[llvm] r337578 - [X86][XOP] Fix SUB constant folding for VPSHA/VPSHL shift lowering
Simon Pilgrim via llvm-commits
llvm-commits at lists.llvm.org
Fri Jul 20 09:55:18 PDT 2018
Author: rksimon
Date: Fri Jul 20 09:55:18 2018
New Revision: 337578
URL: http://llvm.org/viewvc/llvm-project?rev=337578&view=rev
Log:
[X86][XOP] Fix SUB constant folding for VPSHA/VPSHL shift lowering
We can safely use getConstant here as we're still lowering, which allows constant folding to kick in and simplify the vector shift codegen.
Noticed while working on D49562.
Modified:
llvm/trunk/lib/Target/X86/X86ISelLowering.cpp
llvm/trunk/test/CodeGen/X86/combine-sdiv.ll
llvm/trunk/test/CodeGen/X86/vector-shift-ashr-128.ll
llvm/trunk/test/CodeGen/X86/vector-shift-ashr-256.ll
llvm/trunk/test/CodeGen/X86/vector-shift-lshr-128.ll
llvm/trunk/test/CodeGen/X86/vector-shift-lshr-256.ll
Modified: llvm/trunk/lib/Target/X86/X86ISelLowering.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86ISelLowering.cpp?rev=337578&r1=337577&r2=337578&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/X86ISelLowering.cpp (original)
+++ llvm/trunk/lib/Target/X86/X86ISelLowering.cpp Fri Jul 20 09:55:18 2018
@@ -23416,11 +23416,10 @@ static SDValue LowerShift(SDValue Op, co
// XOP has 128-bit variable logical/arithmetic shifts.
// +ve/-ve Amt = shift left/right.
- if (Subtarget.hasXOP() &&
- (VT == MVT::v2i64 || VT == MVT::v4i32 ||
- VT == MVT::v8i16 || VT == MVT::v16i8)) {
+ if (Subtarget.hasXOP() && (VT == MVT::v2i64 || VT == MVT::v4i32 ||
+ VT == MVT::v8i16 || VT == MVT::v16i8)) {
if (Op.getOpcode() == ISD::SRL || Op.getOpcode() == ISD::SRA) {
- SDValue Zero = getZeroVector(VT, Subtarget, DAG, dl);
+ SDValue Zero = DAG.getConstant(0, dl, VT);
Amt = DAG.getNode(ISD::SUB, dl, VT, Zero, Amt);
}
if (Op.getOpcode() == ISD::SHL || Op.getOpcode() == ISD::SRL)
Modified: llvm/trunk/test/CodeGen/X86/combine-sdiv.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/combine-sdiv.ll?rev=337578&r1=337577&r2=337578&view=diff
==============================================================================
--- llvm/trunk/test/CodeGen/X86/combine-sdiv.ll (original)
+++ llvm/trunk/test/CodeGen/X86/combine-sdiv.ll Fri Jul 20 09:55:18 2018
@@ -465,12 +465,10 @@ define <16 x i8> @combine_vec_sdiv_by_po
; XOP-LABEL: combine_vec_sdiv_by_pow2b_v16i8:
; XOP: # %bb.0:
; XOP-NEXT: vpxor %xmm1, %xmm1, %xmm1
-; XOP-NEXT: vpcmpgtb %xmm0, %xmm1, %xmm2
-; XOP-NEXT: vpsubb {{.*}}(%rip), %xmm1, %xmm3
-; XOP-NEXT: vpshlb %xmm3, %xmm2, %xmm2
-; XOP-NEXT: vpaddb %xmm2, %xmm0, %xmm2
-; XOP-NEXT: vpsubb {{.*}}(%rip), %xmm1, %xmm1
-; XOP-NEXT: vpshab %xmm1, %xmm2, %xmm1
+; XOP-NEXT: vpcmpgtb %xmm0, %xmm1, %xmm1
+; XOP-NEXT: vpshlb {{.*}}(%rip), %xmm1, %xmm1
+; XOP-NEXT: vpaddb %xmm1, %xmm0, %xmm1
+; XOP-NEXT: vpshab {{.*}}(%rip), %xmm1, %xmm1
; XOP-NEXT: vmovdqa {{.*#+}} xmm2 = [0,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255]
; XOP-NEXT: vpblendvb %xmm2, %xmm1, %xmm0, %xmm0
; XOP-NEXT: retq
@@ -571,13 +569,10 @@ define <8 x i16> @combine_vec_sdiv_by_po
;
; XOP-LABEL: combine_vec_sdiv_by_pow2b_v8i16:
; XOP: # %bb.0:
-; XOP-NEXT: vpxor %xmm1, %xmm1, %xmm1
-; XOP-NEXT: vpsubw {{.*}}(%rip), %xmm1, %xmm2
-; XOP-NEXT: vpsraw $15, %xmm0, %xmm3
-; XOP-NEXT: vpshlw %xmm2, %xmm3, %xmm2
-; XOP-NEXT: vpaddw %xmm2, %xmm0, %xmm2
-; XOP-NEXT: vpsubw {{.*}}(%rip), %xmm1, %xmm1
-; XOP-NEXT: vpshaw %xmm1, %xmm2, %xmm1
+; XOP-NEXT: vpsraw $15, %xmm0, %xmm1
+; XOP-NEXT: vpshlw {{.*}}(%rip), %xmm1, %xmm1
+; XOP-NEXT: vpaddw %xmm1, %xmm0, %xmm1
+; XOP-NEXT: vpshaw {{.*}}(%rip), %xmm1, %xmm1
