[X86][SSE] Don't use 'sign bit select' vXi8 ROTL lowering for splat rotation amounts

Noticed by @spatel on D55747 - we get much better codegen if we use the regular shift expansion.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@349500 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/lib/Target/X86/X86ISelLowering.cpp b/lib/Target/X86/X86ISelLowering.cpp
index 2596a04a18dd6aa4617b65b8f182b907af3ce30d..3d8b1787bb0bec46c6f362c1f3db8dddfb889467 100644 (file)
--- a/lib/Target/X86/X86ISelLowering.cpp
+++ b/lib/Target/X86/X86ISelLowering.cpp
@@ -24813,9 +24813,11 @@ static SDValue LowerRotate(SDValue Op, const X86Subtarget &Subtarget,
   if (0 <= CstSplatIndex)
     return SDValue();
 
+  bool IsSplatAmt = DAG.isSplatValue(Amt);
+
   // v16i8/v32i8: Split rotation into rot4/rot2/rot1 stages and select by
   // the amount bit.
-  if (EltSizeInBits == 8) {
+  if (EltSizeInBits == 8 && !IsSplatAmt) {
     // We don't need ModuloAmt here as we just peek at individual bits.
     MVT ExtVT = MVT::getVectorVT(MVT::i16, NumElts / 2);
 
@@ -24882,8 +24884,7 @@ static SDValue LowerRotate(SDValue Op, const X86Subtarget &Subtarget,
 
   // Fallback for splats + all supported variable shifts.
   // Fallback for non-constants AVX2 vXi16 as well.
-  if (LegalVarShifts || (Subtarget.hasAVX2() && !ConstantAmt) ||
-      DAG.isSplatValue(Amt)) {
+  if (IsSplatAmt || LegalVarShifts || (Subtarget.hasAVX2() && !ConstantAmt)) {
     SDValue AmtR = DAG.getConstant(EltSizeInBits, DL, VT);
     AmtR = DAG.getNode(ISD::SUB, DL, VT, AmtR, Amt);
     SDValue SHL = DAG.getNode(ISD::SHL, DL, VT, R, Amt);

diff --git a/test/CodeGen/X86/vector-fshl-rot-512.ll b/test/CodeGen/X86/vector-fshl-rot-512.ll
index afa235ae005acd841575bddddde2ffda9c440062..3510a22e13f23accd77c3326ffbc6eed863118ab 100644 (file)
--- a/test/CodeGen/X86/vector-fshl-rot-512.ll
+++ b/test/CodeGen/X86/vector-fshl-rot-512.ll
@@ -418,97 +418,53 @@ define <64 x i8> @splatvar_funnnel_v64i8(<64 x i8> %x, <64 x i8> %amt) nounwind
 ; AVX512F-LABEL: splatvar_funnnel_v64i8:
 ; AVX512F:       # %bb.0:
 ; AVX512F-NEXT:    vpbroadcastb %xmm2, %ymm2
-; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm3
-; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
-; AVX512F-NEXT:    vpand %ymm4, %ymm3, %ymm3
-; AVX512F-NEXT:    vpsllw $4, %ymm0, %ymm5
-; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm6 = [240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240]
-; AVX512F-NEXT:    vpand %ymm6, %ymm5, %ymm5
-; AVX512F-NEXT:    vpor %ymm3, %ymm5, %ymm3
-; AVX512F-NEXT:    vpand {{.*}}(%rip), %ymm2, %ymm2
-; AVX512F-NEXT:    vpsllw $5, %ymm2, %ymm2
-; AVX512F-NEXT:    vpblendvb %ymm2, %ymm3, %ymm0, %ymm0
-; AVX512F-NEXT:    vpsrlw $6, %ymm0, %ymm3
-; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm5 = [3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3]
-; AVX512F-NEXT:    vpand %ymm5, %ymm3, %ymm3
-; AVX512F-NEXT:    vpsllw $2, %ymm0, %ymm7
-; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm8 = [252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252]
-; AVX512F-NEXT:    vpand %ymm8, %ymm7, %ymm7
-; AVX512F-NEXT:    vpor %ymm3, %ymm7, %ymm3
-; AVX512F-NEXT:    vpaddb %ymm2, %ymm2, %ymm7
-; AVX512F-NEXT:    vpblendvb %ymm7, %ymm3, %ymm0, %ymm0
-; AVX512F-NEXT:    vpsrlw $7, %ymm0, %ymm3
-; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm9 = [1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1]
-; AVX512F-NEXT:    vpand %ymm9, %ymm3, %ymm3
-; AVX512F-NEXT:    vpaddb %ymm0, %ymm0, %ymm10
-; AVX512F-NEXT:    vpor %ymm3, %ymm10, %ymm3
-; AVX512F-NEXT:    vpaddb %ymm7, %ymm7, %ymm10
-; AVX512F-NEXT:    vpblendvb %ymm10, %ymm3, %ymm0, %ymm0
-; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm3
-; AVX512F-NEXT:    vpand %ymm4, %ymm3, %ymm3
-; AVX512F-NEXT:    vpsllw $4, %ymm1, %ymm4
+; AVX512F-NEXT:    vpand {{.*}}(%rip), %xmm2, %xmm2
+; AVX512F-NEXT:    vpmovzxbq {{.*#+}} xmm3 = xmm2[0],zero,zero,zero,zero,zero,zero,zero,xmm2[1],zero,zero,zero,zero,zero,zero,zero
+; AVX512F-NEXT:    vpsllw %xmm3, %ymm0, %ymm4
+; AVX512F-NEXT:    vpcmpeqd %ymm5, %ymm5, %ymm5
+; AVX512F-NEXT:    vpsllw %xmm3, %ymm5, %ymm6
+; AVX512F-NEXT:    vpbroadcastb %xmm6, %ymm6
 ; AVX512F-NEXT:    vpand %ymm6, %ymm4, %ymm4
-; AVX512F-NEXT:    vpor %ymm3, %ymm4, %ymm3
-; AVX512F-NEXT:    vpblendvb %ymm2, %ymm3, %ymm1, %ymm1
-; AVX512F-NEXT:    vpsrlw $6, %ymm1, %ymm2
-; AVX512F-NEXT:    vpand %ymm5, %ymm2, %ymm2
-; AVX512F-NEXT:    vpsllw $2, %ymm1, %ymm3
-; AVX512F-NEXT:    vpand %ymm8, %ymm3, %ymm3
-; AVX512F-NEXT:    vpor %ymm2, %ymm3, %ymm2
-; AVX512F-NEXT:    vpblendvb %ymm7, %ymm2, %ymm1, %ymm1
-; AVX512F-NEXT:    vpsrlw $7, %ymm1, %ymm2
-; AVX512F-NEXT:    vpand %ymm9, %ymm2, %ymm2
-; AVX512F-NEXT:    vpaddb %ymm1, %ymm1, %ymm3
-; AVX512F-NEXT:    vpor %ymm2, %ymm3, %ymm2
-; AVX512F-NEXT:    vpblendvb %ymm10, %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} xmm7 = [8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8]
+; AVX512F-NEXT:    vpsubb %xmm2, %xmm7, %xmm2
+; AVX512F-NEXT:    vpmovzxbq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,zero,zero,zero,zero,xmm2[1],zero,zero,zero,zero,zero,zero,zero
+; AVX512F-NEXT:    vpsrlw %xmm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsrlw %xmm2, %ymm5, %ymm5
+; AVX512F-NEXT:    vpsrlw $8, %ymm5, %ymm5
+; AVX512F-NEXT:    vpbroadcastb %xmm5, %ymm5
+; AVX512F-NEXT:    vpand %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpor %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpsllw %xmm3, %ymm1, %ymm3
+; AVX512F-NEXT:    vpand %ymm6, %ymm3, %ymm3
+; AVX512F-NEXT:    vpsrlw %xmm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm5, %ymm1, %ymm1
+; AVX512F-NEXT:    vpor %ymm1, %ymm3, %ymm1
 ; AVX512F-NEXT:    retq
 ;
 ; AVX512VL-LABEL: splatvar_funnnel_v64i8:
 ; AVX512VL:       # %bb.0:
 ; AVX512VL-NEXT:    vpbroadcastb %xmm2, %ymm2
-; AVX512VL-NEXT:    vpsrlw $4, %ymm0, %ymm3
-; AVX512VL-NEXT:    vmovdqa {{.*#+}} ymm4 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
-; AVX512VL-NEXT:    vpand %ymm4, %ymm3, %ymm3
-; AVX512VL-NEXT:    vpsllw $4, %ymm0, %ymm5
-; AVX512VL-NEXT:    vmovdqa {{.*#+}} ymm6 = [240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240]
-; AVX512VL-NEXT:    vpand %ymm6, %ymm5, %ymm5
-; AVX512VL-NEXT:    vpor %ymm3, %ymm5, %ymm3
-; AVX512VL-NEXT:    vpand {{.*}}(%rip), %ymm2, %ymm2
-; AVX512VL-NEXT:    vpsllw $5, %ymm2, %ymm2
-; AVX512VL-NEXT:    vpblendvb %ymm2, %ymm3, %ymm0, %ymm0
-; AVX512VL-NEXT:    vpsrlw $6, %ymm0, %ymm3
-; AVX512VL-NEXT:    vmovdqa {{.*#+}} ymm5 = [3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3]
-; AVX512VL-NEXT:    vpand %ymm5, %ymm3, %ymm3
-; AVX512VL-NEXT:    vpsllw $2, %ymm0, %ymm7
-; AVX512VL-NEXT:    vmovdqa {{.*#+}} ymm8 = [252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252]
-; AVX512VL-NEXT:    vpand %ymm8, %ymm7, %ymm7
-; AVX512VL-NEXT:    vpor %ymm3, %ymm7, %ymm3
-; AVX512VL-NEXT:    vpaddb %ymm2, %ymm2, %ymm7
