEXT: vunpckhps {{.*#+}} ymm1 = ymm3[2],ymm2[2],ymm3[3],ymm2[3],ymm3[6],ymm2[6],ymm3[7],ymm2[7]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[3,3,3,3]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[3,3,3,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4],ymm1[5,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FAST-NEXT: vshufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
; AVX2-FAST-NEXT: # ymm1 = ymm1[3,3],mem[3,3],ymm1[7,7],mem[7,7]
@@ -10237,7 +10237,7 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-FAST-NEXT: vunpckhps {{.*#+}} ymm1 = ymm3[2],ymm2[2],ymm3[3],ymm2[3],ymm3[6],ymm2[6],ymm3[7],ymm2[7]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[3,3,3,3]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[3,3,3,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4],ymm1[5,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FAST-NEXT: vshufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
; AVX2-FAST-NEXT: # ymm1 = ymm1[3,3],mem[3,3],ymm1[7,7],mem[7,7]
@@ -10263,7 +10263,7 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-FAST-NEXT: vunpckhps {{.*#+}} ymm5 = ymm2[2],ymm12[2],ymm2[3],ymm12[3],ymm2[6],ymm12[6],ymm2[7],ymm12[7]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[3,3,3,3]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm5 = ymm5[3,3,3,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm5[0,1,2],ymm0[3,4],ymm5[5,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm5[0,1,2],ymm0[3,4,5,6,7]
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FAST-NEXT: vshufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm5 # 32-byte Folded Reload
; AVX2-FAST-NEXT: # ymm5 = ymm1[3,3],mem[3,3],ymm1[7,7],mem[7,7]
@@ -10289,7 +10289,7 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-FAST-NEXT: vunpckhps {{.*#+}} ymm0 = ymm1[2],ymm9[2],ymm1[3],ymm9[3],ymm1[6],ymm9[6],ymm1[7],ymm9[7]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm5 = ymm5[3,3,3,3]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[3,3,3,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm5[3,4],ymm0[5,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm5[3,4,5,6,7]
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FAST-NEXT: vshufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm5 # 32-byte Folded Reload
; AVX2-FAST-NEXT: # ymm5 = ymm1[3,3],mem[3,3],ymm1[7,7],mem[7,7]
@@ -10315,7 +10315,7 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-FAST-NEXT: vunpckhps {{.*#+}} ymm0 = ymm0[2],ymm2[2],ymm0[3],ymm2[3],ymm0[6],ymm2[6],ymm0[7],ymm2[7]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm9 = ymm9[3,3,3,3]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[3,3,3,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm9[3,4],ymm0[5,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm9[3,4,5,6,7]
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FAST-NEXT: vshufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm9 # 32-byte Folded Reload
; AVX2-FAST-NEXT: # ymm9 = ymm1[3,3],mem[3,3],ymm1[7,7],mem[7,7]
@@ -10633,7 +10633,7 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-FAST-PERLANE-NEXT: vmovaps (%rcx), %ymm1
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vunpckhps {{.*#+}} ymm1 = ymm2[2],ymm1[2],ymm2[3],ymm1[3],ymm2[6],ymm1[6],ymm2[7],ymm1[7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovaps (%r8), %ymm2
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovaps (%r9), %ymm1
@@ -10656,7 +10656,7 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-FAST-PERLANE-NEXT: vmovaps 32(%rcx), %ymm2
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vunpckhps {{.*#+}} ymm1 = ymm1[2],ymm2[2],ymm1[3],ymm2[3],ymm1[6],ymm2[6],ymm1[7],ymm2[7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovaps 32(%r8), %ymm2
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovaps 32(%r9), %ymm1
@@ -10679,7 +10679,7 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-FAST-PERLANE-NEXT: vmovaps 64(%rcx), %ymm2
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vunpckhps {{.*#+}} ymm1 = ymm1[2],ymm2[2],ymm1[3],ymm2[3],ymm1[6],ymm2[6],ymm1[7],ymm2[7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovaps 64(%r8), %ymm2
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovaps 64(%r9), %ymm1
@@ -10702,7 +10702,7 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-FAST-PERLANE-NEXT: vmovaps 96(%rcx), %ymm2
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vunpckhps {{.*#+}} ymm1 = ymm1[2],ymm2[2],ymm1[3],ymm2[3],ymm1[6],ymm2[6],ymm1[7],ymm2[7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovaps 96(%r8), %ymm2
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovaps 96(%r9), %ymm1
@@ -10725,7 +10725,7 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-FAST-PERLANE-NEXT: vmovaps 128(%rcx), %ymm2
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vunpckhps {{.*#+}} ymm1 = ymm1[2],ymm2[2],ymm1[3],ymm2[3],ymm1[6],ymm2[6],ymm1[7],ymm2[7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovaps 128(%r8), %ymm2
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovaps 128(%r9), %ymm1
@@ -10748,7 +10748,7 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-FAST-PERLANE-NEXT: vmovaps 160(%rcx), %ymm2
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vunpckhps {{.*#+}} ymm1 = ymm1[2],ymm2[2],ymm1[3],ymm2[3],ymm1[6],ymm2[6],ymm1[7],ymm2[7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovaps 160(%r8), %ymm2
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovaps 160(%r9), %ymm1
@@ -10771,7 +10771,7 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-FAST-PERLANE-NEXT: vmovaps 192(%rcx), %ymm2
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vunpckhps {{.*#+}} ymm1 = ymm1[2],ymm2[2],ymm1[3],ymm2[3],ymm1[6],ymm2[6],ymm1[7],ymm2[7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovaps 192(%r8), %ymm2
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovaps 192(%r9), %ymm1
@@ -10870,7 +10870,7 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} xmm6 = xmm0[2,2,2,2]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm6 = xmm1[0,1,2],xmm6[3]
; AVX2-FAST-PERLANE-NEXT: vbroadcastsd 8(%rax), %ymm8
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm8[4,5],ymm6[6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm8[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm5[0,1],ymm6[2,3,4],ymm5[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
@@ -10900,7 +10900,7 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} xmm6 = xmm13[2,2,2,2]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm6 = xmm9[0,1,2],xmm6[3]
; AVX2-FAST-PERLANE-NEXT: vbroadcastsd 40(%rax), %ymm7
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm7[4,5],ymm6[6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm7[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm5[0,1],ymm6[2,3,4],ymm5[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
@@ -10930,7 +10930,7 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} xmm6 = xmm0[2,2,2,2]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm6 = xmm1[0,1,2],xmm6[3]
; AVX2-FAST-PERLANE-NEXT: vbroadcastsd 72(%rax), %ymm7
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm7[4,5],ymm6[6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm7[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm5[0,1],ymm6[2,3,4],ymm5[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
@@ -10961,7 +10961,7 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} xmm6 = xmm11[2,2,2,2]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm6 = xmm8[0,1,2],xmm6[3]
; AVX2-FAST-PERLANE-NEXT: vbroadcastsd 104(%rax), %ymm7
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm7[4,5],ymm6[6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm7[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm5[0,1],ymm6[2,3,4],ymm5[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
@@ -10992,7 +10992,7 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} xmm6 = xmm8[2,2,2,2]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm6 = xmm0[0,1,2],xmm6[3]
; AVX2-FAST-PERLANE-NEXT: vbroadcastsd 136(%rax), %ymm7
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm7[4,5],ymm6[6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm7[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm5[0,1],ymm6[2,3,4],ymm5[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
@@ -11023,7 +11023,7 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} xmm6 = xmm11[2,2,2,2]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm6 = xmm0[0,1,2],xmm6[3]
; AVX2-FAST-PERLANE-NEXT: vbroadcastsd 168(%rax), %ymm7
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm7[4,5],ymm6[6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm7[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm5[0,1],ymm6[2,3,4],ymm5[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
@@ -11054,7 +11054,7 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} xmm6 = xmm9[2,2,2,2]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm6 = xmm0[0,1,2],xmm6[3]
; AVX2-FAST-PERLANE-NEXT: vbroadcastsd 200(%rax), %ymm7
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm7[4,5],ymm6[6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm7[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm5[0,1],ymm6[2,3,4],ymm5[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
@@ -11065,7 +11065,7 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vunpckhps {{.*#+}} ymm6 = ymm9[2],ymm14[2],ymm9[3],ymm14[3],ymm9[6],ymm14[6],ymm9[7],ymm14[7]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm6 = ymm6[3,3,3,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm6[0,1,2],ymm5[3,4],ymm6[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm6[0,1,2],ymm5[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vbroadcastss 220(%r8), %ymm6
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4],ymm6[5],ymm5[6,7]
; AVX2-FAST-PERLANE-NEXT: vbroadcastss 220(%r9), %ymm6
@@ -11085,7 +11085,7 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-FAST-PERLANE-NEXT: vunpckhps {{.*#+}} ymm5 = ymm10[2],ymm4[2],ymm10[3],ymm4[3],ymm10[6],ymm4[6],ymm10[7],ymm4[7]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm5 = ymm5[2,2,2,2]
; AVX2-FAST-PERLANE-NEXT: vunpckhps {{.*#+}} ymm6 = ymm12[2],ymm2[2],ymm12[3],ymm2[3],ymm12[6],ymm2[6],ymm12[7],ymm2[7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm6[4,5],ymm5[6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm6[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vblendps $192, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # ymm5 = ymm5[0,1,2,3,4,5],mem[6,7]
; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm7 = mem[1,2,2,3,5,6,6,7]
@@ -11101,7 +11101,7 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-FAST-PERLANE-NEXT: vunpckhps {{.*#+}} ymm1 = ymm4[2],ymm10[2],ymm4[3],ymm10[3],ymm4[6],ymm10[6],ymm4[7],ymm10[7]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[3,3,3,3]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[3,3,3,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4],ymm1[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vbroadcastss 252(%r8), %ymm1
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5],ymm0[6,7]
; AVX2-FAST-PERLANE-NEXT: vbroadcastss 252(%r9), %ymm1
@@ -11255,7 +11255,7 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-FAST-PERLANE-NEXT: vunpckhps {{.*#+}} ymm1 = ymm5[2],ymm4[2],ymm5[3],ymm4[3],ymm5[6],ymm4[6],ymm5[7],ymm4[7]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[3,3,3,3]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[3,3,3,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4],ymm1[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm8[3,3],ymm11[3,3],ymm8[7,7],ymm11[7,7]
; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm2 = mem[2,3,2,3,6,7,6,7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0],ymm1[1,2],ymm2[3,4],ymm1[5,6],ymm2[7]
@@ -11278,7 +11278,7 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-FAST-PERLANE-NEXT: vunpckhps {{.*#+}} ymm1 = ymm12[2],ymm2[2],ymm12[3],ymm2[3],ymm12[6],ymm2[6],ymm12[7],ymm2[7]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[3,3,3,3]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[3,3,3,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4],ymm1[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vshufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # ymm1 = ymm1[3,3],mem[3,3],ymm1[7,7],mem[7,7]
@@ -11305,7 +11305,7 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-FAST-PERLANE-NEXT: vunpckhps {{.*#+}} ymm1 = ymm9[2],ymm2[2],ymm9[3],ymm2[3],ymm9[6],ymm2[6],ymm9[7],ymm2[7]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[3,3,3,3]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[3,3,3,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4],ymm1[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vshufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # ymm1 = ymm1[3,3],mem[3,3],ymm1[7,7],mem[7,7]
@@ -11332,7 +11332,7 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-FAST-PERLANE-NEXT: vunpckhps {{.*#+}} ymm1 = ymm3[2],ymm2[2],ymm3[3],ymm2[3],ymm3[6],ymm2[6],ymm3[7],ymm2[7]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[3,3,3,3]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[3,3,3,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4],ymm1[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vshufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # ymm1 = ymm1[3,3],mem[3,3],ymm1[7,7],mem[7,7]
@@ -11359,7 +11359,7 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-FAST-PERLANE-NEXT: vunpckhps {{.*#+}} ymm0 = ymm2[2],ymm1[2],ymm2[3],ymm1[3],ymm2[6],ymm1[6],ymm2[7],ymm1[7]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm9 = ymm9[3,3,3,3]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[3,3,3,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm9[3,4],ymm0[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm9[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vshufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm9 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # ymm9 = ymm1[3,3],mem[3,3],ymm1[7,7],mem[7,7]
@@ -11386,7 +11386,7 @@ define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-FAST-PERLANE-NEXT: vunpckhps {{.*#+}} ymm0 = ymm12[2],ymm1[2],ymm12[3],ymm1[3],ymm12[6],ymm1[6],ymm12[7],ymm1[7]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm9 = ymm9[3,3,3,3]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[3,3,3,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm9[3,4],ymm0[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm9[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vshufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm9 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # ymm9 = ymm1[3,3],mem[3,3],ymm1[7,7],mem[7,7]
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-store-i64-stride-5.ll b/llvm/test/CodeGen/X86/vector-interleaved-store-i64-stride-5.ll
index 158813076ed54..0d7acc2f45710 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-store-i64-stride-5.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-store-i64-stride-5.ll
@@ -154,12 +154,12 @@ define void @store_i64_stride5_vf4(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm2[0,1],ymm1[2],ymm2[3]
; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm4 = mem[0,0]
; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm5 = ymm0[0],mem[0],ymm0[2],mem[2]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm4 = ymm4[0],ymm5[1,2,3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm4 = ymm4[0,1],ymm5[2,3]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm4 = ymm4[0],ymm1[1],ymm4[2,3]
; AVX1-ONLY-NEXT: vmovapd 16(%rsi), %xmm5
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm5 = xmm5[1],mem[1]
; AVX1-ONLY-NEXT: vbroadcastsd 24(%rcx), %ymm6
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm5 = ymm5[0,1],ymm6[2],ymm5[3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm5 = ymm5[0,1],ymm6[2,3]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm5 = ymm5[0,1,2],ymm1[3]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm0
; AVX1-ONLY-NEXT: vbroadcastsd 8(%rsi), %ymm6
@@ -186,12 +186,12 @@ define void @store_i64_stride5_vf4(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
; AVX2-ONLY-NEXT: vmovaps (%rdx), %xmm4
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm4, %ymm0, %ymm5
; AVX2-ONLY-NEXT: vbroadcastsd 8(%rsi), %ymm6
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm0[0,1,2,3],ymm6[4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm0[0,1,2,3],ymm6[4,5,6,7]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm6[0,1,2,3,4,5],ymm5[6,7]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm3[0,1],ymm5[2,3,4,5,6,7]
; AVX2-ONLY-NEXT: vmovddup {{.*#+}} xmm6 = mem[0,0]
; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm7 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1],ymm7[2,3,4,5,6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm7[4,5,6,7]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1],ymm3[2,3],ymm6[4,5,6,7]
; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm2[2,3],ymm0[2,3]
; AVX2-ONLY-NEXT: vbroadcastsd 16(%rcx), %ymm7
@@ -347,7 +347,7 @@ define void @store_i64_stride5_vf8(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm3[6,7]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm4
; AVX1-ONLY-NEXT: vbroadcastsd 8(%rsi), %ymm7
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm1[0,1,2,3],ymm7[4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm1[0,1,2,3],ymm7[4,5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm7[0,1,2,3,4,5],ymm4[6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm3[0,1],ymm4[2,3,4,5,6,7]
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm7 = mem[2,3,2,3]
@@ -359,7 +359,7 @@ define void @store_i64_stride5_vf8(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm9 = ymm9[0,1],ymm5[2],ymm9[3]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm6, %ymm0, %ymm12
; AVX1-ONLY-NEXT: vbroadcastsd 40(%rsi), %ymm13
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm10 = ymm10[0,1],ymm13[2],ymm10[3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm10 = ymm10[0,1],ymm13[2,3]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm10 = ymm10[0,1,2],ymm12[3]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm10 = ymm5[0],ymm10[1,2,3]
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm12 = mem[2,3,2,3]
@@ -369,10 +369,10 @@ define void @store_i64_stride5_vf8(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
; AVX1-ONLY-NEXT: vmovapd 48(%rsi), %xmm13
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm11 = xmm13[1],xmm11[1]
; AVX1-ONLY-NEXT: vbroadcastsd 56(%rcx), %ymm13
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm11 = ymm11[0,1],ymm13[2],ymm11[3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm11 = ymm11[0,1],ymm13[2,3]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm5 = ymm11[0,1,2],ymm5[3]
; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm8 = xmm8[0],mem[0]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm8[0,1,2,3,4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm8[0,1,2,3],ymm1[4,5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5],ymm1[6,7]
; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm3 = xmm6[0],mem[0]
; AVX1-ONLY-NEXT: vmovaps (%rdi), %xmm6
@@ -407,7 +407,7 @@ define void @store_i64_stride5_vf8(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
; AVX2-ONLY-NEXT: vmovaps 32(%rdx), %xmm8
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
; AVX2-ONLY-NEXT: vbroadcastsd 8(%rsi), %ymm3
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm2[0,1,2,3],ymm3[4,5],ymm2[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm2[0,1,2,3],ymm3[4,5,6,7]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3,4,5],ymm0[6,7]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm5[0,1],ymm0[2,3,4,5,6,7]
; AVX2-ONLY-NEXT: vmovddup {{.*#+}} xmm3 = mem[0,0]
@@ -416,7 +416,7 @@ define void @store_i64_stride5_vf8(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1],ymm9[2,3],ymm3[4,5,6,7]
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm8, %ymm0, %ymm10
; AVX2-ONLY-NEXT: vbroadcastsd 40(%rsi), %ymm11
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm11 = ymm7[0,1,2,3],ymm11[4,5],ymm7[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm11 = ymm7[0,1,2,3],ymm11[4,5,6,7]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm10 = ymm11[0,1,2,3,4,5],ymm10[6,7]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm10 = ymm9[0,1],ymm10[2,3,4,5,6,7]
; AVX2-ONLY-NEXT: vmovddup {{.*#+}} xmm11 = mem[0,0]
@@ -426,7 +426,7 @@ define void @store_i64_stride5_vf8(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
; AVX2-ONLY-NEXT: vbroadcastsd 56(%rsi), %ymm12
; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm12 = xmm12[0,1],mem[2,3]
; AVX2-ONLY-NEXT: vbroadcastsd 56(%rcx), %ymm13
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3],ymm13[4,5],ymm12[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3],ymm13[4,5,6,7]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3,4,5],ymm9[6,7]
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm13 = ymm4[1],ymm6[1],ymm4[3],ymm6[3]
; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm13 = ymm13[0,2,3,3]
@@ -808,7 +808,7 @@ define void @store_i64_stride5_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0
; AVX1-ONLY-NEXT: vbroadcastsd 8(%rsi), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm5[0,1,2,3],ymm1[4,5],ymm5[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm5[0,1,2,3],ymm1[4,5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm0 = mem[2,3,2,3]
@@ -817,7 +817,7 @@ define void @store_i64_stride5_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm0
; AVX1-ONLY-NEXT: vbroadcastsd 40(%rsi), %ymm1
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm6[0,1],ymm1[2],ymm6[3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm6[0,1],ymm1[2,3]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3]
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm0 = mem[2,3,2,3]
@@ -825,7 +825,7 @@ define void @store_i64_stride5_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm14 = ymm0[0,1],ymm1[2,3]
; AVX1-ONLY-NEXT: vmovaps 64(%rdi), %ymm15
; AVX1-ONLY-NEXT: vbroadcastsd 72(%rsi), %ymm0
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm15[0,1,2,3],ymm0[4,5],ymm15[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm15[0,1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps 64(%rdx), %xmm11
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm11, %ymm0, %ymm7
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm0[0,1,2,3,4,5],ymm7[6,7]
@@ -833,12 +833,12 @@ define void @store_i64_stride5_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm10 = ymm15[0],mem[0],ymm15[2],mem[2]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm0[0,1,2,3],ymm10[4,5,6,7]
; AVX1-ONLY-NEXT: vbroadcastsd 104(%rsi), %ymm0
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm10 = ymm8[0,1],ymm0[2],ymm8[3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm10 = ymm8[0,1],ymm0[2,3]
; AVX1-ONLY-NEXT: vmovaps 96(%rdx), %xmm2
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm12
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm3 = ymm10[0,1,2],ymm12[3]
; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm9 = xmm9[0],mem[0]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm9 = ymm9[0,1,2,3,4,5],ymm5[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm9 = ymm9[0,1,2,3],ymm5[4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps (%r8), %ymm0
; AVX1-ONLY-NEXT: vblendps $63, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm1 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm1 = mem[0,1,2,3,4,5],ymm0[6,7]
@@ -857,7 +857,7 @@ define void @store_i64_stride5_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vmovapd 48(%rsi), %xmm6
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm5 = xmm6[1],xmm5[1]
; AVX1-ONLY-NEXT: vbroadcastsd 56(%rcx), %ymm6
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm6 = ymm5[0,1],ymm6[2],ymm5[3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm6 = ymm5[0,1],ymm6[2,3]
; AVX1-ONLY-NEXT: vmovapd 32(%r8), %ymm0
; AVX1-ONLY-NEXT: vblendpd $14, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm5 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm5 = ymm0[0],mem[1,2,3]
@@ -867,7 +867,7 @@ define void @store_i64_stride5_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm6[0,1,2],ymm0[3]
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm1 = xmm13[0],mem[0]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm15[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm15[4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps 64(%r8), %ymm13
; AVX1-ONLY-NEXT: vblendps $63, (%rsp), %ymm13, %ymm0 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm0 = mem[0,1,2,3,4,5],ymm13[6,7]
@@ -882,7 +882,7 @@ define void @store_i64_stride5_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vmovapd 112(%rsi), %xmm13
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm8 = xmm13[1],xmm8[1]
; AVX1-ONLY-NEXT: vbroadcastsd 120(%rcx), %ymm13
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm8 = ymm8[0,1],ymm13[2],ymm8[3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm8 = ymm8[0,1],ymm13[2,3]
; AVX1-ONLY-NEXT: vmovapd 96(%r8), %ymm13
; AVX1-ONLY-NEXT: vblendpd $13, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm5 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm5 = mem[0],ymm13[1],mem[2,3]
@@ -955,7 +955,7 @@ define void @store_i64_stride5_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
; AVX2-ONLY-NEXT: vbroadcastsd 8(%rsi), %ymm3
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm6[0,1,2,3],ymm3[4,5],ymm6[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm6[0,1,2,3],ymm3[4,5,6,7]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3,4,5],ymm0[6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovddup {{.*#+}} xmm0 = mem[0,0]
@@ -964,7 +964,7 @@ define void @store_i64_stride5_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
; AVX2-ONLY-NEXT: vbroadcastsd 40(%rsi), %ymm3
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm5[0,1,2,3],ymm3[4,5],ymm5[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm5[0,1,2,3],ymm3[4,5,6,7]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3,4,5],ymm0[6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovddup {{.*#+}} xmm0 = mem[0,0]
@@ -973,7 +973,7 @@ define void @store_i64_stride5_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0
; AVX2-ONLY-NEXT: vbroadcastsd 72(%rsi), %ymm3
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm11[0,1,2,3],ymm3[4,5],ymm11[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm11[0,1,2,3],ymm3[4,5,6,7]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3,4,5],ymm0[6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovddup {{.*#+}} xmm0 = mem[0,0]
@@ -983,7 +983,7 @@ define void @store_i64_stride5_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovaps 96(%rdx), %xmm9
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm9, %ymm0, %ymm0
; AVX2-ONLY-NEXT: vbroadcastsd 104(%rsi), %ymm3
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm14[0,1,2,3],ymm3[4,5],ymm14[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm14[0,1,2,3],ymm3[4,5,6,7]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3,4,5],ymm0[6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovddup {{.*#+}} xmm0 = mem[0,0]
@@ -993,11 +993,11 @@ define void @store_i64_stride5_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vbroadcastsd 56(%rsi), %ymm0
; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm0[0,1],mem[2,3]
; AVX2-ONLY-NEXT: vbroadcastsd 56(%rcx), %ymm3
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm13 = ymm0[0,1,2,3],ymm3[4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm13 = ymm0[0,1,2,3],ymm3[4,5,6,7]
; AVX2-ONLY-NEXT: vbroadcastsd 120(%rsi), %ymm3
; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm3 = xmm3[0,1],mem[2,3]
; AVX2-ONLY-NEXT: vbroadcastsd 120(%rcx), %ymm15
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm3[0,1,2,3],ymm15[4,5],ymm3[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm3[0,1,2,3],ymm15[4,5,6,7]
; AVX2-ONLY-NEXT: vmovaps (%rdx), %ymm3
; AVX2-ONLY-NEXT: vmovaps (%rcx), %ymm10
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm15 = ymm3[1],ymm10[1],ymm3[3],ymm10[3]
@@ -1767,7 +1767,7 @@ define void @store_i64_stride5_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps (%rdi), %ymm9
; AVX1-ONLY-NEXT: vbroadcastsd 8(%rsi), %ymm0
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm9[0,1,2,3],ymm0[4,5],ymm9[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm9[0,1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps (%rdx), %xmm1
; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
@@ -1779,7 +1779,7 @@ define void @store_i64_stride5_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovapd 32(%rdi), %ymm14
; AVX1-ONLY-NEXT: vbroadcastsd 40(%rsi), %ymm0
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm14[0,1],ymm0[2],ymm14[3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm14[0,1],ymm0[2,3]
; AVX1-ONLY-NEXT: vmovaps 32(%rdx), %xmm1
; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
@@ -1791,7 +1791,7 @@ define void @store_i64_stride5_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 64(%rdi), %ymm2
; AVX1-ONLY-NEXT: vbroadcastsd 72(%rsi), %ymm0
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps 64(%rdx), %xmm1
; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
@@ -1802,7 +1802,7 @@ define void @store_i64_stride5_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, (%rsp) # 32-byte Spill
; AVX1-ONLY-NEXT: vbroadcastsd 104(%rsi), %ymm0
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm6[0,1,2,3],ymm0[4,5],ymm6[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm6[0,1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps 96(%rdx), %xmm1
; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
@@ -1810,7 +1810,7 @@ define void @store_i64_stride5_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 128(%rdi), %ymm1
; AVX1-ONLY-NEXT: vbroadcastsd 136(%rsi), %ymm0
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps 128(%rdx), %xmm3
; AVX1-ONLY-NEXT: vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm3
@@ -1821,19 +1821,19 @@ define void @store_i64_stride5_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vbroadcastsd 168(%rsi), %ymm0
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm11[0,1],ymm0[2],ymm11[3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm11[0,1],ymm0[2,3]
; AVX1-ONLY-NEXT: vmovaps 160(%rdx), %xmm6
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm6, %ymm0, %ymm3
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm12 = ymm0[0,1,2],ymm3[3]
; AVX1-ONLY-NEXT: vbroadcastsd 200(%rsi), %ymm0
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm4[0,1,2,3],ymm0[4,5],ymm4[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm4[0,1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps %ymm4, %ymm5
; AVX1-ONLY-NEXT: vmovaps 192(%rdx), %xmm3
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm4
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm10 = ymm0[0,1,2,3,4,5],ymm4[6,7]
; AVX1-ONLY-NEXT: vmovapd 224(%rdi), %ymm0
; AVX1-ONLY-NEXT: vbroadcastsd 232(%rsi), %ymm4
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm4 = ymm0[0,1],ymm4[2],ymm0[3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm4 = ymm0[0,1],ymm4[2,3]
; AVX1-ONLY-NEXT: vmovaps 224(%rdx), %xmm7
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm7, %ymm0, %ymm8
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm8 = ymm4[0,1,2],ymm8[3]
@@ -1842,7 +1842,7 @@ define void @store_i64_stride5_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm4 = ymm4[0,1],ymm15[2,3]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm15 = xmm15[0],mem[0]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm9 = ymm15[0,1,2,3,4,5],ymm9[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm9 = ymm15[0,1,2,3],ymm9[4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps (%r8), %ymm15
; AVX1-ONLY-NEXT: vblendps $63, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm13 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm13 = mem[0,1,2,3,4,5],ymm15[6,7]
@@ -1861,7 +1861,7 @@ define void @store_i64_stride5_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vmovapd 48(%rsi), %xmm15
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm14 = xmm15[1],xmm14[1]
; AVX1-ONLY-NEXT: vbroadcastsd 56(%rcx), %ymm15
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm14 = ymm14[0,1],ymm15[2],ymm14[3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm14 = ymm14[0,1],ymm15[2,3]
; AVX1-ONLY-NEXT: vmovapd 32(%r8), %ymm15
; AVX1-ONLY-NEXT: vblendpd $14, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm13 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm13 = ymm15[0],mem[1,2,3]
@@ -1875,7 +1875,7 @@ define void @store_i64_stride5_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vmovupd %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm9 = xmm9[0],mem[0]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm9[0,1,2,3,4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm9[0,1,2,3],ymm2[4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps 64(%r8), %ymm9
; AVX1-ONLY-NEXT: vblendps $63, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm13 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm13 = mem[0,1,2,3,4,5],ymm9[6,7]
@@ -1895,7 +1895,7 @@ define void @store_i64_stride5_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vmovapd 112(%rsi), %xmm14
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm9 = xmm14[1],xmm9[1]
; AVX1-ONLY-NEXT: vbroadcastsd 120(%rcx), %ymm14
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm9 = ymm9[0,1],ymm14[2],ymm9[3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm9 = ymm9[0,1],ymm14[2,3]
; AVX1-ONLY-NEXT: vmovapd 96(%r8), %ymm14
; AVX1-ONLY-NEXT: vblendpd $13, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm13 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm13 = mem[0],ymm14[1],mem[2,3]
@@ -1909,7 +1909,7 @@ define void @store_i64_stride5_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm2 = xmm2[0],mem[0]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps 128(%r8), %ymm2
; AVX1-ONLY-NEXT: vblendps $63, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm9 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm9 = mem[0,1,2,3,4,5],ymm2[6,7]
@@ -1928,7 +1928,7 @@ define void @store_i64_stride5_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vmovapd 176(%rsi), %xmm14
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm2 = xmm14[1],xmm2[1]
; AVX1-ONLY-NEXT: vbroadcastsd 184(%rcx), %ymm14
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm2[0,1],ymm14[2],ymm2[3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm2[0,1],ymm14[2,3]
; AVX1-ONLY-NEXT: vmovapd 160(%r8), %ymm14
; AVX1-ONLY-NEXT: vblendpd $13, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm9 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm9 = mem[0],ymm14[1],mem[2,3]
@@ -1941,7 +1941,7 @@ define void @store_i64_stride5_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm2 = xmm1[0],mem[0]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm5[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm5[4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps 192(%r8), %ymm14
; AVX1-ONLY-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm1 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm1 = mem[0,1],ymm14[2,3],mem[4,5,6,7]
@@ -1959,7 +1959,7 @@ define void @store_i64_stride5_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vmovapd 240(%rsi), %xmm14
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm2 = xmm14[1],xmm2[1]
; AVX1-ONLY-NEXT: vbroadcastsd 248(%rcx), %ymm14
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm2[0,1],ymm14[2],ymm2[3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm2[0,1],ymm14[2,3]
; AVX1-ONLY-NEXT: vmovapd 224(%r8), %ymm14
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm14[0],ymm8[1,2,3]
; AVX1-ONLY-NEXT: vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -2101,7 +2101,7 @@ define void @store_i64_stride5_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
; AVX2-ONLY-NEXT: vbroadcastsd 8(%rsi), %ymm1
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm7[0,1,2,3],ymm1[4,5],ymm7[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm7[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovddup {{.*#+}} xmm0 = mem[0,0]
@@ -2111,7 +2111,7 @@ define void @store_i64_stride5_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0
; AVX2-ONLY-NEXT: vbroadcastsd 40(%rsi), %ymm1
; AVX2-ONLY-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm6[0,1,2,3],ymm1[4,5],ymm6[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm6[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovddup {{.*#+}} xmm0 = mem[0,0]
@@ -2121,7 +2121,7 @@ define void @store_i64_stride5_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm0
; AVX2-ONLY-NEXT: vbroadcastsd 72(%rsi), %ymm1
; AVX2-ONLY-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm5[0,1,2,3],ymm1[4,5],ymm5[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm5[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovddup {{.*#+}} xmm0 = mem[0,0]
@@ -2133,7 +2133,7 @@ define void @store_i64_stride5_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
; AVX2-ONLY-NEXT: vbroadcastsd 104(%rsi), %ymm1
; AVX2-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm4[0,1,2,3],ymm1[4,5],ymm4[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm4[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovddup {{.*#+}} xmm0 = mem[0,0]
@@ -2143,7 +2143,7 @@ define void @store_i64_stride5_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovaps 128(%rdi), %ymm2
; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vbroadcastsd 136(%rsi), %ymm0
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5],ymm2[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5,6,7]
; AVX2-ONLY-NEXT: vmovaps 128(%rdx), %xmm1
; AVX2-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
@@ -2155,7 +2155,7 @@ define void @store_i64_stride5_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovaps 160(%rdi), %ymm9
; AVX2-ONLY-NEXT: vbroadcastsd 168(%rsi), %ymm0
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm9[0,1,2,3],ymm0[4,5],ymm9[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm9[0,1,2,3],ymm0[4,5,6,7]
; AVX2-ONLY-NEXT: vmovaps 160(%rdx), %xmm1
; AVX2-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
@@ -2167,7 +2167,7 @@ define void @store_i64_stride5_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovaps 192(%rdi), %ymm5
; AVX2-ONLY-NEXT: vbroadcastsd 200(%rsi), %ymm0
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm5[0,1,2,3],ymm0[4,5],ymm5[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm5[0,1,2,3],ymm0[4,5,6,7]
; AVX2-ONLY-NEXT: vmovaps 192(%rdx), %xmm1
; AVX2-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm4
@@ -2179,7 +2179,7 @@ define void @store_i64_stride5_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovaps 224(%rdi), %ymm3
; AVX2-ONLY-NEXT: vbroadcastsd 232(%rsi), %ymm4
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm3[0,1,2,3],ymm4[4,5],ymm3[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm3[0,1,2,3],ymm4[4,5,6,7]
; AVX2-ONLY-NEXT: vmovaps 224(%rdx), %xmm0
; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm6
@@ -2192,22 +2192,22 @@ define void @store_i64_stride5_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vbroadcastsd 56(%rsi), %ymm4
; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm4 = xmm4[0,1],mem[2,3]
; AVX2-ONLY-NEXT: vbroadcastsd 56(%rcx), %ymm6
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm4[0,1,2,3],ymm6[4,5],ymm4[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm4[0,1,2,3],ymm6[4,5,6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, (%rsp) # 32-byte Spill
; AVX2-ONLY-NEXT: vbroadcastsd 120(%rsi), %ymm4
; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm4 = xmm4[0,1],mem[2,3]
; AVX2-ONLY-NEXT: vbroadcastsd 120(%rcx), %ymm6
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm4[0,1,2,3],ymm6[4,5],ymm4[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm4[0,1,2,3],ymm6[4,5,6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vbroadcastsd 184(%rsi), %ymm4
; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm4 = xmm4[0,1],mem[2,3]
; AVX2-ONLY-NEXT: vbroadcastsd 184(%rcx), %ymm6
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm4[0,1,2,3],ymm6[4,5],ymm4[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm4[0,1,2,3],ymm6[4,5,6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vbroadcastsd 248(%rsi), %ymm6
; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm6 = xmm6[0,1],mem[2,3]
; AVX2-ONLY-NEXT: vbroadcastsd 248(%rcx), %ymm13
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm6[0,1,2,3],ymm13[4,5],ymm6[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm6[0,1,2,3],ymm13[4,5,6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovaps (%rdx), %ymm13
; AVX2-ONLY-NEXT: vmovaps (%rcx), %ymm14
@@ -3815,7 +3815,7 @@ define void @store_i64_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps (%rdi), %ymm9
; AVX1-ONLY-NEXT: vbroadcastsd 8(%rsi), %ymm0
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm9[0,1,2,3],ymm0[4,5],ymm9[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm9[0,1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps (%rdx), %xmm1
; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
@@ -3827,7 +3827,7 @@ define void @store_i64_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovapd 32(%rdi), %ymm7
; AVX1-ONLY-NEXT: vbroadcastsd 40(%rsi), %ymm0
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm7[0,1],ymm0[2],ymm7[3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm7[0,1],ymm0[2,3]
; AVX1-ONLY-NEXT: vmovaps 32(%rdx), %xmm1
; AVX1-ONLY-NEXT: vmovaps %xmm1, (%rsp) # 16-byte Spill
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
@@ -3839,7 +3839,7 @@ define void @store_i64_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 64(%rdi), %ymm6
; AVX1-ONLY-NEXT: vbroadcastsd 72(%rsi), %ymm0
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm6[0,1,2,3],ymm0[4,5],ymm6[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm6[0,1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps 64(%rdx), %xmm1
; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
@@ -3850,7 +3850,7 @@ define void @store_i64_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vbroadcastsd 104(%rsi), %ymm0
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3],ymm0[4,5],ymm3[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps 96(%rdx), %xmm1
; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
@@ -3858,7 +3858,7 @@ define void @store_i64_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 128(%rdi), %ymm5
; AVX1-ONLY-NEXT: vbroadcastsd 136(%rsi), %ymm0
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm5[0,1,2,3],ymm0[4,5],ymm5[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm5[0,1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps 128(%rdx), %xmm1
; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
@@ -3869,14 +3869,14 @@ define void @store_i64_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vbroadcastsd 168(%rsi), %ymm0
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps 160(%rdx), %xmm1
; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vbroadcastsd 200(%rsi), %ymm0
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm4[0,1,2,3],ymm0[4,5],ymm4[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm4[0,1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps 192(%rdx), %xmm1
; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
@@ -3884,7 +3884,7 @@ define void @store_i64_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovapd 224(%rdi), %ymm4
; AVX1-ONLY-NEXT: vbroadcastsd 232(%rsi), %ymm0
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm4[0,1],ymm0[2],ymm4[3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm4[0,1],ymm0[2,3]
; AVX1-ONLY-NEXT: vmovaps 224(%rdx), %xmm1
; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
@@ -3896,7 +3896,7 @@ define void @store_i64_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 256(%rdi), %ymm3
; AVX1-ONLY-NEXT: vbroadcastsd 264(%rsi), %ymm0
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3],ymm0[4,5],ymm3[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps 256(%rdx), %xmm1
; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
@@ -3907,14 +3907,14 @@ define void @store_i64_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vbroadcastsd 296(%rsi), %ymm0
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm10[0,1,2,3],ymm0[4,5],ymm10[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm10[0,1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps 288(%rdx), %xmm1
; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vbroadcastsd 328(%rsi), %ymm0
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm8[0,1,2,3],ymm0[4,5],ymm8[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm8[0,1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps 320(%rdx), %xmm1
; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
@@ -3922,7 +3922,7 @@ define void @store_i64_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovapd 352(%rdi), %ymm2
; AVX1-ONLY-NEXT: vbroadcastsd 360(%rsi), %ymm0
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm2[0,1],ymm0[2],ymm2[3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3]
; AVX1-ONLY-NEXT: vmovaps 352(%rdx), %xmm1
; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
@@ -3933,7 +3933,7 @@ define void @store_i64_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3]
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vbroadcastsd 392(%rsi), %ymm0
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm11[0,1,2,3],ymm0[4,5],ymm11[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm11[0,1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps 384(%rdx), %xmm1
; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
@@ -3941,7 +3941,7 @@ define void @store_i64_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovapd 416(%rdi), %ymm1
; AVX1-ONLY-NEXT: vbroadcastsd 424(%rsi), %ymm0
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2],ymm1[3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3]
; AVX1-ONLY-NEXT: vmovaps 416(%rdx), %xmm8
; AVX1-ONLY-NEXT: vmovaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm8, %ymm0, %ymm8
@@ -3953,7 +3953,7 @@ define void @store_i64_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 448(%rdi), %ymm0
; AVX1-ONLY-NEXT: vbroadcastsd 456(%rsi), %ymm8
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm0[0,1,2,3],ymm8[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm0[0,1,2,3],ymm8[4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps 448(%rdx), %xmm10
; AVX1-ONLY-NEXT: vmovaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm10, %ymm0, %ymm10
@@ -3962,14 +3962,14 @@ define void @store_i64_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm10 = ymm0[0],mem[0],ymm0[2],mem[2]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm12 = ymm8[0,1,2,3],ymm10[4,5,6,7]
; AVX1-ONLY-NEXT: vbroadcastsd 488(%rsi), %ymm8
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm10 = ymm15[0,1],ymm8[2],ymm15[3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm10 = ymm15[0,1],ymm8[2,3]
; AVX1-ONLY-NEXT: vmovapd %ymm15, %ymm11
; AVX1-ONLY-NEXT: vmovaps 480(%rdx), %xmm8
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm8, %ymm0, %ymm15
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm10 = ymm10[0,1,2],ymm15[3]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm15 = xmm15[0],mem[0]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm9 = ymm15[0,1,2,3,4,5],ymm9[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm9 = ymm15[0,1,2,3],ymm9[4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps (%r8), %ymm15
; AVX1-ONLY-NEXT: vblendps $63, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm14 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm14 = mem[0,1,2,3,4,5],ymm15[6,7]
@@ -3988,7 +3988,7 @@ define void @store_i64_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vmovapd 48(%rsi), %xmm15
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm9 = xmm15[1],xmm9[1]
; AVX1-ONLY-NEXT: vbroadcastsd 56(%rcx), %ymm15
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm9 = ymm9[0,1],ymm15[2],ymm9[3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm9 = ymm9[0,1],ymm15[2,3]
; AVX1-ONLY-NEXT: vmovapd 32(%r8), %ymm15
; AVX1-ONLY-NEXT: vblendpd $14, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm14 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm14 = ymm15[0],mem[1,2,3]
@@ -4002,7 +4002,7 @@ define void @store_i64_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vmovupd %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm7 = xmm7[0],mem[0]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm7[0,1,2,3,4,5],ymm6[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm7[0,1,2,3],ymm6[4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps 64(%r8), %ymm7
; AVX1-ONLY-NEXT: vblendps $63, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm9 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm9 = mem[0,1,2,3,4,5],ymm7[6,7]
@@ -4022,7 +4022,7 @@ define void @store_i64_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vmovapd 112(%rsi), %xmm9
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm7 = xmm9[1],xmm7[1]
; AVX1-ONLY-NEXT: vbroadcastsd 120(%rcx), %ymm9
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm7 = ymm7[0,1],ymm9[2],ymm7[3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm7 = ymm7[0,1],ymm9[2,3]
; AVX1-ONLY-NEXT: vmovapd 96(%r8), %ymm9
; AVX1-ONLY-NEXT: vblendpd $13, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm14 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm14 = mem[0],ymm9[1],mem[2,3]
@@ -4036,7 +4036,7 @@ define void @store_i64_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vmovupd %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm6 = xmm6[0],mem[0]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm6[0,1,2,3,4,5],ymm5[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm6[0,1,2,3],ymm5[4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps 128(%r8), %ymm6
; AVX1-ONLY-NEXT: vblendps $63, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm7 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm7 = mem[0,1,2,3,4,5],ymm6[6,7]
@@ -4056,7 +4056,7 @@ define void @store_i64_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vmovapd 176(%rsi), %xmm7
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm6 = xmm7[1],xmm6[1]
; AVX1-ONLY-NEXT: vbroadcastsd 184(%rcx), %ymm7
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm6 = ymm6[0,1],ymm7[2],ymm6[3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm6 = ymm6[0,1],ymm7[2,3]
; AVX1-ONLY-NEXT: vmovapd 160(%r8), %ymm7
; AVX1-ONLY-NEXT: vblendpd $13, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm9 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm9 = mem[0],ymm7[1],mem[2,3]
@@ -4070,8 +4070,8 @@ define void @store_i64_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vmovupd %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm5 = xmm5[0],mem[0]
-; AVX1-ONLY-NEXT: vblendps $192, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm5 = ymm5[0,1,2,3,4,5],mem[6,7]
+; AVX1-ONLY-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm5 = ymm5[0,1,2,3],mem[4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps 192(%r8), %ymm6
; AVX1-ONLY-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm7 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm7 = mem[0,1],ymm6[2,3],mem[4,5,6,7]
@@ -4090,7 +4090,7 @@ define void @store_i64_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vmovapd 240(%rsi), %xmm6
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm5 = xmm6[1],xmm5[1]
; AVX1-ONLY-NEXT: vbroadcastsd 248(%rcx), %ymm6
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm5 = ymm5[0,1],ymm6[2],ymm5[3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm5 = ymm5[0,1],ymm6[2,3]
; AVX1-ONLY-NEXT: vmovapd 224(%r8), %ymm6
; AVX1-ONLY-NEXT: vblendpd $14, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm7 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm7 = ymm6[0],mem[1,2,3]
@@ -4104,7 +4104,7 @@ define void @store_i64_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vmovupd %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm4 = xmm4[0],mem[0]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm4[0,1,2,3,4,5],ymm3[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm4[0,1,2,3],ymm3[4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps 256(%r8), %ymm4
; AVX1-ONLY-NEXT: vblendps $63, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm5 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm5 = mem[0,1,2,3,4,5],ymm4[6,7]
@@ -4124,7 +4124,7 @@ define void @store_i64_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vmovapd 304(%rsi), %xmm5
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm4 = xmm5[1],xmm4[1]
; AVX1-ONLY-NEXT: vbroadcastsd 312(%rcx), %ymm5
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm4 = ymm4[0,1],ymm5[2],ymm4[3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm4 = ymm4[0,1],ymm5[2,3]
; AVX1-ONLY-NEXT: vmovapd 288(%r8), %ymm5
; AVX1-ONLY-NEXT: vblendpd $13, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm6 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm6 = mem[0],ymm5[1],mem[2,3]
@@ -4138,8 +4138,8 @@ define void @store_i64_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vmovupd %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm3 = xmm3[0],mem[0]
-; AVX1-ONLY-NEXT: vblendps $192, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm3 = ymm3[0,1,2,3,4,5],mem[6,7]
+; AVX1-ONLY-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm3 = ymm3[0,1,2,3],mem[4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps 320(%r8), %ymm4
; AVX1-ONLY-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm5 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm5 = mem[0,1],ymm4[2,3],mem[4,5,6,7]
@@ -4158,7 +4158,7 @@ define void @store_i64_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vmovapd 368(%rsi), %xmm4
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm3 = xmm4[1],xmm3[1]
; AVX1-ONLY-NEXT: vbroadcastsd 376(%rcx), %ymm4
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm3 = ymm3[0,1],ymm4[2],ymm3[3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm3 = ymm3[0,1],ymm4[2,3]
; AVX1-ONLY-NEXT: vmovapd 352(%r8), %ymm4
; AVX1-ONLY-NEXT: vblendpd $14, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm5 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm5 = ymm4[0],mem[1,2,3]
@@ -4172,8 +4172,8 @@ define void @store_i64_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm2 = xmm2[0],mem[0]
-; AVX1-ONLY-NEXT: vblendps $192, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm2 = ymm2[0,1,2,3,4,5],mem[6,7]
+; AVX1-ONLY-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm2 = ymm2[0,1,2,3],mem[4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps 384(%r8), %ymm3
; AVX1-ONLY-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm4 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm4 = mem[0,1],ymm3[2,3],mem[4,5,6,7]
@@ -4192,7 +4192,7 @@ define void @store_i64_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vmovapd 432(%rsi), %xmm3
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm2 = xmm3[1],xmm2[1]
; AVX1-ONLY-NEXT: vbroadcastsd 440(%rcx), %ymm3
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm2[0,1],ymm3[2],ymm2[3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm2[0,1],ymm3[2,3]
; AVX1-ONLY-NEXT: vmovapd 416(%r8), %ymm3
; AVX1-ONLY-NEXT: vblendpd $14, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm4 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm4 = ymm3[0],mem[1,2,3]
@@ -4206,7 +4206,7 @@ define void @store_i64_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm1 = xmm1[0],mem[0]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps 448(%r8), %ymm1
; AVX1-ONLY-NEXT: vblendps $63, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm2 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm2 = mem[0,1,2,3,4,5],ymm1[6,7]
@@ -4223,7 +4223,7 @@ define void @store_i64_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vmovapd 496(%rsi), %xmm2
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm2[1],xmm1[1]
; AVX1-ONLY-NEXT: vbroadcastsd 504(%rcx), %ymm2
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm1[0,1],ymm2[2],ymm1[3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm1[0,1],ymm2[2,3]
; AVX1-ONLY-NEXT: vmovapd 480(%r8), %ymm2
; AVX1-ONLY-NEXT: vblendpd $13, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm3 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm3 = mem[0],ymm2[1],mem[2,3]
@@ -4516,7 +4516,7 @@ define void @store_i64_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
; AVX2-ONLY-NEXT: vbroadcastsd 8(%rsi), %ymm1
; AVX2-ONLY-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm7[0,1,2,3],ymm1[4,5],ymm7[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm7[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovddup {{.*#+}} xmm0 = mem[0,0]
@@ -4526,7 +4526,7 @@ define void @store_i64_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0
; AVX2-ONLY-NEXT: vbroadcastsd 40(%rsi), %ymm1
; AVX2-ONLY-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm6[0,1,2,3],ymm1[4,5],ymm6[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm6[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovddup {{.*#+}} xmm0 = mem[0,0]
@@ -4536,7 +4536,7 @@ define void @store_i64_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm0
; AVX2-ONLY-NEXT: vbroadcastsd 72(%rsi), %ymm1
; AVX2-ONLY-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm5[0,1,2,3],ymm1[4,5],ymm5[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm5[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovddup {{.*#+}} xmm0 = mem[0,0]
@@ -4548,7 +4548,7 @@ define void @store_i64_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
; AVX2-ONLY-NEXT: vbroadcastsd 104(%rsi), %ymm1
; AVX2-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm4[0,1,2,3],ymm1[4,5],ymm4[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm4[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovddup {{.*#+}} xmm0 = mem[0,0]
@@ -4558,7 +4558,7 @@ define void @store_i64_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovaps 128(%rdi), %ymm2
; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vbroadcastsd 136(%rsi), %ymm0
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5],ymm2[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5,6,7]
; AVX2-ONLY-NEXT: vmovaps 128(%rdx), %xmm1
; AVX2-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
@@ -4571,7 +4571,7 @@ define void @store_i64_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovaps 160(%rdi), %ymm2
; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vbroadcastsd 168(%rsi), %ymm0
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5],ymm2[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5,6,7]
; AVX2-ONLY-NEXT: vmovaps 160(%rdx), %xmm1
; AVX2-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
@@ -4584,7 +4584,7 @@ define void @store_i64_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovaps 192(%rdi), %ymm2
; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vbroadcastsd 200(%rsi), %ymm0
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5],ymm2[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5,6,7]
; AVX2-ONLY-NEXT: vmovaps 192(%rdx), %xmm1
; AVX2-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
@@ -4597,7 +4597,7 @@ define void @store_i64_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovaps 224(%rdi), %ymm2
; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vbroadcastsd 232(%rsi), %ymm0
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5],ymm2[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5,6,7]
; AVX2-ONLY-NEXT: vmovaps 224(%rdx), %xmm1
; AVX2-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
@@ -4610,7 +4610,7 @@ define void @store_i64_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovaps 256(%rdi), %ymm2
; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vbroadcastsd 264(%rsi), %ymm0
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5],ymm2[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5,6,7]
; AVX2-ONLY-NEXT: vmovaps 256(%rdx), %xmm1
; AVX2-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
@@ -4623,7 +4623,7 @@ define void @store_i64_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovaps 288(%rdi), %ymm2
; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vbroadcastsd 296(%rsi), %ymm0
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5],ymm2[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5,6,7]
; AVX2-ONLY-NEXT: vmovaps 288(%rdx), %xmm1
; AVX2-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
@@ -4636,7 +4636,7 @@ define void @store_i64_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovaps 320(%rdi), %ymm2
; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vbroadcastsd 328(%rsi), %ymm0
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5],ymm2[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5,6,7]
; AVX2-ONLY-NEXT: vmovaps 320(%rdx), %xmm1
; AVX2-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
@@ -4649,7 +4649,7 @@ define void @store_i64_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovaps 352(%rdi), %ymm2
; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vbroadcastsd 360(%rsi), %ymm0
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5],ymm2[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5,6,7]
; AVX2-ONLY-NEXT: vmovaps 352(%rdx), %xmm1
; AVX2-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
@@ -4662,7 +4662,7 @@ define void @store_i64_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovaps 384(%rdi), %ymm2
; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vbroadcastsd 392(%rsi), %ymm0
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5],ymm2[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5,6,7]
; AVX2-ONLY-NEXT: vmovaps 384(%rdx), %xmm1
; AVX2-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
@@ -4675,7 +4675,7 @@ define void @store_i64_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovaps 416(%rdi), %ymm2
; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vbroadcastsd 424(%rsi), %ymm0
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5],ymm2[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5,6,7]
; AVX2-ONLY-NEXT: vmovaps 416(%rdx), %xmm1
; AVX2-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
@@ -4688,7 +4688,7 @@ define void @store_i64_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovaps 448(%rdi), %ymm2
; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vbroadcastsd 456(%rsi), %ymm0
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5],ymm2[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5,6,7]
; AVX2-ONLY-NEXT: vmovaps 448(%rdx), %xmm1
; AVX2-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
@@ -4701,7 +4701,7 @@ define void @store_i64_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovaps 480(%rdi), %ymm2
; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vbroadcastsd 488(%rsi), %ymm0
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5],ymm2[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5,6,7]
; AVX2-ONLY-NEXT: vmovaps 480(%rdx), %xmm1
; AVX2-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
@@ -4714,42 +4714,42 @@ define void @store_i64_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vbroadcastsd 56(%rsi), %ymm0
; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm0[0,1],mem[2,3]
; AVX2-ONLY-NEXT: vbroadcastsd 56(%rcx), %ymm1
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vbroadcastsd 120(%rsi), %ymm0
; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm0[0,1],mem[2,3]
; AVX2-ONLY-NEXT: vbroadcastsd 120(%rcx), %ymm1
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vbroadcastsd 184(%rsi), %ymm0
; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm0[0,1],mem[2,3]
; AVX2-ONLY-NEXT: vbroadcastsd 184(%rcx), %ymm1
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vbroadcastsd 248(%rsi), %ymm0
; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm0[0,1],mem[2,3]
; AVX2-ONLY-NEXT: vbroadcastsd 248(%rcx), %ymm1
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vbroadcastsd 312(%rsi), %ymm0
; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm0[0,1],mem[2,3]
; AVX2-ONLY-NEXT: vbroadcastsd 312(%rcx), %ymm1
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vbroadcastsd 376(%rsi), %ymm0
; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm0[0,1],mem[2,3]
; AVX2-ONLY-NEXT: vbroadcastsd 376(%rcx), %ymm1
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vbroadcastsd 440(%rsi), %ymm0
; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm0[0,1],mem[2,3]
; AVX2-ONLY-NEXT: vbroadcastsd 440(%rcx), %ymm1
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vbroadcastsd 504(%rsi), %ymm0
; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm0[0,1],mem[2,3]
; AVX2-ONLY-NEXT: vbroadcastsd 504(%rcx), %ymm1
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovaps (%rdx), %ymm0
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-store-i64-stride-6.ll b/llvm/test/CodeGen/X86/vector-interleaved-store-i64-stride-6.ll
index 04a826d281d00..825abedac4e63 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-store-i64-stride-6.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-store-i64-stride-6.ll
@@ -183,7 +183,7 @@ define void @store_i64_stride6_vf4(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
; AVX1-ONLY-NEXT: vmovaps (%rdi), %xmm7
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm8 = xmm7[1],xmm6[1]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm8, %ymm0, %ymm8
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm8 = ymm3[0],ymm8[1,2,3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm8 = ymm3[0,1],ymm8[2,3]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm5 = ymm8[0],ymm5[1],ymm8[2,3]
; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm2 = ymm2[0],mem[0],ymm2[2],mem[2]
; AVX1-ONLY-NEXT: vmovaps 16(%rdi), %xmm8
@@ -193,7 +193,7 @@ define void @store_i64_stride6_vf4(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
; AVX1-ONLY-NEXT: vinsertf128 $1, (%r9), %ymm8, %ymm9
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm10 = mem[2,3,2,3]
; AVX1-ONLY-NEXT: vbroadcastsd 8(%r8), %ymm11
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm10 = ymm10[0,1,2,3],ymm11[4,5],ymm10[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm10 = ymm10[0,1,2,3],ymm11[4,5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm9 = ymm10[0,1],ymm9[2,3],ymm10[4,5],ymm9[6,7]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm10 = ymm4[2,3],ymm1[2,3]
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
@@ -202,7 +202,7 @@ define void @store_i64_stride6_vf4(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
; AVX1-ONLY-NEXT: vmovapd 16(%rdx), %xmm1
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm1[1],mem[1]
; AVX1-ONLY-NEXT: vbroadcastsd 24(%r8), %ymm3
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm1[0,1],ymm3[2],ymm1[3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm1[0,1],ymm3[2,3]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm1[0,1,2],ymm4[3]
; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm3 = xmm7[0],xmm6[0]
; AVX1-ONLY-NEXT: vmovaps (%rdx), %xmm4
@@ -237,7 +237,7 @@ define void @store_i64_stride6_vf4(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
; AVX2-ONLY-NEXT: vmovaps (%rdx), %xmm10
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm11 = xmm10[1],xmm9[1]
; AVX2-ONLY-NEXT: vbroadcastsd 8(%r8), %ymm12
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm11 = ymm11[0,1,2,3],ymm12[4,5],ymm11[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm11 = ymm11[0,1,2,3],ymm12[4,5,6,7]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm11[0,1,2,3,4,5],ymm5[6,7]
; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm11 = ymm2[0],ymm3[0],ymm2[2],ymm3[2]
; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm12 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
@@ -418,93 +418,89 @@ define void @store_i64_stride6_vf8(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
;
; AVX1-ONLY-LABEL: store_i64_stride6_vf8:
; AVX1-ONLY: # %bb.0:
-; AVX1-ONLY-NEXT: vmovapd 32(%rdi), %ymm8
-; AVX1-ONLY-NEXT: vmovapd (%r8), %ymm7
-; AVX1-ONLY-NEXT: vmovapd 32(%r8), %ymm9
+; AVX1-ONLY-NEXT: vmovapd (%rdi), %ymm9
+; AVX1-ONLY-NEXT: vmovapd 32(%rdi), %ymm13
+; AVX1-ONLY-NEXT: vmovapd (%rsi), %ymm11
+; AVX1-ONLY-NEXT: vmovapd (%r8), %ymm12
+; AVX1-ONLY-NEXT: vmovapd 32(%r8), %ymm14
; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm0 = mem[0,0]
; AVX1-ONLY-NEXT: vmovaps (%rsi), %xmm1
-; AVX1-ONLY-NEXT: vmovaps 16(%rsi), %xmm10
; AVX1-ONLY-NEXT: vmovaps 32(%rsi), %xmm4
-; AVX1-ONLY-NEXT: vmovaps (%rdi), %xmm2
-; AVX1-ONLY-NEXT: vmovaps 16(%rdi), %xmm11
-; AVX1-ONLY-NEXT: vmovaps 32(%rdi), %xmm5
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm3 = xmm2[1],xmm1[1]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm3
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm3 = ymm7[0],ymm3[1,2,3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm3[0],ymm0[1],ymm3[2,3]
+; AVX1-ONLY-NEXT: vmovaps (%rdi), %xmm3
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm2 = xmm3[1],xmm1[1]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm2
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm12[0,1],ymm2[2,3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm2[0],ymm0[1],ymm2[2,3]
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 48(%rsi), %xmm3
-; AVX1-ONLY-NEXT: vmovaps 48(%rdi), %xmm6
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm3 = ymm6[0],ymm3[0],ymm6[2],ymm3[2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],mem[4,5],ymm3[6,7]
-; AVX1-ONLY-NEXT: vbroadcastsd 48(%rcx), %ymm6
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3,4,5],ymm6[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm6 = mem[0,0]
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm12 = xmm5[1],xmm4[1]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm12, %ymm0, %ymm12
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm12 = ymm9[0],ymm12[1,2,3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm6 = ymm12[0],ymm6[1],ymm12[2,3]
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm10 = ymm11[0],ymm10[0],ymm11[2],ymm10[2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm10 = ymm10[0,1,2,3],mem[4,5],ymm10[6,7]
-; AVX1-ONLY-NEXT: vbroadcastsd 16(%rcx), %ymm11
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm10 = ymm10[0,1,2,3,4,5],ymm11[6,7]
-; AVX1-ONLY-NEXT: vmovaps (%rcx), %xmm12
-; AVX1-ONLY-NEXT: vmovaps 32(%rcx), %xmm13
-; AVX1-ONLY-NEXT: vinsertf128 $1, 32(%r9), %ymm13, %ymm11
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm14 = mem[2,3,2,3]
+; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm2 = mem[0,0]
+; AVX1-ONLY-NEXT: vmovaps 32(%rdi), %xmm5
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm6 = xmm5[1],xmm4[1]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm6, %ymm0, %ymm6
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm6 = ymm14[0,1],ymm6[2,3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm6[0],ymm2[1],ymm6[2,3]
+; AVX1-ONLY-NEXT: vmovaps (%rcx), %xmm7
+; AVX1-ONLY-NEXT: vmovaps 32(%rcx), %xmm8
+; AVX1-ONLY-NEXT: vinsertf128 $1, 32(%r9), %ymm8, %ymm6
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm10 = mem[2,3,2,3]
; AVX1-ONLY-NEXT: vbroadcastsd 40(%r8), %ymm15
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3],ymm15[4,5],ymm14[6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm11 = ymm14[0,1],ymm11[2,3],ymm14[4,5],ymm11[6,7]
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm14 = mem[2,3,2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm10 = ymm10[0,1,2,3],ymm15[4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm10[0,1],ymm6[2,3],ymm10[4,5],ymm6[6,7]
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm10 = mem[2,3,2,3]
; AVX1-ONLY-NEXT: vbroadcastsd 8(%r8), %ymm15
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3],ymm15[4,5],ymm14[6,7]
-; AVX1-ONLY-NEXT: vinsertf128 $1, (%r9), %ymm12, %ymm15
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1],ymm15[2,3],ymm14[4,5],ymm15[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm10 = ymm10[0,1,2,3],ymm15[4,5,6,7]
+; AVX1-ONLY-NEXT: vinsertf128 $1, (%r9), %ymm7, %ymm15
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm10 = ymm10[0,1],ymm15[2,3],ymm10[4,5],ymm15[6,7]
; AVX1-ONLY-NEXT: vmovapd 32(%rsi), %ymm15
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm8 = ymm8[1],ymm15[1],ymm8[3],ymm15[3]
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm8 = ymm9[2,3],ymm8[2,3]
-; AVX1-ONLY-NEXT: vmovapd 32(%r9), %ymm9
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm15 = ymm9[2,3],ymm15[2,3]
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm8 = ymm8[0],ymm15[0],ymm8[2],ymm15[3]
-; AVX1-ONLY-NEXT: vmovapd (%rdi), %ymm15
-; AVX1-ONLY-NEXT: vmovapd (%rsi), %ymm0
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm15 = ymm15[1],ymm0[1],ymm15[3],ymm0[3]
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm7 = ymm7[2,3],ymm15[2,3]
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm13 = ymm13[1],ymm15[1],ymm13[3],ymm15[3]
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm13 = ymm14[2,3],ymm13[2,3]
+; AVX1-ONLY-NEXT: vmovapd 32(%r9), %ymm14
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm15 = ymm14[2,3],ymm15[2,3]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm13 = ymm13[0],ymm15[0],ymm13[2],ymm15[3]
; AVX1-ONLY-NEXT: vmovapd (%r9), %ymm15
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm15[2,3],ymm0[2,3]
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm0 = ymm7[0],ymm0[0],ymm7[2],ymm0[3]
-; AVX1-ONLY-NEXT: vmovapd 48(%rdx), %xmm7
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm7 = xmm7[1],mem[1]
-; AVX1-ONLY-NEXT: vbroadcastsd 56(%r8), %ymm3
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm3 = ymm7[0,1],ymm3[2],ymm7[3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm3 = ymm3[0,1,2],ymm9[3]
-; AVX1-ONLY-NEXT: vmovapd 16(%rdx), %xmm7
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm7 = xmm7[1],mem[1]
-; AVX1-ONLY-NEXT: vbroadcastsd 24(%r8), %ymm9
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm7 = ymm7[0,1],ymm9[2],ymm7[3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm7 = ymm7[0,1,2],ymm15[3]
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm9 = ymm9[1],ymm11[1],ymm9[3],ymm11[3]
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm11 = ymm15[2,3],ymm11[2,3]
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm9 = ymm12[2,3],ymm9[2,3]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm11 = ymm9[0],ymm11[0],ymm9[2],ymm11[3]
+; AVX1-ONLY-NEXT: vmovapd 48(%rdx), %xmm9
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm9 = xmm9[1],mem[1]
+; AVX1-ONLY-NEXT: vbroadcastsd 56(%r8), %ymm12
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm9 = ymm9[0,1],ymm12[2,3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm9 = ymm9[0,1,2],ymm14[3]
+; AVX1-ONLY-NEXT: vmovaps 48(%rdi), %xmm12
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm12 = xmm12[0],mem[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vbroadcastsd 48(%rcx), %ymm14
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3,4,5],ymm14[6,7]
+; AVX1-ONLY-NEXT: vmovapd 16(%rdx), %xmm14
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm14 = xmm14[1],mem[1]
+; AVX1-ONLY-NEXT: vbroadcastsd 24(%r8), %ymm0
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm14[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1,2],ymm15[3]
+; AVX1-ONLY-NEXT: vmovaps 16(%rdi), %xmm14
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm14 = xmm14[0],mem[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vbroadcastsd 16(%rcx), %ymm15
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4,5],ymm15[6,7]
; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm4 = xmm5[0],xmm4[0]
; AVX1-ONLY-NEXT: vmovaps 32(%rdx), %xmm5
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm5 = xmm5[0],xmm13[0]
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm1 = xmm2[0],xmm1[0]
-; AVX1-ONLY-NEXT: vmovaps (%rdx), %xmm2
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm2[0],xmm12[0]
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm5 = xmm5[0],xmm8[0]
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm1 = xmm3[0],xmm1[0]
+; AVX1-ONLY-NEXT: vmovaps (%rdx), %xmm3
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm3 = xmm3[0],xmm7[0]
; AVX1-ONLY-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; AVX1-ONLY-NEXT: vmovaps %xmm2, 16(%rax)
+; AVX1-ONLY-NEXT: vmovaps %xmm3, 16(%rax)
; AVX1-ONLY-NEXT: vmovaps %xmm1, (%rax)
; AVX1-ONLY-NEXT: vmovaps %xmm5, 208(%rax)
; AVX1-ONLY-NEXT: vmovaps %xmm4, 192(%rax)
-; AVX1-ONLY-NEXT: vmovapd %ymm0, 128(%rax)
-; AVX1-ONLY-NEXT: vmovapd %ymm8, 320(%rax)
-; AVX1-ONLY-NEXT: vmovaps %ymm14, 64(%rax)
-; AVX1-ONLY-NEXT: vmovaps %ymm11, 256(%rax)
-; AVX1-ONLY-NEXT: vmovaps %ymm10, 96(%rax)
-; AVX1-ONLY-NEXT: vmovapd %ymm7, 160(%rax)
-; AVX1-ONLY-NEXT: vmovapd %ymm6, 224(%rax)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm0, 288(%rax)
-; AVX1-ONLY-NEXT: vmovapd %ymm3, 352(%rax)
+; AVX1-ONLY-NEXT: vmovapd %ymm11, 128(%rax)
+; AVX1-ONLY-NEXT: vmovapd %ymm13, 320(%rax)
+; AVX1-ONLY-NEXT: vmovaps %ymm10, 64(%rax)
+; AVX1-ONLY-NEXT: vmovaps %ymm6, 256(%rax)
+; AVX1-ONLY-NEXT: vmovaps %ymm14, 96(%rax)
+; AVX1-ONLY-NEXT: vmovapd %ymm0, 160(%rax)
+; AVX1-ONLY-NEXT: vmovapd %ymm2, 224(%rax)
+; AVX1-ONLY-NEXT: vmovaps %ymm12, 288(%rax)
+; AVX1-ONLY-NEXT: vmovapd %ymm9, 352(%rax)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm0, 32(%rax)
; AVX1-ONLY-NEXT: vzeroupper
@@ -535,12 +531,12 @@ define void @store_i64_stride6_vf8(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
; AVX2-ONLY-NEXT: vmovaps 32(%rdx), %xmm11
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm14 = xmm11[1],xmm10[1]
; AVX2-ONLY-NEXT: vbroadcastsd 40(%r8), %ymm15
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3],ymm15[4,5],ymm14[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3],ymm15[4,5,6,7]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm14[0,1,2,3,4,5],ymm5[6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm14 = xmm8[1],xmm6[1]
; AVX2-ONLY-NEXT: vbroadcastsd 8(%r8), %ymm15
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3],ymm15[4,5],ymm14[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3],ymm15[4,5,6,7]
; AVX2-ONLY-NEXT: vmovaps 32(%rdi), %xmm15
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm9, %ymm0, %ymm9
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm14[0,1,2,3,4,5],ymm9[6,7]
@@ -1037,181 +1033,179 @@ define void @store_i64_stride6_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-LABEL: store_i64_stride6_vf16:
; AVX1-ONLY: # %bb.0:
; AVX1-ONLY-NEXT: subq $376, %rsp # imm = 0x178
-; AVX1-ONLY-NEXT: vmovapd (%r8), %ymm0
-; AVX1-ONLY-NEXT: vmovapd 32(%r8), %ymm15
-; AVX1-ONLY-NEXT: vmovapd 64(%r8), %ymm9
-; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm1 = mem[0,0]
+; AVX1-ONLY-NEXT: vmovapd (%r8), %ymm1
+; AVX1-ONLY-NEXT: vmovapd 32(%r8), %ymm0
+; AVX1-ONLY-NEXT: vmovapd 64(%r8), %ymm4
+; AVX1-ONLY-NEXT: vmovapd 96(%r8), %ymm7
+; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm2 = mem[0,0]
; AVX1-ONLY-NEXT: vmovaps (%rsi), %xmm5
; AVX1-ONLY-NEXT: vmovaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 16(%rsi), %xmm2
; AVX1-ONLY-NEXT: vmovaps 32(%rsi), %xmm6
; AVX1-ONLY-NEXT: vmovaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps (%rdi), %xmm4
-; AVX1-ONLY-NEXT: vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 16(%rdi), %xmm3
-; AVX1-ONLY-NEXT: vmovaps 32(%rdi), %xmm7
-; AVX1-ONLY-NEXT: vmovaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm4 = xmm4[1],xmm5[1]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm4, %ymm0, %ymm4
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm4 = ymm0[0],ymm4[1,2,3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm4[0],ymm1[1],ymm4[2,3]
-; AVX1-ONLY-NEXT: vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm3[0],ymm2[0],ymm3[2],ymm2[2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],mem[4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT: vbroadcastsd 16(%rcx), %ymm2
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm1 = mem[0,0]
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm2 = xmm7[1],xmm6[1]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm2
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm15[0],ymm2[1,2,3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm2[0],ymm1[1],ymm2[2,3]
-; AVX1-ONLY-NEXT: vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 48(%rsi), %xmm1
-; AVX1-ONLY-NEXT: vmovaps 48(%rdi), %xmm2
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],mem[4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT: vbroadcastsd 48(%rcx), %ymm2
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm1 = mem[0,0]
-; AVX1-ONLY-NEXT: vmovaps 64(%rsi), %xmm2
-; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 64(%rdi), %xmm14
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm2 = xmm14[1],xmm2[1]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm2
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm9[0],ymm2[1,2,3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm2[0],ymm1[1],ymm2[2,3]
-; AVX1-ONLY-NEXT: vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 80(%rsi), %xmm1
-; AVX1-ONLY-NEXT: vmovaps 80(%rdi), %xmm2
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],mem[4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT: vbroadcastsd 80(%rcx), %ymm2
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 96(%rsi), %xmm12
-; AVX1-ONLY-NEXT: vmovaps 96(%rdi), %xmm11
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm11[1],xmm12[1]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
-; AVX1-ONLY-NEXT: vmovapd 96(%r8), %ymm7
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm7[0],ymm1[1,2,3]
+; AVX1-ONLY-NEXT: vmovaps 64(%rsi), %xmm8
+; AVX1-ONLY-NEXT: vmovaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps (%rdi), %xmm3
+; AVX1-ONLY-NEXT: vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm3 = xmm3[1],xmm5[1]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm3
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm3 = ymm1[0,1],ymm3[2,3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm3[0],ymm2[1],ymm3[2,3]
+; AVX1-ONLY-NEXT: vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm2 = mem[0,0]
+; AVX1-ONLY-NEXT: vmovaps 32(%rdi), %xmm3
+; AVX1-ONLY-NEXT: vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm3 = xmm3[1],xmm6[1]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm3
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm3 = ymm0[0,1],ymm3[2,3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm3[0],ymm2[1],ymm3[2,3]
+; AVX1-ONLY-NEXT: vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm2 = mem[0,0]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm1[0],ymm2[1],ymm1[2,3]
+; AVX1-ONLY-NEXT: vmovaps 64(%rdi), %xmm10
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm3 = xmm10[1],xmm8[1]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm3
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm3 = ymm4[0,1],ymm3[2,3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm3[0],ymm2[1],ymm3[2,3]
+; AVX1-ONLY-NEXT: vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 96(%rsi), %xmm11
+; AVX1-ONLY-NEXT: vmovaps 96(%rdi), %xmm13
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm2 = xmm13[1],xmm11[1]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm2
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm7[0,1],ymm2[2,3]
+; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm3 = mem[0,0]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm2[0],ymm3[1],ymm2[2,3]
+; AVX1-ONLY-NEXT: vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm2 = mem[2,3,2,3]
+; AVX1-ONLY-NEXT: vbroadcastsd 8(%r8), %ymm3
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps (%rcx), %xmm15
+; AVX1-ONLY-NEXT: vinsertf128 $1, (%r9), %ymm15, %ymm3
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm3[2,3],ymm2[4,5],ymm3[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovapd (%rdi), %ymm2
+; AVX1-ONLY-NEXT: vmovapd (%rsi), %ymm3
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm2 = ymm2[1],ymm3[1],ymm2[3],ymm3[3]
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm1[2,3],ymm2[2,3]
+; AVX1-ONLY-NEXT: vmovapd (%r9), %ymm12
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm2 = ymm12[2,3],ymm3[2,3]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm1 = ymm1[0],ymm2[0],ymm1[2],ymm2[3]
; AVX1-ONLY-NEXT: vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 112(%rsi), %xmm1
-; AVX1-ONLY-NEXT: vmovaps 112(%rdi), %xmm2
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],mem[4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT: vbroadcastsd 112(%rcx), %ymm2
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm1 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT: vbroadcastsd 8(%r8), %ymm2
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT: vmovaps (%rcx), %xmm10
-; AVX1-ONLY-NEXT: vinsertf128 $1, (%r9), %ymm10, %ymm2
+; AVX1-ONLY-NEXT: vbroadcastsd 40(%r8), %ymm2
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps 32(%rcx), %xmm5
+; AVX1-ONLY-NEXT: vinsertf128 $1, 32(%r9), %ymm5, %ymm2
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2,3],ymm1[4,5],ymm2[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovapd (%rdi), %ymm1
-; AVX1-ONLY-NEXT: vmovapd (%rsi), %ymm2
+; AVX1-ONLY-NEXT: vmovapd 32(%rdi), %ymm1
+; AVX1-ONLY-NEXT: vmovapd 32(%rsi), %ymm2
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm1 = ymm1[1],ymm2[1],ymm1[3],ymm2[3]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[2,3]
-; AVX1-ONLY-NEXT: vmovapd (%r9), %ymm4
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm4[2,3],ymm2[2,3]
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[3]
-; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm0 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT: vbroadcastsd 40(%r8), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vmovaps 32(%rcx), %xmm8
-; AVX1-ONLY-NEXT: vinsertf128 $1, 32(%r9), %ymm8, %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovapd 32(%rdi), %ymm0
-; AVX1-ONLY-NEXT: vmovapd 32(%rsi), %ymm1
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm15[2,3],ymm0[2,3]
-; AVX1-ONLY-NEXT: vmovapd 32(%r9), %ymm3
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm3[2,3],ymm1[2,3]
+; AVX1-ONLY-NEXT: vmovapd 32(%r9), %ymm6
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm6[2,3],ymm2[2,3]
; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[3]
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm0 = mem[2,3,2,3]
; AVX1-ONLY-NEXT: vbroadcastsd 72(%r8), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vmovaps 64(%rcx), %xmm5
-; AVX1-ONLY-NEXT: vinsertf128 $1, 64(%r9), %ymm5, %ymm6
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm6[2,3],ymm0[4,5],ymm6[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps 64(%rcx), %xmm3
+; AVX1-ONLY-NEXT: vinsertf128 $1, 64(%r9), %ymm3, %ymm1
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovapd 64(%rdi), %ymm0
-; AVX1-ONLY-NEXT: vmovapd 64(%rsi), %ymm6
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm6[1],ymm0[3],ymm6[3]
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm9[2,3],ymm0[2,3]
-; AVX1-ONLY-NEXT: vmovapd 64(%r9), %ymm1
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm6 = ymm1[2,3],ymm6[2,3]
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm13 = ymm0[0],ymm6[0],ymm0[2],ymm6[3]
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm0 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT: vbroadcastsd 104(%r8), %ymm6
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm0[0,1,2,3],ymm6[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT: vmovapd 64(%rsi), %ymm1
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm2 = ymm4[2,3],ymm0[2,3]
+; AVX1-ONLY-NEXT: vmovapd 64(%r9), %ymm0
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm0[2,3],ymm1[2,3]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[2],ymm1[3]
+; AVX1-ONLY-NEXT: vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm1 = mem[2,3,2,3]
+; AVX1-ONLY-NEXT: vbroadcastsd 104(%r8), %ymm2
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps 96(%rcx), %xmm2
-; AVX1-ONLY-NEXT: vinsertf128 $1, 96(%r9), %ymm2, %ymm9
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm9 = ymm6[0,1],ymm9[2,3],ymm6[4,5],ymm9[6,7]
-; AVX1-ONLY-NEXT: vmovapd 96(%rdi), %ymm6
-; AVX1-ONLY-NEXT: vmovapd 96(%rsi), %ymm0
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm6 = ymm6[1],ymm0[1],ymm6[3],ymm0[3]
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm6 = ymm7[2,3],ymm6[2,3]
-; AVX1-ONLY-NEXT: vmovapd 96(%r9), %ymm7
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm7[2,3],ymm0[2,3]
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm6 = ymm6[0],ymm0[0],ymm6[2],ymm0[3]
-; AVX1-ONLY-NEXT: vmovapd 16(%rdx), %xmm0
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],mem[1]
-; AVX1-ONLY-NEXT: vbroadcastsd 24(%r8), %ymm15
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm15[2],ymm0[3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1,2],ymm4[3]
+; AVX1-ONLY-NEXT: vinsertf128 $1, 96(%r9), %ymm2, %ymm4
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm4[2,3],ymm1[4,5],ymm4[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovapd 96(%rdi), %ymm1
+; AVX1-ONLY-NEXT: vmovapd 96(%rsi), %ymm4
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm1 = ymm1[1],ymm4[1],ymm1[3],ymm4[3]
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm7 = ymm7[2,3],ymm1[2,3]
+; AVX1-ONLY-NEXT: vmovapd 96(%r9), %ymm1
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm4 = ymm1[2,3],ymm4[2,3]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm9 = ymm7[0],ymm4[0],ymm7[2],ymm4[3]
+; AVX1-ONLY-NEXT: vmovaps 16(%rdi), %xmm4
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm4 = xmm4[0],mem[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vbroadcastsd 16(%rcx), %ymm7
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5],ymm7[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovapd 16(%rdx), %xmm4
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm4 = xmm4[1],mem[1]
+; AVX1-ONLY-NEXT: vbroadcastsd 24(%r8), %ymm7
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm4 = ymm4[0,1],ymm7[2,3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm4 = ymm4[0,1,2],ymm12[3]
+; AVX1-ONLY-NEXT: vmovupd %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 48(%rdi), %xmm4
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm4 = xmm4[0],mem[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vbroadcastsd 48(%rcx), %ymm7
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm4[0,1,2,3,4,5],ymm7[6,7]
; AVX1-ONLY-NEXT: vmovapd 48(%rdx), %xmm4
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm4 = xmm4[1],mem[1]
-; AVX1-ONLY-NEXT: vbroadcastsd 56(%r8), %ymm15
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm4 = ymm4[0,1],ymm15[2],ymm4[3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm4 = ymm4[0,1,2],ymm3[3]
-; AVX1-ONLY-NEXT: vmovapd 80(%rdx), %xmm3
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm3 = xmm3[1],mem[1]
-; AVX1-ONLY-NEXT: vbroadcastsd 88(%r8), %ymm15
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm3 = ymm3[0,1],ymm15[2],ymm3[3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm3 = ymm3[0,1,2],ymm1[3]
-; AVX1-ONLY-NEXT: vmovapd 112(%rdx), %xmm1
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm1[1],mem[1]
-; AVX1-ONLY-NEXT: vbroadcastsd 120(%r8), %ymm15
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm1[0,1],ymm15[2],ymm1[3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm1[0,1,2],ymm7[3]
-; AVX1-ONLY-NEXT: vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm14, %xmm7 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm7 = xmm14[0],mem[0]
-; AVX1-ONLY-NEXT: vmovaps 64(%rdx), %xmm14
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm5 = xmm14[0],xmm5[0]
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; AVX1-ONLY-NEXT: vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm14, %xmm14 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm14 = xmm14[0],mem[0]
-; AVX1-ONLY-NEXT: vmovaps 32(%rdx), %xmm15
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm8 = xmm15[0],xmm8[0]
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm11 = xmm11[0],xmm12[0]
-; AVX1-ONLY-NEXT: vmovaps 96(%rdx), %xmm12
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm12[0],xmm2[0]
+; AVX1-ONLY-NEXT: vbroadcastsd 56(%r8), %ymm7
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm4 = ymm4[0,1],ymm7[2,3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm7 = ymm4[0,1,2],ymm6[3]
+; AVX1-ONLY-NEXT: vmovaps 80(%rdi), %xmm4
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm4 = xmm4[0],mem[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vbroadcastsd 80(%rcx), %ymm6
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm4[0,1,2,3,4,5],ymm6[6,7]
+; AVX1-ONLY-NEXT: vmovapd 80(%rdx), %xmm4
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm4 = xmm4[1],mem[1]
+; AVX1-ONLY-NEXT: vbroadcastsd 88(%r8), %ymm12
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm4 = ymm4[0,1],ymm12[2,3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm4 = ymm4[0,1,2],ymm0[3]
+; AVX1-ONLY-NEXT: vmovaps 112(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm0 = xmm0[0],mem[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vbroadcastsd 112(%rcx), %ymm12
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm12[6,7]
+; AVX1-ONLY-NEXT: vmovapd 112(%rdx), %xmm12
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm12 = xmm12[1],mem[1]
+; AVX1-ONLY-NEXT: vbroadcastsd 120(%r8), %ymm14
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm12 = ymm12[0,1],ymm14[2,3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm12[0,1,2],ymm1[3]
+; AVX1-ONLY-NEXT: vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm10, %xmm10 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm10 = xmm10[0],mem[0]
+; AVX1-ONLY-NEXT: vmovaps 64(%rdx), %xmm12
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm3 = xmm12[0],xmm3[0]
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
; AVX1-ONLY-NEXT: vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm12, %xmm12 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm12 = xmm12[0],mem[0]
-; AVX1-ONLY-NEXT: vmovaps (%rdx), %xmm15
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm10 = xmm15[0],xmm10[0]
+; AVX1-ONLY-NEXT: vmovaps 32(%rdx), %xmm14
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm5 = xmm14[0],xmm5[0]
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm11 = xmm13[0],xmm11[0]
+; AVX1-ONLY-NEXT: vmovaps 96(%rdx), %xmm13
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm13[0],xmm2[0]
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; AVX1-ONLY-NEXT: vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm13, %xmm13 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm13 = xmm13[0],mem[0]
+; AVX1-ONLY-NEXT: vmovaps (%rdx), %xmm14
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm14 = xmm14[0],xmm15[0]
; AVX1-ONLY-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; AVX1-ONLY-NEXT: vmovaps %xmm10, 16(%rax)
-; AVX1-ONLY-NEXT: vmovaps %xmm12, (%rax)
+; AVX1-ONLY-NEXT: vmovaps %xmm14, 16(%rax)
+; AVX1-ONLY-NEXT: vmovaps %xmm13, (%rax)
; AVX1-ONLY-NEXT: vmovaps %xmm2, 592(%rax)
; AVX1-ONLY-NEXT: vmovaps %xmm11, 576(%rax)
-; AVX1-ONLY-NEXT: vmovaps %xmm8, 208(%rax)
-; AVX1-ONLY-NEXT: vmovaps %xmm14, 192(%rax)
-; AVX1-ONLY-NEXT: vmovaps %xmm5, 400(%rax)
-; AVX1-ONLY-NEXT: vmovaps %xmm7, 384(%rax)
-; AVX1-ONLY-NEXT: vmovapd %ymm6, 704(%rax)
-; AVX1-ONLY-NEXT: vmovaps %ymm9, 640(%rax)
-; AVX1-ONLY-NEXT: vmovapd %ymm13, 512(%rax)
+; AVX1-ONLY-NEXT: vmovaps %xmm5, 208(%rax)
+; AVX1-ONLY-NEXT: vmovaps %xmm12, 192(%rax)
+; AVX1-ONLY-NEXT: vmovaps %xmm3, 400(%rax)
+; AVX1-ONLY-NEXT: vmovaps %xmm10, 384(%rax)
+; AVX1-ONLY-NEXT: vmovapd %ymm9, 704(%rax)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm2, 640(%rax)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm2, 512(%rax)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm2, 448(%rax)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
@@ -1223,21 +1217,19 @@ define void @store_i64_stride6_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm2, 64(%rax)
; AVX1-ONLY-NEXT: vmovapd %ymm1, 736(%rax)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm1, 672(%rax)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm1, 608(%rax)
-; AVX1-ONLY-NEXT: vmovapd %ymm3, 544(%rax)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm1, 480(%rax)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm1, 416(%rax)
-; AVX1-ONLY-NEXT: vmovapd %ymm4, 352(%rax)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm1, 288(%rax)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm1, 224(%rax)
-; AVX1-ONLY-NEXT: vmovapd %ymm0, 160(%rax)
+; AVX1-ONLY-NEXT: vmovaps %ymm0, 672(%rax)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm0, 608(%rax)
+; AVX1-ONLY-NEXT: vmovapd %ymm4, 544(%rax)
+; AVX1-ONLY-NEXT: vmovaps %ymm6, 480(%rax)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm0, 416(%rax)
+; AVX1-ONLY-NEXT: vmovapd %ymm7, 352(%rax)
+; AVX1-ONLY-NEXT: vmovaps %ymm8, 288(%rax)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm0, 224(%rax)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm0, 160(%rax)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm0, 96(%rax)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
@@ -1278,7 +1270,7 @@ define void @store_i64_stride6_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm2 = xmm2[1],xmm3[1]
; AVX2-ONLY-NEXT: vbroadcastsd 8(%r8), %ymm3
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5],ymm2[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovddup {{.*#+}} xmm1 = xmm0[0,0]
@@ -1288,7 +1280,7 @@ define void @store_i64_stride6_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm4[1],xmm7[1]
; AVX2-ONLY-NEXT: vbroadcastsd 40(%r8), %ymm2
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX2-ONLY-NEXT: vmovaps 64(%r8), %ymm6
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3,4,5],ymm0[6,7]
@@ -1307,7 +1299,7 @@ define void @store_i64_stride6_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovaps 64(%rdx), %xmm10
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm10[1],xmm11[1]
; AVX2-ONLY-NEXT: vbroadcastsd 72(%r8), %ymm3
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5,6,7]
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm2
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm2[6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -1324,7 +1316,7 @@ define void @store_i64_stride6_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovaps 96(%rdx), %xmm2
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm12 = xmm2[1],xmm3[1]
; AVX2-ONLY-NEXT: vbroadcastsd 104(%r8), %ymm15
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3],ymm15[4,5],ymm12[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3],ymm15[4,5,6,7]
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm5, %ymm0, %ymm5
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm12[0,1,2,3,4,5],ymm5[6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -2258,133 +2250,84 @@ define void @store_i64_stride6_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
;
; AVX1-ONLY-LABEL: store_i64_stride6_vf32:
; AVX1-ONLY: # %bb.0:
-; AVX1-ONLY-NEXT: subq $1480, %rsp # imm = 0x5C8
-; AVX1-ONLY-NEXT: vmovapd (%r8), %ymm14
-; AVX1-ONLY-NEXT: vmovapd 32(%r8), %ymm5
-; AVX1-ONLY-NEXT: vmovapd 64(%r8), %ymm3
+; AVX1-ONLY-NEXT: subq $1576, %rsp # imm = 0x628
+; AVX1-ONLY-NEXT: vmovapd (%r8), %ymm4
+; AVX1-ONLY-NEXT: vmovapd 32(%r8), %ymm2
+; AVX1-ONLY-NEXT: vmovapd 64(%r8), %ymm1
+; AVX1-ONLY-NEXT: vmovapd 96(%r8), %ymm5
; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm0 = mem[0,0]
; AVX1-ONLY-NEXT: vmovaps (%rsi), %xmm6
; AVX1-ONLY-NEXT: vmovaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 16(%rsi), %xmm1
-; AVX1-ONLY-NEXT: vmovaps 32(%rsi), %xmm7
-; AVX1-ONLY-NEXT: vmovaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps (%rdi), %xmm4
-; AVX1-ONLY-NEXT: vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 16(%rdi), %xmm2
-; AVX1-ONLY-NEXT: vmovaps 32(%rdi), %xmm8
+; AVX1-ONLY-NEXT: vmovaps 32(%rsi), %xmm8
; AVX1-ONLY-NEXT: vmovaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm4 = xmm4[1],xmm6[1]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm4, %ymm0, %ymm4
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm4 = ymm14[0],ymm4[1,2,3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm4[0],ymm0[1],ymm4[2,3]
+; AVX1-ONLY-NEXT: vmovaps 64(%rsi), %xmm7
+; AVX1-ONLY-NEXT: vmovaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps (%rdi), %xmm3
+; AVX1-ONLY-NEXT: vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm3 = xmm3[1],xmm6[1]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm3
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm3 = ymm4[0,1],ymm3[2,3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm3[0],ymm0[1],ymm3[2,3]
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],mem[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vbroadcastsd 16(%rcx), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm0 = mem[0,0]
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm8[1],xmm7[1]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm5[0],ymm1[1,2,3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[2,3]
+; AVX1-ONLY-NEXT: vmovaps 32(%rdi), %xmm3
+; AVX1-ONLY-NEXT: vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm3 = xmm3[1],xmm8[1]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm3
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm3 = ymm2[0,1],ymm3[2,3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm3[0],ymm0[1],ymm3[2,3]
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 48(%rsi), %xmm0
-; AVX1-ONLY-NEXT: vmovaps 48(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],mem[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vbroadcastsd 48(%rcx), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm0 = mem[0,0]
-; AVX1-ONLY-NEXT: vmovaps 64(%rsi), %xmm2
-; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 64(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm2[1]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm3[0],ymm1[1,2,3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[2,3]
+; AVX1-ONLY-NEXT: vmovaps 64(%rdi), %xmm3
+; AVX1-ONLY-NEXT: vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm3 = xmm3[1],xmm7[1]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm3
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm3 = ymm1[0,1],ymm3[2,3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm3[0],ymm0[1],ymm3[2,3]
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 80(%rsi), %xmm0
-; AVX1-ONLY-NEXT: vmovaps 80(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],mem[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vbroadcastsd 80(%rcx), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 96(%rsi), %xmm1
-; AVX1-ONLY-NEXT: vmovaps %xmm1, (%rsp) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 96(%rsi), %xmm3
+; AVX1-ONLY-NEXT: vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vmovaps 96(%rdi), %xmm0
; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm3[1]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
-; AVX1-ONLY-NEXT: vmovapd 96(%r8), %ymm2
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm2[0],ymm0[1,2,3]
-; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm1 = mem[0,0]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm5[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm3 = mem[0,0]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0],ymm3[1],ymm0[2,3]
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 112(%rsi), %xmm0
-; AVX1-ONLY-NEXT: vmovaps 112(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],mem[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vbroadcastsd 112(%rcx), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 128(%rsi), %xmm1
-; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 128(%rsi), %xmm3
+; AVX1-ONLY-NEXT: vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vmovaps 128(%rdi), %xmm0
; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm3[1]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
-; AVX1-ONLY-NEXT: vmovapd 128(%r8), %ymm4
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm4[0],ymm0[1,2,3]
-; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm1 = mem[0,0]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3]
+; AVX1-ONLY-NEXT: vmovapd 128(%r8), %ymm8
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm8[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm3 = mem[0,0]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0],ymm3[1],ymm0[2,3]
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 144(%rsi), %xmm0
-; AVX1-ONLY-NEXT: vmovaps 144(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],mem[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vbroadcastsd 144(%rcx), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 160(%rsi), %xmm1
-; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 160(%rsi), %xmm3
+; AVX1-ONLY-NEXT: vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vmovaps 160(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm1
+; AVX1-ONLY-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm3[1]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm3
; AVX1-ONLY-NEXT: vmovapd 160(%r8), %ymm0
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm0[0],ymm1[1,2,3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm3 = ymm0[0,1],ymm3[2,3]
; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm6 = mem[0,0]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm1[0],ymm6[1],ymm1[2,3]
-; AVX1-ONLY-NEXT: vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 176(%rsi), %xmm1
-; AVX1-ONLY-NEXT: vmovaps 176(%rdi), %xmm6
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm6[0],ymm1[0],ymm6[2],ymm1[2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],mem[4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT: vbroadcastsd 176(%rcx), %ymm6
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm6[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm3 = ymm3[0],ymm6[1],ymm3[2,3]
+; AVX1-ONLY-NEXT: vmovupd %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 192(%rsi), %xmm6
; AVX1-ONLY-NEXT: vmovaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 192(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm6[1]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm6
-; AVX1-ONLY-NEXT: vmovapd 192(%r8), %ymm1
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm6 = ymm1[0],ymm6[1,2,3]
+; AVX1-ONLY-NEXT: vmovaps 192(%rdi), %xmm3
+; AVX1-ONLY-NEXT: vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm3 = xmm3[1],xmm6[1]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm6
+; AVX1-ONLY-NEXT: vmovapd 192(%r8), %ymm3
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm6 = ymm3[0,1],ymm6[2,3]
; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm7 = mem[0,0]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm6 = ymm6[0],ymm7[1],ymm6[2,3]
; AVX1-ONLY-NEXT: vmovupd %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 208(%rsi), %xmm6
-; AVX1-ONLY-NEXT: vmovaps 208(%rdi), %xmm7
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm6 = ymm7[0],ymm6[0],ymm7[2],ymm6[2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],mem[4,5],ymm6[6,7]
-; AVX1-ONLY-NEXT: vbroadcastsd 208(%rcx), %ymm7
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5],ymm7[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 224(%rsi), %xmm7
; AVX1-ONLY-NEXT: vmovaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vmovaps 224(%rdi), %xmm6
@@ -2392,246 +2335,293 @@ define void @store_i64_stride6_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm6 = xmm6[1],xmm7[1]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm6, %ymm0, %ymm6
; AVX1-ONLY-NEXT: vmovapd 224(%r8), %ymm12
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm6 = ymm12[0],ymm6[1,2,3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm6 = ymm12[0,1],ymm6[2,3]
; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm7 = mem[0,0]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm6 = ymm6[0],ymm7[1],ymm6[2,3]
; AVX1-ONLY-NEXT: vmovupd %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 240(%rsi), %xmm6
-; AVX1-ONLY-NEXT: vmovaps 240(%rdi), %xmm7
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm6 = ymm7[0],ymm6[0],ymm7[2],ymm6[2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],mem[4,5],ymm6[6,7]
-; AVX1-ONLY-NEXT: vbroadcastsd 240(%rcx), %ymm7
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5],ymm7[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm6 = mem[2,3,2,3]
; AVX1-ONLY-NEXT: vbroadcastsd 8(%r8), %ymm7
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm7[4,5],ymm6[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm7[4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps (%rcx), %xmm7
; AVX1-ONLY-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vinsertf128 $1, (%r9), %ymm7, %ymm7
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1],ymm7[2,3],ymm6[4,5],ymm7[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovapd (%rdi), %ymm6
-; AVX1-ONLY-NEXT: vmovapd (%rsi), %ymm8
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm6 = ymm6[1],ymm8[1],ymm6[3],ymm8[3]
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm6 = ymm14[2,3],ymm6[2,3]
+; AVX1-ONLY-NEXT: vmovapd (%rsi), %ymm9
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm6 = ymm6[1],ymm9[1],ymm6[3],ymm9[3]
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm4 = ymm4[2,3],ymm6[2,3]
; AVX1-ONLY-NEXT: vmovapd (%r9), %ymm7
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm8 = ymm7[2,3],ymm8[2,3]
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm6 = ymm6[0],ymm8[0],ymm6[2],ymm8[3]
-; AVX1-ONLY-NEXT: vmovupd %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm6 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT: vbroadcastsd 40(%r8), %ymm8
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm8[4,5],ymm6[6,7]
-; AVX1-ONLY-NEXT: vmovaps 32(%rcx), %xmm8
-; AVX1-ONLY-NEXT: vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vinsertf128 $1, 32(%r9), %ymm8, %ymm8
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1],ymm8[2,3],ymm6[4,5],ymm8[6,7]
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm6 = ymm7[2,3],ymm9[2,3]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm4 = ymm4[0],ymm6[0],ymm4[2],ymm6[3]
+; AVX1-ONLY-NEXT: vmovupd %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm4 = mem[2,3,2,3]
+; AVX1-ONLY-NEXT: vbroadcastsd 40(%r8), %ymm6
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm6[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps 32(%rcx), %xmm6
; AVX1-ONLY-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovapd 32(%rdi), %ymm6
-; AVX1-ONLY-NEXT: vmovapd 32(%rsi), %ymm8
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm6 = ymm6[1],ymm8[1],ymm6[3],ymm8[3]
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm5 = ymm5[2,3],ymm6[2,3]
-; AVX1-ONLY-NEXT: vmovapd 32(%r9), %ymm10
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm6 = ymm10[2,3],ymm8[2,3]
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm5 = ymm5[0],ymm6[0],ymm5[2],ymm6[3]
-; AVX1-ONLY-NEXT: vmovupd %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm5 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT: vbroadcastsd 72(%r8), %ymm6
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm6[4,5],ymm5[6,7]
-; AVX1-ONLY-NEXT: vmovaps 64(%rcx), %xmm14
-; AVX1-ONLY-NEXT: vinsertf128 $1, 64(%r9), %ymm14, %ymm6
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1],ymm6[2,3],ymm5[4,5],ymm6[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovapd 64(%rdi), %ymm5
-; AVX1-ONLY-NEXT: vmovapd 64(%rsi), %ymm6
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm5 = ymm5[1],ymm6[1],ymm5[3],ymm6[3]
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm3 = ymm3[2,3],ymm5[2,3]
-; AVX1-ONLY-NEXT: vmovapd 64(%r9), %ymm5
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm6 = ymm5[2,3],ymm6[2,3]
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm3 = ymm3[0],ymm6[0],ymm3[2],ymm6[3]
-; AVX1-ONLY-NEXT: vmovupd %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm3 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT: vbroadcastsd 104(%r8), %ymm6
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm6[4,5],ymm3[6,7]
-; AVX1-ONLY-NEXT: vmovaps 96(%rcx), %xmm13
-; AVX1-ONLY-NEXT: vinsertf128 $1, 96(%r9), %ymm13, %ymm6
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1],ymm6[2,3],ymm3[4,5],ymm6[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovapd 96(%rdi), %ymm3
-; AVX1-ONLY-NEXT: vmovapd 96(%rsi), %ymm6
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm3 = ymm3[1],ymm6[1],ymm3[3],ymm6[3]
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm2 = ymm2[2,3],ymm3[2,3]
-; AVX1-ONLY-NEXT: vmovapd 96(%r9), %ymm8
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm3 = ymm8[2,3],ymm6[2,3]
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm2 = ymm2[0],ymm3[0],ymm2[2],ymm3[3]
-; AVX1-ONLY-NEXT: vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm2 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT: vbroadcastsd 136(%r8), %ymm3
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5],ymm2[6,7]
-; AVX1-ONLY-NEXT: vmovaps 128(%rcx), %xmm9
-; AVX1-ONLY-NEXT: vinsertf128 $1, 128(%r9), %ymm9, %ymm3
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm3[2,3],ymm2[4,5],ymm3[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovapd 128(%rdi), %ymm2
-; AVX1-ONLY-NEXT: vmovapd 128(%rsi), %ymm6
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm2 = ymm2[1],ymm6[1],ymm2[3],ymm6[3]
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm2 = ymm4[2,3],ymm2[2,3]
-; AVX1-ONLY-NEXT: vmovapd 128(%r9), %ymm3
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm4 = ymm3[2,3],ymm6[2,3]
+; AVX1-ONLY-NEXT: vinsertf128 $1, 32(%r9), %ymm6, %ymm6
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm6[2,3],ymm4[4,5],ymm6[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovapd 32(%rdi), %ymm4
+; AVX1-ONLY-NEXT: vmovapd 32(%rsi), %ymm6
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm4 = ymm4[1],ymm6[1],ymm4[3],ymm6[3]
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm2 = ymm2[2,3],ymm4[2,3]
+; AVX1-ONLY-NEXT: vmovapd 32(%r9), %ymm15
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm4 = ymm15[2,3],ymm6[2,3]
; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm2 = ymm2[0],ymm4[0],ymm2[2],ymm4[3]
; AVX1-ONLY-NEXT: vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm2 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT: vbroadcastsd 168(%r8), %ymm4
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm4[4,5],ymm2[6,7]
-; AVX1-ONLY-NEXT: vmovaps 160(%rcx), %xmm11
-; AVX1-ONLY-NEXT: vinsertf128 $1, 160(%r9), %ymm11, %ymm4
+; AVX1-ONLY-NEXT: vbroadcastsd 72(%r8), %ymm4
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm4[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps 64(%rcx), %xmm4
+; AVX1-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vinsertf128 $1, 64(%r9), %ymm4, %ymm4
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm4[2,3],ymm2[4,5],ymm4[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovapd 160(%rdi), %ymm2
-; AVX1-ONLY-NEXT: vmovapd 160(%rsi), %ymm4
+; AVX1-ONLY-NEXT: vmovapd 64(%rdi), %ymm2
+; AVX1-ONLY-NEXT: vmovapd 64(%rsi), %ymm4
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm2 = ymm2[1],ymm4[1],ymm2[3],ymm4[3]
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm2[2,3]
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm1[2,3],ymm2[2,3]
+; AVX1-ONLY-NEXT: vmovapd 64(%r9), %ymm13
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm2 = ymm13[2,3],ymm4[2,3]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm1 = ymm1[0],ymm2[0],ymm1[2],ymm2[3]
+; AVX1-ONLY-NEXT: vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm1 = mem[2,3,2,3]
+; AVX1-ONLY-NEXT: vbroadcastsd 104(%r8), %ymm2
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps 96(%rcx), %xmm2
+; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vinsertf128 $1, 96(%r9), %ymm2, %ymm2
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2,3],ymm1[4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovapd 96(%rdi), %ymm1
+; AVX1-ONLY-NEXT: vmovapd 96(%rsi), %ymm2
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm1 = ymm1[1],ymm2[1],ymm1[3],ymm2[3]
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm5[2,3],ymm1[2,3]
+; AVX1-ONLY-NEXT: vmovapd 96(%r9), %ymm10
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm2 = ymm10[2,3],ymm2[2,3]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm1 = ymm1[0],ymm2[0],ymm1[2],ymm2[3]
+; AVX1-ONLY-NEXT: vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm1 = mem[2,3,2,3]
+; AVX1-ONLY-NEXT: vbroadcastsd 136(%r8), %ymm2
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps 128(%rcx), %xmm11
+; AVX1-ONLY-NEXT: vinsertf128 $1, 128(%r9), %ymm11, %ymm2
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2,3],ymm1[4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovapd 128(%rdi), %ymm1
+; AVX1-ONLY-NEXT: vmovapd 128(%rsi), %ymm2
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm1 = ymm1[1],ymm2[1],ymm1[3],ymm2[3]
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm8[2,3],ymm1[2,3]
+; AVX1-ONLY-NEXT: vmovapd 128(%r9), %ymm8
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm2 = ymm8[2,3],ymm2[2,3]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm1 = ymm1[0],ymm2[0],ymm1[2],ymm2[3]
+; AVX1-ONLY-NEXT: vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm1 = mem[2,3,2,3]
+; AVX1-ONLY-NEXT: vbroadcastsd 168(%r8), %ymm2
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps 160(%rcx), %xmm14
+; AVX1-ONLY-NEXT: vinsertf128 $1, 160(%r9), %ymm14, %ymm2
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2,3],ymm1[4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovapd 160(%rdi), %ymm1
+; AVX1-ONLY-NEXT: vmovapd 160(%rsi), %ymm4
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm1 = ymm1[1],ymm4[1],ymm1[3],ymm4[3]
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[2,3]
; AVX1-ONLY-NEXT: vmovapd 160(%r9), %ymm2
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm4 = ymm2[2,3],ymm4[2,3]
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[0],ymm4[0],ymm0[2],ymm4[3]
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm2[2,3],ymm4[2,3]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[3]
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm0 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT: vbroadcastsd 200(%r8), %ymm4
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm4[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vmovaps 192(%rcx), %xmm6
-; AVX1-ONLY-NEXT: vinsertf128 $1, 192(%r9), %ymm6, %ymm4
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm4[2,3],ymm0[4,5],ymm4[6,7]
+; AVX1-ONLY-NEXT: vbroadcastsd 200(%r8), %ymm1
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps 192(%rcx), %xmm1
+; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vinsertf128 $1, 192(%r9), %ymm1, %ymm1
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovapd 192(%rdi), %ymm0
; AVX1-ONLY-NEXT: vmovapd 192(%rsi), %ymm4
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm4[1],ymm0[3],ymm4[3]
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm3[2,3],ymm0[2,3]
; AVX1-ONLY-NEXT: vmovapd 192(%r9), %ymm1
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm4 = ymm1[2,3],ymm4[2,3]
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[0],ymm4[0],ymm0[2],ymm4[3]
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm3 = ymm1[2,3],ymm4[2,3]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[0],ymm3[0],ymm0[2],ymm3[3]
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm0 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT: vbroadcastsd 232(%r8), %ymm4
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm4[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vmovaps 224(%rcx), %xmm4
-; AVX1-ONLY-NEXT: vinsertf128 $1, 224(%r9), %ymm4, %ymm15
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm15[2,3],ymm0[4,5],ymm15[6,7]
+; AVX1-ONLY-NEXT: vbroadcastsd 232(%r8), %ymm3
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps 224(%rcx), %xmm9
+; AVX1-ONLY-NEXT: vinsertf128 $1, 224(%r9), %ymm9, %ymm3
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm3[2,3],ymm0[4,5],ymm3[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovapd 224(%rdi), %ymm0
-; AVX1-ONLY-NEXT: vmovapd 224(%rsi), %ymm15
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm15[1],ymm0[3],ymm15[3]
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm12 = ymm12[2,3],ymm0[2,3]
+; AVX1-ONLY-NEXT: vmovapd 224(%rsi), %ymm3
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm3[1],ymm0[3],ymm3[3]
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm5 = ymm12[2,3],ymm0[2,3]
; AVX1-ONLY-NEXT: vmovapd 224(%r9), %ymm0
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm15 = ymm0[2,3],ymm15[2,3]
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm12 = ymm12[0],ymm15[0],ymm12[2],ymm15[3]
-; AVX1-ONLY-NEXT: vmovupd %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovapd 16(%rdx), %xmm15
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm15 = xmm15[1],mem[1]
-; AVX1-ONLY-NEXT: vbroadcastsd 24(%r8), %ymm12
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm12 = ymm15[0,1],ymm12[2],ymm15[3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm7 = ymm12[0,1,2],ymm7[3]
-; AVX1-ONLY-NEXT: vmovupd %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovapd 48(%rdx), %xmm7
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm7 = xmm7[1],mem[1]
-; AVX1-ONLY-NEXT: vbroadcastsd 56(%r8), %ymm12
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm7 = ymm7[0,1],ymm12[2],ymm7[3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm7 = ymm7[0,1,2],ymm10[3]
-; AVX1-ONLY-NEXT: vmovupd %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovapd 80(%rdx), %xmm7
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm7 = xmm7[1],mem[1]
-; AVX1-ONLY-NEXT: vbroadcastsd 88(%r8), %ymm10
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm7 = ymm7[0,1],ymm10[2],ymm7[3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm5 = ymm7[0,1,2],ymm5[3]
-; AVX1-ONLY-NEXT: vmovupd %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovapd 112(%rdx), %xmm5
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm5 = xmm5[1],mem[1]
-; AVX1-ONLY-NEXT: vbroadcastsd 120(%r8), %ymm7
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm5 = ymm5[0,1],ymm7[2],ymm5[3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm5 = ymm5[0,1,2],ymm8[3]
-; AVX1-ONLY-NEXT: vmovupd %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovapd 144(%rdx), %xmm5
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm5 = xmm5[1],mem[1]
-; AVX1-ONLY-NEXT: vbroadcastsd 152(%r8), %ymm7
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm5 = ymm5[0,1],ymm7[2],ymm5[3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm3 = ymm5[0,1,2],ymm3[3]
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm3 = ymm0[2,3],ymm3[2,3]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm3 = ymm5[0],ymm3[0],ymm5[2],ymm3[3]
+; AVX1-ONLY-NEXT: vmovupd %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 16(%rdi), %xmm3
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm3 = xmm3[0],mem[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vbroadcastsd 16(%rcx), %ymm5
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm5[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovapd 16(%rdx), %xmm3
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm3 = xmm3[1],mem[1]
+; AVX1-ONLY-NEXT: vbroadcastsd 24(%r8), %ymm5
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm3 = ymm3[0,1],ymm5[2,3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm3 = ymm3[0,1,2],ymm7[3]
+; AVX1-ONLY-NEXT: vmovupd %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 48(%rdi), %xmm3
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm3 = xmm3[0],mem[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vbroadcastsd 48(%rcx), %ymm5
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm5[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovapd 48(%rdx), %xmm3
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm3 = xmm3[1],mem[1]
+; AVX1-ONLY-NEXT: vbroadcastsd 56(%r8), %ymm5
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm3 = ymm3[0,1],ymm5[2,3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm3 = ymm3[0,1,2],ymm15[3]
+; AVX1-ONLY-NEXT: vmovupd %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 80(%rdi), %xmm3
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm3 = xmm3[0],mem[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vbroadcastsd 80(%rcx), %ymm5
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm5[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovapd 80(%rdx), %xmm3
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm3 = xmm3[1],mem[1]
+; AVX1-ONLY-NEXT: vbroadcastsd 88(%r8), %ymm5
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm3 = ymm3[0,1],ymm5[2,3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm3 = ymm3[0,1,2],ymm13[3]
+; AVX1-ONLY-NEXT: vmovupd %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 112(%rdi), %xmm3
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm3 = xmm3[0],mem[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vbroadcastsd 112(%rcx), %ymm5
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm5[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovapd 112(%rdx), %xmm3
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm3 = xmm3[1],mem[1]
+; AVX1-ONLY-NEXT: vbroadcastsd 120(%r8), %ymm5
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm3 = ymm3[0,1],ymm5[2,3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm3 = ymm3[0,1,2],ymm10[3]
+; AVX1-ONLY-NEXT: vmovupd %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 144(%rdi), %xmm3
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm3 = xmm3[0],mem[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vbroadcastsd 144(%rcx), %ymm5
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm5[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovapd 144(%rdx), %xmm3
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm3 = xmm3[1],mem[1]
+; AVX1-ONLY-NEXT: vbroadcastsd 152(%r8), %ymm5
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm3 = ymm3[0,1],ymm5[2,3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm3 = ymm3[0,1,2],ymm8[3]
; AVX1-ONLY-NEXT: vmovupd %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovapd 176(%rdx), %xmm5
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm5 = xmm5[1],mem[1]
-; AVX1-ONLY-NEXT: vbroadcastsd 184(%r8), %ymm7
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm5 = ymm5[0,1],ymm7[2],ymm5[3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm5[0,1,2],ymm2[3]
+; AVX1-ONLY-NEXT: vmovaps 176(%rdi), %xmm3
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm3 = xmm3[0],mem[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vbroadcastsd 176(%rcx), %ymm5
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm5[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovapd 176(%rdx), %xmm3
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm3 = xmm3[1],mem[1]
+; AVX1-ONLY-NEXT: vbroadcastsd 184(%r8), %ymm5
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm3 = ymm3[0,1],ymm5[2,3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3]
; AVX1-ONLY-NEXT: vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 208(%rdi), %xmm2
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm2 = xmm2[0],mem[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vbroadcastsd 208(%rcx), %ymm3