; XOP-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0],xmm1[1,2,3,4,5,6,7]
; XOP-NEXT: retq
%1 = sdiv <8 x i16> %x, <i16 1, i16 4, i16 2, i16 16, i16 8, i16 32, i16 64, i16 2>
@@ -734,19 +729,18 @@ define <16 x i16> @combine_vec_sdiv_by_p
;
; XOP-LABEL: combine_vec_sdiv_by_pow2b_v16i16:
; XOP: # %bb.0:
-; XOP-NEXT: vpxor %xmm1, %xmm1, %xmm1
-; XOP-NEXT: vpsubw {{.*}}(%rip), %xmm1, %xmm2
-; XOP-NEXT: vextractf128 $1, %ymm0, %xmm3
-; XOP-NEXT: vpsraw $15, %xmm3, %xmm4
-; XOP-NEXT: vpshlw %xmm2, %xmm4, %xmm4
-; XOP-NEXT: vpaddw %xmm4, %xmm3, %xmm3
-; XOP-NEXT: vpsubw {{.*}}(%rip), %xmm1, %xmm1
-; XOP-NEXT: vpshaw %xmm1, %xmm3, %xmm3
+; XOP-NEXT: vextractf128 $1, %ymm0, %xmm1
+; XOP-NEXT: vpsraw $15, %xmm1, %xmm2
+; XOP-NEXT: vmovdqa {{.*#+}} xmm3 = [65520,65522,65521,65524,65523,65525,65526,65521]
+; XOP-NEXT: vpshlw %xmm3, %xmm2, %xmm2
+; XOP-NEXT: vpaddw %xmm2, %xmm1, %xmm1
+; XOP-NEXT: vmovdqa {{.*#+}} xmm2 = [0,65534,65535,65532,65533,65531,65530,65535]
+; XOP-NEXT: vpshaw %xmm2, %xmm1, %xmm1
; XOP-NEXT: vpsraw $15, %xmm0, %xmm4
-; XOP-NEXT: vpshlw %xmm2, %xmm4, %xmm2
-; XOP-NEXT: vpaddw %xmm2, %xmm0, %xmm2
-; XOP-NEXT: vpshaw %xmm1, %xmm2, %xmm1
-; XOP-NEXT: vinsertf128 $1, %xmm3, %ymm1, %ymm1
+; XOP-NEXT: vpshlw %xmm3, %xmm4, %xmm3
+; XOP-NEXT: vpaddw %xmm3, %xmm0, %xmm3
+; XOP-NEXT: vpshaw %xmm2, %xmm3, %xmm2
+; XOP-NEXT: vinsertf128 $1, %xmm1, %ymm2, %ymm1
; XOP-NEXT: vpcmov {{.*}}(%rip), %ymm0, %ymm1, %ymm0
; XOP-NEXT: retq
%1 = sdiv <16 x i16> %x, <i16 1, i16 4, i16 2, i16 16, i16 8, i16 32, i16 64, i16 2, i16 1, i16 4, i16 2, i16 16, i16 8, i16 32, i16 64, i16 2>
@@ -1025,31 +1019,30 @@ define <32 x i16> @combine_vec_sdiv_by_p
;
; XOP-LABEL: combine_vec_sdiv_by_pow2b_v32i16:
; XOP: # %bb.0:
-; XOP-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; XOP-NEXT: vpsubw {{.*}}(%rip), %xmm2, %xmm3
-; XOP-NEXT: vextractf128 $1, %ymm0, %xmm4
-; XOP-NEXT: vpsraw $15, %xmm4, %xmm5
-; XOP-NEXT: vpshlw %xmm3, %xmm5, %xmm5
-; XOP-NEXT: vpaddw %xmm5, %xmm4, %xmm4
-; XOP-NEXT: vpsubw {{.*}}(%rip), %xmm2, %xmm2
-; XOP-NEXT: vpshaw %xmm2, %xmm4, %xmm4
+; XOP-NEXT: vextractf128 $1, %ymm0, %xmm2
+; XOP-NEXT: vpsraw $15, %xmm2, %xmm3
+; XOP-NEXT: vmovdqa {{.*#+}} xmm4 = [65520,65522,65521,65524,65523,65525,65526,65521]
+; XOP-NEXT: vpshlw %xmm4, %xmm3, %xmm3
+; XOP-NEXT: vpaddw %xmm3, %xmm2, %xmm2
+; XOP-NEXT: vmovdqa {{.*#+}} xmm3 = [0,65534,65535,65532,65533,65531,65530,65535]
+; XOP-NEXT: vpshaw %xmm3, %xmm2, %xmm2
; XOP-NEXT: vpsraw $15, %xmm0, %xmm5
-; XOP-NEXT: vpshlw %xmm3, %xmm5, %xmm5
+; XOP-NEXT: vpshlw %xmm4, %xmm5, %xmm5
; XOP-NEXT: vpaddw %xmm5, %xmm0, %xmm5
-; XOP-NEXT: vpshaw %xmm2, %xmm5, %xmm5
-; XOP-NEXT: vinsertf128 $1, %xmm4, %ymm5, %ymm4
+; XOP-NEXT: vpshaw %xmm3, %xmm5, %xmm5
+; XOP-NEXT: vinsertf128 $1, %xmm2, %ymm5, %ymm2
; XOP-NEXT: vmovdqa {{.*#+}} ymm5 = [0,65535,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,65535]
-; XOP-NEXT: vpcmov %ymm5, %ymm0, %ymm4, %ymm0
-; XOP-NEXT: vextractf128 $1, %ymm1, %xmm4
-; XOP-NEXT: vpsraw $15, %xmm4, %xmm6
-; XOP-NEXT: vpshlw %xmm3, %xmm6, %xmm6
-; XOP-NEXT: vpaddw %xmm6, %xmm4, %xmm4
-; XOP-NEXT: vpshaw %xmm2, %xmm4, %xmm4
+; XOP-NEXT: vpcmov %ymm5, %ymm0, %ymm2, %ymm0
+; XOP-NEXT: vextractf128 $1, %ymm1, %xmm2
+; XOP-NEXT: vpsraw $15, %xmm2, %xmm6
+; XOP-NEXT: vpshlw %xmm4, %xmm6, %xmm6
+; XOP-NEXT: vpaddw %xmm6, %xmm2, %xmm2
+; XOP-NEXT: vpshaw %xmm3, %xmm2, %xmm2
; XOP-NEXT: vpsraw $15, %xmm1, %xmm6
-; XOP-NEXT: vpshlw %xmm3, %xmm6, %xmm3
-; XOP-NEXT: vpaddw %xmm3, %xmm1, %xmm3
-; XOP-NEXT: vpshaw %xmm2, %xmm3, %xmm2
-; XOP-NEXT: vinsertf128 $1, %xmm4, %ymm2, %ymm2
