[llvm] 09a081f - [X86][SSE] LowerADDSAT_SUBSAT - avoid X86ISD::BLENDV in UADDSAT/USUBSAT custom lowering
Simon Pilgrim via llvm-commits
llvm-commits at lists.llvm.org
Fri Nov 20 09:01:20 PST 2020
Author: Simon Pilgrim
Date: 2020-11-20T16:53:01Z
New Revision: 09a081f2213639cf4ea529baa77352ed5f6f18c8
URL: https://github.com/llvm/llvm-project/commit/09a081f2213639cf4ea529baa77352ed5f6f18c8
DIFF: https://github.com/llvm/llvm-project/commit/09a081f2213639cf4ea529baa77352ed5f6f18c8.diff
LOG: [X86][SSE] LowerADDSAT_SUBSAT - avoid X86ISD::BLENDV in UADDSAT/USUBSAT custom lowering
Use the OR(CMP,ADD) / AND(CMP,SUB) patterns like we do on pre-SSE4 targets.
We're still using X86ISD::BLENDV on some AVX targets as we don't do custom lowering for >= 256-bit vectors.
Really this (and combineVSelectWithAllOnesOrZeros) needs moving to DAGCombiner, but pre-SSE42 we see the vXi64 comparison type as a 2 x 32-bits result so we can't just rely on ComputeNumSignBits to give us the 'all bits' result we need.
Added:
Modified:
llvm/lib/Target/X86/X86ISelLowering.cpp
llvm/test/CodeGen/X86/uadd_sat_vec.ll
llvm/test/CodeGen/X86/usub_sat_vec.ll
Removed:
################################################################################
diff --git a/llvm/lib/Target/X86/X86ISelLowering.cpp b/llvm/lib/Target/X86/X86ISelLowering.cpp
index ae4cc6424ae6..dacd4fb2f1ea 100644
--- a/llvm/lib/Target/X86/X86ISelLowering.cpp
+++ b/llvm/lib/Target/X86/X86ISelLowering.cpp
@@ -26879,12 +26879,20 @@ static SDValue LowerADDSAT_SUBSAT(SDValue Op, SelectionDAG &DAG,
// uaddsat X, Y --> (X >u (X + Y)) ? -1 : X + Y
SDValue Add = DAG.getNode(ISD::ADD, DL, VT, X, Y);
SDValue Cmp = DAG.getSetCC(DL, SetCCResultType, X, Add, ISD::SETUGT);
+ // TODO: Move this to DAGCombiner?
+ if (SetCCResultType == VT &&
+ DAG.ComputeNumSignBits(Cmp) == VT.getScalarSizeInBits())
+ return DAG.getNode(ISD::OR, DL, VT, Cmp, Add);
return DAG.getSelect(DL, VT, Cmp, DAG.getAllOnesConstant(DL, VT), Add);
}
if (Opcode == ISD::USUBSAT && !TLI.isOperationLegal(ISD::UMAX, VT)) {
// usubsat X, Y --> (X >u Y) ? X - Y : 0
SDValue Sub = DAG.getNode(ISD::SUB, DL, VT, X, Y);
SDValue Cmp = DAG.getSetCC(DL, SetCCResultType, X, Y, ISD::SETUGT);
+ // TODO: Move this to DAGCombiner?
+ if (SetCCResultType == VT &&
+ DAG.ComputeNumSignBits(Cmp) == VT.getScalarSizeInBits())
+ return DAG.getNode(ISD::AND, DL, VT, Cmp, Sub);
return DAG.getSelect(DL, VT, Cmp, Sub, DAG.getConstant(0, DL, VT));
}
// Use default expansion.
diff --git a/llvm/test/CodeGen/X86/uadd_sat_vec.ll b/llvm/test/CodeGen/X86/uadd_sat_vec.ll
index 1a50b30d14eb..7c8665ec086d 100644
--- a/llvm/test/CodeGen/X86/uadd_sat_vec.ll
+++ b/llvm/test/CodeGen/X86/uadd_sat_vec.ll
@@ -915,57 +915,22 @@ define <16 x i32> @v16i32(<16 x i32> %x, <16 x i32> %y) nounwind {
}
define <2 x i64> @v2i64(<2 x i64> %x, <2 x i64> %y) nounwind {
-; SSE2-LABEL: v2i64:
-; SSE2: # %bb.0:
-; SSE2-NEXT: movdqa {{.*#+}} xmm2 = [9223372039002259456,9223372039002259456]
-; SSE2-NEXT: paddq %xmm0, %xmm1
-; SSE2-NEXT: pxor %xmm2, %xmm0
-; SSE2-NEXT: pxor %xmm1, %xmm2
-; SSE2-NEXT: movdqa %xmm0, %xmm3
-; SSE2-NEXT: pcmpgtd %xmm2, %xmm3
-; SSE2-NEXT: pshufd {{.*#+}} xmm4 = xmm3[0,0,2,2]
-; SSE2-NEXT: pcmpeqd %xmm0, %xmm2
-; SSE2-NEXT: pshufd {{.*#+}} xmm2 = xmm2[1,1,3,3]
-; SSE2-NEXT: pand %xmm4, %xmm2
-; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm3[1,1,3,3]
-; SSE2-NEXT: por %xmm1, %xmm0
-; SSE2-NEXT: por %xmm2, %xmm0
-; SSE2-NEXT: retq
-;
-; SSSE3-LABEL: v2i64:
-; SSSE3: # %bb.0:
-; SSSE3-NEXT: movdqa {{.*#+}} xmm2 = [9223372039002259456,9223372039002259456]
-; SSSE3-NEXT: paddq %xmm0, %xmm1
-; SSSE3-NEXT: pxor %xmm2, %xmm0
-; SSSE3-NEXT: pxor %xmm1, %xmm2
-; SSSE3-NEXT: movdqa %xmm0, %xmm3
-; SSSE3-NEXT: pcmpgtd %xmm2, %xmm3
-; SSSE3-NEXT: pshufd {{.*#+}} xmm4 = xmm3[0,0,2,2]
-; SSSE3-NEXT: pcmpeqd %xmm0, %xmm2
-; SSSE3-NEXT: pshufd {{.*#+}} xmm2 = xmm2[1,1,3,3]
-; SSSE3-NEXT: pand %xmm4, %xmm2
-; SSSE3-NEXT: pshufd {{.*#+}} xmm0 = xmm3[1,1,3,3]
-; SSSE3-NEXT: por %xmm1, %xmm0
-; SSSE3-NEXT: por %xmm2, %xmm0
-; SSSE3-NEXT: retq
-;
-; SSE41-LABEL: v2i64:
-; SSE41: # %bb.0:
-; SSE41-NEXT: movdqa {{.*#+}} xmm2 = [9223372039002259456,9223372039002259456]
-; SSE41-NEXT: paddq %xmm0, %xmm1
-; SSE41-NEXT: movdqa %xmm0, %xmm3
-; SSE41-NEXT: pxor %xmm2, %xmm3
-; SSE41-NEXT: pxor %xmm1, %xmm2
-; SSE41-NEXT: movdqa %xmm3, %xmm0