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm3[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovapd 208(%rdx), %xmm2
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm2 = xmm2[1],mem[1]
-; AVX1-ONLY-NEXT: vbroadcastsd 216(%r8), %ymm7
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm2[0,1],ymm7[2],ymm2[3]
+; AVX1-ONLY-NEXT: vbroadcastsd 216(%r8), %ymm3
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm2[0,1],ymm3[2,3]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3]
; AVX1-ONLY-NEXT: vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 240(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm1 = xmm1[0],mem[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vbroadcastsd 240(%rcx), %ymm2
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovapd 240(%rdx), %xmm1
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm1[1],mem[1]
-; AVX1-ONLY-NEXT: vbroadcastsd 248(%r8), %ymm7
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm1[0,1],ymm7[2],ymm1[3]
+; AVX1-ONLY-NEXT: vbroadcastsd 248(%r8), %ymm2
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm1[0,1],ymm2[2,3]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3]
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX1-ONLY-NEXT: vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm1 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm1 = xmm0[0],mem[0]
-; AVX1-ONLY-NEXT: vmovaps 128(%rdx), %xmm7
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm7 = xmm7[0],xmm9[0]
+; AVX1-ONLY-NEXT: vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm6 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm6 = xmm0[0],mem[0]
+; AVX1-ONLY-NEXT: vmovaps 128(%rdx), %xmm0
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm11 = xmm0[0],xmm11[0]
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX1-ONLY-NEXT: vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm9 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm9 = xmm0[0],mem[0]
-; AVX1-ONLY-NEXT: vmovaps 64(%rdx), %xmm12
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm12 = xmm12[0],xmm14[0]
+; AVX1-ONLY-NEXT: vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm5 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm5 = xmm0[0],mem[0]
+; AVX1-ONLY-NEXT: vmovaps 64(%rdx), %xmm15
+; AVX1-ONLY-NEXT: vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm15 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm15 = xmm15[0],mem[0]
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT: vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm4 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm4 = xmm0[0],mem[0]
+; AVX1-ONLY-NEXT: vmovaps 32(%rdx), %xmm13
+; AVX1-ONLY-NEXT: vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm13, %xmm13 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm13 = xmm13[0],mem[0]
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT: vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm3 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm3 = xmm0[0],mem[0]
+; AVX1-ONLY-NEXT: vmovaps 96(%rdx), %xmm12
+; AVX1-ONLY-NEXT: vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm12, %xmm12 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm12 = xmm12[0],mem[0]
+; AVX1-ONLY-NEXT: vmovaps (%rsp), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT: vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm2 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm2 = xmm0[0],mem[0]
+; AVX1-ONLY-NEXT: vmovaps 160(%rdx), %xmm10
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm10 = xmm10[0],xmm14[0]
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT: vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm1 = xmm0[0],mem[0]
+; AVX1-ONLY-NEXT: vmovaps 224(%rdx), %xmm14
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm9 = xmm14[0],xmm9[0]
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX1-ONLY-NEXT: vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm14 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm14 = xmm0[0],mem[0]
-; AVX1-ONLY-NEXT: vmovaps 32(%rdx), %xmm15
-; AVX1-ONLY-NEXT: vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm15 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm15 = xmm15[0],mem[0]
+; AVX1-ONLY-NEXT: vmovaps 192(%rdx), %xmm8
+; AVX1-ONLY-NEXT: vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm8, %xmm8 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm8 = xmm8[0],mem[0]
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX1-ONLY-NEXT: vunpcklpd (%rsp), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm0 = xmm0[0],mem[0]
-; AVX1-ONLY-NEXT: vmovaps 96(%rdx), %xmm10
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm10 = xmm10[0],xmm13[0]
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX1-ONLY-NEXT: vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm13 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm13 = xmm2[0],mem[0]
-; AVX1-ONLY-NEXT: vmovaps 160(%rdx), %xmm8
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm8 = xmm8[0],xmm11[0]
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX1-ONLY-NEXT: vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm11 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm11 = xmm2[0],mem[0]
-; AVX1-ONLY-NEXT: vmovaps 224(%rdx), %xmm5
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm4 = xmm5[0],xmm4[0]
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX1-ONLY-NEXT: vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm5 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm5 = xmm2[0],mem[0]
-; AVX1-ONLY-NEXT: vmovaps 192(%rdx), %xmm3
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm3 = xmm3[0],xmm6[0]
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX1-ONLY-NEXT: vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm6 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm6 = xmm2[0],mem[0]
-; AVX1-ONLY-NEXT: vmovaps (%rdx), %xmm2
-; AVX1-ONLY-NEXT: vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm2 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm2 = xmm2[0],mem[0]
+; AVX1-ONLY-NEXT: vmovaps (%rdx), %xmm7
+; AVX1-ONLY-NEXT: vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm7, %xmm7 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm7 = xmm7[0],mem[0]
; AVX1-ONLY-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; AVX1-ONLY-NEXT: vmovaps %xmm2, 16(%rax)
-; AVX1-ONLY-NEXT: vmovaps %xmm6, (%rax)
-; AVX1-ONLY-NEXT: vmovaps %xmm3, 1168(%rax)
-; AVX1-ONLY-NEXT: vmovaps %xmm5, 1152(%rax)
-; AVX1-ONLY-NEXT: vmovaps %xmm4, 1360(%rax)
-; AVX1-ONLY-NEXT: vmovaps %xmm11, 1344(%rax)
-; AVX1-ONLY-NEXT: vmovaps %xmm8, 976(%rax)
-; AVX1-ONLY-NEXT: vmovaps %xmm13, 960(%rax)
-; AVX1-ONLY-NEXT: vmovaps %xmm10, 592(%rax)
-; AVX1-ONLY-NEXT: vmovaps %xmm0, 576(%rax)
-; AVX1-ONLY-NEXT: vmovaps %xmm15, 208(%rax)
-; AVX1-ONLY-NEXT: vmovaps %xmm14, 192(%rax)
-; AVX1-ONLY-NEXT: vmovaps %xmm12, 400(%rax)
-; AVX1-ONLY-NEXT: vmovaps %xmm9, 384(%rax)
-; AVX1-ONLY-NEXT: vmovaps %xmm7, 784(%rax)
-; AVX1-ONLY-NEXT: vmovaps %xmm1, 768(%rax)
+; AVX1-ONLY-NEXT: vmovaps %xmm7, 16(%rax)
+; AVX1-ONLY-NEXT: vmovaps %xmm0, (%rax)
+; AVX1-ONLY-NEXT: vmovaps %xmm8, 1168(%rax)
+; AVX1-ONLY-NEXT: vmovaps %xmm14, 1152(%rax)
+; AVX1-ONLY-NEXT: vmovaps %xmm9, 1360(%rax)
+; AVX1-ONLY-NEXT: vmovaps %xmm1, 1344(%rax)
+; AVX1-ONLY-NEXT: vmovaps %xmm10, 976(%rax)
+; AVX1-ONLY-NEXT: vmovaps %xmm2, 960(%rax)
+; AVX1-ONLY-NEXT: vmovaps %xmm12, 592(%rax)
+; AVX1-ONLY-NEXT: vmovaps %xmm3, 576(%rax)
+; AVX1-ONLY-NEXT: vmovaps %xmm13, 208(%rax)
+; AVX1-ONLY-NEXT: vmovaps %xmm4, 192(%rax)
+; AVX1-ONLY-NEXT: vmovaps %xmm15, 400(%rax)
+; AVX1-ONLY-NEXT: vmovaps %xmm5, 384(%rax)
+; AVX1-ONLY-NEXT: vmovaps %xmm11, 784(%rax)
+; AVX1-ONLY-NEXT: vmovaps %xmm6, 768(%rax)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm0, 1472(%rax)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
@@ -2712,7 +2702,7 @@ define void @store_i64_stride6_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vmovaps %ymm0, 96(%rax)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm0, 32(%rax)
-; AVX1-ONLY-NEXT: addq $1480, %rsp # imm = 0x5C8
+; AVX1-ONLY-NEXT: addq $1576, %rsp # imm = 0x628
; AVX1-ONLY-NEXT: vzeroupper
; AVX1-ONLY-NEXT: retq
;
@@ -2748,7 +2738,7 @@ define void @store_i64_stride6_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm3 = xmm3[1],xmm1[1]
; AVX2-ONLY-NEXT: vbroadcastsd 8(%r8), %ymm4
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm4[4,5],ymm3[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm4[4,5,6,7]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm3[0,1,2,3,4,5],ymm2[6,7]
; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovddup {{.*#+}} xmm2 = xmm0[0,0]
@@ -2758,7 +2748,7 @@ define void @store_i64_stride6_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm2 = xmm7[1],xmm6[1]
; AVX2-ONLY-NEXT: vbroadcastsd 40(%r8), %ymm3
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm2[0,1,2,3],ymm3[4,5],ymm2[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm2[0,1,2,3],ymm3[4,5,6,7]
; AVX2-ONLY-NEXT: vmovaps 64(%r8), %ymm2
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3,4,5],ymm0[6,7]
@@ -2779,7 +2769,7 @@ define void @store_i64_stride6_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX2-ONLY-NEXT: vbroadcastsd 72(%r8), %ymm4
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm4[4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm4[4,5,6,7]
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm3
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm3[6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -2800,7 +2790,7 @@ define void @store_i64_stride6_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
; AVX2-ONLY-NEXT: vbroadcastsd 104(%r8), %ymm5
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm5[4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm5[4,5,6,7]
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm3
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm3[6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -2821,7 +2811,7 @@ define void @store_i64_stride6_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
; AVX2-ONLY-NEXT: vbroadcastsd 136(%r8), %ymm5
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm5[4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm5[4,5,6,7]
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm3
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm3[6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -2842,7 +2832,7 @@ define void @store_i64_stride6_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm3 = xmm1[1],xmm3[1]
; AVX2-ONLY-NEXT: vbroadcastsd 168(%r8), %ymm6
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm6[4,5],ymm3[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm6[4,5,6,7]
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm5, %ymm0, %ymm5
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm3[0,1,2,3,4,5],ymm5[6,7]
; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -2863,7 +2853,7 @@ define void @store_i64_stride6_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm5 = xmm1[1],xmm5[1]
; AVX2-ONLY-NEXT: vbroadcastsd 200(%r8), %ymm8
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm8[4,5],ymm5[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm8[4,5,6,7]
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm6, %ymm0, %ymm6
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm5[0,1,2,3,4,5],ymm6[6,7]
; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -2881,7 +2871,7 @@ define void @store_i64_stride6_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovaps 224(%rdx), %xmm11
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm10 = xmm11[1],xmm8[1]
; AVX2-ONLY-NEXT: vbroadcastsd 232(%r8), %ymm13
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm10 = ymm10[0,1,2,3],ymm13[4,5],ymm10[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm10 = ymm10[0,1,2,3],ymm13[4,5,6,7]
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm9, %ymm0, %ymm9
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm10[0,1,2,3,4,5],ymm9[6,7]
; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -4802,662 +4792,643 @@ define void @store_i64_stride6_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
;
; AVX1-ONLY-LABEL: store_i64_stride6_vf64:
; AVX1-ONLY: # %bb.0:
-; AVX1-ONLY-NEXT: subq $3432, %rsp # imm = 0xD68
-; AVX1-ONLY-NEXT: vmovapd 32(%r8), %ymm3
-; AVX1-ONLY-NEXT: vmovapd 64(%r8), %ymm10
-; AVX1-ONLY-NEXT: vmovapd (%r8), %ymm4
+; AVX1-ONLY-NEXT: subq $3464, %rsp # imm = 0xD88
+; AVX1-ONLY-NEXT: vmovaps 96(%r8), %ymm7
+; AVX1-ONLY-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovapd 32(%r8), %ymm4
+; AVX1-ONLY-NEXT: vmovapd 64(%r8), %ymm14
+; AVX1-ONLY-NEXT: vmovapd (%r8), %ymm5
; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm0 = mem[0,0]
-; AVX1-ONLY-NEXT: vmovaps (%rsi), %xmm6
+; AVX1-ONLY-NEXT: vmovaps (%rsi), %xmm2
+; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 32(%rsi), %xmm3
+; AVX1-ONLY-NEXT: vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 64(%rsi), %xmm6
; AVX1-ONLY-NEXT: vmovaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 16(%rsi), %xmm1
-; AVX1-ONLY-NEXT: vmovaps 32(%rsi), %xmm8
-; AVX1-ONLY-NEXT: vmovaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps (%rdi), %xmm5
-; AVX1-ONLY-NEXT: vmovaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 16(%rdi), %xmm2
-; AVX1-ONLY-NEXT: vmovaps 32(%rdi), %xmm7
-; AVX1-ONLY-NEXT: vmovaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm5 = xmm5[1],xmm6[1]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm5, %ymm0, %ymm5
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm5 = ymm4[0],ymm5[1,2,3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm5[0],ymm0[1],ymm5[2,3]
+; AVX1-ONLY-NEXT: vmovaps (%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm2[1]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm5[0,1],ymm1[2,3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[2,3]
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],mem[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vbroadcastsd 16(%rcx), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm0 = mem[0,0]
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm7[1],xmm8[1]
+; AVX1-ONLY-NEXT: vmovaps 32(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm3[1]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm3[0],ymm1[1,2,3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm4[0,1],ymm1[2,3]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[2,3]
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 48(%rsi), %xmm0
-; AVX1-ONLY-NEXT: vmovaps 48(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],mem[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vbroadcastsd 48(%rcx), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm0 = mem[0,0]
-; AVX1-ONLY-NEXT: vmovaps 64(%rsi), %xmm2
-; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vmovaps 64(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm2[1]
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm6[1]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm10[0],ymm1[1,2,3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm14[0,1],ymm1[2,3]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[2,3]
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 80(%rsi), %xmm0
-; AVX1-ONLY-NEXT: vmovaps 80(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],mem[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vbroadcastsd 80(%rcx), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 96(%rsi), %xmm1
; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vmovaps 96(%rdi), %xmm0
; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
-; AVX1-ONLY-NEXT: vmovapd 96(%r8), %ymm2
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm2[0],ymm0[1,2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm7[0,1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm1 = mem[0,0]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3]
-; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 112(%rsi), %xmm0
-; AVX1-ONLY-NEXT: vmovaps 112(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],mem[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vbroadcastsd 112(%rcx), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 128(%rsi), %xmm1
; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vmovaps 128(%rdi), %xmm0
; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm1
-; AVX1-ONLY-NEXT: vmovapd 128(%r8), %ymm0
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm0[0],ymm1[1,2,3]
-; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm5 = mem[0,0]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm1[0],ymm5[1],ymm1[2,3]
-; AVX1-ONLY-NEXT: vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 144(%rsi), %xmm1
-; AVX1-ONLY-NEXT: vmovaps 144(%rdi), %xmm5
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm5[0],ymm1[0],ymm5[2],ymm1[2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],mem[4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT: vbroadcastsd 144(%rcx), %ymm5
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm5[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 160(%rsi), %xmm5
-; AVX1-ONLY-NEXT: vmovaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 160(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; AVX1-ONLY-NEXT: vmovapd 128(%r8), %ymm7
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm7[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm1 = mem[0,0]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3]
+; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 160(%rsi), %xmm1
; AVX1-ONLY-NEXT: vmovaps %xmm1, (%rsp) # 16-byte Spill
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm5[1]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm5
-; AVX1-ONLY-NEXT: vmovapd 160(%r8), %ymm1
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm5 = ymm1[0],ymm5[1,2,3]
-; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm6 = mem[0,0]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm5 = ymm5[0],ymm6[1],ymm5[2,3]
-; AVX1-ONLY-NEXT: vmovupd %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 176(%rsi), %xmm5
-; AVX1-ONLY-NEXT: vmovaps 176(%rdi), %xmm6
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm5 = ymm6[0],ymm5[0],ymm6[2],ymm5[2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],mem[4,5],ymm5[6,7]
-; AVX1-ONLY-NEXT: vbroadcastsd 176(%rcx), %ymm6
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5],ymm6[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 192(%rsi), %xmm6
-; AVX1-ONLY-NEXT: vmovaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 192(%rdi), %xmm5
-; AVX1-ONLY-NEXT: vmovaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm5 = xmm5[1],xmm6[1]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm5, %ymm0, %ymm6
-; AVX1-ONLY-NEXT: vmovapd 192(%r8), %ymm5
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm6 = ymm5[0],ymm6[1,2,3]
-; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm7 = mem[0,0]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm6 = ymm6[0],ymm7[1],ymm6[2,3]
-; AVX1-ONLY-NEXT: vmovupd %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 208(%rsi), %xmm6
-; AVX1-ONLY-NEXT: vmovaps 208(%rdi), %xmm7
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm6 = ymm7[0],ymm6[0],ymm7[2],ymm6[2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],mem[4,5],ymm6[6,7]
-; AVX1-ONLY-NEXT: vbroadcastsd 208(%rcx), %ymm7
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5],ymm7[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 224(%rsi), %xmm7
-; AVX1-ONLY-NEXT: vmovaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 224(%rdi), %xmm6
-; AVX1-ONLY-NEXT: vmovaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm6 = xmm6[1],xmm7[1]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm6, %ymm0, %ymm7
-; AVX1-ONLY-NEXT: vmovapd 224(%r8), %ymm6
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm7 = ymm6[0],ymm7[1,2,3]
+; AVX1-ONLY-NEXT: vmovaps 160(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm1
+; AVX1-ONLY-NEXT: vmovapd 160(%r8), %ymm6
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm6[0,1],ymm1[2,3]
+; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm2 = mem[0,0]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm1[0],ymm2[1],ymm1[2,3]
+; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 192(%rsi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 192(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX1-ONLY-NEXT: vmovapd 192(%r8), %ymm2
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3]
+; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm3 = mem[0,0]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm1[0],ymm3[1],ymm1[2,3]
+; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 224(%rsi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 224(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX1-ONLY-NEXT: vmovapd 224(%r8), %ymm15
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm15[0,1],ymm1[2,3]
+; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm3 = mem[0,0]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm1[0],ymm3[1],ymm1[2,3]
+; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 256(%rsi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 256(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX1-ONLY-NEXT: vmovapd 256(%r8), %ymm13
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm13[0,1],ymm1[2,3]
+; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm3 = mem[0,0]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm1[0],ymm3[1],ymm1[2,3]
+; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 288(%rsi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 288(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX1-ONLY-NEXT: vmovapd 288(%r8), %ymm11
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm11[0,1],ymm1[2,3]
+; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm3 = mem[0,0]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm1[0],ymm3[1],ymm1[2,3]
+; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 320(%rsi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 320(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm3
+; AVX1-ONLY-NEXT: vmovapd 320(%r8), %ymm1
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm3 = ymm1[0,1],ymm3[2,3]
; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm8 = mem[0,0]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm7 = ymm7[0],ymm8[1],ymm7[2,3]
-; AVX1-ONLY-NEXT: vmovupd %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 240(%rsi), %xmm7
-; AVX1-ONLY-NEXT: vmovaps 240(%rdi), %xmm8
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm7 = ymm8[0],ymm7[0],ymm8[2],ymm7[2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3],mem[4,5],ymm7[6,7]
-; AVX1-ONLY-NEXT: vbroadcastsd 240(%rcx), %ymm8
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5],ymm8[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 256(%rsi), %xmm8
-; AVX1-ONLY-NEXT: vmovaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 256(%rdi), %xmm7
-; AVX1-ONLY-NEXT: vmovaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm7 = xmm7[1],xmm8[1]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm7, %ymm0, %ymm8
-; AVX1-ONLY-NEXT: vmovapd 256(%r8), %ymm7
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm8 = ymm7[0],ymm8[1,2,3]
-; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm9 = mem[0,0]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm8 = ymm8[0],ymm9[1],ymm8[2,3]
-; AVX1-ONLY-NEXT: vmovupd %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 272(%rsi), %xmm8
-; AVX1-ONLY-NEXT: vmovaps 272(%rdi), %xmm9
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm8 = ymm9[0],ymm8[0],ymm9[2],ymm8[2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3],mem[4,5],ymm8[6,7]
-; AVX1-ONLY-NEXT: vbroadcastsd 272(%rcx), %ymm9
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5],ymm9[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 288(%rsi), %xmm9
-; AVX1-ONLY-NEXT: vmovaps %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 288(%rdi), %xmm8
-; AVX1-ONLY-NEXT: vmovaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm8 = xmm8[1],xmm9[1]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm8, %ymm0, %ymm8
-; AVX1-ONLY-NEXT: vmovapd 288(%r8), %ymm13
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm8 = ymm13[0],ymm8[1,2,3]
-; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm9 = mem[0,0]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm8 = ymm8[0],ymm9[1],ymm8[2,3]
-; AVX1-ONLY-NEXT: vmovupd %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 304(%rsi), %xmm8
-; AVX1-ONLY-NEXT: vmovaps 304(%rdi), %xmm9
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm8 = ymm9[0],ymm8[0],ymm9[2],ymm8[2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3],mem[4,5],ymm8[6,7]
-; AVX1-ONLY-NEXT: vbroadcastsd 304(%rcx), %ymm9
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5],ymm9[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 320(%rsi), %xmm9
-; AVX1-ONLY-NEXT: vmovaps %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 320(%rdi), %xmm8
-; AVX1-ONLY-NEXT: vmovaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm8 = xmm8[1],xmm9[1]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm8, %ymm0, %ymm8
-; AVX1-ONLY-NEXT: vmovapd 320(%r8), %ymm12
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm8 = ymm12[0],ymm8[1,2,3]
-; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm9 = mem[0,0]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm8 = ymm8[0],ymm9[1],ymm8[2,3]
-; AVX1-ONLY-NEXT: vmovupd %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 336(%rsi), %xmm8
-; AVX1-ONLY-NEXT: vmovaps 336(%rdi), %xmm9
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm8 = ymm9[0],ymm8[0],ymm9[2],ymm8[2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3],mem[4,5],ymm8[6,7]
-; AVX1-ONLY-NEXT: vbroadcastsd 336(%rcx), %ymm9
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5],ymm9[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 352(%rsi), %xmm9
-; AVX1-ONLY-NEXT: vmovaps %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 352(%rdi), %xmm8
-; AVX1-ONLY-NEXT: vmovaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm8 = xmm8[1],xmm9[1]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm8, %ymm0, %ymm8
-; AVX1-ONLY-NEXT: vmovaps 352(%r8), %ymm9
-; AVX1-ONLY-NEXT: vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm9[0,1],ymm8[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm9 = mem[0,0]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1],ymm9[2,3],ymm8[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 368(%rsi), %xmm8
-; AVX1-ONLY-NEXT: vmovaps 368(%rdi), %xmm9
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm8 = ymm9[0],ymm8[0],ymm9[2],ymm8[2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3],mem[4,5],ymm8[6,7]
-; AVX1-ONLY-NEXT: vbroadcastsd 368(%rcx), %ymm9
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5],ymm9[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 384(%rsi), %xmm9
-; AVX1-ONLY-NEXT: vmovaps %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm3[0],ymm8[1],ymm3[2,3]
+; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 352(%rsi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 352(%rdi), %xmm3
+; AVX1-ONLY-NEXT: vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm3 = xmm3[1],xmm0[1]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm8
+; AVX1-ONLY-NEXT: vmovapd 352(%r8), %ymm3
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm8 = ymm3[0,1],ymm8[2,3]
+; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm10 = mem[0,0]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm8[0],ymm10[1],ymm8[2,3]
+; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 384(%rsi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vmovaps 384(%rdi), %xmm8
; AVX1-ONLY-NEXT: vmovaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm8 = xmm8[1],xmm9[1]
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm8 = xmm8[1],xmm0[1]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm8, %ymm0, %ymm8
-; AVX1-ONLY-NEXT: vmovaps 384(%r8), %ymm9
-; AVX1-ONLY-NEXT: vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm9[0,1],ymm8[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm9 = mem[0,0]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1],ymm9[2,3],ymm8[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 400(%rsi), %xmm8
-; AVX1-ONLY-NEXT: vmovaps 400(%rdi), %xmm9
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm8 = ymm9[0],ymm8[0],ymm9[2],ymm8[2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3],mem[4,5],ymm8[6,7]
-; AVX1-ONLY-NEXT: vbroadcastsd 400(%rcx), %ymm9
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5],ymm9[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 416(%rsi), %xmm9
-; AVX1-ONLY-NEXT: vmovaps %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovapd 384(%r8), %ymm12
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm8 = ymm12[0,1],ymm8[2,3]
+; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm10 = mem[0,0]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm8[0],ymm10[1],ymm8[2,3]
+; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 416(%rsi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vmovaps 416(%rdi), %xmm8
; AVX1-ONLY-NEXT: vmovaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm8 = xmm8[1],xmm9[1]
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm8 = xmm8[1],xmm0[1]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm8, %ymm0, %ymm8
-; AVX1-ONLY-NEXT: vmovaps 416(%r8), %ymm9
-; AVX1-ONLY-NEXT: vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm9[0,1],ymm8[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm9 = mem[0,0]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1],ymm9[2,3],ymm8[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 432(%rsi), %xmm8
-; AVX1-ONLY-NEXT: vmovaps 432(%rdi), %xmm9
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm8 = ymm9[0],ymm8[0],ymm9[2],ymm8[2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3],mem[4,5],ymm8[6,7]
-; AVX1-ONLY-NEXT: vbroadcastsd 432(%rcx), %ymm9
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5],ymm9[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 448(%rsi), %xmm9
-; AVX1-ONLY-NEXT: vmovaps %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 416(%r8), %ymm0
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm0[0,1,2,3],ymm8[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm10 = mem[0,0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm8[0,1],ymm10[2,3],ymm8[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 448(%rsi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vmovaps 448(%rdi), %xmm8
; AVX1-ONLY-NEXT: vmovaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm8 = xmm8[1],xmm9[1]
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm8 = xmm8[1],xmm0[1]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm8, %ymm0, %ymm8
-; AVX1-ONLY-NEXT: vmovaps 448(%r8), %ymm9
-; AVX1-ONLY-NEXT: vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm9[0,1],ymm8[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovapd 448(%r8), %ymm10
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm8 = ymm10[0,1],ymm8[2,3]
; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm9 = mem[0,0]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1],ymm9[2,3],ymm8[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 464(%rsi), %xmm8
-; AVX1-ONLY-NEXT: vmovaps 464(%rdi), %xmm9
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm8 = ymm9[0],ymm8[0],ymm9[2],ymm8[2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3],mem[4,5],ymm8[6,7]
-; AVX1-ONLY-NEXT: vbroadcastsd 464(%rcx), %ymm9
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5],ymm9[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 480(%rsi), %xmm9
-; AVX1-ONLY-NEXT: vmovaps %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm8[0],ymm9[1],ymm8[2,3]
+; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 480(%rsi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vmovaps 480(%rdi), %xmm8
; AVX1-ONLY-NEXT: vmovaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm8 = xmm8[1],xmm9[1]
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm8 = xmm8[1],xmm0[1]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm8, %ymm0, %ymm8
-; AVX1-ONLY-NEXT: vmovaps 480(%r8), %ymm9
-; AVX1-ONLY-NEXT: vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm9[0,1],ymm8[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps 480(%r8), %ymm0
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm0[0,1,2,3],ymm8[4,5,6,7]
; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm9 = mem[0,0]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1],ymm9[2,3],ymm8[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 496(%rsi), %xmm8
-; AVX1-ONLY-NEXT: vmovaps 496(%rdi), %xmm9
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm8 = ymm9[0],ymm8[0],ymm9[2],ymm8[2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3],mem[4,5],ymm8[6,7]
-; AVX1-ONLY-NEXT: vbroadcastsd 496(%rcx), %ymm9
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5],ymm9[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm8[0,1],ymm9[2,3],ymm8[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm8 = mem[2,3,2,3]
; AVX1-ONLY-NEXT: vbroadcastsd 8(%r8), %ymm9
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3],ymm9[4,5],ymm8[6,7]
-; AVX1-ONLY-NEXT: vmovaps (%rcx), %xmm9
-; AVX1-ONLY-NEXT: vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vinsertf128 $1, (%r9), %ymm9, %ymm9
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1],ymm9[2,3],ymm8[4,5],ymm9[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3],ymm9[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps (%rcx), %xmm0
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vinsertf128 $1, (%r9), %ymm0, %ymm9
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm8[0,1],ymm9[2,3],ymm8[4,5],ymm9[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovapd (%rdi), %ymm8
; AVX1-ONLY-NEXT: vmovapd (%rsi), %ymm9
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm8 = ymm8[1],ymm9[1],ymm8[3],ymm9[3]
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm4 = ymm4[2,3],ymm8[2,3]
-; AVX1-ONLY-NEXT: vmovapd (%r9), %ymm8
-; AVX1-ONLY-NEXT: vmovupd %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm8 = ymm8[2,3],ymm9[2,3]
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm4 = ymm4[0],ymm8[0],ymm4[2],ymm8[3]
-; AVX1-ONLY-NEXT: vmovupd %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm5 = ymm5[2,3],ymm8[2,3]
+; AVX1-ONLY-NEXT: vmovapd (%r9), %ymm0
+; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm9 = ymm0[2,3],ymm9[2,3]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm0 = ymm5[0],ymm9[0],ymm5[2],ymm9[3]
+; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm5 = mem[2,3,2,3]
+; AVX1-ONLY-NEXT: vbroadcastsd 40(%r8), %ymm9
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm9[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps 32(%rcx), %xmm0
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vinsertf128 $1, 32(%r9), %ymm0, %ymm9
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm5[0,1],ymm9[2,3],ymm5[4,5],ymm9[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovapd 32(%rdi), %ymm5
+; AVX1-ONLY-NEXT: vmovapd 32(%rsi), %ymm9
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm5 = ymm5[1],ymm9[1],ymm5[3],ymm9[3]
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm4 = ymm4[2,3],ymm5[2,3]
+; AVX1-ONLY-NEXT: vmovapd 32(%r9), %ymm5
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm9 = ymm5[2,3],ymm9[2,3]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm0 = ymm4[0],ymm9[0],ymm4[2],ymm9[3]
+; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm4 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT: vbroadcastsd 40(%r8), %ymm8
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm8[4,5],ymm4[6,7]
-; AVX1-ONLY-NEXT: vmovaps 32(%rcx), %xmm8
+; AVX1-ONLY-NEXT: vbroadcastsd 72(%r8), %ymm9
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm9[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps 64(%rcx), %xmm0
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vinsertf128 $1, 64(%r9), %ymm0, %ymm9
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm4[0,1],ymm9[2,3],ymm4[4,5],ymm9[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovapd 64(%rdi), %ymm4
+; AVX1-ONLY-NEXT: vmovapd 64(%rsi), %ymm9
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm4 = ymm4[1],ymm9[1],ymm4[3],ymm9[3]
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm4 = ymm14[2,3],ymm4[2,3]
+; AVX1-ONLY-NEXT: vmovapd 64(%r9), %ymm14
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm9 = ymm14[2,3],ymm9[2,3]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm0 = ymm4[0],ymm9[0],ymm4[2],ymm9[3]
+; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm4 = mem[2,3,2,3]
+; AVX1-ONLY-NEXT: vbroadcastsd 104(%r8), %ymm9
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm9[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps 96(%rcx), %xmm0
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vinsertf128 $1, 96(%r9), %ymm0, %ymm9
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm4[0,1],ymm9[2,3],ymm4[4,5],ymm9[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovapd 96(%rdi), %ymm4
+; AVX1-ONLY-NEXT: vmovapd 96(%rsi), %ymm9
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm4 = ymm4[1],ymm9[1],ymm4[3],ymm9[3]
+; AVX1-ONLY-NEXT: vperm2f128 $19, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm0 = mem[2,3],ymm4[2,3]
+; AVX1-ONLY-NEXT: vmovapd 96(%r9), %ymm4
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm9 = ymm4[2,3],ymm9[2,3]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[0],ymm9[0],ymm0[2],ymm9[3]
+; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm0 = mem[2,3,2,3]
+; AVX1-ONLY-NEXT: vbroadcastsd 136(%r8), %ymm9
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm9[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps 128(%rcx), %xmm8
; AVX1-ONLY-NEXT: vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vinsertf128 $1, 32(%r9), %ymm8, %ymm8
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm8[2,3],ymm4[4,5],ymm8[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovapd 32(%rdi), %ymm4
-; AVX1-ONLY-NEXT: vmovapd 32(%rsi), %ymm8
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm4 = ymm4[1],ymm8[1],ymm4[3],ymm8[3]
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm3 = ymm3[2,3],ymm4[2,3]
-; AVX1-ONLY-NEXT: vmovapd 32(%r9), %ymm4
-; AVX1-ONLY-NEXT: vmovupd %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm4 = ymm4[2,3],ymm8[2,3]
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm3 = ymm3[0],ymm4[0],ymm3[2],ymm4[3]
-; AVX1-ONLY-NEXT: vmovupd %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm3 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT: vbroadcastsd 72(%r8), %ymm4
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm4[4,5],ymm3[6,7]
-; AVX1-ONLY-NEXT: vmovaps 64(%rcx), %xmm4
-; AVX1-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vinsertf128 $1, 64(%r9), %ymm4, %ymm4
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1],ymm4[2,3],ymm3[4,5],ymm4[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovapd 64(%rdi), %ymm3
-; AVX1-ONLY-NEXT: vmovapd 64(%rsi), %ymm4
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm3 = ymm3[1],ymm4[1],ymm3[3],ymm4[3]
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm3 = ymm10[2,3],ymm3[2,3]
-; AVX1-ONLY-NEXT: vmovapd 64(%r9), %ymm8
-; AVX1-ONLY-NEXT: vmovupd %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm4 = ymm8[2,3],ymm4[2,3]
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm3 = ymm3[0],ymm4[0],ymm3[2],ymm4[3]
-; AVX1-ONLY-NEXT: vmovupd %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm3 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT: vbroadcastsd 104(%r8), %ymm4
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm4[4,5],ymm3[6,7]
-; AVX1-ONLY-NEXT: vmovaps 96(%rcx), %xmm4
-; AVX1-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vinsertf128 $1, 96(%r9), %ymm4, %ymm4
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1],ymm4[2,3],ymm3[4,5],ymm4[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovapd 96(%rdi), %ymm3
-; AVX1-ONLY-NEXT: vmovapd 96(%rsi), %ymm4
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm3 = ymm3[1],ymm4[1],ymm3[3],ymm4[3]
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm2 = ymm2[2,3],ymm3[2,3]
-; AVX1-ONLY-NEXT: vmovapd 96(%r9), %ymm10
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm3 = ymm10[2,3],ymm4[2,3]
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm2 = ymm2[0],ymm3[0],ymm2[2],ymm3[3]
-; AVX1-ONLY-NEXT: vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm2 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT: vbroadcastsd 136(%r8), %ymm3
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5],ymm2[6,7]
-; AVX1-ONLY-NEXT: vmovaps 128(%rcx), %xmm3
-; AVX1-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vinsertf128 $1, 128(%r9), %ymm3, %ymm3
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm3[2,3],ymm2[4,5],ymm3[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovapd 128(%rdi), %ymm2
-; AVX1-ONLY-NEXT: vmovapd 128(%rsi), %ymm3
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm2 = ymm2[1],ymm3[1],ymm2[3],ymm3[3]
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm2[2,3]
-; AVX1-ONLY-NEXT: vmovapd 128(%r9), %ymm14
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm2 = ymm14[2,3],ymm3[2,3]
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[0],ymm2[0],ymm0[2],ymm2[3]
+; AVX1-ONLY-NEXT: vinsertf128 $1, 128(%r9), %ymm8, %ymm9
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm9[2,3],ymm0[4,5],ymm9[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovapd 128(%rdi), %ymm0
+; AVX1-ONLY-NEXT: vmovapd 128(%rsi), %ymm9
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm9[1],ymm0[3],ymm9[3]
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm7[2,3],ymm0[2,3]
+; AVX1-ONLY-NEXT: vmovapd 128(%r9), %ymm7
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm9 = ymm7[2,3],ymm9[2,3]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[0],ymm9[0],ymm0[2],ymm9[3]
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm0 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT: vbroadcastsd 168(%r8), %ymm2
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vmovaps 160(%rcx), %xmm2
-; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vinsertf128 $1, 160(%r9), %ymm2, %ymm2
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm2[2,3],ymm0[4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT: vbroadcastsd 168(%r8), %ymm9
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm9[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps 160(%rcx), %xmm8
+; AVX1-ONLY-NEXT: vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vinsertf128 $1, 160(%r9), %ymm8, %ymm9
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm9[2,3],ymm0[4,5],ymm9[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovapd 160(%rdi), %ymm0
-; AVX1-ONLY-NEXT: vmovapd 160(%rsi), %ymm2
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm2[1],ymm0[3],ymm2[3]
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
-; AVX1-ONLY-NEXT: vmovapd 160(%r9), %ymm11
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm11[2,3],ymm2[2,3]
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[3]
+; AVX1-ONLY-NEXT: vmovapd 160(%rdi), %ymm9
+; AVX1-ONLY-NEXT: vmovapd 160(%rsi), %ymm0
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm9 = ymm9[1],ymm0[1],ymm9[3],ymm0[3]
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm6 = ymm6[2,3],ymm9[2,3]
+; AVX1-ONLY-NEXT: vmovapd 160(%r9), %ymm8
+; AVX1-ONLY-NEXT: vmovupd %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm8[2,3],ymm0[2,3]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm0 = ymm6[0],ymm0[0],ymm6[2],ymm0[3]
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm0 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT: vbroadcastsd 200(%r8), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vmovaps 192(%rcx), %xmm1
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vinsertf128 $1, 192(%r9), %ymm1, %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT: vbroadcastsd 200(%r8), %ymm6
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm6[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps 192(%rcx), %xmm6
+; AVX1-ONLY-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vinsertf128 $1, 192(%r9), %ymm6, %ymm6
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm6[2,3],ymm0[4,5],ymm6[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovapd 192(%rdi), %ymm0
-; AVX1-ONLY-NEXT: vmovapd 192(%rsi), %ymm1
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm5[2,3],ymm0[2,3]
+; AVX1-ONLY-NEXT: vmovapd 192(%rsi), %ymm6
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm6[1],ymm0[3],ymm6[3]
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm2[2,3],ymm0[2,3]
; AVX1-ONLY-NEXT: vmovapd 192(%r9), %ymm8
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm8[2,3],ymm1[2,3]
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[3]
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm6 = ymm8[2,3],ymm6[2,3]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[0],ymm6[0],ymm0[2],ymm6[3]
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm0 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT: vbroadcastsd 232(%r8), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vmovaps 224(%rcx), %xmm1
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vinsertf128 $1, 224(%r9), %ymm1, %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT: vbroadcastsd 232(%r8), %ymm6
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm6[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps 224(%rcx), %xmm2
+; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vinsertf128 $1, 224(%r9), %ymm2, %ymm6
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm6[2,3],ymm0[4,5],ymm6[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovapd 224(%rdi), %ymm0
-; AVX1-ONLY-NEXT: vmovapd 224(%rsi), %ymm1
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm6[2,3],ymm0[2,3]
-; AVX1-ONLY-NEXT: vmovapd 224(%r9), %ymm9
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm9[2,3],ymm1[2,3]
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[3]
+; AVX1-ONLY-NEXT: vmovapd 224(%rsi), %ymm6
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm6[1],ymm0[3],ymm6[3]
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm15[2,3],ymm0[2,3]
+; AVX1-ONLY-NEXT: vmovapd 224(%r9), %ymm15
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm6 = ymm15[2,3],ymm6[2,3]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[0],ymm6[0],ymm0[2],ymm6[3]
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm0 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT: vbroadcastsd 264(%r8), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vmovaps 256(%rcx), %xmm1
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vinsertf128 $1, 256(%r9), %ymm1, %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT: vbroadcastsd 264(%r8), %ymm6
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm6[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps 256(%rcx), %xmm2
+; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vinsertf128 $1, 256(%r9), %ymm2, %ymm6
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm6[2,3],ymm0[4,5],ymm6[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovapd 256(%rdi), %ymm0
-; AVX1-ONLY-NEXT: vmovapd 256(%rsi), %ymm1
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm7[2,3],ymm0[2,3]
-; AVX1-ONLY-NEXT: vmovapd 256(%r9), %ymm7
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm7[2,3],ymm1[2,3]
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[3]
+; AVX1-ONLY-NEXT: vmovapd 256(%rsi), %ymm6
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm6[1],ymm0[3],ymm6[3]
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm13[2,3],ymm0[2,3]
+; AVX1-ONLY-NEXT: vmovapd 256(%r9), %ymm13
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm6 = ymm13[2,3],ymm6[2,3]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[0],ymm6[0],ymm0[2],ymm6[3]
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm0 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT: vbroadcastsd 296(%r8), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vmovaps 288(%rcx), %xmm15
-; AVX1-ONLY-NEXT: vinsertf128 $1, 288(%r9), %ymm15, %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT: vbroadcastsd 296(%r8), %ymm6
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm6[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps 288(%rcx), %xmm2
+; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vinsertf128 $1, 288(%r9), %ymm2, %ymm6
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm6[2,3],ymm0[4,5],ymm6[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovapd 288(%rdi), %ymm0
-; AVX1-ONLY-NEXT: vmovapd 288(%rsi), %ymm1
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm13[2,3],ymm0[2,3]
+; AVX1-ONLY-NEXT: vmovapd 288(%rdi), %ymm6
+; AVX1-ONLY-NEXT: vmovapd 288(%rsi), %ymm0
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm6 = ymm6[1],ymm0[1],ymm6[3],ymm0[3]
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm11 = ymm11[2,3],ymm6[2,3]
; AVX1-ONLY-NEXT: vmovapd 288(%r9), %ymm6
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm6[2,3],ymm1[2,3]
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[3]
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm6[2,3],ymm0[2,3]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm0 = ymm11[0],ymm0[0],ymm11[2],ymm0[3]
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm0 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT: vbroadcastsd 328(%r8), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vmovaps 320(%rcx), %xmm1
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vinsertf128 $1, 320(%r9), %ymm1, %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT: vbroadcastsd 328(%r8), %ymm11
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm11[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps 320(%rcx), %xmm2
+; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vinsertf128 $1, 320(%r9), %ymm2, %ymm11
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm11[2,3],ymm0[4,5],ymm11[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovapd 320(%rdi), %ymm0
-; AVX1-ONLY-NEXT: vmovapd 320(%rsi), %ymm1
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm12[2,3],ymm0[2,3]
-; AVX1-ONLY-NEXT: vmovapd 320(%r9), %ymm5
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm5[2,3],ymm1[2,3]
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[3]
+; AVX1-ONLY-NEXT: vmovapd 320(%rsi), %ymm11
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm11[1],ymm0[3],ymm11[3]
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm1[2,3],ymm0[2,3]
+; AVX1-ONLY-NEXT: vmovapd 320(%r9), %ymm2
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm11 = ymm2[2,3],ymm11[2,3]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm0 = ymm1[0],ymm11[0],ymm1[2],ymm11[3]
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm0 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT: vbroadcastsd 360(%r8), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vmovaps 352(%rcx), %xmm1
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vinsertf128 $1, 352(%r9), %ymm1, %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm1 = mem[2,3,2,3]
+; AVX1-ONLY-NEXT: vbroadcastsd 360(%r8), %ymm11
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm11[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps 352(%rcx), %xmm0
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovapd 352(%rdi), %ymm0
-; AVX1-ONLY-NEXT: vmovapd 352(%rsi), %ymm1
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
-; AVX1-ONLY-NEXT: vperm2f128 $19, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm0 = mem[2,3],ymm0[2,3]
-; AVX1-ONLY-NEXT: vmovapd 352(%r9), %ymm4
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm4[2,3],ymm1[2,3]
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[3]
+; AVX1-ONLY-NEXT: vinsertf128 $1, 352(%r9), %ymm0, %ymm11
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm11[2,3],ymm1[4,5],ymm11[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovapd 352(%rdi), %ymm1
+; AVX1-ONLY-NEXT: vmovapd 352(%rsi), %ymm11
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm1 = ymm1[1],ymm11[1],ymm1[3],ymm11[3]
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm3 = ymm3[2,3],ymm1[2,3]
+; AVX1-ONLY-NEXT: vmovapd 352(%r9), %ymm9
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm11 = ymm9[2,3],ymm11[2,3]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm0 = ymm3[0],ymm11[0],ymm3[2],ymm11[3]
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm0 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT: vbroadcastsd 392(%r8), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vmovaps 384(%rcx), %xmm1
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vinsertf128 $1, 384(%r9), %ymm1, %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm3 = mem[2,3,2,3]
+; AVX1-ONLY-NEXT: vbroadcastsd 392(%r8), %ymm11
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm11[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps 384(%rcx), %xmm0
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovapd 384(%rdi), %ymm0
-; AVX1-ONLY-NEXT: vmovapd 384(%rsi), %ymm1
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
-; AVX1-ONLY-NEXT: vperm2f128 $19, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm0 = mem[2,3],ymm0[2,3]
-; AVX1-ONLY-NEXT: vmovapd 384(%r9), %ymm3
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm3[2,3],ymm1[2,3]
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[3]
+; AVX1-ONLY-NEXT: vinsertf128 $1, 384(%r9), %ymm0, %ymm11
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm3[0,1],ymm11[2,3],ymm3[4,5],ymm11[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovapd 384(%rdi), %ymm3
+; AVX1-ONLY-NEXT: vmovapd 384(%rsi), %ymm0
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm3 = ymm3[1],ymm0[1],ymm3[3],ymm0[3]
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm3 = ymm12[2,3],ymm3[2,3]
+; AVX1-ONLY-NEXT: vmovapd 384(%r9), %ymm11
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm11[2,3],ymm0[2,3]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm0 = ymm3[0],ymm0[0],ymm3[2],ymm0[3]
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm0 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT: vbroadcastsd 424(%r8), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vmovaps 416(%rcx), %xmm1
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vinsertf128 $1, 416(%r9), %ymm1, %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT: vbroadcastsd 424(%r8), %ymm3
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps 416(%rcx), %xmm3
+; AVX1-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vinsertf128 $1, 416(%r9), %ymm3, %ymm3
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm3[2,3],ymm0[4,5],ymm3[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovapd 416(%rdi), %ymm0
-; AVX1-ONLY-NEXT: vmovapd 416(%rsi), %ymm1
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
+; AVX1-ONLY-NEXT: vmovapd 416(%rsi), %ymm3
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm3[1],ymm0[3],ymm3[3]
; AVX1-ONLY-NEXT: vperm2f128 $19, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm0 = mem[2,3],ymm0[2,3]
-; AVX1-ONLY-NEXT: vmovapd 416(%r9), %ymm2
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm2[2,3],ymm1[2,3]
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[3]
+; AVX1-ONLY-NEXT: vmovapd 416(%r9), %ymm12
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm3 = ymm12[2,3],ymm3[2,3]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[0],ymm3[0],ymm0[2],ymm3[3]
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm0 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT: vbroadcastsd 456(%r8), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vmovaps 448(%rcx), %xmm1
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vinsertf128 $1, 448(%r9), %ymm1, %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT: vbroadcastsd 456(%r8), %ymm3
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm0[0,1,2,3],ymm3[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps 448(%rcx), %xmm0
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vinsertf128 $1, 448(%r9), %ymm0, %ymm3
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm3[2,3],ymm1[4,5],ymm3[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovapd 448(%rdi), %ymm0
-; AVX1-ONLY-NEXT: vmovapd 448(%rsi), %ymm12
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm12[1],ymm0[3],ymm12[3]
-; AVX1-ONLY-NEXT: vperm2f128 $19, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm0 = mem[2,3],ymm0[2,3]
-; AVX1-ONLY-NEXT: vmovapd 448(%r9), %ymm1
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm12 = ymm1[2,3],ymm12[2,3]
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[0],ymm12[0],ymm0[2],ymm12[3]
+; AVX1-ONLY-NEXT: vmovapd 448(%rsi), %ymm3
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm3[1],ymm0[3],ymm3[3]
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm10[2,3],ymm0[2,3]
+; AVX1-ONLY-NEXT: vmovapd 448(%r9), %ymm10
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm3 = ymm10[2,3],ymm3[2,3]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[0],ymm3[0],ymm0[2],ymm3[3]
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm0 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT: vbroadcastsd 488(%r8), %ymm12
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm12[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vmovaps 480(%rcx), %xmm12
-; AVX1-ONLY-NEXT: vmovups %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vinsertf128 $1, 480(%r9), %ymm12, %ymm12
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm12[2,3],ymm0[4,5],ymm12[6,7]
+; AVX1-ONLY-NEXT: vbroadcastsd 488(%r8), %ymm3
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps 480(%rcx), %xmm1
+; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vinsertf128 $1, 480(%r9), %ymm1, %ymm3
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm3[2,3],ymm0[4,5],ymm3[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovapd 480(%rdi), %ymm0
-; AVX1-ONLY-NEXT: vmovapd 480(%rsi), %ymm12
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm12[1],ymm0[3],ymm12[3]
-; AVX1-ONLY-NEXT: vperm2f128 $19, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm13 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm13 = mem[2,3],ymm0[2,3]
-; AVX1-ONLY-NEXT: vmovapd 480(%r9), %ymm0
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm12 = ymm0[2,3],ymm12[2,3]
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm12 = ymm13[0],ymm12[0],ymm13[2],ymm12[3]
-; AVX1-ONLY-NEXT: vmovupd %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 16(%rdx), %xmm12
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm12 = xmm12[1],mem[1]
-; AVX1-ONLY-NEXT: vbroadcastsd 24(%r8), %ymm13
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3],ymm13[4,5],ymm12[6,7]
-; AVX1-ONLY-NEXT: vblendps $192, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm12 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm12 = ymm12[0,1,2,3,4,5],mem[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 48(%rdx), %xmm12
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm12 = xmm12[1],mem[1]
-; AVX1-ONLY-NEXT: vbroadcastsd 56(%r8), %ymm13
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3],ymm13[4,5],ymm12[6,7]
-; AVX1-ONLY-NEXT: vblendps $192, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm12 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm12 = ymm12[0,1,2,3,4,5],mem[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 80(%rdx), %xmm12
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm12 = xmm12[1],mem[1]
-; AVX1-ONLY-NEXT: vbroadcastsd 88(%r8), %ymm13
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3],ymm13[4,5],ymm12[6,7]
-; AVX1-ONLY-NEXT: vblendps $192, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm12 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm12 = ymm12[0,1,2,3,4,5],mem[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovapd 112(%rdx), %xmm12
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm12 = xmm12[1],mem[1]
-; AVX1-ONLY-NEXT: vbroadcastsd 120(%r8), %ymm13
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm12 = ymm12[0,1],ymm13[2],ymm12[3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm10 = ymm12[0,1,2],ymm10[3]
-; AVX1-ONLY-NEXT: vmovupd %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovapd 144(%rdx), %xmm10
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm10 = xmm10[1],mem[1]
-; AVX1-ONLY-NEXT: vbroadcastsd 152(%r8), %ymm12
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm10 = ymm10[0,1],ymm12[2],ymm10[3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm10 = ymm10[0,1,2],ymm14[3]
-; AVX1-ONLY-NEXT: vmovupd %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovapd 176(%rdx), %xmm10
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm10 = xmm10[1],mem[1]
-; AVX1-ONLY-NEXT: vbroadcastsd 184(%r8), %ymm12
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm10 = ymm10[0,1],ymm12[2],ymm10[3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm10 = ymm10[0,1,2],ymm11[3]
-; AVX1-ONLY-NEXT: vmovupd %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovapd 208(%rdx), %xmm10
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm10 = xmm10[1],mem[1]
-; AVX1-ONLY-NEXT: vbroadcastsd 216(%r8), %ymm11
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm10 = ymm10[0,1],ymm11[2],ymm10[3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm8 = ymm10[0,1,2],ymm8[3]
-; AVX1-ONLY-NEXT: vmovupd %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovapd 240(%rdx), %xmm8
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm8 = xmm8[1],mem[1]
-; AVX1-ONLY-NEXT: vbroadcastsd 248(%r8), %ymm10
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm8 = ymm8[0,1],ymm10[2],ymm8[3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm8 = ymm8[0,1,2],ymm9[3]
-; AVX1-ONLY-NEXT: vmovupd %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovapd 272(%rdx), %xmm8
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm8 = xmm8[1],mem[1]
-; AVX1-ONLY-NEXT: vbroadcastsd 280(%r8), %ymm9
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm8 = ymm8[0,1],ymm9[2],ymm8[3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm7 = ymm8[0,1,2],ymm7[3]
-; AVX1-ONLY-NEXT: vmovupd %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovapd 304(%rdx), %xmm7
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm7 = xmm7[1],mem[1]
-; AVX1-ONLY-NEXT: vbroadcastsd 312(%r8), %ymm8
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm7 = ymm7[0,1],ymm8[2],ymm7[3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm6 = ymm7[0,1,2],ymm6[3]
-; AVX1-ONLY-NEXT: vmovupd %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovapd 336(%rdx), %xmm6
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm6 = xmm6[1],mem[1]
-; AVX1-ONLY-NEXT: vbroadcastsd 344(%r8), %ymm7
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm6 = ymm6[0,1],ymm7[2],ymm6[3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm5 = ymm6[0,1,2],ymm5[3]
-; AVX1-ONLY-NEXT: vmovupd %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovapd 368(%rdx), %xmm5
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm5 = xmm5[1],mem[1]
-; AVX1-ONLY-NEXT: vbroadcastsd 376(%r8), %ymm6
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm5 = ymm5[0,1],ymm6[2],ymm5[3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm4 = ymm5[0,1,2],ymm4[3]
-; AVX1-ONLY-NEXT: vmovupd %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovapd 400(%rdx), %xmm4
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm4 = xmm4[1],mem[1]
-; AVX1-ONLY-NEXT: vbroadcastsd 408(%r8), %ymm5
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm4 = ymm4[0,1],ymm5[2],ymm4[3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3]
-; AVX1-ONLY-NEXT: vmovupd %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovapd 432(%rdx), %xmm3
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm3 = xmm3[1],mem[1]