+; XOP-NEXT: vpshlw %xmm4, %xmm6, %xmm4
+; XOP-NEXT: vpaddw %xmm4, %xmm1, %xmm4
+; XOP-NEXT: vpshaw %xmm3, %xmm4, %xmm3
+; XOP-NEXT: vinsertf128 $1, %xmm2, %ymm3, %ymm2
; XOP-NEXT: vpcmov %ymm5, %ymm1, %ymm2, %ymm1
; XOP-NEXT: retq
%1 = sdiv <32 x i16> %x, <i16 1, i16 4, i16 2, i16 16, i16 8, i16 32, i16 64, i16 2, i16 1, i16 4, i16 2, i16 16, i16 8, i16 32, i16 64, i16 2, i16 1, i16 4, i16 2, i16 16, i16 8, i16 32, i16 64, i16 2, i16 1, i16 4, i16 2, i16 16, i16 8, i16 32, i16 64, i16 2>
@@ -1544,17 +1537,13 @@ define <2 x i64> @combine_vec_sdiv_by_po
;
; XOP-LABEL: combine_vec_sdiv_by_pow2b_v2i64:
; XOP: # %bb.0:
-; XOP-NEXT: movl $2, %eax
-; XOP-NEXT: vmovq %rax, %xmm1
-; XOP-NEXT: vpslldq {{.*#+}} xmm1 = zero,zero,zero,zero,zero,zero,zero,zero,xmm1[0,1,2,3,4,5,6,7]
-; XOP-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; XOP-NEXT: vpsubq %xmm1, %xmm2, %xmm1
-; XOP-NEXT: vpsubq {{.*}}(%rip), %xmm2, %xmm3
-; XOP-NEXT: vpshaq %xmm3, %xmm0, %xmm3
-; XOP-NEXT: vpsubq {{.*}}(%rip), %xmm2, %xmm2
-; XOP-NEXT: vpshlq %xmm2, %xmm3, %xmm2
-; XOP-NEXT: vpaddq %xmm2, %xmm0, %xmm2
-; XOP-NEXT: vpshaq %xmm1, %xmm2, %xmm1
+; XOP-NEXT: vpshaq {{.*}}(%rip), %xmm0, %xmm1
+; XOP-NEXT: vpshlq {{.*}}(%rip), %xmm1, %xmm1
+; XOP-NEXT: vpaddq %xmm1, %xmm0, %xmm1
+; XOP-NEXT: movq $-2, %rax
+; XOP-NEXT: vmovq %rax, %xmm2
+; XOP-NEXT: vpslldq {{.*#+}} xmm2 = zero,zero,zero,zero,zero,zero,zero,zero,xmm2[0,1,2,3,4,5,6,7]
+; XOP-NEXT: vpshaq %xmm2, %xmm1, %xmm1
; XOP-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3],xmm1[4,5,6,7]
; XOP-NEXT: retq
%1 = sdiv <2 x i64> %x, <i64 1, i64 4>
@@ -1658,25 +1647,20 @@ define <4 x i64> @combine_vec_sdiv_by_po
;
; XOP-LABEL: combine_vec_sdiv_by_pow2b_v4i64:
; XOP: # %bb.0:
-; XOP-NEXT: movl $2, %eax
-; XOP-NEXT: vmovq %rax, %xmm1
-; XOP-NEXT: vpslldq {{.*#+}} xmm1 = zero,zero,zero,zero,zero,zero,zero,zero,xmm1[0,1,2,3,4,5,6,7]
-; XOP-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; XOP-NEXT: vpsubq %xmm1, %xmm2, %xmm1
-; XOP-NEXT: vpsubq {{.*}}(%rip), %xmm2, %xmm3
-; XOP-NEXT: vpshaq %xmm3, %xmm0, %xmm4
-; XOP-NEXT: vpsubq {{.*}}(%rip), %xmm2, %xmm5
-; XOP-NEXT: vpshlq %xmm5, %xmm4, %xmm4
-; XOP-NEXT: vpaddq %xmm4, %xmm0, %xmm4
-; XOP-NEXT: vpshaq %xmm1, %xmm4, %xmm1
-; XOP-NEXT: vextractf128 $1, %ymm0, %xmm4
-; XOP-NEXT: vpshaq %xmm3, %xmm4, %xmm3
-; XOP-NEXT: vpsubq {{.*}}(%rip), %xmm2, %xmm5
-; XOP-NEXT: vpshlq %xmm5, %xmm3, %xmm3
-; XOP-NEXT: vpaddq %xmm3, %xmm4, %xmm3
-; XOP-NEXT: vpsubq {{.*}}(%rip), %xmm2, %xmm2
-; XOP-NEXT: vpshaq %xmm2, %xmm3, %xmm2
-; XOP-NEXT: vinsertf128 $1, %xmm2, %ymm1, %ymm1
+; XOP-NEXT: vmovdqa {{.*#+}} xmm1 = [18446744073709551553,18446744073709551553]
+; XOP-NEXT: vpshaq %xmm1, %xmm0, %xmm2
+; XOP-NEXT: vpshlq {{.*}}(%rip), %xmm2, %xmm2
+; XOP-NEXT: vpaddq %xmm2, %xmm0, %xmm2
+; XOP-NEXT: movq $-2, %rax
+; XOP-NEXT: vmovq %rax, %xmm3
+; XOP-NEXT: vpslldq {{.*#+}} xmm3 = zero,zero,zero,zero,zero,zero,zero,zero,xmm3[0,1,2,3,4,5,6,7]
+; XOP-NEXT: vpshaq %xmm3, %xmm2, %xmm2
+; XOP-NEXT: vextractf128 $1, %ymm0, %xmm3
+; XOP-NEXT: vpshaq %xmm1, %xmm3, %xmm1
+; XOP-NEXT: vpshlq {{.*}}(%rip), %xmm1, %xmm1
+; XOP-NEXT: vpaddq %xmm1, %xmm3, %xmm1
+; XOP-NEXT: vpshaq {{.*}}(%rip), %xmm1, %xmm1
+; XOP-NEXT: vinsertf128 $1, %xmm1, %ymm2, %ymm1
; XOP-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4,5,6,7]
; XOP-NEXT: retq
%1 = sdiv <4 x i64> %x, <i64 1, i64 4, i64 8, i64 16>
@@ -1841,36 +1825,34 @@ define <8 x i64> @combine_vec_sdiv_by_po
;
; XOP-LABEL: combine_vec_sdiv_by_pow2b_v8i64:
; XOP: # %bb.0:
-; XOP-NEXT: movl $2, %eax
-; XOP-NEXT: vmovq %rax, %xmm2