-; SSE41-NEXT: pcmpeqd %xmm2, %xmm0
-; SSE41-NEXT: pcmpgtd %xmm2, %xmm3
-; SSE41-NEXT: pshufd {{.*#+}} xmm2 = xmm3[0,0,2,2]
-; SSE41-NEXT: pand %xmm2, %xmm0
-; SSE41-NEXT: por %xmm3, %xmm0
-; SSE41-NEXT: pcmpeqd %xmm2, %xmm2
-; SSE41-NEXT: blendvpd %xmm0, %xmm2, %xmm1
-; SSE41-NEXT: movapd %xmm1, %xmm0
-; SSE41-NEXT: retq
+; SSE-LABEL: v2i64:
+; SSE: # %bb.0:
+; SSE-NEXT: movdqa {{.*#+}} xmm2 = [9223372039002259456,9223372039002259456]
+; SSE-NEXT: paddq %xmm0, %xmm1
+; SSE-NEXT: pxor %xmm2, %xmm0
+; SSE-NEXT: pxor %xmm1, %xmm2
+; SSE-NEXT: movdqa %xmm0, %xmm3
+; SSE-NEXT: pcmpgtd %xmm2, %xmm3
+; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm3[0,0,2,2]
+; SSE-NEXT: pcmpeqd %xmm0, %xmm2
+; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm2[1,1,3,3]
+; SSE-NEXT: pand %xmm4, %xmm2
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm3[1,1,3,3]
+; SSE-NEXT: por %xmm1, %xmm0
+; SSE-NEXT: por %xmm2, %xmm0
+; SSE-NEXT: retq
;
; AVX1-LABEL: v2i64:
; AVX1: # %bb.0:
@@ -1010,95 +975,35 @@ define <2 x i64> @v2i64(<2 x i64> %x, <2 x i64> %y) nounwind {
}
define <4 x i64> @v4i64(<4 x i64> %x, <4 x i64> %y) nounwind {
-; SSE2-LABEL: v4i64:
-; SSE2: # %bb.0:
-; SSE2-NEXT: movdqa {{.*#+}} xmm4 = [9223372039002259456,9223372039002259456]
-; SSE2-NEXT: paddq %xmm0, %xmm2
-; SSE2-NEXT: pxor %xmm4, %xmm0
-; SSE2-NEXT: movdqa %xmm2, %xmm5
-; SSE2-NEXT: pxor %xmm4, %xmm5
-; SSE2-NEXT: movdqa %xmm0, %xmm6
-; SSE2-NEXT: pcmpgtd %xmm5, %xmm6
-; SSE2-NEXT: pshufd {{.*#+}} xmm7 = xmm6[0,0,2,2]
-; SSE2-NEXT: pcmpeqd %xmm0, %xmm5
-; SSE2-NEXT: pshufd {{.*#+}} xmm5 = xmm5[1,1,3,3]
-; SSE2-NEXT: pand %xmm7, %xmm5
-; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm6[1,1,3,3]
-; SSE2-NEXT: por %xmm2, %xmm0
-; SSE2-NEXT: por %xmm5, %xmm0
-; SSE2-NEXT: paddq %xmm1, %xmm3
-; SSE2-NEXT: pxor %xmm4, %xmm1
-; SSE2-NEXT: pxor %xmm3, %xmm4
-; SSE2-NEXT: movdqa %xmm1, %xmm2
-; SSE2-NEXT: pcmpgtd %xmm4, %xmm2
-; SSE2-NEXT: pshufd {{.*#+}} xmm5 = xmm2[0,0,2,2]
-; SSE2-NEXT: pcmpeqd %xmm1, %xmm4
-; SSE2-NEXT: pshufd {{.*#+}} xmm4 = xmm4[1,1,3,3]
-; SSE2-NEXT: pand %xmm5, %xmm4
-; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm2[1,1,3,3]
-; SSE2-NEXT: por %xmm3, %xmm1
-; SSE2-NEXT: por %xmm4, %xmm1
-; SSE2-NEXT: retq
-;
-; SSSE3-LABEL: v4i64:
-; SSSE3: # %bb.0:
-; SSSE3-NEXT: movdqa {{.*#+}} xmm4 = [9223372039002259456,9223372039002259456]
-; SSSE3-NEXT: paddq %xmm0, %xmm2
-; SSSE3-NEXT: pxor %xmm4, %xmm0
-; SSSE3-NEXT: movdqa %xmm2, %xmm5
-; SSSE3-NEXT: pxor %xmm4, %xmm5
-; SSSE3-NEXT: movdqa %xmm0, %xmm6
-; SSSE3-NEXT: pcmpgtd %xmm5, %xmm6
-; SSSE3-NEXT: pshufd {{.*#+}} xmm7 = xmm6[0,0,2,2]
-; SSSE3-NEXT: pcmpeqd %xmm0, %xmm5
-; SSSE3-NEXT: pshufd {{.*#+}} xmm5 = xmm5[1,1,3,3]
-; SSSE3-NEXT: pand %xmm7, %xmm5
-; SSSE3-NEXT: pshufd {{.*#+}} xmm0 = xmm6[1,1,3,3]
-; SSSE3-NEXT: por %xmm2, %xmm0
-; SSSE3-NEXT: por %xmm5, %xmm0
-; SSSE3-NEXT: paddq %xmm1, %xmm3
-; SSSE3-NEXT: pxor %xmm4, %xmm1
-; SSSE3-NEXT: pxor %xmm3, %xmm4
-; SSSE3-NEXT: movdqa %xmm1, %xmm2
-; SSSE3-NEXT: pcmpgtd %xmm4, %xmm2
-; SSSE3-NEXT: pshufd {{.*#+}} xmm5 = xmm2[0,0,2,2]
-; SSSE3-NEXT: pcmpeqd %xmm1, %xmm4
-; SSSE3-NEXT: pshufd {{.*#+}} xmm4 = xmm4[1,1,3,3]
-; SSSE3-NEXT: pand %xmm5, %xmm4
-; SSSE3-NEXT: pshufd {{.*#+}} xmm1 = xmm2[1,1,3,3]
-; SSSE3-NEXT: por %xmm3, %xmm1
-; SSSE3-NEXT: por %xmm4, %xmm1
-; SSSE3-NEXT: retq
-;
-; SSE41-LABEL: v4i64:
-; SSE41: # %bb.0:
-; SSE41-NEXT: movdqa {{.*#+}} xmm4 = [9223372039002259456,9223372039002259456]
-; SSE41-NEXT: paddq %xmm0, %xmm2
-; SSE41-NEXT: movdqa %xmm0, %xmm5
-; SSE41-NEXT: pxor %xmm4, %xmm5
-; SSE41-NEXT: movdqa %xmm2, %xmm6
-; SSE41-NEXT: pxor %xmm4, %xmm6
-; SSE41-NEXT: movdqa %xmm5, %xmm0
-; SSE41-NEXT: pcmpeqd %xmm6, %xmm0
-; SSE41-NEXT: pcmpgtd %xmm6, %xmm5
-; SSE41-NEXT: pshufd {{.*#+}} xmm6 = xmm5[0,0,2,2]
-; SSE41-NEXT: pand %xmm6, %xmm0
-; SSE41-NEXT: por %xmm5, %xmm0
-; SSE41-NEXT: pcmpeqd %xmm5, %xmm5
-; SSE41-NEXT: blendvpd %xmm0, %xmm5, %xmm2
-; SSE41-NEXT: paddq %xmm1, %xmm3
-; SSE41-NEXT: pxor %xmm4, %xmm1
-; SSE41-NEXT: pxor %xmm3, %xmm4
-; SSE41-NEXT: movdqa %xmm1, %xmm6
-; SSE41-NEXT: pcmpeqd %xmm4, %xmm6
-; SSE41-NEXT: pcmpgtd %xmm4, %xmm1
-; SSE41-NEXT: pshufd {{.*#+}} xmm0 = xmm1[0,0,2,2]
-; SSE41-NEXT: pand %xmm6, %xmm0
-; SSE41-NEXT: por %xmm1, %xmm0
-; SSE41-NEXT: blendvpd %xmm0, %xmm5, %xmm3
-; SSE41-NEXT: movapd %xmm2, %xmm0
-; SSE41-NEXT: movapd %xmm3, %xmm1