-; AVX1-ONLY-NEXT: vbroadcastsd 440(%r8), %ymm4
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm3 = ymm3[0,1],ymm4[2],ymm3[3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3]
-; AVX1-ONLY-NEXT: vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovapd 464(%rdx), %xmm2
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm2 = xmm2[1],mem[1]
-; AVX1-ONLY-NEXT: vbroadcastsd 472(%r8), %ymm3
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm2[0,1],ymm3[2],ymm2[3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3]
+; AVX1-ONLY-NEXT: vmovapd 480(%rsi), %ymm3
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm3[1],ymm0[3],ymm3[3]
+; AVX1-ONLY-NEXT: vperm2f128 $19, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm0 = mem[2,3],ymm0[2,3]
+; AVX1-ONLY-NEXT: vmovapd 480(%r9), %ymm1
; AVX1-ONLY-NEXT: vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovapd 496(%rdx), %xmm1
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm1[1],mem[1]
-; AVX1-ONLY-NEXT: vbroadcastsd 504(%r8), %ymm2
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm1[0,1],ymm2[2],ymm1[3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3]
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm3 = ymm1[2,3],ymm3[2,3]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[0],ymm3[0],ymm0[2],ymm3[3]
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 16(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm0 = xmm0[0],mem[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vbroadcastsd 16(%rcx), %ymm3
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm3[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 16(%rdx), %xmm0
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],mem[1]
+; AVX1-ONLY-NEXT: vbroadcastsd 24(%r8), %ymm3
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps $192, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm0 = ymm0[0,1,2,3,4,5],mem[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 48(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm0 = xmm0[0],mem[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vbroadcastsd 48(%rcx), %ymm3
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm3[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovapd 48(%rdx), %xmm0
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],mem[1]
+; AVX1-ONLY-NEXT: vbroadcastsd 56(%r8), %ymm3
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm3[2,3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1,2],ymm5[3]
+; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 80(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm0 = xmm0[0],mem[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vbroadcastsd 80(%rcx), %ymm3
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm3[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovapd 80(%rdx), %xmm0
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],mem[1]
+; AVX1-ONLY-NEXT: vbroadcastsd 88(%r8), %ymm3
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm3[2,3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1,2],ymm14[3]
+; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 112(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm0 = xmm0[0],mem[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vbroadcastsd 112(%rcx), %ymm3
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm3[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovapd 112(%rdx), %xmm0
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],mem[1]
+; AVX1-ONLY-NEXT: vbroadcastsd 120(%r8), %ymm3
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm3[2,3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1,2],ymm4[3]
+; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 144(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm0 = xmm0[0],mem[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vbroadcastsd 144(%rcx), %ymm3
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm3[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovapd 144(%rdx), %xmm0
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],mem[1]
+; AVX1-ONLY-NEXT: vbroadcastsd 152(%r8), %ymm3
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm3[2,3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1,2],ymm7[3]
+; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 176(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm0 = xmm0[0],mem[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vbroadcastsd 176(%rcx), %ymm3
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm3[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 176(%rdx), %xmm0
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],mem[1]
+; AVX1-ONLY-NEXT: vbroadcastsd 184(%r8), %ymm3
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps $192, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm0 = ymm0[0,1,2,3,4,5],mem[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 208(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm0 = xmm0[0],mem[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vbroadcastsd 208(%rcx), %ymm3
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm3[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovapd 208(%rdx), %xmm0
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],mem[1]
+; AVX1-ONLY-NEXT: vbroadcastsd 216(%r8), %ymm3
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm3[2,3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1,2],ymm8[3]
+; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 240(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm0 = xmm0[0],mem[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vbroadcastsd 240(%rcx), %ymm3
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm3[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovapd 240(%rdx), %xmm0
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],mem[1]
+; AVX1-ONLY-NEXT: vbroadcastsd 248(%r8), %ymm3
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm3[2,3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1,2],ymm15[3]
+; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 272(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm0 = xmm0[0],mem[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vbroadcastsd 272(%rcx), %ymm3
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm3[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovapd 272(%rdx), %xmm0
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],mem[1]
+; AVX1-ONLY-NEXT: vbroadcastsd 280(%r8), %ymm3
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm3[2,3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1,2],ymm13[3]
+; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 304(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm0 = xmm0[0],mem[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vbroadcastsd 304(%rcx), %ymm3
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm3[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovapd 304(%rdx), %xmm0
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],mem[1]
+; AVX1-ONLY-NEXT: vbroadcastsd 312(%r8), %ymm3
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm3[2,3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1,2],ymm6[3]
+; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 336(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm0 = xmm0[0],mem[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vbroadcastsd 336(%rcx), %ymm3
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm3[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovapd 336(%rdx), %xmm0
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],mem[1]
+; AVX1-ONLY-NEXT: vbroadcastsd 344(%r8), %ymm3
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm3[2,3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1,2],ymm2[3]
+; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 368(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm0 = xmm0[0],mem[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vbroadcastsd 368(%rcx), %ymm2
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovapd 368(%rdx), %xmm0
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],mem[1]
+; AVX1-ONLY-NEXT: vbroadcastsd 376(%r8), %ymm2
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm2[2,3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1,2],ymm9[3]
+; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 400(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm0 = xmm0[0],mem[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vbroadcastsd 400(%rcx), %ymm1
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovapd 400(%rdx), %xmm0
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],mem[1]
+; AVX1-ONLY-NEXT: vbroadcastsd 408(%r8), %ymm1
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1,2],ymm11[3]
+; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 432(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm0 = xmm0[0],mem[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vbroadcastsd 432(%rcx), %ymm1
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovapd 432(%rdx), %xmm0
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],mem[1]
+; AVX1-ONLY-NEXT: vbroadcastsd 440(%r8), %ymm1
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1,2],ymm12[3]
+; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 464(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm0 = xmm0[0],mem[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vbroadcastsd 464(%rcx), %ymm1
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovapd 464(%rdx), %xmm0
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],mem[1]
+; AVX1-ONLY-NEXT: vbroadcastsd 472(%r8), %ymm1
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1,2],ymm10[3]
+; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 496(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm0 = xmm0[0],mem[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vbroadcastsd 496(%rcx), %ymm1
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 496(%rdx), %xmm0
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],mem[1]
+; AVX1-ONLY-NEXT: vbroadcastsd 504(%r8), %ymm1
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps $192, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm0 = ymm0[0,1,2,3,4,5],mem[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX1-ONLY-NEXT: vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm0 = xmm0[0],mem[0]
@@ -5498,14 +5469,14 @@ define void @store_i64_stride6_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm0 = xmm0[0],mem[0]
; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps (%rsp), %xmm0 # 16-byte Reload
-; AVX1-ONLY-NEXT: vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT: vunpcklpd (%rsp), %xmm0, %xmm0 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm0 = xmm0[0],mem[0]
-; AVX1-ONLY-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vmovaps 160(%rdx), %xmm0
; AVX1-ONLY-NEXT: vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm0 = xmm0[0],mem[0]
-; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX1-ONLY-NEXT: vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm0 = xmm0[0],mem[0]
@@ -5526,7 +5497,8 @@ define void @store_i64_stride6_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm8 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm8 = xmm0[0],mem[0]
; AVX1-ONLY-NEXT: vmovaps 288(%rdx), %xmm0
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm7 = xmm0[0],xmm15[0]
+; AVX1-ONLY-NEXT: vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm7 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm7 = xmm0[0],mem[0]
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX1-ONLY-NEXT: vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm6 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm6 = xmm0[0],mem[0]
@@ -5594,9 +5566,9 @@ define void @store_i64_stride6_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vmovaps %xmm0, 1360(%rax)
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX1-ONLY-NEXT: vmovaps %xmm0, 1344(%rax)
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm0, 976(%rax)
; AVX1-ONLY-NEXT: vmovaps (%rsp), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %xmm0, 976(%rax)
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX1-ONLY-NEXT: vmovaps %xmm0, 960(%rax)
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX1-ONLY-NEXT: vmovaps %xmm0, 592(%rax)
@@ -5778,7 +5750,7 @@ define void @store_i64_stride6_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vmovaps %ymm0, 96(%rax)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm0, 32(%rax)
-; AVX1-ONLY-NEXT: addq $3432, %rsp # imm = 0xD68
+; AVX1-ONLY-NEXT: addq $3464, %rsp # imm = 0xD88
; AVX1-ONLY-NEXT: vzeroupper
; AVX1-ONLY-NEXT: retq
;
@@ -5812,7 +5784,7 @@ define void @store_i64_stride6_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm2[1]
; AVX2-ONLY-NEXT: vbroadcastsd 8(%r8), %ymm2
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovaps 32(%r9), %xmm0
@@ -5825,7 +5797,7 @@ define void @store_i64_stride6_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm3[1],xmm1[1]
; AVX2-ONLY-NEXT: vbroadcastsd 40(%r8), %ymm2
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX2-ONLY-NEXT: vmovaps 64(%r8), %ymm2
; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
@@ -5847,7 +5819,7 @@ define void @store_i64_stride6_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm2[1]
; AVX2-ONLY-NEXT: vbroadcastsd 72(%r8), %ymm2
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5,6,7]
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -5869,7 +5841,7 @@ define void @store_i64_stride6_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm2[1]
; AVX2-ONLY-NEXT: vbroadcastsd 104(%r8), %ymm2
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5,6,7]
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -5891,7 +5863,7 @@ define void @store_i64_stride6_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm2[1]
; AVX2-ONLY-NEXT: vbroadcastsd 136(%r8), %ymm2
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5,6,7]
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -5912,7 +5884,7 @@ define void @store_i64_stride6_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm2[1]
; AVX2-ONLY-NEXT: vbroadcastsd 168(%r8), %ymm2
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5,6,7]
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -5933,7 +5905,7 @@ define void @store_i64_stride6_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm2[1]
; AVX2-ONLY-NEXT: vbroadcastsd 200(%r8), %ymm2
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5,6,7]
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -5954,7 +5926,7 @@ define void @store_i64_stride6_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm2[1]
; AVX2-ONLY-NEXT: vbroadcastsd 232(%r8), %ymm2
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5,6,7]
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -5975,7 +5947,7 @@ define void @store_i64_stride6_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm4[1]
; AVX2-ONLY-NEXT: vbroadcastsd 264(%r8), %ymm4
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm4[4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm4[4,5,6,7]
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm2
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6,7]
; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -5996,7 +5968,7 @@ define void @store_i64_stride6_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm6[1]
; AVX2-ONLY-NEXT: vbroadcastsd 296(%r8), %ymm6
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm6[4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm6[4,5,6,7]
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm4, %ymm0, %ymm4
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm4[6,7]
; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -6017,7 +5989,7 @@ define void @store_i64_stride6_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm4 = xmm4[1],xmm7[1]
; AVX2-ONLY-NEXT: vbroadcastsd 328(%r8), %ymm7
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm7[4,5],ymm4[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm7[4,5,6,7]
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm6, %ymm0, %ymm6
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5],ymm6[6,7]
; AVX2-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -6038,7 +6010,7 @@ define void @store_i64_stride6_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm6 = xmm6[1],xmm8[1]
; AVX2-ONLY-NEXT: vbroadcastsd 360(%r8), %ymm8
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm8[4,5],ymm6[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm8[4,5,6,7]
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm7, %ymm0, %ymm7
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5],ymm7[6,7]
; AVX2-ONLY-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -6059,7 +6031,7 @@ define void @store_i64_stride6_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm7 = xmm7[1],xmm9[1]
; AVX2-ONLY-NEXT: vbroadcastsd 392(%r8), %ymm9
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3],ymm9[4,5],ymm7[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3],ymm9[4,5,6,7]
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm8, %ymm0, %ymm8
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5],ymm8[6,7]
; AVX2-ONLY-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -6080,7 +6052,7 @@ define void @store_i64_stride6_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm7 = xmm7[1],xmm10[1]
; AVX2-ONLY-NEXT: vbroadcastsd 424(%r8), %ymm10
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3],ymm10[4,5],ymm7[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3],ymm10[4,5,6,7]
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm9, %ymm0, %ymm9
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5],ymm9[6,7]
; AVX2-ONLY-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -6101,7 +6073,7 @@ define void @store_i64_stride6_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovaps %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm9 = xmm9[1],xmm11[1]
; AVX2-ONLY-NEXT: vbroadcastsd 456(%r8), %ymm11
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm9 = ymm9[0,1,2,3],ymm11[4,5],ymm9[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm9 = ymm9[0,1,2,3],ymm11[4,5,6,7]
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm10, %ymm0, %ymm10
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm9 = ymm9[0,1,2,3,4,5],ymm10[6,7]
; AVX2-ONLY-NEXT: vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -6122,7 +6094,7 @@ define void @store_i64_stride6_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm10 = xmm10[1],xmm12[1]
; AVX2-ONLY-NEXT: vbroadcastsd 488(%r8), %ymm12
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm10 = ymm10[0,1,2,3],ymm12[4,5],ymm10[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm10 = ymm10[0,1,2,3],ymm12[4,5,6,7]
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm11, %ymm0, %ymm11
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm10 = ymm10[0,1,2,3,4,5],ymm11[6,7]
; AVX2-ONLY-NEXT: vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-store-i64-stride-7.ll b/llvm/test/CodeGen/X86/vector-interleaved-store-i64-stride-7.ll
index 4c5332d26b2f2..c6a0e42022b98 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-store-i64-stride-7.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-store-i64-stride-7.ll
@@ -218,7 +218,7 @@ define void @store_i64_stride7_vf4(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
; AVX1-ONLY-NEXT: vmovaps (%rsi), %xmm10
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm11 = xmm10[1],xmm9[1]
; AVX1-ONLY-NEXT: vbroadcastsd 8(%rcx), %ymm12
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm11 = ymm11[0,1,2,3],ymm12[4,5],ymm11[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm11 = ymm11[0,1,2,3],ymm12[4,5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm11[0,1,2,3,4,5],ymm6[6,7]
; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm4 = ymm4[0],mem[0],ymm4[2],mem[2]
; AVX1-ONLY-NEXT: vmovaps 16(%rcx), %xmm11
@@ -230,7 +230,7 @@ define void @store_i64_stride7_vf4(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2,3],ymm2[4,5,6,7]
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm3 = xmm11[1],mem[1]
; AVX1-ONLY-NEXT: vbroadcastsd 24(%r9), %ymm5
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm5[4,5],ymm3[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm5[4,5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],mem[6,7]
; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm1 = xmm1[0],xmm10[0]
; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm5 = xmm9[0],mem[0]
@@ -266,7 +266,7 @@ define void @store_i64_stride7_vf4(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
; AVX2-ONLY-NEXT: vmovaps (%rsi), %xmm10
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm11 = xmm10[1],xmm9[1]
; AVX2-ONLY-NEXT: vbroadcastsd 8(%rcx), %ymm12
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm11 = ymm11[0,1,2,3],ymm12[4,5],ymm11[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm11 = ymm11[0,1,2,3],ymm12[4,5,6,7]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm11[0,1,2,3,4,5],ymm8[6,7]
; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm1[0],ymm2[0],ymm1[2],ymm2[2]
; AVX2-ONLY-NEXT: vmovaps (%r9), %xmm2
@@ -554,7 +554,7 @@ define void @store_i64_stride7_vf8(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
; AVX1-ONLY-NEXT: vmovaps 32(%rsi), %xmm13
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm14 = xmm13[1],xmm10[1]
; AVX1-ONLY-NEXT: vbroadcastsd 40(%rcx), %ymm15
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3],ymm15[4,5],ymm14[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3],ymm15[4,5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm14[0,1,2,3,4,5],ymm2[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm7, %ymm0, %ymm15
@@ -569,7 +569,7 @@ define void @store_i64_stride7_vf8(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm0 = ymm7[0],ymm13[1],ymm7[2],ymm13[2]
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vbroadcastsd 8(%rcx), %ymm10
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3],ymm10[4,5],ymm8[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3],ymm10[4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps (%rsi), %xmm10
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm13 = xmm10[2,3,2,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm13[0,1],ymm8[2,3,4,5,6,7]
@@ -614,7 +614,7 @@ define void @store_i64_stride7_vf8(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm5[0,1,2,3],ymm4[4,5,6,7]
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],mem[1]
; AVX1-ONLY-NEXT: vbroadcastsd 56(%r9), %ymm5
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm5[2],ymm0[3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm5[2,3]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3]
; AVX1-ONLY-NEXT: movq {{[0-9]+}}(%rsp), %rax
; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm1 = xmm7[0],xmm10[0]
@@ -668,7 +668,7 @@ define void @store_i64_stride7_vf8(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
; AVX2-ONLY-NEXT: vmovddup {{.*#+}} xmm3 = mem[0,0]
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm3 = xmm3[1],xmm9[1]
; AVX2-ONLY-NEXT: vbroadcastsd 40(%rcx), %ymm14
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm14[4,5],ymm3[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm14[4,5,6,7]
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm15, %ymm0, %ymm14
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm14[6,7]
; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm5 = ymm5[0],mem[0],ymm5[2],mem[2]
@@ -695,7 +695,7 @@ define void @store_i64_stride7_vf8(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm10 = ymm11[0,1,2,3],ymm10[4,5,6,7]
; AVX2-ONLY-NEXT: vmovddup {{.*#+}} xmm11 = mem[0,0]
; AVX2-ONLY-NEXT: vbroadcastsd 8(%rcx), %ymm13
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm13 = ymm7[0,1,2,3],ymm13[4,5],ymm7[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm13 = ymm7[0,1,2,3],ymm13[4,5,6,7]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm11 = ymm11[0,1],ymm13[2,3,4,5,6,7]
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm15, %ymm0, %ymm13
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm11 = ymm11[0,1,2,3,4,5],ymm13[6,7]
@@ -705,7 +705,7 @@ define void @store_i64_stride7_vf8(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
; AVX2-ONLY-NEXT: vbroadcastsd 56(%rcx), %ymm13
; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm13 = xmm13[0,1],mem[2,3]
; AVX2-ONLY-NEXT: vbroadcastsd 56(%r9), %ymm0
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm13[0,1,2,3],ymm0[4,5],ymm13[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm13[0,1,2,3],ymm0[4,5,6,7]
; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} xmm13 = xmm14[0],mem[0]
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm9, %ymm13, %ymm9
; AVX2-ONLY-NEXT: vbroadcastsd 32(%rcx), %ymm13
@@ -823,7 +823,7 @@ define void @store_i64_stride7_vf8(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
; AVX512F-ONLY-SLOW-NEXT: # zmm13 = mem[0,1,2,3,0,1,2,3]
; AVX512F-ONLY-SLOW-NEXT: vpermi2q %zmm0, %zmm1, %zmm13
; AVX512F-ONLY-SLOW-NEXT: vpbroadcastq 8(%rcx), %ymm14
-; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm14 = mem[0,1,2,3],ymm14[4,5],mem[6,7]
+; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm14 = mem[0,1,2,3],ymm14[4,5,6,7]
; AVX512F-ONLY-SLOW-NEXT: movb $6, %cl
; AVX512F-ONLY-SLOW-NEXT: kmovw %ecx, %k2
; AVX512F-ONLY-SLOW-NEXT: vinserti64x4 $0, %ymm14, %zmm0, %zmm13 {%k2}
@@ -957,7 +957,7 @@ define void @store_i64_stride7_vf8(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
; AVX512F-ONLY-FAST-NEXT: # zmm9 = mem[0,1,2,3,0,1,2,3]
; AVX512F-ONLY-FAST-NEXT: vpermi2q %zmm0, %zmm1, %zmm9
; AVX512F-ONLY-FAST-NEXT: vpbroadcastq 8(%rcx), %ymm12
-; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm12 = mem[0,1,2,3],ymm12[4,5],mem[6,7]
+; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm12 = mem[0,1,2,3],ymm12[4,5,6,7]
; AVX512F-ONLY-FAST-NEXT: movb $6, %cl
; AVX512F-ONLY-FAST-NEXT: kmovw %ecx, %k2
; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $0, %ymm12, %zmm0, %zmm9 {%k2}
@@ -1054,7 +1054,7 @@ define void @store_i64_stride7_vf8(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
; AVX512DQ-SLOW-NEXT: # zmm8 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-SLOW-NEXT: vpermi2q %zmm0, %zmm1, %zmm8
; AVX512DQ-SLOW-NEXT: vpbroadcastq 8(%rcx), %ymm11
-; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm11 = mem[0,1,2,3],ymm11[4,5],mem[6,7]
+; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm11 = mem[0,1,2,3],ymm11[4,5,6,7]
; AVX512DQ-SLOW-NEXT: movb $6, %sil
; AVX512DQ-SLOW-NEXT: kmovw %esi, %k2
; AVX512DQ-SLOW-NEXT: vinserti64x4 $0, %ymm11, %zmm0, %zmm8 {%k2}
@@ -1187,7 +1187,7 @@ define void @store_i64_stride7_vf8(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
; AVX512DQ-FAST-NEXT: # zmm8 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-FAST-NEXT: vpermi2q %zmm0, %zmm1, %zmm8
; AVX512DQ-FAST-NEXT: vpbroadcastq 8(%rcx), %ymm10
-; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm10 = mem[0,1,2,3],ymm10[4,5],mem[6,7]
+; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm10 = mem[0,1,2,3],ymm10[4,5,6,7]
; AVX512DQ-FAST-NEXT: movb $6, %sil
; AVX512DQ-FAST-NEXT: kmovw %esi, %k2
; AVX512DQ-FAST-NEXT: vinserti64x4 $0, %ymm10, %zmm0, %zmm8 {%k2}
@@ -1309,7 +1309,7 @@ define void @store_i64_stride7_vf8(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
; AVX512BW-ONLY-SLOW-NEXT: # zmm13 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-ONLY-SLOW-NEXT: vpermi2q %zmm0, %zmm1, %zmm13
; AVX512BW-ONLY-SLOW-NEXT: vpbroadcastq 8(%rcx), %ymm14
-; AVX512BW-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm14 = mem[0,1,2,3],ymm14[4,5],mem[6,7]
+; AVX512BW-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm14 = mem[0,1,2,3],ymm14[4,5,6,7]
; AVX512BW-ONLY-SLOW-NEXT: movb $6, %cl
; AVX512BW-ONLY-SLOW-NEXT: kmovd %ecx, %k2
; AVX512BW-ONLY-SLOW-NEXT: vinserti64x4 $0, %ymm14, %zmm0, %zmm13 {%k2}
@@ -1443,7 +1443,7 @@ define void @store_i64_stride7_vf8(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
; AVX512BW-ONLY-FAST-NEXT: # zmm9 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-ONLY-FAST-NEXT: vpermi2q %zmm0, %zmm1, %zmm9
; AVX512BW-ONLY-FAST-NEXT: vpbroadcastq 8(%rcx), %ymm12
-; AVX512BW-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm12 = mem[0,1,2,3],ymm12[4,5],mem[6,7]
+; AVX512BW-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm12 = mem[0,1,2,3],ymm12[4,5,6,7]
; AVX512BW-ONLY-FAST-NEXT: movb $6, %cl
; AVX512BW-ONLY-FAST-NEXT: kmovd %ecx, %k2
; AVX512BW-ONLY-FAST-NEXT: vinserti64x4 $0, %ymm12, %zmm0, %zmm9 {%k2}
@@ -1540,7 +1540,7 @@ define void @store_i64_stride7_vf8(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
; AVX512DQBW-SLOW-NEXT: # zmm8 = mem[0,1,2,3,0,1,2,3]
; AVX512DQBW-SLOW-NEXT: vpermi2q %zmm0, %zmm1, %zmm8
; AVX512DQBW-SLOW-NEXT: vpbroadcastq 8(%rcx), %ymm11
-; AVX512DQBW-SLOW-NEXT: vpblendd {{.*#+}} ymm11 = mem[0,1,2,3],ymm11[4,5],mem[6,7]
+; AVX512DQBW-SLOW-NEXT: vpblendd {{.*#+}} ymm11 = mem[0,1,2,3],ymm11[4,5,6,7]
; AVX512DQBW-SLOW-NEXT: movb $6, %sil
; AVX512DQBW-SLOW-NEXT: kmovd %esi, %k2
; AVX512DQBW-SLOW-NEXT: vinserti64x4 $0, %ymm11, %zmm0, %zmm8 {%k2}
@@ -1673,7 +1673,7 @@ define void @store_i64_stride7_vf8(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
; AVX512DQBW-FAST-NEXT: # zmm8 = mem[0,1,2,3,0,1,2,3]
; AVX512DQBW-FAST-NEXT: vpermi2q %zmm0, %zmm1, %zmm8
; AVX512DQBW-FAST-NEXT: vpbroadcastq 8(%rcx), %ymm10
-; AVX512DQBW-FAST-NEXT: vpblendd {{.*#+}} ymm10 = mem[0,1,2,3],ymm10[4,5],mem[6,7]
+; AVX512DQBW-FAST-NEXT: vpblendd {{.*#+}} ymm10 = mem[0,1,2,3],ymm10[4,5,6,7]
; AVX512DQBW-FAST-NEXT: movb $6, %sil
; AVX512DQBW-FAST-NEXT: kmovd %esi, %k2
; AVX512DQBW-FAST-NEXT: vinserti64x4 $0, %ymm10, %zmm0, %zmm8 {%k2}
@@ -2004,7 +2004,7 @@ define void @store_i64_stride7_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
;
; AVX1-ONLY-LABEL: store_i64_stride7_vf16:
; AVX1-ONLY: # %bb.0:
-; AVX1-ONLY-NEXT: subq $552, %rsp # imm = 0x228
+; AVX1-ONLY-NEXT: subq $520, %rsp # imm = 0x208
; AVX1-ONLY-NEXT: movq {{[0-9]+}}(%rsp), %rax
; AVX1-ONLY-NEXT: vmovaps (%rsi), %ymm9
; AVX1-ONLY-NEXT: vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -2028,7 +2028,7 @@ define void @store_i64_stride7_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1],ymm6[2,3],ymm5[4,5],ymm6[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vbroadcastsd 8(%rcx), %ymm5
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm0[0,1,2,3],ymm5[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm0[0,1,2,3],ymm5[4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps (%rsi), %xmm6
; AVX1-ONLY-NEXT: vmovaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm6 = xmm6[2,3,2,3]
@@ -2040,8 +2040,9 @@ define void @store_i64_stride7_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3,4,5],ymm0[6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm4[0,1],ymm0[2,3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 16(%rcx), %xmm14
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm0 = xmm14[2,3,2,3]
+; AVX1-ONLY-NEXT: vmovaps 16(%rcx), %xmm0
+; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[2,3,2,3]
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm1 = ymm10[1],ymm1[1],ymm10[3],ymm1[3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],mem[6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4,5,6,7]
@@ -2055,171 +2056,168 @@ define void @store_i64_stride7_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 32(%r8), %xmm1
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm2
-; AVX1-ONLY-NEXT: vmovaps 32(%rax), %xmm5
-; AVX1-ONLY-NEXT: vmovaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 32(%r9), %xmm13
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm3 = xmm1[0],xmm13[0]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm5, %ymm3, %ymm3
+; AVX1-ONLY-NEXT: vmovaps 32(%rax), %xmm12
+; AVX1-ONLY-NEXT: vmovaps 32(%r9), %xmm11
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm3 = xmm1[0],xmm11[0]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm12, %ymm3, %ymm3
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm2[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm4[1]
; AVX1-ONLY-NEXT: vbroadcastsd 40(%rcx), %ymm2
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5,6,7]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 32(%rsi), %ymm11
+; AVX1-ONLY-NEXT: vmovaps 32(%rsi), %ymm10
; AVX1-ONLY-NEXT: vmovaps 48(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm11[1],ymm0[1],ymm11[3],ymm0[3]
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm10[1],ymm0[1],ymm10[3],ymm0[3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],mem[6,7]
; AVX1-ONLY-NEXT: vmovaps 48(%rax), %xmm1
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 32(%r9), %ymm10
+; AVX1-ONLY-NEXT: vmovaps 32(%r9), %ymm9
; AVX1-ONLY-NEXT: vmovaps 48(%r8), %xmm0
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm10[1],ymm0[1],ymm10[3],ymm0[3]
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm9[1],ymm0[1],ymm9[3],ymm0[3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],mem[6,7]
; AVX1-ONLY-NEXT: vmovaps 48(%rcx), %xmm8
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm1 = xmm8[2,3,2,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 64(%r9), %xmm9
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm0 = xmm9[0,1,0,1]
+; AVX1-ONLY-NEXT: vmovaps 64(%r9), %xmm7
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm0 = xmm7[0,1,0,1]
; AVX1-ONLY-NEXT: vmovaps 64(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
; AVX1-ONLY-NEXT: vmovaps 64(%r8), %xmm1
-; AVX1-ONLY-NEXT: vmovaps 64(%rax), %xmm7
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm7, %ymm1, %ymm2
+; AVX1-ONLY-NEXT: vmovaps 64(%rax), %xmm6
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm6, %ymm1, %ymm2
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3],ymm2[4,5],ymm0[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 64(%rdx), %ymm0
; AVX1-ONLY-NEXT: vbroadcastsd 72(%rcx), %ymm2
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm0[0,1,2,3],ymm2[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vmovaps 64(%rsi), %xmm12
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm3 = xmm12[2,3,2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm0[0,1,2,3],ymm2[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps 64(%rsi), %xmm15
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm13 = xmm15[2,3,2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm13[0,1],ymm2[2,3,4,5,6,7]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 64(%rsi), %ymm6
+; AVX1-ONLY-NEXT: vmovaps 64(%rsi), %ymm5
; AVX1-ONLY-NEXT: vmovaps 80(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm1 = ymm6[1],ymm1[1],ymm6[3],ymm1[3]
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm1 = ymm5[1],ymm1[1],ymm5[3],ymm1[3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX1-ONLY-NEXT: vmovaps 80(%rax), %xmm1
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 64(%r9), %ymm5
+; AVX1-ONLY-NEXT: vmovaps 64(%r9), %ymm4
; AVX1-ONLY-NEXT: vmovaps 80(%r8), %xmm0
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm5[1],ymm0[1],ymm5[3],ymm0[3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],mem[6,7]
-; AVX1-ONLY-NEXT: vmovaps 80(%rcx), %xmm4
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm15 = xmm4[2,3,2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm15[0,1],ymm0[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm4[1],ymm0[1],ymm4[3],ymm0[3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm13 = ymm0[0,1,2,3,4,5],mem[6,7]
+; AVX1-ONLY-NEXT: vmovaps 80(%rcx), %xmm3
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm14 = xmm3[2,3,2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm14[0,1],ymm13[2,3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 96(%rsi), %xmm2
-; AVX1-ONLY-NEXT: vmovaps 96(%rdi), %xmm15
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm0 = xmm15[0],xmm2[0]
-; AVX1-ONLY-NEXT: vinsertf128 $1, 96(%rcx), %ymm0, %ymm3
+; AVX1-ONLY-NEXT: vmovaps 96(%rsi), %xmm13
+; AVX1-ONLY-NEXT: vmovaps 96(%rdi), %xmm14
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm0 = xmm14[0],xmm13[0]
+; AVX1-ONLY-NEXT: vinsertf128 $1, 96(%rcx), %ymm0, %ymm2
; AVX1-ONLY-NEXT: vmovaps 96(%rdx), %xmm1
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[0],ymm3[1],ymm0[2],ymm3[2]
-; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm15, %ymm0, %ymm0
-; AVX1-ONLY-NEXT: vmovaps 96(%r8), %xmm3
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm15 = xmm3[0],mem[0]
-; AVX1-ONLY-NEXT: vinsertf128 $1, 96(%rax), %ymm15, %ymm15
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm15[0,1,2,3,4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, (%rsp) # 32-byte Spill
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm2[1],xmm1[1]
-; AVX1-ONLY-NEXT: vbroadcastsd 104(%rcx), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[0],ymm2[1],ymm0[2],ymm2[2]
+; AVX1-ONLY-NEXT: vmovupd %ymm0, (%rsp) # 32-byte Spill
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm14, %ymm0, %ymm0
+; AVX1-ONLY-NEXT: vmovaps 96(%r8), %xmm2
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm14 = xmm2[0],mem[0]
+; AVX1-ONLY-NEXT: vinsertf128 $1, 96(%rax), %ymm14, %ymm14
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm14[0,1,2,3,4,5],ymm0[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovapd 112(%rcx), %xmm0
-; AVX1-ONLY-NEXT: vmovapd 112(%rdx), %xmm1
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm1[0,1],mem[2],ymm1[3]
-; AVX1-ONLY-NEXT: vbroadcastsd 112(%r9), %ymm2
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm15 = ymm1[0,1,2],ymm2[3]
-; AVX1-ONLY-NEXT: vmovaps (%rdi), %ymm1
-; AVX1-ONLY-NEXT: vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm1 = ymm1[0],mem[0],ymm1[2],mem[2]
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX1-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm2 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm2 = xmm2[1],mem[1]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovaps (%r8), %ymm2
-; AVX1-ONLY-NEXT: vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm2 = ymm2[0],mem[0],ymm2[2],mem[2]
-; AVX1-ONLY-NEXT: vmovaps 16(%rdx), %xmm3
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm3 = xmm3[0],xmm14[0]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm14 = ymm3[0,1,2,3],ymm2[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovaps 32(%rdi), %ymm2
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm2 = ymm2[0],ymm11[0],ymm2[2],ymm11[2]
-; AVX1-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm13, %xmm3 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm3 = xmm13[1],mem[1]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2,3],ymm2[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovaps 32(%r8), %ymm3
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm3 = ymm3[0],ymm10[0],ymm3[2],ymm10[2]
-; AVX1-ONLY-NEXT: vmovaps 48(%rdx), %xmm10
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm8 = xmm10[0],xmm8[0]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3],ymm3[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovaps 64(%rdi), %ymm3
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm3 = ymm3[0],ymm6[0],ymm3[2],ymm6[2]
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm6 = xmm9[1],xmm7[1]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm6[0,1,2,3],ymm3[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovaps 64(%r8), %ymm6
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm5 = ymm6[0],ymm5[0],ymm6[2],ymm5[2]
-; AVX1-ONLY-NEXT: vmovaps 80(%rdx), %xmm6
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm4 = xmm6[0],xmm4[0]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm5[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovapd 96(%rdi), %ymm5
-; AVX1-ONLY-NEXT: vmovapd 96(%rsi), %ymm6
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm5 = ymm5[0],ymm6[0],ymm5[2],ymm6[2]
-; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm7 = mem[0,0]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm5 = ymm7[0],ymm5[1,2,3]
-; AVX1-ONLY-NEXT: vmovapd 112(%rdi), %xmm7
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm7 = ymm7[0,1],mem[2,3]
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm6 = ymm6[0,0,3,2]
-; AVX1-ONLY-NEXT: vmovapd 96(%rax), %ymm9
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm6 = ymm9[2,3],ymm6[2,3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm6 = ymm6[0],ymm7[1],ymm6[2],ymm7[3]
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],mem[1]
-; AVX1-ONLY-NEXT: vbroadcastsd 120(%r9), %ymm7
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm7[2],ymm0[3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm5 = ymm5[0],ymm9[1],ymm5[2,3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1,2],ymm9[3]
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm13[1],xmm1[1]
+; AVX1-ONLY-NEXT: vbroadcastsd 104(%rcx), %ymm1
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm1
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm13 = ymm0[0,1,2,3,4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT: vmovaps (%rdi), %ymm0
+; AVX1-ONLY-NEXT: vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm0 = ymm0[0],mem[0],ymm0[2],mem[2]
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm1 = xmm1[1],mem[1]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm14 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps (%r8), %ymm0
+; AVX1-ONLY-NEXT: vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm0 = ymm0[0],mem[0],ymm0[2],mem[2]
+; AVX1-ONLY-NEXT: vmovaps 16(%rdx), %xmm1
+; AVX1-ONLY-NEXT: vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm1 = xmm1[0],mem[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps 32(%rdi), %ymm0
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm10[0],ymm0[2],ymm10[2]
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm2 = xmm11[1],xmm12[1]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps 32(%r8), %ymm0
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm9[0],ymm0[2],ymm9[2]
+; AVX1-ONLY-NEXT: vmovaps 48(%rdx), %xmm9
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm8 = xmm9[0],xmm8[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps 64(%rdi), %ymm0
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm5[0],ymm0[2],ymm5[2]
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm5 = xmm7[1],xmm6[1]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps 64(%r8), %ymm0
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm4[0],ymm0[2],ymm4[2]
+; AVX1-ONLY-NEXT: vmovaps 80(%rdx), %xmm4
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm3 = xmm4[0],xmm3[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovapd 96(%rdi), %ymm3
+; AVX1-ONLY-NEXT: vmovapd 96(%rsi), %ymm4
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm3 = ymm3[0],ymm4[0],ymm3[2],ymm4[2]
+; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm6 = mem[0,0]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm3 = ymm6[0,1],ymm3[2,3]
+; AVX1-ONLY-NEXT: vmovapd 112(%rdi), %xmm6
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm6 = ymm6[0,1],mem[2,3]
+; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm4 = ymm4[0,0,3,2]
+; AVX1-ONLY-NEXT: vmovapd 96(%rax), %ymm7
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm4 = ymm7[2,3],ymm4[2,3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm4 = ymm4[0],ymm6[1],ymm4[2],ymm6[3]
+; AVX1-ONLY-NEXT: vmovapd 112(%rcx), %xmm6
+; AVX1-ONLY-NEXT: vmovapd 112(%rdx), %xmm9
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm9 = xmm9[0],xmm6[0]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm9 = ymm9[0,1],mem[2,3]
+; AVX1-ONLY-NEXT: vbroadcastsd 112(%r9), %ymm10
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm9 = ymm9[0,1,2],ymm10[3]
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm6 = xmm6[1],mem[1]
+; AVX1-ONLY-NEXT: vbroadcastsd 120(%r9), %ymm10
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm6 = ymm6[0,1],ymm10[2,3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm3 = ymm3[0],ymm7[1],ymm3[2,3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm6 = ymm6[0,1,2],ymm7[3]
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm7 = xmm7[0],xmm12[0]
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; AVX1-ONLY-NEXT: vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm9, %xmm9 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm9 = xmm9[0],mem[0]
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm7 = xmm7[0],xmm15[0]
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; AVX1-ONLY-NEXT: vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm10, %xmm10 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm10 = xmm10[0],mem[0]
; AVX1-ONLY-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; AVX1-ONLY-NEXT: vmovaps 64(%rdx), %xmm10
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm10 = xmm10[0],mem[0]
-; AVX1-ONLY-NEXT: vmovaps (%rdx), %xmm11
+; AVX1-ONLY-NEXT: vmovaps 64(%rdx), %xmm11
; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm11 = xmm11[0],mem[0]
-; AVX1-ONLY-NEXT: vmovaps %xmm11, 16(%rax)
-; AVX1-ONLY-NEXT: vmovaps %xmm9, (%rax)
-; AVX1-ONLY-NEXT: vmovaps %xmm10, 464(%rax)
+; AVX1-ONLY-NEXT: vmovaps (%rdx), %xmm12
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm12 = xmm12[0],mem[0]
+; AVX1-ONLY-NEXT: vmovaps %xmm12, 16(%rax)
+; AVX1-ONLY-NEXT: vmovaps %xmm10, (%rax)
+; AVX1-ONLY-NEXT: vmovaps %xmm11, 464(%rax)
; AVX1-ONLY-NEXT: vmovaps %xmm7, 448(%rax)
-; AVX1-ONLY-NEXT: vmovapd %ymm6, 832(%rax)
-; AVX1-ONLY-NEXT: vmovapd %ymm5, 768(%rax)
-; AVX1-ONLY-NEXT: vmovaps %ymm4, 576(%rax)
-; AVX1-ONLY-NEXT: vmovaps %ymm3, 544(%rax)
+; AVX1-ONLY-NEXT: vmovapd %ymm4, 832(%rax)
+; AVX1-ONLY-NEXT: vmovapd %ymm3, 768(%rax)
+; AVX1-ONLY-NEXT: vmovaps %ymm0, 576(%rax)
+; AVX1-ONLY-NEXT: vmovaps %ymm5, 544(%rax)
; AVX1-ONLY-NEXT: vmovaps %ymm8, 352(%rax)
; AVX1-ONLY-NEXT: vmovaps %ymm2, 320(%rax)
-; AVX1-ONLY-NEXT: vmovaps %ymm14, 128(%rax)
-; AVX1-ONLY-NEXT: vmovaps %ymm1, 96(%rax)
-; AVX1-ONLY-NEXT: vmovapd %ymm0, 864(%rax)
-; AVX1-ONLY-NEXT: vmovapd %ymm15, 800(%rax)
+; AVX1-ONLY-NEXT: vmovaps %ymm1, 128(%rax)
+; AVX1-ONLY-NEXT: vmovaps %ymm14, 96(%rax)
+; AVX1-ONLY-NEXT: vmovapd %ymm6, 864(%rax)
+; AVX1-ONLY-NEXT: vmovapd %ymm9, 800(%rax)
+; AVX1-ONLY-NEXT: vmovaps %ymm13, 736(%rax)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm0, 736(%rax)
-; AVX1-ONLY-NEXT: vmovups (%rsp), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm0, 704(%rax)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovups (%rsp), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm0, 672(%rax)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm0, 640(%rax)
@@ -2247,7 +2245,7 @@ define void @store_i64_stride7_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vmovaps %ymm0, 64(%rax)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm0, 32(%rax)
-; AVX1-ONLY-NEXT: addq $552, %rsp # imm = 0x228
+; AVX1-ONLY-NEXT: addq $520, %rsp # imm = 0x208
; AVX1-ONLY-NEXT: vzeroupper
; AVX1-ONLY-NEXT: retq
;
@@ -2284,7 +2282,7 @@ define void @store_i64_stride7_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovddup {{.*#+}} xmm0 = mem[0,0]
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm4[1]
; AVX2-ONLY-NEXT: vbroadcastsd 40(%rcx), %ymm4
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm4[4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm4[4,5,6,7]
; AVX2-ONLY-NEXT: vmovaps 48(%rax), %xmm4
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm2
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm2[6,7]
@@ -2314,7 +2312,7 @@ define void @store_i64_stride7_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovddup {{.*#+}} xmm4 = mem[0,0]
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm4 = xmm4[1],xmm5[1]
; AVX2-ONLY-NEXT: vbroadcastsd 104(%rcx), %ymm5
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm5[4,5],ymm4[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm5[4,5,6,7]
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm2
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm4[0,1,2,3,4,5],ymm2[6,7]
; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -2328,7 +2326,7 @@ define void @store_i64_stride7_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm4[0,1],ymm2[2,3,4,5,6,7]
; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vbroadcastsd 8(%rcx), %ymm2
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm9[0,1,2,3],ymm2[4,5],ymm9[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm9[0,1,2,3],ymm2[4,5,6,7]
; AVX2-ONLY-NEXT: vmovddup {{.*#+}} xmm4 = mem[0,0]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm4[0,1],ymm2[2,3,4,5,6,7]
; AVX2-ONLY-NEXT: vmovaps (%r8), %xmm12
@@ -2359,7 +2357,7 @@ define void @store_i64_stride7_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm1[2,3],ymm9[2,3]
; AVX2-ONLY-NEXT: vmovups %ymm1, (%rsp) # 32-byte Spill
; AVX2-ONLY-NEXT: vbroadcastsd 72(%rcx), %ymm1
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT: vmovddup {{.*#+}} xmm9 = mem[0,0]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm9[0,1],ymm1[2,3,4,5,6,7]
; AVX2-ONLY-NEXT: vmovaps 64(%r8), %xmm11
@@ -2380,11 +2378,11 @@ define void @store_i64_stride7_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm15 = ymm15[2,3],ymm13[2,3]
; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm6 = ymm6[0],ymm10[0],ymm6[2],ymm10[2]
; AVX2-ONLY-NEXT: vmovddup {{.*#+}} xmm10 = mem[0,0]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm10[0,1],ymm6[2,3,4,5,6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm10[0,1,2,3],ymm6[4,5,6,7]
; AVX2-ONLY-NEXT: vbroadcastsd 120(%rcx), %ymm10
; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm10 = xmm10[0,1],mem[2,3]
; AVX2-ONLY-NEXT: vbroadcastsd 120(%r9), %ymm13
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm13 = ymm10[0,1,2,3],ymm13[4,5],ymm10[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm13 = ymm10[0,1,2,3],ymm13[4,5,6,7]
; AVX2-ONLY-NEXT: vmovaps 96(%rax), %ymm0
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm10 = ymm6[0,1],ymm0[2,3],ymm6[4,5,6,7]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3,4,5],ymm0[6,7]
@@ -2644,7 +2642,7 @@ define void @store_i64_stride7_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX512F-ONLY-SLOW-NEXT: kmovw %esi, %k3
; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 %zmm6, %zmm10 {%k3}
; AVX512F-ONLY-SLOW-NEXT: vpbroadcastq 72(%rcx), %ymm2
-; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = mem[0,1,2,3],ymm2[4,5],mem[6,7]
+; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = mem[0,1,2,3],ymm2[4,5,6,7]
; AVX512F-ONLY-SLOW-NEXT: movb $6, %sil
; AVX512F-ONLY-SLOW-NEXT: kmovw %esi, %k4
; AVX512F-ONLY-SLOW-NEXT: vinserti64x4 $0, %ymm2, %zmm0, %zmm0 {%k4}
@@ -2664,7 +2662,7 @@ define void @store_i64_stride7_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 %zmm9, %zmm22 {%k1}
; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 %zmm22, %zmm15 {%k3}
; AVX512F-ONLY-SLOW-NEXT: vpbroadcastq 8(%rcx), %ymm3
-; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = mem[0,1,2,3],ymm3[4,5],mem[6,7]
+; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = mem[0,1,2,3],ymm3[4,5,6,7]
; AVX512F-ONLY-SLOW-NEXT: vinserti64x4 $0, %ymm3, %zmm0, %zmm28 {%k4}
; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 %zmm16, %zmm28 {%k5}
; AVX512F-ONLY-SLOW-NEXT: vpunpckhqdq {{.*#+}} ymm3 = ymm19[1],ymm21[1],ymm19[3],ymm21[3]
@@ -2849,7 +2847,7 @@ define void @store_i64_stride7_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX512F-ONLY-FAST-NEXT: vpunpcklqdq {{.*#+}} ymm5 = ymm8[0],ymm22[0],ymm8[2],ymm22[2]
; AVX512F-ONLY-FAST-NEXT: vshufi64x2 {{.*#+}} zmm1 {%k1} = zmm5[0,1,2,3],zmm13[2,3,0,1]
; AVX512F-ONLY-FAST-NEXT: vpbroadcastq 72(%rcx), %ymm5
-; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm5 = mem[0,1,2,3],ymm5[4,5],mem[6,7]
+; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm5 = mem[0,1,2,3],ymm5[4,5,6,7]
; AVX512F-ONLY-FAST-NEXT: movb $6, %al
; AVX512F-ONLY-FAST-NEXT: kmovw %eax, %k1
; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $0, %ymm5, %zmm0, %zmm11 {%k1}
@@ -2863,7 +2861,7 @@ define void @store_i64_stride7_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX512F-ONLY-FAST-NEXT: vpermt2q %zmm13, %zmm5, %zmm12
; AVX512F-ONLY-FAST-NEXT: vpermt2q %zmm10, %zmm5, %zmm8
; AVX512F-ONLY-FAST-NEXT: vpbroadcastq 8(%rcx), %ymm5
-; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm5 = mem[0,1,2,3],ymm5[4,5],mem[6,7]
+; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm5 = mem[0,1,2,3],ymm5[4,5,6,7]
; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $0, %ymm5, %zmm0, %zmm0 {%k1}
; AVX512F-ONLY-FAST-NEXT: movb $64, %al
; AVX512F-ONLY-FAST-NEXT: kmovw %eax, %k1
@@ -3060,7 +3058,7 @@ define void @store_i64_stride7_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX512DQ-SLOW-NEXT: kmovw %esi, %k5
; AVX512DQ-SLOW-NEXT: vmovdqa64 %zmm6, %zmm10 {%k5}
; AVX512DQ-SLOW-NEXT: vpbroadcastq 72(%rcx), %ymm2
-; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = mem[0,1,2,3],ymm2[4,5],mem[6,7]
+; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = mem[0,1,2,3],ymm2[4,5,6,7]
; AVX512DQ-SLOW-NEXT: movb $6, %sil
; AVX512DQ-SLOW-NEXT: kmovw %esi, %k6
; AVX512DQ-SLOW-NEXT: vinserti64x4 $0, %ymm2, %zmm0, %zmm8 {%k6}
@@ -3084,7 +3082,7 @@ define void @store_i64_stride7_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX512DQ-SLOW-NEXT: vmovdqa64 %zmm9, %zmm16 {%k1}
; AVX512DQ-SLOW-NEXT: vmovdqa64 %zmm16, %zmm11 {%k5}
; AVX512DQ-SLOW-NEXT: vpbroadcastq 8(%rcx), %ymm3
-; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = mem[0,1,2,3],ymm3[4,5],mem[6,7]
+; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = mem[0,1,2,3],ymm3[4,5,6,7]
; AVX512DQ-SLOW-NEXT: vinserti64x4 $0, %ymm3, %zmm0, %zmm23 {%k6}
; AVX512DQ-SLOW-NEXT: vmovdqa64 %zmm15, %zmm23 {%k7}
; AVX512DQ-SLOW-NEXT: vpunpckhqdq {{.*#+}} ymm3 = ymm19[1],ymm21[1],ymm19[3],ymm21[3]
@@ -3267,7 +3265,7 @@ define void @store_i64_stride7_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX512DQ-FAST-NEXT: vpunpcklqdq {{.*#+}} ymm5 = ymm7[0],ymm21[0],ymm7[2],ymm21[2]
; AVX512DQ-FAST-NEXT: vshufi64x2 {{.*#+}} zmm1 {%k1} = zmm5[0,1,2,3],zmm13[2,3,0,1]
; AVX512DQ-FAST-NEXT: vpbroadcastq 72(%rcx), %ymm5
-; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm5 = mem[0,1,2,3],ymm5[4,5],mem[6,7]
+; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm5 = mem[0,1,2,3],ymm5[4,5,6,7]
; AVX512DQ-FAST-NEXT: movb $6, %al
; AVX512DQ-FAST-NEXT: kmovw %eax, %k1
; AVX512DQ-FAST-NEXT: vinserti64x4 $0, %ymm5, %zmm0, %zmm11 {%k1}
@@ -3281,7 +3279,7 @@ define void @store_i64_stride7_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX512DQ-FAST-NEXT: vpermt2q %zmm13, %zmm5, %zmm12
; AVX512DQ-FAST-NEXT: vpermt2q %zmm10, %zmm5, %zmm7
; AVX512DQ-FAST-NEXT: vpbroadcastq 8(%rcx), %ymm5
-; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm5 = mem[0,1,2,3],ymm5[4,5],mem[6,7]
+; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm5 = mem[0,1,2,3],ymm5[4,5,6,7]
; AVX512DQ-FAST-NEXT: vinserti64x4 $0, %ymm5, %zmm0, %zmm0 {%k1}
; AVX512DQ-FAST-NEXT: movb $64, %al
; AVX512DQ-FAST-NEXT: kmovw %eax, %k1
@@ -3482,7 +3480,7 @@ define void @store_i64_stride7_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX512BW-ONLY-SLOW-NEXT: kmovd %esi, %k3
; AVX512BW-ONLY-SLOW-NEXT: vmovdqa64 %zmm6, %zmm10 {%k3}
; AVX512BW-ONLY-SLOW-NEXT: vpbroadcastq 72(%rcx), %ymm2
-; AVX512BW-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = mem[0,1,2,3],ymm2[4,5],mem[6,7]
+; AVX512BW-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = mem[0,1,2,3],ymm2[4,5,6,7]
; AVX512BW-ONLY-SLOW-NEXT: movb $6, %sil
; AVX512BW-ONLY-SLOW-NEXT: kmovd %esi, %k4
; AVX512BW-ONLY-SLOW-NEXT: vinserti64x4 $0, %ymm2, %zmm0, %zmm0 {%k4}
@@ -3502,7 +3500,7 @@ define void @store_i64_stride7_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX512BW-ONLY-SLOW-NEXT: vmovdqa64 %zmm9, %zmm22 {%k1}
; AVX512BW-ONLY-SLOW-NEXT: vmovdqa64 %zmm22, %zmm15 {%k3}
; AVX512BW-ONLY-SLOW-NEXT: vpbroadcastq 8(%rcx), %ymm3
-; AVX512BW-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = mem[0,1,2,3],ymm3[4,5],mem[6,7]
+; AVX512BW-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = mem[0,1,2,3],ymm3[4,5,6,7]
; AVX512BW-ONLY-SLOW-NEXT: vinserti64x4 $0, %ymm3, %zmm0, %zmm28 {%k4}
; AVX512BW-ONLY-SLOW-NEXT: vmovdqa64 %zmm16, %zmm28 {%k5}
; AVX512BW-ONLY-SLOW-NEXT: vpunpckhqdq {{.*#+}} ymm3 = ymm19[1],ymm21[1],ymm19[3],ymm21[3]
@@ -3687,7 +3685,7 @@ define void @store_i64_stride7_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX512BW-ONLY-FAST-NEXT: vpunpcklqdq {{.*#+}} ymm5 = ymm8[0],ymm22[0],ymm8[2],ymm22[2]
; AVX512BW-ONLY-FAST-NEXT: vshufi64x2 {{.*#+}} zmm1 {%k1} = zmm5[0,1,2,3],zmm13[2,3,0,1]
; AVX512BW-ONLY-FAST-NEXT: vpbroadcastq 72(%rcx), %ymm5
-; AVX512BW-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm5 = mem[0,1,2,3],ymm5[4,5],mem[6,7]
+; AVX512BW-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm5 = mem[0,1,2,3],ymm5[4,5,6,7]
; AVX512BW-ONLY-FAST-NEXT: movb $6, %al
; AVX512BW-ONLY-FAST-NEXT: kmovd %eax, %k1
; AVX512BW-ONLY-FAST-NEXT: vinserti64x4 $0, %ymm5, %zmm0, %zmm11 {%k1}
@@ -3701,7 +3699,7 @@ define void @store_i64_stride7_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX512BW-ONLY-FAST-NEXT: vpermt2q %zmm13, %zmm5, %zmm12
; AVX512BW-ONLY-FAST-NEXT: vpermt2q %zmm10, %zmm5, %zmm8
; AVX512BW-ONLY-FAST-NEXT: vpbroadcastq 8(%rcx), %ymm5
-; AVX512BW-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm5 = mem[0,1,2,3],ymm5[4,5],mem[6,7]
+; AVX512BW-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm5 = mem[0,1,2,3],ymm5[4,5,6,7]
; AVX512BW-ONLY-FAST-NEXT: vinserti64x4 $0, %ymm5, %zmm0, %zmm0 {%k1}
; AVX512BW-ONLY-FAST-NEXT: movb $64, %al
; AVX512BW-ONLY-FAST-NEXT: kmovd %eax, %k1
@@ -3898,7 +3896,7 @@ define void @store_i64_stride7_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX512DQBW-SLOW-NEXT: kmovd %esi, %k5
; AVX512DQBW-SLOW-NEXT: vmovdqa64 %zmm6, %zmm10 {%k5}
; AVX512DQBW-SLOW-NEXT: vpbroadcastq 72(%rcx), %ymm2
-; AVX512DQBW-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = mem[0,1,2,3],ymm2[4,5],mem[6,7]
+; AVX512DQBW-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = mem[0,1,2,3],ymm2[4,5,6,7]
; AVX512DQBW-SLOW-NEXT: movb $6, %sil
; AVX512DQBW-SLOW-NEXT: kmovd %esi, %k6
; AVX512DQBW-SLOW-NEXT: vinserti64x4 $0, %ymm2, %zmm0, %zmm8 {%k6}
@@ -3922,7 +3920,7 @@ define void @store_i64_stride7_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX512DQBW-SLOW-NEXT: vmovdqa64 %zmm9, %zmm16 {%k1}
; AVX512DQBW-SLOW-NEXT: vmovdqa64 %zmm16, %zmm11 {%k5}
; AVX512DQBW-SLOW-NEXT: vpbroadcastq 8(%rcx), %ymm3
-; AVX512DQBW-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = mem[0,1,2,3],ymm3[4,5],mem[6,7]
+; AVX512DQBW-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = mem[0,1,2,3],ymm3[4,5,6,7]
; AVX512DQBW-SLOW-NEXT: vinserti64x4 $0, %ymm3, %zmm0, %zmm23 {%k6}
; AVX512DQBW-SLOW-NEXT: vmovdqa64 %zmm15, %zmm23 {%k7}
; AVX512DQBW-SLOW-NEXT: vpunpckhqdq {{.*#+}} ymm3 = ymm19[1],ymm21[1],ymm19[3],ymm21[3]
@@ -4105,7 +4103,7 @@ define void @store_i64_stride7_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX512DQBW-FAST-NEXT: vpunpcklqdq {{.*#+}} ymm5 = ymm7[0],ymm21[0],ymm7[2],ymm21[2]
; AVX512DQBW-FAST-NEXT: vshufi64x2 {{.*#+}} zmm1 {%k1} = zmm5[0,1,2,3],zmm13[2,3,0,1]
; AVX512DQBW-FAST-NEXT: vpbroadcastq 72(%rcx), %ymm5
-; AVX512DQBW-FAST-NEXT: vpblendd {{.*#+}} ymm5 = mem[0,1,2,3],ymm5[4,5],mem[6,7]
+; AVX512DQBW-FAST-NEXT: vpblendd {{.*#+}} ymm5 = mem[0,1,2,3],ymm5[4,5,6,7]
; AVX512DQBW-FAST-NEXT: movb $6, %al
; AVX512DQBW-FAST-NEXT: kmovd %eax, %k1
; AVX512DQBW-FAST-NEXT: vinserti64x4 $0, %ymm5, %zmm0, %zmm11 {%k1}
@@ -4119,7 +4117,7 @@ define void @store_i64_stride7_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX512DQBW-FAST-NEXT: vpermt2q %zmm13, %zmm5, %zmm12
; AVX512DQBW-FAST-NEXT: vpermt2q %zmm10, %zmm5, %zmm7
; AVX512DQBW-FAST-NEXT: vpbroadcastq 8(%rcx), %ymm5
-; AVX512DQBW-FAST-NEXT: vpblendd {{.*#+}} ymm5 = mem[0,1,2,3],ymm5[4,5],mem[6,7]
+; AVX512DQBW-FAST-NEXT: vpblendd {{.*#+}} ymm5 = mem[0,1,2,3],ymm5[4,5,6,7]
; AVX512DQBW-FAST-NEXT: vinserti64x4 $0, %ymm5, %zmm0, %zmm0 {%k1}
; AVX512DQBW-FAST-NEXT: movb $64, %al
; AVX512DQBW-FAST-NEXT: kmovd %eax, %k1
@@ -4755,7 +4753,7 @@ define void @store_i64_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1],ymm12[2,3],ymm5[4,5],ymm12[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vbroadcastsd 8(%rcx), %ymm5
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm2[0,1,2,3],ymm5[4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm2[0,1,2,3],ymm5[4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps (%rsi), %xmm12
; AVX1-ONLY-NEXT: vmovaps %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm12 = xmm12[2,3,2,3]
@@ -4801,7 +4799,7 @@ define void @store_i64_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vmovaps 32(%rdi), %ymm2
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm6[1]
; AVX1-ONLY-NEXT: vbroadcastsd 40(%rcx), %ymm4
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm4[4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm4[4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps 32(%rsi), %ymm4
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
@@ -4842,7 +4840,7 @@ define void @store_i64_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 64(%rdx), %ymm1
; AVX1-ONLY-NEXT: vbroadcastsd 72(%rcx), %ymm4
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm1[0,1,2,3],ymm4[4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm1[0,1,2,3],ymm4[4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps 64(%rsi), %xmm11
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm5 = xmm11[2,3,2,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm5[0,1],ymm4[2,3,4,5,6,7]
@@ -4893,7 +4891,7 @@ define void @store_i64_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm4[1]
; AVX1-ONLY-NEXT: vbroadcastsd 104(%rcx), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm1
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -4934,7 +4932,7 @@ define void @store_i64_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 128(%rdx), %ymm1
; AVX1-ONLY-NEXT: vbroadcastsd 136(%rcx), %ymm4
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm1[0,1,2,3],ymm4[4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm1[0,1,2,3],ymm4[4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps 128(%rsi), %xmm10
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm5 = xmm10[2,3,2,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm5[0,1],ymm4[2,3,4,5,6,7]
@@ -4985,7 +4983,7 @@ define void @store_i64_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm4[1]
; AVX1-ONLY-NEXT: vbroadcastsd 168(%rcx), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm1
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -5026,7 +5024,7 @@ define void @store_i64_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 192(%rdx), %ymm1
; AVX1-ONLY-NEXT: vbroadcastsd 200(%rcx), %ymm2
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps 192(%rsi), %xmm8
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm7 = xmm8[2,3,2,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm7[0,1],ymm6[2,3,4,5,6,7]
@@ -5041,9 +5039,9 @@ define void @store_i64_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 192(%r8), %ymm0
; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm4 = ymm0[0],mem[0],ymm0[2],mem[2]
-; AVX1-ONLY-NEXT: vmovaps 208(%rcx), %xmm7
+; AVX1-ONLY-NEXT: vmovaps 208(%rcx), %xmm3
; AVX1-ONLY-NEXT: vmovaps 208(%rdx), %xmm6
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm6 = xmm6[0],xmm7[0]
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm6 = xmm6[0],xmm3[0]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm6[0,1,2,3],ymm4[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 208(%rdi), %xmm4
@@ -5058,50 +5056,51 @@ define void @store_i64_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vinsertf128 $1, 224(%rcx), %ymm5, %ymm6
; AVX1-ONLY-NEXT: vmovaps 224(%rdx), %xmm0
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm5, %ymm5
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm6 = ymm5[0],ymm6[1],ymm5[2],ymm6[2]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm7 = ymm5[0],ymm6[1],ymm5[2],ymm6[2]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm4, %ymm0, %ymm4
; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm5 = mem[0,0]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm4, %ymm5, %ymm4
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm5[2,3],ymm4[4,5,6,7]
; AVX1-ONLY-NEXT: vmovapd 224(%r8), %ymm2
; AVX1-ONLY-NEXT: vinsertf128 $1, 224(%rax), %ymm2, %ymm5
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1],ymm4[2,3],ymm5[4,5],ymm4[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm5[0,1],ymm4[2,3],ymm5[4,5],ymm4[6,7]
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm1 = xmm1[2,3,2,3]
; AVX1-ONLY-NEXT: vbroadcastsd 232(%rcx), %ymm4
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm4[4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm4[4,5,6,7]
; AVX1-ONLY-NEXT: vinsertf128 $1, 224(%r8), %ymm0, %ymm0
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
; AVX1-ONLY-NEXT: vmovapd 224(%rdi), %ymm0
; AVX1-ONLY-NEXT: vmovapd 224(%rsi), %ymm1
; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm15 = mem[0,0]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm15 = ymm15[0],ymm0[1,2,3]
-; AVX1-ONLY-NEXT: vmovapd 240(%rcx), %xmm3
-; AVX1-ONLY-NEXT: vmovapd 240(%rdx), %xmm14
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm14 = ymm14[0],ymm3[0],ymm14[2],ymm3[2]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm14[0,1],ymm2[2],ymm14[3]
-; AVX1-ONLY-NEXT: vbroadcastsd 240(%r9), %ymm14
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm2[0,1,2],ymm14[3]
-; AVX1-ONLY-NEXT: vmovapd 240(%rdi), %xmm14
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm14 = ymm14[0,1],mem[2,3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm4 = ymm15[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT: vmovapd 240(%rdi), %xmm15
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm15 = ymm15[0,1],mem[2,3]
; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm1 = ymm1[0,0,3,2]
; AVX1-ONLY-NEXT: vmovapd 224(%rax), %ymm0
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm0[2,3],ymm1[2,3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm1[0],ymm14[1],ymm1[2],ymm14[3]
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm7 = xmm7[1],mem[1]
-; AVX1-ONLY-NEXT: vbroadcastsd 216(%r9), %ymm14
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3],ymm14[4,5],ymm7[6,7]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm1[0],ymm15[1],ymm1[2],ymm15[3]
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm3 = xmm3[1],mem[1]
-; AVX1-ONLY-NEXT: vbroadcastsd 248(%r9), %ymm14
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm3 = ymm3[0,1],ymm14[2],ymm3[3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm14 = ymm15[0],ymm0[1],ymm15[2,3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm3[0,1,2],ymm0[3]
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm3 = xmm12[0],xmm10[0]
+; AVX1-ONLY-NEXT: vbroadcastsd 216(%r9), %ymm15
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm15[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovapd 240(%rcx), %xmm15
+; AVX1-ONLY-NEXT: vmovapd 240(%rdx), %xmm14
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm14 = xmm14[0],xmm15[0]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm14[0,1],ymm2[2,3]
+; AVX1-ONLY-NEXT: vbroadcastsd 240(%r9), %ymm14
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm2[0,1,2],ymm14[3]
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm14 = xmm15[1],mem[1]
+; AVX1-ONLY-NEXT: vbroadcastsd 248(%r9), %ymm15
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm14 = ymm14[0,1],ymm15[2,3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm4 = ymm4[0],ymm0[1],ymm4[2,3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm14[0,1,2],ymm0[3]
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm14 = xmm12[0],xmm10[0]
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm15 = xmm10[0],xmm11[0]
; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm8 = xmm9[0],xmm8[0]
; AVX1-ONLY-NEXT: vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm13, %xmm9 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm9 = xmm13[0],mem[0]
; AVX1-ONLY-NEXT: movq {{[0-9]+}}(%rsp), %rsi
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5],mem[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],mem[6,7]
; AVX1-ONLY-NEXT: vmovaps 128(%rdx), %xmm13
; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm13 = xmm13[0],mem[0]
; AVX1-ONLY-NEXT: vmovaps 64(%rdx), %xmm12
@@ -5117,15 +5116,15 @@ define void @store_i64_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vmovaps %xmm12, 464(%rsi)
; AVX1-ONLY-NEXT: vmovaps %xmm15, 448(%rsi)
; AVX1-ONLY-NEXT: vmovaps %xmm13, 912(%rsi)
-; AVX1-ONLY-NEXT: vmovaps %xmm3, 896(%rsi)
+; AVX1-ONLY-NEXT: vmovaps %xmm14, 896(%rsi)
; AVX1-ONLY-NEXT: vmovapd %ymm0, 1760(%rsi)
; AVX1-ONLY-NEXT: vmovapd %ymm1, 1728(%rsi)
; AVX1-ONLY-NEXT: vmovapd %ymm2, 1696(%rsi)
-; AVX1-ONLY-NEXT: vmovapd %ymm14, 1664(%rsi)
-; AVX1-ONLY-NEXT: vmovaps %ymm4, 1632(%rsi)
-; AVX1-ONLY-NEXT: vmovaps %ymm5, 1600(%rsi)
-; AVX1-ONLY-NEXT: vmovapd %ymm6, 1568(%rsi)
-; AVX1-ONLY-NEXT: vmovaps %ymm7, 1536(%rsi)
+; AVX1-ONLY-NEXT: vmovapd %ymm4, 1664(%rsi)
+; AVX1-ONLY-NEXT: vmovaps %ymm5, 1632(%rsi)
+; AVX1-ONLY-NEXT: vmovaps %ymm6, 1600(%rsi)
+; AVX1-ONLY-NEXT: vmovapd %ymm7, 1568(%rsi)
+; AVX1-ONLY-NEXT: vmovaps %ymm3, 1536(%rsi)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm0, 1504(%rsi)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
@@ -5231,7 +5230,7 @@ define void @store_i64_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovddup {{.*#+}} xmm4 = mem[0,0]
; AVX2-ONLY-NEXT: vbroadcastsd 8(%rcx), %ymm5
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm3[0,1,2,3],ymm5[4,5],ymm3[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm3[0,1,2,3],ymm5[4,5,6,7]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm4[0,1],ymm5[2,3,4,5,6,7]
; AVX2-ONLY-NEXT: vmovaps (%r8), %xmm6
; AVX2-ONLY-NEXT: vmovaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -5269,7 +5268,7 @@ define void @store_i64_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovddup {{.*#+}} xmm1 = mem[0,0]
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm2[1]
; AVX2-ONLY-NEXT: vbroadcastsd 40(%rcx), %ymm2
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX2-ONLY-NEXT: vmovaps 32(%rsi), %ymm2
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm4, %ymm0, %ymm3
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm3[6,7]
@@ -5296,7 +5295,7 @@ define void @store_i64_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovaps 64(%rdx), %ymm0
; AVX2-ONLY-NEXT: vbroadcastsd 72(%rcx), %ymm1
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT: vmovddup {{.*#+}} xmm2 = mem[0,0]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5,6,7]
; AVX2-ONLY-NEXT: vmovaps 64(%r8), %xmm2
@@ -5341,7 +5340,7 @@ define void @store_i64_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovddup {{.*#+}} xmm1 = mem[0,0]
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm3[1]
; AVX2-ONLY-NEXT: vbroadcastsd 104(%rcx), %ymm3
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5,6,7]
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -5369,7 +5368,7 @@ define void @store_i64_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovaps 128(%rdx), %ymm0
; AVX2-ONLY-NEXT: vbroadcastsd 136(%rcx), %ymm1
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT: vmovddup {{.*#+}} xmm2 = mem[0,0]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5,6,7]
; AVX2-ONLY-NEXT: vmovaps 128(%r8), %xmm14
@@ -5409,7 +5408,7 @@ define void @store_i64_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovddup {{.*#+}} xmm1 = mem[0,0]
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm9[1]
; AVX2-ONLY-NEXT: vbroadcastsd 168(%rcx), %ymm3
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5,6,7]
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -5435,7 +5434,7 @@ define void @store_i64_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovaps 192(%rdx), %ymm0
; AVX2-ONLY-NEXT: vbroadcastsd 200(%rcx), %ymm1
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT: vmovddup {{.*#+}} xmm2 = mem[0,0]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5,6,7]
; AVX2-ONLY-NEXT: vmovaps 192(%r8), %xmm4
@@ -5473,14 +5472,14 @@ define void @store_i64_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovddup {{.*#+}} xmm1 = mem[0,0]
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm2[1]
; AVX2-ONLY-NEXT: vbroadcastsd 232(%rcx), %ymm3
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5,6,7]
; AVX2-ONLY-NEXT: vinsertf128 $1, 224(%r8), %ymm0, %ymm3
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm3[6,7]
; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovaps 224(%rsi), %ymm1
; AVX2-ONLY-NEXT: vmovddup {{.*#+}} xmm3 = mem[0,0]
; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm15 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1],ymm15[2,3,4,5,6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm15[4,5,6,7]
; AVX2-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,2,3,3]
@@ -5575,7 +5574,7 @@ define void @store_i64_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vbroadcastsd 216(%rcx), %ymm5
; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm5 = xmm5[0,1],mem[2,3]
; AVX2-ONLY-NEXT: vbroadcastsd 216(%r9), %ymm10
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm10[4,5],ymm5[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm10[4,5,6,7]
; AVX2-ONLY-NEXT: vmovaps 224(%rdi), %xmm10
; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} xmm10 = xmm10[0],mem[0]
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm10, %ymm2
@@ -5588,7 +5587,7 @@ define void @store_i64_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vbroadcastsd 248(%rcx), %ymm6
; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm6 = xmm6[0,1],mem[2,3]
; AVX2-ONLY-NEXT: vbroadcastsd 248(%r9), %ymm10
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm10[4,5],ymm6[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm10[4,5,6,7]
; AVX2-ONLY-NEXT: vmovaps 224(%rax), %ymm10
; AVX2-ONLY-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm12 # 32-byte Folded Reload
; AVX2-ONLY-NEXT: # ymm12 = mem[0,1],ymm10[2,3],mem[4,5,6,7]
@@ -6037,18 +6036,18 @@ define void @store_i64_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX512F-ONLY-SLOW-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 %zmm0, %zmm9 {%k4}
; AVX512F-ONLY-SLOW-NEXT: vpbroadcastq 8(%rcx), %ymm4
-; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = mem[0,1,2,3],ymm4[4,5],mem[6,7]
+; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = mem[0,1,2,3],ymm4[4,5,6,7]
; AVX512F-ONLY-SLOW-NEXT: movb $6, %sil
; AVX512F-ONLY-SLOW-NEXT: kmovw %esi, %k4
; AVX512F-ONLY-SLOW-NEXT: vinserti64x4 $0, %ymm4, %zmm0, %zmm31 {%k4}
; AVX512F-ONLY-SLOW-NEXT: vpbroadcastq 72(%rcx), %ymm4
-; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = mem[0,1,2,3],ymm4[4,5],mem[6,7]
+; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = mem[0,1,2,3],ymm4[4,5,6,7]
; AVX512F-ONLY-SLOW-NEXT: vinserti64x4 $0, %ymm4, %zmm0, %zmm8 {%k4}
; AVX512F-ONLY-SLOW-NEXT: vpbroadcastq 136(%rcx), %ymm4
-; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = mem[0,1,2,3],ymm4[4,5],mem[6,7]
+; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = mem[0,1,2,3],ymm4[4,5,6,7]
; AVX512F-ONLY-SLOW-NEXT: vinserti64x4 $0, %ymm4, %zmm0, %zmm10 {%k4}
; AVX512F-ONLY-SLOW-NEXT: vpbroadcastq 200(%rcx), %ymm4
-; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = mem[0,1,2,3],ymm4[4,5],mem[6,7]
+; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = mem[0,1,2,3],ymm4[4,5,6,7]
; AVX512F-ONLY-SLOW-NEXT: vinserti64x4 $0, %ymm4, %zmm0, %zmm3 {%k4}
; AVX512F-ONLY-SLOW-NEXT: movb $56, %cl
; AVX512F-ONLY-SLOW-NEXT: kmovw %ecx, %k4
@@ -6497,18 +6496,18 @@ define void @store_i64_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %zmm16, %zmm28 {%k3}
; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %zmm12, %zmm28 {%k4}
; AVX512F-ONLY-FAST-NEXT: vpbroadcastq 8(%rcx), %ymm8
-; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm8 = mem[0,1,2,3],ymm8[4,5],mem[6,7]
+; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm8 = mem[0,1,2,3],ymm8[4,5,6,7]
; AVX512F-ONLY-FAST-NEXT: movb $6, %sil
; AVX512F-ONLY-FAST-NEXT: kmovw %esi, %k4
; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $0, %ymm8, %zmm0, %zmm15 {%k4}
; AVX512F-ONLY-FAST-NEXT: vpbroadcastq 72(%rcx), %ymm8
-; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm8 = mem[0,1,2,3],ymm8[4,5],mem[6,7]
+; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm8 = mem[0,1,2,3],ymm8[4,5,6,7]
; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $0, %ymm8, %zmm0, %zmm11 {%k4}
; AVX512F-ONLY-FAST-NEXT: vpbroadcastq 136(%rcx), %ymm8
-; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm8 = mem[0,1,2,3],ymm8[4,5],mem[6,7]
+; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm8 = mem[0,1,2,3],ymm8[4,5,6,7]
; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $0, %ymm8, %zmm0, %zmm2 {%k4}
; AVX512F-ONLY-FAST-NEXT: vpbroadcastq 200(%rcx), %ymm8
-; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm8 = mem[0,1,2,3],ymm8[4,5],mem[6,7]
+; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm8 = mem[0,1,2,3],ymm8[4,5,6,7]
; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $0, %ymm8, %zmm0, %zmm5 {%k4}
; AVX512F-ONLY-FAST-NEXT: movb $56, %cl
; AVX512F-ONLY-FAST-NEXT: kmovw %ecx, %k4
@@ -6915,18 +6914,18 @@ define void @store_i64_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX512DQ-SLOW-NEXT: vpermi2q %zmm7, %zmm6, %zmm3
; AVX512DQ-SLOW-NEXT: vinserti64x2 $3, 192(%rax), %zmm3, %zmm4 {%k4}
; AVX512DQ-SLOW-NEXT: vpbroadcastq 8(%rcx), %ymm3
-; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = mem[0,1,2,3],ymm3[4,5],mem[6,7]
+; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = mem[0,1,2,3],ymm3[4,5,6,7]
; AVX512DQ-SLOW-NEXT: movb $6, %sil
; AVX512DQ-SLOW-NEXT: kmovw %esi, %k4
; AVX512DQ-SLOW-NEXT: vinserti64x4 $0, %ymm3, %zmm0, %zmm19 {%k4}
; AVX512DQ-SLOW-NEXT: vpbroadcastq 72(%rcx), %ymm3
-; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = mem[0,1,2,3],ymm3[4,5],mem[6,7]
+; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = mem[0,1,2,3],ymm3[4,5,6,7]
; AVX512DQ-SLOW-NEXT: vinserti64x4 $0, %ymm3, %zmm0, %zmm15 {%k4}
; AVX512DQ-SLOW-NEXT: vpbroadcastq 136(%rcx), %ymm3
-; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = mem[0,1,2,3],ymm3[4,5],mem[6,7]
+; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = mem[0,1,2,3],ymm3[4,5,6,7]
; AVX512DQ-SLOW-NEXT: vinserti64x4 $0, %ymm3, %zmm0, %zmm13 {%k4}
; AVX512DQ-SLOW-NEXT: vpbroadcastq 200(%rcx), %ymm3
-; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = mem[0,1,2,3],ymm3[4,5],mem[6,7]
+; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = mem[0,1,2,3],ymm3[4,5,6,7]
; AVX512DQ-SLOW-NEXT: vinserti64x4 $0, %ymm3, %zmm0, %zmm5 {%k4}
; AVX512DQ-SLOW-NEXT: movb $56, %cl
; AVX512DQ-SLOW-NEXT: kmovw %ecx, %k4
@@ -7369,18 +7368,18 @@ define void @store_i64_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX512DQ-FAST-NEXT: vpermi2q %zmm9, %zmm8, %zmm5
; AVX512DQ-FAST-NEXT: vinserti64x2 $3, 192(%rax), %zmm5, %zmm7 {%k4}
; AVX512DQ-FAST-NEXT: vpbroadcastq 8(%rcx), %ymm5
-; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm5 = mem[0,1,2,3],ymm5[4,5],mem[6,7]
+; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm5 = mem[0,1,2,3],ymm5[4,5,6,7]
; AVX512DQ-FAST-NEXT: movb $6, %sil
; AVX512DQ-FAST-NEXT: kmovw %esi, %k4
; AVX512DQ-FAST-NEXT: vinserti64x4 $0, %ymm5, %zmm0, %zmm28 {%k4}
; AVX512DQ-FAST-NEXT: vpbroadcastq 72(%rcx), %ymm5
-; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm5 = mem[0,1,2,3],ymm5[4,5],mem[6,7]
+; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm5 = mem[0,1,2,3],ymm5[4,5,6,7]
; AVX512DQ-FAST-NEXT: vinserti64x4 $0, %ymm5, %zmm0, %zmm14 {%k4}
; AVX512DQ-FAST-NEXT: vpbroadcastq 136(%rcx), %ymm5
-; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm5 = mem[0,1,2,3],ymm5[4,5],mem[6,7]
+; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm5 = mem[0,1,2,3],ymm5[4,5,6,7]
; AVX512DQ-FAST-NEXT: vinserti64x4 $0, %ymm5, %zmm0, %zmm4 {%k4}
; AVX512DQ-FAST-NEXT: vpbroadcastq 200(%rcx), %ymm5
-; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm5 = mem[0,1,2,3],ymm5[4,5],mem[6,7]
+; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm5 = mem[0,1,2,3],ymm5[4,5,6,7]
; AVX512DQ-FAST-NEXT: vinserti64x4 $0, %ymm5, %zmm0, %zmm6 {%k4}
; AVX512DQ-FAST-NEXT: movb $56, %cl
; AVX512DQ-FAST-NEXT: kmovw %ecx, %k4
@@ -7792,18 +7791,18 @@ define void @store_i64_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX512BW-ONLY-SLOW-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-ONLY-SLOW-NEXT: vmovdqa64 %zmm0, %zmm9 {%k4}
; AVX512BW-ONLY-SLOW-NEXT: vpbroadcastq 8(%rcx), %ymm4
-; AVX512BW-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = mem[0,1,2,3],ymm4[4,5],mem[6,7]
+; AVX512BW-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = mem[0,1,2,3],ymm4[4,5,6,7]
; AVX512BW-ONLY-SLOW-NEXT: movb $6, %sil
; AVX512BW-ONLY-SLOW-NEXT: kmovd %esi, %k4
; AVX512BW-ONLY-SLOW-NEXT: vinserti64x4 $0, %ymm4, %zmm0, %zmm31 {%k4}
; AVX512BW-ONLY-SLOW-NEXT: vpbroadcastq 72(%rcx), %ymm4
-; AVX512BW-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = mem[0,1,2,3],ymm4[4,5],mem[6,7]
+; AVX512BW-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = mem[0,1,2,3],ymm4[4,5,6,7]
; AVX512BW-ONLY-SLOW-NEXT: vinserti64x4 $0, %ymm4, %zmm0, %zmm8 {%k4}
; AVX512BW-ONLY-SLOW-NEXT: vpbroadcastq 136(%rcx), %ymm4
-; AVX512BW-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = mem[0,1,2,3],ymm4[4,5],mem[6,7]
+; AVX512BW-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = mem[0,1,2,3],ymm4[4,5,6,7]
; AVX512BW-ONLY-SLOW-NEXT: vinserti64x4 $0, %ymm4, %zmm0, %zmm10 {%k4}
; AVX512BW-ONLY-SLOW-NEXT: vpbroadcastq 200(%rcx), %ymm4
-; AVX512BW-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = mem[0,1,2,3],ymm4[4,5],mem[6,7]
+; AVX512BW-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = mem[0,1,2,3],ymm4[4,5,6,7]
; AVX512BW-ONLY-SLOW-NEXT: vinserti64x4 $0, %ymm4, %zmm0, %zmm3 {%k4}
; AVX512BW-ONLY-SLOW-NEXT: movb $56, %cl
; AVX512BW-ONLY-SLOW-NEXT: kmovd %ecx, %k4
@@ -8252,18 +8251,18 @@ define void @store_i64_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX512BW-ONLY-FAST-NEXT: vmovdqa64 %zmm16, %zmm28 {%k3}
; AVX512BW-ONLY-FAST-NEXT: vmovdqa64 %zmm12, %zmm28 {%k4}
; AVX512BW-ONLY-FAST-NEXT: vpbroadcastq 8(%rcx), %ymm8
-; AVX512BW-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm8 = mem[0,1,2,3],ymm8[4,5],mem[6,7]
+; AVX512BW-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm8 = mem[0,1,2,3],ymm8[4,5,6,7]
; AVX512BW-ONLY-FAST-NEXT: movb $6, %sil
; AVX512BW-ONLY-FAST-NEXT: kmovd %esi, %k4
; AVX512BW-ONLY-FAST-NEXT: vinserti64x4 $0, %ymm8, %zmm0, %zmm15 {%k4}
; AVX512BW-ONLY-FAST-NEXT: vpbroadcastq 72(%rcx), %ymm8
-; AVX512BW-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm8 = mem[0,1,2,3],ymm8[4,5],mem[6,7]
+; AVX512BW-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm8 = mem[0,1,2,3],ymm8[4,5,6,7]
; AVX512BW-ONLY-FAST-NEXT: vinserti64x4 $0, %ymm8, %zmm0, %zmm11 {%k4}
; AVX512BW-ONLY-FAST-NEXT: vpbroadcastq 136(%rcx), %ymm8
-; AVX512BW-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm8 = mem[0,1,2,3],ymm8[4,5],mem[6,7]
+; AVX512BW-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm8 = mem[0,1,2,3],ymm8[4,5,6,7]
; AVX512BW-ONLY-FAST-NEXT: vinserti64x4 $0, %ymm8, %zmm0, %zmm2 {%k4}
; AVX512BW-ONLY-FAST-NEXT: vpbroadcastq 200(%rcx), %ymm8
-; AVX512BW-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm8 = mem[0,1,2,3],ymm8[4,5],mem[6,7]
+; AVX512BW-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm8 = mem[0,1,2,3],ymm8[4,5,6,7]
; AVX512BW-ONLY-FAST-NEXT: vinserti64x4 $0, %ymm8, %zmm0, %zmm5 {%k4}
; AVX512BW-ONLY-FAST-NEXT: movb $56, %cl
; AVX512BW-ONLY-FAST-NEXT: kmovd %ecx, %k4
@@ -8670,18 +8669,18 @@ define void @store_i64_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX512DQBW-SLOW-NEXT: vpermi2q %zmm7, %zmm6, %zmm3
; AVX512DQBW-SLOW-NEXT: vinserti64x2 $3, 192(%rax), %zmm3, %zmm4 {%k4}
; AVX512DQBW-SLOW-NEXT: vpbroadcastq 8(%rcx), %ymm3
-; AVX512DQBW-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = mem[0,1,2,3],ymm3[4,5],mem[6,7]
+; AVX512DQBW-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = mem[0,1,2,3],ymm3[4,5,6,7]
; AVX512DQBW-SLOW-NEXT: movb $6, %sil
; AVX512DQBW-SLOW-NEXT: kmovd %esi, %k4
; AVX512DQBW-SLOW-NEXT: vinserti64x4 $0, %ymm3, %zmm0, %zmm19 {%k4}
; AVX512DQBW-SLOW-NEXT: vpbroadcastq 72(%rcx), %ymm3
-; AVX512DQBW-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = mem[0,1,2,3],ymm3[4,5],mem[6,7]
+; AVX512DQBW-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = mem[0,1,2,3],ymm3[4,5,6,7]
; AVX512DQBW-SLOW-NEXT: vinserti64x4 $0, %ymm3, %zmm0, %zmm15 {%k4}
; AVX512DQBW-SLOW-NEXT: vpbroadcastq 136(%rcx), %ymm3
-; AVX512DQBW-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = mem[0,1,2,3],ymm3[4,5],mem[6,7]
+; AVX512DQBW-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = mem[0,1,2,3],ymm3[4,5,6,7]
; AVX512DQBW-SLOW-NEXT: vinserti64x4 $0, %ymm3, %zmm0, %zmm13 {%k4}
; AVX512DQBW-SLOW-NEXT: vpbroadcastq 200(%rcx), %ymm3
-; AVX512DQBW-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = mem[0,1,2,3],ymm3[4,5],mem[6,7]
+; AVX512DQBW-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = mem[0,1,2,3],ymm3[4,5,6,7]
; AVX512DQBW-SLOW-NEXT: vinserti64x4 $0, %ymm3, %zmm0, %zmm5 {%k4}
; AVX512DQBW-SLOW-NEXT: movb $56, %cl
; AVX512DQBW-SLOW-NEXT: kmovd %ecx, %k4
@@ -9124,18 +9123,18 @@ define void @store_i64_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX512DQBW-FAST-NEXT: vpermi2q %zmm9, %zmm8, %zmm5
; AVX512DQBW-FAST-NEXT: vinserti64x2 $3, 192(%rax), %zmm5, %zmm7 {%k4}
; AVX512DQBW-FAST-NEXT: vpbroadcastq 8(%rcx), %ymm5
-; AVX512DQBW-FAST-NEXT: vpblendd {{.*#+}} ymm5 = mem[0,1,2,3],ymm5[4,5],mem[6,7]
+; AVX512DQBW-FAST-NEXT: vpblendd {{.*#+}} ymm5 = mem[0,1,2,3],ymm5[4,5,6,7]
; AVX512DQBW-FAST-NEXT: movb $6, %sil
; AVX512DQBW-FAST-NEXT: kmovd %esi, %k4
; AVX512DQBW-FAST-NEXT: vinserti64x4 $0, %ymm5, %zmm0, %zmm28 {%k4}
; AVX512DQBW-FAST-NEXT: vpbroadcastq 72(%rcx), %ymm5
-; AVX512DQBW-FAST-NEXT: vpblendd {{.*#+}} ymm5 = mem[0,1,2,3],ymm5[4,5],mem[6,7]
+; AVX512DQBW-FAST-NEXT: vpblendd {{.*#+}} ymm5 = mem[0,1,2,3],ymm5[4,5,6,7]
; AVX512DQBW-FAST-NEXT: vinserti64x4 $0, %ymm5, %zmm0, %zmm14 {%k4}
; AVX512DQBW-FAST-NEXT: vpbroadcastq 136(%rcx), %ymm5
-; AVX512DQBW-FAST-NEXT: vpblendd {{.*#+}} ymm5 = mem[0,1,2,3],ymm5[4,5],mem[6,7]
+; AVX512DQBW-FAST-NEXT: vpblendd {{.*#+}} ymm5 = mem[0,1,2,3],ymm5[4,5,6,7]
; AVX512DQBW-FAST-NEXT: vinserti64x4 $0, %ymm5, %zmm0, %zmm4 {%k4}
; AVX512DQBW-FAST-NEXT: vpbroadcastq 200(%rcx), %ymm5
-; AVX512DQBW-FAST-NEXT: vpblendd {{.*#+}} ymm5 = mem[0,1,2,3],ymm5[4,5],mem[6,7]
+; AVX512DQBW-FAST-NEXT: vpblendd {{.*#+}} ymm5 = mem[0,1,2,3],ymm5[4,5,6,7]
; AVX512DQBW-FAST-NEXT: vinserti64x4 $0, %ymm5, %zmm0, %zmm6 {%k4}
; AVX512DQBW-FAST-NEXT: movb $56, %cl
; AVX512DQBW-FAST-NEXT: kmovd %ecx, %k4
@@ -10377,7 +10376,7 @@ define void @store_i64_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
;
; AVX1-ONLY-LABEL: store_i64_stride7_vf64:
; AVX1-ONLY: # %bb.0:
-; AVX1-ONLY-NEXT: subq $3960, %rsp # imm = 0xF78
+; AVX1-ONLY-NEXT: subq $3816, %rsp # imm = 0xEE8
; AVX1-ONLY-NEXT: movq {{[0-9]+}}(%rsp), %rax
; AVX1-ONLY-NEXT: vmovaps (%rsi), %ymm7
; AVX1-ONLY-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -10399,7 +10398,7 @@ define void @store_i64_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1],ymm4[2,3],ymm3[4,5],ymm4[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vbroadcastsd 8(%rcx), %ymm3
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm0[0,1,2,3],ymm3[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm0[0,1,2,3],ymm3[4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps (%rsi), %xmm4
; AVX1-ONLY-NEXT: vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm4 = xmm4[2,3,2,3]
@@ -10439,7 +10438,7 @@ define void @store_i64_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm4[1]
; AVX1-ONLY-NEXT: vbroadcastsd 40(%rcx), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm1
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -10475,7 +10474,7 @@ define void @store_i64_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 64(%rdx), %ymm0
; AVX1-ONLY-NEXT: vbroadcastsd 72(%rcx), %ymm2
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm0[0,1,2,3],ymm2[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm0[0,1,2,3],ymm2[4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps 64(%rsi), %xmm3
; AVX1-ONLY-NEXT: vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm3 = xmm3[2,3,2,3]
@@ -10521,7 +10520,7 @@ define void @store_i64_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm4[1]
; AVX1-ONLY-NEXT: vbroadcastsd 104(%rcx), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm1
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -10557,7 +10556,7 @@ define void @store_i64_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 128(%rdx), %ymm0
; AVX1-ONLY-NEXT: vbroadcastsd 136(%rcx), %ymm2
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm0[0,1,2,3],ymm2[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm0[0,1,2,3],ymm2[4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps 128(%rsi), %xmm3
; AVX1-ONLY-NEXT: vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm3 = xmm3[2,3,2,3]
@@ -10603,7 +10602,7 @@ define void @store_i64_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm4[1]
; AVX1-ONLY-NEXT: vbroadcastsd 168(%rcx), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm1
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -10639,7 +10638,7 @@ define void @store_i64_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 192(%rdx), %ymm0
; AVX1-ONLY-NEXT: vbroadcastsd 200(%rcx), %ymm2
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm0[0,1,2,3],ymm2[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm0[0,1,2,3],ymm2[4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps 192(%rsi), %xmm3
; AVX1-ONLY-NEXT: vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm3 = xmm3[2,3,2,3]
@@ -10685,7 +10684,7 @@ define void @store_i64_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm4[1]
; AVX1-ONLY-NEXT: vbroadcastsd 232(%rcx), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm1
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -10721,7 +10720,7 @@ define void @store_i64_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 256(%rdx), %ymm0
; AVX1-ONLY-NEXT: vbroadcastsd 264(%rcx), %ymm2
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm0[0,1,2,3],ymm2[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm0[0,1,2,3],ymm2[4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps 256(%rsi), %xmm3
; AVX1-ONLY-NEXT: vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm3 = xmm3[2,3,2,3]
@@ -10767,7 +10766,7 @@ define void @store_i64_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm4[1]
; AVX1-ONLY-NEXT: vbroadcastsd 296(%rcx), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm1
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -10803,7 +10802,7 @@ define void @store_i64_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 320(%rdx), %ymm0
; AVX1-ONLY-NEXT: vbroadcastsd 328(%rcx), %ymm2
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm0[0,1,2,3],ymm2[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm0[0,1,2,3],ymm2[4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps 320(%rsi), %xmm3
; AVX1-ONLY-NEXT: vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm3 = xmm3[2,3,2,3]
@@ -10849,124 +10848,105 @@ define void @store_i64_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm4[1]
; AVX1-ONLY-NEXT: vbroadcastsd 360(%rcx), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm1
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 352(%rsi), %ymm1
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 352(%rsi), %ymm15
; AVX1-ONLY-NEXT: vmovaps 368(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm1[1],ymm0[1],ymm1[3],ymm0[3]
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm15[1],ymm0[1],ymm15[3],ymm0[3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],mem[6,7]
; AVX1-ONLY-NEXT: vmovaps 368(%rax), %xmm1
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 352(%r9), %ymm1
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 352(%r9), %ymm14
; AVX1-ONLY-NEXT: vmovaps 368(%r8), %xmm0
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm1[1],ymm0[1],ymm1[3],ymm0[3]
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm14[1],ymm0[1],ymm14[3],ymm0[3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],mem[6,7]
-; AVX1-ONLY-NEXT: vmovaps 368(%rcx), %xmm15
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm1 = xmm15[2,3,2,3]
+; AVX1-ONLY-NEXT: vmovaps 368(%rcx), %xmm13
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm1 = xmm13[2,3,2,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 384(%r9), %xmm14
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm1 = xmm14[0,1,0,1]
-; AVX1-ONLY-NEXT: vmovaps 384(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm1
-; AVX1-ONLY-NEXT: vmovaps 384(%r8), %xmm2
-; AVX1-ONLY-NEXT: vmovaps 384(%rax), %xmm13
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm13, %ymm2, %ymm3
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm3[0,1],ymm1[2,3],ymm3[4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 384(%rdx), %ymm1
-; AVX1-ONLY-NEXT: vbroadcastsd 392(%rcx), %ymm3
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm1[0,1,2,3],ymm3[4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT: vmovaps 384(%rsi), %xmm0
-; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm4 = xmm0[2,3,2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm4[0,1],ymm3[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm2
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3,4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT: vmovaps 384(%r9), %xmm12
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm0 = xmm12[0,1,0,1]
+; AVX1-ONLY-NEXT: vmovaps 384(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; AVX1-ONLY-NEXT: vmovaps 384(%r8), %xmm1
+; AVX1-ONLY-NEXT: vmovaps 384(%rax), %xmm11
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm11, %ymm1, %ymm2
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3],ymm2[4,5],ymm0[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 384(%rsi), %ymm10
-; AVX1-ONLY-NEXT: vmovaps 400(%rdi), %xmm2
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm2 = ymm10[1],ymm2[1],ymm10[3],ymm2[3]
+; AVX1-ONLY-NEXT: vmovaps 384(%rdx), %ymm0
+; AVX1-ONLY-NEXT: vbroadcastsd 392(%rcx), %ymm2
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm0[0,1,2,3],ymm2[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps 384(%rsi), %xmm3
+; AVX1-ONLY-NEXT: vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm3 = xmm3[2,3,2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT: vmovaps 400(%rax), %xmm2
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm1[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 384(%rsi), %ymm10
+; AVX1-ONLY-NEXT: vmovaps 400(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm1 = ymm10[1],ymm1[1],ymm10[3],ymm1[3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT: vmovaps 400(%rax), %xmm1
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 384(%r9), %ymm8
-; AVX1-ONLY-NEXT: vmovaps 400(%r8), %xmm1
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm1 = ymm8[1],ymm1[1],ymm8[3],ymm1[3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],mem[6,7]
+; AVX1-ONLY-NEXT: vmovaps 400(%r8), %xmm0
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm8[1],ymm0[1],ymm8[3],ymm0[3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],mem[6,7]
; AVX1-ONLY-NEXT: vmovaps 400(%rcx), %xmm7
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm2 = xmm7[2,3,2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm1[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 416(%rsi), %xmm1
-; AVX1-ONLY-NEXT: vmovaps 416(%rdi), %xmm2
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm3 = xmm2[0],xmm1[0]
-; AVX1-ONLY-NEXT: vinsertf128 $1, 416(%rcx), %ymm3, %ymm4
-; AVX1-ONLY-NEXT: vmovaps 416(%rdx), %xmm6
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm6, %ymm3, %ymm3
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm0 = ymm3[0],ymm4[1],ymm3[2],ymm4[2]
-; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm4
-; AVX1-ONLY-NEXT: vmovapd 416(%r9), %xmm5
-; AVX1-ONLY-NEXT: vmovapd 416(%r8), %xmm9
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm11 = xmm9[0],xmm5[0]
-; AVX1-ONLY-NEXT: vmovapd 416(%rax), %xmm2
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm11, %ymm11
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm11[0,1,2,3,4,5],ymm4[6,7]
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm1 = xmm7[2,3,2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm6[1]
-; AVX1-ONLY-NEXT: vbroadcastsd 424(%rcx), %ymm4
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm4[4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm9, %ymm0, %ymm4
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm4[6,7]
+; AVX1-ONLY-NEXT: vmovaps 416(%rsi), %xmm3
+; AVX1-ONLY-NEXT: vmovaps 416(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm1 = xmm0[0],xmm3[0]
+; AVX1-ONLY-NEXT: vinsertf128 $1, 416(%rcx), %ymm1, %ymm5
+; AVX1-ONLY-NEXT: vmovaps 416(%rdx), %xmm2
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm1, %ymm1
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm1 = ymm1[0],ymm5[1],ymm1[2],ymm5[2]
+; AVX1-ONLY-NEXT: vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm5
+; AVX1-ONLY-NEXT: vmovapd 416(%r9), %xmm6
+; AVX1-ONLY-NEXT: vmovapd 416(%r8), %xmm0
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm1 = xmm0[0],xmm6[0]
+; AVX1-ONLY-NEXT: vmovapd 416(%rax), %xmm4
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm4, %ymm1, %ymm1
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm5[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm3[1],xmm2[1]
+; AVX1-ONLY-NEXT: vbroadcastsd 424(%rcx), %ymm2
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 448(%rsi), %xmm0
; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm1 = xmm0[2,3,2,3]
-; AVX1-ONLY-NEXT: vbroadcastsd 456(%rcx), %ymm4
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm4[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovapd 448(%rdx), %ymm6
-; AVX1-ONLY-NEXT: vinsertf128 $1, 448(%r8), %ymm6, %ymm4
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm4[2,3],ymm1[4,5],ymm4[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 464(%rcx), %xmm0
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 464(%rdx), %xmm1
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm4 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
-; AVX1-ONLY-NEXT: vmovaps 448(%r8), %ymm3
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm3[4,5],ymm4[6,7]
-; AVX1-ONLY-NEXT: vbroadcastsd 464(%r9), %ymm9
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm4[0,1,2,3,4,5],ymm9[6,7]
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[2,3,2,3]
+; AVX1-ONLY-NEXT: vbroadcastsd 456(%rcx), %ymm1
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovapd 448(%rdx), %ymm5
+; AVX1-ONLY-NEXT: vinsertf128 $1, 448(%r8), %ymm5, %ymm1
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 480(%rsi), %xmm9
-; AVX1-ONLY-NEXT: vmovaps 480(%rdi), %xmm12
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm11 = xmm12[0],xmm9[0]
-; AVX1-ONLY-NEXT: vinsertf128 $1, 480(%rcx), %ymm11, %ymm1
+; AVX1-ONLY-NEXT: vmovaps 480(%rsi), %xmm2
+; AVX1-ONLY-NEXT: vmovaps 480(%rdi), %xmm9
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm1 = xmm9[0],xmm2[0]
+; AVX1-ONLY-NEXT: vinsertf128 $1, 480(%rcx), %ymm1, %ymm3
; AVX1-ONLY-NEXT: vmovaps 480(%rdx), %xmm0
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm11, %ymm11
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm1 = ymm11[0],ymm1[1],ymm11[2],ymm1[2]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm1
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm1 = ymm1[0],ymm3[1],ymm1[2],ymm3[2]
; AVX1-ONLY-NEXT: vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm1 = xmm9[2,3,2,3]
-; AVX1-ONLY-NEXT: vbroadcastsd 488(%rcx), %ymm9
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm9[4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm1 = xmm2[2,3,2,3]
+; AVX1-ONLY-NEXT: vbroadcastsd 488(%rcx), %ymm2
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX1-ONLY-NEXT: vinsertf128 $1, 480(%r8), %ymm0, %ymm0
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovapd 496(%rcx), %xmm9
-; AVX1-ONLY-NEXT: vmovapd 496(%rdx), %xmm0
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm9[0],ymm0[2],ymm9[2]
-; AVX1-ONLY-NEXT: vmovapd 480(%r8), %ymm11
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm11[2],ymm0[3]
-; AVX1-ONLY-NEXT: vbroadcastsd 496(%r9), %ymm1
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3]
-; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps (%rdi), %ymm0
; AVX1-ONLY-NEXT: vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm0 = ymm0[0],mem[0],ymm0[2],mem[2]
@@ -11144,23 +11124,21 @@ define void @store_i64_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 352(%rdi), %ymm0
-; AVX1-ONLY-NEXT: vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm0 = ymm0[0],mem[0],ymm0[2],mem[2]
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm15[0],ymm0[2],ymm15[2]
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX1-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm1 = xmm1[1],mem[1]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 352(%r8), %ymm0
-; AVX1-ONLY-NEXT: vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm0 = ymm0[0],mem[0],ymm0[2],mem[2]
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm14[0],ymm0[2],ymm14[2]
; AVX1-ONLY-NEXT: vmovaps 368(%rdx), %xmm1
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm1 = xmm1[0],xmm15[0]
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm1 = xmm1[0],xmm13[0]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 384(%rdi), %ymm0
; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm10[0],ymm0[2],ymm10[2]
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm14[1],xmm13[1]
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm12[1],xmm11[1]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 384(%r8), %ymm0
@@ -11169,81 +11147,98 @@ define void @store_i64_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm1 = xmm1[0],xmm7[0]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm5[1],xmm2[1]
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm6[1],xmm4[1]
; AVX1-ONLY-NEXT: vmovapd 416(%rdi), %ymm1
-; AVX1-ONLY-NEXT: vmovapd 416(%rsi), %ymm5
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm1[0],ymm5[0],ymm1[2],ymm5[2]
+; AVX1-ONLY-NEXT: vmovapd 416(%rsi), %ymm2
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm1[0],ymm2[0],ymm1[2],ymm2[2]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3]
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovapd 416(%r8), %ymm0
; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm0[0],mem[0],ymm0[2],mem[2]
-; AVX1-ONLY-NEXT: vmovapd 432(%rcx), %xmm4
+; AVX1-ONLY-NEXT: vmovapd 432(%rcx), %xmm7
; AVX1-ONLY-NEXT: vmovapd 432(%rdx), %xmm1
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm1 = xmm1[0],xmm4[0]
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm1 = xmm1[0],xmm7[0]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3]
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovapd 432(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm0[0,1],mem[2,3]
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm5 = ymm5[0,0,3,2]
-; AVX1-ONLY-NEXT: vmovapd 416(%rax), %ymm2
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm5 = ymm2[2,3],ymm5[2,3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm5[0],ymm1[1],ymm5[2],ymm1[3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],mem[2,3]
+; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm2 = ymm2[0,0,3,2]
+; AVX1-ONLY-NEXT: vmovapd 416(%rax), %ymm6
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm2 = ymm6[2,3],ymm2[2,3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm2[0],ymm0[1],ymm2[2],ymm0[3]
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vinsertf128 $1, 448(%rax), %ymm3, %ymm1
+; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm0 = mem[0,0]
+; AVX1-ONLY-NEXT: vmovaps 448(%rdi), %xmm15
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm15, %ymm0, %ymm2
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3],ymm2[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovapd 448(%r8), %ymm13
+; AVX1-ONLY-NEXT: vinsertf128 $1, 448(%rax), %ymm13, %ymm2
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3],ymm2[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovapd 448(%rdi), %ymm0
+; AVX1-ONLY-NEXT: vmovapd 448(%rsi), %ymm2
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm2[0],ymm0[2],ymm2[2]
; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm3 = mem[0,0]
-; AVX1-ONLY-NEXT: vmovaps 448(%rdi), %xmm14
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm14, %ymm3, %ymm3
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm3[2,3],ymm1[4,5],ymm3[6,7]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm8 = ymm3[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT: vmovapd 464(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm5[2,3]
+; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm2 = ymm2[0,0,3,2]
+; AVX1-ONLY-NEXT: vmovapd 448(%rax), %ymm4
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm2 = ymm4[2,3],ymm2[2,3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm2[0],ymm0[1],ymm2[2],ymm0[3]
+; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm9, %ymm0, %ymm0
+; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm2 = mem[0,0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm0[0,1],ymm2[2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovapd 480(%r8), %ymm1
+; AVX1-ONLY-NEXT: vinsertf128 $1, 480(%rax), %ymm1, %ymm3
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm3[0,1],ymm2[2,3],ymm3[4,5],ymm2[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovapd 448(%rdi), %ymm1
-; AVX1-ONLY-NEXT: vmovapd 448(%rsi), %ymm3
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm1[0],ymm3[0],ymm1[2],ymm3[2]
-; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm10 = mem[0,0]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm15 = ymm10[0],ymm1[1,2,3]
-; AVX1-ONLY-NEXT: vmovapd 464(%rdi), %xmm10
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm6 = ymm10[0,1],ymm6[2,3]
+; AVX1-ONLY-NEXT: vmovapd 480(%rdi), %ymm2
+; AVX1-ONLY-NEXT: vmovapd 480(%rsi), %ymm3
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm2 = ymm2[0],ymm3[0],ymm2[2],ymm3[2]
+; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm5 = mem[0,0]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm5[0,1],ymm2[2,3]
+; AVX1-ONLY-NEXT: vmovapd 496(%rdi), %xmm5
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm5 = ymm5[0,1],mem[2,3]
; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm3 = ymm3[0,0,3,2]
-; AVX1-ONLY-NEXT: vmovapd 448(%rax), %ymm0
+; AVX1-ONLY-NEXT: vmovapd 480(%rax), %ymm0
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm3 = ymm0[2,3],ymm3[2,3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm3[0],ymm6[1],ymm3[2],ymm6[3]
-; AVX1-ONLY-NEXT: vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm3 = mem[0,0]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm12, %ymm3, %ymm3
-; AVX1-ONLY-NEXT: vinsertf128 $1, 480(%rax), %ymm11, %ymm6
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm6[0,1],ymm3[2,3],ymm6[4,5],ymm3[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovapd 480(%rdi), %ymm3
-; AVX1-ONLY-NEXT: vmovapd 480(%rsi), %ymm6
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm3 = ymm3[0],ymm6[0],ymm3[2],ymm6[2]
-; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm12 = mem[0,0]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm12 = ymm12[0],ymm3[1,2,3]
-; AVX1-ONLY-NEXT: vmovapd 496(%rdi), %xmm3
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm3 = ymm3[0,1],mem[2,3]
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm6 = ymm6[0,0,3,2]
-; AVX1-ONLY-NEXT: vmovapd 480(%rax), %ymm1
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm6 = ymm1[2,3],ymm6[2,3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm3 = ymm6[0],ymm3[1],ymm6[2],ymm3[3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm3 = ymm3[0],ymm5[1],ymm3[2],ymm5[3]
; AVX1-ONLY-NEXT: vmovupd %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm4 = xmm4[1],mem[1]
-; AVX1-ONLY-NEXT: vbroadcastsd 440(%r9), %ymm6
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm4 = ymm4[0,1],ymm6[2],ymm4[3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm4[0,1,2],ymm2[3]
-; AVX1-ONLY-NEXT: vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovupd {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm4 = xmm2[1],mem[1]
-; AVX1-ONLY-NEXT: vbroadcastsd 472(%r9), %ymm6
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm4 = ymm4[0,1],ymm6[2],ymm4[3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm15[0],ymm0[1],ymm15[2,3]
-; AVX1-ONLY-NEXT: vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm3 = xmm7[1],mem[1]
+; AVX1-ONLY-NEXT: vbroadcastsd 440(%r9), %ymm7
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm3 = ymm3[0,1],ymm7[2,3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm3 = ymm3[0,1,2],ymm6[3]
+; AVX1-ONLY-NEXT: vmovupd %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovapd 464(%rcx), %xmm3
+; AVX1-ONLY-NEXT: vmovapd 464(%rdx), %xmm6
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm6 = xmm6[0],xmm3[0]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm6 = ymm6[0,1],ymm13[2,3]
+; AVX1-ONLY-NEXT: vbroadcastsd 464(%r9), %ymm7
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm5 = ymm6[0,1,2],ymm7[3]
+; AVX1-ONLY-NEXT: vmovupd %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm3 = xmm3[1],mem[1]
+; AVX1-ONLY-NEXT: vbroadcastsd 472(%r9), %ymm7
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm7 = ymm3[0,1],ymm7[2,3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm3 = ymm8[0],ymm4[1],ymm8[2,3]
+; AVX1-ONLY-NEXT: vmovupd %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm3 = ymm7[0,1,2],ymm4[3]
+; AVX1-ONLY-NEXT: vmovupd %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovapd 496(%rcx), %xmm4
+; AVX1-ONLY-NEXT: vmovapd 496(%rdx), %xmm7
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm7 = xmm7[0],xmm4[0]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm7[0,1],ymm1[2,3]
+; AVX1-ONLY-NEXT: vbroadcastsd 496(%r9), %ymm7
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm1[0,1,2],ymm7[3]
+; AVX1-ONLY-NEXT: vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm4[1],mem[1]
+; AVX1-ONLY-NEXT: vbroadcastsd 504(%r9), %ymm4
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm4 = ymm1[0,1],ymm4[2,3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm2[0],ymm0[1],ymm2[2,3]
+; AVX1-ONLY-NEXT: vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm4[0,1,2],ymm0[3]
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm9[1],mem[1]
-; AVX1-ONLY-NEXT: vbroadcastsd 504(%r9), %ymm6
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm6[2],ymm0[3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm12[0],ymm1[1],ymm12[2,3]
-; AVX1-ONLY-NEXT: vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3]
-; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX1-ONLY-NEXT: vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm5 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm5 = xmm0[0],mem[0]
@@ -11251,16 +11246,16 @@ define void @store_i64_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm4 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm4 = xmm0[0],mem[0]
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX1-ONLY-NEXT: vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm12 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm12 = xmm0[0],mem[0]
+; AVX1-ONLY-NEXT: vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm7 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm7 = xmm0[0],mem[0]
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX1-ONLY-NEXT: vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm3 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm3 = xmm0[0],mem[0]
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX1-ONLY-NEXT: vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm2 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm2 = xmm0[0],mem[0]
-; AVX1-ONLY-NEXT: vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm14, %xmm14 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm14 = xmm14[0],mem[0]
+; AVX1-ONLY-NEXT: vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm15 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm15 = xmm15[0],mem[0]
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX1-ONLY-NEXT: vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm1 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm1 = xmm0[0],mem[0]
@@ -11268,37 +11263,37 @@ define void @store_i64_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm0 = xmm0[0],mem[0]
; AVX1-ONLY-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; AVX1-ONLY-NEXT: vmovaps 256(%rdx), %xmm13
+; AVX1-ONLY-NEXT: vmovaps 256(%rdx), %xmm14
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm14 = xmm14[0],mem[0]
+; AVX1-ONLY-NEXT: vmovaps 128(%rdx), %xmm11
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm11 = xmm11[0],mem[0]
+; AVX1-ONLY-NEXT: vmovaps 64(%rdx), %xmm12
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm12 = xmm12[0],mem[0]
+; AVX1-ONLY-NEXT: vmovaps 192(%rdx), %xmm13
; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm13 = xmm13[0],mem[0]
-; AVX1-ONLY-NEXT: vmovaps 128(%rdx), %xmm8
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm8 = xmm8[0],mem[0]
-; AVX1-ONLY-NEXT: vmovaps 64(%rdx), %xmm6
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm6 = xmm6[0],mem[0]
-; AVX1-ONLY-NEXT: vmovaps 192(%rdx), %xmm10
+; AVX1-ONLY-NEXT: vmovaps 320(%rdx), %xmm10
; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm10 = xmm10[0],mem[0]
-; AVX1-ONLY-NEXT: vmovaps 320(%rdx), %xmm11
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm11 = xmm11[0],mem[0]
-; AVX1-ONLY-NEXT: vmovaps 448(%rdx), %xmm7
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm7 = xmm7[0],mem[0]
-; AVX1-ONLY-NEXT: vmovaps 384(%rdx), %xmm15
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm15 = xmm15[0],mem[0]
-; AVX1-ONLY-NEXT: vmovaps (%rdx), %xmm9
+; AVX1-ONLY-NEXT: vmovaps 448(%rdx), %xmm9
; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm9 = xmm9[0],mem[0]
-; AVX1-ONLY-NEXT: vmovaps %xmm9, 16(%rax)
+; AVX1-ONLY-NEXT: vmovaps 384(%rdx), %xmm8
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm8 = xmm8[0],mem[0]
+; AVX1-ONLY-NEXT: vmovaps (%rdx), %xmm6
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm6 = xmm6[0],mem[0]
+; AVX1-ONLY-NEXT: vmovaps %xmm6, 16(%rax)
; AVX1-ONLY-NEXT: vmovaps %xmm0, (%rax)
-; AVX1-ONLY-NEXT: vmovaps %xmm15, 2704(%rax)
+; AVX1-ONLY-NEXT: vmovaps %xmm8, 2704(%rax)
; AVX1-ONLY-NEXT: vmovaps %xmm1, 2688(%rax)
-; AVX1-ONLY-NEXT: vmovaps %xmm7, 3152(%rax)
-; AVX1-ONLY-NEXT: vmovaps %xmm14, 3136(%rax)
-; AVX1-ONLY-NEXT: vmovaps %xmm11, 2256(%rax)
+; AVX1-ONLY-NEXT: vmovaps %xmm9, 3152(%rax)
+; AVX1-ONLY-NEXT: vmovaps %xmm15, 3136(%rax)
+; AVX1-ONLY-NEXT: vmovaps %xmm10, 2256(%rax)
; AVX1-ONLY-NEXT: vmovaps %xmm2, 2240(%rax)
-; AVX1-ONLY-NEXT: vmovaps %xmm10, 1360(%rax)
+; AVX1-ONLY-NEXT: vmovaps %xmm13, 1360(%rax)
; AVX1-ONLY-NEXT: vmovaps %xmm3, 1344(%rax)
-; AVX1-ONLY-NEXT: vmovaps %xmm6, 464(%rax)
-; AVX1-ONLY-NEXT: vmovaps %xmm12, 448(%rax)
-; AVX1-ONLY-NEXT: vmovaps %xmm8, 912(%rax)
+; AVX1-ONLY-NEXT: vmovaps %xmm12, 464(%rax)
+; AVX1-ONLY-NEXT: vmovaps %xmm7, 448(%rax)
+; AVX1-ONLY-NEXT: vmovaps %xmm11, 912(%rax)
; AVX1-ONLY-NEXT: vmovaps %xmm4, 896(%rax)
-; AVX1-ONLY-NEXT: vmovaps %xmm13, 1808(%rax)
+; AVX1-ONLY-NEXT: vmovaps %xmm14, 1808(%rax)
; AVX1-ONLY-NEXT: vmovaps %xmm5, 1792(%rax)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm0, 3520(%rax)
@@ -11508,7 +11503,7 @@ define void @store_i64_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vmovaps %ymm0, 64(%rax)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm0, 32(%rax)
-; AVX1-ONLY-NEXT: addq $3960, %rsp # imm = 0xF78
+; AVX1-ONLY-NEXT: addq $3816, %rsp # imm = 0xEE8
; AVX1-ONLY-NEXT: vzeroupper
; AVX1-ONLY-NEXT: retq
;
@@ -11526,7 +11521,7 @@ define void @store_i64_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovddup {{.*#+}} xmm4 = mem[0,0]
; AVX2-ONLY-NEXT: vbroadcastsd 8(%rcx), %ymm5
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm3[0,1,2,3],ymm5[4,5],ymm3[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm3[0,1,2,3],ymm5[4,5,6,7]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm5[2,3,4,5,6,7]
; AVX2-ONLY-NEXT: vmovaps (%r8), %xmm9
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm9, %ymm0, %ymm5
@@ -11561,7 +11556,7 @@ define void @store_i64_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovddup {{.*#+}} xmm1 = mem[0,0]
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm3[1]
; AVX2-ONLY-NEXT: vbroadcastsd 40(%rcx), %ymm3
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5,6,7]
; AVX2-ONLY-NEXT: vmovaps 32(%rsi), %ymm3
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm2
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6,7]
@@ -11588,7 +11583,7 @@ define void @store_i64_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovaps 64(%rdx), %ymm0
; AVX2-ONLY-NEXT: vbroadcastsd 72(%rcx), %ymm1
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT: vmovddup {{.*#+}} xmm2 = mem[0,0]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5,6,7]
; AVX2-ONLY-NEXT: vmovaps 64(%r8), %xmm2
@@ -11633,7 +11628,7 @@ define void @store_i64_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovddup {{.*#+}} xmm1 = mem[0,0]
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm3[1]
; AVX2-ONLY-NEXT: vbroadcastsd 104(%rcx), %ymm3
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5,6,7]
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -11661,7 +11656,7 @@ define void @store_i64_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovaps 128(%rdx), %ymm0
; AVX2-ONLY-NEXT: vbroadcastsd 136(%rcx), %ymm1
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT: vmovddup {{.*#+}} xmm2 = mem[0,0]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5,6,7]
; AVX2-ONLY-NEXT: vmovaps 128(%r8), %xmm2
@@ -11706,7 +11701,7 @@ define void @store_i64_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovddup {{.*#+}} xmm1 = mem[0,0]
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm3[1]
; AVX2-ONLY-NEXT: vbroadcastsd 168(%rcx), %ymm3
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5,6,7]
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -11734,7 +11729,7 @@ define void @store_i64_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovaps 192(%rdx), %ymm0
; AVX2-ONLY-NEXT: vbroadcastsd 200(%rcx), %ymm1
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT: vmovddup {{.*#+}} xmm2 = mem[0,0]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5,6,7]
; AVX2-ONLY-NEXT: vmovaps 192(%r8), %xmm2
@@ -11779,7 +11774,7 @@ define void @store_i64_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovddup {{.*#+}} xmm1 = mem[0,0]
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm3[1]
; AVX2-ONLY-NEXT: vbroadcastsd 232(%rcx), %ymm3
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5,6,7]
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -11807,7 +11802,7 @@ define void @store_i64_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovaps 256(%rdx), %ymm0
; AVX2-ONLY-NEXT: vbroadcastsd 264(%rcx), %ymm1
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT: vmovddup {{.*#+}} xmm2 = mem[0,0]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5,6,7]
; AVX2-ONLY-NEXT: vmovaps 256(%r8), %xmm2
@@ -11852,7 +11847,7 @@ define void @store_i64_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovddup {{.*#+}} xmm1 = mem[0,0]
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm3[1]
; AVX2-ONLY-NEXT: vbroadcastsd 296(%rcx), %ymm3
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5,6,7]
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -11880,7 +11875,7 @@ define void @store_i64_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovaps 320(%rdx), %ymm0
; AVX2-ONLY-NEXT: vbroadcastsd 328(%rcx), %ymm1
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT: vmovddup {{.*#+}} xmm2 = mem[0,0]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5,6,7]
; AVX2-ONLY-NEXT: vmovaps 320(%r8), %xmm7
@@ -11924,7 +11919,7 @@ define void @store_i64_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovddup {{.*#+}} xmm1 = mem[0,0]
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm3[1]
; AVX2-ONLY-NEXT: vbroadcastsd 360(%rcx), %ymm3
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5,6,7]
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -11952,7 +11947,7 @@ define void @store_i64_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovaps 384(%rdx), %ymm0
; AVX2-ONLY-NEXT: vbroadcastsd 392(%rcx), %ymm1
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT: vmovddup {{.*#+}} xmm2 = mem[0,0]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5,6,7]
; AVX2-ONLY-NEXT: vmovaps 384(%r8), %xmm2
@@ -11994,7 +11989,7 @@ define void @store_i64_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovddup {{.*#+}} xmm1 = mem[0,0]
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm12[1]
; AVX2-ONLY-NEXT: vbroadcastsd 424(%rcx), %ymm4
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm4[4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm4[4,5,6,7]
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -12027,7 +12022,7 @@ define void @store_i64_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovaps 448(%rdx), %ymm8
; AVX2-ONLY-NEXT: vbroadcastsd 456(%rcx), %ymm3
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm8[0,1,2,3],ymm3[4,5],ymm8[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm8[0,1,2,3],ymm3[4,5,6,7]
; AVX2-ONLY-NEXT: vmovddup {{.*#+}} xmm4 = mem[0,0]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm4[0,1],ymm3[2,3,4,5,6,7]
; AVX2-ONLY-NEXT: vinsertf128 $1, 448(%r8), %ymm0, %ymm4
@@ -12036,7 +12031,7 @@ define void @store_i64_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovaps 448(%rsi), %ymm3
; AVX2-ONLY-NEXT: vmovddup {{.*#+}} xmm4 = mem[0,0]
; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm5 = ymm1[0],ymm3[0],ymm1[2],ymm3[2]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm4[0,1],ymm5[2,3,4,5,6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm4[0,1,2,3],ymm5[4,5,6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm1 = ymm1[1],ymm3[1],ymm1[3],ymm3[3]
; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,3,3]
@@ -12056,14 +12051,14 @@ define void @store_i64_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovddup {{.*#+}} xmm3 = mem[0,0]
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm3 = xmm3[1],xmm10[1]
; AVX2-ONLY-NEXT: vbroadcastsd 488(%rcx), %ymm4
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm4[4,5],ymm3[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm4[4,5,6,7]
; AVX2-ONLY-NEXT: vinsertf128 $1, 480(%r8), %ymm0, %ymm4
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3,4,5],ymm4[6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovaps 480(%rsi), %ymm3
; AVX2-ONLY-NEXT: vmovddup {{.*#+}} xmm4 = mem[0,0]
; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm5 = ymm1[0],ymm3[0],ymm1[2],ymm3[2]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm4[0,1],ymm5[2,3,4,5,6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm4[0,1,2,3],ymm5[4,5,6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm1 = ymm1[1],ymm3[1],ymm1[3],ymm3[3]
; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,3,3]
@@ -12268,7 +12263,7 @@ define void @store_i64_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vbroadcastsd 440(%rcx), %ymm2
; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm2[0,1],mem[2,3]
; AVX2-ONLY-NEXT: vbroadcastsd 440(%r9), %ymm3
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm3[4,5],ymm2[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm3[4,5,6,7]
; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm2 = ymm8[0],mem[0],ymm8[2],mem[2]
; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm2 = ymm2[2,3],ymm13[2,3]
; AVX2-ONLY-NEXT: vbroadcastsd 464(%r9), %ymm3
@@ -12277,7 +12272,7 @@ define void @store_i64_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vbroadcastsd 472(%rcx), %ymm2
; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm2[0,1],mem[2,3]
; AVX2-ONLY-NEXT: vbroadcastsd 472(%r9), %ymm3
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5],ymm2[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7]
; AVX2-ONLY-NEXT: vmovaps 448(%rax), %ymm3
; AVX2-ONLY-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm1 # 32-byte Folded Reload
; AVX2-ONLY-NEXT: # ymm1 = mem[0,1],ymm3[2,3],mem[4,5,6,7]
@@ -12298,7 +12293,7 @@ define void @store_i64_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vbroadcastsd 504(%rcx), %ymm2
; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm2[0,1],mem[2,3]
; AVX2-ONLY-NEXT: vbroadcastsd 504(%r9), %ymm3
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5],ymm2[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7]
; AVX2-ONLY-NEXT: vmovaps 480(%rax), %ymm8
; AVX2-ONLY-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm1 # 32-byte Folded Reload
; AVX2-ONLY-NEXT: # ymm1 = mem[0,1],ymm8[2,3],mem[4,5,6,7]
@@ -13092,7 +13087,7 @@ define void @store_i64_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX512F-ONLY-SLOW-NEXT: movb $6, %sil
; AVX512F-ONLY-SLOW-NEXT: kmovw %esi, %k3
; AVX512F-ONLY-SLOW-NEXT: vpbroadcastq 456(%rcx), %ymm14
-; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm14 = mem[0,1,2,3],ymm14[4,5],mem[6,7]
+; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm14 = mem[0,1,2,3],ymm14[4,5,6,7]
; AVX512F-ONLY-SLOW-NEXT: vinserti64x4 $0, %ymm14, %zmm0, %zmm1 {%k3}
; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 {{.*#+}} zmm14 = <0,1,2,9,u,u,6,7>
; AVX512F-ONLY-SLOW-NEXT: vpermi2q %zmm3, %zmm1, %zmm14
@@ -13257,29 +13252,29 @@ define void @store_i64_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX512F-ONLY-SLOW-NEXT: vinserti32x4 $3, 384(%rax), %zmm9, %zmm0
; AVX512F-ONLY-SLOW-NEXT: vmovdqa64 %zmm0, %zmm22 {%k2}
; AVX512F-ONLY-SLOW-NEXT: vpbroadcastq 8(%rcx), %ymm0
-; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = mem[0,1,2,3],ymm0[4,5],mem[6,7]
+; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = mem[0,1,2,3],ymm0[4,5,6,7]
; AVX512F-ONLY-SLOW-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
; AVX512F-ONLY-SLOW-NEXT: vinserti64x4 $0, %ymm0, %zmm0, %zmm1 {%k3}
; AVX512F-ONLY-SLOW-NEXT: vpbroadcastq 72(%rcx), %ymm0
-; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = mem[0,1,2,3],ymm0[4,5],mem[6,7]
+; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = mem[0,1,2,3],ymm0[4,5,6,7]
; AVX512F-ONLY-SLOW-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm3 # 64-byte Reload
; AVX512F-ONLY-SLOW-NEXT: vinserti64x4 $0, %ymm0, %zmm0, %zmm3 {%k3}
; AVX512F-ONLY-SLOW-NEXT: vpbroadcastq 136(%rcx), %ymm0
-; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = mem[0,1,2,3],ymm0[4,5],mem[6,7]
+; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = mem[0,1,2,3],ymm0[4,5,6,7]
; AVX512F-ONLY-SLOW-NEXT: vmovdqu64 (%rsp), %zmm4 # 64-byte Reload
; AVX512F-ONLY-SLOW-NEXT: vinserti64x4 $0, %ymm0, %zmm0, %zmm4 {%k3}
; AVX512F-ONLY-SLOW-NEXT: vpbroadcastq 200(%rcx), %ymm0
-; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = mem[0,1,2,3],ymm0[4,5],mem[6,7]
+; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = mem[0,1,2,3],ymm0[4,5,6,7]
; AVX512F-ONLY-SLOW-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
; AVX512F-ONLY-SLOW-NEXT: vinserti64x4 $0, %ymm0, %zmm0, %zmm2 {%k3}
; AVX512F-ONLY-SLOW-NEXT: vpbroadcastq 264(%rcx), %ymm0
-; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = mem[0,1,2,3],ymm0[4,5],mem[6,7]
+; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = mem[0,1,2,3],ymm0[4,5,6,7]
