[llvm] 72fddda - [X86] ReplaceNodeResults - widen vector truncate nodes on pre-SSSE3 targets
Simon Pilgrim via llvm-commits
llvm-commits at lists.llvm.org
Tue Jul 11 09:21:42 PDT 2023
Author: Simon Pilgrim
Date: 2023-07-11T17:21:20+01:00
New Revision: 72fddda8af473b893c7c81af78364982fcb64c86
URL: https://github.com/llvm/llvm-project/commit/72fddda8af473b893c7c81af78364982fcb64c86
DIFF: https://github.com/llvm/llvm-project/commit/72fddda8af473b893c7c81af78364982fcb64c86.diff
LOG: [X86] ReplaceNodeResults - widen vector truncate nodes on pre-SSSE3 targets
Building on the support for wider input vector types from D154592, try to more aggressively widen inputs instead of scalarizing them.
Added:
Modified:
llvm/lib/Target/X86/X86ISelLowering.cpp
llvm/test/CodeGen/X86/masked_store_trunc.ll
llvm/test/CodeGen/X86/masked_store_trunc_ssat.ll
llvm/test/CodeGen/X86/masked_store_trunc_usat.ll
llvm/test/CodeGen/X86/pmulh.ll
llvm/test/CodeGen/X86/vector-trunc-packus.ll
llvm/test/CodeGen/X86/vector-trunc-ssat.ll
llvm/test/CodeGen/X86/vector-trunc-usat.ll
llvm/test/CodeGen/X86/vector-trunc.ll
Removed:
################################################################################
diff --git a/llvm/lib/Target/X86/X86ISelLowering.cpp b/llvm/lib/Target/X86/X86ISelLowering.cpp
index 88478ce0874e0f..75254d23059b55 100644
--- a/llvm/lib/Target/X86/X86ISelLowering.cpp
+++ b/llvm/lib/Target/X86/X86ISelLowering.cpp
@@ -34560,15 +34560,15 @@ void X86TargetLowering::ReplaceNodeResults(SDNode *N,
MVT WidenVT = getTypeToTransformTo(*DAG.getContext(), VT).getSimpleVT();
SDValue In = N->getOperand(0);
EVT InVT = In.getValueType();
+ EVT InEltVT = InVT.getVectorElementType();
+ EVT EltVT = VT.getVectorElementType();
+ unsigned WidenNumElts = WidenVT.getVectorNumElements();
unsigned InBits = InVT.getSizeInBits();
if (128 % InBits == 0) {
// 128 bit and smaller inputs should avoid truncate all together and
// just use a build_vector that will become a shuffle.
// TODO: Widen and use a shuffle directly?
- MVT InEltVT = InVT.getSimpleVT().getVectorElementType();
- EVT EltVT = VT.getVectorElementType();
- unsigned WidenNumElts = WidenVT.getVectorNumElements();
SmallVector<SDValue, 16> Ops(WidenNumElts, DAG.getUNDEF(EltVT));
// Use the original element count so we don't do more scalar opts than
// necessary.
@@ -34615,6 +34615,18 @@ void X86TargetLowering::ReplaceNodeResults(SDNode *N,
return;
}
+ // Pre-SSSE3 (or v4i64 -> v4i16) widen the truncation input vector to let
+ // LowerTRUNCATE handle this via type legalization.
+ if ((InEltVT == MVT::i16 || InEltVT == MVT::i32 || InEltVT == MVT::i64) &&
+ (EltVT == MVT::i8 || EltVT == MVT::i16 || EltVT == MVT::i32) &&
+ (!Subtarget.hasSSSE3() || (InVT == MVT::v4i64 && VT == MVT::v4i16)) &&
+ !Subtarget.hasAVX()) {
+ SDValue WidenIn = widenSubVector(In, false, Subtarget, DAG, dl,
+ InEltVT.getSizeInBits() * WidenNumElts);
+ Results.push_back(DAG.getNode(ISD::TRUNCATE, dl, WidenVT, WidenIn));
+ return;
+ }
+
return;
}
case ISD::ANY_EXTEND:
diff --git a/llvm/test/CodeGen/X86/masked_store_trunc.ll b/llvm/test/CodeGen/X86/masked_store_trunc.ll
index d650a08200a9bb..8458b88cdefd17 100644
--- a/llvm/test/CodeGen/X86/masked_store_trunc.ll
+++ b/llvm/test/CodeGen/X86/masked_store_trunc.ll
@@ -1133,13 +1133,12 @@ define void @truncstore_v4i64_v4i16(<4 x i64> %x, ptr %p, <4 x i32> %mask) {
; SSE2-LABEL: truncstore_v4i64_v4i16:
; SSE2: # %bb.0:
; SSE2-NEXT: pxor %xmm3, %xmm3
-; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
-; SSE2-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[0,2,2,3,4,5,6,7]
-; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
-; SSE2-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[0,2,2,3,4,5,6,7]
-; SSE2-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; SSE2-NEXT: pcmpeqd %xmm2, %xmm3
-; SSE2-NEXT: movmskps %xmm3, %eax
+; SSE2-NEXT: shufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2]
+; SSE2-NEXT: pslld $16, %xmm0
+; SSE2-NEXT: psrad $16, %xmm0
+; SSE2-NEXT: packssdw %xmm3, %xmm0
+; SSE2-NEXT: pcmpeqd %xmm3, %xmm2
+; SSE2-NEXT: movmskps %xmm2, %eax
; SSE2-NEXT: xorl $15, %eax
; SSE2-NEXT: testb $1, %al
; SSE2-NEXT: jne .LBB4_1
@@ -1180,7 +1179,7 @@ define void @truncstore_v4i64_v4i16(<4 x i64> %x, ptr %p, <4 x i32> %mask) {
; SSE4-NEXT: pblendw {{.*#+}} xmm1 = xmm1[0],xmm3[1,2,3],xmm1[4],xmm3[5,6,7]
; SSE4-NEXT: pblendw {{.*#+}} xmm0 = xmm0[0],xmm3[1,2,3],xmm0[4],xmm3[5,6,7]
; SSE4-NEXT: packusdw %xmm1, %xmm0
-; SSE4-NEXT: packusdw %xmm0, %xmm0
+; SSE4-NEXT: packusdw %xmm3, %xmm0
; SSE4-NEXT: pcmpeqd %xmm3, %xmm2
; SSE4-NEXT: movmskps %xmm2, %eax
; SSE4-NEXT: xorl $15, %eax
@@ -1363,14 +1362,14 @@ define void @truncstore_v4i64_v4i8(<4 x i64> %x, ptr %p, <4 x i32> %mask) {
; SSE2-LABEL: truncstore_v4i64_v4i8:
; SSE2: # %bb.0:
; SSE2-NEXT: pxor %xmm3, %xmm3
-; SSE2-NEXT: movdqa {{.*#+}} xmm4 = [255,0,0,0,255,0,0,0]