-; XOP-NEXT: vpslldq {{.*#+}} xmm2 = zero,zero,zero,zero,zero,zero,zero,zero,xmm2[0,1,2,3,4,5,6,7]
-; XOP-NEXT: vpxor %xmm3, %xmm3, %xmm3
-; XOP-NEXT: vpsubq %xmm2, %xmm3, %xmm9
-; XOP-NEXT: vpsubq {{.*}}(%rip), %xmm3, %xmm4
-; XOP-NEXT: vpshaq %xmm4, %xmm0, %xmm5
-; XOP-NEXT: vpsubq {{.*}}(%rip), %xmm3, %xmm6
-; XOP-NEXT: vpshlq %xmm6, %xmm5, %xmm5
-; XOP-NEXT: vpaddq %xmm5, %xmm0, %xmm5
-; XOP-NEXT: vpshaq %xmm9, %xmm5, %xmm8
-; XOP-NEXT: vextractf128 $1, %ymm0, %xmm7
-; XOP-NEXT: vpshaq %xmm4, %xmm7, %xmm5
-; XOP-NEXT: vpsubq {{.*}}(%rip), %xmm3, %xmm2
-; XOP-NEXT: vpshlq %xmm2, %xmm5, %xmm5
-; XOP-NEXT: vpaddq %xmm5, %xmm7, %xmm5
-; XOP-NEXT: vpsubq {{.*}}(%rip), %xmm3, %xmm3
-; XOP-NEXT: vpshaq %xmm3, %xmm5, %xmm5
-; XOP-NEXT: vinsertf128 $1, %xmm5, %ymm8, %ymm5
-; XOP-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm5[2,3,4,5,6,7]
-; XOP-NEXT: vpshaq %xmm4, %xmm1, %xmm5
-; XOP-NEXT: vpshlq %xmm6, %xmm5, %xmm5
-; XOP-NEXT: vpaddq %xmm5, %xmm1, %xmm5
-; XOP-NEXT: vpshaq %xmm9, %xmm5, %xmm5
-; XOP-NEXT: vextractf128 $1, %ymm1, %xmm6
-; XOP-NEXT: vpshaq %xmm4, %xmm6, %xmm4
-; XOP-NEXT: vpshlq %xmm2, %xmm4, %xmm2
-; XOP-NEXT: vpaddq %xmm2, %xmm6, %xmm2
-; XOP-NEXT: vpshaq %xmm3, %xmm2, %xmm2
-; XOP-NEXT: vinsertf128 $1, %xmm2, %ymm5, %ymm2
+; XOP-NEXT: vextractf128 $1, %ymm0, %xmm2
+; XOP-NEXT: vmovdqa {{.*#+}} xmm3 = [18446744073709551553,18446744073709551553]
+; XOP-NEXT: vpshaq %xmm3, %xmm2, %xmm4
+; XOP-NEXT: vmovdqa {{.*#+}} xmm8 = [18446744073709551555,18446744073709551556]
+; XOP-NEXT: vpshlq %xmm8, %xmm4, %xmm4
+; XOP-NEXT: vpaddq %xmm4, %xmm2, %xmm2
+; XOP-NEXT: vmovdqa {{.*#+}} xmm4 = [18446744073709551613,18446744073709551612]
+; XOP-NEXT: vpshaq %xmm4, %xmm2, %xmm2
+; XOP-NEXT: vpshaq %xmm3, %xmm0, %xmm6
+; XOP-NEXT: vmovdqa {{.*#+}} xmm7 = [18446744073709551552,18446744073709551554]
+; XOP-NEXT: vpshlq %xmm7, %xmm6, %xmm6
+; XOP-NEXT: vpaddq %xmm6, %xmm0, %xmm6
+; XOP-NEXT: movq $-2, %rax
+; XOP-NEXT: vmovq %rax, %xmm5
+; XOP-NEXT: vpslldq {{.*#+}} xmm5 = zero,zero,zero,zero,zero,zero,zero,zero,xmm5[0,1,2,3,4,5,6,7]
+; XOP-NEXT: vpshaq %xmm5, %xmm6, %xmm6
+; XOP-NEXT: vinsertf128 $1, %xmm2, %ymm6, %ymm2
+; XOP-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm2[2,3,4,5,6,7]
+; XOP-NEXT: vextractf128 $1, %ymm1, %xmm2
+; XOP-NEXT: vpshaq %xmm3, %xmm2, %xmm6
+; XOP-NEXT: vpshlq %xmm8, %xmm6, %xmm6
+; XOP-NEXT: vpaddq %xmm6, %xmm2, %xmm2
+; XOP-NEXT: vpshaq %xmm4, %xmm2, %xmm2
+; XOP-NEXT: vpshaq %xmm3, %xmm1, %xmm3
+; XOP-NEXT: vpshlq %xmm7, %xmm3, %xmm3
+; XOP-NEXT: vpaddq %xmm3, %xmm1, %xmm3
+; XOP-NEXT: vpshaq %xmm5, %xmm3, %xmm3
+; XOP-NEXT: vinsertf128 $1, %xmm2, %ymm3, %ymm2
; XOP-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2,3,4,5,6,7]
; XOP-NEXT: retq
%1 = sdiv <8 x i64> %x, <i64 1, i64 4, i64 8, i64 16, i64 1, i64 4, i64 8, i64 16>
@@ -2235,11 +2217,9 @@ define <16 x i8> @non_splat_minus_one_di
; XOP: # %bb.0:
; XOP-NEXT: vpxor %xmm1, %xmm1, %xmm1
; XOP-NEXT: vpcmpgtb %xmm0, %xmm1, %xmm2
-; XOP-NEXT: vpsubb {{.*}}(%rip), %xmm1, %xmm3
-; XOP-NEXT: vpshlb %xmm3, %xmm2, %xmm2
+; XOP-NEXT: vpshlb {{.*}}(%rip), %xmm2, %xmm2
; XOP-NEXT: vpaddb %xmm2, %xmm0, %xmm2
-; XOP-NEXT: vpsubb {{.*}}(%rip), %xmm1, %xmm3
-; XOP-NEXT: vpshab %xmm3, %xmm2, %xmm2
+; XOP-NEXT: vpshab {{.*}}(%rip), %xmm2, %xmm2
; XOP-NEXT: vmovdqa {{.*#+}} xmm3 = [0,0,255,0,0,0,255,0,0,255,255,255,255,255,255,255]
; XOP-NEXT: vpblendvb %xmm3, %xmm2, %xmm0, %xmm0
; XOP-NEXT: vpsubb %xmm0, %xmm1, %xmm1
Modified: llvm/trunk/test/CodeGen/X86/vector-shift-ashr-128.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/vector-shift-ashr-128.ll?rev=337578&r1=337577&r2=337578&view=diff