-; AVX512VL-NEXT:    vpblendvb %ymm7, %ymm3, %ymm0, %ymm0
-; AVX512VL-NEXT:    vpsrlw $7, %ymm0, %ymm3
-; AVX512VL-NEXT:    vmovdqa {{.*#+}} ymm9 = [1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1]
-; AVX512VL-NEXT:    vpand %ymm9, %ymm3, %ymm3
-; AVX512VL-NEXT:    vpaddb %ymm0, %ymm0, %ymm10
-; AVX512VL-NEXT:    vpor %ymm3, %ymm10, %ymm3
-; AVX512VL-NEXT:    vpaddb %ymm7, %ymm7, %ymm10
-; AVX512VL-NEXT:    vpblendvb %ymm10, %ymm3, %ymm0, %ymm0
-; AVX512VL-NEXT:    vpsrlw $4, %ymm1, %ymm3
-; AVX512VL-NEXT:    vpand %ymm4, %ymm3, %ymm3
-; AVX512VL-NEXT:    vpsllw $4, %ymm1, %ymm4
+; AVX512VL-NEXT:    vpand {{.*}}(%rip), %xmm2, %xmm2
+; AVX512VL-NEXT:    vpmovzxbq {{.*#+}} xmm3 = xmm2[0],zero,zero,zero,zero,zero,zero,zero,xmm2[1],zero,zero,zero,zero,zero,zero,zero
+; AVX512VL-NEXT:    vpsllw %xmm3, %ymm0, %ymm4
+; AVX512VL-NEXT:    vpcmpeqd %ymm5, %ymm5, %ymm5
+; AVX512VL-NEXT:    vpsllw %xmm3, %ymm5, %ymm6
+; AVX512VL-NEXT:    vpbroadcastb %xmm6, %ymm6
 ; AVX512VL-NEXT:    vpand %ymm6, %ymm4, %ymm4
-; AVX512VL-NEXT:    vpor %ymm3, %ymm4, %ymm3
-; AVX512VL-NEXT:    vpblendvb %ymm2, %ymm3, %ymm1, %ymm1
-; AVX512VL-NEXT:    vpsrlw $6, %ymm1, %ymm2
-; AVX512VL-NEXT:    vpand %ymm5, %ymm2, %ymm2
-; AVX512VL-NEXT:    vpsllw $2, %ymm1, %ymm3
-; AVX512VL-NEXT:    vpand %ymm8, %ymm3, %ymm3
-; AVX512VL-NEXT:    vpor %ymm2, %ymm3, %ymm2
-; AVX512VL-NEXT:    vpblendvb %ymm7, %ymm2, %ymm1, %ymm1
-; AVX512VL-NEXT:    vpsrlw $7, %ymm1, %ymm2
-; AVX512VL-NEXT:    vpand %ymm9, %ymm2, %ymm2
-; AVX512VL-NEXT:    vpaddb %ymm1, %ymm1, %ymm3
-; AVX512VL-NEXT:    vpor %ymm2, %ymm3, %ymm2
-; AVX512VL-NEXT:    vpblendvb %ymm10, %ymm2, %ymm1, %ymm1
+; AVX512VL-NEXT:    vmovdqa {{.*#+}} xmm7 = [8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8]
+; AVX512VL-NEXT:    vpsubb %xmm2, %xmm7, %xmm2
+; AVX512VL-NEXT:    vpmovzxbq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,zero,zero,zero,zero,xmm2[1],zero,zero,zero,zero,zero,zero,zero
+; AVX512VL-NEXT:    vpsrlw %xmm2, %ymm0, %ymm0
+; AVX512VL-NEXT:    vpsrlw %xmm2, %ymm5, %ymm5
+; AVX512VL-NEXT:    vpsrlw $8, %ymm5, %ymm5
+; AVX512VL-NEXT:    vpbroadcastb %xmm5, %ymm5
+; AVX512VL-NEXT:    vpand %ymm5, %ymm0, %ymm0
+; AVX512VL-NEXT:    vpor %ymm0, %ymm4, %ymm0
+; AVX512VL-NEXT:    vpsllw %xmm3, %ymm1, %ymm3
+; AVX512VL-NEXT:    vpand %ymm6, %ymm3, %ymm3
+; AVX512VL-NEXT:    vpsrlw %xmm2, %ymm1, %ymm1
+; AVX512VL-NEXT:    vpand %ymm5, %ymm1, %ymm1
+; AVX512VL-NEXT:    vpor %ymm1, %ymm3, %ymm1
 ; AVX512VL-NEXT:    retq
 ;
 ; AVX512BW-LABEL: splatvar_funnnel_v64i8:

diff --git a/test/CodeGen/X86/vector-fshr-rot-512.ll b/test/CodeGen/X86/vector-fshr-rot-512.ll
index 53b027f688266596450db87fc6378be8d47fe7e7..792f172e4d169b94014281349b1dfe45bd7040f4 100644 (file)
--- a/test/CodeGen/X86/vector-fshr-rot-512.ll
+++ b/test/CodeGen/X86/vector-fshr-rot-512.ll
@@ -434,101 +434,57 @@ define <64 x i8> @splatvar_funnnel_v64i8(<64 x i8> %x, <64 x i8> %amt) nounwind
 ; AVX512F-LABEL: splatvar_funnnel_v64i8:
 ; AVX512F:       # %bb.0:
 ; AVX512F-NEXT:    vpbroadcastb %xmm2, %ymm2
-; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm3
-; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
-; AVX512F-NEXT:    vpand %ymm4, %ymm3, %ymm3
-; AVX512F-NEXT:    vpsllw $4, %ymm0, %ymm5
-; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm6 = [240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240]
-; AVX512F-NEXT:    vpand %ymm6, %ymm5, %ymm5
-; AVX512F-NEXT:    vpor %ymm3, %ymm5, %ymm3
-; AVX512F-NEXT:    vpxor %xmm5, %xmm5, %xmm5
-; AVX512F-NEXT:    vpsubb %ymm2, %ymm5, %ymm2
-; AVX512F-NEXT:    vpand {{.*}}(%rip), %ymm2, %ymm2
-; AVX512F-NEXT:    vpsllw $5, %ymm2, %ymm2
-; AVX512F-NEXT:    vpblendvb %ymm2, %ymm3, %ymm0, %ymm0
-; AVX512F-NEXT:    vpsrlw $6, %ymm0, %ymm3
-; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm5 = [3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3]
-; AVX512F-NEXT:    vpand %ymm5, %ymm3, %ymm3
-; AVX512F-NEXT:    vpsllw $2, %ymm0, %ymm7
-; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm8 = [252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252]
-; AVX512F-NEXT:    vpand %ymm8, %ymm7, %ymm7
-; AVX512F-NEXT:    vpor %ymm3, %ymm7, %ymm3
-; AVX512F-NEXT:    vpaddb %ymm2, %ymm2, %ymm7
-; AVX512F-NEXT:    vpblendvb %ymm7, %ymm3, %ymm0, %ymm0
-; AVX512F-NEXT:    vpsrlw $7, %ymm0, %ymm3
-; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm9 = [1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1]
-; AVX512F-NEXT:    vpand %ymm9, %ymm3, %ymm3
-; AVX512F-NEXT:    vpaddb %ymm0, %ymm0, %ymm10
-; AVX512F-NEXT:    vpor %ymm3, %ymm10, %ymm3
-; AVX512F-NEXT:    vpaddb %ymm7, %ymm7, %ymm10
-; AVX512F-NEXT:    vpblendvb %ymm10, %ymm3, %ymm0, %ymm0
-; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm3
-; AVX512F-NEXT:    vpand %ymm4, %ymm3, %ymm3
-; AVX512F-NEXT:    vpsllw $4, %ymm1, %ymm4
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsubb %xmm2, %xmm3, %xmm2
+; AVX512F-NEXT:    vpand {{.*}}(%rip), %xmm2, %xmm2
+; AVX512F-NEXT:    vpmovzxbq {{.*#+}} xmm3 = xmm2[0],zero,zero,zero,zero,zero,zero,zero,xmm2[1],zero,zero,zero,zero,zero,zero,zero
+; AVX512F-NEXT:    vpsllw %xmm3, %ymm0, %ymm4
+; AVX512F-NEXT:    vpcmpeqd %ymm5, %ymm5, %ymm5
+; AVX512F-NEXT:    vpsllw %xmm3, %ymm5, %ymm6
+; AVX512F-NEXT:    vpbroadcastb %xmm6, %ymm6
 ; AVX512F-NEXT:    vpand %ymm6, %ymm4, %ymm4
-; AVX512F-NEXT:    vpor %ymm3, %ymm4, %ymm3
-; AVX512F-NEXT:    vpblendvb %ymm2, %ymm3, %ymm1, %ymm1
-; AVX512F-NEXT:    vpsrlw $6, %ymm1, %ymm2
-; AVX512F-NEXT:    vpand %ymm5, %ymm2, %ymm2
-; AVX512F-NEXT:    vpsllw $2, %ymm1, %ymm3
-; AVX512F-NEXT:    vpand %ymm8, %ymm3, %ymm3
-; AVX512F-NEXT:    vpor %ymm2, %ymm3, %ymm2
-; AVX512F-NEXT:    vpblendvb %ymm7, %ymm2, %ymm1, %ymm1
-; AVX512F-NEXT:    vpsrlw $7, %ymm1, %ymm2
-; AVX512F-NEXT:    vpand %ymm9, %ymm2, %ymm2
-; AVX512F-NEXT:    vpaddb %ymm1, %ymm1, %ymm3
-; AVX512F-NEXT:    vpor %ymm2, %ymm3, %ymm2
-; AVX512F-NEXT:    vpblendvb %ymm10, %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} xmm7 = [8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8]
+; AVX512F-NEXT:    vpsubb %xmm2, %xmm7, %xmm2
+; AVX512F-NEXT:    vpmovzxbq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,zero,zero,zero,zero,xmm2[1],zero,zero,zero,zero,zero,zero,zero
+; AVX512F-NEXT:    vpsrlw %xmm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsrlw %xmm2, %ymm5, %ymm5
+; AVX512F-NEXT:    vpsrlw $8, %ymm5, %ymm5
+; AVX512F-NEXT:    vpbroadcastb %xmm5, %ymm5
+; AVX512F-NEXT:    vpand %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpor %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpsllw %xmm3, %ymm1, %ymm3
+; AVX512F-NEXT:    vpand %ymm6, %ymm3, %ymm3
+; AVX512F-NEXT:    vpsrlw %xmm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm5, %ymm1, %ymm1
+; AVX512F-NEXT:    vpor %ymm1, %ymm3, %ymm1
 ; AVX512F-NEXT:    retq
 ;
 ; AVX512VL-LABEL: splatvar_funnnel_v64i8:
 ; AVX512VL:       # %bb.0:
 ; AVX512VL-NEXT:    vpbroadcastb %xmm2, %ymm2
-; AVX512VL-NEXT:    vpsrlw $4, %ymm0, %ymm3
-; AVX512VL-NEXT:    vmovdqa {{.*#+}} ymm4 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
-; AVX512VL-NEXT:    vpand %ymm4, %ymm3, %ymm3
-; AVX512VL-NEXT:    vpsllw $4, %ymm0, %ymm5
-; AVX512VL-NEXT:    vmovdqa {{.*#+}} ymm6 = [240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240]
-; AVX512VL-NEXT:    vpand %ymm6, %ymm5, %ymm5
-; AVX512VL-NEXT:    vpor %ymm3, %ymm5, %ymm3
-; AVX512VL-NEXT:    vpxor %xmm5, %xmm5, %xmm5
-; AVX512VL-NEXT:    vpsubb %ymm2, %ymm5, %ymm2
-; AVX512VL-NEXT:    vpand {{.*}}(%rip), %ymm2, %ymm2
-; AVX512VL-NEXT:    vpsllw $5, %ymm2, %ymm2
-; AVX512VL-NEXT:    vpblendvb %ymm2, %ymm3, %ymm0, %ymm0
-; AVX512VL-NEXT:    vpsrlw $6, %ymm0, %ymm3
-; AVX512VL-NEXT:    vmovdqa {{.*#+}} ymm5 = [3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3]
-; AVX512VL-NEXT:    vpand %ymm5, %ymm3, %ymm3
-; AVX512VL-NEXT:    vpsllw $2, %ymm0, %ymm7
-; AVX512VL-NEXT:    vmovdqa {{.*#+}} ymm8 = [252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252]
-; AVX512VL-NEXT:    vpand %ymm8, %ymm7, %ymm7
-; AVX512VL-NEXT:    vpor %ymm3, %ymm7, %ymm3
-; AVX512VL-NEXT:    vpaddb %ymm2, %ymm2, %ymm7
-; AVX512VL-NEXT:    vpblendvb %ymm7, %ymm3, %ymm0, %ymm0
-; AVX512VL-NEXT:    vpsrlw $7, %ymm0, %ymm3
-; AVX512VL-NEXT:    vmovdqa {{.*#+}} ymm9 = [1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1]
-; AVX512VL-NEXT:    vpand %ymm9, %ymm3, %ymm3
-; AVX512VL-NEXT:    vpaddb %ymm0, %ymm0, %ymm10
-; AVX512VL-NEXT:    vpor %ymm3, %ymm10, %ymm3
-; AVX512VL-NEXT:    vpaddb %ymm7, %ymm7, %ymm10
-; AVX512VL-NEXT:    vpblendvb %ymm10, %ymm3, %ymm0, %ymm0
-; AVX512VL-NEXT:    vpsrlw $4, %ymm1, %ymm3
-; AVX512VL-NEXT:    vpand %ymm4, %ymm3, %ymm3
-; AVX512VL-NEXT:    vpsllw $4, %ymm1, %ymm4
+; AVX512VL-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512VL-NEXT:    vpsubb %xmm2, %xmm3, %xmm2
+; AVX512VL-NEXT:    vpand {{.*}}(%rip), %xmm2, %xmm2
+; AVX512VL-NEXT:    vpmovzxbq {{.*#+}} xmm3 = xmm2[0],zero,zero,zero,zero,zero,zero,zero,xmm2[1],zero,zero,zero,zero,zero,zero,zero
+; AVX512VL-NEXT:    vpsllw %xmm3, %ymm0, %ymm4
+; AVX512VL-NEXT:    vpcmpeqd %ymm5, %ymm5, %ymm5
+; AVX512VL-NEXT:    vpsllw %xmm3, %ymm5, %ymm6
+; AVX512VL-NEXT:    vpbroadcastb %xmm6, %ymm6
 ; AVX512VL-NEXT:    vpand %ymm6, %ymm4, %ymm4
-; AVX512VL-NEXT:    vpor %ymm3, %ymm4, %ymm3
-; AVX512VL-NEXT:    vpblendvb %ymm2, %ymm3, %ymm1, %ymm1
-; AVX512VL-NEXT:    vpsrlw $6, %ymm1, %ymm2
-; AVX512VL-NEXT:    vpand %ymm5, %ymm2, %ymm2
-; AVX512VL-NEXT:    vpsllw $2, %ymm1, %ymm3
-; AVX512VL-NEXT:    vpand %ymm8, %ymm3, %ymm3
-; AVX512VL-NEXT:    vpor %ymm2, %ymm3, %ymm2
-; AVX512VL-NEXT:    vpblendvb %ymm7, %ymm2, %ymm1, %ymm1
-; AVX512VL-NEXT:    vpsrlw $7, %ymm1, %ymm2
-; AVX512VL-NEXT:    vpand %ymm9, %ymm2, %ymm2
-; AVX512VL-NEXT:    vpaddb %ymm1, %ymm1, %ymm3
-; AVX512VL-NEXT:    vpor %ymm2, %ymm3, %ymm2
-; AVX512VL-NEXT:    vpblendvb %ymm10, %ymm2, %ymm1, %ymm1
+; AVX512VL-NEXT:    vmovdqa {{.*#+}} xmm7 = [8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8]
+; AVX512VL-NEXT:    vpsubb %xmm2, %xmm7, %xmm2
+; AVX512VL-NEXT:    vpmovzxbq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,zero,zero,zero,zero,xmm2[1],zero,zero,zero,zero,zero,zero,zero
+; AVX512VL-NEXT:    vpsrlw %xmm2, %ymm0, %ymm0
+; AVX512VL-NEXT:    vpsrlw %xmm2, %ymm5, %ymm5
+; AVX512VL-NEXT:    vpsrlw $8, %ymm5, %ymm5
+; AVX512VL-NEXT:    vpbroadcastb %xmm5, %ymm5
+; AVX512VL-NEXT:    vpand %ymm5, %ymm0, %ymm0
+; AVX512VL-NEXT:    vpor %ymm0, %ymm4, %ymm0
+; AVX512VL-NEXT:    vpsllw %xmm3, %ymm1, %ymm3
+; AVX512VL-NEXT:    vpand %ymm6, %ymm3, %ymm3
+; AVX512VL-NEXT:    vpsrlw %xmm2, %ymm1, %ymm1
+; AVX512VL-NEXT:    vpand %ymm5, %ymm1, %ymm1
+; AVX512VL-NEXT:    vpor %ymm1, %ymm3, %ymm1
 ; AVX512VL-NEXT:    retq
 ;
 ; AVX512BW-LABEL: splatvar_funnnel_v64i8:

diff --git a/test/CodeGen/X86/vector-rotate-128.ll b/test/CodeGen/X86/vector-rotate-128.ll
index a1e0b621c9b44ca5b5e2d54c61bbcaf2ba9725c3..650fe3b616ce97860151bc282a7d0991f622a605 100644 (file)
--- a/test/CodeGen/X86/vector-rotate-128.ll
+++ b/test/CodeGen/X86/vector-rotate-128.ll
@@ -991,187 +991,133 @@ define <8 x i16> @splatvar_rotate_v8i16(<8 x i16> %a, <8 x i16> %b) nounwind {
 define <16 x i8> @splatvar_rotate_v16i8(<16 x i8> %a, <16 x i8> %b) nounwind {
 ; SSE2-LABEL: splatvar_rotate_v16i8:
 ; SSE2:       # %bb.0:
-; SSE2-NEXT:    movdqa %xmm0, %xmm2
 ; SSE2-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSE2-NEXT:    pshuflw {{.*#+}} xmm0 = xmm1[0,0,2,3,4,5,6,7]
-; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[0,0,0,0]
-; SSE2-NEXT:    movdqa %xmm2, %xmm0
-; SSE2-NEXT:    psrlw $4, %xmm0
-; SSE2-NEXT:    pand {{.*}}(%rip), %xmm0
-; SSE2-NEXT:    movdqa %xmm2, %xmm3
-; SSE2-NEXT:    psllw $4, %xmm3
-; SSE2-NEXT:    pand {{.*}}(%rip), %xmm3
-; SSE2-NEXT:    por %xmm0, %xmm3
-; SSE2-NEXT:    psllw $5, %xmm1
-; SSE2-NEXT:    pxor %xmm0, %xmm0
-; SSE2-NEXT:    pxor %xmm4, %xmm4
-; SSE2-NEXT:    pcmpgtb %xmm1, %xmm4
-; SSE2-NEXT:    pand %xmm4, %xmm3
-; SSE2-NEXT:    pandn %xmm2, %xmm4
-; SSE2-NEXT:    por %xmm3, %xmm4
-; SSE2-NEXT:    movdqa %xmm4, %xmm2
-; SSE2-NEXT:    psrlw $6, %xmm2
-; SSE2-NEXT:    pand {{.*}}(%rip), %xmm2
-; SSE2-NEXT:    movdqa %xmm4, %xmm3
-; SSE2-NEXT:    psllw $2, %xmm3
-; SSE2-NEXT:    pand {{.*}}(%rip), %xmm3
-; SSE2-NEXT:    por %xmm2, %xmm3
-; SSE2-NEXT:    paddb %xmm1, %xmm1
-; SSE2-NEXT:    pxor %xmm2, %xmm2
-; SSE2-NEXT:    pcmpgtb %xmm1, %xmm2
-; SSE2-NEXT:    pand %xmm2, %xmm3
-; SSE2-NEXT:    pandn %xmm4, %xmm2
-; SSE2-NEXT:    por %xmm3, %xmm2
-; SSE2-NEXT:    movdqa %xmm2, %xmm3
-; SSE2-NEXT:    paddb %xmm2, %xmm3
-; SSE2-NEXT:    movdqa %xmm2, %xmm4
-; SSE2-NEXT:    psrlw $7, %xmm4
-; SSE2-NEXT:    pand {{.*}}(%rip), %xmm4
-; SSE2-NEXT:    por %xmm3, %xmm4
-; SSE2-NEXT:    paddb %xmm1, %xmm1
-; SSE2-NEXT:    pcmpgtb %xmm1, %xmm0
-; SSE2-NEXT:    pand %xmm0, %xmm4
-; SSE2-NEXT:    pandn %xmm2, %xmm0
-; SSE2-NEXT:    por %xmm4, %xmm0
+; SSE2-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,0,2,3,4,5,6,7]
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,0,0,0]
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    movdqa {{.*#+}} xmm2 = [8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8]
+; SSE2-NEXT:    psubb %xmm1, %xmm2
+; SSE2-NEXT:    movdqa %xmm1, %xmm3
+; SSE2-NEXT:    pslldq {{.*#+}} xmm3 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm3[0]
+; SSE2-NEXT:    psrldq {{.*#+}} xmm3 = xmm3[15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psllw %xmm3, %xmm1
+; SSE2-NEXT:    pcmpeqd %xmm4, %xmm4
+; SSE2-NEXT:    pcmpeqd %xmm5, %xmm5
+; SSE2-NEXT:    psllw %xmm3, %xmm5
+; SSE2-NEXT:    punpcklbw {{.*#+}} xmm5 = xmm5[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
+; SSE2-NEXT:    pshuflw {{.*#+}} xmm3 = xmm5[0,0,2,3,4,5,6,7]