-; SSE41-NEXT: retq
+; SSE-LABEL: v4i64:
+; SSE: # %bb.0:
+; SSE-NEXT: movdqa {{.*#+}} xmm4 = [9223372039002259456,9223372039002259456]
+; SSE-NEXT: paddq %xmm0, %xmm2
+; SSE-NEXT: pxor %xmm4, %xmm0
+; SSE-NEXT: movdqa %xmm2, %xmm5
+; SSE-NEXT: pxor %xmm4, %xmm5
+; SSE-NEXT: movdqa %xmm0, %xmm6
+; SSE-NEXT: pcmpgtd %xmm5, %xmm6
+; SSE-NEXT: pshufd {{.*#+}} xmm7 = xmm6[0,0,2,2]
+; SSE-NEXT: pcmpeqd %xmm0, %xmm5
+; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm5[1,1,3,3]
+; SSE-NEXT: pand %xmm7, %xmm5
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm6[1,1,3,3]
+; SSE-NEXT: por %xmm2, %xmm0
+; SSE-NEXT: por %xmm5, %xmm0
+; SSE-NEXT: paddq %xmm1, %xmm3
+; SSE-NEXT: pxor %xmm4, %xmm1
+; SSE-NEXT: pxor %xmm3, %xmm4
+; SSE-NEXT: movdqa %xmm1, %xmm2
+; SSE-NEXT: pcmpgtd %xmm4, %xmm2
+; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm2[0,0,2,2]
+; SSE-NEXT: pcmpeqd %xmm1, %xmm4
+; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm4[1,1,3,3]
+; SSE-NEXT: pand %xmm5, %xmm4
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm2[1,1,3,3]
+; SSE-NEXT: por %xmm3, %xmm1
+; SSE-NEXT: por %xmm4, %xmm1
+; SSE-NEXT: retq
;
; AVX1-LABEL: v4i64:
; AVX1: # %bb.0:
@@ -1157,175 +1062,61 @@ define <4 x i64> @v4i64(<4 x i64> %x, <4 x i64> %y) nounwind {
}
define <8 x i64> @v8i64(<8 x i64> %x, <8 x i64> %y) nounwind {
-; SSE2-LABEL: v8i64:
-; SSE2: # %bb.0:
-; SSE2-NEXT: movdqa {{.*#+}} xmm8 = [9223372039002259456,9223372039002259456]
-; SSE2-NEXT: paddq %xmm0, %xmm4
-; SSE2-NEXT: pxor %xmm8, %xmm0
-; SSE2-NEXT: movdqa %xmm4, %xmm9
-; SSE2-NEXT: pxor %xmm8, %xmm9
-; SSE2-NEXT: movdqa %xmm0, %xmm10
-; SSE2-NEXT: pcmpgtd %xmm9, %xmm10
-; SSE2-NEXT: pshufd {{.*#+}} xmm11 = xmm10[0,0,2,2]
-; SSE2-NEXT: pcmpeqd %xmm0, %xmm9
-; SSE2-NEXT: pshufd {{.*#+}} xmm9 = xmm9[1,1,3,3]
-; SSE2-NEXT: pand %xmm11, %xmm9
-; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm10[1,1,3,3]
-; SSE2-NEXT: por %xmm4, %xmm0
-; SSE2-NEXT: por %xmm9, %xmm0
-; SSE2-NEXT: paddq %xmm1, %xmm5
-; SSE2-NEXT: pxor %xmm8, %xmm1
-; SSE2-NEXT: movdqa %xmm5, %xmm4
-; SSE2-NEXT: pxor %xmm8, %xmm4
-; SSE2-NEXT: movdqa %xmm1, %xmm9
-; SSE2-NEXT: pcmpgtd %xmm4, %xmm9
-; SSE2-NEXT: pshufd {{.*#+}} xmm10 = xmm9[0,0,2,2]
-; SSE2-NEXT: pcmpeqd %xmm1, %xmm4
-; SSE2-NEXT: pshufd {{.*#+}} xmm4 = xmm4[1,1,3,3]
-; SSE2-NEXT: pand %xmm10, %xmm4
-; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm9[1,1,3,3]
-; SSE2-NEXT: por %xmm5, %xmm1
-; SSE2-NEXT: por %xmm4, %xmm1
-; SSE2-NEXT: paddq %xmm2, %xmm6
-; SSE2-NEXT: pxor %xmm8, %xmm2
-; SSE2-NEXT: movdqa %xmm6, %xmm4
-; SSE2-NEXT: pxor %xmm8, %xmm4
-; SSE2-NEXT: movdqa %xmm2, %xmm5
-; SSE2-NEXT: pcmpgtd %xmm4, %xmm5
-; SSE2-NEXT: pshufd {{.*#+}} xmm9 = xmm5[0,0,2,2]
-; SSE2-NEXT: pcmpeqd %xmm2, %xmm4
-; SSE2-NEXT: pshufd {{.*#+}} xmm4 = xmm4[1,1,3,3]
-; SSE2-NEXT: pand %xmm9, %xmm4
-; SSE2-NEXT: pshufd {{.*#+}} xmm2 = xmm5[1,1,3,3]
-; SSE2-NEXT: por %xmm6, %xmm2
-; SSE2-NEXT: por %xmm4, %xmm2
-; SSE2-NEXT: paddq %xmm3, %xmm7
-; SSE2-NEXT: pxor %xmm8, %xmm3
-; SSE2-NEXT: pxor %xmm7, %xmm8
-; SSE2-NEXT: movdqa %xmm3, %xmm4
-; SSE2-NEXT: pcmpgtd %xmm8, %xmm4
-; SSE2-NEXT: pshufd {{.*#+}} xmm5 = xmm4[0,0,2,2]
-; SSE2-NEXT: pcmpeqd %xmm3, %xmm8
-; SSE2-NEXT: pshufd {{.*#+}} xmm6 = xmm8[1,1,3,3]
-; SSE2-NEXT: pand %xmm5, %xmm6
-; SSE2-NEXT: pshufd {{.*#+}} xmm3 = xmm4[1,1,3,3]
-; SSE2-NEXT: por %xmm7, %xmm3
-; SSE2-NEXT: por %xmm6, %xmm3
-; SSE2-NEXT: retq
-;
-; SSSE3-LABEL: v8i64:
-; SSSE3: # %bb.0:
-; SSSE3-NEXT: movdqa {{.*#+}} xmm8 = [9223372039002259456,9223372039002259456]
-; SSSE3-NEXT: paddq %xmm0, %xmm4
-; SSSE3-NEXT: pxor %xmm8, %xmm0
-; SSSE3-NEXT: movdqa %xmm4, %xmm9
-; SSSE3-NEXT: pxor %xmm8, %xmm9
-; SSSE3-NEXT: movdqa %xmm0, %xmm10
-; SSSE3-NEXT: pcmpgtd %xmm9, %xmm10
-; SSSE3-NEXT: pshufd {{.*#+}} xmm11 = xmm10[0,0,2,2]
-; SSSE3-NEXT: pcmpeqd %xmm0, %xmm9
-; SSSE3-NEXT: pshufd {{.*#+}} xmm9 = xmm9[1,1,3,3]
-; SSSE3-NEXT: pand %xmm11, %xmm9
-; SSSE3-NEXT: pshufd {{.*#+}} xmm0 = xmm10[1,1,3,3]
-; SSSE3-NEXT: por %xmm4, %xmm0
-; SSSE3-NEXT: por %xmm9, %xmm0
-; SSSE3-NEXT: paddq %xmm1, %xmm5
-; SSSE3-NEXT: pxor %xmm8, %xmm1
-; SSSE3-NEXT: movdqa %xmm5, %xmm4
-; SSSE3-NEXT: pxor %xmm8, %xmm4
-; SSSE3-NEXT: movdqa %xmm1, %xmm9
-; SSSE3-NEXT: pcmpgtd %xmm4, %xmm9
-; SSSE3-NEXT: pshufd {{.*#+}} xmm10 = xmm9[0,0,2,2]
-; SSSE3-NEXT: pcmpeqd %xmm1, %xmm4