==============================================================================
--- llvm/trunk/test/CodeGen/X86/vector-shift-ashr-128.ll (original)
+++ llvm/trunk/test/CodeGen/X86/vector-shift-ashr-128.ll Fri Jul 20 09:55:18 2018
@@ -1110,9 +1110,7 @@ define <2 x i64> @constant_shift_v2i64(<
;
; XOP-LABEL: constant_shift_v2i64:
; XOP: # %bb.0:
-; XOP-NEXT: vpxor %xmm1, %xmm1, %xmm1
-; XOP-NEXT: vpsubq {{.*}}(%rip), %xmm1, %xmm1
-; XOP-NEXT: vpshaq %xmm1, %xmm0, %xmm0
+; XOP-NEXT: vpshaq {{.*}}(%rip), %xmm0, %xmm0
; XOP-NEXT: retq
;
; AVX512-LABEL: constant_shift_v2i64:
@@ -1281,9 +1279,7 @@ define <8 x i16> @constant_shift_v8i16(<
;
; XOP-LABEL: constant_shift_v8i16:
; XOP: # %bb.0:
-; XOP-NEXT: vpxor %xmm1, %xmm1, %xmm1
-; XOP-NEXT: vpsubw {{.*}}(%rip), %xmm1, %xmm1
-; XOP-NEXT: vpshaw %xmm1, %xmm0, %xmm0
+; XOP-NEXT: vpshaw {{.*}}(%rip), %xmm0, %xmm0
; XOP-NEXT: retq
;
; AVX512DQ-LABEL: constant_shift_v8i16:
@@ -1462,9 +1458,7 @@ define <16 x i8> @constant_shift_v16i8(<
;
; XOP-LABEL: constant_shift_v16i8:
; XOP: # %bb.0:
-; XOP-NEXT: vpxor %xmm1, %xmm1, %xmm1
-; XOP-NEXT: vpsubb {{.*}}(%rip), %xmm1, %xmm1
-; XOP-NEXT: vpshab %xmm1, %xmm0, %xmm0
+; XOP-NEXT: vpshab {{.*}}(%rip), %xmm0, %xmm0
; XOP-NEXT: retq
;
; AVX512DQ-LABEL: constant_shift_v16i8:
@@ -1601,9 +1595,7 @@ define <2 x i64> @splatconstant_shift_v2
;
; XOP-LABEL: splatconstant_shift_v2i64:
; XOP: # %bb.0:
-; XOP-NEXT: vpxor %xmm1, %xmm1, %xmm1
-; XOP-NEXT: vpsubq {{.*}}(%rip), %xmm1, %xmm1
-; XOP-NEXT: vpshaq %xmm1, %xmm0, %xmm0
+; XOP-NEXT: vpshaq {{.*}}(%rip), %xmm0, %xmm0
; XOP-NEXT: retq
;
; AVX512-LABEL: splatconstant_shift_v2i64:
@@ -1721,9 +1713,7 @@ define <16 x i8> @splatconstant_shift_v1
;
; XOP-LABEL: splatconstant_shift_v16i8:
; XOP: # %bb.0:
-; XOP-NEXT: vpxor %xmm1, %xmm1, %xmm1
-; XOP-NEXT: vpsubb {{.*}}(%rip), %xmm1, %xmm1
-; XOP-NEXT: vpshab %xmm1, %xmm0, %xmm0
+; XOP-NEXT: vpshab {{.*}}(%rip), %xmm0, %xmm0
; XOP-NEXT: retq
;
; AVX512-LABEL: splatconstant_shift_v16i8:
Modified: llvm/trunk/test/CodeGen/X86/vector-shift-ashr-256.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/vector-shift-ashr-256.ll?rev=337578&r1=337577&r2=337578&view=diff
==============================================================================
--- llvm/trunk/test/CodeGen/X86/vector-shift-ashr-256.ll (original)
+++ llvm/trunk/test/CodeGen/X86/vector-shift-ashr-256.ll Fri Jul 20 09:55:18 2018
@@ -1151,13 +1151,10 @@ define <4 x i64> @constant_shift_v4i64(<
;
; XOPAVX1-LABEL: constant_shift_v4i64:
; XOPAVX1: # %bb.0:
-; XOPAVX1-NEXT: vpxor %xmm1, %xmm1, %xmm1
-; XOPAVX1-NEXT: vpsubq {{.*}}(%rip), %xmm1, %xmm2
-; XOPAVX1-NEXT: vextractf128 $1, %ymm0, %xmm3
-; XOPAVX1-NEXT: vpshaq %xmm2, %xmm3, %xmm2
-; XOPAVX1-NEXT: vpsubq {{.*}}(%rip), %xmm1, %xmm1
-; XOPAVX1-NEXT: vpshaq %xmm1, %xmm0, %xmm0
-; XOPAVX1-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0
+; XOPAVX1-NEXT: vpshaq {{.*}}(%rip), %xmm0, %xmm1
+; XOPAVX1-NEXT: vextractf128 $1, %ymm0, %xmm0
+; XOPAVX1-NEXT: vpshaq {{.*}}(%rip), %xmm0, %xmm0
+; XOPAVX1-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
; XOPAVX1-NEXT: retq
;
; XOPAVX2-LABEL: constant_shift_v4i64:
@@ -1331,24 +1328,18 @@ define <16 x i16> @constant_shift_v16i16
;
; XOPAVX1-LABEL: constant_shift_v16i16:
; XOPAVX1: # %bb.0:
-; XOPAVX1-NEXT: vpxor %xmm1, %xmm1, %xmm1
-; XOPAVX1-NEXT: vpsubw {{.*}}(%rip), %xmm1, %xmm2
-; XOPAVX1-NEXT: vextractf128 $1, %ymm0, %xmm3
-; XOPAVX1-NEXT: vpshaw %xmm2, %xmm3, %xmm2
-; XOPAVX1-NEXT: vpsubw {{.*}}(%rip), %xmm1, %xmm1