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,0,0,0]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pslldq {{.*#+}} xmm2 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm2[0]
+; SSE2-NEXT:    psrldq {{.*#+}} xmm2 = xmm2[15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; SSE2-NEXT:    psrlw %xmm2, %xmm0
+; SSE2-NEXT:    psrlw %xmm2, %xmm4
+; SSE2-NEXT:    psrlw $8, %xmm4
+; SSE2-NEXT:    punpcklbw {{.*#+}} xmm4 = xmm4[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
+; SSE2-NEXT:    pshuflw {{.*#+}} xmm2 = xmm4[0,0,2,3,4,5,6,7]
+; SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,0,0,0]
+; SSE2-NEXT:    pand %xmm0, %xmm2
+; SSE2-NEXT:    por %xmm2, %xmm1
+; SSE2-NEXT:    movdqa %xmm1, %xmm0
 ; SSE2-NEXT:    retq
 ;
 ; SSE41-LABEL: splatvar_rotate_v16i8:
 ; SSE41:       # %bb.0:
-; SSE41-NEXT:    movdqa %xmm1, %xmm2
-; SSE41-NEXT:    movdqa %xmm0, %xmm1
-; SSE41-NEXT:    pxor %xmm0, %xmm0
-; SSE41-NEXT:    pshufb %xmm0, %xmm2
-; SSE41-NEXT:    movdqa %xmm1, %xmm0
-; SSE41-NEXT:    psrlw $4, %xmm0
-; SSE41-NEXT:    pand {{.*}}(%rip), %xmm0
-; SSE41-NEXT:    movdqa %xmm1, %xmm3
-; SSE41-NEXT:    psllw $4, %xmm3
-; SSE41-NEXT:    pand {{.*}}(%rip), %xmm3
-; SSE41-NEXT:    por %xmm0, %xmm3
-; SSE41-NEXT:    psllw $5, %xmm2
-; SSE41-NEXT:    movdqa %xmm2, %xmm0
-; SSE41-NEXT:    pblendvb %xmm0, %xmm3, %xmm1
-; SSE41-NEXT:    movdqa %xmm1, %xmm0
-; SSE41-NEXT:    psrlw $6, %xmm0
-; SSE41-NEXT:    pand {{.*}}(%rip), %xmm0
-; SSE41-NEXT:    movdqa %xmm1, %xmm3
-; SSE41-NEXT:    psllw $2, %xmm3
-; SSE41-NEXT:    pand {{.*}}(%rip), %xmm3
-; SSE41-NEXT:    por %xmm0, %xmm3
-; SSE41-NEXT:    paddb %xmm2, %xmm2
-; SSE41-NEXT:    movdqa %xmm2, %xmm0
-; SSE41-NEXT:    pblendvb %xmm0, %xmm3, %xmm1
-; SSE41-NEXT:    movdqa %xmm1, %xmm0
-; SSE41-NEXT:    paddb %xmm1, %xmm0
-; SSE41-NEXT:    movdqa %xmm1, %xmm3
-; SSE41-NEXT:    psrlw $7, %xmm3
-; SSE41-NEXT:    pand {{.*}}(%rip), %xmm3
-; SSE41-NEXT:    por %xmm0, %xmm3
-; SSE41-NEXT:    paddb %xmm2, %xmm2
+; SSE41-NEXT:    pxor %xmm3, %xmm3
+; SSE41-NEXT:    pshufb %xmm3, %xmm1
+; SSE41-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE41-NEXT:    pmovzxbq {{.*#+}} xmm4 = xmm1[0],zero,zero,zero,zero,zero,zero,zero,xmm1[1],zero,zero,zero,zero,zero,zero,zero
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    psllw %xmm4, %xmm2
+; SSE41-NEXT:    pcmpeqd %xmm5, %xmm5
+; SSE41-NEXT:    pcmpeqd %xmm6, %xmm6
+; SSE41-NEXT:    psllw %xmm4, %xmm6
+; SSE41-NEXT:    pshufb %xmm3, %xmm6
+; SSE41-NEXT:    pand %xmm6, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8]
+; SSE41-NEXT:    psubb %xmm1, %xmm3
+; SSE41-NEXT:    pmovzxbq {{.*#+}} xmm1 = xmm3[0],zero,zero,zero,zero,zero,zero,zero,xmm3[1],zero,zero,zero,zero,zero,zero,zero
+; SSE41-NEXT:    psrlw %xmm1, %xmm0
+; SSE41-NEXT:    psrlw %xmm1, %xmm5
+; SSE41-NEXT:    pshufb {{.*#+}} xmm5 = xmm5[1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1]
+; SSE41-NEXT:    pand %xmm0, %xmm5
+; SSE41-NEXT:    por %xmm5, %xmm2
 ; SSE41-NEXT:    movdqa %xmm2, %xmm0
-; SSE41-NEXT:    pblendvb %xmm0, %xmm3, %xmm1
-; SSE41-NEXT:    movdqa %xmm1, %xmm0
 ; SSE41-NEXT:    retq
 ;
 ; AVX1-LABEL: splatvar_rotate_v16i8:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vpxor %xmm2, %xmm2, %xmm2
 ; AVX1-NEXT:    vpshufb %xmm2, %xmm1, %xmm1
-; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm2
-; AVX1-NEXT:    vpand {{.*}}(%rip), %xmm2, %xmm2
-; AVX1-NEXT:    vpsllw $4, %xmm0, %xmm3
-; AVX1-NEXT:    vpand {{.*}}(%rip), %xmm3, %xmm3
-; AVX1-NEXT:    vpor %xmm2, %xmm3, %xmm2
-; AVX1-NEXT:    vpsllw $5, %xmm1, %xmm1
-; AVX1-NEXT:    vpblendvb %xmm1, %xmm2, %xmm0, %xmm0
-; AVX1-NEXT:    vpsrlw $6, %xmm0, %xmm2
-; AVX1-NEXT:    vpand {{.*}}(%rip), %xmm2, %xmm2
-; AVX1-NEXT:    vpsllw $2, %xmm0, %xmm3
-; AVX1-NEXT:    vpand {{.*}}(%rip), %xmm3, %xmm3
-; AVX1-NEXT:    vpor %xmm2, %xmm3, %xmm2
-; AVX1-NEXT:    vpaddb %xmm1, %xmm1, %xmm1
-; AVX1-NEXT:    vpblendvb %xmm1, %xmm2, %xmm0, %xmm0
-; AVX1-NEXT:    vpaddb %xmm0, %xmm0, %xmm2
-; AVX1-NEXT:    vpsrlw $7, %xmm0, %xmm3
-; AVX1-NEXT:    vpand {{.*}}(%rip), %xmm3, %xmm3
-; AVX1-NEXT:    vpor %xmm3, %xmm2, %xmm2
-; AVX1-NEXT:    vpaddb %xmm1, %xmm1, %xmm1
-; AVX1-NEXT:    vpblendvb %xmm1, %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpand {{.*}}(%rip), %xmm1, %xmm1
+; AVX1-NEXT:    vpmovzxbq {{.*#+}} xmm3 = xmm1[0],zero,zero,zero,zero,zero,zero,zero,xmm1[1],zero,zero,zero,zero,zero,zero,zero
+; AVX1-NEXT:    vpsllw %xmm3, %xmm0, %xmm4
+; AVX1-NEXT:    vpcmpeqd %xmm5, %xmm5, %xmm5
+; AVX1-NEXT:    vpsllw %xmm3, %xmm5, %xmm3
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpand %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8]
+; AVX1-NEXT:    vpsubb %xmm1, %xmm3, %xmm1
+; AVX1-NEXT:    vpmovzxbq {{.*#+}} xmm1 = xmm1[0],zero,zero,zero,zero,zero,zero,zero,xmm1[1],zero,zero,zero,zero,zero,zero,zero
+; AVX1-NEXT:    vpsrlw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpsrlw %xmm1, %xmm5, %xmm1
+; AVX1-NEXT:    vpshufb {{.*#+}} xmm1 = xmm1[1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpor %xmm0, %xmm2, %xmm0
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: splatvar_rotate_v16i8:
 ; AVX2:       # %bb.0:
 ; AVX2-NEXT:    vpbroadcastb %xmm1, %xmm1
-; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm2
-; AVX2-NEXT:    vpand {{.*}}(%rip), %xmm2, %xmm2
-; AVX2-NEXT:    vpsllw $4, %xmm0, %xmm3
-; AVX2-NEXT:    vpand {{.*}}(%rip), %xmm3, %xmm3
-; AVX2-NEXT:    vpor %xmm2, %xmm3, %xmm2
-; AVX2-NEXT:    vpsllw $5, %xmm1, %xmm1
-; AVX2-NEXT:    vpblendvb %xmm1, %xmm2, %xmm0, %xmm0
-; AVX2-NEXT:    vpsrlw $6, %xmm0, %xmm2
-; AVX2-NEXT:    vpand {{.*}}(%rip), %xmm2, %xmm2
-; AVX2-NEXT:    vpsllw $2, %xmm0, %xmm3
-; AVX2-NEXT:    vpand {{.*}}(%rip), %xmm3, %xmm3
-; AVX2-NEXT:    vpor %xmm2, %xmm3, %xmm2
-; AVX2-NEXT:    vpaddb %xmm1, %xmm1, %xmm1
-; AVX2-NEXT:    vpblendvb %xmm1, %xmm2, %xmm0, %xmm0
-; AVX2-NEXT:    vpaddb %xmm0, %xmm0, %xmm2
-; AVX2-NEXT:    vpsrlw $7, %xmm0, %xmm3
-; AVX2-NEXT:    vpand {{.*}}(%rip), %xmm3, %xmm3
-; AVX2-NEXT:    vpor %xmm3, %xmm2, %xmm2
-; AVX2-NEXT:    vpaddb %xmm1, %xmm1, %xmm1
-; AVX2-NEXT:    vpblendvb %xmm1, %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpand {{.*}}(%rip), %xmm1, %xmm1
+; AVX2-NEXT:    vpmovzxbq {{.*#+}} xmm2 = xmm1[0],zero,zero,zero,zero,zero,zero,zero,xmm1[1],zero,zero,zero,zero,zero,zero,zero
+; AVX2-NEXT:    vpsllw %xmm2, %xmm0, %xmm3
+; AVX2-NEXT:    vpcmpeqd %xmm4, %xmm4, %xmm4
+; AVX2-NEXT:    vpsllw %xmm2, %xmm4, %xmm2
+; AVX2-NEXT:    vpbroadcastb %xmm2, %xmm2
+; AVX2-NEXT:    vpand %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8]
+; AVX2-NEXT:    vpsubb %xmm1, %xmm3, %xmm1
+; AVX2-NEXT:    vpmovzxbq {{.*#+}} xmm1 = xmm1[0],zero,zero,zero,zero,zero,zero,zero,xmm1[1],zero,zero,zero,zero,zero,zero,zero
+; AVX2-NEXT:    vpsrlw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpsrlw %xmm1, %xmm4, %xmm1
+; AVX2-NEXT:    vpsrlw $8, %xmm1, %xmm1
+; AVX2-NEXT:    vpbroadcastb %xmm1, %xmm1