-; SSSE3-NEXT: pshufd {{.*#+}} xmm4 = xmm4[1,1,3,3]
-; SSSE3-NEXT: pand %xmm10, %xmm4
-; SSSE3-NEXT: pshufd {{.*#+}} xmm1 = xmm9[1,1,3,3]
-; SSSE3-NEXT: por %xmm5, %xmm1
-; SSSE3-NEXT: por %xmm4, %xmm1
-; SSSE3-NEXT: paddq %xmm2, %xmm6
-; SSSE3-NEXT: pxor %xmm8, %xmm2
-; SSSE3-NEXT: movdqa %xmm6, %xmm4
-; SSSE3-NEXT: pxor %xmm8, %xmm4
-; SSSE3-NEXT: movdqa %xmm2, %xmm5
-; SSSE3-NEXT: pcmpgtd %xmm4, %xmm5
-; SSSE3-NEXT: pshufd {{.*#+}} xmm9 = xmm5[0,0,2,2]
-; SSSE3-NEXT: pcmpeqd %xmm2, %xmm4
-; SSSE3-NEXT: pshufd {{.*#+}} xmm4 = xmm4[1,1,3,3]
-; SSSE3-NEXT: pand %xmm9, %xmm4
-; SSSE3-NEXT: pshufd {{.*#+}} xmm2 = xmm5[1,1,3,3]
-; SSSE3-NEXT: por %xmm6, %xmm2
-; SSSE3-NEXT: por %xmm4, %xmm2
-; SSSE3-NEXT: paddq %xmm3, %xmm7
-; SSSE3-NEXT: pxor %xmm8, %xmm3
-; SSSE3-NEXT: pxor %xmm7, %xmm8
-; SSSE3-NEXT: movdqa %xmm3, %xmm4
-; SSSE3-NEXT: pcmpgtd %xmm8, %xmm4
-; SSSE3-NEXT: pshufd {{.*#+}} xmm5 = xmm4[0,0,2,2]
-; SSSE3-NEXT: pcmpeqd %xmm3, %xmm8
-; SSSE3-NEXT: pshufd {{.*#+}} xmm6 = xmm8[1,1,3,3]
-; SSSE3-NEXT: pand %xmm5, %xmm6
-; SSSE3-NEXT: pshufd {{.*#+}} xmm3 = xmm4[1,1,3,3]
-; SSSE3-NEXT: por %xmm7, %xmm3
-; SSSE3-NEXT: por %xmm6, %xmm3
-; SSSE3-NEXT: retq
-;
-; SSE41-LABEL: v8i64:
-; SSE41: # %bb.0:
-; SSE41-NEXT: movdqa %xmm7, %xmm8
-; SSE41-NEXT: movdqa %xmm6, %xmm10
-; SSE41-NEXT: movdqa %xmm5, %xmm11
-; SSE41-NEXT: movdqa %xmm4, %xmm5
-; SSE41-NEXT: movdqa {{.*#+}} xmm7 = [9223372039002259456,9223372039002259456]
-; SSE41-NEXT: paddq %xmm0, %xmm5
-; SSE41-NEXT: movdqa %xmm0, %xmm6
-; SSE41-NEXT: pxor %xmm7, %xmm6
-; SSE41-NEXT: movdqa %xmm5, %xmm4
-; SSE41-NEXT: pxor %xmm7, %xmm4
-; SSE41-NEXT: movdqa %xmm6, %xmm0
-; SSE41-NEXT: pcmpeqd %xmm4, %xmm0
-; SSE41-NEXT: pcmpgtd %xmm4, %xmm6
-; SSE41-NEXT: pshufd {{.*#+}} xmm4 = xmm6[0,0,2,2]
-; SSE41-NEXT: pand %xmm4, %xmm0
-; SSE41-NEXT: por %xmm6, %xmm0
-; SSE41-NEXT: pcmpeqd %xmm9, %xmm9
-; SSE41-NEXT: blendvpd %xmm0, %xmm9, %xmm5
-; SSE41-NEXT: paddq %xmm1, %xmm11
-; SSE41-NEXT: pxor %xmm7, %xmm1
-; SSE41-NEXT: movdqa %xmm11, %xmm0
-; SSE41-NEXT: pxor %xmm7, %xmm0
-; SSE41-NEXT: movdqa %xmm1, %xmm4
-; SSE41-NEXT: pcmpeqd %xmm0, %xmm4
-; SSE41-NEXT: pcmpgtd %xmm0, %xmm1
-; SSE41-NEXT: pshufd {{.*#+}} xmm0 = xmm1[0,0,2,2]
-; SSE41-NEXT: pand %xmm4, %xmm0
-; SSE41-NEXT: por %xmm1, %xmm0
-; SSE41-NEXT: blendvpd %xmm0, %xmm9, %xmm11
-; SSE41-NEXT: paddq %xmm2, %xmm10
-; SSE41-NEXT: pxor %xmm7, %xmm2
-; SSE41-NEXT: movdqa %xmm10, %xmm0
-; SSE41-NEXT: pxor %xmm7, %xmm0
-; SSE41-NEXT: movdqa %xmm2, %xmm1
-; SSE41-NEXT: pcmpeqd %xmm0, %xmm1
-; SSE41-NEXT: pcmpgtd %xmm0, %xmm2
-; SSE41-NEXT: pshufd {{.*#+}} xmm0 = xmm2[0,0,2,2]
-; SSE41-NEXT: pand %xmm1, %xmm0
-; SSE41-NEXT: por %xmm2, %xmm0
-; SSE41-NEXT: blendvpd %xmm0, %xmm9, %xmm10
-; SSE41-NEXT: paddq %xmm3, %xmm8
-; SSE41-NEXT: pxor %xmm7, %xmm3
-; SSE41-NEXT: pxor %xmm8, %xmm7
-; SSE41-NEXT: movdqa %xmm3, %xmm1
-; SSE41-NEXT: pcmpeqd %xmm7, %xmm1
-; SSE41-NEXT: pcmpgtd %xmm7, %xmm3
-; SSE41-NEXT: pshufd {{.*#+}} xmm0 = xmm3[0,0,2,2]
-; SSE41-NEXT: pand %xmm1, %xmm0
-; SSE41-NEXT: por %xmm3, %xmm0
-; SSE41-NEXT: blendvpd %xmm0, %xmm9, %xmm8
-; SSE41-NEXT: movapd %xmm5, %xmm0
-; SSE41-NEXT: movapd %xmm11, %xmm1
-; SSE41-NEXT: movapd %xmm10, %xmm2
-; SSE41-NEXT: movapd %xmm8, %xmm3
-; SSE41-NEXT: retq
+; SSE-LABEL: v8i64:
+; SSE: # %bb.0:
+; SSE-NEXT: movdqa {{.*#+}} xmm8 = [9223372039002259456,9223372039002259456]
+; SSE-NEXT: paddq %xmm0, %xmm4
+; SSE-NEXT: pxor %xmm8, %xmm0
+; SSE-NEXT: movdqa %xmm4, %xmm9
+; SSE-NEXT: pxor %xmm8, %xmm9
+; SSE-NEXT: movdqa %xmm0, %xmm10
+; SSE-NEXT: pcmpgtd %xmm9, %xmm10
+; SSE-NEXT: pshufd {{.*#+}} xmm11 = xmm10[0,0,2,2]
+; SSE-NEXT: pcmpeqd %xmm0, %xmm9
+; SSE-NEXT: pshufd {{.*#+}} xmm9 = xmm9[1,1,3,3]
+; SSE-NEXT: pand %xmm11, %xmm9
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm10[1,1,3,3]
+; SSE-NEXT: por %xmm4, %xmm0
+; SSE-NEXT: por %xmm9, %xmm0
+; SSE-NEXT: paddq %xmm1, %xmm5
+; SSE-NEXT: pxor %xmm8, %xmm1
+; SSE-NEXT: movdqa %xmm5, %xmm4
+; SSE-NEXT: pxor %xmm8, %xmm4
+; SSE-NEXT: movdqa %xmm1, %xmm9
+; SSE-NEXT: pcmpgtd %xmm4, %xmm9
+; SSE-NEXT: pshufd {{.*#+}} xmm10 = xmm9[0,0,2,2]
+; SSE-NEXT: pcmpeqd %xmm1, %xmm4
+; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm4[1,1,3,3]
+; SSE-NEXT: pand %xmm10, %xmm4
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm9[1,1,3,3]
+; SSE-NEXT: por %xmm5, %xmm1
+; SSE-NEXT: por %xmm4, %xmm1