-; XOPAVX1-NEXT: vpshaw %xmm1, %xmm0, %xmm0
-; XOPAVX1-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0
+; XOPAVX1-NEXT: vpshaw {{.*}}(%rip), %xmm0, %xmm1
+; XOPAVX1-NEXT: vextractf128 $1, %ymm0, %xmm0
+; XOPAVX1-NEXT: vpshaw {{.*}}(%rip), %xmm0, %xmm0
+; XOPAVX1-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
; XOPAVX1-NEXT: retq
;
; XOPAVX2-LABEL: constant_shift_v16i16:
; XOPAVX2: # %bb.0:
-; XOPAVX2-NEXT: vpxor %xmm1, %xmm1, %xmm1
-; XOPAVX2-NEXT: vpsubw {{.*}}(%rip), %xmm1, %xmm2
-; XOPAVX2-NEXT: vextracti128 $1, %ymm0, %xmm3
-; XOPAVX2-NEXT: vpshaw %xmm2, %xmm3, %xmm2
-; XOPAVX2-NEXT: vpsubw {{.*}}(%rip), %xmm1, %xmm1
-; XOPAVX2-NEXT: vpshaw %xmm1, %xmm0, %xmm0
-; XOPAVX2-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm0
+; XOPAVX2-NEXT: vpshaw {{.*}}(%rip), %xmm0, %xmm1
+; XOPAVX2-NEXT: vextracti128 $1, %ymm0, %xmm0
+; XOPAVX2-NEXT: vpshaw {{.*}}(%rip), %xmm0, %xmm0
+; XOPAVX2-NEXT: vinserti128 $1, %xmm0, %ymm1, %ymm0
; XOPAVX2-NEXT: retq
;
; AVX512DQ-LABEL: constant_shift_v16i16:
@@ -1493,22 +1484,20 @@ define <32 x i8> @constant_shift_v32i8(<
;
; XOPAVX1-LABEL: constant_shift_v32i8:
; XOPAVX1: # %bb.0:
-; XOPAVX1-NEXT: vpxor %xmm1, %xmm1, %xmm1
-; XOPAVX1-NEXT: vpsubb {{.*}}(%rip), %xmm1, %xmm1
-; XOPAVX1-NEXT: vextractf128 $1, %ymm0, %xmm2
-; XOPAVX1-NEXT: vpshab %xmm1, %xmm2, %xmm2
-; XOPAVX1-NEXT: vpshab %xmm1, %xmm0, %xmm0
-; XOPAVX1-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0
+; XOPAVX1-NEXT: vextractf128 $1, %ymm0, %xmm1
+; XOPAVX1-NEXT: vmovdqa {{.*#+}} xmm2 = [0,255,254,253,252,251,250,249,249,250,251,252,253,254,255,0]
+; XOPAVX1-NEXT: vpshab %xmm2, %xmm1, %xmm1
+; XOPAVX1-NEXT: vpshab %xmm2, %xmm0, %xmm0
+; XOPAVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
; XOPAVX1-NEXT: retq
;
; XOPAVX2-LABEL: constant_shift_v32i8:
; XOPAVX2: # %bb.0:
-; XOPAVX2-NEXT: vpxor %xmm1, %xmm1, %xmm1
-; XOPAVX2-NEXT: vpsubb {{.*}}(%rip), %xmm1, %xmm1
-; XOPAVX2-NEXT: vextracti128 $1, %ymm0, %xmm2
-; XOPAVX2-NEXT: vpshab %xmm1, %xmm2, %xmm2
-; XOPAVX2-NEXT: vpshab %xmm1, %xmm0, %xmm0
-; XOPAVX2-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm0
+; XOPAVX2-NEXT: vextracti128 $1, %ymm0, %xmm1
+; XOPAVX2-NEXT: vmovdqa {{.*#+}} xmm2 = [0,255,254,253,252,251,250,249,249,250,251,252,253,254,255,0]
+; XOPAVX2-NEXT: vpshab %xmm2, %xmm1, %xmm1
+; XOPAVX2-NEXT: vpshab %xmm2, %xmm0, %xmm0
+; XOPAVX2-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm0
; XOPAVX2-NEXT: retq
;
; AVX512DQ-LABEL: constant_shift_v32i8:
@@ -1685,12 +1674,11 @@ define <4 x i64> @splatconstant_shift_v4
;
; XOPAVX1-LABEL: splatconstant_shift_v4i64:
; XOPAVX1: # %bb.0:
-; XOPAVX1-NEXT: vpxor %xmm1, %xmm1, %xmm1
-; XOPAVX1-NEXT: vpsubq {{.*}}(%rip), %xmm1, %xmm1
-; XOPAVX1-NEXT: vextractf128 $1, %ymm0, %xmm2
-; XOPAVX1-NEXT: vpshaq %xmm1, %xmm2, %xmm2
-; XOPAVX1-NEXT: vpshaq %xmm1, %xmm0, %xmm0
-; XOPAVX1-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0
+; XOPAVX1-NEXT: vextractf128 $1, %ymm0, %xmm1
+; XOPAVX1-NEXT: vmovdqa {{.*#+}} xmm2 = [18446744073709551609,18446744073709551609]
+; XOPAVX1-NEXT: vpshaq %xmm2, %xmm1, %xmm1
+; XOPAVX1-NEXT: vpshaq %xmm2, %xmm0, %xmm0
+; XOPAVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
; XOPAVX1-NEXT: retq
;
; XOPAVX2-LABEL: splatconstant_shift_v4i64:
@@ -1868,12 +1856,11 @@ define <32 x i8> @splatconstant_shift_v3
;
; XOPAVX1-LABEL: splatconstant_shift_v32i8:
; XOPAVX1: # %bb.0:
-; XOPAVX1-NEXT: vpxor %xmm1, %xmm1, %xmm1
-; XOPAVX1-NEXT: vpsubb {{.*}}(%rip), %xmm1, %xmm1
-; XOPAVX1-NEXT: vextractf128 $1, %ymm0, %xmm2
-; XOPAVX1-NEXT: vpshab %xmm1, %xmm2, %xmm2