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpor %xmm0, %xmm2, %xmm0
 ; AVX2-NEXT:    retq
 ;
 ; AVX512F-LABEL: splatvar_rotate_v16i8:
 ; AVX512F:       # %bb.0:
 ; AVX512F-NEXT:    vpbroadcastb %xmm1, %xmm1
-; AVX512F-NEXT:    vpsrlw $4, %xmm0, %xmm2
-; AVX512F-NEXT:    vpand {{.*}}(%rip), %xmm2, %xmm2
-; AVX512F-NEXT:    vpsllw $4, %xmm0, %xmm3
-; AVX512F-NEXT:    vpand {{.*}}(%rip), %xmm3, %xmm3
-; AVX512F-NEXT:    vpor %xmm2, %xmm3, %xmm2
-; AVX512F-NEXT:    vpsllw $5, %xmm1, %xmm1
-; AVX512F-NEXT:    vpblendvb %xmm1, %xmm2, %xmm0, %xmm0
-; AVX512F-NEXT:    vpsrlw $6, %xmm0, %xmm2
-; AVX512F-NEXT:    vpand {{.*}}(%rip), %xmm2, %xmm2
-; AVX512F-NEXT:    vpsllw $2, %xmm0, %xmm3
-; AVX512F-NEXT:    vpand {{.*}}(%rip), %xmm3, %xmm3
-; AVX512F-NEXT:    vpor %xmm2, %xmm3, %xmm2
-; AVX512F-NEXT:    vpaddb %xmm1, %xmm1, %xmm1
-; AVX512F-NEXT:    vpblendvb %xmm1, %xmm2, %xmm0, %xmm0
-; AVX512F-NEXT:    vpaddb %xmm0, %xmm0, %xmm2
-; AVX512F-NEXT:    vpsrlw $7, %xmm0, %xmm3
-; AVX512F-NEXT:    vpand {{.*}}(%rip), %xmm3, %xmm3
-; AVX512F-NEXT:    vpor %xmm3, %xmm2, %xmm2
-; AVX512F-NEXT:    vpaddb %xmm1, %xmm1, %xmm1
-; AVX512F-NEXT:    vpblendvb %xmm1, %xmm2, %xmm0, %xmm0
+; AVX512F-NEXT:    vpmovzxbd {{.*#+}} zmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero,xmm0[4],zero,zero,zero,xmm0[5],zero,zero,zero,xmm0[6],zero,zero,zero,xmm0[7],zero,zero,zero,xmm0[8],zero,zero,zero,xmm0[9],zero,zero,zero,xmm0[10],zero,zero,zero,xmm0[11],zero,zero,zero,xmm0[12],zero,zero,zero,xmm0[13],zero,zero,zero,xmm0[14],zero,zero,zero,xmm0[15],zero,zero,zero
+; AVX512F-NEXT:    vpand {{.*}}(%rip), %xmm1, %xmm1
+; AVX512F-NEXT:    vpmovzxbd {{.*#+}} zmm2 = xmm1[0],zero,zero,zero,xmm1[1],zero,zero,zero,xmm1[2],zero,zero,zero,xmm1[3],zero,zero,zero,xmm1[4],zero,zero,zero,xmm1[5],zero,zero,zero,xmm1[6],zero,zero,zero,xmm1[7],zero,zero,zero,xmm1[8],zero,zero,zero,xmm1[9],zero,zero,zero,xmm1[10],zero,zero,zero,xmm1[11],zero,zero,zero,xmm1[12],zero,zero,zero,xmm1[13],zero,zero,zero,xmm1[14],zero,zero,zero,xmm1[15],zero,zero,zero
+; AVX512F-NEXT:    vpsllvd %zmm2, %zmm0, %zmm2
+; AVX512F-NEXT:    vmovdqa {{.*#+}} xmm3 = [8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8]
+; AVX512F-NEXT:    vpsubb %xmm1, %xmm3, %xmm1
+; AVX512F-NEXT:    vpmovzxbd {{.*#+}} zmm1 = xmm1[0],zero,zero,zero,xmm1[1],zero,zero,zero,xmm1[2],zero,zero,zero,xmm1[3],zero,zero,zero,xmm1[4],zero,zero,zero,xmm1[5],zero,zero,zero,xmm1[6],zero,zero,zero,xmm1[7],zero,zero,zero,xmm1[8],zero,zero,zero,xmm1[9],zero,zero,zero,xmm1[10],zero,zero,zero,xmm1[11],zero,zero,zero,xmm1[12],zero,zero,zero,xmm1[13],zero,zero,zero,xmm1[14],zero,zero,zero,xmm1[15],zero,zero,zero
+; AVX512F-NEXT:    vpsrlvd %zmm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpord %zmm0, %zmm2, %zmm0
+; AVX512F-NEXT:    vpmovdb %zmm0, %xmm0
+; AVX512F-NEXT:    vzeroupper
 ; AVX512F-NEXT:    retq
 ;
 ; AVX512VL-LABEL: splatvar_rotate_v16i8:
 ; AVX512VL:       # %bb.0:
 ; AVX512VL-NEXT:    vpbroadcastb %xmm1, %xmm1
-; AVX512VL-NEXT:    vpsrlw $4, %xmm0, %xmm2
-; AVX512VL-NEXT:    vpand {{.*}}(%rip), %xmm2, %xmm2
-; AVX512VL-NEXT:    vpsllw $4, %xmm0, %xmm3
-; AVX512VL-NEXT:    vpand {{.*}}(%rip), %xmm3, %xmm3
-; AVX512VL-NEXT:    vpor %xmm2, %xmm3, %xmm2
-; AVX512VL-NEXT:    vpsllw $5, %xmm1, %xmm1
-; AVX512VL-NEXT:    vpblendvb %xmm1, %xmm2, %xmm0, %xmm0
-; AVX512VL-NEXT:    vpsrlw $6, %xmm0, %xmm2
-; AVX512VL-NEXT:    vpand {{.*}}(%rip), %xmm2, %xmm2
-; AVX512VL-NEXT:    vpsllw $2, %xmm0, %xmm3
-; AVX512VL-NEXT:    vpand {{.*}}(%rip), %xmm3, %xmm3
-; AVX512VL-NEXT:    vpor %xmm2, %xmm3, %xmm2
-; AVX512VL-NEXT:    vpaddb %xmm1, %xmm1, %xmm1
-; AVX512VL-NEXT:    vpblendvb %xmm1, %xmm2, %xmm0, %xmm0
-; AVX512VL-NEXT:    vpaddb %xmm0, %xmm0, %xmm2
-; AVX512VL-NEXT:    vpsrlw $7, %xmm0, %xmm3
-; AVX512VL-NEXT:    vpand {{.*}}(%rip), %xmm3, %xmm3
-; AVX512VL-NEXT:    vpor %xmm3, %xmm2, %xmm2
-; AVX512VL-NEXT:    vpaddb %xmm1, %xmm1, %xmm1
-; AVX512VL-NEXT:    vpblendvb %xmm1, %xmm2, %xmm0, %xmm0
+; AVX512VL-NEXT:    vpmovzxbd {{.*#+}} zmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero,xmm0[4],zero,zero,zero,xmm0[5],zero,zero,zero,xmm0[6],zero,zero,zero,xmm0[7],zero,zero,zero,xmm0[8],zero,zero,zero,xmm0[9],zero,zero,zero,xmm0[10],zero,zero,zero,xmm0[11],zero,zero,zero,xmm0[12],zero,zero,zero,xmm0[13],zero,zero,zero,xmm0[14],zero,zero,zero,xmm0[15],zero,zero,zero
+; AVX512VL-NEXT:    vpand {{.*}}(%rip), %xmm1, %xmm1
+; AVX512VL-NEXT:    vpmovzxbd {{.*#+}} zmm2 = xmm1[0],zero,zero,zero,xmm1[1],zero,zero,zero,xmm1[2],zero,zero,zero,xmm1[3],zero,zero,zero,xmm1[4],zero,zero,zero,xmm1[5],zero,zero,zero,xmm1[6],zero,zero,zero,xmm1[7],zero,zero,zero,xmm1[8],zero,zero,zero,xmm1[9],zero,zero,zero,xmm1[10],zero,zero,zero,xmm1[11],zero,zero,zero,xmm1[12],zero,zero,zero,xmm1[13],zero,zero,zero,xmm1[14],zero,zero,zero,xmm1[15],zero,zero,zero
+; AVX512VL-NEXT:    vpsllvd %zmm2, %zmm0, %zmm2
+; AVX512VL-NEXT:    vmovdqa {{.*#+}} xmm3 = [8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8]
+; AVX512VL-NEXT:    vpsubb %xmm1, %xmm3, %xmm1
+; AVX512VL-NEXT:    vpmovzxbd {{.*#+}} zmm1 = xmm1[0],zero,zero,zero,xmm1[1],zero,zero,zero,xmm1[2],zero,zero,zero,xmm1[3],zero,zero,zero,xmm1[4],zero,zero,zero,xmm1[5],zero,zero,zero,xmm1[6],zero,zero,zero,xmm1[7],zero,zero,zero,xmm1[8],zero,zero,zero,xmm1[9],zero,zero,zero,xmm1[10],zero,zero,zero,xmm1[11],zero,zero,zero,xmm1[12],zero,zero,zero,xmm1[13],zero,zero,zero,xmm1[14],zero,zero,zero,xmm1[15],zero,zero,zero
+; AVX512VL-NEXT:    vpsrlvd %zmm1, %zmm0, %zmm0
+; AVX512VL-NEXT:    vpord %zmm0, %zmm2, %zmm0
+; AVX512VL-NEXT:    vpmovdb %zmm0, %xmm0
+; AVX512VL-NEXT:    vzeroupper
 ; AVX512VL-NEXT:    retq
 ;
 ; AVX512BW-LABEL: splatvar_rotate_v16i8:
@@ -1220,48 +1166,35 @@ define <16 x i8> @splatvar_rotate_v16i8(<16 x i8> %a, <16 x i8> %b) nounwind {
 ;
 ; X32-SSE-LABEL: splatvar_rotate_v16i8:
 ; X32-SSE:       # %bb.0:
-; X32-SSE-NEXT:    movdqa %xmm0, %xmm2
 ; X32-SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; X32-SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm1[0,0,2,3,4,5,6,7]
-; X32-SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[0,0,0,0]
-; X32-SSE-NEXT:    movdqa %xmm2, %xmm0
-; X32-SSE-NEXT:    psrlw $4, %xmm0
-; X32-SSE-NEXT:    pand {{\.LCPI.*}}, %xmm0
-; X32-SSE-NEXT:    movdqa %xmm2, %xmm3
-; X32-SSE-NEXT:    psllw $4, %xmm3
-; X32-SSE-NEXT:    pand {{\.LCPI.*}}, %xmm3
-; X32-SSE-NEXT:    por %xmm0, %xmm3
-; X32-SSE-NEXT:    psllw $5, %xmm1
-; X32-SSE-NEXT:    pxor %xmm0, %xmm0
-; X32-SSE-NEXT:    pxor %xmm4, %xmm4
-; X32-SSE-NEXT:    pcmpgtb %xmm1, %xmm4
-; X32-SSE-NEXT:    pand %xmm4, %xmm3
-; X32-SSE-NEXT:    pandn %xmm2, %xmm4
-; X32-SSE-NEXT:    por %xmm3, %xmm4
-; X32-SSE-NEXT:    movdqa %xmm4, %xmm2
-; X32-SSE-NEXT:    psrlw $6, %xmm2
-; X32-SSE-NEXT:    pand {{\.LCPI.*}}, %xmm2
-; X32-SSE-NEXT:    movdqa %xmm4, %xmm3
-; X32-SSE-NEXT:    psllw $2, %xmm3
-; X32-SSE-NEXT:    pand {{\.LCPI.*}}, %xmm3