+; SSE-NEXT: paddq %xmm2, %xmm6
+; SSE-NEXT: pxor %xmm8, %xmm2
+; SSE-NEXT: movdqa %xmm6, %xmm4
+; SSE-NEXT: pxor %xmm8, %xmm4
+; SSE-NEXT: movdqa %xmm2, %xmm5
+; SSE-NEXT: pcmpgtd %xmm4, %xmm5
+; SSE-NEXT: pshufd {{.*#+}} xmm9 = xmm5[0,0,2,2]
+; SSE-NEXT: pcmpeqd %xmm2, %xmm4
+; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm4[1,1,3,3]
+; SSE-NEXT: pand %xmm9, %xmm4
+; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm5[1,1,3,3]
+; SSE-NEXT: por %xmm6, %xmm2
+; SSE-NEXT: por %xmm4, %xmm2
+; SSE-NEXT: paddq %xmm3, %xmm7
+; SSE-NEXT: pxor %xmm8, %xmm3
+; SSE-NEXT: pxor %xmm7, %xmm8
+; SSE-NEXT: movdqa %xmm3, %xmm4
+; SSE-NEXT: pcmpgtd %xmm8, %xmm4
+; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm4[0,0,2,2]
+; SSE-NEXT: pcmpeqd %xmm3, %xmm8
+; SSE-NEXT: pshufd {{.*#+}} xmm6 = xmm8[1,1,3,3]
+; SSE-NEXT: pand %xmm5, %xmm6
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm4[1,1,3,3]
+; SSE-NEXT: por %xmm7, %xmm3
+; SSE-NEXT: por %xmm6, %xmm3
+; SSE-NEXT: retq
;
; AVX1-LABEL: v8i64:
; AVX1: # %bb.0:
diff --git a/llvm/test/CodeGen/X86/usub_sat_vec.ll b/llvm/test/CodeGen/X86/usub_sat_vec.ll
index 9d3a519d6b67..5a8c889165d6 100644
--- a/llvm/test/CodeGen/X86/usub_sat_vec.ll
+++ b/llvm/test/CodeGen/X86/usub_sat_vec.ll
@@ -823,60 +823,23 @@ define <16 x i32> @v16i32(<16 x i32> %x, <16 x i32> %y) nounwind {
}
define <2 x i64> @v2i64(<2 x i64> %x, <2 x i64> %y) nounwind {
-; SSE2-LABEL: v2i64:
-; SSE2: # %bb.0:
-; SSE2-NEXT: movdqa {{.*#+}} xmm2 = [9223372039002259456,9223372039002259456]
-; SSE2-NEXT: movdqa %xmm1, %xmm3
-; SSE2-NEXT: pxor %xmm2, %xmm3
-; SSE2-NEXT: pxor %xmm0, %xmm2
-; SSE2-NEXT: movdqa %xmm2, %xmm4
-; SSE2-NEXT: pcmpgtd %xmm3, %xmm4
-; SSE2-NEXT: pshufd {{.*#+}} xmm5 = xmm4[0,0,2,2]
-; SSE2-NEXT: pcmpeqd %xmm3, %xmm2
-; SSE2-NEXT: pshufd {{.*#+}} xmm2 = xmm2[1,1,3,3]
-; SSE2-NEXT: pand %xmm5, %xmm2
-; SSE2-NEXT: pshufd {{.*#+}} xmm3 = xmm4[1,1,3,3]
-; SSE2-NEXT: por %xmm2, %xmm3
-; SSE2-NEXT: psubq %xmm1, %xmm0
-; SSE2-NEXT: pand %xmm3, %xmm0
-; SSE2-NEXT: retq
-;
-; SSSE3-LABEL: v2i64:
-; SSSE3: # %bb.0:
-; SSSE3-NEXT: movdqa {{.*#+}} xmm2 = [9223372039002259456,9223372039002259456]
-; SSSE3-NEXT: movdqa %xmm1, %xmm3
-; SSSE3-NEXT: pxor %xmm2, %xmm3
-; SSSE3-NEXT: pxor %xmm0, %xmm2
-; SSSE3-NEXT: movdqa %xmm2, %xmm4
-; SSSE3-NEXT: pcmpgtd %xmm3, %xmm4
-; SSSE3-NEXT: pshufd {{.*#+}} xmm5 = xmm4[0,0,2,2]
-; SSSE3-NEXT: pcmpeqd %xmm3, %xmm2
-; SSSE3-NEXT: pshufd {{.*#+}} xmm2 = xmm2[1,1,3,3]
-; SSSE3-NEXT: pand %xmm5, %xmm2
-; SSSE3-NEXT: pshufd {{.*#+}} xmm3 = xmm4[1,1,3,3]
-; SSSE3-NEXT: por %xmm2, %xmm3
-; SSSE3-NEXT: psubq %xmm1, %xmm0
-; SSSE3-NEXT: pand %xmm3, %xmm0
-; SSSE3-NEXT: retq
-;
-; SSE41-LABEL: v2i64:
-; SSE41: # %bb.0:
-; SSE41-NEXT: movdqa %xmm0, %xmm2
-; SSE41-NEXT: movdqa {{.*#+}} xmm3 = [9223372039002259456,9223372039002259456]
-; SSE41-NEXT: movdqa %xmm1, %xmm0
-; SSE41-NEXT: pxor %xmm3, %xmm0
-; SSE41-NEXT: pxor %xmm2, %xmm3
-; SSE41-NEXT: movdqa %xmm3, %xmm4
-; SSE41-NEXT: pcmpeqd %xmm0, %xmm4
-; SSE41-NEXT: pcmpgtd %xmm0, %xmm3
-; SSE41-NEXT: pshufd {{.*#+}} xmm0 = xmm3[0,0,2,2]
-; SSE41-NEXT: pand %xmm4, %xmm0
-; SSE41-NEXT: por %xmm3, %xmm0
-; SSE41-NEXT: psubq %xmm1, %xmm2
-; SSE41-NEXT: pxor %xmm1, %xmm1
-; SSE41-NEXT: blendvpd %xmm0, %xmm2, %xmm1
-; SSE41-NEXT: movapd %xmm1, %xmm0
-; SSE41-NEXT: retq
+; SSE-LABEL: v2i64:
+; SSE: # %bb.0:
+; SSE-NEXT: movdqa {{.*#+}} xmm2 = [9223372039002259456,9223372039002259456]
+; SSE-NEXT: movdqa %xmm1, %xmm3
+; SSE-NEXT: pxor %xmm2, %xmm3
+; SSE-NEXT: pxor %xmm0, %xmm2
+; SSE-NEXT: movdqa %xmm2, %xmm4
+; SSE-NEXT: pcmpgtd %xmm3, %xmm4
+; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm4[0,0,2,2]
+; SSE-NEXT: pcmpeqd %xmm3, %xmm2
+; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm2[1,1,3,3]
+; SSE-NEXT: pand %xmm5, %xmm2
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm4[1,1,3,3]
+; SSE-NEXT: por %xmm2, %xmm3
+; SSE-NEXT: psubq %xmm1, %xmm0
+; SSE-NEXT: pand %xmm3, %xmm0
+; SSE-NEXT: retq
;
; AVX1-LABEL: v2i64:
; AVX1: # %bb.0:
@@ -917,99 +880,36 @@ define <2 x i64> @v2i64(<2 x i64> %x, <2 x i64> %y) nounwind {
}
define <4 x i64> @v4i64(<4 x i64> %x, <4 x i64> %y) nounwind {
-; SSE2-LABEL: v4i64:
-; SSE2: # %bb.0:
-; SSE2-NEXT: movdqa {{.*#+}} xmm4 = [9223372039002259456,9223372039002259456]
-; SSE2-NEXT: movdqa %xmm0, %xmm5
-; SSE2-NEXT: psubq %xmm2, %xmm0