-; XOPAVX1-NEXT: vpshab %xmm1, %xmm0, %xmm0
-; XOPAVX1-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0
+; XOPAVX1-NEXT: vextractf128 $1, %ymm0, %xmm1
+; XOPAVX1-NEXT: vmovdqa {{.*#+}} xmm2 = [253,253,253,253,253,253,253,253,253,253,253,253,253,253,253,253]
+; XOPAVX1-NEXT: vpshab %xmm2, %xmm1, %xmm1
+; XOPAVX1-NEXT: vpshab %xmm2, %xmm0, %xmm0
+; XOPAVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
; XOPAVX1-NEXT: retq
;
; XOPAVX2-LABEL: splatconstant_shift_v32i8:
Modified: llvm/trunk/test/CodeGen/X86/vector-shift-lshr-128.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/vector-shift-lshr-128.ll?rev=337578&r1=337577&r2=337578&view=diff
==============================================================================
--- llvm/trunk/test/CodeGen/X86/vector-shift-lshr-128.ll (original)
+++ llvm/trunk/test/CodeGen/X86/vector-shift-lshr-128.ll Fri Jul 20 09:55:18 2018
@@ -876,9 +876,7 @@ define <2 x i64> @constant_shift_v2i64(<
;
; XOPAVX1-LABEL: constant_shift_v2i64:
; XOPAVX1: # %bb.0:
-; XOPAVX1-NEXT: vpxor %xmm1, %xmm1, %xmm1
-; XOPAVX1-NEXT: vpsubq {{.*}}(%rip), %xmm1, %xmm1
-; XOPAVX1-NEXT: vpshlq %xmm1, %xmm0, %xmm0
+; XOPAVX1-NEXT: vpshlq {{.*}}(%rip), %xmm0, %xmm0
; XOPAVX1-NEXT: retq
;
; XOPAVX2-LABEL: constant_shift_v2i64:
@@ -1042,9 +1040,7 @@ define <8 x i16> @constant_shift_v8i16(<
;
; XOP-LABEL: constant_shift_v8i16:
; XOP: # %bb.0:
-; XOP-NEXT: vpxor %xmm1, %xmm1, %xmm1
-; XOP-NEXT: vpsubw {{.*}}(%rip), %xmm1, %xmm1
-; XOP-NEXT: vpshlw %xmm1, %xmm0, %xmm0
+; XOP-NEXT: vpshlw {{.*}}(%rip), %xmm0, %xmm0
; XOP-NEXT: retq
;
; AVX512DQ-LABEL: constant_shift_v8i16:
@@ -1169,9 +1165,7 @@ define <16 x i8> @constant_shift_v16i8(<
;
; XOP-LABEL: constant_shift_v16i8:
; XOP: # %bb.0:
-; XOP-NEXT: vpxor %xmm1, %xmm1, %xmm1
-; XOP-NEXT: vpsubb {{.*}}(%rip), %xmm1, %xmm1
-; XOP-NEXT: vpshlb %xmm1, %xmm0, %xmm0
+; XOP-NEXT: vpshlb {{.*}}(%rip), %xmm0, %xmm0
; XOP-NEXT: retq
;
; AVX512DQ-LABEL: constant_shift_v16i8:
@@ -1363,9 +1357,7 @@ define <16 x i8> @splatconstant_shift_v1
;
; XOP-LABEL: splatconstant_shift_v16i8:
; XOP: # %bb.0:
-; XOP-NEXT: vpxor %xmm1, %xmm1, %xmm1
-; XOP-NEXT: vpsubb {{.*}}(%rip), %xmm1, %xmm1
-; XOP-NEXT: vpshlb %xmm1, %xmm0, %xmm0
+; XOP-NEXT: vpshlb {{.*}}(%rip), %xmm0, %xmm0
; XOP-NEXT: retq
;
; AVX512-LABEL: splatconstant_shift_v16i8:
Modified: llvm/trunk/test/CodeGen/X86/vector-shift-lshr-256.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/vector-shift-lshr-256.ll?rev=337578&r1=337577&r2=337578&view=diff
==============================================================================
--- llvm/trunk/test/CodeGen/X86/vector-shift-lshr-256.ll (original)
+++ llvm/trunk/test/CodeGen/X86/vector-shift-lshr-256.ll Fri Jul 20 09:55:18 2018
@@ -904,13 +904,10 @@ define <4 x i64> @constant_shift_v4i64(<
;
; XOPAVX1-LABEL: constant_shift_v4i64:
; XOPAVX1: # %bb.0:
-; XOPAVX1-NEXT: vpxor %xmm1, %xmm1, %xmm1
-; XOPAVX1-NEXT: vpsubq {{.*}}(%rip), %xmm1, %xmm2
-; XOPAVX1-NEXT: vextractf128 $1, %ymm0, %xmm3
-; XOPAVX1-NEXT: vpshlq %xmm2, %xmm3, %xmm2
-; XOPAVX1-NEXT: vpsubq {{.*}}(%rip), %xmm1, %xmm1
-; XOPAVX1-NEXT: vpshlq %xmm1, %xmm0, %xmm0
-; XOPAVX1-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0
+; XOPAVX1-NEXT: vpshlq {{.*}}(%rip), %xmm0, %xmm1
+; XOPAVX1-NEXT: vextractf128 $1, %ymm0, %xmm0
+; XOPAVX1-NEXT: vpshlq {{.*}}(%rip), %xmm0, %xmm0
+; XOPAVX1-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
; XOPAVX1-NEXT: retq
;
; XOPAVX2-LABEL: constant_shift_v4i64:
@@ -1062,24 +1059,18 @@ define <16 x i16> @constant_shift_v16i16
;
; XOPAVX1-LABEL: constant_shift_v16i16:
; XOPAVX1: # %bb.0:
-; XOPAVX1-NEXT: vpxor %xmm1, %xmm1, %xmm1
-; XOPAVX1-NEXT: vpsubw {{.*}}(%rip), %xmm1, %xmm2
-; XOPAVX1-NEXT: vextractf128 $1, %ymm0, %xmm3
-; XOPAVX1-NEXT: vpshlw %xmm2, %xmm3, %xmm2
-; XOPAVX1-NEXT: vpsubw {{.*}}(%rip), %xmm1, %xmm1
-; XOPAVX1-NEXT: vpshlw %xmm1, %xmm0, %xmm0
-; XOPAVX1-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0
+; XOPAVX1-NEXT: vpshlw {{.*}}(%rip), %xmm0, %xmm1
+; XOPAVX1-NEXT: vextractf128 $1, %ymm0, %xmm0
+; XOPAVX1-NEXT: vpshlw {{.*}}(%rip), %xmm0, %xmm0
+; XOPAVX1-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
; XOPAVX1-NEXT: retq
;
; XOPAVX2-LABEL: constant_shift_v16i16:
; XOPAVX2: # %bb.0:
-; XOPAVX2-NEXT: vpxor %xmm1, %xmm1, %xmm1
-; XOPAVX2-NEXT: vpsubw {{.*}}(%rip), %xmm1, %xmm2
-; XOPAVX2-NEXT: vextracti128 $1, %ymm0, %xmm3
-; XOPAVX2-NEXT: vpshlw %xmm2, %xmm3, %xmm2
-; XOPAVX2-NEXT: vpsubw {{.*}}(%rip), %xmm1, %xmm1
-; XOPAVX2-NEXT: vpshlw %xmm1, %xmm0, %xmm0
-; XOPAVX2-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm0
+; XOPAVX2-NEXT: vpshlw {{.*}}(%rip), %xmm0, %xmm1
+; XOPAVX2-NEXT: vextracti128 $1, %ymm0, %xmm0
+; XOPAVX2-NEXT: vpshlw {{.*}}(%rip), %xmm0, %xmm0
+; XOPAVX2-NEXT: vinserti128 $1, %xmm0, %ymm1, %ymm0
; XOPAVX2-NEXT: retq
;
; AVX512DQ-LABEL: constant_shift_v16i16:
@@ -1195,22 +1186,20 @@ define <32 x i8> @constant_shift_v32i8(<
;
; XOPAVX1-LABEL: constant_shift_v32i8:
; XOPAVX1: # %bb.0:
-; XOPAVX1-NEXT: vpxor %xmm1, %xmm1, %xmm1
-; XOPAVX1-NEXT: vpsubb {{.*}}(%rip), %xmm1, %xmm1
-; XOPAVX1-NEXT: vextractf128 $1, %ymm0, %xmm2
-; XOPAVX1-NEXT: vpshlb %xmm1, %xmm2, %xmm2
-; XOPAVX1-NEXT: vpshlb %xmm1, %xmm0, %xmm0
-; XOPAVX1-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0
+; XOPAVX1-NEXT: vextractf128 $1, %ymm0, %xmm1
+; XOPAVX1-NEXT: vmovdqa {{.*#+}} xmm2 = [0,255,254,253,252,251,250,249,249,250,251,252,253,254,255,0]
+; XOPAVX1-NEXT: vpshlb %xmm2, %xmm1, %xmm1
+; XOPAVX1-NEXT: vpshlb %xmm2, %xmm0, %xmm0
+; XOPAVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
; XOPAVX1-NEXT: retq
;
; XOPAVX2-LABEL: constant_shift_v32i8:
; XOPAVX2: # %bb.0:
-; XOPAVX2-NEXT: vpxor %xmm1, %xmm1, %xmm1
-; XOPAVX2-NEXT: vpsubb {{.*}}(%rip), %xmm1, %xmm1
-; XOPAVX2-NEXT: vextracti128 $1, %ymm0, %xmm2
-; XOPAVX2-NEXT: vpshlb %xmm1, %xmm2, %xmm2
-; XOPAVX2-NEXT: vpshlb %xmm1, %xmm0, %xmm0
-; XOPAVX2-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm0
+; XOPAVX2-NEXT: vextracti128 $1, %ymm0, %xmm1
+; XOPAVX2-NEXT: vmovdqa {{.*#+}} xmm2 = [0,255,254,253,252,251,250,249,249,250,251,252,253,254,255,0]
+; XOPAVX2-NEXT: vpshlb %xmm2, %xmm1, %xmm1
+; XOPAVX2-NEXT: vpshlb %xmm2, %xmm0, %xmm0
+; XOPAVX2-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm0
; XOPAVX2-NEXT: retq
;
; AVX512DQ-LABEL: constant_shift_v32i8:
@@ -1491,12 +1480,11 @@ define <32 x i8> @splatconstant_shift_v3
;
; XOPAVX1-LABEL: splatconstant_shift_v32i8:
; XOPAVX1: # %bb.0:
-; XOPAVX1-NEXT: vpxor %xmm1, %xmm1, %xmm1
-; XOPAVX1-NEXT: vpsubb {{.*}}(%rip), %xmm1, %xmm1
-; XOPAVX1-NEXT: vextractf128 $1, %ymm0, %xmm2
-; XOPAVX1-NEXT: vpshlb %xmm1, %xmm2, %xmm2
-; XOPAVX1-NEXT: vpshlb %xmm1, %xmm0, %xmm0
-; XOPAVX1-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0
+; XOPAVX1-NEXT: vextractf128 $1, %ymm0, %xmm1
+; XOPAVX1-NEXT: vmovdqa {{.*#+}} xmm2 = [253,253,253,253,253,253,253,253,253,253,253,253,253,253,253,253]
+; XOPAVX1-NEXT: vpshlb %xmm2, %xmm1, %xmm1
+; XOPAVX1-NEXT: vpshlb %xmm2, %xmm0, %xmm0
+; XOPAVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
; XOPAVX1-NEXT: retq
;
; XOPAVX2-LABEL: splatconstant_shift_v32i8:
More information about the llvm-commits
mailing list