-; X32-SSE-NEXT:    por %xmm2, %xmm3
-; X32-SSE-NEXT:    paddb %xmm1, %xmm1
-; X32-SSE-NEXT:    pxor %xmm2, %xmm2
-; X32-SSE-NEXT:    pcmpgtb %xmm1, %xmm2
-; X32-SSE-NEXT:    pand %xmm2, %xmm3
-; X32-SSE-NEXT:    pandn %xmm4, %xmm2
-; X32-SSE-NEXT:    por %xmm3, %xmm2
-; X32-SSE-NEXT:    movdqa %xmm2, %xmm3
-; X32-SSE-NEXT:    paddb %xmm2, %xmm3
-; X32-SSE-NEXT:    movdqa %xmm2, %xmm4
-; X32-SSE-NEXT:    psrlw $7, %xmm4
-; X32-SSE-NEXT:    pand {{\.LCPI.*}}, %xmm4
-; X32-SSE-NEXT:    por %xmm3, %xmm4
-; X32-SSE-NEXT:    paddb %xmm1, %xmm1
-; X32-SSE-NEXT:    pcmpgtb %xmm1, %xmm0
-; X32-SSE-NEXT:    pand %xmm0, %xmm4
-; X32-SSE-NEXT:    pandn %xmm2, %xmm0
-; X32-SSE-NEXT:    por %xmm4, %xmm0
+; X32-SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,0,2,3,4,5,6,7]
+; X32-SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,0,0,0]
+; X32-SSE-NEXT:    pand {{\.LCPI.*}}, %xmm1
+; X32-SSE-NEXT:    movdqa {{.*#+}} xmm2 = [8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8]
+; X32-SSE-NEXT:    psubb %xmm1, %xmm2
+; X32-SSE-NEXT:    movdqa %xmm1, %xmm3
+; X32-SSE-NEXT:    pslldq {{.*#+}} xmm3 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm3[0]
+; X32-SSE-NEXT:    psrldq {{.*#+}} xmm3 = xmm3[15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X32-SSE-NEXT:    movdqa %xmm0, %xmm1
+; X32-SSE-NEXT:    psllw %xmm3, %xmm1
+; X32-SSE-NEXT:    pcmpeqd %xmm4, %xmm4
+; X32-SSE-NEXT:    pcmpeqd %xmm5, %xmm5
+; X32-SSE-NEXT:    psllw %xmm3, %xmm5
+; X32-SSE-NEXT:    punpcklbw {{.*#+}} xmm5 = xmm5[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
+; X32-SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm5[0,0,2,3,4,5,6,7]
+; X32-SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,0,0,0]
+; X32-SSE-NEXT:    pand %xmm3, %xmm1
+; X32-SSE-NEXT:    pslldq {{.*#+}} xmm2 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm2[0]
+; X32-SSE-NEXT:    psrldq {{.*#+}} xmm2 = xmm2[15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X32-SSE-NEXT:    psrlw %xmm2, %xmm0
+; X32-SSE-NEXT:    psrlw %xmm2, %xmm4
+; X32-SSE-NEXT:    psrlw $8, %xmm4
+; X32-SSE-NEXT:    punpcklbw {{.*#+}} xmm4 = xmm4[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
+; X32-SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm4[0,0,2,3,4,5,6,7]
+; X32-SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,0,0,0]
+; X32-SSE-NEXT:    pand %xmm0, %xmm2
+; X32-SSE-NEXT:    por %xmm2, %xmm1
+; X32-SSE-NEXT:    movdqa %xmm1, %xmm0
 ; X32-SSE-NEXT:    retl
   %splat = shufflevector <16 x i8> %b, <16 x i8> undef, <16 x i32> zeroinitializer
   %splat8 = sub <16 x i8> <i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8>, %splat

diff --git a/test/CodeGen/X86/vector-rotate-256.ll b/test/CodeGen/X86/vector-rotate-256.ll
index e9d15a08ef6a2ecdb27517dec7275ef3afa5eb8e..49dd2d2661d09b65064233e2b3cd198d5225fea4 100644 (file)
--- a/test/CodeGen/X86/vector-rotate-256.ll
+++ b/test/CodeGen/X86/vector-rotate-256.ll
@@ -755,125 +755,91 @@ define <32 x i8> @splatvar_rotate_v32i8(<32 x i8> %a, <32 x i8> %b) nounwind {
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vpxor %xmm2, %xmm2, %xmm2
 ; AVX1-NEXT:    vpshufb %xmm2, %xmm1, %xmm1
-; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm2
-; AVX1-NEXT:    vpsrlw $4, %xmm2, %xmm3
-; AVX1-NEXT:    vmovdqa {{.*#+}} xmm8 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
-; AVX1-NEXT:    vpand %xmm8, %xmm3, %xmm3
-; AVX1-NEXT:    vpsllw $4, %xmm2, %xmm5
-; AVX1-NEXT:    vmovdqa {{.*#+}} xmm9 = [240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240]
-; AVX1-NEXT:    vpand %xmm9, %xmm5, %xmm5
-; AVX1-NEXT:    vpor %xmm3, %xmm5, %xmm3
-; AVX1-NEXT:    vpsllw $5, %xmm1, %xmm1
-; AVX1-NEXT:    vpblendvb %xmm1, %xmm3, %xmm2, %xmm2
-; AVX1-NEXT:    vpsrlw $6, %xmm2, %xmm3
-; AVX1-NEXT:    vmovdqa {{.*#+}} xmm10 = [3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3]
-; AVX1-NEXT:    vpand %xmm10, %xmm3, %xmm3
-; AVX1-NEXT:    vpsllw $2, %xmm2, %xmm7
-; AVX1-NEXT:    vmovdqa {{.*#+}} xmm11 = [252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252]
-; AVX1-NEXT:    vpand %xmm11, %xmm7, %xmm7
-; AVX1-NEXT:    vpor %xmm3, %xmm7, %xmm3
-; AVX1-NEXT:    vpaddb %xmm1, %xmm1, %xmm7
-; AVX1-NEXT:    vpblendvb %xmm7, %xmm3, %xmm2, %xmm2
-; AVX1-NEXT:    vpsrlw $7, %xmm2, %xmm3
-; AVX1-NEXT:    vmovdqa {{.*#+}} xmm6 = [1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1]
-; AVX1-NEXT:    vpand %xmm6, %xmm3, %xmm3
-; AVX1-NEXT:    vpaddb %xmm2, %xmm2, %xmm5
-; AVX1-NEXT:    vpor %xmm3, %xmm5, %xmm3
-; AVX1-NEXT:    vpaddb %xmm7, %xmm7, %xmm5
-; AVX1-NEXT:    vpblendvb %xmm5, %xmm3, %xmm2, %xmm2
-; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm3
-; AVX1-NEXT:    vpand %xmm8, %xmm3, %xmm3
-; AVX1-NEXT:    vpsllw $4, %xmm0, %xmm4
-; AVX1-NEXT:    vpand %xmm9, %xmm4, %xmm4
-; AVX1-NEXT:    vpor %xmm3, %xmm4, %xmm3
-; AVX1-NEXT:    vpblendvb %xmm1, %xmm3, %xmm0, %xmm0
-; AVX1-NEXT:    vpsrlw $6, %xmm0, %xmm1
-; AVX1-NEXT:    vpand %xmm10, %xmm1, %xmm1
-; AVX1-NEXT:    vpsllw $2, %xmm0, %xmm3
-; AVX1-NEXT:    vpand %xmm11, %xmm3, %xmm3
-; AVX1-NEXT:    vpor %xmm1, %xmm3, %xmm1
-; AVX1-NEXT:    vpblendvb %xmm7, %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vpsrlw $7, %xmm0, %xmm1
-; AVX1-NEXT:    vpand %xmm6, %xmm1, %xmm1
-; AVX1-NEXT:    vpaddb %xmm0, %xmm0, %xmm3
-; AVX1-NEXT:    vpor %xmm1, %xmm3, %xmm1
-; AVX1-NEXT:    vpblendvb %xmm5, %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
+; AVX1-NEXT:    vpand {{.*}}(%rip), %xmm1, %xmm1
+; AVX1-NEXT:    vpmovzxbq {{.*#+}} xmm3 = xmm1[0],zero,zero,zero,zero,zero,zero,zero,xmm1[1],zero,zero,zero,zero,zero,zero,zero
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm4
+; AVX1-NEXT:    vpsllw %xmm3, %xmm4, %xmm5
+; AVX1-NEXT:    vpcmpeqd %xmm6, %xmm6, %xmm6
+; AVX1-NEXT:    vpsllw %xmm3, %xmm6, %xmm7
+; AVX1-NEXT:    vpshufb %xmm2, %xmm7, %xmm2
+; AVX1-NEXT:    vpand %xmm2, %xmm5, %xmm5
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm7 = [8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8]
+; AVX1-NEXT:    vpsubb %xmm1, %xmm7, %xmm1
+; AVX1-NEXT:    vpmovzxbq {{.*#+}} xmm1 = xmm1[0],zero,zero,zero,zero,zero,zero,zero,xmm1[1],zero,zero,zero,zero,zero,zero,zero
+; AVX1-NEXT:    vpsrlw %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpsrlw %xmm1, %xmm6, %xmm6
+; AVX1-NEXT:    vpshufb {{.*#+}} xmm6 = xmm6[1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1]
+; AVX1-NEXT:    vpand %xmm6, %xmm4, %xmm4
+; AVX1-NEXT:    vpor %xmm4, %xmm5, %xmm4
+; AVX1-NEXT:    vpsllw %xmm3, %xmm0, %xmm3
+; AVX1-NEXT:    vpand %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm6, %xmm0, %xmm0
+; AVX1-NEXT:    vpor %xmm0, %xmm2, %xmm0
+; AVX1-NEXT:    vinsertf128 $1, %xmm4, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: splatvar_rotate_v32i8:
 ; AVX2:       # %bb.0:
 ; AVX2-NEXT:    vpbroadcastb %xmm1, %ymm1
-; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm2
-; AVX2-NEXT:    vpand {{.*}}(%rip), %ymm2, %ymm2