-; SSE2-NEXT: pxor %xmm4, %xmm2
-; SSE2-NEXT: pxor %xmm4, %xmm5
-; SSE2-NEXT: movdqa %xmm5, %xmm6
-; SSE2-NEXT: pcmpgtd %xmm2, %xmm6
-; SSE2-NEXT: pshufd {{.*#+}} xmm7 = xmm6[0,0,2,2]
-; SSE2-NEXT: pcmpeqd %xmm2, %xmm5
-; SSE2-NEXT: pshufd {{.*#+}} xmm2 = xmm5[1,1,3,3]
-; SSE2-NEXT: pand %xmm7, %xmm2
-; SSE2-NEXT: pshufd {{.*#+}} xmm5 = xmm6[1,1,3,3]
-; SSE2-NEXT: por %xmm2, %xmm5
-; SSE2-NEXT: pand %xmm5, %xmm0
-; SSE2-NEXT: movdqa %xmm3, %xmm2
-; SSE2-NEXT: pxor %xmm4, %xmm2
-; SSE2-NEXT: pxor %xmm1, %xmm4
-; SSE2-NEXT: movdqa %xmm4, %xmm5
-; SSE2-NEXT: pcmpgtd %xmm2, %xmm5
-; SSE2-NEXT: pshufd {{.*#+}} xmm6 = xmm5[0,0,2,2]
-; SSE2-NEXT: pcmpeqd %xmm2, %xmm4
-; SSE2-NEXT: pshufd {{.*#+}} xmm2 = xmm4[1,1,3,3]
-; SSE2-NEXT: pand %xmm6, %xmm2
-; SSE2-NEXT: pshufd {{.*#+}} xmm4 = xmm5[1,1,3,3]
-; SSE2-NEXT: por %xmm2, %xmm4
-; SSE2-NEXT: psubq %xmm3, %xmm1
-; SSE2-NEXT: pand %xmm4, %xmm1
-; SSE2-NEXT: retq
-;
-; SSSE3-LABEL: v4i64:
-; SSSE3: # %bb.0:
-; SSSE3-NEXT: movdqa {{.*#+}} xmm4 = [9223372039002259456,9223372039002259456]
-; SSSE3-NEXT: movdqa %xmm0, %xmm5
-; SSSE3-NEXT: psubq %xmm2, %xmm0
-; SSSE3-NEXT: pxor %xmm4, %xmm2
-; SSSE3-NEXT: pxor %xmm4, %xmm5
-; SSSE3-NEXT: movdqa %xmm5, %xmm6
-; SSSE3-NEXT: pcmpgtd %xmm2, %xmm6
-; SSSE3-NEXT: pshufd {{.*#+}} xmm7 = xmm6[0,0,2,2]
-; SSSE3-NEXT: pcmpeqd %xmm2, %xmm5
-; SSSE3-NEXT: pshufd {{.*#+}} xmm2 = xmm5[1,1,3,3]
-; SSSE3-NEXT: pand %xmm7, %xmm2
-; SSSE3-NEXT: pshufd {{.*#+}} xmm5 = xmm6[1,1,3,3]
-; SSSE3-NEXT: por %xmm2, %xmm5
-; SSSE3-NEXT: pand %xmm5, %xmm0
-; SSSE3-NEXT: movdqa %xmm3, %xmm2
-; SSSE3-NEXT: pxor %xmm4, %xmm2
-; SSSE3-NEXT: pxor %xmm1, %xmm4
-; SSSE3-NEXT: movdqa %xmm4, %xmm5
-; SSSE3-NEXT: pcmpgtd %xmm2, %xmm5
-; SSSE3-NEXT: pshufd {{.*#+}} xmm6 = xmm5[0,0,2,2]
-; SSSE3-NEXT: pcmpeqd %xmm2, %xmm4
-; SSSE3-NEXT: pshufd {{.*#+}} xmm2 = xmm4[1,1,3,3]
-; SSSE3-NEXT: pand %xmm6, %xmm2
-; SSSE3-NEXT: pshufd {{.*#+}} xmm4 = xmm5[1,1,3,3]
-; SSSE3-NEXT: por %xmm2, %xmm4
-; SSSE3-NEXT: psubq %xmm3, %xmm1
-; SSSE3-NEXT: pand %xmm4, %xmm1
-; SSSE3-NEXT: retq
-;
-; SSE41-LABEL: v4i64:
-; SSE41: # %bb.0:
-; SSE41-NEXT: movdqa %xmm0, %xmm4
-; SSE41-NEXT: movdqa {{.*#+}} xmm6 = [9223372039002259456,9223372039002259456]
-; SSE41-NEXT: movdqa %xmm0, %xmm5
-; SSE41-NEXT: psubq %xmm2, %xmm4
-; SSE41-NEXT: pxor %xmm6, %xmm2
-; SSE41-NEXT: pxor %xmm6, %xmm5
-; SSE41-NEXT: movdqa %xmm5, %xmm0
-; SSE41-NEXT: pcmpeqd %xmm2, %xmm0
-; SSE41-NEXT: pcmpgtd %xmm2, %xmm5
-; SSE41-NEXT: pshufd {{.*#+}} xmm2 = xmm5[0,0,2,2]
-; SSE41-NEXT: pand %xmm2, %xmm0
-; SSE41-NEXT: por %xmm5, %xmm0
-; SSE41-NEXT: pxor %xmm2, %xmm2
-; SSE41-NEXT: pxor %xmm5, %xmm5
-; SSE41-NEXT: blendvpd %xmm0, %xmm4, %xmm5
-; SSE41-NEXT: movdqa %xmm3, %xmm0
-; SSE41-NEXT: pxor %xmm6, %xmm0
-; SSE41-NEXT: pxor %xmm1, %xmm6
-; SSE41-NEXT: movdqa %xmm6, %xmm4
-; SSE41-NEXT: pcmpeqd %xmm0, %xmm4
-; SSE41-NEXT: pcmpgtd %xmm0, %xmm6
-; SSE41-NEXT: pshufd {{.*#+}} xmm0 = xmm6[0,0,2,2]
-; SSE41-NEXT: pand %xmm4, %xmm0
-; SSE41-NEXT: por %xmm6, %xmm0
-; SSE41-NEXT: psubq %xmm3, %xmm1
-; SSE41-NEXT: blendvpd %xmm0, %xmm1, %xmm2
-; SSE41-NEXT: movapd %xmm5, %xmm0
-; SSE41-NEXT: movapd %xmm2, %xmm1
-; SSE41-NEXT: retq
+; SSE-LABEL: v4i64:
+; SSE: # %bb.0:
+; SSE-NEXT: movdqa {{.*#+}} xmm4 = [9223372039002259456,9223372039002259456]
+; SSE-NEXT: movdqa %xmm0, %xmm5
+; SSE-NEXT: psubq %xmm2, %xmm0
+; SSE-NEXT: pxor %xmm4, %xmm2
+; SSE-NEXT: pxor %xmm4, %xmm5
+; SSE-NEXT: movdqa %xmm5, %xmm6
+; SSE-NEXT: pcmpgtd %xmm2, %xmm6
+; SSE-NEXT: pshufd {{.*#+}} xmm7 = xmm6[0,0,2,2]
+; SSE-NEXT: pcmpeqd %xmm2, %xmm5
+; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm5[1,1,3,3]
+; SSE-NEXT: pand %xmm7, %xmm2
+; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm6[1,1,3,3]
+; SSE-NEXT: por %xmm2, %xmm5
+; SSE-NEXT: pand %xmm5, %xmm0
+; SSE-NEXT: movdqa %xmm3, %xmm2
+; SSE-NEXT: pxor %xmm4, %xmm2
+; SSE-NEXT: pxor %xmm1, %xmm4
+; SSE-NEXT: movdqa %xmm4, %xmm5
+; SSE-NEXT: pcmpgtd %xmm2, %xmm5
+; SSE-NEXT: pshufd {{.*#+}} xmm6 = xmm5[0,0,2,2]
+; SSE-NEXT: pcmpeqd %xmm2, %xmm4
+; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm4[1,1,3,3]
+; SSE-NEXT: pand %xmm6, %xmm2
+; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm5[1,1,3,3]
+; SSE-NEXT: por %xmm2, %xmm4
+; SSE-NEXT: psubq %xmm3, %xmm1
+; SSE-NEXT: pand %xmm4, %xmm1
+; SSE-NEXT: retq
;
; AVX1-LABEL: v4i64:
; AVX1: # %bb.0:
@@ -1057,177 +957,62 @@ define <4 x i64> @v4i64(<4 x i64> %x, <4 x i64> %y) nounwind {
}
define <8 x i64> @v8i64(<8 x i64> %x, <8 x i64> %y) nounwind {
-; SSE2-LABEL: v8i64:
-; SSE2: # %bb.0:
-; SSE2-NEXT: movdqa {{.*#+}} xmm8 = [9223372039002259456,9223372039002259456]
-; SSE2-NEXT: movdqa %xmm0, %xmm9
-; SSE2-NEXT: psubq %xmm4, %xmm0
-; SSE2-NEXT: pxor %xmm8, %xmm4
-; SSE2-NEXT: pxor %xmm8, %xmm9
-; SSE2-NEXT: movdqa %xmm9, %xmm10
-; SSE2-NEXT: pcmpgtd %xmm4, %xmm10
-; SSE2-NEXT: pshufd {{.*#+}} xmm11 = xmm10[0,0,2,2]
-; SSE2-NEXT: pcmpeqd %xmm4, %xmm9
-; SSE2-NEXT: pshufd {{.*#+}} xmm9 = xmm9[1,1,3,3]
-; SSE2-NEXT: pand %xmm11, %xmm9
-; SSE2-NEXT: pshufd {{.*#+}} xmm4 = xmm10[1,1,3,3]
-; SSE2-NEXT: por %xmm9, %xmm4
-; SSE2-NEXT: pand %xmm4, %xmm0
-; SSE2-NEXT: movdqa %xmm1, %xmm9
-; SSE2-NEXT: psubq %xmm5, %xmm1
-; SSE2-NEXT: pxor %xmm8, %xmm5
-; SSE2-NEXT: pxor %xmm8, %xmm9
-; SSE2-NEXT: movdqa %xmm9, %xmm4
-; SSE2-NEXT: pcmpgtd %xmm5, %xmm4
-; SSE2-NEXT: pshufd {{.*#+}} xmm10 = xmm4[0,0,2,2]
-; SSE2-NEXT: pcmpeqd %xmm5, %xmm9
-; SSE2-NEXT: pshufd {{.*#+}} xmm5 = xmm9[1,1,3,3]
-; SSE2-NEXT: pand %xmm10, %xmm5
-; SSE2-NEXT: pshufd {{.*#+}} xmm4 = xmm4[1,1,3,3]
-; SSE2-NEXT: por %xmm5, %xmm4
-; SSE2-NEXT: pand %xmm4, %xmm1
-; SSE2-NEXT: movdqa %xmm2, %xmm4
-; SSE2-NEXT: psubq %xmm6, %xmm2
-; SSE2-NEXT: pxor %xmm8, %xmm6
-; SSE2-NEXT: pxor %xmm8, %xmm4
-; SSE2-NEXT: movdqa %xmm4, %xmm5
-; SSE2-NEXT: pcmpgtd %xmm6, %xmm5
-; SSE2-NEXT: pshufd {{.*#+}} xmm9 = xmm5[0,0,2,2]
-; SSE2-NEXT: pcmpeqd %xmm6, %xmm4
-; SSE2-NEXT: pshufd {{.*#+}} xmm4 = xmm4[1,1,3,3]
-; SSE2-NEXT: pand %xmm9, %xmm4
-; SSE2-NEXT: pshufd {{.*#+}} xmm5 = xmm5[1,1,3,3]
-; SSE2-NEXT: por %xmm4, %xmm5
-; SSE2-NEXT: pand %xmm5, %xmm2
-; SSE2-NEXT: movdqa %xmm7, %xmm4
-; SSE2-NEXT: pxor %xmm8, %xmm4
-; SSE2-NEXT: pxor %xmm3, %xmm8
-; SSE2-NEXT: movdqa %xmm8, %xmm5
-; SSE2-NEXT: pcmpgtd %xmm4, %xmm5
-; SSE2-NEXT: pshufd {{.*#+}} xmm6 = xmm5[0,0,2,2]
-; SSE2-NEXT: pcmpeqd %xmm4, %xmm8
-; SSE2-NEXT: pshufd {{.*#+}} xmm4 = xmm8[1,1,3,3]
-; SSE2-NEXT: pand %xmm6, %xmm4
-; SSE2-NEXT: pshufd {{.*#+}} xmm5 = xmm5[1,1,3,3]
-; SSE2-NEXT: por %xmm4, %xmm5
-; SSE2-NEXT: psubq %xmm7, %xmm3
-; SSE2-NEXT: pand %xmm5, %xmm3
-; SSE2-NEXT: retq
-;
-; SSSE3-LABEL: v8i64:
-; SSSE3: # %bb.0:
-; SSSE3-NEXT: movdqa {{.*#+}} xmm8 = [9223372039002259456,9223372039002259456]
-; SSSE3-NEXT: movdqa %xmm0, %xmm9
-; SSSE3-NEXT: psubq %xmm4, %xmm0
-; SSSE3-NEXT: pxor %xmm8, %xmm4
-; SSSE3-NEXT: pxor %xmm8, %xmm9
-; SSSE3-NEXT: movdqa %xmm9, %xmm10
-; SSSE3-NEXT: pcmpgtd %xmm4, %xmm10
-; SSSE3-NEXT: pshufd {{.*#+}} xmm11 = xmm10[0,0,2,2]
-; SSSE3-NEXT: pcmpeqd %xmm4, %xmm9
-; SSSE3-NEXT: pshufd {{.*#+}} xmm9 = xmm9[1,1,3,3]
-; SSSE3-NEXT: pand %xmm11, %xmm9
-; SSSE3-NEXT: pshufd {{.*#+}} xmm4 = xmm10[1,1,3,3]
-; SSSE3-NEXT: por %xmm9, %xmm4
-; SSSE3-NEXT: pand %xmm4, %xmm0
-; SSSE3-NEXT: movdqa %xmm1, %xmm9
-; SSSE3-NEXT: psubq %xmm5, %xmm1
-; SSSE3-NEXT: pxor %xmm8, %xmm5
-; SSSE3-NEXT: pxor %xmm8, %xmm9
-; SSSE3-NEXT: movdqa %xmm9, %xmm4
-; SSSE3-NEXT: pcmpgtd %xmm5, %xmm4
-; SSSE3-NEXT: pshufd {{.*#+}} xmm10 = xmm4[0,0,2,2]
-; SSSE3-NEXT: pcmpeqd %xmm5, %xmm9
-; SSSE3-NEXT: pshufd {{.*#+}} xmm5 = xmm9[1,1,3,3]
-; SSSE3-NEXT: pand %xmm10, %xmm5
-; SSSE3-NEXT: pshufd {{.*#+}} xmm4 = xmm4[1,1,3,3]
-; SSSE3-NEXT: por %xmm5, %xmm4
-; SSSE3-NEXT: pand %xmm4, %xmm1
-; SSSE3-NEXT: movdqa %xmm2, %xmm4
-; SSSE3-NEXT: psubq %xmm6, %xmm2
-; SSSE3-NEXT: pxor %xmm8, %xmm6
-; SSSE3-NEXT: pxor %xmm8, %xmm4
-; SSSE3-NEXT: movdqa %xmm4, %xmm5
-; SSSE3-NEXT: pcmpgtd %xmm6, %xmm5
-; SSSE3-NEXT: pshufd {{.*#+}} xmm9 = xmm5[0,0,2,2]
-; SSSE3-NEXT: pcmpeqd %xmm6, %xmm4
-; SSSE3-NEXT: pshufd {{.*#+}} xmm4 = xmm4[1,1,3,3]
-; SSSE3-NEXT: pand %xmm9, %xmm4
-; SSSE3-NEXT: pshufd {{.*#+}} xmm5 = xmm5[1,1,3,3]
-; SSSE3-NEXT: por %xmm4, %xmm5
-; SSSE3-NEXT: pand %xmm5, %xmm2
-; SSSE3-NEXT: movdqa %xmm7, %xmm4
-; SSSE3-NEXT: pxor %xmm8, %xmm4
-; SSSE3-NEXT: pxor %xmm3, %xmm8
-; SSSE3-NEXT: movdqa %xmm8, %xmm5
-; SSSE3-NEXT: pcmpgtd %xmm4, %xmm5
-; SSSE3-NEXT: pshufd {{.*#+}} xmm6 = xmm5[0,0,2,2]
-; SSSE3-NEXT: pcmpeqd %xmm4, %xmm8
-; SSSE3-NEXT: pshufd {{.*#+}} xmm4 = xmm8[1,1,3,3]
-; SSSE3-NEXT: pand %xmm6, %xmm4
-; SSSE3-NEXT: pshufd {{.*#+}} xmm5 = xmm5[1,1,3,3]
-; SSSE3-NEXT: por %xmm4, %xmm5
-; SSSE3-NEXT: psubq %xmm7, %xmm3
-; SSSE3-NEXT: pand %xmm5, %xmm3