+; SSE2-NEXT: movdqa {{.*#+}} xmm4 = [255,0,0,0,0,0,0,0,255,0,0,0,0,0,0,0]
; SSE2-NEXT: pand %xmm4, %xmm1
; SSE2-NEXT: pand %xmm4, %xmm0
; SSE2-NEXT: packuswb %xmm1, %xmm0
-; SSE2-NEXT: packuswb %xmm0, %xmm0
-; SSE2-NEXT: packuswb %xmm0, %xmm0
-; SSE2-NEXT: pcmpeqd %xmm2, %xmm3
-; SSE2-NEXT: movmskps %xmm3, %ecx
+; SSE2-NEXT: packuswb %xmm3, %xmm0
+; SSE2-NEXT: packuswb %xmm3, %xmm0
+; SSE2-NEXT: pcmpeqd %xmm3, %xmm2
+; SSE2-NEXT: movmskps %xmm2, %ecx
; SSE2-NEXT: xorl $15, %ecx
; SSE2-NEXT: testb $1, %cl
; SSE2-NEXT: movd %xmm0, %eax
diff --git a/llvm/test/CodeGen/X86/masked_store_trunc_ssat.ll b/llvm/test/CodeGen/X86/masked_store_trunc_ssat.ll
index 3379beaf838fe8..40eed08cd5c851 100644
--- a/llvm/test/CodeGen/X86/masked_store_trunc_ssat.ll
+++ b/llvm/test/CodeGen/X86/masked_store_trunc_ssat.ll
@@ -2088,12 +2088,12 @@ define void @truncstore_v4i64_v4i8(<4 x i64> %x, ptr %p, <4 x i32> %mask) {
; SSE2-NEXT: pand %xmm4, %xmm5
; SSE2-NEXT: pandn %xmm1, %xmm4
; SSE2-NEXT: por %xmm5, %xmm4
-; SSE2-NEXT: movdqa {{.*#+}} xmm1 = [255,0,0,0,255,0,0,0]
+; SSE2-NEXT: movdqa {{.*#+}} xmm1 = [255,0,0,0,0,0,0,0,255,0,0,0,0,0,0,0]
; SSE2-NEXT: pand %xmm1, %xmm4
; SSE2-NEXT: pand %xmm1, %xmm0
; SSE2-NEXT: packuswb %xmm4, %xmm0
-; SSE2-NEXT: packuswb %xmm0, %xmm0
-; SSE2-NEXT: packuswb %xmm0, %xmm0
+; SSE2-NEXT: packuswb %xmm3, %xmm0
+; SSE2-NEXT: packuswb %xmm3, %xmm0
; SSE2-NEXT: pcmpeqd %xmm3, %xmm2
; SSE2-NEXT: movmskps %xmm2, %ecx
; SSE2-NEXT: xorl $15, %ecx
diff --git a/llvm/test/CodeGen/X86/masked_store_trunc_usat.ll b/llvm/test/CodeGen/X86/masked_store_trunc_usat.ll
index 57588985094ad1..b21d6cf79ad810 100644
--- a/llvm/test/CodeGen/X86/masked_store_trunc_usat.ll
+++ b/llvm/test/CodeGen/X86/masked_store_trunc_usat.ll
@@ -1471,7 +1471,7 @@ define void @truncstore_v4i64_v4i16(<4 x i64> %x, ptr %p, <4 x i32> %mask) {
; SSE2: # %bb.0:
; SSE2-NEXT: pxor %xmm3, %xmm3
; SSE2-NEXT: movdqa {{.*#+}} xmm4 = [9223372039002259456,9223372039002259456]
-; SSE2-NEXT: movdqa %xmm0, %xmm5
+; SSE2-NEXT: movdqa %xmm1, %xmm5
; SSE2-NEXT: pxor %xmm4, %xmm5
; SSE2-NEXT: pshufd {{.*#+}} xmm6 = xmm5[0,0,2,2]
; SSE2-NEXT: movdqa {{.*#+}} xmm7 = [2147549183,2147549183,2147549183,2147549183]
@@ -1481,26 +1481,25 @@ define void @truncstore_v4i64_v4i16(<4 x i64> %x, ptr %p, <4 x i32> %mask) {
; SSE2-NEXT: pcmpeqd %xmm4, %xmm5
; SSE2-NEXT: pand %xmm8, %xmm5
; SSE2-NEXT: pcmpeqd %xmm6, %xmm6
-; SSE2-NEXT: pand %xmm5, %xmm0
+; SSE2-NEXT: pand %xmm5, %xmm1
; SSE2-NEXT: pxor %xmm6, %xmm5
-; SSE2-NEXT: por %xmm0, %xmm5
-; SSE2-NEXT: movdqa %xmm1, %xmm0
-; SSE2-NEXT: pxor %xmm4, %xmm0
-; SSE2-NEXT: pshufd {{.*#+}} xmm8 = xmm0[0,0,2,2]
+; SSE2-NEXT: por %xmm1, %xmm5
+; SSE2-NEXT: movdqa %xmm0, %xmm1
+; SSE2-NEXT: pxor %xmm4, %xmm1
+; SSE2-NEXT: pshufd {{.*#+}} xmm8 = xmm1[0,0,2,2]
; SSE2-NEXT: pcmpgtd %xmm8, %xmm7
-; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm0[1,1,3,3]
-; SSE2-NEXT: pcmpeqd %xmm4, %xmm0
-; SSE2-NEXT: pand %xmm7, %xmm0
-; SSE2-NEXT: pxor %xmm0, %xmm6
-; SSE2-NEXT: pand %xmm1, %xmm0
-; SSE2-NEXT: por %xmm6, %xmm0
-; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
-; SSE2-NEXT: pshuflw {{.*#+}} xmm1 = xmm0[0,2,2,3,4,5,6,7]
-; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm5[0,2,2,3]
-; SSE2-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[0,2,2,3,4,5,6,7]
-; SSE2-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; SSE2-NEXT: pcmpeqd %xmm2, %xmm3
-; SSE2-NEXT: movmskps %xmm3, %eax
+; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
+; SSE2-NEXT: pcmpeqd %xmm4, %xmm1
+; SSE2-NEXT: pand %xmm7, %xmm1
+; SSE2-NEXT: pxor %xmm1, %xmm6
+; SSE2-NEXT: pand %xmm0, %xmm1
+; SSE2-NEXT: por %xmm6, %xmm1
+; SSE2-NEXT: shufps {{.*#+}} xmm1 = xmm1[0,2],xmm5[0,2]
+; SSE2-NEXT: pslld $16, %xmm1
+; SSE2-NEXT: psrad $16, %xmm1
+; SSE2-NEXT: packssdw %xmm3, %xmm1
+; SSE2-NEXT: pcmpeqd %xmm3, %xmm2
+; SSE2-NEXT: movmskps %xmm2, %eax
; SSE2-NEXT: xorl $15, %eax
; SSE2-NEXT: testb $1, %al
; SSE2-NEXT: jne .LBB4_1
@@ -1516,22 +1515,22 @@ define void @truncstore_v4i64_v4i16(<4 x i64> %x, ptr %p, <4 x i32> %mask) {
; SSE2-NEXT: .LBB4_8: # %else6
; SSE2-NEXT: retq
; SSE2-NEXT: .LBB4_1: # %cond.store
-; SSE2-NEXT: movd %xmm0, %ecx
+; SSE2-NEXT: movd %xmm1, %ecx
; SSE2-NEXT: movw %cx, (%rdi)
; SSE2-NEXT: testb $2, %al
; SSE2-NEXT: je .LBB4_4
; SSE2-NEXT: .LBB4_3: # %cond.store1
-; SSE2-NEXT: pextrw $1, %xmm0, %ecx
+; SSE2-NEXT: pextrw $1, %xmm1, %ecx
; SSE2-NEXT: movw %cx, 2(%rdi)
; SSE2-NEXT: testb $4, %al
; SSE2-NEXT: je .LBB4_6
; SSE2-NEXT: .LBB4_5: # %cond.store3
-; SSE2-NEXT: pextrw $2, %xmm0, %ecx
+; SSE2-NEXT: pextrw $2, %xmm1, %ecx
; SSE2-NEXT: movw %cx, 4(%rdi)
; SSE2-NEXT: testb $8, %al
; SSE2-NEXT: je .LBB4_8
; SSE2-NEXT: .LBB4_7: # %cond.store5
-; SSE2-NEXT: pextrw $3, %xmm0, %eax
+; SSE2-NEXT: pextrw $3, %xmm1, %eax
; SSE2-NEXT: movw %ax, 6(%rdi)
; SSE2-NEXT: retq
;
@@ -1777,14 +1776,14 @@ define void @truncstore_v4i64_v4i8(<4 x i64> %x, ptr %p, <4 x i32> %mask) {
; SSE2-NEXT: pxor %xmm0, %xmm6
; SSE2-NEXT: pand %xmm1, %xmm0
; SSE2-NEXT: por %xmm6, %xmm0
-; SSE2-NEXT: movdqa {{.*#+}} xmm1 = [255,0,0,0,255,0,0,0]
+; SSE2-NEXT: movdqa {{.*#+}} xmm1 = [255,0,0,0,0,0,0,0,255,0,0,0,0,0,0,0]
; SSE2-NEXT: pand %xmm1, %xmm0
; SSE2-NEXT: pand %xmm1, %xmm4
; SSE2-NEXT: packuswb %xmm0, %xmm4
-; SSE2-NEXT: packuswb %xmm4, %xmm4
-; SSE2-NEXT: packuswb %xmm4, %xmm4
-; SSE2-NEXT: pcmpeqd %xmm2, %xmm3
-; SSE2-NEXT: movmskps %xmm3, %ecx
+; SSE2-NEXT: packuswb %xmm3, %xmm4
+; SSE2-NEXT: packuswb %xmm3, %xmm4
+; SSE2-NEXT: pcmpeqd %xmm3, %xmm2
+; SSE2-NEXT: movmskps %xmm2, %ecx
; SSE2-NEXT: xorl $15, %ecx
; SSE2-NEXT: testb $1, %cl
; SSE2-NEXT: movd %xmm4, %eax
diff --git a/llvm/test/CodeGen/X86/pmulh.ll b/llvm/test/CodeGen/X86/pmulh.ll
index 96613b63b47e76..503a294f2f9300 100644
--- a/llvm/test/CodeGen/X86/pmulh.ll
+++ b/llvm/test/CodeGen/X86/pmulh.ll
@@ -26,16 +26,15 @@ define <4 x i16> @zext_mulhuw_v4i16(<4 x i16> %a, <4 x i16> %b) {
define <4 x i16> @and_mulhuw_v4i16(<4 x i64> %a, <4 x i64> %b) {
; SSE2-LABEL: and_mulhuw_v4i16:
; SSE2: # %bb.0:
-; SSE2-NEXT: pshufd {{.*#+}} xmm3 = xmm3[0,2,2,3]
-; SSE2-NEXT: pshuflw {{.*#+}} xmm3 = xmm3[0,2,2,3,4,5,6,7]
-; SSE2-NEXT: pshufd {{.*#+}} xmm2 = xmm2[0,2,2,3]
-; SSE2-NEXT: pshuflw {{.*#+}} xmm2 = xmm2[0,2,2,3,4,5,6,7]
-; SSE2-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1]
-; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
-; SSE2-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[0,2,2,3,4,5,6,7]
-; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
-; SSE2-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[0,2,2,3,4,5,6,7]
-; SSE2-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE2-NEXT: shufps {{.*#+}} xmm2 = xmm2[0,2],xmm3[0,2]
+; SSE2-NEXT: pslld $16, %xmm2
+; SSE2-NEXT: psrad $16, %xmm2
+; SSE2-NEXT: xorps %xmm3, %xmm3
+; SSE2-NEXT: packssdw %xmm3, %xmm2
+; SSE2-NEXT: shufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2]
+; SSE2-NEXT: pslld $16, %xmm0
+; SSE2-NEXT: psrad $16, %xmm0
+; SSE2-NEXT: packssdw %xmm3, %xmm0
; SSE2-NEXT: pmulhuw %xmm2, %xmm0
; SSE2-NEXT: retq
;
@@ -45,11 +44,11 @@ define <4 x i16> @and_mulhuw_v4i16(<4 x i64> %a, <4 x i64> %b) {
; SSE41-NEXT: pblendw {{.*#+}} xmm3 = xmm3[0],xmm4[1,2,3],xmm3[4],xmm4[5,6,7]
; SSE41-NEXT: pblendw {{.*#+}} xmm2 = xmm2[0],xmm4[1,2,3],xmm2[4],xmm4[5,6,7]
; SSE41-NEXT: packusdw %xmm3, %xmm2
-; SSE41-NEXT: packusdw %xmm2, %xmm2
+; SSE41-NEXT: packusdw %xmm4, %xmm2
; SSE41-NEXT: pblendw {{.*#+}} xmm1 = xmm1[0],xmm4[1,2,3],xmm1[4],xmm4[5,6,7]
; SSE41-NEXT: pblendw {{.*#+}} xmm0 = xmm0[0],xmm4[1,2,3],xmm0[4],xmm4[5,6,7]
; SSE41-NEXT: packusdw %xmm1, %xmm0
-; SSE41-NEXT: packusdw %xmm0, %xmm0
+; SSE41-NEXT: packusdw %xmm4, %xmm0
; SSE41-NEXT: pmulhuw %xmm2, %xmm0
; SSE41-NEXT: retq
;
diff --git a/llvm/test/CodeGen/X86/vector-trunc-packus.ll b/llvm/test/CodeGen/X86/vector-trunc-packus.ll
index 50185d849b926c..bb9c1b472540d5 100644
--- a/llvm/test/CodeGen/X86/vector-trunc-packus.ll
+++ b/llvm/test/CodeGen/X86/vector-trunc-packus.ll
@@ -1114,61 +1114,58 @@ define <4 x i16> @trunc_packus_v4i64_v4i16(<4 x i64> %a0) {
; SSE2-SSSE3-LABEL: trunc_packus_v4i64_v4i16:
; SSE2-SSSE3: # %bb.0:
; SSE2-SSSE3-NEXT: movdqa {{.*#+}} xmm4 = [65535,65535]
-; SSE2-SSSE3-NEXT: movdqa {{.*#+}} xmm2 = [2147483648,2147483648]
-; SSE2-SSSE3-NEXT: movdqa %xmm1, %xmm3
-; SSE2-SSSE3-NEXT: pxor %xmm2, %xmm3
-; SSE2-SSSE3-NEXT: pshufd {{.*#+}} xmm5 = xmm3[1,1,3,3]
-; SSE2-SSSE3-NEXT: pxor %xmm6, %xmm6
-; SSE2-SSSE3-NEXT: pcmpeqd %xmm6, %xmm5
+; SSE2-SSSE3-NEXT: movdqa {{.*#+}} xmm3 = [2147483648,2147483648]
+; SSE2-SSSE3-NEXT: movdqa %xmm0, %xmm5
+; SSE2-SSSE3-NEXT: pxor %xmm3, %xmm5
+; SSE2-SSSE3-NEXT: pshufd {{.*#+}} xmm6 = xmm5[1,1,3,3]
+; SSE2-SSSE3-NEXT: pxor %xmm2, %xmm2
+; SSE2-SSSE3-NEXT: pcmpeqd %xmm2, %xmm6
; SSE2-SSSE3-NEXT: movdqa {{.*#+}} xmm7 = [2147549183,2147549183]
; SSE2-SSSE3-NEXT: movdqa %xmm7, %xmm8
-; SSE2-SSSE3-NEXT: pcmpgtd %xmm3, %xmm8
-; SSE2-SSSE3-NEXT: pshufd {{.*#+}} xmm9 = xmm8[0,0,2,2]
-; SSE2-SSSE3-NEXT: pand %xmm5, %xmm9
-; SSE2-SSSE3-NEXT: pshufd {{.*#+}} xmm3 = xmm8[1,1,3,3]
-; SSE2-SSSE3-NEXT: por %xmm9, %xmm3
-; SSE2-SSSE3-NEXT: pand %xmm3, %xmm1
-; SSE2-SSSE3-NEXT: pandn %xmm4, %xmm3
-; SSE2-SSSE3-NEXT: por %xmm1, %xmm3
-; SSE2-SSSE3-NEXT: movdqa %xmm0, %xmm1
-; SSE2-SSSE3-NEXT: pxor %xmm2, %xmm1
-; SSE2-SSSE3-NEXT: pshufd {{.*#+}} xmm5 = xmm1[1,1,3,3]
-; SSE2-SSSE3-NEXT: pcmpeqd %xmm6, %xmm5
-; SSE2-SSSE3-NEXT: pcmpgtd %xmm1, %xmm7
-; SSE2-SSSE3-NEXT: pshufd {{.*#+}} xmm1 = xmm7[0,0,2,2]
-; SSE2-SSSE3-NEXT: pand %xmm5, %xmm1
-; SSE2-SSSE3-NEXT: pshufd {{.*#+}} xmm5 = xmm7[1,1,3,3]
-; SSE2-SSSE3-NEXT: por %xmm1, %xmm5
-; SSE2-SSSE3-NEXT: pand %xmm5, %xmm0
-; SSE2-SSSE3-NEXT: pandn %xmm4, %xmm5
-; SSE2-SSSE3-NEXT: por %xmm0, %xmm5
-; SSE2-SSSE3-NEXT: movdqa %xmm5, %xmm0
-; SSE2-SSSE3-NEXT: pxor %xmm2, %xmm0
+; SSE2-SSSE3-NEXT: pcmpgtd %xmm5, %xmm8
+; SSE2-SSSE3-NEXT: pshufd {{.*#+}} xmm5 = xmm8[0,0,2,2]
+; SSE2-SSSE3-NEXT: pand %xmm6, %xmm5
+; SSE2-SSSE3-NEXT: pshufd {{.*#+}} xmm6 = xmm8[1,1,3,3]
+; SSE2-SSSE3-NEXT: por %xmm5, %xmm6
+; SSE2-SSSE3-NEXT: pand %xmm6, %xmm0
+; SSE2-SSSE3-NEXT: pandn %xmm4, %xmm6
+; SSE2-SSSE3-NEXT: por %xmm6, %xmm0
+; SSE2-SSSE3-NEXT: movdqa %xmm1, %xmm5
+; SSE2-SSSE3-NEXT: pxor %xmm3, %xmm5
+; SSE2-SSSE3-NEXT: pshufd {{.*#+}} xmm6 = xmm5[1,1,3,3]
+; SSE2-SSSE3-NEXT: pcmpeqd %xmm2, %xmm6
+; SSE2-SSSE3-NEXT: pcmpgtd %xmm5, %xmm7
+; SSE2-SSSE3-NEXT: pshufd {{.*#+}} xmm5 = xmm7[0,0,2,2]
+; SSE2-SSSE3-NEXT: pand %xmm6, %xmm5
+; SSE2-SSSE3-NEXT: pshufd {{.*#+}} xmm6 = xmm7[1,1,3,3]
+; SSE2-SSSE3-NEXT: por %xmm5, %xmm6
+; SSE2-SSSE3-NEXT: pand %xmm6, %xmm1
+; SSE2-SSSE3-NEXT: pandn %xmm4, %xmm6
+; SSE2-SSSE3-NEXT: por %xmm1, %xmm6
+; SSE2-SSSE3-NEXT: movdqa %xmm6, %xmm1
+; SSE2-SSSE3-NEXT: pxor %xmm3, %xmm1
+; SSE2-SSSE3-NEXT: movdqa %xmm1, %xmm4
+; SSE2-SSSE3-NEXT: pcmpgtd %xmm3, %xmm4
+; SSE2-SSSE3-NEXT: pcmpeqd %xmm3, %xmm1
+; SSE2-SSSE3-NEXT: pshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
+; SSE2-SSSE3-NEXT: pand %xmm4, %xmm1
+; SSE2-SSSE3-NEXT: pshufd {{.*#+}} xmm4 = xmm4[1,1,3,3]
+; SSE2-SSSE3-NEXT: por %xmm1, %xmm4
+; SSE2-SSSE3-NEXT: pand %xmm6, %xmm4
; SSE2-SSSE3-NEXT: movdqa %xmm0, %xmm1
-; SSE2-SSSE3-NEXT: pcmpgtd %xmm2, %xmm1
-; SSE2-SSSE3-NEXT: pshufd {{.*#+}} xmm4 = xmm1[0,0,2,2]
-; SSE2-SSSE3-NEXT: pcmpeqd %xmm2, %xmm0
-; SSE2-SSSE3-NEXT: pshufd {{.*#+}} xmm0 = xmm0[1,1,3,3]
-; SSE2-SSSE3-NEXT: pand %xmm4, %xmm0
+; SSE2-SSSE3-NEXT: pxor %xmm3, %xmm1
+; SSE2-SSSE3-NEXT: movdqa %xmm1, %xmm5
+; SSE2-SSSE3-NEXT: pcmpgtd %xmm3, %xmm5
+; SSE2-SSSE3-NEXT: pcmpeqd %xmm3, %xmm1
; SSE2-SSSE3-NEXT: pshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
-; SSE2-SSSE3-NEXT: por %xmm0, %xmm1
; SSE2-SSSE3-NEXT: pand %xmm5, %xmm1
-; SSE2-SSSE3-NEXT: movdqa %xmm3, %xmm0
-; SSE2-SSSE3-NEXT: pxor %xmm2, %xmm0
-; SSE2-SSSE3-NEXT: movdqa %xmm0, %xmm4
-; SSE2-SSSE3-NEXT: pcmpgtd %xmm2, %xmm4
-; SSE2-SSSE3-NEXT: pshufd {{.*#+}} xmm5 = xmm4[0,0,2,2]
-; SSE2-SSSE3-NEXT: pcmpeqd %xmm2, %xmm0
-; SSE2-SSSE3-NEXT: pshufd {{.*#+}} xmm0 = xmm0[1,1,3,3]
-; SSE2-SSSE3-NEXT: pand %xmm5, %xmm0
-; SSE2-SSSE3-NEXT: pshufd {{.*#+}} xmm2 = xmm4[1,1,3,3]
-; SSE2-SSSE3-NEXT: por %xmm0, %xmm2
-; SSE2-SSSE3-NEXT: pand %xmm3, %xmm2
-; SSE2-SSSE3-NEXT: pshufd {{.*#+}} xmm0 = xmm2[0,2,2,3]
-; SSE2-SSSE3-NEXT: pshuflw {{.*#+}} xmm2 = xmm0[0,2,2,3,4,5,6,7]
-; SSE2-SSSE3-NEXT: pshufd {{.*#+}} xmm0 = xmm1[0,2,2,3]
-; SSE2-SSSE3-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[0,2,2,3,4,5,6,7]
-; SSE2-SSSE3-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
+; SSE2-SSSE3-NEXT: pshufd {{.*#+}} xmm3 = xmm5[1,1,3,3]
+; SSE2-SSSE3-NEXT: por %xmm1, %xmm3
+; SSE2-SSSE3-NEXT: pand %xmm3, %xmm0
+; SSE2-SSSE3-NEXT: shufps {{.*#+}} xmm0 = xmm0[0,2],xmm4[0,2]
+; SSE2-SSSE3-NEXT: pslld $16, %xmm0
+; SSE2-SSSE3-NEXT: psrad $16, %xmm0
+; SSE2-SSSE3-NEXT: packssdw %xmm2, %xmm0
; SSE2-SSSE3-NEXT: retq
;
; SSE41-LABEL: trunc_packus_v4i64_v4i16:
@@ -1308,7 +1305,7 @@ define void @trunc_packus_v4i64_v4i16_store(<4 x i64> %a0, ptr%p1) {
; SSE2-SSSE3: # %bb.0:
; SSE2-SSSE3-NEXT: movdqa {{.*#+}} xmm4 = [65535,65535]
; SSE2-SSSE3-NEXT: movdqa {{.*#+}} xmm2 = [2147483648,2147483648]
-; SSE2-SSSE3-NEXT: movdqa %xmm1, %xmm3
+; SSE2-SSSE3-NEXT: movdqa %xmm0, %xmm3
; SSE2-SSSE3-NEXT: pxor %xmm2, %xmm3
; SSE2-SSSE3-NEXT: pshufd {{.*#+}} xmm5 = xmm3[1,1,3,3]
; SSE2-SSSE3-NEXT: pxor %xmm6, %xmm6
@@ -1320,29 +1317,28 @@ define void @trunc_packus_v4i64_v4i16_store(<4 x i64> %a0, ptr%p1) {
; SSE2-SSSE3-NEXT: pand %xmm5, %xmm9
; SSE2-SSSE3-NEXT: pshufd {{.*#+}} xmm3 = xmm8[1,1,3,3]
; SSE2-SSSE3-NEXT: por %xmm9, %xmm3
-; SSE2-SSSE3-NEXT: pand %xmm3, %xmm1
+; SSE2-SSSE3-NEXT: pand %xmm3, %xmm0
; SSE2-SSSE3-NEXT: pandn %xmm4, %xmm3
-; SSE2-SSSE3-NEXT: por %xmm1, %xmm3
-; SSE2-SSSE3-NEXT: movdqa %xmm0, %xmm1
-; SSE2-SSSE3-NEXT: pxor %xmm2, %xmm1
-; SSE2-SSSE3-NEXT: pshufd {{.*#+}} xmm5 = xmm1[1,1,3,3]
+; SSE2-SSSE3-NEXT: por %xmm0, %xmm3
+; SSE2-SSSE3-NEXT: movdqa %xmm1, %xmm0
+; SSE2-SSSE3-NEXT: pxor %xmm2, %xmm0
+; SSE2-SSSE3-NEXT: pshufd {{.*#+}} xmm5 = xmm0[1,1,3,3]
; SSE2-SSSE3-NEXT: pcmpeqd %xmm6, %xmm5
-; SSE2-SSSE3-NEXT: pcmpgtd %xmm1, %xmm7
-; SSE2-SSSE3-NEXT: pshufd {{.*#+}} xmm1 = xmm7[0,0,2,2]
-; SSE2-SSSE3-NEXT: pand %xmm5, %xmm1
-; SSE2-SSSE3-NEXT: pshufd {{.*#+}} xmm5 = xmm7[1,1,3,3]
-; SSE2-SSSE3-NEXT: por %xmm1, %xmm5
+; SSE2-SSSE3-NEXT: pcmpgtd %xmm0, %xmm7
+; SSE2-SSSE3-NEXT: pshufd {{.*#+}} xmm0 = xmm7[0,0,2,2]
; SSE2-SSSE3-NEXT: pand %xmm5, %xmm0
-; SSE2-SSSE3-NEXT: pandn %xmm4, %xmm5
+; SSE2-SSSE3-NEXT: pshufd {{.*#+}} xmm5 = xmm7[1,1,3,3]
; SSE2-SSSE3-NEXT: por %xmm0, %xmm5
+; SSE2-SSSE3-NEXT: pand %xmm5, %xmm1
+; SSE2-SSSE3-NEXT: pandn %xmm4, %xmm5
+; SSE2-SSSE3-NEXT: por %xmm1, %xmm5
; SSE2-SSSE3-NEXT: movdqa %xmm5, %xmm0
; SSE2-SSSE3-NEXT: pxor %xmm2, %xmm0
; SSE2-SSSE3-NEXT: movdqa %xmm0, %xmm1
; SSE2-SSSE3-NEXT: pcmpgtd %xmm2, %xmm1
-; SSE2-SSSE3-NEXT: pshufd {{.*#+}} xmm4 = xmm1[0,0,2,2]
; SSE2-SSSE3-NEXT: pcmpeqd %xmm2, %xmm0
; SSE2-SSSE3-NEXT: pshufd {{.*#+}} xmm0 = xmm0[1,1,3,3]
-; SSE2-SSSE3-NEXT: pand %xmm4, %xmm0
+; SSE2-SSSE3-NEXT: pand %xmm1, %xmm0
; SSE2-SSSE3-NEXT: pshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
; SSE2-SSSE3-NEXT: por %xmm0, %xmm1
; SSE2-SSSE3-NEXT: pand %xmm5, %xmm1
@@ -1350,19 +1346,17 @@ define void @trunc_packus_v4i64_v4i16_store(<4 x i64> %a0, ptr%p1) {
; SSE2-SSSE3-NEXT: pxor %xmm2, %xmm0
; SSE2-SSSE3-NEXT: movdqa %xmm0, %xmm4
; SSE2-SSSE3-NEXT: pcmpgtd %xmm2, %xmm4
-; SSE2-SSSE3-NEXT: pshufd {{.*#+}} xmm5 = xmm4[0,0,2,2]
; SSE2-SSSE3-NEXT: pcmpeqd %xmm2, %xmm0
; SSE2-SSSE3-NEXT: pshufd {{.*#+}} xmm0 = xmm0[1,1,3,3]
-; SSE2-SSSE3-NEXT: pand %xmm5, %xmm0
+; SSE2-SSSE3-NEXT: pand %xmm4, %xmm0
; SSE2-SSSE3-NEXT: pshufd {{.*#+}} xmm2 = xmm4[1,1,3,3]
; SSE2-SSSE3-NEXT: por %xmm0, %xmm2
; SSE2-SSSE3-NEXT: pand %xmm3, %xmm2
-; SSE2-SSSE3-NEXT: pshufd {{.*#+}} xmm0 = xmm2[0,2,2,3]
-; SSE2-SSSE3-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[0,2,2,3,4,5,6,7]
-; SSE2-SSSE3-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
-; SSE2-SSSE3-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[0,2,2,3,4,5,6,7]
-; SSE2-SSSE3-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE2-SSSE3-NEXT: movq %xmm1, (%rdi)
+; SSE2-SSSE3-NEXT: shufps {{.*#+}} xmm2 = xmm2[0,2],xmm1[0,2]
+; SSE2-SSSE3-NEXT: pslld $16, %xmm2
+; SSE2-SSSE3-NEXT: psrad $16, %xmm2
+; SSE2-SSSE3-NEXT: packssdw %xmm2, %xmm2
+; SSE2-SSSE3-NEXT: movq %xmm2, (%rdi)
; SSE2-SSSE3-NEXT: retq
;
; SSE41-LABEL: trunc_packus_v4i64_v4i16_store:
@@ -2493,60 +2487,60 @@ define void @trunc_packus_v2i64_v2i8_store(<2 x i64> %a0, ptr%p1) {
define <4 x i8> @trunc_packus_v4i64_v4i8(<4 x i64> %a0) {
; SSE2-LABEL: trunc_packus_v4i64_v4i8:
; SSE2: # %bb.0:
-; SSE2-NEXT: movdqa {{.*#+}} xmm2 = [255,255]
-; SSE2-NEXT: movdqa {{.*#+}} xmm4 = [2147483648,2147483648]
-; SSE2-NEXT: movdqa %xmm1, %xmm3
-; SSE2-NEXT: pxor %xmm4, %xmm3
-; SSE2-NEXT: pshufd {{.*#+}} xmm5 = xmm3[1,1,3,3]
-; SSE2-NEXT: pxor %xmm6, %xmm6
-; SSE2-NEXT: pcmpeqd %xmm6, %xmm5
+; SSE2-NEXT: movdqa {{.*#+}} xmm3 = [255,255]
+; SSE2-NEXT: movdqa {{.*#+}} xmm5 = [2147483648,2147483648]
+; SSE2-NEXT: movdqa %xmm1, %xmm4
+; SSE2-NEXT: pxor %xmm5, %xmm4
+; SSE2-NEXT: pshufd {{.*#+}} xmm6 = xmm4[1,1,3,3]
+; SSE2-NEXT: pxor %xmm2, %xmm2
+; SSE2-NEXT: pcmpeqd %xmm2, %xmm6
; SSE2-NEXT: movdqa {{.*#+}} xmm7 = [2147483903,2147483903]
; SSE2-NEXT: movdqa %xmm7, %xmm8
-; SSE2-NEXT: pcmpgtd %xmm3, %xmm8
+; SSE2-NEXT: pcmpgtd %xmm4, %xmm8
; SSE2-NEXT: pshufd {{.*#+}} xmm9 = xmm8[0,0,2,2]
-; SSE2-NEXT: pand %xmm5, %xmm9
-; SSE2-NEXT: pshufd {{.*#+}} xmm3 = xmm8[1,1,3,3]
-; SSE2-NEXT: por %xmm9, %xmm3
-; SSE2-NEXT: pand %xmm3, %xmm1
-; SSE2-NEXT: pandn %xmm2, %xmm3
-; SSE2-NEXT: por %xmm1, %xmm3
+; SSE2-NEXT: pand %xmm6, %xmm9
+; SSE2-NEXT: pshufd {{.*#+}} xmm4 = xmm8[1,1,3,3]
+; SSE2-NEXT: por %xmm9, %xmm4
+; SSE2-NEXT: pand %xmm4, %xmm1
+; SSE2-NEXT: pandn %xmm3, %xmm4
+; SSE2-NEXT: por %xmm1, %xmm4
; SSE2-NEXT: movdqa %xmm0, %xmm1
-; SSE2-NEXT: pxor %xmm4, %xmm1
-; SSE2-NEXT: pshufd {{.*#+}} xmm5 = xmm1[1,1,3,3]
-; SSE2-NEXT: pcmpeqd %xmm6, %xmm5
+; SSE2-NEXT: pxor %xmm5, %xmm1
+; SSE2-NEXT: pshufd {{.*#+}} xmm6 = xmm1[1,1,3,3]
+; SSE2-NEXT: pcmpeqd %xmm2, %xmm6
; SSE2-NEXT: pcmpgtd %xmm1, %xmm7
; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm7[0,0,2,2]
-; SSE2-NEXT: pand %xmm5, %xmm1
-; SSE2-NEXT: pshufd {{.*#+}} xmm5 = xmm7[1,1,3,3]
-; SSE2-NEXT: por %xmm1, %xmm5
-; SSE2-NEXT: pand %xmm5, %xmm0
-; SSE2-NEXT: pandn %xmm2, %xmm5
-; SSE2-NEXT: por %xmm5, %xmm0
+; SSE2-NEXT: pand %xmm6, %xmm1
+; SSE2-NEXT: pshufd {{.*#+}} xmm6 = xmm7[1,1,3,3]
+; SSE2-NEXT: por %xmm1, %xmm6
+; SSE2-NEXT: pand %xmm6, %xmm0
+; SSE2-NEXT: pandn %xmm3, %xmm6
+; SSE2-NEXT: por %xmm6, %xmm0
; SSE2-NEXT: movdqa %xmm0, %xmm1
-; SSE2-NEXT: pxor %xmm4, %xmm1
-; SSE2-NEXT: movdqa %xmm1, %xmm5
-; SSE2-NEXT: pcmpgtd %xmm4, %xmm5
-; SSE2-NEXT: pcmpeqd %xmm4, %xmm1
-; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
-; SSE2-NEXT: pand %xmm5, %xmm1
-; SSE2-NEXT: pshufd {{.*#+}} xmm5 = xmm5[1,1,3,3]
-; SSE2-NEXT: por %xmm1, %xmm5
-; SSE2-NEXT: movdqa %xmm3, %xmm1
-; SSE2-NEXT: pxor %xmm4, %xmm1
+; SSE2-NEXT: pxor %xmm5, %xmm1
; SSE2-NEXT: movdqa %xmm1, %xmm6
-; SSE2-NEXT: pcmpgtd %xmm4, %xmm6
-; SSE2-NEXT: pcmpeqd %xmm4, %xmm1
+; SSE2-NEXT: pcmpgtd %xmm5, %xmm6
+; SSE2-NEXT: pcmpeqd %xmm5, %xmm1
; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
; SSE2-NEXT: pand %xmm6, %xmm1
-; SSE2-NEXT: pshufd {{.*#+}} xmm4 = xmm6[1,1,3,3]
-; SSE2-NEXT: por %xmm1, %xmm4
-; SSE2-NEXT: pand %xmm2, %xmm3
-; SSE2-NEXT: pand %xmm4, %xmm3
-; SSE2-NEXT: pand %xmm2, %xmm0
-; SSE2-NEXT: pand %xmm5, %xmm0
-; SSE2-NEXT: packuswb %xmm3, %xmm0
-; SSE2-NEXT: packuswb %xmm0, %xmm0
-; SSE2-NEXT: packuswb %xmm0, %xmm0
+; SSE2-NEXT: pshufd {{.*#+}} xmm6 = xmm6[1,1,3,3]
+; SSE2-NEXT: por %xmm1, %xmm6
+; SSE2-NEXT: movdqa %xmm4, %xmm1
+; SSE2-NEXT: pxor %xmm5, %xmm1
+; SSE2-NEXT: movdqa %xmm1, %xmm7
+; SSE2-NEXT: pcmpgtd %xmm5, %xmm7
+; SSE2-NEXT: pcmpeqd %xmm5, %xmm1
+; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
+; SSE2-NEXT: pand %xmm7, %xmm1
+; SSE2-NEXT: pshufd {{.*#+}} xmm5 = xmm7[1,1,3,3]
+; SSE2-NEXT: por %xmm1, %xmm5
+; SSE2-NEXT: pand %xmm3, %xmm4
+; SSE2-NEXT: pand %xmm5, %xmm4
+; SSE2-NEXT: pand %xmm3, %xmm0
+; SSE2-NEXT: pand %xmm6, %xmm0
+; SSE2-NEXT: packuswb %xmm4, %xmm0
+; SSE2-NEXT: packuswb %xmm2, %xmm0
+; SSE2-NEXT: packuswb %xmm2, %xmm0
; SSE2-NEXT: retq
;
; SSSE3-LABEL: trunc_packus_v4i64_v4i8:
diff --git a/llvm/test/CodeGen/X86/vector-trunc-ssat.ll b/llvm/test/CodeGen/X86/vector-trunc-ssat.ll
index baed531bc9330f..0a34487b22637c 100644
--- a/llvm/test/CodeGen/X86/vector-trunc-ssat.ll
+++ b/llvm/test/CodeGen/X86/vector-trunc-ssat.ll
@@ -2808,67 +2808,67 @@ define void @trunc_ssat_v2i64_v2i8_store(<2 x i64> %a0, ptr%p1) {
define <4 x i8> @trunc_ssat_v4i64_v4i8(<4 x i64> %a0) {
; SSE2-LABEL: trunc_ssat_v4i64_v4i8:
; SSE2: # %bb.0:
-; SSE2-NEXT: movdqa {{.*#+}} xmm4 = [127,127]
-; SSE2-NEXT: movdqa {{.*#+}} xmm2 = [2147483648,2147483648]
-; SSE2-NEXT: movdqa %xmm1, %xmm3
-; SSE2-NEXT: pxor %xmm2, %xmm3
-; SSE2-NEXT: pshufd {{.*#+}} xmm5 = xmm3[1,1,3,3]
-; SSE2-NEXT: pxor %xmm6, %xmm6
-; SSE2-NEXT: pcmpeqd %xmm6, %xmm5
+; SSE2-NEXT: movdqa {{.*#+}} xmm5 = [127,127]
+; SSE2-NEXT: movdqa {{.*#+}} xmm3 = [2147483648,2147483648]
+; SSE2-NEXT: movdqa %xmm1, %xmm4
+; SSE2-NEXT: pxor %xmm3, %xmm4
+; SSE2-NEXT: pshufd {{.*#+}} xmm6 = xmm4[1,1,3,3]
+; SSE2-NEXT: pxor %xmm2, %xmm2
+; SSE2-NEXT: pcmpeqd %xmm2, %xmm6
; SSE2-NEXT: movdqa {{.*#+}} xmm7 = [2147483775,2147483775]
; SSE2-NEXT: movdqa %xmm7, %xmm8
-; SSE2-NEXT: pcmpgtd %xmm3, %xmm8
+; SSE2-NEXT: pcmpgtd %xmm4, %xmm8
; SSE2-NEXT: pshufd {{.*#+}} xmm9 = xmm8[0,0,2,2]
-; SSE2-NEXT: pand %xmm5, %xmm9
-; SSE2-NEXT: pshufd {{.*#+}} xmm3 = xmm8[1,1,3,3]
-; SSE2-NEXT: por %xmm9, %xmm3
-; SSE2-NEXT: pand %xmm3, %xmm1
-; SSE2-NEXT: pandn %xmm4, %xmm3
-; SSE2-NEXT: por %xmm1, %xmm3
+; SSE2-NEXT: pand %xmm6, %xmm9
+; SSE2-NEXT: pshufd {{.*#+}} xmm4 = xmm8[1,1,3,3]
+; SSE2-NEXT: por %xmm9, %xmm4
+; SSE2-NEXT: pand %xmm4, %xmm1
+; SSE2-NEXT: pandn %xmm5, %xmm4
+; SSE2-NEXT: por %xmm1, %xmm4
; SSE2-NEXT: movdqa %xmm0, %xmm1
-; SSE2-NEXT: pxor %xmm2, %xmm1
-; SSE2-NEXT: pshufd {{.*#+}} xmm5 = xmm1[1,1,3,3]
-; SSE2-NEXT: pcmpeqd %xmm6, %xmm5
+; SSE2-NEXT: pxor %xmm3, %xmm1
+; SSE2-NEXT: pshufd {{.*#+}} xmm6 = xmm1[1,1,3,3]
+; SSE2-NEXT: pcmpeqd %xmm2, %xmm6
; SSE2-NEXT: pcmpgtd %xmm1, %xmm7
; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm7[0,0,2,2]
-; SSE2-NEXT: pand %xmm5, %xmm1
-; SSE2-NEXT: pshufd {{.*#+}} xmm5 = xmm7[1,1,3,3]
-; SSE2-NEXT: por %xmm1, %xmm5
+; SSE2-NEXT: pand %xmm6, %xmm1
+; SSE2-NEXT: pshufd {{.*#+}} xmm6 = xmm7[1,1,3,3]
+; SSE2-NEXT: por %xmm1, %xmm6
+; SSE2-NEXT: pand %xmm6, %xmm0
+; SSE2-NEXT: pandn %xmm5, %xmm6
+; SSE2-NEXT: por %xmm6, %xmm0
+; SSE2-NEXT: movdqa {{.*#+}} xmm1 = [18446744073709551488,18446744073709551488]
+; SSE2-NEXT: movdqa %xmm0, %xmm5
+; SSE2-NEXT: pxor %xmm3, %xmm5
+; SSE2-NEXT: pshufd {{.*#+}} xmm6 = xmm5[1,1,3,3]
+; SSE2-NEXT: pcmpeqd %xmm7, %xmm7
+; SSE2-NEXT: pcmpeqd %xmm7, %xmm6
+; SSE2-NEXT: movdqa {{.*#+}} xmm8 = [18446744071562067840,18446744071562067840]
+; SSE2-NEXT: pcmpgtd %xmm8, %xmm5
+; SSE2-NEXT: pshufd {{.*#+}} xmm9 = xmm5[0,0,2,2]
+; SSE2-NEXT: pand %xmm6, %xmm9
+; SSE2-NEXT: pshufd {{.*#+}} xmm5 = xmm5[1,1,3,3]
+; SSE2-NEXT: por %xmm9, %xmm5
; SSE2-NEXT: pand %xmm5, %xmm0
-; SSE2-NEXT: pandn %xmm4, %xmm5
+; SSE2-NEXT: pandn %xmm1, %xmm5
; SSE2-NEXT: por %xmm5, %xmm0
-; SSE2-NEXT: movdqa {{.*#+}} xmm1 = [18446744073709551488,18446744073709551488]
-; SSE2-NEXT: movdqa %xmm0, %xmm4
-; SSE2-NEXT: pxor %xmm2, %xmm4
-; SSE2-NEXT: pshufd {{.*#+}} xmm5 = xmm4[1,1,3,3]
-; SSE2-NEXT: pcmpeqd %xmm6, %xmm6
-; SSE2-NEXT: pcmpeqd %xmm6, %xmm5
-; SSE2-NEXT: movdqa {{.*#+}} xmm7 = [18446744071562067840,18446744071562067840]
-; SSE2-NEXT: pcmpgtd %xmm7, %xmm4
-; SSE2-NEXT: pshufd {{.*#+}} xmm8 = xmm4[0,0,2,2]
-; SSE2-NEXT: pand %xmm5, %xmm8
-; SSE2-NEXT: pshufd {{.*#+}} xmm4 = xmm4[1,1,3,3]
-; SSE2-NEXT: por %xmm8, %xmm4
-; SSE2-NEXT: pand %xmm4, %xmm0
-; SSE2-NEXT: pandn %xmm1, %xmm4
-; SSE2-NEXT: por %xmm4, %xmm0
-; SSE2-NEXT: pxor %xmm3, %xmm2
-; SSE2-NEXT: pshufd {{.*#+}} xmm4 = xmm2[1,1,3,3]
-; SSE2-NEXT: pcmpeqd %xmm6, %xmm4
-; SSE2-NEXT: pcmpgtd %xmm7, %xmm2
-; SSE2-NEXT: pshufd {{.*#+}} xmm5 = xmm2[0,0,2,2]
-; SSE2-NEXT: pand %xmm4, %xmm5
-; SSE2-NEXT: pshufd {{.*#+}} xmm2 = xmm2[1,1,3,3]
-; SSE2-NEXT: por %xmm5, %xmm2
-; SSE2-NEXT: pand %xmm2, %xmm3
-; SSE2-NEXT: pandn %xmm1, %xmm2
-; SSE2-NEXT: por %xmm3, %xmm2
-; SSE2-NEXT: movdqa {{.*#+}} xmm1 = [255,0,0,0,255,0,0,0]
-; SSE2-NEXT: pand %xmm1, %xmm2
+; SSE2-NEXT: pxor %xmm4, %xmm3
+; SSE2-NEXT: pshufd {{.*#+}} xmm5 = xmm3[1,1,3,3]
+; SSE2-NEXT: pcmpeqd %xmm7, %xmm5
+; SSE2-NEXT: pcmpgtd %xmm8, %xmm3
+; SSE2-NEXT: pshufd {{.*#+}} xmm6 = xmm3[0,0,2,2]
+; SSE2-NEXT: pand %xmm5, %xmm6
+; SSE2-NEXT: pshufd {{.*#+}} xmm3 = xmm3[1,1,3,3]
+; SSE2-NEXT: por %xmm6, %xmm3
+; SSE2-NEXT: pand %xmm3, %xmm4
+; SSE2-NEXT: pandn %xmm1, %xmm3
+; SSE2-NEXT: por %xmm4, %xmm3
+; SSE2-NEXT: movdqa {{.*#+}} xmm1 = [255,0,0,0,0,0,0,0,255,0,0,0,0,0,0,0]
+; SSE2-NEXT: pand %xmm1, %xmm3
; SSE2-NEXT: pand %xmm1, %xmm0
+; SSE2-NEXT: packuswb %xmm3, %xmm0
+; SSE2-NEXT: packuswb %xmm2, %xmm0
; SSE2-NEXT: packuswb %xmm2, %xmm0
-; SSE2-NEXT: packuswb %xmm0, %xmm0
-; SSE2-NEXT: packuswb %xmm0, %xmm0
; SSE2-NEXT: retq
;
; SSSE3-LABEL: trunc_ssat_v4i64_v4i8:
@@ -3122,7 +3122,7 @@ define void @trunc_ssat_v4i64_v4i8_store(<4 x i64> %a0, ptr%p1) {
; SSE2-NEXT: pand %xmm2, %xmm3
; SSE2-NEXT: pandn %xmm1, %xmm2
; SSE2-NEXT: por %xmm3, %xmm2
-; SSE2-NEXT: movdqa {{.*#+}} xmm1 = [255,0,0,0,255,0,0,0]
+; SSE2-NEXT: movdqa {{.*#+}} xmm1 = [255,0,0,0,0,0,0,0,255,0,0,0,0,0,0,0]
; SSE2-NEXT: pand %xmm1, %xmm2
; SSE2-NEXT: pand %xmm1, %xmm0
; SSE2-NEXT: packuswb %xmm2, %xmm0
diff --git a/llvm/test/CodeGen/X86/vector-trunc-usat.ll b/llvm/test/CodeGen/X86/vector-trunc-usat.ll
index b65316763aad77..cfbde2aeccf70b 100644
--- a/llvm/test/CodeGen/X86/vector-trunc-usat.ll
+++ b/llvm/test/CodeGen/X86/vector-trunc-usat.ll
@@ -817,7 +817,7 @@ define <4 x i16> @trunc_usat_v4i64_v4i16(<4 x i64> %a0) {
; SSE2-SSSE3-LABEL: trunc_usat_v4i64_v4i16:
; SSE2-SSSE3: # %bb.0:
; SSE2-SSSE3-NEXT: movdqa {{.*#+}} xmm2 = [9223372039002259456,9223372039002259456]
-; SSE2-SSSE3-NEXT: movdqa %xmm0, %xmm3
+; SSE2-SSSE3-NEXT: movdqa %xmm1, %xmm3
; SSE2-SSSE3-NEXT: pxor %xmm2, %xmm3
; SSE2-SSSE3-NEXT: pshufd {{.*#+}} xmm4 = xmm3[0,0,2,2]
; SSE2-SSSE3-NEXT: movdqa {{.*#+}} xmm5 = [2147549183,2147549183,2147549183,2147549183]
@@ -827,24 +827,24 @@ define <4 x i16> @trunc_usat_v4i64_v4i16(<4 x i64> %a0) {
; SSE2-SSSE3-NEXT: pcmpeqd %xmm2, %xmm3
; SSE2-SSSE3-NEXT: pand %xmm6, %xmm3
; SSE2-SSSE3-NEXT: pcmpeqd %xmm4, %xmm4
-; SSE2-SSSE3-NEXT: pand %xmm3, %xmm0
+; SSE2-SSSE3-NEXT: pand %xmm3, %xmm1
; SSE2-SSSE3-NEXT: pxor %xmm4, %xmm3
-; SSE2-SSSE3-NEXT: por %xmm0, %xmm3
-; SSE2-SSSE3-NEXT: movdqa %xmm1, %xmm0
-; SSE2-SSSE3-NEXT: pxor %xmm2, %xmm0
-; SSE2-SSSE3-NEXT: pshufd {{.*#+}} xmm6 = xmm0[0,0,2,2]
+; SSE2-SSSE3-NEXT: por %xmm1, %xmm3
+; SSE2-SSSE3-NEXT: movdqa %xmm0, %xmm1
+; SSE2-SSSE3-NEXT: pxor %xmm2, %xmm1
+; SSE2-SSSE3-NEXT: pshufd {{.*#+}} xmm6 = xmm1[0,0,2,2]
; SSE2-SSSE3-NEXT: pcmpgtd %xmm6, %xmm5
-; SSE2-SSSE3-NEXT: pshufd {{.*#+}} xmm0 = xmm0[1,1,3,3]
-; SSE2-SSSE3-NEXT: pcmpeqd %xmm2, %xmm0
-; SSE2-SSSE3-NEXT: pand %xmm5, %xmm0
-; SSE2-SSSE3-NEXT: pxor %xmm0, %xmm4
+; SSE2-SSSE3-NEXT: pshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
+; SSE2-SSSE3-NEXT: pcmpeqd %xmm2, %xmm1
+; SSE2-SSSE3-NEXT: pand %xmm5, %xmm1
+; SSE2-SSSE3-NEXT: pxor %xmm1, %xmm4
; SSE2-SSSE3-NEXT: pand %xmm1, %xmm0
; SSE2-SSSE3-NEXT: por %xmm4, %xmm0
-; SSE2-SSSE3-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
-; SSE2-SSSE3-NEXT: pshuflw {{.*#+}} xmm1 = xmm0[0,2,2,3,4,5,6,7]
-; SSE2-SSSE3-NEXT: pshufd {{.*#+}} xmm0 = xmm3[0,2,2,3]
-; SSE2-SSSE3-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[0,2,2,3,4,5,6,7]
-; SSE2-SSSE3-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE2-SSSE3-NEXT: shufps {{.*#+}} xmm0 = xmm0[0,2],xmm3[0,2]
+; SSE2-SSSE3-NEXT: pslld $16, %xmm0
+; SSE2-SSSE3-NEXT: psrad $16, %xmm0
+; SSE2-SSSE3-NEXT: pxor %xmm1, %xmm1
+; SSE2-SSSE3-NEXT: packssdw %xmm1, %xmm0
; SSE2-SSSE3-NEXT: retq
;
; SSE41-LABEL: trunc_usat_v4i64_v4i16:
@@ -951,7 +951,7 @@ define void @trunc_usat_v4i64_v4i16_store(<4 x i64> %a0, ptr%p1) {
; SSE2-SSSE3-LABEL: trunc_usat_v4i64_v4i16_store:
; SSE2-SSSE3: # %bb.0:
; SSE2-SSSE3-NEXT: movdqa {{.*#+}} xmm2 = [9223372039002259456,9223372039002259456]
-; SSE2-SSSE3-NEXT: movdqa %xmm0, %xmm3
+; SSE2-SSSE3-NEXT: movdqa %xmm1, %xmm3
; SSE2-SSSE3-NEXT: pxor %xmm2, %xmm3
; SSE2-SSSE3-NEXT: pshufd {{.*#+}} xmm4 = xmm3[0,0,2,2]
; SSE2-SSSE3-NEXT: movdqa {{.*#+}} xmm5 = [2147549183,2147549183,2147549183,2147549183]
@@ -961,24 +961,23 @@ define void @trunc_usat_v4i64_v4i16_store(<4 x i64> %a0, ptr%p1) {
; SSE2-SSSE3-NEXT: pcmpeqd %xmm2, %xmm3
; SSE2-SSSE3-NEXT: pand %xmm6, %xmm3
; SSE2-SSSE3-NEXT: pcmpeqd %xmm4, %xmm4
-; SSE2-SSSE3-NEXT: pand %xmm3, %xmm0
+; SSE2-SSSE3-NEXT: pand %xmm3, %xmm1
; SSE2-SSSE3-NEXT: pxor %xmm4, %xmm3
-; SSE2-SSSE3-NEXT: por %xmm0, %xmm3
-; SSE2-SSSE3-NEXT: movdqa %xmm1, %xmm0
-; SSE2-SSSE3-NEXT: pxor %xmm2, %xmm0
-; SSE2-SSSE3-NEXT: pshufd {{.*#+}} xmm6 = xmm0[0,0,2,2]
+; SSE2-SSSE3-NEXT: por %xmm1, %xmm3
+; SSE2-SSSE3-NEXT: movdqa %xmm0, %xmm1
+; SSE2-SSSE3-NEXT: pxor %xmm2, %xmm1
+; SSE2-SSSE3-NEXT: pshufd {{.*#+}} xmm6 = xmm1[0,0,2,2]
; SSE2-SSSE3-NEXT: pcmpgtd %xmm6, %xmm5
-; SSE2-SSSE3-NEXT: pshufd {{.*#+}} xmm0 = xmm0[1,1,3,3]
-; SSE2-SSSE3-NEXT: pcmpeqd %xmm2, %xmm0
-; SSE2-SSSE3-NEXT: pand %xmm5, %xmm0
-; SSE2-SSSE3-NEXT: pxor %xmm0, %xmm4
-; SSE2-SSSE3-NEXT: pand %xmm1, %xmm0
-; SSE2-SSSE3-NEXT: por %xmm4, %xmm0
-; SSE2-SSSE3-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
-; SSE2-SSSE3-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[0,2,2,3,4,5,6,7]
-; SSE2-SSSE3-NEXT: pshufd {{.*#+}} xmm1 = xmm3[0,2,2,3]
-; SSE2-SSSE3-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[0,2,2,3,4,5,6,7]
-; SSE2-SSSE3-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE2-SSSE3-NEXT: pshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
+; SSE2-SSSE3-NEXT: pcmpeqd %xmm2, %xmm1
+; SSE2-SSSE3-NEXT: pand %xmm5, %xmm1
+; SSE2-SSSE3-NEXT: pxor %xmm1, %xmm4
+; SSE2-SSSE3-NEXT: pand %xmm0, %xmm1
+; SSE2-SSSE3-NEXT: por %xmm4, %xmm1
+; SSE2-SSSE3-NEXT: shufps {{.*#+}} xmm1 = xmm1[0,2],xmm3[0,2]
+; SSE2-SSSE3-NEXT: pslld $16, %xmm1
+; SSE2-SSSE3-NEXT: psrad $16, %xmm1
+; SSE2-SSSE3-NEXT: packssdw %xmm1, %xmm1
; SSE2-SSSE3-NEXT: movq %xmm1, (%rdi)
; SSE2-SSSE3-NEXT: retq
;
@@ -1922,12 +1921,13 @@ define <4 x i8> @trunc_usat_v4i64_v4i8(<4 x i64> %a0) {
; SSE2-NEXT: pxor %xmm3, %xmm4
; SSE2-NEXT: pand %xmm1, %xmm3
; SSE2-NEXT: por %xmm4, %xmm3
-; SSE2-NEXT: movdqa {{.*#+}} xmm1 = [255,0,0,0,255,0,0,0]
+; SSE2-NEXT: movdqa {{.*#+}} xmm1 = [255,0,0,0,0,0,0,0,255,0,0,0,0,0,0,0]
; SSE2-NEXT: pand %xmm1, %xmm3
; SSE2-NEXT: pand %xmm1, %xmm0
; SSE2-NEXT: packuswb %xmm3, %xmm0
-; SSE2-NEXT: packuswb %xmm0, %xmm0
-; SSE2-NEXT: packuswb %xmm0, %xmm0
+; SSE2-NEXT: pxor %xmm1, %xmm1
+; SSE2-NEXT: packuswb %xmm1, %xmm0
+; SSE2-NEXT: packuswb %xmm1, %xmm0
; SSE2-NEXT: retq
;
; SSSE3-LABEL: trunc_usat_v4i64_v4i8:
@@ -2094,7 +2094,7 @@ define void @trunc_usat_v4i64_v4i8_store(<4 x i64> %a0, ptr%p1) {
; SSE2-NEXT: pxor %xmm0, %xmm4
; SSE2-NEXT: pand %xmm1, %xmm0
; SSE2-NEXT: por %xmm4, %xmm0
-; SSE2-NEXT: movdqa {{.*#+}} xmm1 = [255,0,0,0,255,0,0,0]
+; SSE2-NEXT: movdqa {{.*#+}} xmm1 = [255,0,0,0,0,0,0,0,255,0,0,0,0,0,0,0]
; SSE2-NEXT: pand %xmm1, %xmm0
; SSE2-NEXT: pand %xmm1, %xmm2
; SSE2-NEXT: packuswb %xmm0, %xmm2
diff --git a/llvm/test/CodeGen/X86/vector-trunc.ll b/llvm/test/CodeGen/X86/vector-trunc.ll
index ee7d7ca7e38fdf..92ee258c13a743 100644
--- a/llvm/test/CodeGen/X86/vector-trunc.ll
+++ b/llvm/test/CodeGen/X86/vector-trunc.ll
@@ -1284,28 +1284,24 @@ entry:
define <8 x i16> @trunc2x4i64_8i16(<4 x i64> %a, <4 x i64> %b) {
; SSE2-SSSE3-LABEL: trunc2x4i64_8i16:
; SSE2-SSSE3: # %bb.0: # %entry
-; SSE2-SSSE3-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
-; SSE2-SSSE3-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[0,2,2,3,4,5,6,7]
-; SSE2-SSSE3-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
-; SSE2-SSSE3-NEXT: pshuflw {{.*#+}} xmm4 = xmm0[0,2,2,3,4,5,6,7]
-; SSE2-SSSE3-NEXT: punpckldq {{.*#+}} xmm4 = xmm4[0],xmm1[0],xmm4[1],xmm1[1]
-; SSE2-SSSE3-NEXT: pshufd {{.*#+}} xmm0 = xmm3[0,2,2,3]
-; SSE2-SSSE3-NEXT: pshuflw {{.*#+}} xmm1 = xmm0[0,1,0,2,4,5,6,7]
-; SSE2-SSSE3-NEXT: pshufd {{.*#+}} xmm0 = xmm2[0,2,2,3]
-; SSE2-SSSE3-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[0,1,0,2,4,5,6,7]
-; SSE2-SSSE3-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; SSE2-SSSE3-NEXT: movsd {{.*#+}} xmm0 = xmm4[0],xmm0[1]
+; SSE2-SSSE3-NEXT: shufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2]
+; SSE2-SSSE3-NEXT: pslld $16, %xmm0
+; SSE2-SSSE3-NEXT: psrad $16, %xmm0
+; SSE2-SSSE3-NEXT: shufps {{.*#+}} xmm2 = xmm2[0,2],xmm3[0,2]
+; SSE2-SSSE3-NEXT: pslld $16, %xmm2
+; SSE2-SSSE3-NEXT: psrad $16, %xmm2
+; SSE2-SSSE3-NEXT: packssdw %xmm2, %xmm0
; SSE2-SSSE3-NEXT: retq
;
; SSE41-LABEL: trunc2x4i64_8i16:
; SSE41: # %bb.0: # %entry
; SSE41-NEXT: pxor %xmm4, %xmm4
-; SSE41-NEXT: pblendw {{.*#+}} xmm3 = xmm3[0],xmm4[1,2,3],xmm3[4],xmm4[5,6,7]
-; SSE41-NEXT: pblendw {{.*#+}} xmm2 = xmm2[0],xmm4[1,2,3],xmm2[4],xmm4[5,6,7]
-; SSE41-NEXT: packusdw %xmm3, %xmm2
; SSE41-NEXT: pblendw {{.*#+}} xmm1 = xmm1[0],xmm4[1,2,3],xmm1[4],xmm4[5,6,7]
; SSE41-NEXT: pblendw {{.*#+}} xmm0 = xmm0[0],xmm4[1,2,3],xmm0[4],xmm4[5,6,7]
; SSE41-NEXT: packusdw %xmm1, %xmm0
+; SSE41-NEXT: pblendw {{.*#+}} xmm3 = xmm3[0],xmm4[1,2,3],xmm3[4],xmm4[5,6,7]
+; SSE41-NEXT: pblendw {{.*#+}} xmm2 = xmm2[0],xmm4[1,2,3],xmm2[4],xmm4[5,6,7]
+; SSE41-NEXT: packusdw %xmm3, %xmm2
; SSE41-NEXT: packusdw %xmm2, %xmm0
; SSE41-NEXT: retq
;
More information about the llvm-commits
mailing list