-; AVX2-NEXT:    vpsllw $4, %ymm0, %ymm3
-; AVX2-NEXT:    vpand {{.*}}(%rip), %ymm3, %ymm3
-; AVX2-NEXT:    vpor %ymm2, %ymm3, %ymm2
-; AVX2-NEXT:    vpsllw $5, %ymm1, %ymm1
-; AVX2-NEXT:    vpblendvb %ymm1, %ymm2, %ymm0, %ymm0
-; AVX2-NEXT:    vpsrlw $6, %ymm0, %ymm2
-; AVX2-NEXT:    vpand {{.*}}(%rip), %ymm2, %ymm2
-; AVX2-NEXT:    vpsllw $2, %ymm0, %ymm3
-; AVX2-NEXT:    vpand {{.*}}(%rip), %ymm3, %ymm3
-; AVX2-NEXT:    vpor %ymm2, %ymm3, %ymm2
-; AVX2-NEXT:    vpaddb %ymm1, %ymm1, %ymm1
-; AVX2-NEXT:    vpblendvb %ymm1, %ymm2, %ymm0, %ymm0
-; AVX2-NEXT:    vpaddb %ymm0, %ymm0, %ymm2
-; AVX2-NEXT:    vpsrlw $7, %ymm0, %ymm3
-; AVX2-NEXT:    vpand {{.*}}(%rip), %ymm3, %ymm3
-; AVX2-NEXT:    vpor %ymm3, %ymm2, %ymm2
-; AVX2-NEXT:    vpaddb %ymm1, %ymm1, %ymm1
-; AVX2-NEXT:    vpblendvb %ymm1, %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpand {{.*}}(%rip), %xmm1, %xmm1
+; AVX2-NEXT:    vpmovzxbq {{.*#+}} xmm2 = xmm1[0],zero,zero,zero,zero,zero,zero,zero,xmm1[1],zero,zero,zero,zero,zero,zero,zero
+; AVX2-NEXT:    vpsllw %xmm2, %ymm0, %ymm3
+; AVX2-NEXT:    vpcmpeqd %ymm4, %ymm4, %ymm4
+; AVX2-NEXT:    vpsllw %xmm2, %ymm4, %ymm2
+; AVX2-NEXT:    vpbroadcastb %xmm2, %ymm2
+; AVX2-NEXT:    vpand %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8]
+; AVX2-NEXT:    vpsubb %xmm1, %xmm3, %xmm1
+; AVX2-NEXT:    vpmovzxbq {{.*#+}} xmm1 = xmm1[0],zero,zero,zero,zero,zero,zero,zero,xmm1[1],zero,zero,zero,zero,zero,zero,zero
+; AVX2-NEXT:    vpsrlw %xmm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpsrlw %xmm1, %ymm4, %ymm1
+; AVX2-NEXT:    vpsrlw $8, %ymm1, %ymm1
+; AVX2-NEXT:    vpbroadcastb %xmm1, %ymm1
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpor %ymm0, %ymm2, %ymm0
 ; AVX2-NEXT:    retq
 ;
 ; AVX512F-LABEL: splatvar_rotate_v32i8:
 ; AVX512F:       # %bb.0:
 ; AVX512F-NEXT:    vpbroadcastb %xmm1, %ymm1
-; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm2
-; AVX512F-NEXT:    vpand {{.*}}(%rip), %ymm2, %ymm2
-; AVX512F-NEXT:    vpsllw $4, %ymm0, %ymm3
-; AVX512F-NEXT:    vpand {{.*}}(%rip), %ymm3, %ymm3
-; AVX512F-NEXT:    vpor %ymm2, %ymm3, %ymm2
-; AVX512F-NEXT:    vpsllw $5, %ymm1, %ymm1
-; AVX512F-NEXT:    vpblendvb %ymm1, %ymm2, %ymm0, %ymm0
-; AVX512F-NEXT:    vpsrlw $6, %ymm0, %ymm2
-; AVX512F-NEXT:    vpand {{.*}}(%rip), %ymm2, %ymm2
-; AVX512F-NEXT:    vpsllw $2, %ymm0, %ymm3
-; AVX512F-NEXT:    vpand {{.*}}(%rip), %ymm3, %ymm3
-; AVX512F-NEXT:    vpor %ymm2, %ymm3, %ymm2
-; AVX512F-NEXT:    vpaddb %ymm1, %ymm1, %ymm1
-; AVX512F-NEXT:    vpblendvb %ymm1, %ymm2, %ymm0, %ymm0
-; AVX512F-NEXT:    vpaddb %ymm0, %ymm0, %ymm2
-; AVX512F-NEXT:    vpsrlw $7, %ymm0, %ymm3
-; AVX512F-NEXT:    vpand {{.*}}(%rip), %ymm3, %ymm3
-; AVX512F-NEXT:    vpor %ymm3, %ymm2, %ymm2
-; AVX512F-NEXT:    vpaddb %ymm1, %ymm1, %ymm1
-; AVX512F-NEXT:    vpblendvb %ymm1, %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand {{.*}}(%rip), %xmm1, %xmm1
+; AVX512F-NEXT:    vpmovzxbq {{.*#+}} xmm2 = xmm1[0],zero,zero,zero,zero,zero,zero,zero,xmm1[1],zero,zero,zero,zero,zero,zero,zero
+; AVX512F-NEXT:    vpsllw %xmm2, %ymm0, %ymm3
+; AVX512F-NEXT:    vpcmpeqd %ymm4, %ymm4, %ymm4
+; AVX512F-NEXT:    vpsllw %xmm2, %ymm4, %ymm2
+; AVX512F-NEXT:    vpbroadcastb %xmm2, %ymm2
+; AVX512F-NEXT:    vpand %ymm2, %ymm3, %ymm2
+; AVX512F-NEXT:    vmovdqa {{.*#+}} xmm3 = [8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8]
+; AVX512F-NEXT:    vpsubb %xmm1, %xmm3, %xmm1
+; AVX512F-NEXT:    vpmovzxbq {{.*#+}} xmm1 = xmm1[0],zero,zero,zero,zero,zero,zero,zero,xmm1[1],zero,zero,zero,zero,zero,zero,zero
+; AVX512F-NEXT:    vpsrlw %xmm1, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsrlw %xmm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpsrlw $8, %ymm1, %ymm1
+; AVX512F-NEXT:    vpbroadcastb %xmm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX512F-NEXT:    vpor %ymm0, %ymm2, %ymm0
 ; AVX512F-NEXT:    retq
 ;
 ; AVX512VL-LABEL: splatvar_rotate_v32i8:
 ; AVX512VL:       # %bb.0:
 ; AVX512VL-NEXT:    vpbroadcastb %xmm1, %ymm1
-; AVX512VL-NEXT:    vpsrlw $4, %ymm0, %ymm2
-; AVX512VL-NEXT:    vpand {{.*}}(%rip), %ymm2, %ymm2
-; AVX512VL-NEXT:    vpsllw $4, %ymm0, %ymm3
-; AVX512VL-NEXT:    vpand {{.*}}(%rip), %ymm3, %ymm3
-; AVX512VL-NEXT:    vpor %ymm2, %ymm3, %ymm2
-; AVX512VL-NEXT:    vpsllw $5, %ymm1, %ymm1
-; AVX512VL-NEXT:    vpblendvb %ymm1, %ymm2, %ymm0, %ymm0
-; AVX512VL-NEXT:    vpsrlw $6, %ymm0, %ymm2
-; AVX512VL-NEXT:    vpand {{.*}}(%rip), %ymm2, %ymm2
-; AVX512VL-NEXT:    vpsllw $2, %ymm0, %ymm3
-; AVX512VL-NEXT:    vpand {{.*}}(%rip), %ymm3, %ymm3
-; AVX512VL-NEXT:    vpor %ymm2, %ymm3, %ymm2
-; AVX512VL-NEXT:    vpaddb %ymm1, %ymm1, %ymm1
-; AVX512VL-NEXT:    vpblendvb %ymm1, %ymm2, %ymm0, %ymm0
-; AVX512VL-NEXT:    vpaddb %ymm0, %ymm0, %ymm2
-; AVX512VL-NEXT:    vpsrlw $7, %ymm0, %ymm3
-; AVX512VL-NEXT:    vpand {{.*}}(%rip), %ymm3, %ymm3
-; AVX512VL-NEXT:    vpor %ymm3, %ymm2, %ymm2
-; AVX512VL-NEXT:    vpaddb %ymm1, %ymm1, %ymm1
-; AVX512VL-NEXT:    vpblendvb %ymm1, %ymm2, %ymm0, %ymm0
+; AVX512VL-NEXT:    vpand {{.*}}(%rip), %xmm1, %xmm1
+; AVX512VL-NEXT:    vpmovzxbq {{.*#+}} xmm2 = xmm1[0],zero,zero,zero,zero,zero,zero,zero,xmm1[1],zero,zero,zero,zero,zero,zero,zero
+; AVX512VL-NEXT:    vpsllw %xmm2, %ymm0, %ymm3
+; AVX512VL-NEXT:    vpcmpeqd %ymm4, %ymm4, %ymm4
+; AVX512VL-NEXT:    vpsllw %xmm2, %ymm4, %ymm2
+; AVX512VL-NEXT:    vpbroadcastb %xmm2, %ymm2
+; AVX512VL-NEXT:    vpand %ymm2, %ymm3, %ymm2
+; AVX512VL-NEXT:    vmovdqa {{.*#+}} xmm3 = [8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8]
+; AVX512VL-NEXT:    vpsubb %xmm1, %xmm3, %xmm1
+; AVX512VL-NEXT:    vpmovzxbq {{.*#+}} xmm1 = xmm1[0],zero,zero,zero,zero,zero,zero,zero,xmm1[1],zero,zero,zero,zero,zero,zero,zero
+; AVX512VL-NEXT:    vpsrlw %xmm1, %ymm0, %ymm0
+; AVX512VL-NEXT:    vpsrlw %xmm1, %ymm4, %ymm1
+; AVX512VL-NEXT:    vpsrlw $8, %ymm1, %ymm1
+; AVX512VL-NEXT:    vpbroadcastb %xmm1, %ymm1
+; AVX512VL-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX512VL-NEXT:    vpor %ymm0, %ymm2, %ymm0
 ; AVX512VL-NEXT:    retq
 ;
 ; AVX512BW-LABEL: splatvar_rotate_v32i8:

diff --git a/test/CodeGen/X86/vector-rotate-512.ll b/test/CodeGen/X86/vector-rotate-512.ll
index 4f2f6f127fc0dd8b290c8dbbc322b6d067adcf49..0ff6fc1bed1f6aba6147ba8c372e8a8a001bdf10 100644 (file)
--- a/test/CodeGen/X86/vector-rotate-512.ll
+++ b/test/CodeGen/X86/vector-rotate-512.ll
@@ -378,95 +378,53 @@ define <64 x i8> @splatvar_rotate_v64i8(<64 x i8> %a, <64 x i8> %b) nounwind {
 ; AVX512F-LABEL: splatvar_rotate_v64i8:
 ; AVX512F:       # %bb.0:
 ; AVX512F-NEXT:    vpbroadcastb %xmm2, %ymm2
-; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm3
-; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
-; AVX512F-NEXT:    vpand %ymm4, %ymm3, %ymm3
-; AVX512F-NEXT:    vpsllw $4, %ymm0, %ymm5
-; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm6 = [240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240]