-; SSSE3-NEXT: retq
-;
-; SSE41-LABEL: v8i64:
-; SSE41: # %bb.0:
-; SSE41-NEXT: movdqa %xmm0, %xmm9
-; SSE41-NEXT: movdqa {{.*#+}} xmm10 = [9223372039002259456,9223372039002259456]
-; SSE41-NEXT: movdqa %xmm0, %xmm8
-; SSE41-NEXT: psubq %xmm4, %xmm9
-; SSE41-NEXT: pxor %xmm10, %xmm4
-; SSE41-NEXT: pxor %xmm10, %xmm8
-; SSE41-NEXT: movdqa %xmm8, %xmm0
-; SSE41-NEXT: pcmpeqd %xmm4, %xmm0
-; SSE41-NEXT: pcmpgtd %xmm4, %xmm8
-; SSE41-NEXT: pshufd {{.*#+}} xmm4 = xmm8[0,0,2,2]
-; SSE41-NEXT: pand %xmm4, %xmm0
-; SSE41-NEXT: por %xmm8, %xmm0
-; SSE41-NEXT: pxor %xmm8, %xmm8
-; SSE41-NEXT: xorpd %xmm11, %xmm11
-; SSE41-NEXT: blendvpd %xmm0, %xmm9, %xmm11
-; SSE41-NEXT: movdqa %xmm1, %xmm4
-; SSE41-NEXT: psubq %xmm5, %xmm1
-; SSE41-NEXT: pxor %xmm10, %xmm5
-; SSE41-NEXT: pxor %xmm10, %xmm4
-; SSE41-NEXT: movdqa %xmm4, %xmm9
-; SSE41-NEXT: pcmpeqd %xmm5, %xmm9
-; SSE41-NEXT: pcmpgtd %xmm5, %xmm4
-; SSE41-NEXT: pshufd {{.*#+}} xmm0 = xmm4[0,0,2,2]
-; SSE41-NEXT: pand %xmm9, %xmm0
-; SSE41-NEXT: por %xmm4, %xmm0
-; SSE41-NEXT: pxor %xmm5, %xmm5
-; SSE41-NEXT: blendvpd %xmm0, %xmm1, %xmm5
-; SSE41-NEXT: movdqa %xmm2, %xmm1
-; SSE41-NEXT: psubq %xmm6, %xmm2
-; SSE41-NEXT: pxor %xmm10, %xmm6
-; SSE41-NEXT: pxor %xmm10, %xmm1
-; SSE41-NEXT: movdqa %xmm1, %xmm4
-; SSE41-NEXT: pcmpeqd %xmm6, %xmm4
-; SSE41-NEXT: pcmpgtd %xmm6, %xmm1
-; SSE41-NEXT: pshufd {{.*#+}} xmm0 = xmm1[0,0,2,2]
-; SSE41-NEXT: pand %xmm4, %xmm0
-; SSE41-NEXT: por %xmm1, %xmm0
-; SSE41-NEXT: pxor %xmm6, %xmm6
-; SSE41-NEXT: blendvpd %xmm0, %xmm2, %xmm6
-; SSE41-NEXT: movdqa %xmm7, %xmm0
-; SSE41-NEXT: pxor %xmm10, %xmm0
-; SSE41-NEXT: pxor %xmm3, %xmm10
-; SSE41-NEXT: movdqa %xmm10, %xmm1
-; SSE41-NEXT: pcmpeqd %xmm0, %xmm1
-; SSE41-NEXT: pcmpgtd %xmm0, %xmm10
-; SSE41-NEXT: pshufd {{.*#+}} xmm0 = xmm10[0,0,2,2]
-; SSE41-NEXT: pand %xmm1, %xmm0
-; SSE41-NEXT: por %xmm10, %xmm0
-; SSE41-NEXT: psubq %xmm7, %xmm3
-; SSE41-NEXT: blendvpd %xmm0, %xmm3, %xmm8
-; SSE41-NEXT: movapd %xmm11, %xmm0
-; SSE41-NEXT: movapd %xmm5, %xmm1
-; SSE41-NEXT: movapd %xmm6, %xmm2
-; SSE41-NEXT: movapd %xmm8, %xmm3
-; SSE41-NEXT: retq
+; SSE-LABEL: v8i64:
+; SSE: # %bb.0:
+; SSE-NEXT: movdqa {{.*#+}} xmm8 = [9223372039002259456,9223372039002259456]
+; SSE-NEXT: movdqa %xmm0, %xmm9
+; SSE-NEXT: psubq %xmm4, %xmm0
+; SSE-NEXT: pxor %xmm8, %xmm4
+; SSE-NEXT: pxor %xmm8, %xmm9
+; SSE-NEXT: movdqa %xmm9, %xmm10
+; SSE-NEXT: pcmpgtd %xmm4, %xmm10
+; SSE-NEXT: pshufd {{.*#+}} xmm11 = xmm10[0,0,2,2]
+; SSE-NEXT: pcmpeqd %xmm4, %xmm9
+; SSE-NEXT: pshufd {{.*#+}} xmm9 = xmm9[1,1,3,3]
+; SSE-NEXT: pand %xmm11, %xmm9
+; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm10[1,1,3,3]
+; SSE-NEXT: por %xmm9, %xmm4
+; SSE-NEXT: pand %xmm4, %xmm0
+; SSE-NEXT: movdqa %xmm1, %xmm9
+; SSE-NEXT: psubq %xmm5, %xmm1
+; SSE-NEXT: pxor %xmm8, %xmm5
+; SSE-NEXT: pxor %xmm8, %xmm9
+; SSE-NEXT: movdqa %xmm9, %xmm4
+; SSE-NEXT: pcmpgtd %xmm5, %xmm4
+; SSE-NEXT: pshufd {{.*#+}} xmm10 = xmm4[0,0,2,2]
+; SSE-NEXT: pcmpeqd %xmm5, %xmm9
+; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm9[1,1,3,3]
+; SSE-NEXT: pand %xmm10, %xmm5
+; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm4[1,1,3,3]
+; SSE-NEXT: por %xmm5, %xmm4
+; SSE-NEXT: pand %xmm4, %xmm1
+; SSE-NEXT: movdqa %xmm2, %xmm4
+; SSE-NEXT: psubq %xmm6, %xmm2
+; SSE-NEXT: pxor %xmm8, %xmm6
+; SSE-NEXT: pxor %xmm8, %xmm4
+; SSE-NEXT: movdqa %xmm4, %xmm5
+; SSE-NEXT: pcmpgtd %xmm6, %xmm5
+; SSE-NEXT: pshufd {{.*#+}} xmm9 = xmm5[0,0,2,2]
+; SSE-NEXT: pcmpeqd %xmm6, %xmm4
+; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm4[1,1,3,3]
+; SSE-NEXT: pand %xmm9, %xmm4
+; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm5[1,1,3,3]
+; SSE-NEXT: por %xmm4, %xmm5
+; SSE-NEXT: pand %xmm5, %xmm2
+; SSE-NEXT: movdqa %xmm7, %xmm4
+; SSE-NEXT: pxor %xmm8, %xmm4
+; SSE-NEXT: pxor %xmm3, %xmm8
+; SSE-NEXT: movdqa %xmm8, %xmm5
+; SSE-NEXT: pcmpgtd %xmm4, %xmm5
+; SSE-NEXT: pshufd {{.*#+}} xmm6 = xmm5[0,0,2,2]
+; SSE-NEXT: pcmpeqd %xmm4, %xmm8
+; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm8[1,1,3,3]
+; SSE-NEXT: pand %xmm6, %xmm4
+; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm5[1,1,3,3]
+; SSE-NEXT: por %xmm4, %xmm5
+; SSE-NEXT: psubq %xmm7, %xmm3
+; SSE-NEXT: pand %xmm5, %xmm3
+; SSE-NEXT: retq
;
; AVX1-LABEL: v8i64:
; AVX1: # %bb.0:
More information about the llvm-commits
mailing list