; AVX512F-ONLY-SLOW-NEXT: vinserti64x4 $0, %ymm0, %zmm0, %zmm31 {%k3}
; AVX512F-ONLY-SLOW-NEXT: vpbroadcastq 328(%rcx), %ymm0
-; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = mem[0,1,2,3],ymm0[4,5],mem[6,7]
+; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = mem[0,1,2,3],ymm0[4,5,6,7]
; AVX512F-ONLY-SLOW-NEXT: vinserti64x4 $0, %ymm0, %zmm0, %zmm28 {%k3}
; AVX512F-ONLY-SLOW-NEXT: vpbroadcastq 392(%rcx), %ymm0
-; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = mem[0,1,2,3],ymm0[4,5],mem[6,7]
+; AVX512F-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = mem[0,1,2,3],ymm0[4,5,6,7]
; AVX512F-ONLY-SLOW-NEXT: vinserti64x4 $0, %ymm0, %zmm0, %zmm23 {%k3}
; AVX512F-ONLY-SLOW-NEXT: movb $56, %cl
; AVX512F-ONLY-SLOW-NEXT: kmovw %ecx, %k2
@@ -14063,7 +14058,7 @@ define void @store_i64_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX512F-ONLY-FAST-NEXT: movb $6, %sil
; AVX512F-ONLY-FAST-NEXT: kmovw %esi, %k5
; AVX512F-ONLY-FAST-NEXT: vpbroadcastq 456(%rcx), %ymm1
-; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm1 = mem[0,1,2,3],ymm1[4,5],mem[6,7]
+; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm1 = mem[0,1,2,3],ymm1[4,5,6,7]
; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $0, %ymm1, %zmm0, %zmm15 {%k5}
; AVX512F-ONLY-FAST-NEXT: vmovdqa64 {{.*#+}} zmm1 = <0,1,2,9,u,u,6,7>
; AVX512F-ONLY-FAST-NEXT: vpermi2q %zmm2, %zmm15, %zmm1
@@ -14100,29 +14095,29 @@ define void @store_i64_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX512F-ONLY-FAST-NEXT: vpermi2q %zmm11, %zmm0, %zmm27
; AVX512F-ONLY-FAST-NEXT: vpermt2q %zmm11, %zmm2, %zmm6
; AVX512F-ONLY-FAST-NEXT: vpbroadcastq 8(%rcx), %ymm0
-; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm0 = mem[0,1,2,3],ymm0[4,5],mem[6,7]
+; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm0 = mem[0,1,2,3],ymm0[4,5,6,7]
; AVX512F-ONLY-FAST-NEXT: vmovdqu64 (%rsp), %zmm14 # 64-byte Reload
; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $0, %ymm0, %zmm0, %zmm14 {%k5}
; AVX512F-ONLY-FAST-NEXT: vpbroadcastq 72(%rcx), %ymm0
-; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm0 = mem[0,1,2,3],ymm0[4,5],mem[6,7]
+; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm0 = mem[0,1,2,3],ymm0[4,5,6,7]
; AVX512F-ONLY-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm11 # 64-byte Reload
; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $0, %ymm0, %zmm0, %zmm11 {%k5}
; AVX512F-ONLY-FAST-NEXT: vpbroadcastq 136(%rcx), %ymm0
-; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm0 = mem[0,1,2,3],ymm0[4,5],mem[6,7]
+; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm0 = mem[0,1,2,3],ymm0[4,5,6,7]
; AVX512F-ONLY-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm17 # 64-byte Reload
; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $0, %ymm0, %zmm0, %zmm17 {%k5}
; AVX512F-ONLY-FAST-NEXT: vpbroadcastq 200(%rcx), %ymm0
-; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm0 = mem[0,1,2,3],ymm0[4,5],mem[6,7]
+; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm0 = mem[0,1,2,3],ymm0[4,5,6,7]
; AVX512F-ONLY-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm5 # 64-byte Reload
; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $0, %ymm0, %zmm0, %zmm5 {%k5}
; AVX512F-ONLY-FAST-NEXT: vpbroadcastq 264(%rcx), %ymm0
-; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm0 = mem[0,1,2,3],ymm0[4,5],mem[6,7]
+; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm0 = mem[0,1,2,3],ymm0[4,5,6,7]
; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $0, %ymm0, %zmm0, %zmm30 {%k5}
; AVX512F-ONLY-FAST-NEXT: vpbroadcastq 328(%rcx), %ymm0
-; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm0 = mem[0,1,2,3],ymm0[4,5],mem[6,7]
+; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm0 = mem[0,1,2,3],ymm0[4,5,6,7]
; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $0, %ymm0, %zmm0, %zmm24 {%k5}
; AVX512F-ONLY-FAST-NEXT: vpbroadcastq 392(%rcx), %ymm0
-; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm0 = mem[0,1,2,3],ymm0[4,5],mem[6,7]
+; AVX512F-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm0 = mem[0,1,2,3],ymm0[4,5,6,7]
; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $0, %ymm0, %zmm0, %zmm20 {%k5}
; AVX512F-ONLY-FAST-NEXT: vmovdqa64 448(%rax), %zmm2
; AVX512F-ONLY-FAST-NEXT: vmovdqa64 {{.*#+}} zmm0 = [0,1,2,3,10,5,6,7]
@@ -14994,7 +14989,7 @@ define void @store_i64_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX512DQ-SLOW-NEXT: movb $6, %sil
; AVX512DQ-SLOW-NEXT: kmovw %esi, %k3
; AVX512DQ-SLOW-NEXT: vpbroadcastq 456(%rcx), %ymm13
-; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm13 = mem[0,1,2,3],ymm13[4,5],mem[6,7]
+; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm13 = mem[0,1,2,3],ymm13[4,5,6,7]
; AVX512DQ-SLOW-NEXT: vinserti64x4 $0, %ymm13, %zmm0, %zmm1 {%k3}
; AVX512DQ-SLOW-NEXT: vmovdqa64 {{.*#+}} zmm13 = <0,1,2,9,u,u,6,7>
; AVX512DQ-SLOW-NEXT: vpermi2q %zmm3, %zmm1, %zmm13
@@ -15153,29 +15148,29 @@ define void @store_i64_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX512DQ-SLOW-NEXT: vinserti64x2 $3, 320(%rax), %zmm1, %zmm26 {%k2}
; AVX512DQ-SLOW-NEXT: vinserti64x2 $3, 384(%rax), %zmm9, %zmm22 {%k2}
; AVX512DQ-SLOW-NEXT: vpbroadcastq 8(%rcx), %ymm1
-; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = mem[0,1,2,3],ymm1[4,5],mem[6,7]
+; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = mem[0,1,2,3],ymm1[4,5,6,7]
; AVX512DQ-SLOW-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
; AVX512DQ-SLOW-NEXT: vinserti64x4 $0, %ymm1, %zmm0, %zmm2 {%k3}
; AVX512DQ-SLOW-NEXT: vpbroadcastq 72(%rcx), %ymm1
-; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = mem[0,1,2,3],ymm1[4,5],mem[6,7]
+; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = mem[0,1,2,3],ymm1[4,5,6,7]
; AVX512DQ-SLOW-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm4 # 64-byte Reload
; AVX512DQ-SLOW-NEXT: vinserti64x4 $0, %ymm1, %zmm0, %zmm4 {%k3}
; AVX512DQ-SLOW-NEXT: vpbroadcastq 136(%rcx), %ymm1
-; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = mem[0,1,2,3],ymm1[4,5],mem[6,7]
+; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = mem[0,1,2,3],ymm1[4,5,6,7]
; AVX512DQ-SLOW-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm6 # 64-byte Reload
; AVX512DQ-SLOW-NEXT: vinserti64x4 $0, %ymm1, %zmm0, %zmm6 {%k3}
; AVX512DQ-SLOW-NEXT: vpbroadcastq 200(%rcx), %ymm1
-; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = mem[0,1,2,3],ymm1[4,5],mem[6,7]
+; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = mem[0,1,2,3],ymm1[4,5,6,7]
; AVX512DQ-SLOW-NEXT: vmovdqu64 (%rsp), %zmm3 # 64-byte Reload
; AVX512DQ-SLOW-NEXT: vinserti64x4 $0, %ymm1, %zmm0, %zmm3 {%k3}
; AVX512DQ-SLOW-NEXT: vpbroadcastq 264(%rcx), %ymm1
-; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = mem[0,1,2,3],ymm1[4,5],mem[6,7]
+; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = mem[0,1,2,3],ymm1[4,5,6,7]
; AVX512DQ-SLOW-NEXT: vinserti64x4 $0, %ymm1, %zmm0, %zmm31 {%k3}
; AVX512DQ-SLOW-NEXT: vpbroadcastq 328(%rcx), %ymm1
-; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = mem[0,1,2,3],ymm1[4,5],mem[6,7]
+; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = mem[0,1,2,3],ymm1[4,5,6,7]
; AVX512DQ-SLOW-NEXT: vinserti64x4 $0, %ymm1, %zmm0, %zmm30 {%k3}
; AVX512DQ-SLOW-NEXT: vpbroadcastq 392(%rcx), %ymm1
-; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = mem[0,1,2,3],ymm1[4,5],mem[6,7]
+; AVX512DQ-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = mem[0,1,2,3],ymm1[4,5,6,7]
; AVX512DQ-SLOW-NEXT: vinserti64x4 $0, %ymm1, %zmm0, %zmm24 {%k3}
; AVX512DQ-SLOW-NEXT: movb $56, %cl
; AVX512DQ-SLOW-NEXT: kmovw %ecx, %k2
@@ -15961,7 +15956,7 @@ define void @store_i64_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX512DQ-FAST-NEXT: movb $6, %sil
; AVX512DQ-FAST-NEXT: kmovw %esi, %k5
; AVX512DQ-FAST-NEXT: vpbroadcastq 456(%rcx), %ymm1
-; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm1 = mem[0,1,2,3],ymm1[4,5],mem[6,7]
+; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm1 = mem[0,1,2,3],ymm1[4,5,6,7]
; AVX512DQ-FAST-NEXT: vinserti64x4 $0, %ymm1, %zmm0, %zmm17 {%k5}
; AVX512DQ-FAST-NEXT: vmovdqa64 {{.*#+}} zmm1 = <0,1,2,9,u,u,6,7>
; AVX512DQ-FAST-NEXT: vpermi2q %zmm2, %zmm17, %zmm1
@@ -15998,28 +15993,28 @@ define void @store_i64_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX512DQ-FAST-NEXT: vpermi2q %zmm6, %zmm0, %zmm28
; AVX512DQ-FAST-NEXT: vpermt2q %zmm6, %zmm2, %zmm11
; AVX512DQ-FAST-NEXT: vpbroadcastq 8(%rcx), %ymm0
-; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm0 = mem[0,1,2,3],ymm0[4,5],mem[6,7]
+; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm0 = mem[0,1,2,3],ymm0[4,5,6,7]
; AVX512DQ-FAST-NEXT: vmovdqu64 (%rsp), %zmm6 # 64-byte Reload
; AVX512DQ-FAST-NEXT: vinserti64x4 $0, %ymm0, %zmm0, %zmm6 {%k5}
; AVX512DQ-FAST-NEXT: vpbroadcastq 72(%rcx), %ymm0
-; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm0 = mem[0,1,2,3],ymm0[4,5],mem[6,7]
+; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm0 = mem[0,1,2,3],ymm0[4,5,6,7]
; AVX512DQ-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm5 # 64-byte Reload
; AVX512DQ-FAST-NEXT: vinserti64x4 $0, %ymm0, %zmm0, %zmm5 {%k5}
; AVX512DQ-FAST-NEXT: vpbroadcastq 136(%rcx), %ymm0
-; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm0 = mem[0,1,2,3],ymm0[4,5],mem[6,7]
+; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm0 = mem[0,1,2,3],ymm0[4,5,6,7]
; AVX512DQ-FAST-NEXT: vinserti64x4 $0, %ymm0, %zmm0, %zmm15 {%k5}
; AVX512DQ-FAST-NEXT: vpbroadcastq 200(%rcx), %ymm0
-; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm0 = mem[0,1,2,3],ymm0[4,5],mem[6,7]
+; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm0 = mem[0,1,2,3],ymm0[4,5,6,7]
; AVX512DQ-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm4 # 64-byte Reload
; AVX512DQ-FAST-NEXT: vinserti64x4 $0, %ymm0, %zmm0, %zmm4 {%k5}
; AVX512DQ-FAST-NEXT: vpbroadcastq 264(%rcx), %ymm0
-; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm0 = mem[0,1,2,3],ymm0[4,5],mem[6,7]
+; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm0 = mem[0,1,2,3],ymm0[4,5,6,7]
; AVX512DQ-FAST-NEXT: vinserti64x4 $0, %ymm0, %zmm0, %zmm30 {%k5}
; AVX512DQ-FAST-NEXT: vpbroadcastq 328(%rcx), %ymm0
-; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm0 = mem[0,1,2,3],ymm0[4,5],mem[6,7]
+; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm0 = mem[0,1,2,3],ymm0[4,5,6,7]
; AVX512DQ-FAST-NEXT: vinserti64x4 $0, %ymm0, %zmm0, %zmm25 {%k5}
; AVX512DQ-FAST-NEXT: vpbroadcastq 392(%rcx), %ymm0
-; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm0 = mem[0,1,2,3],ymm0[4,5],mem[6,7]
+; AVX512DQ-FAST-NEXT: vpblendd {{.*#+}} ymm0 = mem[0,1,2,3],ymm0[4,5,6,7]
; AVX512DQ-FAST-NEXT: vinserti64x4 $0, %ymm0, %zmm0, %zmm20 {%k5}
; AVX512DQ-FAST-NEXT: vmovdqa64 448(%rax), %zmm2
; AVX512DQ-FAST-NEXT: vmovdqa64 {{.*#+}} zmm0 = [0,1,2,3,10,5,6,7]
@@ -16883,7 +16878,7 @@ define void @store_i64_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX512BW-ONLY-SLOW-NEXT: movb $6, %sil
; AVX512BW-ONLY-SLOW-NEXT: kmovd %esi, %k3
; AVX512BW-ONLY-SLOW-NEXT: vpbroadcastq 456(%rcx), %ymm14
-; AVX512BW-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm14 = mem[0,1,2,3],ymm14[4,5],mem[6,7]
+; AVX512BW-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm14 = mem[0,1,2,3],ymm14[4,5,6,7]
; AVX512BW-ONLY-SLOW-NEXT: vinserti64x4 $0, %ymm14, %zmm0, %zmm1 {%k3}
; AVX512BW-ONLY-SLOW-NEXT: vmovdqa64 {{.*#+}} zmm14 = <0,1,2,9,u,u,6,7>
; AVX512BW-ONLY-SLOW-NEXT: vpermi2q %zmm3, %zmm1, %zmm14
@@ -17048,29 +17043,29 @@ define void @store_i64_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX512BW-ONLY-SLOW-NEXT: vinserti32x4 $3, 384(%rax), %zmm9, %zmm0
; AVX512BW-ONLY-SLOW-NEXT: vmovdqa64 %zmm0, %zmm22 {%k2}
; AVX512BW-ONLY-SLOW-NEXT: vpbroadcastq 8(%rcx), %ymm0
-; AVX512BW-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = mem[0,1,2,3],ymm0[4,5],mem[6,7]
+; AVX512BW-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = mem[0,1,2,3],ymm0[4,5,6,7]
; AVX512BW-ONLY-SLOW-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
; AVX512BW-ONLY-SLOW-NEXT: vinserti64x4 $0, %ymm0, %zmm0, %zmm1 {%k3}
; AVX512BW-ONLY-SLOW-NEXT: vpbroadcastq 72(%rcx), %ymm0
-; AVX512BW-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = mem[0,1,2,3],ymm0[4,5],mem[6,7]
+; AVX512BW-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = mem[0,1,2,3],ymm0[4,5,6,7]
; AVX512BW-ONLY-SLOW-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm3 # 64-byte Reload
; AVX512BW-ONLY-SLOW-NEXT: vinserti64x4 $0, %ymm0, %zmm0, %zmm3 {%k3}
; AVX512BW-ONLY-SLOW-NEXT: vpbroadcastq 136(%rcx), %ymm0
-; AVX512BW-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = mem[0,1,2,3],ymm0[4,5],mem[6,7]
+; AVX512BW-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = mem[0,1,2,3],ymm0[4,5,6,7]
; AVX512BW-ONLY-SLOW-NEXT: vmovdqu64 (%rsp), %zmm4 # 64-byte Reload
; AVX512BW-ONLY-SLOW-NEXT: vinserti64x4 $0, %ymm0, %zmm0, %zmm4 {%k3}
; AVX512BW-ONLY-SLOW-NEXT: vpbroadcastq 200(%rcx), %ymm0
-; AVX512BW-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = mem[0,1,2,3],ymm0[4,5],mem[6,7]
+; AVX512BW-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = mem[0,1,2,3],ymm0[4,5,6,7]
; AVX512BW-ONLY-SLOW-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
; AVX512BW-ONLY-SLOW-NEXT: vinserti64x4 $0, %ymm0, %zmm0, %zmm2 {%k3}
; AVX512BW-ONLY-SLOW-NEXT: vpbroadcastq 264(%rcx), %ymm0
-; AVX512BW-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = mem[0,1,2,3],ymm0[4,5],mem[6,7]
+; AVX512BW-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = mem[0,1,2,3],ymm0[4,5,6,7]
; AVX512BW-ONLY-SLOW-NEXT: vinserti64x4 $0, %ymm0, %zmm0, %zmm31 {%k3}
; AVX512BW-ONLY-SLOW-NEXT: vpbroadcastq 328(%rcx), %ymm0
-; AVX512BW-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = mem[0,1,2,3],ymm0[4,5],mem[6,7]
+; AVX512BW-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = mem[0,1,2,3],ymm0[4,5,6,7]
; AVX512BW-ONLY-SLOW-NEXT: vinserti64x4 $0, %ymm0, %zmm0, %zmm28 {%k3}
; AVX512BW-ONLY-SLOW-NEXT: vpbroadcastq 392(%rcx), %ymm0
-; AVX512BW-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = mem[0,1,2,3],ymm0[4,5],mem[6,7]
+; AVX512BW-ONLY-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = mem[0,1,2,3],ymm0[4,5,6,7]
; AVX512BW-ONLY-SLOW-NEXT: vinserti64x4 $0, %ymm0, %zmm0, %zmm23 {%k3}
; AVX512BW-ONLY-SLOW-NEXT: movb $56, %cl
; AVX512BW-ONLY-SLOW-NEXT: kmovd %ecx, %k2
@@ -17854,7 +17849,7 @@ define void @store_i64_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX512BW-ONLY-FAST-NEXT: movb $6, %sil
; AVX512BW-ONLY-FAST-NEXT: kmovd %esi, %k5
; AVX512BW-ONLY-FAST-NEXT: vpbroadcastq 456(%rcx), %ymm1
-; AVX512BW-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm1 = mem[0,1,2,3],ymm1[4,5],mem[6,7]
+; AVX512BW-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm1 = mem[0,1,2,3],ymm1[4,5,6,7]
; AVX512BW-ONLY-FAST-NEXT: vinserti64x4 $0, %ymm1, %zmm0, %zmm15 {%k5}
; AVX512BW-ONLY-FAST-NEXT: vmovdqa64 {{.*#+}} zmm1 = <0,1,2,9,u,u,6,7>
; AVX512BW-ONLY-FAST-NEXT: vpermi2q %zmm2, %zmm15, %zmm1
@@ -17891,29 +17886,29 @@ define void @store_i64_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX512BW-ONLY-FAST-NEXT: vpermi2q %zmm11, %zmm0, %zmm27
; AVX512BW-ONLY-FAST-NEXT: vpermt2q %zmm11, %zmm2, %zmm6
; AVX512BW-ONLY-FAST-NEXT: vpbroadcastq 8(%rcx), %ymm0
-; AVX512BW-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm0 = mem[0,1,2,3],ymm0[4,5],mem[6,7]
+; AVX512BW-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm0 = mem[0,1,2,3],ymm0[4,5,6,7]
; AVX512BW-ONLY-FAST-NEXT: vmovdqu64 (%rsp), %zmm14 # 64-byte Reload
; AVX512BW-ONLY-FAST-NEXT: vinserti64x4 $0, %ymm0, %zmm0, %zmm14 {%k5}
; AVX512BW-ONLY-FAST-NEXT: vpbroadcastq 72(%rcx), %ymm0
-; AVX512BW-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm0 = mem[0,1,2,3],ymm0[4,5],mem[6,7]
+; AVX512BW-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm0 = mem[0,1,2,3],ymm0[4,5,6,7]
; AVX512BW-ONLY-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm11 # 64-byte Reload
; AVX512BW-ONLY-FAST-NEXT: vinserti64x4 $0, %ymm0, %zmm0, %zmm11 {%k5}
; AVX512BW-ONLY-FAST-NEXT: vpbroadcastq 136(%rcx), %ymm0
-; AVX512BW-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm0 = mem[0,1,2,3],ymm0[4,5],mem[6,7]
+; AVX512BW-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm0 = mem[0,1,2,3],ymm0[4,5,6,7]
; AVX512BW-ONLY-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm17 # 64-byte Reload
; AVX512BW-ONLY-FAST-NEXT: vinserti64x4 $0, %ymm0, %zmm0, %zmm17 {%k5}
; AVX512BW-ONLY-FAST-NEXT: vpbroadcastq 200(%rcx), %ymm0
-; AVX512BW-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm0 = mem[0,1,2,3],ymm0[4,5],mem[6,7]
+; AVX512BW-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm0 = mem[0,1,2,3],ymm0[4,5,6,7]
; AVX512BW-ONLY-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm5 # 64-byte Reload
; AVX512BW-ONLY-FAST-NEXT: vinserti64x4 $0, %ymm0, %zmm0, %zmm5 {%k5}
; AVX512BW-ONLY-FAST-NEXT: vpbroadcastq 264(%rcx), %ymm0
-; AVX512BW-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm0 = mem[0,1,2,3],ymm0[4,5],mem[6,7]
+; AVX512BW-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm0 = mem[0,1,2,3],ymm0[4,5,6,7]
; AVX512BW-ONLY-FAST-NEXT: vinserti64x4 $0, %ymm0, %zmm0, %zmm30 {%k5}
; AVX512BW-ONLY-FAST-NEXT: vpbroadcastq 328(%rcx), %ymm0
-; AVX512BW-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm0 = mem[0,1,2,3],ymm0[4,5],mem[6,7]
+; AVX512BW-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm0 = mem[0,1,2,3],ymm0[4,5,6,7]
; AVX512BW-ONLY-FAST-NEXT: vinserti64x4 $0, %ymm0, %zmm0, %zmm24 {%k5}
; AVX512BW-ONLY-FAST-NEXT: vpbroadcastq 392(%rcx), %ymm0
-; AVX512BW-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm0 = mem[0,1,2,3],ymm0[4,5],mem[6,7]
+; AVX512BW-ONLY-FAST-NEXT: vpblendd {{.*#+}} ymm0 = mem[0,1,2,3],ymm0[4,5,6,7]
; AVX512BW-ONLY-FAST-NEXT: vinserti64x4 $0, %ymm0, %zmm0, %zmm20 {%k5}
; AVX512BW-ONLY-FAST-NEXT: vmovdqa64 448(%rax), %zmm2
; AVX512BW-ONLY-FAST-NEXT: vmovdqa64 {{.*#+}} zmm0 = [0,1,2,3,10,5,6,7]
@@ -18785,7 +18780,7 @@ define void @store_i64_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX512DQBW-SLOW-NEXT: movb $6, %sil
; AVX512DQBW-SLOW-NEXT: kmovd %esi, %k3
; AVX512DQBW-SLOW-NEXT: vpbroadcastq 456(%rcx), %ymm13
-; AVX512DQBW-SLOW-NEXT: vpblendd {{.*#+}} ymm13 = mem[0,1,2,3],ymm13[4,5],mem[6,7]
+; AVX512DQBW-SLOW-NEXT: vpblendd {{.*#+}} ymm13 = mem[0,1,2,3],ymm13[4,5,6,7]
; AVX512DQBW-SLOW-NEXT: vinserti64x4 $0, %ymm13, %zmm0, %zmm1 {%k3}
; AVX512DQBW-SLOW-NEXT: vmovdqa64 {{.*#+}} zmm13 = <0,1,2,9,u,u,6,7>
; AVX512DQBW-SLOW-NEXT: vpermi2q %zmm3, %zmm1, %zmm13
@@ -18944,29 +18939,29 @@ define void @store_i64_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX512DQBW-SLOW-NEXT: vinserti64x2 $3, 320(%rax), %zmm1, %zmm26 {%k2}
; AVX512DQBW-SLOW-NEXT: vinserti64x2 $3, 384(%rax), %zmm9, %zmm22 {%k2}
; AVX512DQBW-SLOW-NEXT: vpbroadcastq 8(%rcx), %ymm1
-; AVX512DQBW-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = mem[0,1,2,3],ymm1[4,5],mem[6,7]
+; AVX512DQBW-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = mem[0,1,2,3],ymm1[4,5,6,7]
; AVX512DQBW-SLOW-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
; AVX512DQBW-SLOW-NEXT: vinserti64x4 $0, %ymm1, %zmm0, %zmm2 {%k3}
; AVX512DQBW-SLOW-NEXT: vpbroadcastq 72(%rcx), %ymm1
-; AVX512DQBW-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = mem[0,1,2,3],ymm1[4,5],mem[6,7]
+; AVX512DQBW-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = mem[0,1,2,3],ymm1[4,5,6,7]
; AVX512DQBW-SLOW-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm4 # 64-byte Reload
; AVX512DQBW-SLOW-NEXT: vinserti64x4 $0, %ymm1, %zmm0, %zmm4 {%k3}
; AVX512DQBW-SLOW-NEXT: vpbroadcastq 136(%rcx), %ymm1
-; AVX512DQBW-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = mem[0,1,2,3],ymm1[4,5],mem[6,7]
+; AVX512DQBW-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = mem[0,1,2,3],ymm1[4,5,6,7]
; AVX512DQBW-SLOW-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm6 # 64-byte Reload
; AVX512DQBW-SLOW-NEXT: vinserti64x4 $0, %ymm1, %zmm0, %zmm6 {%k3}
; AVX512DQBW-SLOW-NEXT: vpbroadcastq 200(%rcx), %ymm1
-; AVX512DQBW-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = mem[0,1,2,3],ymm1[4,5],mem[6,7]
+; AVX512DQBW-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = mem[0,1,2,3],ymm1[4,5,6,7]
; AVX512DQBW-SLOW-NEXT: vmovdqu64 (%rsp), %zmm3 # 64-byte Reload
; AVX512DQBW-SLOW-NEXT: vinserti64x4 $0, %ymm1, %zmm0, %zmm3 {%k3}
; AVX512DQBW-SLOW-NEXT: vpbroadcastq 264(%rcx), %ymm1
-; AVX512DQBW-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = mem[0,1,2,3],ymm1[4,5],mem[6,7]
+; AVX512DQBW-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = mem[0,1,2,3],ymm1[4,5,6,7]
; AVX512DQBW-SLOW-NEXT: vinserti64x4 $0, %ymm1, %zmm0, %zmm31 {%k3}
; AVX512DQBW-SLOW-NEXT: vpbroadcastq 328(%rcx), %ymm1
-; AVX512DQBW-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = mem[0,1,2,3],ymm1[4,5],mem[6,7]
+; AVX512DQBW-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = mem[0,1,2,3],ymm1[4,5,6,7]
; AVX512DQBW-SLOW-NEXT: vinserti64x4 $0, %ymm1, %zmm0, %zmm30 {%k3}
; AVX512DQBW-SLOW-NEXT: vpbroadcastq 392(%rcx), %ymm1
-; AVX512DQBW-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = mem[0,1,2,3],ymm1[4,5],mem[6,7]
+; AVX512DQBW-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = mem[0,1,2,3],ymm1[4,5,6,7]
; AVX512DQBW-SLOW-NEXT: vinserti64x4 $0, %ymm1, %zmm0, %zmm24 {%k3}
; AVX512DQBW-SLOW-NEXT: movb $56, %cl
; AVX512DQBW-SLOW-NEXT: kmovd %ecx, %k2
@@ -19752,7 +19747,7 @@ define void @store_i64_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX512DQBW-FAST-NEXT: movb $6, %sil
; AVX512DQBW-FAST-NEXT: kmovd %esi, %k5
; AVX512DQBW-FAST-NEXT: vpbroadcastq 456(%rcx), %ymm1
-; AVX512DQBW-FAST-NEXT: vpblendd {{.*#+}} ymm1 = mem[0,1,2,3],ymm1[4,5],mem[6,7]
+; AVX512DQBW-FAST-NEXT: vpblendd {{.*#+}} ymm1 = mem[0,1,2,3],ymm1[4,5,6,7]
; AVX512DQBW-FAST-NEXT: vinserti64x4 $0, %ymm1, %zmm0, %zmm17 {%k5}
; AVX512DQBW-FAST-NEXT: vmovdqa64 {{.*#+}} zmm1 = <0,1,2,9,u,u,6,7>
; AVX512DQBW-FAST-NEXT: vpermi2q %zmm2, %zmm17, %zmm1
@@ -19789,28 +19784,28 @@ define void @store_i64_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX512DQBW-FAST-NEXT: vpermi2q %zmm6, %zmm0, %zmm28
; AVX512DQBW-FAST-NEXT: vpermt2q %zmm6, %zmm2, %zmm11
; AVX512DQBW-FAST-NEXT: vpbroadcastq 8(%rcx), %ymm0
-; AVX512DQBW-FAST-NEXT: vpblendd {{.*#+}} ymm0 = mem[0,1,2,3],ymm0[4,5],mem[6,7]
+; AVX512DQBW-FAST-NEXT: vpblendd {{.*#+}} ymm0 = mem[0,1,2,3],ymm0[4,5,6,7]
; AVX512DQBW-FAST-NEXT: vmovdqu64 (%rsp), %zmm6 # 64-byte Reload
; AVX512DQBW-FAST-NEXT: vinserti64x4 $0, %ymm0, %zmm0, %zmm6 {%k5}
; AVX512DQBW-FAST-NEXT: vpbroadcastq 72(%rcx), %ymm0
-; AVX512DQBW-FAST-NEXT: vpblendd {{.*#+}} ymm0 = mem[0,1,2,3],ymm0[4,5],mem[6,7]
+; AVX512DQBW-FAST-NEXT: vpblendd {{.*#+}} ymm0 = mem[0,1,2,3],ymm0[4,5,6,7]
; AVX512DQBW-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm5 # 64-byte Reload
; AVX512DQBW-FAST-NEXT: vinserti64x4 $0, %ymm0, %zmm0, %zmm5 {%k5}
; AVX512DQBW-FAST-NEXT: vpbroadcastq 136(%rcx), %ymm0
-; AVX512DQBW-FAST-NEXT: vpblendd {{.*#+}} ymm0 = mem[0,1,2,3],ymm0[4,5],mem[6,7]
+; AVX512DQBW-FAST-NEXT: vpblendd {{.*#+}} ymm0 = mem[0,1,2,3],ymm0[4,5,6,7]
; AVX512DQBW-FAST-NEXT: vinserti64x4 $0, %ymm0, %zmm0, %zmm15 {%k5}
; AVX512DQBW-FAST-NEXT: vpbroadcastq 200(%rcx), %ymm0
-; AVX512DQBW-FAST-NEXT: vpblendd {{.*#+}} ymm0 = mem[0,1,2,3],ymm0[4,5],mem[6,7]
+; AVX512DQBW-FAST-NEXT: vpblendd {{.*#+}} ymm0 = mem[0,1,2,3],ymm0[4,5,6,7]
; AVX512DQBW-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm4 # 64-byte Reload
; AVX512DQBW-FAST-NEXT: vinserti64x4 $0, %ymm0, %zmm0, %zmm4 {%k5}
; AVX512DQBW-FAST-NEXT: vpbroadcastq 264(%rcx), %ymm0
-; AVX512DQBW-FAST-NEXT: vpblendd {{.*#+}} ymm0 = mem[0,1,2,3],ymm0[4,5],mem[6,7]
+; AVX512DQBW-FAST-NEXT: vpblendd {{.*#+}} ymm0 = mem[0,1,2,3],ymm0[4,5,6,7]
; AVX512DQBW-FAST-NEXT: vinserti64x4 $0, %ymm0, %zmm0, %zmm30 {%k5}
; AVX512DQBW-FAST-NEXT: vpbroadcastq 328(%rcx), %ymm0
-; AVX512DQBW-FAST-NEXT: vpblendd {{.*#+}} ymm0 = mem[0,1,2,3],ymm0[4,5],mem[6,7]
+; AVX512DQBW-FAST-NEXT: vpblendd {{.*#+}} ymm0 = mem[0,1,2,3],ymm0[4,5,6,7]
; AVX512DQBW-FAST-NEXT: vinserti64x4 $0, %ymm0, %zmm0, %zmm25 {%k5}
; AVX512DQBW-FAST-NEXT: vpbroadcastq 392(%rcx), %ymm0
-; AVX512DQBW-FAST-NEXT: vpblendd {{.*#+}} ymm0 = mem[0,1,2,3],ymm0[4,5],mem[6,7]
+; AVX512DQBW-FAST-NEXT: vpblendd {{.*#+}} ymm0 = mem[0,1,2,3],ymm0[4,5,6,7]
; AVX512DQBW-FAST-NEXT: vinserti64x4 $0, %ymm0, %zmm0, %zmm20 {%k5}
; AVX512DQBW-FAST-NEXT: vmovdqa64 448(%rax), %zmm2
; AVX512DQBW-FAST-NEXT: vmovdqa64 {{.*#+}} zmm0 = [0,1,2,3,10,5,6,7]
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-store-i64-stride-8.ll b/llvm/test/CodeGen/X86/vector-interleaved-store-i64-stride-8.ll
index 7e6168dae6e85..bdf4e3069c29e 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-store-i64-stride-8.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-store-i64-stride-8.ll
@@ -544,129 +544,123 @@ define void @store_i64_stride8_vf8(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
;
; AVX1-ONLY-LABEL: store_i64_stride8_vf8:
; AVX1-ONLY: # %bb.0:
-; AVX1-ONLY-NEXT: subq $104, %rsp
-; AVX1-ONLY-NEXT: movq {{[0-9]+}}(%rsp), %rax
+; AVX1-ONLY-NEXT: pushq %rax
; AVX1-ONLY-NEXT: movq {{[0-9]+}}(%rsp), %r10
-; AVX1-ONLY-NEXT: vmovaps (%r9), %xmm10
-; AVX1-ONLY-NEXT: vmovaps 16(%r9), %xmm1
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 32(%r9), %xmm8
-; AVX1-ONLY-NEXT: vmovaps 48(%r9), %xmm7
-; AVX1-ONLY-NEXT: vmovaps (%r8), %xmm12
-; AVX1-ONLY-NEXT: vmovaps 16(%r8), %xmm0
-; AVX1-ONLY-NEXT: vmovups %ymm0, (%rsp) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 32(%r8), %xmm9
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],mem[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vbroadcastsd 16(%rax), %ymm2
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm2[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 16(%rsi), %xmm0
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 32(%rsi), %xmm11
-; AVX1-ONLY-NEXT: vmovaps 16(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 32(%rdi), %xmm13
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm2 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],mem[4,5],ymm2[6,7]
-; AVX1-ONLY-NEXT: vbroadcastsd 16(%rcx), %ymm5
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3,4,5],ymm5[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm5 = xmm9[0],xmm8[0]
-; AVX1-ONLY-NEXT: vmovaps 32(%rax), %xmm14
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm14, %ymm5, %ymm15
-; AVX1-ONLY-NEXT: vinsertf128 $1, 32(%r10), %ymm5, %ymm5
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm0 = ymm5[0],ymm15[1],ymm5[2],ymm15[2]
+; AVX1-ONLY-NEXT: movq {{[0-9]+}}(%rsp), %rax
+; AVX1-ONLY-NEXT: vmovaps 32(%r9), %xmm1
+; AVX1-ONLY-NEXT: vmovaps 32(%r8), %xmm2
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm0 = xmm2[0],xmm1[0]
+; AVX1-ONLY-NEXT: vmovaps (%rax), %xmm5
+; AVX1-ONLY-NEXT: vmovaps 32(%rax), %xmm3
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm4
+; AVX1-ONLY-NEXT: vinsertf128 $1, 32(%r10), %ymm0, %ymm0
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[0],ymm4[1],ymm0[2],ymm4[2]
+; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm1, %ymm1
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm2 = xmm2[2,3,2,3]
+; AVX1-ONLY-NEXT: vbroadcastsd 40(%r10), %ymm3
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm1[2,3],ymm2[4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 32(%rsi), %xmm3
+; AVX1-ONLY-NEXT: vmovaps 32(%rdi), %xmm4
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm4[0],xmm3[0]
+; AVX1-ONLY-NEXT: vmovaps (%rcx), %xmm7
+; AVX1-ONLY-NEXT: vmovaps 32(%rcx), %xmm6
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm6, %ymm2, %ymm8
+; AVX1-ONLY-NEXT: vinsertf128 $1, 32(%rdx), %ymm2, %ymm2
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm0 = ymm2[0],ymm8[1],ymm2[2],ymm8[2]
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm14, %ymm8, %ymm8
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm9 = xmm9[2,3,2,3]
-; AVX1-ONLY-NEXT: vbroadcastsd 40(%r10), %ymm14
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm9 = ymm9[0,1,2,3],ymm14[4,5],ymm9[6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm9[0,1],ymm8[2,3],ymm9[4,5],ymm8[6,7]
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm9 = xmm13[0],xmm11[0]
-; AVX1-ONLY-NEXT: vmovaps 32(%rcx), %xmm14
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm14, %ymm9, %ymm15
-; AVX1-ONLY-NEXT: vinsertf128 $1, 32(%rdx), %ymm9, %ymm9
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm9 = ymm9[0],ymm15[1],ymm9[2],ymm15[2]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm6, %ymm3, %ymm3
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm4 = xmm4[2,3,2,3]
+; AVX1-ONLY-NEXT: vbroadcastsd 40(%rdx), %ymm6
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm6[4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm4[0,1],ymm3[2,3],ymm4[4,5],ymm3[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps (%r9), %xmm6
+; AVX1-ONLY-NEXT: vmovaps (%r8), %xmm8
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm4 = xmm8[0],xmm6[0]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm5, %ymm4, %ymm9
+; AVX1-ONLY-NEXT: vinsertf128 $1, (%r10), %ymm4, %ymm4
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm4 = ymm4[0],ymm9[1],ymm4[2],ymm9[2]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm5, %ymm6, %ymm5
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm6 = xmm8[2,3,2,3]
+; AVX1-ONLY-NEXT: vbroadcastsd 8(%r10), %ymm8
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm8[4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm6[0,1],ymm5[2,3],ymm6[4,5],ymm5[6,7]
+; AVX1-ONLY-NEXT: vmovaps (%rsi), %xmm8
+; AVX1-ONLY-NEXT: vmovaps (%rdi), %xmm9
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm6 = xmm9[0],xmm8[0]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm7, %ymm6, %ymm10
+; AVX1-ONLY-NEXT: vinsertf128 $1, (%rdx), %ymm6, %ymm6
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm6 = ymm6[0],ymm10[1],ymm6[2],ymm10[2]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm7, %ymm8, %ymm7
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm8 = xmm9[2,3,2,3]
+; AVX1-ONLY-NEXT: vbroadcastsd 8(%rdx), %ymm9
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3],ymm9[4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1],ymm7[2,3],ymm8[4,5],ymm7[6,7]
+; AVX1-ONLY-NEXT: vmovaps 16(%r9), %xmm10
+; AVX1-ONLY-NEXT: vmovaps 16(%r8), %xmm11
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm7 = xmm11[0],xmm10[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vbroadcastsd 16(%rax), %ymm9
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5],ymm9[6,7]
+; AVX1-ONLY-NEXT: vmovaps 16(%rsi), %xmm12
+; AVX1-ONLY-NEXT: vmovaps 16(%rdi), %xmm13
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm9 = xmm13[0],xmm12[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm9 = ymm9[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vbroadcastsd 16(%rcx), %ymm14
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm9 = ymm9[0,1,2,3,4,5],ymm14[6,7]
+; AVX1-ONLY-NEXT: vmovaps 48(%r9), %xmm14
; AVX1-ONLY-NEXT: vmovaps 48(%r8), %xmm15
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm14, %ymm11, %ymm11
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm13 = xmm13[2,3,2,3]
-; AVX1-ONLY-NEXT: vbroadcastsd 40(%rdx), %ymm14
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3],ymm14[4,5],ymm13[6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm11 = ymm13[0,1],ymm11[2,3],ymm13[4,5],ymm11[6,7]
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm13 = ymm15[0],ymm7[0],ymm15[2],ymm7[2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3],mem[4,5],ymm13[6,7]
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm15[1],xmm14[1]
+; AVX1-ONLY-NEXT: vbroadcastsd 56(%r10), %ymm1
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm1 = xmm15[0],xmm14[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],mem[4,5,6,7]
; AVX1-ONLY-NEXT: vbroadcastsd 48(%rax), %ymm14
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3,4,5],ymm14[6,7]
-; AVX1-ONLY-NEXT: vmovaps (%rax), %xmm0
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm14 = xmm12[0],xmm10[0]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm14, %ymm1
-; AVX1-ONLY-NEXT: vinsertf128 $1, (%r10), %ymm14, %ymm14
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm14 = ymm14[0],ymm1[1],ymm14[2],ymm1[2]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm10, %ymm0
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm1 = xmm12[2,3,2,3]
-; AVX1-ONLY-NEXT: vbroadcastsd 8(%r10), %ymm10
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm10[4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm10 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vmovaps 48(%rsi), %xmm6
-; AVX1-ONLY-NEXT: vmovaps 48(%rdi), %xmm5
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm12 = ymm5[0],ymm6[0],ymm5[2],ymm6[2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3],mem[4,5],ymm12[6,7]
-; AVX1-ONLY-NEXT: vbroadcastsd 48(%rcx), %ymm2
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3,4,5],ymm2[6,7]
-; AVX1-ONLY-NEXT: vmovaps (%rsi), %xmm2
-; AVX1-ONLY-NEXT: vmovaps (%rdi), %xmm3
-; AVX1-ONLY-NEXT: vmovaps (%rcx), %xmm1
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm0 = xmm3[0],xmm2[0]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm4
-; AVX1-ONLY-NEXT: vinsertf128 $1, (%rdx), %ymm0, %ymm0
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[0],ymm4[1],ymm0[2],ymm4[2]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm2, %ymm1
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm2 = xmm3[2,3,2,3]
-; AVX1-ONLY-NEXT: vbroadcastsd 8(%rdx), %ymm3
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5],ymm2[6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3],ymm2[4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm2 = xmm15[1],xmm7[1]
-; AVX1-ONLY-NEXT: vbroadcastsd 56(%r10), %ymm3
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5],ymm2[6,7]
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm3 = xmm5[1],xmm6[1]
-; AVX1-ONLY-NEXT: vbroadcastsd 56(%rdx), %ymm4
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm4[4,5],ymm3[6,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovlps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4, %xmm4 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm4 = mem[0,1],xmm4[2,3]
-; AVX1-ONLY-NEXT: vbroadcastsd 24(%r10), %ymm5
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm5[4,5],ymm4[6,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovlps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm5 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm5 = mem[0,1],xmm5[2,3]
-; AVX1-ONLY-NEXT: vbroadcastsd 24(%rdx), %ymm6
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm6[4,5],ymm5[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm14[6,7]
+; AVX1-ONLY-NEXT: vmovaps 48(%rsi), %xmm14
+; AVX1-ONLY-NEXT: vmovaps 48(%rdi), %xmm15
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm15[1],xmm14[1]
+; AVX1-ONLY-NEXT: vbroadcastsd 56(%rdx), %ymm3
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm3 = xmm15[0],xmm14[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vbroadcastsd 48(%rcx), %ymm14
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm14[6,7]
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm10 = xmm11[1],xmm10[1]
+; AVX1-ONLY-NEXT: vbroadcastsd 24(%r10), %ymm11
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm10 = ymm10[0,1,2,3],ymm11[4,5,6,7]
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm11 = xmm13[1],xmm12[1]
+; AVX1-ONLY-NEXT: vbroadcastsd 24(%rdx), %ymm12
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm11 = ymm11[0,1,2,3],ymm12[4,5,6,7]
; AVX1-ONLY-NEXT: movq {{[0-9]+}}(%rsp), %rdx
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],mem[6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],mem[6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5],mem[6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5],mem[6,7]
-; AVX1-ONLY-NEXT: vmovaps %ymm5, 192(%rdx)
-; AVX1-ONLY-NEXT: vmovaps %ymm4, 224(%rdx)
-; AVX1-ONLY-NEXT: vmovaps %ymm1, 64(%rdx)
-; AVX1-ONLY-NEXT: vmovapd %ymm0, (%rdx)
-; AVX1-ONLY-NEXT: vmovaps %ymm12, 384(%rdx)
-; AVX1-ONLY-NEXT: vmovaps %ymm3, 448(%rdx)
-; AVX1-ONLY-NEXT: vmovaps %ymm10, 96(%rdx)
-; AVX1-ONLY-NEXT: vmovapd %ymm14, 32(%rdx)
-; AVX1-ONLY-NEXT: vmovaps %ymm13, 416(%rdx)
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],mem[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm10 = ymm10[0,1,2,3,4,5],mem[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm11 = ymm11[0,1,2,3,4,5],mem[6,7]
+; AVX1-ONLY-NEXT: vmovaps %ymm11, 192(%rdx)
+; AVX1-ONLY-NEXT: vmovaps %ymm10, 224(%rdx)
+; AVX1-ONLY-NEXT: vmovaps %ymm8, 64(%rdx)
+; AVX1-ONLY-NEXT: vmovapd %ymm6, (%rdx)
+; AVX1-ONLY-NEXT: vmovaps %ymm3, 384(%rdx)
+; AVX1-ONLY-NEXT: vmovaps %ymm0, 448(%rdx)
+; AVX1-ONLY-NEXT: vmovaps %ymm5, 96(%rdx)
+; AVX1-ONLY-NEXT: vmovapd %ymm4, 32(%rdx)
+; AVX1-ONLY-NEXT: vmovaps %ymm1, 416(%rdx)
; AVX1-ONLY-NEXT: vmovaps %ymm2, 480(%rdx)
-; AVX1-ONLY-NEXT: vmovaps %ymm11, 320(%rdx)
-; AVX1-ONLY-NEXT: vmovapd %ymm9, 256(%rdx)
-; AVX1-ONLY-NEXT: vmovaps %ymm8, 352(%rdx)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm0, 288(%rdx)
+; AVX1-ONLY-NEXT: vmovaps %ymm0, 320(%rdx)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm0, 128(%rdx)
+; AVX1-ONLY-NEXT: vmovaps %ymm0, 256(%rdx)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm0, 160(%rdx)
-; AVX1-ONLY-NEXT: addq $104, %rsp
+; AVX1-ONLY-NEXT: vmovaps %ymm0, 352(%rdx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm0, 288(%rdx)
+; AVX1-ONLY-NEXT: vmovaps %ymm9, 128(%rdx)
+; AVX1-ONLY-NEXT: vmovaps %ymm7, 160(%rdx)
+; AVX1-ONLY-NEXT: popq %rax
; AVX1-ONLY-NEXT: vzeroupper
; AVX1-ONLY-NEXT: retq
;
@@ -684,7 +678,7 @@ define void @store_i64_stride8_vf8(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
; AVX2-ONLY-NEXT: vmovaps 32(%r8), %xmm7
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm7[1],xmm5[1]
; AVX2-ONLY-NEXT: vbroadcastsd 40(%r10), %ymm2
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovaps 32(%rcx), %xmm12
@@ -695,18 +689,18 @@ define void @store_i64_stride8_vf8(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
; AVX2-ONLY-NEXT: vmovaps 32(%rdi), %xmm15
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm2 = xmm15[1],xmm14[1]
; AVX2-ONLY-NEXT: vbroadcastsd 40(%rdx), %ymm4
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm4[4,5],ymm2[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm4[4,5,6,7]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3,4,5],ymm1[6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm6, %ymm0, %ymm2
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm4 = xmm10[1],xmm9[1]
; AVX2-ONLY-NEXT: vbroadcastsd 8(%r10), %ymm8
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm8[4,5],ymm4[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm8[4,5,6,7]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm4[0,1,2,3,4,5],ymm2[6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm4 = xmm13[1],xmm11[1]
; AVX2-ONLY-NEXT: vbroadcastsd 8(%rdx), %ymm8
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm8[4,5],ymm4[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm8[4,5,6,7]
; AVX2-ONLY-NEXT: vmovaps (%rcx), %xmm0
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm8
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5],ymm8[6,7]
@@ -1375,234 +1369,221 @@ define void @store_i64_stride8_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
;
; AVX1-ONLY-LABEL: store_i64_stride8_vf16:
; AVX1-ONLY: # %bb.0:
-; AVX1-ONLY-NEXT: subq $712, %rsp # imm = 0x2C8
+; AVX1-ONLY-NEXT: subq $488, %rsp # imm = 0x1E8
; AVX1-ONLY-NEXT: movq {{[0-9]+}}(%rsp), %r10
; AVX1-ONLY-NEXT: movq {{[0-9]+}}(%rsp), %rax
; AVX1-ONLY-NEXT: vmovaps (%rsi), %xmm4
-; AVX1-ONLY-NEXT: vmovaps 16(%rsi), %xmm13
-; AVX1-ONLY-NEXT: vmovaps 32(%rsi), %xmm0
; AVX1-ONLY-NEXT: vmovaps (%rdi), %xmm5
-; AVX1-ONLY-NEXT: vmovaps 16(%rdi), %xmm14
-; AVX1-ONLY-NEXT: vmovaps 32(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm3 = xmm5[0],xmm4[0]
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm1 = xmm5[0],xmm4[0]
; AVX1-ONLY-NEXT: vmovaps (%rcx), %xmm6
; AVX1-ONLY-NEXT: vmovaps 32(%rcx), %xmm2
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm6, %ymm3, %ymm7
-; AVX1-ONLY-NEXT: vinsertf128 $1, (%rdx), %ymm3, %ymm3
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm3 = ymm3[0],ymm7[1],ymm3[2],ymm7[2]
-; AVX1-ONLY-NEXT: vmovupd %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 64(%rcx), %xmm0
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm6, %ymm1, %ymm3
+; AVX1-ONLY-NEXT: vinsertf128 $1, (%rdx), %ymm1, %ymm1
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm1 = ymm1[0],ymm3[1],ymm1[2],ymm3[2]
+; AVX1-ONLY-NEXT: vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps (%r9), %xmm7
-; AVX1-ONLY-NEXT: vmovaps 16(%r9), %xmm12
-; AVX1-ONLY-NEXT: vmovups %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 32(%r9), %xmm3
; AVX1-ONLY-NEXT: vmovaps (%r8), %xmm8
; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm9 = xmm8[0],xmm7[0]
; AVX1-ONLY-NEXT: vmovaps (%rax), %xmm10
+; AVX1-ONLY-NEXT: vmovaps 32(%rax), %xmm3
+; AVX1-ONLY-NEXT: vmovaps 64(%rax), %xmm1
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm10, %ymm9, %ymm11
; AVX1-ONLY-NEXT: vinsertf128 $1, (%r10), %ymm9, %ymm9
; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm9 = ymm9[0],ymm11[1],ymm9[2],ymm11[2]
; AVX1-ONLY-NEXT: vmovupd %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 16(%r8), %xmm9
-; AVX1-ONLY-NEXT: vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm6, %ymm4, %ymm4
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm5 = xmm5[2,3,2,3]
; AVX1-ONLY-NEXT: vbroadcastsd 8(%rdx), %ymm6
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm6[4,5],ymm5[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm6[4,5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm5[0,1],ymm4[2,3],ymm5[4,5],ymm4[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm10, %ymm7, %ymm4
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm5 = xmm8[2,3,2,3]
; AVX1-ONLY-NEXT: vbroadcastsd 8(%r10), %ymm6
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm6[4,5],ymm5[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm6[4,5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm5[0,1],ymm4[2,3],ymm5[4,5],ymm4[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm4 = ymm14[0],ymm13[0],ymm14[2],ymm13[2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],mem[4,5],ymm4[6,7]
-; AVX1-ONLY-NEXT: vbroadcastsd 16(%rcx), %ymm5
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5],ymm5[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm4 = ymm9[0],ymm12[0],ymm9[2],ymm12[2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],mem[4,5],ymm4[6,7]
-; AVX1-ONLY-NEXT: vbroadcastsd 16(%rax), %ymm5
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5],ymm5[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm4 = xmm1[0],xmm0[0]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm4, %ymm5
-; AVX1-ONLY-NEXT: vinsertf128 $1, 32(%rdx), %ymm4, %ymm4
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm4 = ymm4[0],ymm5[1],ymm4[2],ymm5[2]
-; AVX1-ONLY-NEXT: vmovupd %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 32(%r8), %xmm4
-; AVX1-ONLY-NEXT: vmovaps 32(%rax), %xmm5
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm6 = xmm4[0],xmm3[0]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm5, %ymm6, %ymm7
-; AVX1-ONLY-NEXT: vinsertf128 $1, 32(%r10), %ymm6, %ymm6
+; AVX1-ONLY-NEXT: vmovaps 32(%rsi), %xmm4
+; AVX1-ONLY-NEXT: vmovaps 32(%rdi), %xmm5
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm6 = xmm5[0],xmm4[0]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm6, %ymm7
+; AVX1-ONLY-NEXT: vinsertf128 $1, 32(%rdx), %ymm6, %ymm6
; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm6 = ymm6[0],ymm7[1],ymm6[2],ymm7[2]
; AVX1-ONLY-NEXT: vmovupd %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm1 = xmm1[2,3,2,3]
-; AVX1-ONLY-NEXT: vbroadcastsd 40(%rdx), %ymm2
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm5, %ymm3, %ymm0
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm1 = xmm4[2,3,2,3]
-; AVX1-ONLY-NEXT: vbroadcastsd 40(%r10), %ymm2
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT: vmovaps 48(%rsi), %xmm2
+; AVX1-ONLY-NEXT: vmovaps 32(%r9), %xmm6
+; AVX1-ONLY-NEXT: vmovaps 32(%r8), %xmm7
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm8 = xmm7[0],xmm6[0]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm8, %ymm9
+; AVX1-ONLY-NEXT: vinsertf128 $1, 32(%r10), %ymm8, %ymm8
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm8 = ymm8[0],ymm9[1],ymm8[2],ymm9[2]
+; AVX1-ONLY-NEXT: vmovupd %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm4, %ymm2
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm4 = xmm5[2,3,2,3]
+; AVX1-ONLY-NEXT: vbroadcastsd 40(%rdx), %ymm5
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm5[4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm4[0,1],ymm2[2,3],ymm4[4,5],ymm2[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 48(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm2[0],ymm0[2],ymm2[2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],mem[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vbroadcastsd 48(%rcx), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 48(%r9), %xmm0
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 48(%r8), %xmm12
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm12[0],ymm0[0],ymm12[2],ymm0[2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],mem[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vbroadcastsd 48(%rax), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 64(%rcx), %xmm0
-; AVX1-ONLY-NEXT: vmovaps 64(%rsi), %xmm1
-; AVX1-ONLY-NEXT: vmovaps 64(%rdi), %xmm2
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm3 = xmm2[0],xmm1[0]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm3, %ymm4
-; AVX1-ONLY-NEXT: vinsertf128 $1, 64(%rdx), %ymm3, %ymm3
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm3 = ymm3[0],ymm4[1],ymm3[2],ymm4[2]
-; AVX1-ONLY-NEXT: vmovupd %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 64(%rax), %xmm3
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm6, %ymm2
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm3 = xmm7[2,3,2,3]
+; AVX1-ONLY-NEXT: vbroadcastsd 40(%r10), %ymm4
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm4[4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3],ymm3[4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 64(%rsi), %xmm2
+; AVX1-ONLY-NEXT: vmovaps 64(%rdi), %xmm3
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm4 = xmm3[0],xmm2[0]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm4, %ymm5
+; AVX1-ONLY-NEXT: vinsertf128 $1, 64(%rdx), %ymm4, %ymm4
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm4 = ymm4[0],ymm5[1],ymm4[2],ymm5[2]
+; AVX1-ONLY-NEXT: vmovupd %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 64(%r9), %xmm4
; AVX1-ONLY-NEXT: vmovaps 64(%r8), %xmm5
; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm6 = xmm5[0],xmm4[0]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm6, %ymm7
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm6, %ymm7
; AVX1-ONLY-NEXT: vinsertf128 $1, 64(%r10), %ymm6, %ymm6
; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm6 = ymm6[0],ymm7[1],ymm6[2],ymm7[2]
; AVX1-ONLY-NEXT: vmovupd %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm1 = xmm2[2,3,2,3]
-; AVX1-ONLY-NEXT: vbroadcastsd 72(%rdx), %ymm2
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm2, %ymm0
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm2 = xmm3[2,3,2,3]
+; AVX1-ONLY-NEXT: vbroadcastsd 72(%rdx), %ymm3
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps 96(%rcx), %xmm3
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3],ymm2[4,5],ymm0[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm4, %ymm0
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm4, %ymm0
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm1 = xmm5[2,3,2,3]
; AVX1-ONLY-NEXT: vbroadcastsd 72(%r10), %ymm2
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 80(%rsi), %xmm9
-; AVX1-ONLY-NEXT: vmovaps 80(%rdi), %xmm7
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm7[0],ymm9[0],ymm7[2],ymm9[2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],mem[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vbroadcastsd 80(%rcx), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, (%rsp) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 80(%r9), %xmm6
-; AVX1-ONLY-NEXT: vmovaps 80(%r8), %xmm5
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm5[0],ymm6[0],ymm5[2],ymm6[2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],mem[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vbroadcastsd 80(%rax), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 96(%rcx), %xmm0
-; AVX1-ONLY-NEXT: vmovaps 96(%rsi), %xmm1
-; AVX1-ONLY-NEXT: vmovaps 96(%rdi), %xmm2
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm3 = xmm2[0],xmm1[0]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm3, %ymm4
-; AVX1-ONLY-NEXT: vinsertf128 $1, 96(%rdx), %ymm3, %ymm3
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm3 = ymm3[0],ymm4[1],ymm3[2],ymm4[2]
-; AVX1-ONLY-NEXT: vmovupd %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 96(%rax), %xmm3
+; AVX1-ONLY-NEXT: vmovaps 96(%rsi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps 96(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm2, %ymm4
+; AVX1-ONLY-NEXT: vinsertf128 $1, 96(%rdx), %ymm2, %ymm2
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm2 = ymm2[0],ymm4[1],ymm2[2],ymm4[2]
+; AVX1-ONLY-NEXT: vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 96(%rax), %xmm2
; AVX1-ONLY-NEXT: vmovaps 96(%r9), %xmm4
-; AVX1-ONLY-NEXT: vmovaps 96(%r8), %xmm8
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm10 = xmm8[0],xmm4[0]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm10, %ymm11
-; AVX1-ONLY-NEXT: vinsertf128 $1, 96(%r10), %ymm10, %ymm10
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm10 = ymm10[0],ymm11[1],ymm10[2],ymm11[2]
-; AVX1-ONLY-NEXT: vmovupd %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm1 = xmm2[2,3,2,3]
-; AVX1-ONLY-NEXT: vbroadcastsd 104(%rdx), %ymm2
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT: vmovaps 96(%r8), %xmm5
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm6 = xmm5[0],xmm4[0]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm6, %ymm7
+; AVX1-ONLY-NEXT: vinsertf128 $1, 96(%r10), %ymm6, %ymm6
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm6 = ymm6[0],ymm7[1],ymm6[2],ymm7[2]
+; AVX1-ONLY-NEXT: vmovupd %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm0
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm1 = xmm1[2,3,2,3]
+; AVX1-ONLY-NEXT: vbroadcastsd 104(%rdx), %ymm3
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm4, %ymm0
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm1 = xmm8[2,3,2,3]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm4, %ymm0
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm1 = xmm5[2,3,2,3]
; AVX1-ONLY-NEXT: vbroadcastsd 104(%r10), %ymm2
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm11 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vmovaps 112(%rsi), %xmm3
-; AVX1-ONLY-NEXT: vmovaps 112(%rdi), %xmm2
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm2[0],ymm3[0],ymm2[2],ymm3[2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],mem[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vbroadcastsd 112(%rcx), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm10 = ymm0[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT: vmovaps 112(%r9), %xmm1
-; AVX1-ONLY-NEXT: vmovaps 112(%r8), %xmm0
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm4 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],mem[4,5],ymm4[6,7]
-; AVX1-ONLY-NEXT: vbroadcastsd 112(%rax), %ymm8
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm4[0,1,2,3,4,5],ymm8[6,7]
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm4 = xmm14[1],xmm13[1]
-; AVX1-ONLY-NEXT: vbroadcastsd 24(%rdx), %ymm15
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm15[4,5],ymm4[6,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovlps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13, %xmm15 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm15 = mem[0,1],xmm13[2,3]
-; AVX1-ONLY-NEXT: vbroadcastsd 24(%r10), %ymm14
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm14 = ymm15[0,1,2,3],ymm14[4,5],ymm15[6,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovlps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13, %xmm15 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm15 = mem[0,1],xmm13[2,3]
-; AVX1-ONLY-NEXT: vbroadcastsd 56(%rdx), %ymm13
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm13 = ymm15[0,1,2,3],ymm13[4,5],ymm15[6,7]
-; AVX1-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm12, %xmm12 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm12 = xmm12[1],mem[1]
-; AVX1-ONLY-NEXT: vbroadcastsd 56(%r10), %ymm15
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3],ymm15[4,5],ymm12[6,7]
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm7 = xmm7[1],xmm9[1]
-; AVX1-ONLY-NEXT: vbroadcastsd 88(%rdx), %ymm9
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3],ymm9[4,5],ymm7[6,7]
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm5 = xmm5[1],xmm6[1]
-; AVX1-ONLY-NEXT: vbroadcastsd 88(%r10), %ymm6
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm6[4,5],ymm5[6,7]
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm2 = xmm2[1],xmm3[1]
-; AVX1-ONLY-NEXT: vbroadcastsd 120(%rdx), %ymm3
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 16(%rsi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps 16(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vbroadcastsd 16(%rcx), %ymm3
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm3[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm2, (%rsp) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 16(%r9), %xmm2
+; AVX1-ONLY-NEXT: vmovaps 16(%r8), %xmm3
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm4 = xmm3[0],xmm2[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vbroadcastsd 16(%rax), %ymm5
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5],ymm5[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT: vbroadcastsd 24(%rdx), %ymm1
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm11 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm3[1],xmm2[1]
+; AVX1-ONLY-NEXT: vbroadcastsd 24(%r10), %ymm1
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm9 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps 48(%rsi), %xmm4
+; AVX1-ONLY-NEXT: vmovaps 48(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm4[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vbroadcastsd 48(%rcx), %ymm3
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3,4,5],ymm3[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 48(%r9), %xmm2
+; AVX1-ONLY-NEXT: vmovaps 48(%r8), %xmm0
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm3 = xmm0[0],xmm2[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vbroadcastsd 48(%rax), %ymm6
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm10 = ymm3[0,1,2,3,4,5],ymm6[6,7]
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm4[1]
+; AVX1-ONLY-NEXT: vbroadcastsd 56(%rdx), %ymm3
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm1[0,1,2,3],ymm3[4,5,6,7]
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm2[1]
+; AVX1-ONLY-NEXT: vbroadcastsd 56(%r10), %ymm1
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps 80(%rsi), %xmm4
+; AVX1-ONLY-NEXT: vmovaps 80(%rdi), %xmm2
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm1 = xmm2[0],xmm4[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vbroadcastsd 80(%rcx), %ymm3
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm1[0,1,2,3,4,5],ymm3[6,7]
+; AVX1-ONLY-NEXT: vmovaps 80(%r9), %xmm1
+; AVX1-ONLY-NEXT: vmovaps 80(%r8), %xmm0
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm3 = xmm0[0],xmm1[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vbroadcastsd 80(%rax), %ymm14
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm3[0,1,2,3,4,5],ymm14[6,7]
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm2 = xmm2[1],xmm4[1]
+; AVX1-ONLY-NEXT: vbroadcastsd 88(%rdx), %ymm4
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm2[0,1,2,3],ymm4[4,5,6,7]
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
-; AVX1-ONLY-NEXT: vbroadcastsd 120(%r10), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT: vbroadcastsd 88(%r10), %ymm1
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps 112(%rsi), %xmm1
+; AVX1-ONLY-NEXT: vmovaps 112(%rdi), %xmm4
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm14 = xmm4[0],xmm1[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vbroadcastsd 112(%rcx), %ymm13
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm13 = ymm14[0,1,2,3,4,5],ymm13[6,7]
+; AVX1-ONLY-NEXT: vmovaps 112(%r9), %xmm14
+; AVX1-ONLY-NEXT: vmovaps 112(%r8), %xmm0
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm15 = xmm0[0],xmm14[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm15 = ymm15[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vbroadcastsd 112(%rax), %ymm12
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm12 = ymm15[0,1,2,3,4,5],ymm12[6,7]
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm4[1],xmm1[1]
+; AVX1-ONLY-NEXT: vbroadcastsd 120(%rdx), %ymm4
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm4[4,5,6,7]
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm14[1]
+; AVX1-ONLY-NEXT: vbroadcastsd 120(%r10), %ymm4
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm4[4,5,6,7]
; AVX1-ONLY-NEXT: movq {{[0-9]+}}(%rsp), %rdx
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm4[0,1,2,3,4,5],mem[6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm14[0,1,2,3,4,5],mem[6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm13[0,1,2,3,4,5],mem[6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm12[0,1,2,3,4,5],mem[6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5],mem[6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5],mem[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm11[0,1,2,3,4,5],mem[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm9 = ymm9[0,1,2,3,4,5],mem[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5],mem[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5],mem[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],mem[6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],mem[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],mem[6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],mem[6,7]
; AVX1-ONLY-NEXT: vmovaps %ymm0, 992(%rdx)
-; AVX1-ONLY-NEXT: vmovaps %ymm2, 960(%rdx)
-; AVX1-ONLY-NEXT: vmovaps %ymm8, 928(%rdx)
-; AVX1-ONLY-NEXT: vmovaps %ymm10, 896(%rdx)
-; AVX1-ONLY-NEXT: vmovaps %ymm11, 864(%rdx)
+; AVX1-ONLY-NEXT: vmovaps %ymm1, 960(%rdx)
+; AVX1-ONLY-NEXT: vmovaps %ymm12, 928(%rdx)
+; AVX1-ONLY-NEXT: vmovaps %ymm13, 896(%rdx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm0, 864(%rdx)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm0, 832(%rdx)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm0, 800(%rdx)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm0, 768(%rdx)
-; AVX1-ONLY-NEXT: vmovaps %ymm5, 736(%rdx)
-; AVX1-ONLY-NEXT: vmovaps %ymm7, 704(%rdx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm0, 672(%rdx)
-; AVX1-ONLY-NEXT: vmovups (%rsp), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm0, 640(%rdx)
+; AVX1-ONLY-NEXT: vmovaps %ymm2, 736(%rdx)
+; AVX1-ONLY-NEXT: vmovaps %ymm3, 704(%rdx)
+; AVX1-ONLY-NEXT: vmovaps %ymm5, 672(%rdx)
+; AVX1-ONLY-NEXT: vmovaps %ymm7, 640(%rdx)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm0, 608(%rdx)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
@@ -1612,9 +1593,8 @@ define void @store_i64_stride8_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm0, 512(%rdx)
; AVX1-ONLY-NEXT: vmovaps %ymm6, 480(%rdx)
-; AVX1-ONLY-NEXT: vmovaps %ymm4, 448(%rdx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm0, 416(%rdx)
+; AVX1-ONLY-NEXT: vmovaps %ymm8, 448(%rdx)
+; AVX1-ONLY-NEXT: vmovaps %ymm10, 416(%rdx)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm0, 384(%rdx)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
@@ -1625,11 +1605,11 @@ define void @store_i64_stride8_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vmovaps %ymm0, 288(%rdx)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm0, 256(%rdx)
-; AVX1-ONLY-NEXT: vmovaps %ymm3, 224(%rdx)
-; AVX1-ONLY-NEXT: vmovaps %ymm1, 192(%rdx)
+; AVX1-ONLY-NEXT: vmovaps %ymm9, 224(%rdx)
+; AVX1-ONLY-NEXT: vmovaps %ymm4, 192(%rdx)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm0, 160(%rdx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovups (%rsp), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm0, 128(%rdx)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm0, 96(%rdx)
@@ -1639,7 +1619,7 @@ define void @store_i64_stride8_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vmovaps %ymm0, 32(%rdx)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm0, (%rdx)
-; AVX1-ONLY-NEXT: addq $712, %rsp # imm = 0x2C8
+; AVX1-ONLY-NEXT: addq $488, %rsp # imm = 0x1E8
; AVX1-ONLY-NEXT: vzeroupper
; AVX1-ONLY-NEXT: retq
;
@@ -1663,7 +1643,7 @@ define void @store_i64_stride8_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm2[1]
; AVX2-ONLY-NEXT: vbroadcastsd 8(%rdx), %ymm2
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovaps (%rax), %xmm0
@@ -1678,18 +1658,18 @@ define void @store_i64_stride8_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm13[1],xmm1[1]
; AVX2-ONLY-NEXT: vbroadcastsd 8(%r10), %ymm2
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm0
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm5[1],xmm4[1]
; AVX2-ONLY-NEXT: vbroadcastsd 40(%rdx), %ymm2
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm7[1],xmm6[1]
; AVX2-ONLY-NEXT: vbroadcastsd 40(%r10), %ymm1
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT: vmovaps 32(%rax), %xmm9
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm9, %ymm0, %ymm1
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
@@ -1698,7 +1678,7 @@ define void @store_i64_stride8_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovaps 64(%rdi), %xmm11
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm11[1],xmm12[1]
; AVX2-ONLY-NEXT: vbroadcastsd 72(%rdx), %ymm1
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT: vmovaps 64(%rcx), %xmm7
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm7, %ymm0, %ymm1
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
@@ -1707,7 +1687,7 @@ define void @store_i64_stride8_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovaps 64(%r8), %xmm8
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm8[1],xmm10[1]
; AVX2-ONLY-NEXT: vbroadcastsd 72(%r10), %ymm1
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT: vmovaps 64(%rax), %xmm6
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm6, %ymm0, %ymm1
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
@@ -1716,7 +1696,7 @@ define void @store_i64_stride8_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovaps 96(%rdi), %xmm4
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm4[1],xmm5[1]
; AVX2-ONLY-NEXT: vbroadcastsd 104(%rdx), %ymm1
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT: vmovaps 96(%rcx), %xmm3
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm1
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
@@ -1725,7 +1705,7 @@ define void @store_i64_stride8_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovaps 96(%r8), %xmm1
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm2[1]
; AVX2-ONLY-NEXT: vbroadcastsd 104(%r10), %ymm14
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm14 = ymm0[0,1,2,3],ymm14[4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm14 = ymm0[0,1,2,3],ymm14[4,5,6,7]
; AVX2-ONLY-NEXT: vmovaps 96(%rax), %xmm0
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm15
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4,5],ymm15[6,7]
@@ -3039,190 +3019,121 @@ define void @store_i64_stride8_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
;
; AVX1-ONLY-LABEL: store_i64_stride8_vf32:
; AVX1-ONLY: # %bb.0:
-; AVX1-ONLY-NEXT: subq $2056, %rsp # imm = 0x808
+; AVX1-ONLY-NEXT: subq $1704, %rsp # imm = 0x6A8
; AVX1-ONLY-NEXT: movq {{[0-9]+}}(%rsp), %r10
; AVX1-ONLY-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; AVX1-ONLY-NEXT: vmovaps (%rsi), %xmm4
-; AVX1-ONLY-NEXT: vmovaps 16(%rsi), %xmm12
-; AVX1-ONLY-NEXT: vmovups %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 32(%rsi), %xmm0
-; AVX1-ONLY-NEXT: vmovaps (%rdi), %xmm5
-; AVX1-ONLY-NEXT: vmovaps 16(%rdi), %xmm13
-; AVX1-ONLY-NEXT: vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 32(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm3 = xmm5[0],xmm4[0]
-; AVX1-ONLY-NEXT: vmovaps (%rcx), %xmm6
-; AVX1-ONLY-NEXT: vmovaps 32(%rcx), %xmm2
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm6, %ymm3, %ymm7
-; AVX1-ONLY-NEXT: vinsertf128 $1, (%rdx), %ymm3, %ymm3
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm3 = ymm3[0],ymm7[1],ymm3[2],ymm7[2]
-; AVX1-ONLY-NEXT: vmovupd %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps (%r9), %xmm7
-; AVX1-ONLY-NEXT: vmovaps 16(%r9), %xmm14
-; AVX1-ONLY-NEXT: vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 32(%r9), %xmm3
-; AVX1-ONLY-NEXT: vmovaps (%r8), %xmm8
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm9 = xmm8[0],xmm7[0]
-; AVX1-ONLY-NEXT: vmovaps (%rax), %xmm10
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm10, %ymm9, %ymm11
-; AVX1-ONLY-NEXT: vinsertf128 $1, (%r10), %ymm9, %ymm9
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm9 = ymm9[0],ymm11[1],ymm9[2],ymm11[2]
-; AVX1-ONLY-NEXT: vmovupd %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 16(%r8), %xmm9
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm6, %ymm4, %ymm4
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm5 = xmm5[2,3,2,3]
-; AVX1-ONLY-NEXT: vbroadcastsd 8(%rdx), %ymm6
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm6[4,5],ymm5[6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm5[0,1],ymm4[2,3],ymm5[4,5],ymm4[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm10, %ymm7, %ymm4
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm5 = xmm8[2,3,2,3]
-; AVX1-ONLY-NEXT: vbroadcastsd 8(%r10), %ymm6
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm6[4,5],ymm5[6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm5[0,1],ymm4[2,3],ymm5[4,5],ymm4[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm4 = ymm13[0],ymm12[0],ymm13[2],ymm12[2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],mem[4,5],ymm4[6,7]
-; AVX1-ONLY-NEXT: vbroadcastsd 16(%rcx), %ymm5
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5],ymm5[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm4 = ymm9[0],ymm14[0],ymm9[2],ymm14[2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],mem[4,5],ymm4[6,7]
-; AVX1-ONLY-NEXT: vbroadcastsd 16(%rax), %ymm5
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5],ymm5[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm4 = xmm1[0],xmm0[0]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm4, %ymm5
-; AVX1-ONLY-NEXT: vinsertf128 $1, 32(%rdx), %ymm4, %ymm4
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm4 = ymm4[0],ymm5[1],ymm4[2],ymm5[2]
+; AVX1-ONLY-NEXT: vmovaps (%rsi), %xmm2
+; AVX1-ONLY-NEXT: vmovaps (%rdi), %xmm3
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm4 = xmm3[0],xmm2[0]
+; AVX1-ONLY-NEXT: vmovaps (%rcx), %xmm5
+; AVX1-ONLY-NEXT: vmovaps 32(%rcx), %xmm1
+; AVX1-ONLY-NEXT: vmovaps 64(%rcx), %xmm0
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm5, %ymm4, %ymm6
+; AVX1-ONLY-NEXT: vinsertf128 $1, (%rdx), %ymm4, %ymm4
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm4 = ymm4[0],ymm6[1],ymm4[2],ymm6[2]
; AVX1-ONLY-NEXT: vmovupd %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 32(%r8), %xmm4
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm5 = xmm4[0],xmm3[0]
-; AVX1-ONLY-NEXT: vmovaps 32(%rax), %xmm6
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm6, %ymm5, %ymm7
-; AVX1-ONLY-NEXT: vinsertf128 $1, 32(%r10), %ymm5, %ymm5
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm5 = ymm5[0],ymm7[1],ymm5[2],ymm7[2]
-; AVX1-ONLY-NEXT: vmovupd %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm1 = xmm1[2,3,2,3]
-; AVX1-ONLY-NEXT: vbroadcastsd 40(%rdx), %ymm2
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm6, %ymm3, %ymm0
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm1 = xmm4[2,3,2,3]
-; AVX1-ONLY-NEXT: vbroadcastsd 40(%r10), %ymm2
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT: vmovaps 48(%rsi), %xmm8
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 48(%rdi), %xmm10
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm10[0],ymm8[0],ymm10[2],ymm8[2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],mem[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vbroadcastsd 48(%rcx), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 48(%r9), %xmm0
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 48(%r8), %xmm1
+; AVX1-ONLY-NEXT: vmovaps (%r9), %xmm4
+; AVX1-ONLY-NEXT: vmovaps (%r8), %xmm6
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm7 = xmm6[0],xmm4[0]
+; AVX1-ONLY-NEXT: vmovaps (%rax), %xmm8
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm8, %ymm7, %ymm9
+; AVX1-ONLY-NEXT: vinsertf128 $1, (%r10), %ymm7, %ymm7
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm7 = ymm7[0],ymm9[1],ymm7[2],ymm9[2]
+; AVX1-ONLY-NEXT: vmovupd %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm5, %ymm2, %ymm2
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm3 = xmm3[2,3,2,3]
+; AVX1-ONLY-NEXT: vbroadcastsd 8(%rdx), %ymm5
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm5[4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3],ymm3[4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm8, %ymm4, %ymm2
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm3 = xmm6[2,3,2,3]
+; AVX1-ONLY-NEXT: vbroadcastsd 8(%r10), %ymm4
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm4[4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3],ymm3[4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 32(%rsi), %xmm3
+; AVX1-ONLY-NEXT: vmovaps 32(%rdi), %xmm4
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm4[0],xmm3[0]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm2, %ymm5
+; AVX1-ONLY-NEXT: vinsertf128 $1, 32(%rdx), %ymm2, %ymm2
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm2 = ymm2[0],ymm5[1],ymm2[2],ymm5[2]
+; AVX1-ONLY-NEXT: vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 32(%r9), %xmm5
+; AVX1-ONLY-NEXT: vmovaps 32(%r8), %xmm6
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm7 = xmm6[0],xmm5[0]
+; AVX1-ONLY-NEXT: vmovaps 32(%rax), %xmm8
+; AVX1-ONLY-NEXT: vmovaps 64(%rax), %xmm2
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm8, %ymm7, %ymm9
+; AVX1-ONLY-NEXT: vinsertf128 $1, 32(%r10), %ymm7, %ymm7
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm7 = ymm7[0],ymm9[1],ymm7[2],ymm9[2]
+; AVX1-ONLY-NEXT: vmovupd %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm3, %ymm1
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm3 = xmm4[2,3,2,3]
+; AVX1-ONLY-NEXT: vbroadcastsd 40(%rdx), %ymm4
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm4[4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm3[0,1],ymm1[2,3],ymm3[4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm8, %ymm5, %ymm1
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm3 = xmm6[2,3,2,3]
+; AVX1-ONLY-NEXT: vbroadcastsd 40(%r10), %ymm4
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm4[4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm3[0,1],ymm1[2,3],ymm3[4,5],ymm1[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],mem[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vbroadcastsd 48(%rax), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 64(%rcx), %xmm0
; AVX1-ONLY-NEXT: vmovaps 64(%rsi), %xmm1
-; AVX1-ONLY-NEXT: vmovaps 64(%rdi), %xmm2
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm3 = xmm2[0],xmm1[0]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm3, %ymm4
-; AVX1-ONLY-NEXT: vinsertf128 $1, 64(%rdx), %ymm3, %ymm3
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm3 = ymm3[0],ymm4[1],ymm3[2],ymm4[2]
-; AVX1-ONLY-NEXT: vmovupd %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 64(%rax), %xmm3
+; AVX1-ONLY-NEXT: vmovaps 64(%rdi), %xmm3
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm4 = xmm3[0],xmm1[0]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm4, %ymm5
+; AVX1-ONLY-NEXT: vinsertf128 $1, 64(%rdx), %ymm4, %ymm4
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm4 = ymm4[0],ymm5[1],ymm4[2],ymm5[2]
+; AVX1-ONLY-NEXT: vmovupd %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 64(%r9), %xmm4
; AVX1-ONLY-NEXT: vmovaps 64(%r8), %xmm5
; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm6 = xmm5[0],xmm4[0]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm6, %ymm7
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm6, %ymm7
; AVX1-ONLY-NEXT: vinsertf128 $1, 64(%r10), %ymm6, %ymm6
; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm6 = ymm6[0],ymm7[1],ymm6[2],ymm7[2]
; AVX1-ONLY-NEXT: vmovupd %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm1 = xmm2[2,3,2,3]
-; AVX1-ONLY-NEXT: vbroadcastsd 72(%rdx), %ymm2
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm1 = xmm3[2,3,2,3]
+; AVX1-ONLY-NEXT: vbroadcastsd 72(%rdx), %ymm3
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps 96(%rcx), %xmm3
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm4, %ymm0
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm4, %ymm0
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm1 = xmm5[2,3,2,3]
; AVX1-ONLY-NEXT: vbroadcastsd 72(%r10), %ymm2
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 80(%rsi), %xmm1
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 80(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],mem[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vbroadcastsd 80(%rcx), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 80(%r9), %xmm1
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 80(%r8), %xmm0
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],mem[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vbroadcastsd 80(%rax), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 96(%rcx), %xmm0
-; AVX1-ONLY-NEXT: vmovaps 96(%rsi), %xmm1
-; AVX1-ONLY-NEXT: vmovaps 96(%rdi), %xmm2
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm3 = xmm2[0],xmm1[0]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm3, %ymm4
-; AVX1-ONLY-NEXT: vinsertf128 $1, 96(%rdx), %ymm3, %ymm3
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm3 = ymm3[0],ymm4[1],ymm3[2],ymm4[2]
-; AVX1-ONLY-NEXT: vmovupd %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 96(%r9), %xmm3
+; AVX1-ONLY-NEXT: vmovaps 96(%rsi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps 96(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm2, %ymm4
+; AVX1-ONLY-NEXT: vinsertf128 $1, 96(%rdx), %ymm2, %ymm2
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm2 = ymm2[0],ymm4[1],ymm2[2],ymm4[2]
+; AVX1-ONLY-NEXT: vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 96(%r9), %xmm2
; AVX1-ONLY-NEXT: vmovaps 96(%r8), %xmm4
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm5 = xmm4[0],xmm3[0]
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm5 = xmm4[0],xmm2[0]
; AVX1-ONLY-NEXT: vmovaps 96(%rax), %xmm6
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm6, %ymm5, %ymm7
; AVX1-ONLY-NEXT: vinsertf128 $1, 96(%r10), %ymm5, %ymm5
; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm5 = ymm5[0],ymm7[1],ymm5[2],ymm7[2]
; AVX1-ONLY-NEXT: vmovupd %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm1 = xmm2[2,3,2,3]
-; AVX1-ONLY-NEXT: vbroadcastsd 104(%rdx), %ymm2
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm0
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm1 = xmm1[2,3,2,3]
+; AVX1-ONLY-NEXT: vbroadcastsd 104(%rdx), %ymm3
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm6, %ymm3, %ymm0
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm6, %ymm2, %ymm0
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm1 = xmm4[2,3,2,3]
; AVX1-ONLY-NEXT: vbroadcastsd 104(%r10), %ymm2
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 112(%rsi), %xmm1
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 112(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],mem[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vbroadcastsd 112(%rcx), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 112(%r9), %xmm1
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 112(%r8), %xmm0
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],mem[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vbroadcastsd 112(%rax), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 128(%rsi), %xmm0
; AVX1-ONLY-NEXT: vmovaps 128(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
@@ -3242,32 +3153,15 @@ define void @store_i64_stride8_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm0
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm1 = xmm1[2,3,2,3]
; AVX1-ONLY-NEXT: vbroadcastsd 136(%rdx), %ymm3
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm6, %ymm2, %ymm0
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm1 = xmm4[2,3,2,3]
; AVX1-ONLY-NEXT: vbroadcastsd 136(%r10), %ymm2
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 144(%rsi), %xmm1
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 144(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],mem[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vbroadcastsd 144(%rcx), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 144(%r9), %xmm14
-; AVX1-ONLY-NEXT: vmovaps 144(%r8), %xmm0
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm14[0],ymm0[2],ymm14[2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],mem[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vbroadcastsd 144(%rax), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 160(%rsi), %xmm0
; AVX1-ONLY-NEXT: vmovaps 160(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
@@ -3287,31 +3181,15 @@ define void @store_i64_stride8_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm0
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm1 = xmm1[2,3,2,3]
; AVX1-ONLY-NEXT: vbroadcastsd 168(%rdx), %ymm3
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm4, %ymm0
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm1 = xmm5[2,3,2,3]
; AVX1-ONLY-NEXT: vbroadcastsd 168(%r10), %ymm2
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 176(%rsi), %xmm15
-; AVX1-ONLY-NEXT: vmovaps 176(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm15[0],ymm0[2],ymm15[2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],mem[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vbroadcastsd 176(%rcx), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 176(%r9), %xmm13
-; AVX1-ONLY-NEXT: vmovaps 176(%r8), %xmm0
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm13[0],ymm0[2],ymm13[2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],mem[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vbroadcastsd 176(%rax), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 192(%rsi), %xmm0
; AVX1-ONLY-NEXT: vmovaps 192(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
@@ -3331,31 +3209,15 @@ define void @store_i64_stride8_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm0
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm1 = xmm1[2,3,2,3]
; AVX1-ONLY-NEXT: vbroadcastsd 200(%rdx), %ymm3
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm4, %ymm0
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm1 = xmm5[2,3,2,3]
; AVX1-ONLY-NEXT: vbroadcastsd 200(%r10), %ymm2
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 208(%rsi), %xmm12
-; AVX1-ONLY-NEXT: vmovaps 208(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm12[0],ymm0[2],ymm12[2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],mem[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vbroadcastsd 208(%rcx), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 208(%r9), %xmm11
-; AVX1-ONLY-NEXT: vmovaps 208(%r8), %xmm0
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm11[0],ymm0[2],ymm11[2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],mem[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vbroadcastsd 208(%rax), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 224(%rsi), %xmm0
; AVX1-ONLY-NEXT: vmovaps 224(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
@@ -3375,164 +3237,251 @@ define void @store_i64_stride8_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm0
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm1 = xmm1[2,3,2,3]
; AVX1-ONLY-NEXT: vbroadcastsd 232(%rdx), %ymm3
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm4, %ymm0
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm1 = xmm5[2,3,2,3]
; AVX1-ONLY-NEXT: vbroadcastsd 232(%r10), %ymm2
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 240(%rsi), %xmm7
-; AVX1-ONLY-NEXT: vmovaps 240(%rdi), %xmm5
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm5[0],ymm7[0],ymm5[2],ymm7[2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],mem[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vbroadcastsd 240(%rcx), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, (%rsp) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 240(%r9), %xmm4
-; AVX1-ONLY-NEXT: vmovaps 240(%r8), %xmm3
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm3[0],ymm4[0],ymm3[2],ymm4[2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],mem[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vbroadcastsd 240(%rax), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm0 = xmm0[1],mem[1]
+; AVX1-ONLY-NEXT: vmovaps 16(%rsi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps 16(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vbroadcastsd 16(%rcx), %ymm3
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm3[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 16(%r9), %xmm2
+; AVX1-ONLY-NEXT: vmovaps 16(%r8), %xmm3
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm4 = xmm3[0],xmm2[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vbroadcastsd 16(%rax), %ymm5
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5],ymm5[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX1-ONLY-NEXT: vbroadcastsd 24(%rdx), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm9, %xmm0 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm0 = xmm9[1],mem[1]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm3[1],xmm2[1]
; AVX1-ONLY-NEXT: vbroadcastsd 24(%r10), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm10[1],xmm8[1]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm13 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps 48(%rsi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps 48(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vbroadcastsd 48(%rcx), %ymm3
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm3[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 48(%r9), %xmm2
+; AVX1-ONLY-NEXT: vmovaps 48(%r8), %xmm3
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm4 = xmm3[0],xmm2[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vbroadcastsd 48(%rax), %ymm5
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5],ymm5[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX1-ONLY-NEXT: vbroadcastsd 56(%rdx), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm0 = xmm0[1],mem[1]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm15 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm3[1],xmm2[1]
; AVX1-ONLY-NEXT: vbroadcastsd 56(%r10), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm10 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovlps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm0 = mem[0,1],xmm0[2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 80(%rsi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps 80(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vbroadcastsd 80(%rcx), %ymm3
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm3[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 80(%r9), %xmm2
+; AVX1-ONLY-NEXT: vmovaps 80(%r8), %xmm3
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm4 = xmm3[0],xmm2[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vbroadcastsd 80(%rax), %ymm5
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5],ymm5[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX1-ONLY-NEXT: vbroadcastsd 88(%rdx), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm9 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovlps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm0 = mem[0,1],xmm0[2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm3[1],xmm2[1]
; AVX1-ONLY-NEXT: vbroadcastsd 88(%r10), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovlps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm0 = mem[0,1],xmm0[2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 112(%rsi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps 112(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vbroadcastsd 112(%rcx), %ymm3
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm3[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 112(%r9), %xmm2
+; AVX1-ONLY-NEXT: vmovaps 112(%r8), %xmm3
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm4 = xmm3[0],xmm2[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vbroadcastsd 112(%rax), %ymm5
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5],ymm5[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX1-ONLY-NEXT: vbroadcastsd 120(%rdx), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovlps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm0 = mem[0,1],xmm0[2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm3[1],xmm2[1]
; AVX1-ONLY-NEXT: vbroadcastsd 120(%r10), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovlps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm0 = mem[0,1],xmm0[2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 144(%rsi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps 144(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vbroadcastsd 144(%rcx), %ymm3
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm3[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 144(%r9), %xmm2
+; AVX1-ONLY-NEXT: vmovaps 144(%r8), %xmm3
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm4 = xmm3[0],xmm2[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vbroadcastsd 144(%rax), %ymm5
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5],ymm5[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX1-ONLY-NEXT: vbroadcastsd 152(%rdx), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vmovlps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14, %xmm0 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm0 = mem[0,1],xmm14[2,3]
-; AVX1-ONLY-NEXT: vbroadcastsd 152(%r10), %ymm14
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm14[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vmovlps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm14 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm14 = mem[0,1],xmm15[2,3]
-; AVX1-ONLY-NEXT: vbroadcastsd 184(%rdx), %ymm15
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3],ymm15[4,5],ymm14[6,7]
-; AVX1-ONLY-NEXT: vmovlps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13, %xmm15 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm15 = mem[0,1],xmm13[2,3]
-; AVX1-ONLY-NEXT: vbroadcastsd 184(%r10), %ymm13
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm13 = ymm15[0,1,2,3],ymm13[4,5],ymm15[6,7]
-; AVX1-ONLY-NEXT: vmovlps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12, %xmm15 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm15 = mem[0,1],xmm12[2,3]
-; AVX1-ONLY-NEXT: vbroadcastsd 216(%rdx), %ymm12
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm15 = ymm15[0,1,2,3],ymm12[4,5],ymm15[6,7]
-; AVX1-ONLY-NEXT: vmovlps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11, %xmm12 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm12 = mem[0,1],xmm11[2,3]
-; AVX1-ONLY-NEXT: vbroadcastsd 216(%r10), %ymm11
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm11 = ymm12[0,1,2,3],ymm11[4,5],ymm12[6,7]
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm5 = xmm5[1],xmm7[1]
-; AVX1-ONLY-NEXT: vbroadcastsd 248(%rdx), %ymm7
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm5[0,1,2,3],ymm7[4,5],ymm5[6,7]
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm3 = xmm3[1],xmm4[1]
-; AVX1-ONLY-NEXT: vbroadcastsd 248(%r10), %ymm4
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm3[0,1,2,3],ymm4[4,5],ymm3[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm3[1],xmm2[1]
+; AVX1-ONLY-NEXT: vbroadcastsd 152(%r10), %ymm1
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 176(%rsi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps 176(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vbroadcastsd 176(%rcx), %ymm3
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm3[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 176(%r9), %xmm2
+; AVX1-ONLY-NEXT: vmovaps 176(%r8), %xmm3
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm4 = xmm3[0],xmm2[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vbroadcastsd 176(%rax), %ymm5
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5],ymm5[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT: vbroadcastsd 184(%rdx), %ymm1
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, (%rsp) # 32-byte Spill
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm3[1],xmm2[1]
+; AVX1-ONLY-NEXT: vbroadcastsd 184(%r10), %ymm1
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 208(%rsi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps 208(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vbroadcastsd 208(%rcx), %ymm3
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm3[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 208(%r9), %xmm2
+; AVX1-ONLY-NEXT: vmovaps 208(%r8), %xmm4
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm3 = xmm4[0],xmm2[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vbroadcastsd 208(%rax), %ymm5
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm14 = ymm3[0,1,2,3,4,5],ymm5[6,7]
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT: vbroadcastsd 216(%rdx), %ymm1
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm11 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm4[1],xmm2[1]
+; AVX1-ONLY-NEXT: vbroadcastsd 216(%r10), %ymm2
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm9 = ymm1[0,1,2,3],ymm2[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps 240(%rsi), %xmm1
+; AVX1-ONLY-NEXT: vmovaps 240(%rdi), %xmm2
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm4 = xmm2[0],xmm1[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vbroadcastsd 240(%rcx), %ymm6
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm4[0,1,2,3,4,5],ymm6[6,7]
+; AVX1-ONLY-NEXT: vmovaps 240(%r9), %xmm4
+; AVX1-ONLY-NEXT: vmovaps 240(%r8), %xmm6
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm10 = xmm6[0],xmm4[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm10 = ymm10[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vbroadcastsd 240(%rax), %ymm12
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm10[0,1,2,3,4,5],ymm12[6,7]
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm2[1],xmm1[1]
+; AVX1-ONLY-NEXT: vbroadcastsd 248(%rdx), %ymm2
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm1[0,1,2,3],ymm2[4,5,6,7]
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm6[1],xmm4[1]
+; AVX1-ONLY-NEXT: vbroadcastsd 248(%r10), %ymm2
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX1-ONLY-NEXT: movq {{[0-9]+}}(%rsp), %rdx
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],mem[6,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm12 = ymm5[0,1,2,3,4,5],mem[6,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5],mem[6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm10 = ymm10[0,1,2,3,4,5],mem[6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm9 = ymm9[0,1,2,3,4,5],mem[6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5],mem[6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5],mem[6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],mem[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm8[0,1,2,3,4,5],mem[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm13[0,1,2,3,4,5],mem[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm15[0,1,2,3,4,5],mem[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm15 = ymm1[0,1,2,3,4,5],mem[6,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm13 = ymm1[0,1,2,3,4,5],mem[6,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm12 = ymm1[0,1,2,3,4,5],mem[6,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm10 = ymm1[0,1,2,3,4,5],mem[6,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm1[0,1,2,3,4,5],mem[6,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm1[0,1,2,3,4,5],mem[6,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm1[0,1,2,3,4,5],mem[6,7]
+; AVX1-ONLY-NEXT: vmovups (%rsp), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm1[0,1,2,3,4,5],mem[6,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],mem[6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],mem[6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4,5],mem[6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3,4,5],mem[6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm15 = ymm15[0,1,2,3,4,5],mem[6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm11 = ymm11[0,1,2,3,4,5],mem[6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5],mem[6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5],mem[6,7]
-; AVX1-ONLY-NEXT: vmovaps %ymm4, 2016(%rdx)
-; AVX1-ONLY-NEXT: vmovaps %ymm7, 1984(%rdx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm4, 1952(%rdx)
-; AVX1-ONLY-NEXT: vmovups (%rsp), %ymm4 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm4, 1920(%rdx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm4, 1888(%rdx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm4, 1856(%rdx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm4, 1824(%rdx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm4, 1792(%rdx)
-; AVX1-ONLY-NEXT: vmovaps %ymm11, 1760(%rdx)
-; AVX1-ONLY-NEXT: vmovaps %ymm15, 1728(%rdx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm4, 1696(%rdx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm4, 1664(%rdx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm4, 1632(%rdx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm4, 1600(%rdx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm4, 1568(%rdx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm4, 1536(%rdx)
-; AVX1-ONLY-NEXT: vmovaps %ymm13, 1504(%rdx)
-; AVX1-ONLY-NEXT: vmovaps %ymm14, 1472(%rdx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm4, 1440(%rdx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm4, 1408(%rdx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm4, 1376(%rdx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm4, 1344(%rdx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm4, 1312(%rdx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm4, 1280(%rdx)
-; AVX1-ONLY-NEXT: vmovaps %ymm0, 1248(%rdx)
-; AVX1-ONLY-NEXT: vmovaps %ymm1, 1216(%rdx)
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm9 = ymm9[0,1,2,3,4,5],mem[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],mem[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],mem[6,7]
+; AVX1-ONLY-NEXT: vmovaps %ymm0, 2016(%rdx)
+; AVX1-ONLY-NEXT: vmovaps %ymm3, 1984(%rdx)
+; AVX1-ONLY-NEXT: vmovaps %ymm5, 1952(%rdx)
+; AVX1-ONLY-NEXT: vmovaps %ymm7, 1920(%rdx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm0, 1888(%rdx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm0, 1856(%rdx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm0, 1824(%rdx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm0, 1792(%rdx)
+; AVX1-ONLY-NEXT: vmovaps %ymm9, 1760(%rdx)
+; AVX1-ONLY-NEXT: vmovaps %ymm11, 1728(%rdx)
+; AVX1-ONLY-NEXT: vmovaps %ymm14, 1696(%rdx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm0, 1664(%rdx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm0, 1632(%rdx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm0, 1600(%rdx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm0, 1568(%rdx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm0, 1536(%rdx)
+; AVX1-ONLY-NEXT: vmovaps %ymm1, 1504(%rdx)
+; AVX1-ONLY-NEXT: vmovaps %ymm2, 1472(%rdx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm0, 1440(%rdx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm0, 1408(%rdx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm0, 1376(%rdx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm0, 1344(%rdx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm0, 1312(%rdx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm0, 1280(%rdx)
+; AVX1-ONLY-NEXT: vmovaps %ymm4, 1248(%rdx)
+; AVX1-ONLY-NEXT: vmovaps %ymm6, 1216(%rdx)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm0, 1184(%rdx)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
@@ -3545,8 +3494,8 @@ define void @store_i64_stride8_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vmovaps %ymm0, 1056(%rdx)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm0, 1024(%rdx)
-; AVX1-ONLY-NEXT: vmovaps %ymm2, 992(%rdx)
-; AVX1-ONLY-NEXT: vmovaps %ymm6, 960(%rdx)
+; AVX1-ONLY-NEXT: vmovaps %ymm8, 992(%rdx)
+; AVX1-ONLY-NEXT: vmovaps %ymm10, 960(%rdx)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm0, 928(%rdx)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
@@ -3559,8 +3508,8 @@ define void @store_i64_stride8_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vmovaps %ymm0, 800(%rdx)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm0, 768(%rdx)
-; AVX1-ONLY-NEXT: vmovaps %ymm8, 736(%rdx)
-; AVX1-ONLY-NEXT: vmovaps %ymm9, 704(%rdx)
+; AVX1-ONLY-NEXT: vmovaps %ymm12, 736(%rdx)
+; AVX1-ONLY-NEXT: vmovaps %ymm13, 704(%rdx)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm0, 672(%rdx)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
@@ -3573,8 +3522,9 @@ define void @store_i64_stride8_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vmovaps %ymm0, 544(%rdx)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm0, 512(%rdx)
-; AVX1-ONLY-NEXT: vmovaps %ymm10, 480(%rdx)
-; AVX1-ONLY-NEXT: vmovaps %ymm5, 448(%rdx)
+; AVX1-ONLY-NEXT: vmovaps %ymm15, 480(%rdx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm0, 448(%rdx)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm0, 416(%rdx)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
@@ -3587,8 +3537,10 @@ define void @store_i64_stride8_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vmovaps %ymm0, 288(%rdx)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm0, 256(%rdx)
-; AVX1-ONLY-NEXT: vmovaps %ymm12, 224(%rdx)
-; AVX1-ONLY-NEXT: vmovaps %ymm3, 192(%rdx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm0, 224(%rdx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm0, 192(%rdx)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm0, 160(%rdx)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
@@ -3601,7 +3553,7 @@ define void @store_i64_stride8_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vmovaps %ymm0, 32(%rdx)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm0, (%rdx)
-; AVX1-ONLY-NEXT: addq $2056, %rsp # imm = 0x808
+; AVX1-ONLY-NEXT: addq $1704, %rsp # imm = 0x6A8
; AVX1-ONLY-NEXT: vzeroupper
; AVX1-ONLY-NEXT: retq
;
@@ -3625,7 +3577,7 @@ define void @store_i64_stride8_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm2[1]
; AVX2-ONLY-NEXT: vbroadcastsd 8(%rdx), %ymm2
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovaps (%rax), %xmm0
@@ -3641,18 +3593,18 @@ define void @store_i64_stride8_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm2[1]
; AVX2-ONLY-NEXT: vbroadcastsd 8(%r10), %ymm2
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm0
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm5[1],xmm4[1]
; AVX2-ONLY-NEXT: vbroadcastsd 40(%rdx), %ymm2
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm7[1],xmm6[1]
; AVX2-ONLY-NEXT: vbroadcastsd 40(%r10), %ymm1
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT: vmovaps 32(%rax), %xmm1
; AVX2-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
@@ -3664,7 +3616,7 @@ define void @store_i64_stride8_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX2-ONLY-NEXT: vbroadcastsd 72(%rdx), %ymm1
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT: vmovaps 64(%rcx), %xmm1
; AVX2-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
@@ -3676,7 +3628,7 @@ define void @store_i64_stride8_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX2-ONLY-NEXT: vbroadcastsd 72(%r10), %ymm1
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT: vmovaps 64(%rax), %xmm1
; AVX2-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
@@ -3688,7 +3640,7 @@ define void @store_i64_stride8_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
; AVX2-ONLY-NEXT: vbroadcastsd 104(%rdx), %ymm1
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT: vmovaps 96(%rcx), %xmm1
; AVX2-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
@@ -3700,7 +3652,7 @@ define void @store_i64_stride8_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
; AVX2-ONLY-NEXT: vbroadcastsd 104(%r10), %ymm1
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT: vmovaps 96(%rax), %xmm1
; AVX2-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
@@ -3712,7 +3664,7 @@ define void @store_i64_stride8_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
; AVX2-ONLY-NEXT: vbroadcastsd 136(%rdx), %ymm1
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT: vmovaps 128(%rcx), %xmm1
; AVX2-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
@@ -3724,7 +3676,7 @@ define void @store_i64_stride8_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
; AVX2-ONLY-NEXT: vbroadcastsd 136(%r10), %ymm1
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT: vmovaps 128(%rax), %xmm1
; AVX2-ONLY-NEXT: vmovaps %xmm1, (%rsp) # 16-byte Spill
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
@@ -3736,7 +3688,7 @@ define void @store_i64_stride8_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
; AVX2-ONLY-NEXT: vbroadcastsd 168(%rdx), %ymm1
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT: vmovaps 160(%rcx), %xmm1
; AVX2-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
@@ -3747,7 +3699,7 @@ define void @store_i64_stride8_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovaps 160(%r8), %xmm13
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm13[1],xmm0[1]
; AVX2-ONLY-NEXT: vbroadcastsd 168(%r10), %ymm1
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT: vmovaps 160(%rax), %xmm12
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm12, %ymm0, %ymm1
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
@@ -3756,7 +3708,7 @@ define void @store_i64_stride8_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovaps 192(%rdi), %xmm10
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm10[1],xmm11[1]
; AVX2-ONLY-NEXT: vbroadcastsd 200(%rdx), %ymm1
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT: vmovaps 192(%rcx), %xmm9
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm9, %ymm0, %ymm1
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
@@ -3765,7 +3717,7 @@ define void @store_i64_stride8_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovaps 192(%r8), %xmm7
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm7[1],xmm8[1]
; AVX2-ONLY-NEXT: vbroadcastsd 200(%r10), %ymm1
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT: vmovaps 192(%rax), %xmm6
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm6, %ymm0, %ymm1
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
@@ -3774,7 +3726,7 @@ define void @store_i64_stride8_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovaps 224(%rdi), %xmm4
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm4[1],xmm5[1]
; AVX2-ONLY-NEXT: vbroadcastsd 232(%rdx), %ymm1
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT: vmovaps 224(%rcx), %xmm3
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm1
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
@@ -3783,7 +3735,7 @@ define void @store_i64_stride8_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovaps 224(%r8), %xmm1
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm2[1]
; AVX2-ONLY-NEXT: vbroadcastsd 232(%r10), %ymm15
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm15 = ymm0[0,1,2,3],ymm15[4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm15 = ymm0[0,1,2,3],ymm15[4,5,6,7]
; AVX2-ONLY-NEXT: vmovaps 224(%rax), %xmm0
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm14
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm14 = ymm15[0,1,2,3,4,5],ymm14[6,7]
@@ -6607,147 +6559,93 @@ define void @store_i64_stride8_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
;
; AVX1-ONLY-LABEL: store_i64_stride8_vf64:
; AVX1-ONLY: # %bb.0:
-; AVX1-ONLY-NEXT: subq $4712, %rsp # imm = 0x1268
+; AVX1-ONLY-NEXT: subq $3784, %rsp # imm = 0xEC8
; AVX1-ONLY-NEXT: movq {{[0-9]+}}(%rsp), %r10
; AVX1-ONLY-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; AVX1-ONLY-NEXT: vmovaps (%rsi), %xmm2
-; AVX1-ONLY-NEXT: vmovaps 16(%rsi), %xmm10
-; AVX1-ONLY-NEXT: vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 32(%rsi), %xmm0
-; AVX1-ONLY-NEXT: vmovaps (%rdi), %xmm3
-; AVX1-ONLY-NEXT: vmovaps 16(%rdi), %xmm11
-; AVX1-ONLY-NEXT: vmovups %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 32(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm4 = xmm3[0],xmm2[0]
-; AVX1-ONLY-NEXT: vmovaps (%rcx), %xmm5
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm5, %ymm4, %ymm6
-; AVX1-ONLY-NEXT: vinsertf128 $1, (%rdx), %ymm4, %ymm4
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm4 = ymm4[0],ymm6[1],ymm4[2],ymm6[2]
-; AVX1-ONLY-NEXT: vmovupd %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps (%r9), %xmm4
-; AVX1-ONLY-NEXT: vmovaps (%r8), %xmm6
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm7 = xmm6[0],xmm4[0]
-; AVX1-ONLY-NEXT: vmovaps (%rax), %xmm8
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm8, %ymm7, %ymm9
-; AVX1-ONLY-NEXT: vinsertf128 $1, (%r10), %ymm7, %ymm7
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm7 = ymm7[0],ymm9[1],ymm7[2],ymm9[2]
-; AVX1-ONLY-NEXT: vmovupd %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm5, %ymm2, %ymm2
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm3 = xmm3[2,3,2,3]
-; AVX1-ONLY-NEXT: vbroadcastsd 8(%rdx), %ymm5
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm5[4,5],ymm3[6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3],ymm3[4,5],ymm2[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm8, %ymm4, %ymm2
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm3 = xmm6[2,3,2,3]
-; AVX1-ONLY-NEXT: vbroadcastsd 8(%r10), %ymm4
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm4[4,5],ymm3[6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3],ymm3[4,5],ymm2[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm2 = ymm11[0],ymm10[0],ymm11[2],ymm10[2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],mem[4,5],ymm2[6,7]
-; AVX1-ONLY-NEXT: vbroadcastsd 16(%rcx), %ymm3
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm3[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 16(%r9), %xmm3
-; AVX1-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 32(%r9), %xmm2
-; AVX1-ONLY-NEXT: vmovaps 16(%r8), %xmm4
-; AVX1-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm3 = ymm4[0],ymm3[0],ymm4[2],ymm3[2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],mem[4,5],ymm3[6,7]
-; AVX1-ONLY-NEXT: vbroadcastsd 16(%rax), %ymm4
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm4[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm3 = xmm1[0],xmm0[0]
-; AVX1-ONLY-NEXT: vmovaps 32(%rcx), %xmm4
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm4, %ymm3, %ymm5
-; AVX1-ONLY-NEXT: vinsertf128 $1, 32(%rdx), %ymm3, %ymm3
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm3 = ymm3[0],ymm5[1],ymm3[2],ymm5[2]
-; AVX1-ONLY-NEXT: vmovupd %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 32(%r8), %xmm3
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm5 = xmm3[0],xmm2[0]
-; AVX1-ONLY-NEXT: vmovaps 32(%rax), %xmm6
+; AVX1-ONLY-NEXT: vmovaps (%rsi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps (%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT: vmovaps (%rcx), %xmm3
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm2, %ymm4
+; AVX1-ONLY-NEXT: vinsertf128 $1, (%rdx), %ymm2, %ymm2
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm2 = ymm2[0],ymm4[1],ymm2[2],ymm4[2]
+; AVX1-ONLY-NEXT: vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps (%r9), %xmm2
+; AVX1-ONLY-NEXT: vmovaps (%r8), %xmm4
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm5 = xmm4[0],xmm2[0]
+; AVX1-ONLY-NEXT: vmovaps (%rax), %xmm6
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm6, %ymm5, %ymm7
-; AVX1-ONLY-NEXT: vinsertf128 $1, 32(%r10), %ymm5, %ymm5
+; AVX1-ONLY-NEXT: vinsertf128 $1, (%r10), %ymm5, %ymm5
; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm5 = ymm5[0],ymm7[1],ymm5[2],ymm7[2]
; AVX1-ONLY-NEXT: vmovupd %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm4, %ymm0, %ymm0
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm0
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm1 = xmm1[2,3,2,3]
-; AVX1-ONLY-NEXT: vbroadcastsd 40(%rdx), %ymm4
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm4[4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT: vbroadcastsd 8(%rdx), %ymm3
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm6, %ymm2, %ymm0
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm1 = xmm3[2,3,2,3]
-; AVX1-ONLY-NEXT: vbroadcastsd 40(%r10), %ymm2
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT: vmovaps 48(%rsi), %xmm2
-; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm1 = xmm4[2,3,2,3]
+; AVX1-ONLY-NEXT: vbroadcastsd 8(%r10), %ymm2
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 48(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm2[0],ymm0[2],ymm2[2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],mem[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vbroadcastsd 48(%rcx), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 48(%r9), %xmm0
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 48(%r8), %xmm1
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],mem[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vbroadcastsd 48(%rax), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 64(%rcx), %xmm0
-; AVX1-ONLY-NEXT: vmovaps 64(%rsi), %xmm1
-; AVX1-ONLY-NEXT: vmovaps 64(%rdi), %xmm2
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm3 = xmm2[0],xmm1[0]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm3, %ymm4
-; AVX1-ONLY-NEXT: vinsertf128 $1, 64(%rdx), %ymm3, %ymm3
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm3 = ymm3[0],ymm4[1],ymm3[2],ymm4[2]
+; AVX1-ONLY-NEXT: vmovaps 32(%rsi), %xmm4
+; AVX1-ONLY-NEXT: vmovaps 32(%rdi), %xmm5
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm0 = xmm5[0],xmm4[0]
+; AVX1-ONLY-NEXT: vmovaps 32(%rcx), %xmm6
+; AVX1-ONLY-NEXT: vmovaps 64(%rcx), %xmm2
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm6, %ymm0, %ymm1
+; AVX1-ONLY-NEXT: vinsertf128 $1, 32(%rdx), %ymm0, %ymm0
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[2]
+; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 32(%r9), %xmm7
+; AVX1-ONLY-NEXT: vmovaps 64(%r9), %xmm0
+; AVX1-ONLY-NEXT: vmovaps 32(%r8), %xmm8
+; AVX1-ONLY-NEXT: vmovaps 64(%r8), %xmm1
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm3 = xmm8[0],xmm7[0]
+; AVX1-ONLY-NEXT: vmovaps 32(%rax), %xmm9
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm9, %ymm3, %ymm10
+; AVX1-ONLY-NEXT: vinsertf128 $1, 32(%r10), %ymm3, %ymm3
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm3 = ymm3[0],ymm10[1],ymm3[2],ymm10[2]
; AVX1-ONLY-NEXT: vmovupd %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 64(%rax), %xmm3
-; AVX1-ONLY-NEXT: vmovaps 64(%r9), %xmm4
-; AVX1-ONLY-NEXT: vmovaps 64(%r8), %xmm5
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm6, %ymm4, %ymm4
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm5 = xmm5[2,3,2,3]
+; AVX1-ONLY-NEXT: vbroadcastsd 40(%rdx), %ymm6
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm6[4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm5[0,1],ymm4[2,3],ymm5[4,5],ymm4[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm9, %ymm7, %ymm4
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm5 = xmm8[2,3,2,3]
+; AVX1-ONLY-NEXT: vbroadcastsd 40(%r10), %ymm6
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm6[4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm5[0,1],ymm4[2,3],ymm5[4,5],ymm4[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 64(%rsi), %xmm4
+; AVX1-ONLY-NEXT: vmovaps 64(%rdi), %xmm5
; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm6 = xmm5[0],xmm4[0]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm6, %ymm7
+; AVX1-ONLY-NEXT: vinsertf128 $1, 64(%rdx), %ymm6, %ymm6
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm6 = ymm6[0],ymm7[1],ymm6[2],ymm7[2]
+; AVX1-ONLY-NEXT: vmovupd %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm6 = xmm1[0],xmm0[0]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm6, %ymm7
; AVX1-ONLY-NEXT: vinsertf128 $1, 64(%r10), %ymm6, %ymm6
; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm6 = ymm6[0],ymm7[1],ymm6[2],ymm7[2]
; AVX1-ONLY-NEXT: vmovupd %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm1 = xmm2[2,3,2,3]
-; AVX1-ONLY-NEXT: vbroadcastsd 72(%rdx), %ymm2
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm4, %ymm0
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm1 = xmm5[2,3,2,3]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm4, %ymm2
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm4 = xmm5[2,3,2,3]
+; AVX1-ONLY-NEXT: vbroadcastsd 72(%rdx), %ymm5
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm5[4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm4[0,1],ymm2[2,3],ymm4[4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm0
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm1 = xmm1[2,3,2,3]
; AVX1-ONLY-NEXT: vbroadcastsd 72(%r10), %ymm2
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 80(%rsi), %xmm1
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 80(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],mem[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vbroadcastsd 80(%rcx), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 80(%r9), %xmm1
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 80(%r8), %xmm0
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],mem[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vbroadcastsd 80(%rax), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 96(%rsi), %xmm0
; AVX1-ONLY-NEXT: vmovaps 96(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
@@ -6767,33 +6665,15 @@ define void @store_i64_stride8_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm0
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm1 = xmm1[2,3,2,3]
; AVX1-ONLY-NEXT: vbroadcastsd 104(%rdx), %ymm3
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm6, %ymm2, %ymm0
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm1 = xmm4[2,3,2,3]
; AVX1-ONLY-NEXT: vbroadcastsd 104(%r10), %ymm2
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 112(%rsi), %xmm1
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 112(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],mem[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vbroadcastsd 112(%rcx), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 112(%r9), %xmm1
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 112(%r8), %xmm0
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],mem[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vbroadcastsd 112(%rax), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 128(%rsi), %xmm0
; AVX1-ONLY-NEXT: vmovaps 128(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
@@ -6813,33 +6693,15 @@ define void @store_i64_stride8_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm0
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm1 = xmm1[2,3,2,3]
; AVX1-ONLY-NEXT: vbroadcastsd 136(%rdx), %ymm3
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm6, %ymm2, %ymm0
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm1 = xmm4[2,3,2,3]
; AVX1-ONLY-NEXT: vbroadcastsd 136(%r10), %ymm2
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 144(%rsi), %xmm1
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 144(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],mem[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vbroadcastsd 144(%rcx), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 144(%r9), %xmm1
-; AVX1-ONLY-NEXT: vmovups %ymm1, (%rsp) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 144(%r8), %xmm0
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],mem[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vbroadcastsd 144(%rax), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 160(%rcx), %xmm0
; AVX1-ONLY-NEXT: vmovaps 160(%rsi), %xmm1
; AVX1-ONLY-NEXT: vmovaps 160(%rdi), %xmm2
@@ -6848,44 +6710,26 @@ define void @store_i64_stride8_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vinsertf128 $1, 160(%rdx), %ymm3, %ymm3
; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm3 = ymm3[0],ymm4[1],ymm3[2],ymm4[2]
; AVX1-ONLY-NEXT: vmovupd %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 160(%rax), %xmm3
-; AVX1-ONLY-NEXT: vmovaps 160(%r9), %xmm4
-; AVX1-ONLY-NEXT: vmovaps 160(%r8), %xmm5
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm6 = xmm5[0],xmm4[0]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm6, %ymm7
+; AVX1-ONLY-NEXT: vmovaps 160(%r9), %xmm3
+; AVX1-ONLY-NEXT: vmovaps 160(%r8), %xmm4
+; AVX1-ONLY-NEXT: vmovaps 160(%rax), %xmm5
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm6 = xmm4[0],xmm3[0]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm5, %ymm6, %ymm7
; AVX1-ONLY-NEXT: vinsertf128 $1, 160(%r10), %ymm6, %ymm6
; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm6 = ymm6[0],ymm7[1],ymm6[2],ymm7[2]
; AVX1-ONLY-NEXT: vmovupd %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm1 = xmm2[2,3,2,3]
; AVX1-ONLY-NEXT: vbroadcastsd 168(%rdx), %ymm2
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm4, %ymm0
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm1 = xmm5[2,3,2,3]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm5, %ymm3, %ymm0
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm1 = xmm4[2,3,2,3]
; AVX1-ONLY-NEXT: vbroadcastsd 168(%r10), %ymm2
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 176(%rsi), %xmm1
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 176(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],mem[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vbroadcastsd 176(%rcx), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 176(%r9), %xmm1
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 176(%r8), %xmm0
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],mem[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vbroadcastsd 176(%rax), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 192(%rsi), %xmm0
; AVX1-ONLY-NEXT: vmovaps 192(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
@@ -6905,33 +6749,15 @@ define void @store_i64_stride8_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm0
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm1 = xmm1[2,3,2,3]
; AVX1-ONLY-NEXT: vbroadcastsd 200(%rdx), %ymm3
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm6, %ymm2, %ymm0
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm1 = xmm4[2,3,2,3]
; AVX1-ONLY-NEXT: vbroadcastsd 200(%r10), %ymm2
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 208(%rsi), %xmm1
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 208(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],mem[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vbroadcastsd 208(%rcx), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 208(%r9), %xmm1
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 208(%r8), %xmm0
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],mem[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vbroadcastsd 208(%rax), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 224(%rcx), %xmm0
; AVX1-ONLY-NEXT: vmovaps 224(%rsi), %xmm1
; AVX1-ONLY-NEXT: vmovaps 224(%rdi), %xmm2
@@ -6940,44 +6766,26 @@ define void @store_i64_stride8_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vinsertf128 $1, 224(%rdx), %ymm3, %ymm3
; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm3 = ymm3[0],ymm4[1],ymm3[2],ymm4[2]
; AVX1-ONLY-NEXT: vmovupd %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 224(%rax), %xmm3
-; AVX1-ONLY-NEXT: vmovaps 224(%r9), %xmm4
-; AVX1-ONLY-NEXT: vmovaps 224(%r8), %xmm5
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm6 = xmm5[0],xmm4[0]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm6, %ymm7
+; AVX1-ONLY-NEXT: vmovaps 224(%r9), %xmm3
+; AVX1-ONLY-NEXT: vmovaps 224(%r8), %xmm4
+; AVX1-ONLY-NEXT: vmovaps 224(%rax), %xmm5
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm6 = xmm4[0],xmm3[0]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm5, %ymm6, %ymm7
; AVX1-ONLY-NEXT: vinsertf128 $1, 224(%r10), %ymm6, %ymm6
; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm6 = ymm6[0],ymm7[1],ymm6[2],ymm7[2]
; AVX1-ONLY-NEXT: vmovupd %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm1 = xmm2[2,3,2,3]
; AVX1-ONLY-NEXT: vbroadcastsd 232(%rdx), %ymm2
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm4, %ymm0
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm1 = xmm5[2,3,2,3]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm5, %ymm3, %ymm0
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm1 = xmm4[2,3,2,3]
; AVX1-ONLY-NEXT: vbroadcastsd 232(%r10), %ymm2
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 240(%rsi), %xmm1
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 240(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],mem[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vbroadcastsd 240(%rcx), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 240(%r9), %xmm1
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 240(%r8), %xmm0
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],mem[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vbroadcastsd 240(%rax), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 256(%rsi), %xmm0
; AVX1-ONLY-NEXT: vmovaps 256(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
@@ -6997,33 +6805,15 @@ define void @store_i64_stride8_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm0
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm1 = xmm1[2,3,2,3]
; AVX1-ONLY-NEXT: vbroadcastsd 264(%rdx), %ymm3
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm6, %ymm2, %ymm0
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm1 = xmm4[2,3,2,3]
; AVX1-ONLY-NEXT: vbroadcastsd 264(%r10), %ymm2
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 272(%rsi), %xmm1
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 272(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],mem[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vbroadcastsd 272(%rcx), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 272(%r9), %xmm1
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 272(%r8), %xmm0
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],mem[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vbroadcastsd 272(%rax), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 288(%rsi), %xmm0
; AVX1-ONLY-NEXT: vmovaps 288(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
@@ -7043,32 +6833,15 @@ define void @store_i64_stride8_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm0
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm1 = xmm1[2,3,2,3]
; AVX1-ONLY-NEXT: vbroadcastsd 296(%rdx), %ymm3
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm6, %ymm2, %ymm0
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm1 = xmm4[2,3,2,3]
; AVX1-ONLY-NEXT: vbroadcastsd 296(%r10), %ymm2
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 304(%rsi), %xmm1
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 304(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],mem[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vbroadcastsd 304(%rcx), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 304(%r9), %xmm11
-; AVX1-ONLY-NEXT: vmovaps 304(%r8), %xmm0
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm11[0],ymm0[2],ymm11[2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],mem[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vbroadcastsd 304(%rax), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 320(%rsi), %xmm0
; AVX1-ONLY-NEXT: vmovaps 320(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
@@ -7088,33 +6861,15 @@ define void @store_i64_stride8_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm0
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm1 = xmm1[2,3,2,3]
; AVX1-ONLY-NEXT: vbroadcastsd 328(%rdx), %ymm3
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm6, %ymm2, %ymm0
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm1 = xmm4[2,3,2,3]
; AVX1-ONLY-NEXT: vbroadcastsd 328(%r10), %ymm2
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 336(%rsi), %xmm1
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 336(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],mem[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vbroadcastsd 336(%rcx), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 336(%r9), %xmm1
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 336(%r8), %xmm0
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],mem[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vbroadcastsd 336(%rax), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 352(%rcx), %xmm0
; AVX1-ONLY-NEXT: vmovaps 352(%rsi), %xmm1
; AVX1-ONLY-NEXT: vmovaps 352(%rdi), %xmm2
@@ -7123,44 +6878,26 @@ define void @store_i64_stride8_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vinsertf128 $1, 352(%rdx), %ymm3, %ymm3
; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm3 = ymm3[0],ymm4[1],ymm3[2],ymm4[2]
; AVX1-ONLY-NEXT: vmovupd %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 352(%rax), %xmm3
-; AVX1-ONLY-NEXT: vmovaps 352(%r9), %xmm4
-; AVX1-ONLY-NEXT: vmovaps 352(%r8), %xmm5
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm6 = xmm5[0],xmm4[0]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm6, %ymm7
+; AVX1-ONLY-NEXT: vmovaps 352(%r9), %xmm3
+; AVX1-ONLY-NEXT: vmovaps 352(%r8), %xmm4
+; AVX1-ONLY-NEXT: vmovaps 352(%rax), %xmm5
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm6 = xmm4[0],xmm3[0]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm5, %ymm6, %ymm7
; AVX1-ONLY-NEXT: vinsertf128 $1, 352(%r10), %ymm6, %ymm6
; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm6 = ymm6[0],ymm7[1],ymm6[2],ymm7[2]
; AVX1-ONLY-NEXT: vmovupd %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm1 = xmm2[2,3,2,3]
; AVX1-ONLY-NEXT: vbroadcastsd 360(%rdx), %ymm2
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm4, %ymm0
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm1 = xmm5[2,3,2,3]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm5, %ymm3, %ymm0
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm1 = xmm4[2,3,2,3]
; AVX1-ONLY-NEXT: vbroadcastsd 360(%r10), %ymm2
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 368(%rsi), %xmm1
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 368(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],mem[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vbroadcastsd 368(%rcx), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 368(%r9), %xmm1
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 368(%r8), %xmm0
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],mem[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vbroadcastsd 368(%rax), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 384(%rcx), %xmm0
; AVX1-ONLY-NEXT: vmovaps 384(%rsi), %xmm1
; AVX1-ONLY-NEXT: vmovaps 384(%rdi), %xmm2
@@ -7169,42 +6906,26 @@ define void @store_i64_stride8_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vinsertf128 $1, 384(%rdx), %ymm3, %ymm3
; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm3 = ymm3[0],ymm4[1],ymm3[2],ymm4[2]
; AVX1-ONLY-NEXT: vmovupd %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 384(%rax), %xmm3
-; AVX1-ONLY-NEXT: vmovaps 384(%r9), %xmm4
-; AVX1-ONLY-NEXT: vmovaps 384(%r8), %xmm5
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm6 = xmm5[0],xmm4[0]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm6, %ymm7
+; AVX1-ONLY-NEXT: vmovaps 384(%r9), %xmm3
+; AVX1-ONLY-NEXT: vmovaps 384(%r8), %xmm4
+; AVX1-ONLY-NEXT: vmovaps 384(%rax), %xmm5
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm6 = xmm4[0],xmm3[0]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm5, %ymm6, %ymm7
; AVX1-ONLY-NEXT: vinsertf128 $1, 384(%r10), %ymm6, %ymm6
; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm6 = ymm6[0],ymm7[1],ymm6[2],ymm7[2]
; AVX1-ONLY-NEXT: vmovupd %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm1 = xmm2[2,3,2,3]
; AVX1-ONLY-NEXT: vbroadcastsd 392(%rdx), %ymm2
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm4, %ymm0
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm1 = xmm5[2,3,2,3]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm5, %ymm3, %ymm0
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm1 = xmm4[2,3,2,3]
; AVX1-ONLY-NEXT: vbroadcastsd 392(%r10), %ymm2
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 400(%rsi), %xmm14
-; AVX1-ONLY-NEXT: vmovaps 400(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm14[0],ymm0[2],ymm14[2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],mem[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vbroadcastsd 400(%rcx), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 400(%r9), %xmm12
-; AVX1-ONLY-NEXT: vmovaps 400(%r8), %xmm0
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm12[0],ymm0[2],ymm12[2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],mem[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vbroadcastsd 400(%rax), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 416(%rcx), %xmm0
; AVX1-ONLY-NEXT: vmovaps 416(%rsi), %xmm1
; AVX1-ONLY-NEXT: vmovaps 416(%rdi), %xmm2
@@ -7213,42 +6934,26 @@ define void @store_i64_stride8_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vinsertf128 $1, 416(%rdx), %ymm3, %ymm3
; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm3 = ymm3[0],ymm4[1],ymm3[2],ymm4[2]
; AVX1-ONLY-NEXT: vmovupd %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 416(%rax), %xmm3
-; AVX1-ONLY-NEXT: vmovaps 416(%r9), %xmm4
-; AVX1-ONLY-NEXT: vmovaps 416(%r8), %xmm5
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm6 = xmm5[0],xmm4[0]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm6, %ymm7
+; AVX1-ONLY-NEXT: vmovaps 416(%r9), %xmm3
+; AVX1-ONLY-NEXT: vmovaps 416(%r8), %xmm4
+; AVX1-ONLY-NEXT: vmovaps 416(%rax), %xmm5
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm6 = xmm4[0],xmm3[0]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm5, %ymm6, %ymm7
; AVX1-ONLY-NEXT: vinsertf128 $1, 416(%r10), %ymm6, %ymm6
; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm6 = ymm6[0],ymm7[1],ymm6[2],ymm7[2]
; AVX1-ONLY-NEXT: vmovupd %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm1 = xmm2[2,3,2,3]
; AVX1-ONLY-NEXT: vbroadcastsd 424(%rdx), %ymm2
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm4, %ymm0
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm1 = xmm5[2,3,2,3]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm5, %ymm3, %ymm0
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm1 = xmm4[2,3,2,3]
; AVX1-ONLY-NEXT: vbroadcastsd 424(%r10), %ymm2
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 432(%rsi), %xmm9
-; AVX1-ONLY-NEXT: vmovaps 432(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm9[0],ymm0[2],ymm9[2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],mem[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vbroadcastsd 432(%rcx), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 432(%r9), %xmm8
-; AVX1-ONLY-NEXT: vmovaps 432(%r8), %xmm0
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm8[0],ymm0[2],ymm8[2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],mem[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vbroadcastsd 432(%rax), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 448(%rcx), %xmm0
; AVX1-ONLY-NEXT: vmovaps 448(%rsi), %xmm1
; AVX1-ONLY-NEXT: vmovaps 448(%rdi), %xmm2
@@ -7257,43 +6962,26 @@ define void @store_i64_stride8_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vinsertf128 $1, 448(%rdx), %ymm3, %ymm3
; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm3 = ymm3[0],ymm4[1],ymm3[2],ymm4[2]
; AVX1-ONLY-NEXT: vmovupd %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 448(%rax), %xmm3
-; AVX1-ONLY-NEXT: vmovaps 448(%r9), %xmm4
-; AVX1-ONLY-NEXT: vmovaps 448(%r8), %xmm5
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm6 = xmm5[0],xmm4[0]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm6, %ymm7
+; AVX1-ONLY-NEXT: vmovaps 448(%r9), %xmm3
+; AVX1-ONLY-NEXT: vmovaps 448(%r8), %xmm4
+; AVX1-ONLY-NEXT: vmovaps 448(%rax), %xmm5
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm6 = xmm4[0],xmm3[0]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm5, %ymm6, %ymm7
; AVX1-ONLY-NEXT: vinsertf128 $1, 448(%r10), %ymm6, %ymm6
; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm6 = ymm6[0],ymm7[1],ymm6[2],ymm7[2]
; AVX1-ONLY-NEXT: vmovupd %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm1 = xmm2[2,3,2,3]
; AVX1-ONLY-NEXT: vbroadcastsd 456(%rdx), %ymm2
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm4, %ymm0
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm1 = xmm5[2,3,2,3]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm5, %ymm3, %ymm0
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm1 = xmm4[2,3,2,3]
; AVX1-ONLY-NEXT: vbroadcastsd 456(%r10), %ymm2
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 464(%rsi), %xmm15
-; AVX1-ONLY-NEXT: vmovaps 464(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm15[0],ymm0[2],ymm15[2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],mem[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vbroadcastsd 464(%rcx), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 464(%r9), %xmm1
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 464(%r8), %xmm0
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],mem[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vbroadcastsd 464(%rax), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 480(%rcx), %xmm0
; AVX1-ONLY-NEXT: vmovaps 480(%rsi), %xmm1
; AVX1-ONLY-NEXT: vmovaps 480(%rdi), %xmm2
@@ -7302,313 +6990,473 @@ define void @store_i64_stride8_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vinsertf128 $1, 480(%rdx), %ymm3, %ymm3
; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm3 = ymm3[0],ymm4[1],ymm3[2],ymm4[2]
; AVX1-ONLY-NEXT: vmovupd %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 480(%rax), %xmm3
-; AVX1-ONLY-NEXT: vmovaps 480(%r9), %xmm4
-; AVX1-ONLY-NEXT: vmovaps 480(%r8), %xmm5
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm6 = xmm5[0],xmm4[0]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm6, %ymm7
+; AVX1-ONLY-NEXT: vmovaps 480(%r9), %xmm3
+; AVX1-ONLY-NEXT: vmovaps 480(%r8), %xmm4
+; AVX1-ONLY-NEXT: vmovaps 480(%rax), %xmm5
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm6 = xmm4[0],xmm3[0]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm5, %ymm6, %ymm7
; AVX1-ONLY-NEXT: vinsertf128 $1, 480(%r10), %ymm6, %ymm6
; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm6 = ymm6[0],ymm7[1],ymm6[2],ymm7[2]
; AVX1-ONLY-NEXT: vmovupd %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm1 = xmm2[2,3,2,3]
; AVX1-ONLY-NEXT: vbroadcastsd 488(%rdx), %ymm2
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm4, %ymm0
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm1 = xmm5[2,3,2,3]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm5, %ymm3, %ymm0
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm1 = xmm4[2,3,2,3]
; AVX1-ONLY-NEXT: vbroadcastsd 488(%r10), %ymm2
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 496(%rsi), %xmm7
-; AVX1-ONLY-NEXT: vmovaps 496(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm7[0],ymm0[2],ymm7[2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],mem[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vbroadcastsd 496(%rcx), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 496(%r9), %xmm10
-; AVX1-ONLY-NEXT: vmovaps 496(%r8), %xmm13
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm13[0],ymm10[0],ymm13[2],ymm10[2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],mem[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vbroadcastsd 496(%rax), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm0 = xmm0[1],mem[1]
+; AVX1-ONLY-NEXT: vmovaps 16(%rsi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps 16(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vbroadcastsd 16(%rcx), %ymm3
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm3[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 16(%r9), %xmm2
+; AVX1-ONLY-NEXT: vmovaps 16(%r8), %xmm3
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm4 = xmm3[0],xmm2[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vbroadcastsd 16(%rax), %ymm5
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5],ymm5[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX1-ONLY-NEXT: vbroadcastsd 24(%rdx), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm0 = xmm0[1],mem[1]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm10 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm3[1],xmm2[1]
; AVX1-ONLY-NEXT: vbroadcastsd 24(%r10), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm0 = xmm0[1],mem[1]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm13 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps 48(%rsi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps 48(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vbroadcastsd 48(%rcx), %ymm3
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm3[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 48(%r9), %xmm2
+; AVX1-ONLY-NEXT: vmovaps 48(%r8), %xmm3
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm4 = xmm3[0],xmm2[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vbroadcastsd 48(%rax), %ymm5
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5],ymm5[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX1-ONLY-NEXT: vbroadcastsd 56(%rdx), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm0 = xmm0[1],mem[1]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm14 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm3[1],xmm2[1]
; AVX1-ONLY-NEXT: vbroadcastsd 56(%r10), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovlps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm0 = mem[0,1],xmm0[2,3]
+; AVX1-ONLY-NEXT: vmovaps 80(%rsi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps 80(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vbroadcastsd 80(%rcx), %ymm3
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm3[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 80(%r9), %xmm2
+; AVX1-ONLY-NEXT: vmovaps 80(%r8), %xmm3
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm4 = xmm3[0],xmm2[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vbroadcastsd 80(%rax), %ymm5
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5],ymm5[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX1-ONLY-NEXT: vbroadcastsd 88(%rdx), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovlps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm0 = mem[0,1],xmm0[2,3]
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm3[1],xmm2[1]
; AVX1-ONLY-NEXT: vbroadcastsd 88(%r10), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovlps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm0 = mem[0,1],xmm0[2,3]
+; AVX1-ONLY-NEXT: vmovaps 112(%rsi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps 112(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vbroadcastsd 112(%rcx), %ymm3
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm3[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 112(%r9), %xmm2
+; AVX1-ONLY-NEXT: vmovaps 112(%r8), %xmm3
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm4 = xmm3[0],xmm2[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vbroadcastsd 112(%rax), %ymm5
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5],ymm5[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX1-ONLY-NEXT: vbroadcastsd 120(%rdx), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovlps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm0 = mem[0,1],xmm0[2,3]
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm3[1],xmm2[1]
; AVX1-ONLY-NEXT: vbroadcastsd 120(%r10), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovlps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm0 = mem[0,1],xmm0[2,3]
+; AVX1-ONLY-NEXT: vmovaps 144(%rsi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps 144(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vbroadcastsd 144(%rcx), %ymm3
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm3[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 144(%r9), %xmm2
+; AVX1-ONLY-NEXT: vmovaps 144(%r8), %xmm3
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm4 = xmm3[0],xmm2[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vbroadcastsd 144(%rax), %ymm5
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5],ymm5[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX1-ONLY-NEXT: vbroadcastsd 152(%rdx), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups (%rsp), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovlps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm0 = mem[0,1],xmm0[2,3]
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm3[1],xmm2[1]
; AVX1-ONLY-NEXT: vbroadcastsd 152(%r10), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovlps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm0 = mem[0,1],xmm0[2,3]
+; AVX1-ONLY-NEXT: vmovaps 176(%rsi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps 176(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vbroadcastsd 176(%rcx), %ymm3
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm3[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 176(%r9), %xmm2
+; AVX1-ONLY-NEXT: vmovaps 176(%r8), %xmm3
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm4 = xmm3[0],xmm2[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vbroadcastsd 176(%rax), %ymm5
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5],ymm5[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX1-ONLY-NEXT: vbroadcastsd 184(%rdx), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovlps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm0 = mem[0,1],xmm0[2,3]
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm3[1],xmm2[1]
; AVX1-ONLY-NEXT: vbroadcastsd 184(%r10), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, (%rsp) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovlps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm0 = mem[0,1],xmm0[2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 208(%rsi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps 208(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vbroadcastsd 208(%rcx), %ymm3
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm3[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 208(%r9), %xmm2
+; AVX1-ONLY-NEXT: vmovaps 208(%r8), %xmm3
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm4 = xmm3[0],xmm2[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vbroadcastsd 208(%rax), %ymm5
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5],ymm5[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX1-ONLY-NEXT: vbroadcastsd 216(%rdx), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovlps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm0 = mem[0,1],xmm0[2,3]
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm3[1],xmm2[1]
; AVX1-ONLY-NEXT: vbroadcastsd 216(%r10), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovlps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm0 = mem[0,1],xmm0[2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, (%rsp) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 240(%rsi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps 240(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vbroadcastsd 240(%rcx), %ymm3
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm3[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 240(%r9), %xmm2
+; AVX1-ONLY-NEXT: vmovaps 240(%r8), %xmm3
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm4 = xmm3[0],xmm2[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vbroadcastsd 240(%rax), %ymm5
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5],ymm5[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX1-ONLY-NEXT: vbroadcastsd 248(%rdx), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovlps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm0 = mem[0,1],xmm0[2,3]
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm3[1],xmm2[1]
; AVX1-ONLY-NEXT: vbroadcastsd 248(%r10), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovlps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm0 = mem[0,1],xmm0[2,3]
+; AVX1-ONLY-NEXT: vmovaps 272(%rsi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps 272(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vbroadcastsd 272(%rcx), %ymm3
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm3[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 272(%r9), %xmm2
+; AVX1-ONLY-NEXT: vmovaps 272(%r8), %xmm3
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm4 = xmm3[0],xmm2[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vbroadcastsd 272(%rax), %ymm5
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5],ymm5[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX1-ONLY-NEXT: vbroadcastsd 280(%rdx), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovlps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm0 = mem[0,1],xmm0[2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm3[1],xmm2[1]
; AVX1-ONLY-NEXT: vbroadcastsd 280(%r10), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovlps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm0 = mem[0,1],xmm0[2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 304(%rsi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps 304(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vbroadcastsd 304(%rcx), %ymm3
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm3[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 304(%r9), %xmm2
+; AVX1-ONLY-NEXT: vmovaps 304(%r8), %xmm3
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm4 = xmm3[0],xmm2[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vbroadcastsd 304(%rax), %ymm5
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5],ymm5[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX1-ONLY-NEXT: vbroadcastsd 312(%rdx), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovlps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11, %xmm0 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm0 = mem[0,1],xmm11[2,3]
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm3[1],xmm2[1]
; AVX1-ONLY-NEXT: vbroadcastsd 312(%r10), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm11 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovlps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm0 = mem[0,1],xmm0[2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 336(%rsi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps 336(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vbroadcastsd 336(%rcx), %ymm3
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm3[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 336(%r9), %xmm2
+; AVX1-ONLY-NEXT: vmovaps 336(%r8), %xmm3
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm4 = xmm3[0],xmm2[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vbroadcastsd 336(%rax), %ymm5
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5],ymm5[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX1-ONLY-NEXT: vbroadcastsd 344(%rdx), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovlps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm0 = mem[0,1],xmm0[2,3]
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm3[1],xmm2[1]
; AVX1-ONLY-NEXT: vbroadcastsd 344(%r10), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovlps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm0 = mem[0,1],xmm0[2,3]
+; AVX1-ONLY-NEXT: vmovaps 368(%rsi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps 368(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vbroadcastsd 368(%rcx), %ymm3
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm3[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 368(%r9), %xmm2
+; AVX1-ONLY-NEXT: vmovaps 368(%r8), %xmm3
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm4 = xmm3[0],xmm2[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vbroadcastsd 368(%rax), %ymm5
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5],ymm5[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX1-ONLY-NEXT: vbroadcastsd 376(%rdx), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovlps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm0 = mem[0,1],xmm0[2,3]
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm3[1],xmm2[1]
; AVX1-ONLY-NEXT: vbroadcastsd 376(%r10), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovlps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14, %xmm0 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm0 = mem[0,1],xmm14[2,3]
+; AVX1-ONLY-NEXT: vmovaps 400(%rsi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps 400(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vbroadcastsd 400(%rcx), %ymm3
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm3[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 400(%r9), %xmm2
+; AVX1-ONLY-NEXT: vmovaps 400(%r8), %xmm3
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm4 = xmm3[0],xmm2[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vbroadcastsd 400(%rax), %ymm5
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5],ymm5[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX1-ONLY-NEXT: vbroadcastsd 408(%rdx), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm14 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vmovlps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12, %xmm0 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm0 = mem[0,1],xmm12[2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm3[1],xmm2[1]
; AVX1-ONLY-NEXT: vbroadcastsd 408(%r10), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm12 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vmovlps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9, %xmm0 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm0 = mem[0,1],xmm9[2,3]
-; AVX1-ONLY-NEXT: vbroadcastsd 440(%rdx), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm9 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vmovlps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8, %xmm0 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm0 = mem[0,1],xmm8[2,3]
-; AVX1-ONLY-NEXT: vbroadcastsd 440(%r10), %ymm2
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm0[0,1,2,3],ymm2[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vmovlps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm2 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm2 = mem[0,1],xmm15[2,3]
-; AVX1-ONLY-NEXT: vbroadcastsd 472(%rdx), %ymm15
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm2[0,1,2,3],ymm15[4,5],ymm2[6,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovlps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm2 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm2 = mem[0,1],xmm0[2,3]
-; AVX1-ONLY-NEXT: vbroadcastsd 472(%r10), %ymm15
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm2[0,1,2,3],ymm15[4,5],ymm2[6,7]
-; AVX1-ONLY-NEXT: vmovlps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7, %xmm2 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm2 = mem[0,1],xmm7[2,3]
-; AVX1-ONLY-NEXT: vbroadcastsd 504(%rdx), %ymm15
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm15[4,5],ymm2[6,7]
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm2 = xmm13[1],xmm10[1]
-; AVX1-ONLY-NEXT: vbroadcastsd 504(%r10), %ymm13
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm13[4,5],ymm2[6,7]
-; AVX1-ONLY-NEXT: movq {{[0-9]+}}(%rsp), %rdx
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],mem[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],mem[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],mem[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],mem[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],mem[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],mem[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],mem[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],mem[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],mem[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],mem[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],mem[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups (%rsp), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],mem[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm2, (%rsp) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],mem[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],mem[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],mem[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 432(%rsi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps 432(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vbroadcastsd 432(%rcx), %ymm3
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm3[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],mem[6,7]
+; AVX1-ONLY-NEXT: vmovaps 432(%r9), %xmm2
+; AVX1-ONLY-NEXT: vmovaps 432(%r8), %xmm3
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm4 = xmm3[0],xmm2[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vbroadcastsd 432(%rax), %ymm5
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5],ymm5[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT: vbroadcastsd 440(%rdx), %ymm1
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm3[1],xmm2[1]
+; AVX1-ONLY-NEXT: vbroadcastsd 440(%r10), %ymm1
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 464(%rsi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps 464(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vbroadcastsd 464(%rcx), %ymm3
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm3[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm15 = ymm6[0,1,2,3,4,5],mem[6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm13 = ymm3[0,1,2,3,4,5],mem[6,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm10 = ymm2[0,1,2,3,4,5],mem[6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm11 = ymm11[0,1,2,3,4,5],mem[6,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm2[0,1,2,3,4,5],mem[6,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm2[0,1,2,3,4,5],mem[6,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm2[0,1,2,3,4,5],mem[6,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],mem[6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4,5],mem[6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3,4,5],mem[6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm9 = ymm9[0,1,2,3,4,5],mem[6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5],mem[6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5],mem[6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5],mem[6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],mem[6,7]
+; AVX1-ONLY-NEXT: vmovaps 464(%r9), %xmm2
+; AVX1-ONLY-NEXT: vmovaps 464(%r8), %xmm3
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm4 = xmm3[0],xmm2[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vbroadcastsd 464(%rax), %ymm5
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5],ymm5[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT: vbroadcastsd 472(%rdx), %ymm1
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm15 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm3[1],xmm2[1]
+; AVX1-ONLY-NEXT: vbroadcastsd 472(%r10), %ymm1
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm12 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps 496(%rsi), %xmm2
+; AVX1-ONLY-NEXT: vmovaps 496(%rdi), %xmm3
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm0 = xmm3[0],xmm2[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vbroadcastsd 496(%rcx), %ymm5
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm11 = ymm0[0,1,2,3,4,5],ymm5[6,7]
+; AVX1-ONLY-NEXT: vmovaps 496(%r9), %xmm5
+; AVX1-ONLY-NEXT: vmovaps 496(%r8), %xmm6
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm8 = xmm6[0],xmm5[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vbroadcastsd 496(%rax), %ymm9
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5],ymm9[6,7]
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm2 = xmm3[1],xmm2[1]
+; AVX1-ONLY-NEXT: vbroadcastsd 504(%rdx), %ymm3
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm2[0,1,2,3],ymm3[4,5,6,7]
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm2 = xmm6[1],xmm5[1]
+; AVX1-ONLY-NEXT: vbroadcastsd 504(%r10), %ymm3
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm2[0,1,2,3],ymm3[4,5,6,7]
+; AVX1-ONLY-NEXT: movq {{[0-9]+}}(%rsp), %rdx
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm10[0,1,2,3,4,5],mem[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm13[0,1,2,3,4,5],mem[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm14[0,1,2,3,4,5],mem[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],mem[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],mem[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],mem[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],mem[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],mem[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],mem[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],mem[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],mem[6,7]
-; AVX1-ONLY-NEXT: vmovaps %ymm0, 4064(%rdx)
-; AVX1-ONLY-NEXT: vmovaps %ymm1, 4032(%rdx)
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm0, 4000(%rdx)
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],mem[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm0, 3968(%rdx)
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],mem[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovups (%rsp), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],mem[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, (%rsp) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],mem[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],mem[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],mem[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],mem[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm0, 3936(%rdx)
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm14 = ymm0[0,1,2,3,4,5],mem[6,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm0, 3904(%rdx)
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm13 = ymm0[0,1,2,3,4,5],mem[6,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm0, 3872(%rdx)
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm10 = ymm0[0,1,2,3,4,5],mem[6,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm0, 3840(%rdx)
-; AVX1-ONLY-NEXT: vmovaps %ymm4, 3808(%rdx)
-; AVX1-ONLY-NEXT: vmovaps %ymm5, 3776(%rdx)
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm9 = ymm0[0,1,2,3,4,5],mem[6,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm0, 3744(%rdx)
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm0[0,1,2,3,4,5],mem[6,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm0, 3712(%rdx)
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm0[0,1,2,3,4,5],mem[6,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm0, 3680(%rdx)
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm0[0,1,2,3,4,5],mem[6,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm0, 3648(%rdx)
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm0[0,1,2,3,4,5],mem[6,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm0, 3616(%rdx)
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm0[0,1,2,3,4,5],mem[6,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm0, 3584(%rdx)
-; AVX1-ONLY-NEXT: vmovaps %ymm8, 3552(%rdx)
-; AVX1-ONLY-NEXT: vmovaps %ymm9, 3520(%rdx)
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],mem[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm15 = ymm15[0,1,2,3,4,5],mem[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3,4,5],mem[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5],mem[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],mem[6,7]
+; AVX1-ONLY-NEXT: vmovaps %ymm3, 4064(%rdx)
+; AVX1-ONLY-NEXT: vmovaps %ymm7, 4032(%rdx)
+; AVX1-ONLY-NEXT: vmovaps %ymm8, 4000(%rdx)
+; AVX1-ONLY-NEXT: vmovaps %ymm11, 3968(%rdx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm3, 3936(%rdx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm3, 3904(%rdx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm3, 3872(%rdx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm3, 3840(%rdx)
+; AVX1-ONLY-NEXT: vmovaps %ymm12, 3808(%rdx)
+; AVX1-ONLY-NEXT: vmovaps %ymm15, 3776(%rdx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm3, 3744(%rdx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm3, 3712(%rdx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm3, 3680(%rdx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm3, 3648(%rdx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm3, 3616(%rdx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm3, 3584(%rdx)
+; AVX1-ONLY-NEXT: vmovaps %ymm0, 3552(%rdx)
+; AVX1-ONLY-NEXT: vmovaps %ymm1, 3520(%rdx)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm0, 3488(%rdx)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
@@ -7621,8 +7469,8 @@ define void @store_i64_stride8_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vmovaps %ymm0, 3360(%rdx)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm0, 3328(%rdx)
-; AVX1-ONLY-NEXT: vmovaps %ymm12, 3296(%rdx)
-; AVX1-ONLY-NEXT: vmovaps %ymm14, 3264(%rdx)
+; AVX1-ONLY-NEXT: vmovaps %ymm2, 3296(%rdx)
+; AVX1-ONLY-NEXT: vmovaps %ymm4, 3264(%rdx)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm0, 3232(%rdx)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
@@ -7635,8 +7483,8 @@ define void @store_i64_stride8_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vmovaps %ymm0, 3104(%rdx)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm0, 3072(%rdx)
-; AVX1-ONLY-NEXT: vmovaps %ymm2, 3040(%rdx)
-; AVX1-ONLY-NEXT: vmovaps %ymm3, 3008(%rdx)
+; AVX1-ONLY-NEXT: vmovaps %ymm5, 3040(%rdx)
+; AVX1-ONLY-NEXT: vmovaps %ymm6, 3008(%rdx)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm0, 2976(%rdx)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
@@ -7649,8 +7497,8 @@ define void @store_i64_stride8_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vmovaps %ymm0, 2848(%rdx)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm0, 2816(%rdx)
-; AVX1-ONLY-NEXT: vmovaps %ymm6, 2784(%rdx)
-; AVX1-ONLY-NEXT: vmovaps %ymm7, 2752(%rdx)
+; AVX1-ONLY-NEXT: vmovaps %ymm9, 2784(%rdx)
+; AVX1-ONLY-NEXT: vmovaps %ymm10, 2752(%rdx)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm0, 2720(%rdx)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
@@ -7663,8 +7511,8 @@ define void @store_i64_stride8_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vmovaps %ymm0, 2592(%rdx)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm0, 2560(%rdx)
-; AVX1-ONLY-NEXT: vmovaps %ymm11, 2528(%rdx)
-; AVX1-ONLY-NEXT: vmovaps %ymm10, 2496(%rdx)
+; AVX1-ONLY-NEXT: vmovaps %ymm13, 2528(%rdx)
+; AVX1-ONLY-NEXT: vmovaps %ymm14, 2496(%rdx)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm0, 2464(%rdx)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
@@ -7677,8 +7525,10 @@ define void @store_i64_stride8_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vmovaps %ymm0, 2336(%rdx)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm0, 2304(%rdx)
-; AVX1-ONLY-NEXT: vmovaps %ymm13, 2272(%rdx)
-; AVX1-ONLY-NEXT: vmovaps %ymm15, 2240(%rdx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm0, 2272(%rdx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm0, 2240(%rdx)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm0, 2208(%rdx)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
@@ -7707,7 +7557,7 @@ define void @store_i64_stride8_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vmovaps %ymm0, 1824(%rdx)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm0, 1792(%rdx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovups (%rsp), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm0, 1760(%rdx)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm0, 1728(%rdx)
@@ -7723,7 +7573,7 @@ define void @store_i64_stride8_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vmovaps %ymm0, 1568(%rdx)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm0, 1536(%rdx)
-; AVX1-ONLY-NEXT: vmovups (%rsp), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm0, 1504(%rdx)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm0, 1472(%rdx)
@@ -7819,7 +7669,7 @@ define void @store_i64_stride8_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX1-ONLY-NEXT: vmovaps %ymm0, 32(%rdx)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm0, (%rdx)
-; AVX1-ONLY-NEXT: addq $4712, %rsp # imm = 0x1268
+; AVX1-ONLY-NEXT: addq $3784, %rsp # imm = 0xEC8
; AVX1-ONLY-NEXT: vzeroupper
; AVX1-ONLY-NEXT: retq
;
@@ -7841,7 +7691,7 @@ define void @store_i64_stride8_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm2[1]
; AVX2-ONLY-NEXT: vbroadcastsd 8(%rdx), %ymm2
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovaps (%rax), %xmm0
@@ -7853,7 +7703,7 @@ define void @store_i64_stride8_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm2[1]
; AVX2-ONLY-NEXT: vbroadcastsd 8(%r10), %ymm2
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovaps 32(%rcx), %xmm0
@@ -7861,7 +7711,7 @@ define void @store_i64_stride8_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm4[1],xmm3[1]
; AVX2-ONLY-NEXT: vbroadcastsd 40(%rdx), %ymm2
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovaps 32(%r9), %xmm1
@@ -7870,7 +7720,7 @@ define void @store_i64_stride8_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
; AVX2-ONLY-NEXT: vbroadcastsd 40(%r10), %ymm1
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT: vmovaps 32(%rax), %xmm1
; AVX2-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
@@ -7882,7 +7732,7 @@ define void @store_i64_stride8_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX2-ONLY-NEXT: vbroadcastsd 72(%rdx), %ymm1
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT: vmovaps 64(%rcx), %xmm1
; AVX2-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
@@ -7894,7 +7744,7 @@ define void @store_i64_stride8_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX2-ONLY-NEXT: vbroadcastsd 72(%r10), %ymm1
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT: vmovaps 64(%rax), %xmm1
; AVX2-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
@@ -7906,7 +7756,7 @@ define void @store_i64_stride8_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
; AVX2-ONLY-NEXT: vbroadcastsd 104(%rdx), %ymm1
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT: vmovaps 96(%rcx), %xmm1
; AVX2-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
@@ -7918,7 +7768,7 @@ define void @store_i64_stride8_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
; AVX2-ONLY-NEXT: vbroadcastsd 104(%r10), %ymm1
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT: vmovaps 96(%rax), %xmm1
; AVX2-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
@@ -7930,7 +7780,7 @@ define void @store_i64_stride8_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
; AVX2-ONLY-NEXT: vbroadcastsd 136(%rdx), %ymm1
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT: vmovaps 128(%rcx), %xmm1
; AVX2-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
@@ -7942,7 +7792,7 @@ define void @store_i64_stride8_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
; AVX2-ONLY-NEXT: vbroadcastsd 136(%r10), %ymm1
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT: vmovaps 128(%rax), %xmm1
; AVX2-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
@@ -7954,7 +7804,7 @@ define void @store_i64_stride8_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
; AVX2-ONLY-NEXT: vbroadcastsd 168(%rdx), %ymm1
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT: vmovaps 160(%rcx), %xmm1
; AVX2-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
@@ -7966,7 +7816,7 @@ define void @store_i64_stride8_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX2-ONLY-NEXT: vbroadcastsd 168(%r10), %ymm1
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT: vmovaps 160(%rax), %xmm1
; AVX2-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
@@ -7978,7 +7828,7 @@ define void @store_i64_stride8_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
; AVX2-ONLY-NEXT: vbroadcastsd 200(%rdx), %ymm1
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT: vmovaps 192(%rcx), %xmm1
; AVX2-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
@@ -7990,7 +7840,7 @@ define void @store_i64_stride8_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
; AVX2-ONLY-NEXT: vbroadcastsd 200(%r10), %ymm1
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT: vmovaps 192(%rax), %xmm1
; AVX2-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
@@ -8002,7 +7852,7 @@ define void @store_i64_stride8_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
; AVX2-ONLY-NEXT: vbroadcastsd 232(%rdx), %ymm1
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT: vmovaps 224(%rcx), %xmm1
; AVX2-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
@@ -8014,7 +7864,7 @@ define void @store_i64_stride8_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX2-ONLY-NEXT: vbroadcastsd 232(%r10), %ymm1
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT: vmovaps 224(%rax), %xmm1
; AVX2-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
@@ -8026,7 +7876,7 @@ define void @store_i64_stride8_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
; AVX2-ONLY-NEXT: vbroadcastsd 264(%rdx), %ymm1
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT: vmovaps 256(%rcx), %xmm1
; AVX2-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
@@ -8038,7 +7888,7 @@ define void @store_i64_stride8_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
; AVX2-ONLY-NEXT: vbroadcastsd 264(%r10), %ymm1
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT: vmovaps 256(%rax), %xmm1
; AVX2-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
@@ -8050,7 +7900,7 @@ define void @store_i64_stride8_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
; AVX2-ONLY-NEXT: vbroadcastsd 296(%rdx), %ymm1
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT: vmovaps 288(%rcx), %xmm1
; AVX2-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
@@ -8062,7 +7912,7 @@ define void @store_i64_stride8_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
; AVX2-ONLY-NEXT: vbroadcastsd 296(%r10), %ymm1
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT: vmovaps 288(%rax), %xmm1
; AVX2-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
@@ -8074,7 +7924,7 @@ define void @store_i64_stride8_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
; AVX2-ONLY-NEXT: vbroadcastsd 328(%rdx), %ymm1
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT: vmovaps 320(%rcx), %xmm1
; AVX2-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
@@ -8086,7 +7936,7 @@ define void @store_i64_stride8_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
; AVX2-ONLY-NEXT: vbroadcastsd 328(%r10), %ymm1
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT: vmovaps 320(%rax), %xmm1
; AVX2-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
@@ -8098,7 +7948,7 @@ define void @store_i64_stride8_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
; AVX2-ONLY-NEXT: vbroadcastsd 360(%rdx), %ymm1
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT: vmovaps 352(%rcx), %xmm1
; AVX2-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
@@ -8110,7 +7960,7 @@ define void @store_i64_stride8_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX2-ONLY-NEXT: vbroadcastsd 360(%r10), %ymm1
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT: vmovaps 352(%rax), %xmm1
; AVX2-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
@@ -8122,7 +7972,7 @@ define void @store_i64_stride8_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
; AVX2-ONLY-NEXT: vbroadcastsd 392(%rdx), %ymm1
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT: vmovaps 384(%rcx), %xmm1
; AVX2-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
@@ -8134,7 +7984,7 @@ define void @store_i64_stride8_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX2-ONLY-NEXT: vbroadcastsd 392(%r10), %ymm1
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT: vmovaps 384(%rax), %xmm1
; AVX2-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
@@ -8146,7 +7996,7 @@ define void @store_i64_stride8_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
; AVX2-ONLY-NEXT: vbroadcastsd 424(%rdx), %ymm1
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT: vmovaps 416(%rcx), %xmm13
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm13, %ymm0, %ymm1
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
@@ -8157,7 +8007,7 @@ define void @store_i64_stride8_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX2-ONLY-NEXT: vbroadcastsd 424(%r10), %ymm1
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT: vmovaps 416(%rax), %xmm12
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm12, %ymm0, %ymm1
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
@@ -8166,7 +8016,7 @@ define void @store_i64_stride8_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovaps 448(%rdi), %xmm10
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm10[1],xmm11[1]
; AVX2-ONLY-NEXT: vbroadcastsd 456(%rdx), %ymm1
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT: vmovaps 448(%rcx), %xmm7
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm7, %ymm0, %ymm1
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
@@ -8175,7 +8025,7 @@ define void @store_i64_stride8_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovaps 448(%r8), %xmm8
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm8[1],xmm9[1]
; AVX2-ONLY-NEXT: vbroadcastsd 456(%r10), %ymm1
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT: vmovaps 448(%rax), %xmm6
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm6, %ymm0, %ymm1
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
@@ -8184,7 +8034,7 @@ define void @store_i64_stride8_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovaps 480(%rdi), %xmm4
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm4[1],xmm5[1]
; AVX2-ONLY-NEXT: vbroadcastsd 488(%rdx), %ymm1
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT: vmovaps 480(%rcx), %xmm1
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm2
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm2[6,7]
@@ -8193,7 +8043,7 @@ define void @store_i64_stride8_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovaps 480(%r8), %xmm2
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm2[1],xmm3[1]
; AVX2-ONLY-NEXT: vbroadcastsd 488(%r10), %ymm15
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm15 = ymm0[0,1,2,3],ymm15[4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm15 = ymm0[0,1,2,3],ymm15[4,5,6,7]
; AVX2-ONLY-NEXT: vmovaps 480(%rax), %xmm0
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm14
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm14 = ymm15[0,1,2,3,4,5],ymm14[6,7]
diff --git a/llvm/test/CodeGen/X86/vector-shuffle-256-v4.ll b/llvm/test/CodeGen/X86/vector-shuffle-256-v4.ll
index 1504076a588fc..2afdee01e453d 100644
--- a/llvm/test/CodeGen/X86/vector-shuffle-256-v4.ll
+++ b/llvm/test/CodeGen/X86/vector-shuffle-256-v4.ll
@@ -1054,7 +1054,7 @@ define <4 x i64> @shuffle_v4i64_0142(<4 x i64> %a, <4 x i64> %b) {
define <4 x i64> @shuffle_v4i64_0412(<4 x i64> %a, <4 x i64> %b) {
; AVX1-LABEL: shuffle_v4i64_0412:
; AVX1: # %bb.0:
-; AVX1-NEXT: vblendpd {{.*#+}} ymm1 = ymm1[0,1],ymm0[2],ymm1[3]
+; AVX1-NEXT: vblendpd {{.*#+}} ymm1 = ymm1[0,1],ymm0[2,3]
; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
; AVX1-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[3],ymm1[2]
; AVX1-NEXT: retq
@@ -1185,13 +1185,13 @@ define <4 x i64> @shuffle_v4i64_2u35(<4 x i64> %a, <4 x i64> %b) {
;
; AVX2-LABEL: shuffle_v4i64_2u35:
; AVX2: # %bb.0:
-; AVX2-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5,6,7]
+; AVX2-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX2-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[2,1,3,1]
; AVX2-NEXT: retq
;
; AVX512VL-SLOW-LABEL: shuffle_v4i64_2u35:
; AVX512VL-SLOW: # %bb.0:
-; AVX512VL-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5,6,7]
+; AVX512VL-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX512VL-SLOW-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[2,1,3,1]
; AVX512VL-SLOW-NEXT: retq
;
@@ -1203,7 +1203,7 @@ define <4 x i64> @shuffle_v4i64_2u35(<4 x i64> %a, <4 x i64> %b) {
;
; AVX512VL-FAST-PERLANE-LABEL: shuffle_v4i64_2u35:
; AVX512VL-FAST-PERLANE: # %bb.0:
-; AVX512VL-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5,6,7]
+; AVX512VL-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX512VL-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[2,1,3,1]
; AVX512VL-FAST-PERLANE-NEXT: retq
%shuffle = shufflevector <4 x i64> %a, <4 x i64> %b, <4 x i32> <i32 2, i32 undef, i32 3, i32 5>
diff --git a/llvm/test/CodeGen/X86/vector-shuffle-256-v8.ll b/llvm/test/CodeGen/X86/vector-shuffle-256-v8.ll
index 21cd0ca1380eb..1546a58bc8401 100644
--- a/llvm/test/CodeGen/X86/vector-shuffle-256-v8.ll
+++ b/llvm/test/CodeGen/X86/vector-shuffle-256-v8.ll
@@ -3807,7 +3807,7 @@ define <8 x i32> @lowhalf_v8i32(<8 x i32> %x, <8 x i32> %y) {
;
; AVX2-LABEL: lowhalf_v8i32:
; AVX2: # %bb.0:
-; AVX2-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-NEXT: vmovaps {{.*#+}} xmm1 = [2,6,3,6]
; AVX2-NEXT: vpermps %ymm0, %ymm1, %ymm0
; AVX2-NEXT: retq
@@ -3833,7 +3833,7 @@ define <8 x float> @lowhalf_v8f32(<8 x float> %x, <8 x float> %y) {
;
; AVX2-LABEL: lowhalf_v8f32:
; AVX2: # %bb.0:
-; AVX2-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-NEXT: vmovaps {{.*#+}} xmm1 = [2,6,3,6]
; AVX2-NEXT: vpermps %ymm0, %ymm1, %ymm0
; AVX2-NEXT: retq
diff --git a/llvm/test/CodeGen/X86/vector-shuffle-combining-avx.ll b/llvm/test/CodeGen/X86/vector-shuffle-combining-avx.ll
index ca791bd765264..5baf7bb788d2e 100644
--- a/llvm/test/CodeGen/X86/vector-shuffle-combining-avx.ll
+++ b/llvm/test/CodeGen/X86/vector-shuffle-combining-avx.ll
@@ -448,7 +448,7 @@ define void @PR48908(<4 x double> %v0, <4 x double> %v1, <4 x double> %v2, ptr n
; X86-AVX1-NEXT: vblendpd {{.*#+}} ymm3 = ymm5[0],ymm3[1],ymm5[2],ymm3[3]
; X86-AVX1-NEXT: vmovapd %ymm3, (%edx)
; X86-AVX1-NEXT: vperm2f128 {{.*#+}} ymm3 = ymm4[2,3,0,1]
-; X86-AVX1-NEXT: vblendpd {{.*#+}} ymm4 = ymm4[0,1],ymm0[2],ymm4[3]
+; X86-AVX1-NEXT: vblendpd {{.*#+}} ymm4 = ymm4[0,1],ymm0[2,3]
; X86-AVX1-NEXT: vblendpd {{.*#+}} ymm3 = ymm4[0],ymm3[1],ymm4[2],ymm3[3]
; X86-AVX1-NEXT: vmovapd %ymm3, (%ecx)
; X86-AVX1-NEXT: vextractf128 $1, %ymm0, %xmm0
@@ -471,7 +471,7 @@ define void @PR48908(<4 x double> %v0, <4 x double> %v1, <4 x double> %v2, ptr n
; X86-AVX2-NEXT: vpermpd {{.*#+}} ymm3 = ymm3[0,2,2,1]
; X86-AVX2-NEXT: vblendpd {{.*#+}} ymm3 = ymm6[0],ymm3[1],ymm6[2],ymm3[3]
; X86-AVX2-NEXT: vmovapd %ymm3, (%edx)
-; X86-AVX2-NEXT: vblendpd {{.*#+}} ymm3 = ymm5[0,1],ymm0[2],ymm5[3]
+; X86-AVX2-NEXT: vblendpd {{.*#+}} ymm3 = ymm5[0,1],ymm0[2,3]
; X86-AVX2-NEXT: vpermpd {{.*#+}} ymm4 = ymm4[0,3,2,0]
; X86-AVX2-NEXT: vblendpd {{.*#+}} ymm3 = ymm3[0],ymm4[1],ymm3[2],ymm4[3]
; X86-AVX2-NEXT: vmovapd %ymm3, (%ecx)
@@ -491,15 +491,15 @@ define void @PR48908(<4 x double> %v0, <4 x double> %v1, <4 x double> %v2, ptr n
; X86-AVX512-NEXT: movl {{[0-9]+}}(%esp), %eax
; X86-AVX512-NEXT: movl {{[0-9]+}}(%esp), %ecx
; X86-AVX512-NEXT: movl {{[0-9]+}}(%esp), %edx
-; X86-AVX512-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm3
-; X86-AVX512-NEXT: vshufpd {{.*#+}} ymm3 = ymm0[0],ymm3[1],ymm0[2],ymm3[2]
-; X86-AVX512-NEXT: vperm2f128 {{.*#+}} ymm4 = ymm1[2,3],ymm2[0,1]
-; X86-AVX512-NEXT: vshufpd {{.*#+}} ymm4 = ymm1[1],ymm4[0],ymm1[2],ymm4[3]
-; X86-AVX512-NEXT: vmovapd {{.*#+}} ymm5 = [0,0,3,0,8,0,1,0]
-; X86-AVX512-NEXT: vpermt2pd %zmm2, %zmm5, %zmm3
-; X86-AVX512-NEXT: vmovapd %ymm3, (%edx)
-; X86-AVX512-NEXT: vmovapd {{.*#+}} ymm3 = [0,0,3,0,10,0,1,0]
-; X86-AVX512-NEXT: vpermt2pd %zmm0, %zmm3, %zmm4
+; X86-AVX512-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm4
+; X86-AVX512-NEXT: vmovapd {{.*#+}} ymm3 = [1,0,2,0,8,0,9,0]
+; X86-AVX512-NEXT: vpermi2pd %zmm2, %zmm1, %zmm3
+; X86-AVX512-NEXT: vmovapd {{.*#+}} ymm5 = [0,0,10,0,2,0,9,0]
+; X86-AVX512-NEXT: vperm2f128 {{.*#+}} ymm6 = ymm0[0,1],ymm2[0,1]
+; X86-AVX512-NEXT: vpermt2pd %zmm4, %zmm5, %zmm6
+; X86-AVX512-NEXT: vmovapd %ymm6, (%edx)
+; X86-AVX512-NEXT: vmovapd {{.*#+}} ymm4 = [0,0,3,0,10,0,1,0]
+; X86-AVX512-NEXT: vpermi2pd %zmm0, %zmm3, %zmm4
; X86-AVX512-NEXT: vmovapd %ymm4, (%ecx)
; X86-AVX512-NEXT: vmovapd {{.*#+}} ymm3 = <3,0,11,0,u,u,u,u>
; X86-AVX512-NEXT: vpermi2pd %zmm1, %zmm0, %zmm3
@@ -521,7 +521,7 @@ define void @PR48908(<4 x double> %v0, <4 x double> %v1, <4 x double> %v2, ptr n
; X64-AVX1-NEXT: vblendpd {{.*#+}} ymm3 = ymm5[0],ymm3[1],ymm5[2],ymm3[3]
; X64-AVX1-NEXT: vmovapd %ymm3, (%rdi)
; X64-AVX1-NEXT: vperm2f128 {{.*#+}} ymm3 = ymm4[2,3,0,1]
-; X64-AVX1-NEXT: vblendpd {{.*#+}} ymm4 = ymm4[0,1],ymm0[2],ymm4[3]
+; X64-AVX1-NEXT: vblendpd {{.*#+}} ymm4 = ymm4[0,1],ymm0[2,3]
; X64-AVX1-NEXT: vblendpd {{.*#+}} ymm3 = ymm4[0],ymm3[1],ymm4[2],ymm3[3]
; X64-AVX1-NEXT: vmovapd %ymm3, (%rsi)
; X64-AVX1-NEXT: vextractf128 $1, %ymm0, %xmm0
@@ -541,7 +541,7 @@ define void @PR48908(<4 x double> %v0, <4 x double> %v1, <4 x double> %v2, ptr n
; X64-AVX2-NEXT: vpermpd {{.*#+}} ymm3 = ymm3[0,2,2,1]
; X64-AVX2-NEXT: vblendpd {{.*#+}} ymm3 = ymm6[0],ymm3[1],ymm6[2],ymm3[3]
; X64-AVX2-NEXT: vmovapd %ymm3, (%rdi)
-; X64-AVX2-NEXT: vblendpd {{.*#+}} ymm3 = ymm5[0,1],ymm0[2],ymm5[3]
+; X64-AVX2-NEXT: vblendpd {{.*#+}} ymm3 = ymm5[0,1],ymm0[2,3]
; X64-AVX2-NEXT: vpermpd {{.*#+}} ymm4 = ymm4[0,3,2,0]
; X64-AVX2-NEXT: vblendpd {{.*#+}} ymm3 = ymm3[0],ymm4[1],ymm3[2],ymm4[3]
; X64-AVX2-NEXT: vmovapd %ymm3, (%rsi)
@@ -559,15 +559,15 @@ define void @PR48908(<4 x double> %v0, <4 x double> %v1, <4 x double> %v2, ptr n
; X64-AVX512-NEXT: # kill: def $ymm1 killed $ymm1 def $zmm1
; X64-AVX512-NEXT: # kill: def $ymm0 killed $ymm0 def $zmm0
; X64-AVX512-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm3
-; X64-AVX512-NEXT: vshufpd {{.*#+}} ymm3 = ymm0[0],ymm3[1],ymm0[2],ymm3[2]
-; X64-AVX512-NEXT: vperm2f128 {{.*#+}} ymm4 = ymm1[2,3],ymm2[0,1]
-; X64-AVX512-NEXT: vshufpd {{.*#+}} ymm4 = ymm1[1],ymm4[0],ymm1[2],ymm4[3]
-; X64-AVX512-NEXT: vmovapd {{.*#+}} ymm5 = [0,3,8,1]
-; X64-AVX512-NEXT: vpermt2pd %zmm2, %zmm5, %zmm3
-; X64-AVX512-NEXT: vmovapd %ymm3, (%rdi)
+; X64-AVX512-NEXT: vmovapd {{.*#+}} ymm4 = [1,2,8,9]
+; X64-AVX512-NEXT: vpermi2pd %zmm2, %zmm1, %zmm4
+; X64-AVX512-NEXT: vmovapd {{.*#+}} ymm5 = [0,10,2,9]
+; X64-AVX512-NEXT: vperm2f128 {{.*#+}} ymm6 = ymm0[0,1],ymm2[0,1]
+; X64-AVX512-NEXT: vpermt2pd %zmm3, %zmm5, %zmm6
+; X64-AVX512-NEXT: vmovapd %ymm6, (%rdi)
; X64-AVX512-NEXT: vmovapd {{.*#+}} ymm3 = [0,3,10,1]
-; X64-AVX512-NEXT: vpermt2pd %zmm0, %zmm3, %zmm4
-; X64-AVX512-NEXT: vmovapd %ymm4, (%rsi)
+; X64-AVX512-NEXT: vpermi2pd %zmm0, %zmm4, %zmm3
+; X64-AVX512-NEXT: vmovapd %ymm3, (%rsi)
; X64-AVX512-NEXT: vmovapd {{.*#+}} ymm3 = <3,11,u,u>
; X64-AVX512-NEXT: vpermi2pd %zmm1, %zmm0, %zmm3
; X64-AVX512-NEXT: vmovapd {{.*#+}} ymm0 = [2,8,9,3]
diff --git a/llvm/test/CodeGen/X86/vector-shuffle-combining-avx2.ll b/llvm/test/CodeGen/X86/vector-shuffle-combining-avx2.ll
index 068cfff72a7d5..4e8bab11f979d 100644
--- a/llvm/test/CodeGen/X86/vector-shuffle-combining-avx2.ll
+++ b/llvm/test/CodeGen/X86/vector-shuffle-combining-avx2.ll
@@ -843,7 +843,7 @@ define internal fastcc <8 x float> @PR34577(<8 x float> %inp0, <8 x float> %inp1
; AVX512-NEXT: # kill: def $ymm1 killed $ymm1 def $zmm1
; AVX512-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[1,1,1,1]
; AVX512-NEXT: vxorps %xmm2, %xmm2, %xmm2
-; AVX512-NEXT: vblendps {{.*#+}} ymm2 = ymm0[0,1],ymm2[2,3],ymm0[4,5],ymm2[6,7]
+; AVX512-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm0[4,5],ymm2[6,7]
; AVX512-NEXT: vmovaps {{.*#+}} ymm0 = <23,18,7,2,20,u,3,2>
; AVX512-NEXT: vpermi2ps %zmm2, %zmm1, %zmm0
; AVX512-NEXT: # kill: def $ymm0 killed $ymm0 killed $zmm0
diff --git a/llvm/test/CodeGen/X86/zero_extend_vector_inreg_of_broadcast.ll b/llvm/test/CodeGen/X86/zero_extend_vector_inreg_of_broadcast.ll
index ca68479d60ddc..869068fad0888 100644
--- a/llvm/test/CodeGen/X86/zero_extend_vector_inreg_of_broadcast.ll
+++ b/llvm/test/CodeGen/X86/zero_extend_vector_inreg_of_broadcast.ll
@@ -4879,7 +4879,7 @@ define void @vec384_i32_widen_to_i96_factor3_broadcast_to_v4i96_factor4(ptr %in.
; AVX2-SLOW-NEXT: vmovdqa 32(%rdi), %ymm1
; AVX2-SLOW-NEXT: vpaddb (%rsi), %ymm0, %ymm0
; AVX2-SLOW-NEXT: vpaddb 32(%rsi), %ymm1, %ymm1
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm0[0,1],ymm1[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} xmm2 = [0,5,6,0]
; AVX2-SLOW-NEXT: vpermd %ymm1, %ymm2, %ymm1
; AVX2-SLOW-NEXT: vpxor %xmm2, %xmm2, %xmm2
@@ -4900,7 +4900,7 @@ define void @vec384_i32_widen_to_i96_factor3_broadcast_to_v4i96_factor4(ptr %in.
; AVX2-FAST-PERLANE-NEXT: vmovdqa 32(%rdi), %ymm1
; AVX2-FAST-PERLANE-NEXT: vpaddb (%rsi), %ymm0, %ymm0
; AVX2-FAST-PERLANE-NEXT: vpaddb 32(%rsi), %ymm1, %ymm1
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm0[0,1],ymm1[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm2 = [0,5,6,0]
; AVX2-FAST-PERLANE-NEXT: vpermd %ymm1, %ymm2, %ymm1
; AVX2-FAST-PERLANE-NEXT: vpxor %xmm2, %xmm2, %xmm2
@@ -4919,7 +4919,7 @@ define void @vec384_i32_widen_to_i96_factor3_broadcast_to_v4i96_factor4(ptr %in.
; AVX2-FAST-NEXT: vmovdqa 32(%rdi), %ymm1
; AVX2-FAST-NEXT: vpaddb (%rsi), %ymm0, %ymm0
; AVX2-FAST-NEXT: vpaddb 32(%rsi), %ymm1, %ymm1
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm0[0,1],ymm1[2,3,4,5,6,7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm2 = [0,5,6,0]
; AVX2-FAST-NEXT: vpermd %ymm1, %ymm2, %ymm1
; AVX2-FAST-NEXT: vpxor %xmm2, %xmm2, %xmm2
@@ -5078,7 +5078,7 @@ define void @vec384_i32_widen_to_i128_factor4_broadcast_to_v3i128_factor3(ptr %i
; AVX2-NEXT: vmovdqa 48(%rdi), %xmm1
; AVX2-NEXT: vpaddb 48(%rsi), %xmm1, %xmm1
; AVX2-NEXT: vpermq {{.*#+}} ymm2 = ymm0[0,1,0,1]
-; AVX2-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0],ymm1[1,2,3],ymm2[4],ymm1[5,6,7]
+; AVX2-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0],ymm1[1,2,3],ymm2[4,5,6,7]
; AVX2-NEXT: vpxor %xmm2, %xmm2, %xmm2
; AVX2-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
; AVX2-NEXT: vpxor %xmm2, %xmm2, %xmm2
@@ -5129,8 +5129,8 @@ define void @vec384_i32_widen_to_i128_factor4_broadcast_to_v3i128_factor3(ptr %i
; AVX512BW-LABEL: vec384_i32_widen_to_i128_factor4_broadcast_to_v3i128_factor3:
; AVX512BW: # %bb.0:
; AVX512BW-NEXT: vmovdqa64 (%rdi), %zmm0
+; AVX512BW-NEXT: vmovdqa {{.*#+}} ymm1 = [0,13,14,15,0,1,2,3]
; AVX512BW-NEXT: vpaddb (%rsi), %zmm0, %zmm0
-; AVX512BW-NEXT: vmovdqa {{.*#+}} xmm1 = [0,13,14,15]
; AVX512BW-NEXT: vpermd %zmm0, %zmm1, %zmm1
; AVX512BW-NEXT: vpxor %xmm2, %xmm2, %xmm2
; AVX512BW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
@@ -5218,7 +5218,7 @@ define void @vec384_i32_widen_to_i192_factor6_broadcast_to_v2i192_factor2(ptr %i
; AVX2-NEXT: vmovdqa 32(%rdi), %ymm1
; AVX2-NEXT: vpaddb (%rsi), %ymm0, %ymm0
; AVX2-NEXT: vpaddb 32(%rsi), %ymm1, %ymm1
-; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4,5,6,7]
+; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-NEXT: vmovdqa {{.*#+}} xmm1 = [0,5,6,7]
; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
; AVX2-NEXT: vpxor %xmm1, %xmm1, %xmm1
diff --git a/llvm/test/CodeGen/X86/zero_extend_vector_inreg_of_broadcast_from_memory.ll b/llvm/test/CodeGen/X86/zero_extend_vector_inreg_of_broadcast_from_memory.ll
index 74271a48e0bac..5050327a89fda 100644
--- a/llvm/test/CodeGen/X86/zero_extend_vector_inreg_of_broadcast_from_memory.ll
+++ b/llvm/test/CodeGen/X86/zero_extend_vector_inreg_of_broadcast_from_memory.ll
@@ -3929,7 +3929,7 @@ define void @vec384_i32_widen_to_i96_factor3_broadcast_to_v4i96_factor4(ptr %in.
; AVX2-SLOW-NEXT: vmovdqa 32(%rdi), %ymm0
; AVX2-SLOW-NEXT: vmovd {{.*#+}} xmm1 = mem[0],zero,zero,zero
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[1,0,1,1]
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = mem[0,1],ymm0[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = mem[0,1,2,3],ymm0[4,5,6,7]
; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} xmm2 = [0,5,6,0]
; AVX2-SLOW-NEXT: vpermd %ymm0, %ymm2, %ymm0
; AVX2-SLOW-NEXT: vpxor %xmm2, %xmm2, %xmm2
@@ -3946,7 +3946,7 @@ define void @vec384_i32_widen_to_i96_factor3_broadcast_to_v4i96_factor4(ptr %in.
; AVX2-FAST-PERLANE-NEXT: vmovdqa 32(%rdi), %ymm0
; AVX2-FAST-PERLANE-NEXT: vmovdqa (%rdi), %xmm1
; AVX2-FAST-PERLANE-NEXT: vpshufb {{.*#+}} xmm1 = zero,zero,zero,zero,xmm1[0,1,2,3],zero,zero,zero,zero,zero,zero,zero,zero
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = mem[0,1],ymm0[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = mem[0,1,2,3],ymm0[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm2 = [0,5,6,0]
; AVX2-FAST-PERLANE-NEXT: vpermd %ymm0, %ymm2, %ymm0
; AVX2-FAST-PERLANE-NEXT: vpxor %xmm2, %xmm2, %xmm2
@@ -3963,7 +3963,7 @@ define void @vec384_i32_widen_to_i96_factor3_broadcast_to_v4i96_factor4(ptr %in.
; AVX2-FAST-NEXT: vmovdqa 32(%rdi), %ymm0
; AVX2-FAST-NEXT: vmovdqa (%rdi), %xmm1
; AVX2-FAST-NEXT: vpshufb {{.*#+}} xmm1 = zero,zero,zero,zero,xmm1[0,1,2,3],zero,zero,zero,zero,zero,zero,zero,zero
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = mem[0,1],ymm0[2,3,4,5,6,7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = mem[0,1,2,3],ymm0[4,5,6,7]
; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm2 = [0,5,6,0]
; AVX2-FAST-NEXT: vpermd %ymm0, %ymm2, %ymm0
; AVX2-FAST-NEXT: vpxor %xmm2, %xmm2, %xmm2
@@ -4079,7 +4079,7 @@ define void @vec384_i32_widen_to_i128_factor4_broadcast_to_v3i128_factor3(ptr %i
; AVX2-NEXT: vmovd {{.*#+}} xmm0 = mem[0],zero,zero,zero
; AVX2-NEXT: vmovdqa 48(%rdi), %xmm1
; AVX2-NEXT: vbroadcasti128 {{.*#+}} ymm2 = mem[0,1,0,1]
-; AVX2-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0],ymm1[1,2,3],ymm2[4],ymm1[5,6,7]
+; AVX2-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0],ymm1[1,2,3],ymm2[4,5,6,7]
; AVX2-NEXT: vpxor %xmm2, %xmm2, %xmm2
; AVX2-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
; AVX2-NEXT: vpaddb (%rsi), %ymm1, %ymm1
@@ -4191,7 +4191,7 @@ define void @vec384_i32_widen_to_i192_factor6_broadcast_to_v2i192_factor2(ptr %i
; AVX2-LABEL: vec384_i32_widen_to_i192_factor6_broadcast_to_v2i192_factor2:
; AVX2: # %bb.0:
; AVX2-NEXT: vmovdqa 32(%rdi), %ymm0
-; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = mem[0,1],ymm0[2,3,4,5,6,7]
+; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = mem[0,1,2,3],ymm0[4,5,6,7]
; AVX2-NEXT: vmovdqa {{.*#+}} xmm1 = [0,5,6,7]
; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
; AVX2-NEXT: vpxor %xmm1, %xmm1, %xmm1
More information about the llvm-commits
mailing list