-; AVX512F-NEXT:    vpand %ymm6, %ymm5, %ymm5
-; AVX512F-NEXT:    vpor %ymm3, %ymm5, %ymm3
-; AVX512F-NEXT:    vpsllw $5, %ymm2, %ymm2
-; AVX512F-NEXT:    vpblendvb %ymm2, %ymm3, %ymm0, %ymm0
-; AVX512F-NEXT:    vpsrlw $6, %ymm0, %ymm3
-; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm5 = [3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3]
-; AVX512F-NEXT:    vpand %ymm5, %ymm3, %ymm3
-; AVX512F-NEXT:    vpsllw $2, %ymm0, %ymm7
-; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm8 = [252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252]
-; AVX512F-NEXT:    vpand %ymm8, %ymm7, %ymm7
-; AVX512F-NEXT:    vpor %ymm3, %ymm7, %ymm3
-; AVX512F-NEXT:    vpaddb %ymm2, %ymm2, %ymm7
-; AVX512F-NEXT:    vpblendvb %ymm7, %ymm3, %ymm0, %ymm0
-; AVX512F-NEXT:    vpsrlw $7, %ymm0, %ymm3
-; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm9 = [1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1]
-; AVX512F-NEXT:    vpand %ymm9, %ymm3, %ymm3
-; AVX512F-NEXT:    vpaddb %ymm0, %ymm0, %ymm10
-; AVX512F-NEXT:    vpor %ymm3, %ymm10, %ymm3
-; AVX512F-NEXT:    vpaddb %ymm7, %ymm7, %ymm10
-; AVX512F-NEXT:    vpblendvb %ymm10, %ymm3, %ymm0, %ymm0
-; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm3
-; AVX512F-NEXT:    vpand %ymm4, %ymm3, %ymm3
-; AVX512F-NEXT:    vpsllw $4, %ymm1, %ymm4
+; AVX512F-NEXT:    vpand {{.*}}(%rip), %xmm2, %xmm2
+; AVX512F-NEXT:    vpmovzxbq {{.*#+}} xmm3 = xmm2[0],zero,zero,zero,zero,zero,zero,zero,xmm2[1],zero,zero,zero,zero,zero,zero,zero
+; AVX512F-NEXT:    vpsllw %xmm3, %ymm0, %ymm4
+; AVX512F-NEXT:    vpcmpeqd %ymm5, %ymm5, %ymm5
+; AVX512F-NEXT:    vpsllw %xmm3, %ymm5, %ymm6
+; AVX512F-NEXT:    vpbroadcastb %xmm6, %ymm6
 ; AVX512F-NEXT:    vpand %ymm6, %ymm4, %ymm4
-; AVX512F-NEXT:    vpor %ymm3, %ymm4, %ymm3
-; AVX512F-NEXT:    vpblendvb %ymm2, %ymm3, %ymm1, %ymm1
-; AVX512F-NEXT:    vpsrlw $6, %ymm1, %ymm2
-; AVX512F-NEXT:    vpand %ymm5, %ymm2, %ymm2
-; AVX512F-NEXT:    vpsllw $2, %ymm1, %ymm3
-; AVX512F-NEXT:    vpand %ymm8, %ymm3, %ymm3
-; AVX512F-NEXT:    vpor %ymm2, %ymm3, %ymm2
-; AVX512F-NEXT:    vpblendvb %ymm7, %ymm2, %ymm1, %ymm1
-; AVX512F-NEXT:    vpsrlw $7, %ymm1, %ymm2
-; AVX512F-NEXT:    vpand %ymm9, %ymm2, %ymm2
-; AVX512F-NEXT:    vpaddb %ymm1, %ymm1, %ymm3
-; AVX512F-NEXT:    vpor %ymm2, %ymm3, %ymm2
-; AVX512F-NEXT:    vpblendvb %ymm10, %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} xmm7 = [8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8]
+; AVX512F-NEXT:    vpsubb %xmm2, %xmm7, %xmm2
+; AVX512F-NEXT:    vpmovzxbq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,zero,zero,zero,zero,xmm2[1],zero,zero,zero,zero,zero,zero,zero
+; AVX512F-NEXT:    vpsrlw %xmm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsrlw %xmm2, %ymm5, %ymm5
+; AVX512F-NEXT:    vpsrlw $8, %ymm5, %ymm5
+; AVX512F-NEXT:    vpbroadcastb %xmm5, %ymm5
+; AVX512F-NEXT:    vpand %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpor %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpsllw %xmm3, %ymm1, %ymm3
+; AVX512F-NEXT:    vpand %ymm6, %ymm3, %ymm3
+; AVX512F-NEXT:    vpsrlw %xmm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm5, %ymm1, %ymm1
+; AVX512F-NEXT:    vpor %ymm1, %ymm3, %ymm1
 ; AVX512F-NEXT:    retq
 ;
 ; AVX512VL-LABEL: splatvar_rotate_v64i8:
 ; AVX512VL:       # %bb.0:
 ; AVX512VL-NEXT:    vpbroadcastb %xmm2, %ymm2
-; AVX512VL-NEXT:    vpsrlw $4, %ymm0, %ymm3
-; AVX512VL-NEXT:    vmovdqa {{.*#+}} ymm4 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
-; AVX512VL-NEXT:    vpand %ymm4, %ymm3, %ymm3
-; AVX512VL-NEXT:    vpsllw $4, %ymm0, %ymm5
-; AVX512VL-NEXT:    vmovdqa {{.*#+}} ymm6 = [240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240,240]
-; AVX512VL-NEXT:    vpand %ymm6, %ymm5, %ymm5
-; AVX512VL-NEXT:    vpor %ymm3, %ymm5, %ymm3
-; AVX512VL-NEXT:    vpsllw $5, %ymm2, %ymm2
-; AVX512VL-NEXT:    vpblendvb %ymm2, %ymm3, %ymm0, %ymm0
-; AVX512VL-NEXT:    vpsrlw $6, %ymm0, %ymm3
-; AVX512VL-NEXT:    vmovdqa {{.*#+}} ymm5 = [3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3]
-; AVX512VL-NEXT:    vpand %ymm5, %ymm3, %ymm3
-; AVX512VL-NEXT:    vpsllw $2, %ymm0, %ymm7
-; AVX512VL-NEXT:    vmovdqa {{.*#+}} ymm8 = [252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252,252]
-; AVX512VL-NEXT:    vpand %ymm8, %ymm7, %ymm7
-; AVX512VL-NEXT:    vpor %ymm3, %ymm7, %ymm3
-; AVX512VL-NEXT:    vpaddb %ymm2, %ymm2, %ymm7
-; AVX512VL-NEXT:    vpblendvb %ymm7, %ymm3, %ymm0, %ymm0
-; AVX512VL-NEXT:    vpsrlw $7, %ymm0, %ymm3
-; AVX512VL-NEXT:    vmovdqa {{.*#+}} ymm9 = [1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1]
-; AVX512VL-NEXT:    vpand %ymm9, %ymm3, %ymm3
-; AVX512VL-NEXT:    vpaddb %ymm0, %ymm0, %ymm10
-; AVX512VL-NEXT:    vpor %ymm3, %ymm10, %ymm3
-; AVX512VL-NEXT:    vpaddb %ymm7, %ymm7, %ymm10
-; AVX512VL-NEXT:    vpblendvb %ymm10, %ymm3, %ymm0, %ymm0
-; AVX512VL-NEXT:    vpsrlw $4, %ymm1, %ymm3
-; AVX512VL-NEXT:    vpand %ymm4, %ymm3, %ymm3
-; AVX512VL-NEXT:    vpsllw $4, %ymm1, %ymm4
+; AVX512VL-NEXT:    vpand {{.*}}(%rip), %xmm2, %xmm2
+; AVX512VL-NEXT:    vpmovzxbq {{.*#+}} xmm3 = xmm2[0],zero,zero,zero,zero,zero,zero,zero,xmm2[1],zero,zero,zero,zero,zero,zero,zero
+; AVX512VL-NEXT:    vpsllw %xmm3, %ymm0, %ymm4
+; AVX512VL-NEXT:    vpcmpeqd %ymm5, %ymm5, %ymm5
+; AVX512VL-NEXT:    vpsllw %xmm3, %ymm5, %ymm6
+; AVX512VL-NEXT:    vpbroadcastb %xmm6, %ymm6
 ; AVX512VL-NEXT:    vpand %ymm6, %ymm4, %ymm4
-; AVX512VL-NEXT:    vpor %ymm3, %ymm4, %ymm3
-; AVX512VL-NEXT:    vpblendvb %ymm2, %ymm3, %ymm1, %ymm1
-; AVX512VL-NEXT:    vpsrlw $6, %ymm1, %ymm2
-; AVX512VL-NEXT:    vpand %ymm5, %ymm2, %ymm2
-; AVX512VL-NEXT:    vpsllw $2, %ymm1, %ymm3
-; AVX512VL-NEXT:    vpand %ymm8, %ymm3, %ymm3
-; AVX512VL-NEXT:    vpor %ymm2, %ymm3, %ymm2
-; AVX512VL-NEXT:    vpblendvb %ymm7, %ymm2, %ymm1, %ymm1
-; AVX512VL-NEXT:    vpsrlw $7, %ymm1, %ymm2
-; AVX512VL-NEXT:    vpand %ymm9, %ymm2, %ymm2
-; AVX512VL-NEXT:    vpaddb %ymm1, %ymm1, %ymm3
-; AVX512VL-NEXT:    vpor %ymm2, %ymm3, %ymm2
-; AVX512VL-NEXT:    vpblendvb %ymm10, %ymm2, %ymm1, %ymm1
+; AVX512VL-NEXT:    vmovdqa {{.*#+}} xmm7 = [8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8]
+; AVX512VL-NEXT:    vpsubb %xmm2, %xmm7, %xmm2
+; AVX512VL-NEXT:    vpmovzxbq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,zero,zero,zero,zero,xmm2[1],zero,zero,zero,zero,zero,zero,zero
+; AVX512VL-NEXT:    vpsrlw %xmm2, %ymm0, %ymm0
+; AVX512VL-NEXT:    vpsrlw %xmm2, %ymm5, %ymm5
+; AVX512VL-NEXT:    vpsrlw $8, %ymm5, %ymm5
+; AVX512VL-NEXT:    vpbroadcastb %xmm5, %ymm5
+; AVX512VL-NEXT:    vpand %ymm5, %ymm0, %ymm0
+; AVX512VL-NEXT:    vpor %ymm0, %ymm4, %ymm0
+; AVX512VL-NEXT:    vpsllw %xmm3, %ymm1, %ymm3
+; AVX512VL-NEXT:    vpand %ymm6, %ymm3, %ymm3
+; AVX512VL-NEXT:    vpsrlw %xmm2, %ymm1, %ymm1
+; AVX512VL-NEXT:    vpand %ymm5, %ymm1, %ymm1
+; AVX512VL-NEXT:    vpor %ymm1, %ymm3, %ymm1
 ; AVX512VL-NEXT:    retq
 ;
 ; AVX512BW-LABEL: splatvar_rotate_v64i8: