[X86] Fix perf bug in `permilps` -> `shufd` in X86FixupInstTuning.

[goldstein.w.n]

Summary

We shouldn't do the transformation if we either have bypass delay OR
the new opcode has worse performance. Previous code was incorrectly
using AND.

[goldstein.w.n]

Rebase

[RKSimon]

LGTM

[goldstein.w.n]

Rebase

[goldstein.w.n]

Rebase

[goldstein.w.n]

Rebase

llvm/lib/Target/X86/X86FixupInstTuning.cpp

bool X86FixupInstTuningPass::processInstruction(
   };
 
   // `vpermilps m, i` -> `vpshufd m, i` iff no domain delay penalty on shuffles.
   // `vpshufd` is always as fast or faster than `vpermilps` and takes 1 less
   // byte of code size.
   auto ProcessVPERMILPSmi = [&](unsigned NewOpc) -> bool {
     // TODO: Might be work adding bypass delay if -Os/-Oz is enabled as
     // `vpshufd` saves a byte of code size.
-    if (!ST->hasNoDomainDelayShuffle() &&
+    if (!ST->hasNoDomainDelayShuffle() ||
         !NewOpcPreferable(NewOpc, /*ReplaceInTie*/ false))
       return false;
     MI.setDesc(TII->get(NewOpc));
     return true;
   };
 
   // `vunpcklpd/vmovlhps r, r` -> `vshufpd r, r, 0x00`
   // `vunpckhpd/vmovlhps r, r` -> `vshufpd r, r, 0xff`

llvm/test/CodeGen/X86/tuning-shuffle-permilps-avx512.ll

 ; CHECK-NEXT:    retq
   %mask = bitcast i4 %mask_int to <4 x i1>
   %shufp = shufflevector <4 x float> %a, <4 x float> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
   %res = select <4 x i1> %mask, <4 x float> %shufp, <4 x float> %b
   ret <4 x float> %res
 }
 
 define <16 x float> @transform_VPERMILPSZrm(ptr %ap) nounwind {
-; CHECK-ICX-NO-BYPASS-DELAY-LABEL: transform_VPERMILPSZrm:
-; CHECK-ICX-NO-BYPASS-DELAY:       # %bb.0:
-; CHECK-ICX-NO-BYPASS-DELAY-NEXT:    vpshufd {{.*#+}} zmm0 = mem[3,2,1,0,7,6,5,4,11,10,9,8,15,14,13,12]
-; CHECK-ICX-NO-BYPASS-DELAY-NEXT:    retq
-;
-; CHECK-ICX-BYPASS-DELAY-LABEL: transform_VPERMILPSZrm:
-; CHECK-ICX-BYPASS-DELAY:       # %bb.0:
-; CHECK-ICX-BYPASS-DELAY-NEXT:    vpermilps {{.*#+}} zmm0 = mem[3,2,1,0,7,6,5,4,11,10,9,8,15,14,13,12]
-; CHECK-ICX-BYPASS-DELAY-NEXT:    retq
-;
-; CHECK-V4-LABEL: transform_VPERMILPSZrm:
-; CHECK-V4:       # %bb.0:
-; CHECK-V4-NEXT:    vpermilps {{.*#+}} zmm0 = mem[3,2,1,0,7,6,5,4,11,10,9,8,15,14,13,12]
-; CHECK-V4-NEXT:    retq
-;
-; CHECK-AVX512-LABEL: transform_VPERMILPSZrm:
-; CHECK-AVX512:       # %bb.0:
-; CHECK-AVX512-NEXT:    vpermilps {{.*#+}} zmm0 = mem[3,2,1,0,7,6,5,4,11,10,9,8,15,14,13,12]
-; CHECK-AVX512-NEXT:    retq
-;
-; CHECK-ZNVER4-LABEL: transform_VPERMILPSZrm:
-; CHECK-ZNVER4:       # %bb.0:
-; CHECK-ZNVER4-NEXT:    vpermilps {{.*#+}} zmm0 = mem[3,2,1,0,7,6,5,4,11,10,9,8,15,14,13,12]
-; CHECK-ZNVER4-NEXT:    retq
+; CHECK-LABEL: transform_VPERMILPSZrm:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vpermilps {{.*#+}} zmm0 = mem[3,2,1,0,7,6,5,4,11,10,9,8,15,14,13,12]
+; CHECK-NEXT:    retq
   %a = load <16 x float>, ptr %ap
   %shufp = shufflevector <16 x float> %a, <16 x float> poison, <16 x i32> <i32 3, i32 2, i32 1, i32 0, i32 7, i32 6, i32 5, i32 4, i32 11, i32 10, i32 9, i32 8, i32 15, i32 14, i32 13, i32 12>
   ret <16 x float> %shufp
 }
 
 define <8 x float> @transform_VPERMILPSYrm(ptr %ap) nounwind {
-; CHECK-ICX-LABEL: transform_VPERMILPSYrm:
-; CHECK-ICX:       # %bb.0:
-; CHECK-ICX-NEXT:    vpshufd {{.*#+}} ymm0 = mem[3,2,1,0,7,6,5,4]
-; CHECK-ICX-NEXT:    retq
+; CHECK-ICX-NO-BYPASS-DELAY-LABEL: transform_VPERMILPSYrm:
+; CHECK-ICX-NO-BYPASS-DELAY:       # %bb.0:
+; CHECK-ICX-NO-BYPASS-DELAY-NEXT:    vpshufd {{.*#+}} ymm0 = mem[3,2,1,0,7,6,5,4]
+; CHECK-ICX-NO-BYPASS-DELAY-NEXT:    retq
+;
+; CHECK-ICX-BYPASS-DELAY-LABEL: transform_VPERMILPSYrm:
+; CHECK-ICX-BYPASS-DELAY:       # %bb.0:
+; CHECK-ICX-BYPASS-DELAY-NEXT:    vpermilps {{.*#+}} ymm0 = mem[3,2,1,0,7,6,5,4]
+; CHECK-ICX-BYPASS-DELAY-NEXT:    retq
 ;
 ; CHECK-V4-LABEL: transform_VPERMILPSYrm:
 ; CHECK-V4:       # %bb.0:
 ; CHECK-V4-NEXT:    vpermilps {{.*#+}} ymm0 = mem[3,2,1,0,7,6,5,4]
 ; CHECK-V4-NEXT:    retq
 ;
 ; CHECK-AVX512-LABEL: transform_VPERMILPSYrm:
 ; CHECK-AVX512:       # %bb.0:
 ; CHECK-AVX512-NEXT:    vpermilps {{.*#+}} ymm0 = mem[3,2,1,0,7,6,5,4]
 ; CHECK-AVX512-NEXT:    retq
 ;
 ; CHECK-ZNVER4-LABEL: transform_VPERMILPSYrm:
 ; CHECK-ZNVER4:       # %bb.0:
 ; CHECK-ZNVER4-NEXT:    vpermilps {{.*#+}} ymm0 = mem[3,2,1,0,7,6,5,4]
 ; CHECK-ZNVER4-NEXT:    retq
   %a = load <8 x float>, ptr %ap
   %shufp = shufflevector <8 x float> %a, <8 x float> poison, <8 x i32> <i32 3, i32 2, i32 1, i32 0, i32 7, i32 6, i32 5, i32 4>
   ret <8 x float> %shufp
 }
 
 define <4 x float> @transform_VPERMILPSrm(ptr %ap) nounwind {
-; CHECK-ICX-LABEL: transform_VPERMILPSrm:
-; CHECK-ICX:       # %bb.0:
-; CHECK-ICX-NEXT:    vpshufd {{.*#+}} xmm0 = mem[3,2,1,0]
-; CHECK-ICX-NEXT:    retq
+; CHECK-ICX-NO-BYPASS-DELAY-LABEL: transform_VPERMILPSrm:
+; CHECK-ICX-NO-BYPASS-DELAY:       # %bb.0:
+; CHECK-ICX-NO-BYPASS-DELAY-NEXT:    vpshufd {{.*#+}} xmm0 = mem[3,2,1,0]
+; CHECK-ICX-NO-BYPASS-DELAY-NEXT:    retq
+;
+; CHECK-ICX-BYPASS-DELAY-LABEL: transform_VPERMILPSrm:
+; CHECK-ICX-BYPASS-DELAY:       # %bb.0:
+; CHECK-ICX-BYPASS-DELAY-NEXT:    vpermilps {{.*#+}} xmm0 = mem[3,2,1,0]
+; CHECK-ICX-BYPASS-DELAY-NEXT:    retq
 ;
 ; CHECK-V4-LABEL: transform_VPERMILPSrm:
 ; CHECK-V4:       # %bb.0:
 ; CHECK-V4-NEXT:    vpermilps {{.*#+}} xmm0 = mem[3,2,1,0]
 ; CHECK-V4-NEXT:    retq
 ;
 ; CHECK-AVX512-LABEL: transform_VPERMILPSrm:
 ; CHECK-AVX512:       # %bb.0:
 ; CHECK-AVX512-NEXT:    vpermilps {{.*#+}} xmm0 = mem[3,2,1,0]
 ; CHECK-AVX512-NEXT:    retq
 ;
 ; CHECK-ZNVER4-LABEL: transform_VPERMILPSrm:
 ; CHECK-ZNVER4:       # %bb.0:
 ; CHECK-ZNVER4-NEXT:    vpermilps {{.*#+}} xmm0 = mem[3,2,1,0]
 ; CHECK-ZNVER4-NEXT:    retq
   %a = load <4 x float>, ptr %ap
   %shufp = shufflevector <4 x float> %a, <4 x float> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
   ret <4 x float> %shufp
 }
 
 define <16 x float> @transform_VPERMILPSZrmkz(ptr %ap, i16 %mask_int) nounwind {
-; CHECK-ICX-NO-BYPASS-DELAY-LABEL: transform_VPERMILPSZrmkz:
-; CHECK-ICX-NO-BYPASS-DELAY:       # %bb.0:
-; CHECK-ICX-NO-BYPASS-DELAY-NEXT:    kmovd %esi, %k1
-; CHECK-ICX-NO-BYPASS-DELAY-NEXT:    vpshufd {{.*#+}} zmm0 {%k1} {z} = mem[3,2,1,0,7,6,5,4,11,10,9,8,15,14,13,12]
-; CHECK-ICX-NO-BYPASS-DELAY-NEXT:    retq
-;
-; CHECK-ICX-BYPASS-DELAY-LABEL: transform_VPERMILPSZrmkz:
-; CHECK-ICX-BYPASS-DELAY:       # %bb.0:
-; CHECK-ICX-BYPASS-DELAY-NEXT:    kmovd %esi, %k1
-; CHECK-ICX-BYPASS-DELAY-NEXT:    vpermilps {{.*#+}} zmm0 {%k1} {z} = mem[3,2,1,0,7,6,5,4,11,10,9,8,15,14,13,12]
-; CHECK-ICX-BYPASS-DELAY-NEXT:    retq
-;
-; CHECK-V4-LABEL: transform_VPERMILPSZrmkz:
-; CHECK-V4:       # %bb.0:
-; CHECK-V4-NEXT:    kmovd %esi, %k1
-; CHECK-V4-NEXT:    vpermilps {{.*#+}} zmm0 {%k1} {z} = mem[3,2,1,0,7,6,5,4,11,10,9,8,15,14,13,12]
-; CHECK-V4-NEXT:    retq
-;
-; CHECK-AVX512-LABEL: transform_VPERMILPSZrmkz:
-; CHECK-AVX512:       # %bb.0:
-; CHECK-AVX512-NEXT:    kmovd %esi, %k1
-; CHECK-AVX512-NEXT:    vpermilps {{.*#+}} zmm0 {%k1} {z} = mem[3,2,1,0,7,6,5,4,11,10,9,8,15,14,13,12]
-; CHECK-AVX512-NEXT:    retq
-;
-; CHECK-ZNVER4-LABEL: transform_VPERMILPSZrmkz:
-; CHECK-ZNVER4:       # %bb.0:
-; CHECK-ZNVER4-NEXT:    kmovd %esi, %k1
-; CHECK-ZNVER4-NEXT:    vpermilps {{.*#+}} zmm0 {%k1} {z} = mem[3,2,1,0,7,6,5,4,11,10,9,8,15,14,13,12]
-; CHECK-ZNVER4-NEXT:    retq
+; CHECK-LABEL: transform_VPERMILPSZrmkz:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    kmovd %esi, %k1
+; CHECK-NEXT:    vpermilps {{.*#+}} zmm0 {%k1} {z} = mem[3,2,1,0,7,6,5,4,11,10,9,8,15,14,13,12]
+; CHECK-NEXT:    retq
   %mask = bitcast i16 %mask_int to <16 x i1>
   %a = load <16 x float>, ptr %ap
   %shufp = shufflevector <16 x float> %a, <16 x float> poison, <16 x i32> <i32 3, i32 2, i32 1, i32 0, i32 7, i32 6, i32 5, i32 4, i32 11, i32 10, i32 9, i32 8, i32 15, i32 14, i32 13, i32 12>
   %res = select <16 x i1> %mask, <16 x float> %shufp, <16 x float> zeroinitializer
   ret <16 x float> %res
 }
 
 define <8 x float> @transform_VPERMILPSYrmkz(ptr %ap, i8 %mask_int) nounwind {
-; CHECK-ICX-LABEL: transform_VPERMILPSYrmkz:
-; CHECK-ICX:       # %bb.0:
-; CHECK-ICX-NEXT:    kmovd %esi, %k1
-; CHECK-ICX-NEXT:    vpshufd {{.*#+}} ymm0 {%k1} {z} = mem[3,2,1,0,7,6,5,4]
-; CHECK-ICX-NEXT:    retq
+; CHECK-ICX-NO-BYPASS-DELAY-LABEL: transform_VPERMILPSYrmkz:
+; CHECK-ICX-NO-BYPASS-DELAY:       # %bb.0:
+; CHECK-ICX-NO-BYPASS-DELAY-NEXT:    kmovd %esi, %k1
+; CHECK-ICX-NO-BYPASS-DELAY-NEXT:    vpshufd {{.*#+}} ymm0 {%k1} {z} = mem[3,2,1,0,7,6,5,4]
+; CHECK-ICX-NO-BYPASS-DELAY-NEXT:    retq
+;
+; CHECK-ICX-BYPASS-DELAY-LABEL: transform_VPERMILPSYrmkz:
+; CHECK-ICX-BYPASS-DELAY:       # %bb.0:
+; CHECK-ICX-BYPASS-DELAY-NEXT:    kmovd %esi, %k1
+; CHECK-ICX-BYPASS-DELAY-NEXT:    vpermilps {{.*#+}} ymm0 {%k1} {z} = mem[3,2,1,0,7,6,5,4]
+; CHECK-ICX-BYPASS-DELAY-NEXT:    retq
 ;
 ; CHECK-V4-LABEL: transform_VPERMILPSYrmkz:
 ; CHECK-V4:       # %bb.0:
 ; CHECK-V4-NEXT:    kmovd %esi, %k1
 ; CHECK-V4-NEXT:    vpermilps {{.*#+}} ymm0 {%k1} {z} = mem[3,2,1,0,7,6,5,4]
 ; CHECK-V4-NEXT:    retq
 ;
 ; CHECK-AVX512-LABEL: transform_VPERMILPSYrmkz:
; CHECK-ZNVER4-NEXT:    retq
   %mask = bitcast i8 %mask_int to <8 x i1>
   %a = load <8 x float>, ptr %ap
   %shufp = shufflevector <8 x float> %a, <8 x float> poison, <8 x i32> <i32 3, i32 2, i32 1, i32 0, i32 7, i32 6, i32 5, i32 4>
   %res = select <8 x i1> %mask, <8 x float> %shufp, <8 x float> zeroinitializer
   ret <8 x float> %res
 }
 
 define <4 x float> @transform_VPERMILPSrmkz(ptr %ap, i4 %mask_int) nounwind {
-; CHECK-ICX-LABEL: transform_VPERMILPSrmkz:
-; CHECK-ICX:       # %bb.0:
-; CHECK-ICX-NEXT:    kmovd %esi, %k1
-; CHECK-ICX-NEXT:    vpshufd {{.*#+}} xmm0 {%k1} {z} = mem[3,2,1,0]
-; CHECK-ICX-NEXT:    retq
+; CHECK-ICX-NO-BYPASS-DELAY-LABEL: transform_VPERMILPSrmkz:
+; CHECK-ICX-NO-BYPASS-DELAY:       # %bb.0:
+; CHECK-ICX-NO-BYPASS-DELAY-NEXT:    kmovd %esi, %k1
+; CHECK-ICX-NO-BYPASS-DELAY-NEXT:    vpshufd {{.*#+}} xmm0 {%k1} {z} = mem[3,2,1,0]
+; CHECK-ICX-NO-BYPASS-DELAY-NEXT:    retq
+;
+; CHECK-ICX-BYPASS-DELAY-LABEL: transform_VPERMILPSrmkz:
+; CHECK-ICX-BYPASS-DELAY:       # %bb.0:
+; CHECK-ICX-BYPASS-DELAY-NEXT:    kmovd %esi, %k1
+; CHECK-ICX-BYPASS-DELAY-NEXT:    vpermilps {{.*#+}} xmm0 {%k1} {z} = mem[3,2,1,0]
+; CHECK-ICX-BYPASS-DELAY-NEXT:    retq
 ;
 ; CHECK-V4-LABEL: transform_VPERMILPSrmkz:
 ; CHECK-V4:       # %bb.0:
 ; CHECK-V4-NEXT:    kmovd %esi, %k1
 ; CHECK-V4-NEXT:    vpermilps {{.*#+}} xmm0 {%k1} {z} = mem[3,2,1,0]
 ; CHECK-V4-NEXT:    retq
 ;
 ; CHECK-AVX512-LABEL: transform_VPERMILPSrmkz:
; CHECK-ZNVER4-NEXT:    retq
   %mask = bitcast i4 %mask_int to <4 x i1>
   %a = load <4 x float>, ptr %ap
   %shufp = shufflevector <4 x float> %a, <4 x float> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
   %res = select <4 x i1> %mask, <4 x float> %shufp, <4 x float> zeroinitializer
   ret <4 x float> %res
 }
 
 define <16 x float> @transform_VPERMILPSZrmk(ptr %ap, <16 x float> %b, i16 %mask_int) nounwind {
-; CHECK-ICX-NO-BYPASS-DELAY-LABEL: transform_VPERMILPSZrmk:
-; CHECK-ICX-NO-BYPASS-DELAY:       # %bb.0:
-; CHECK-ICX-NO-BYPASS-DELAY-NEXT:    kmovd %esi, %k1
-; CHECK-ICX-NO-BYPASS-DELAY-NEXT:    vpshufd {{.*#+}} zmm0 {%k1} = mem[3,2,1,0,7,6,5,4,11,10,9,8,15,14,13,12]
-; CHECK-ICX-NO-BYPASS-DELAY-NEXT:    retq
-;
-; CHECK-ICX-BYPASS-DELAY-LABEL: transform_VPERMILPSZrmk:
-; CHECK-ICX-BYPASS-DELAY:       # %bb.0:
-; CHECK-ICX-BYPASS-DELAY-NEXT:    kmovd %esi, %k1
-; CHECK-ICX-BYPASS-DELAY-NEXT:    vpermilps {{.*#+}} zmm0 {%k1} = mem[3,2,1,0,7,6,5,4,11,10,9,8,15,14,13,12]
-; CHECK-ICX-BYPASS-DELAY-NEXT:    retq
-;
-; CHECK-V4-LABEL: transform_VPERMILPSZrmk:
-; CHECK-V4:       # %bb.0:
-; CHECK-V4-NEXT:    kmovd %esi, %k1
-; CHECK-V4-NEXT:    vpermilps {{.*#+}} zmm0 {%k1} = mem[3,2,1,0,7,6,5,4,11,10,9,8,15,14,13,12]
-; CHECK-V4-NEXT:    retq
-;
-; CHECK-AVX512-LABEL: transform_VPERMILPSZrmk:
-; CHECK-AVX512:       # %bb.0:
-; CHECK-AVX512-NEXT:    kmovd %esi, %k1
-; CHECK-AVX512-NEXT:    vpermilps {{.*#+}} zmm0 {%k1} = mem[3,2,1,0,7,6,5,4,11,10,9,8,15,14,13,12]
-; CHECK-AVX512-NEXT:    retq
-;
-; CHECK-ZNVER4-LABEL: transform_VPERMILPSZrmk:
-; CHECK-ZNVER4:       # %bb.0:
-; CHECK-ZNVER4-NEXT:    kmovd %esi, %k1
-; CHECK-ZNVER4-NEXT:    vpermilps {{.*#+}} zmm0 {%k1} = mem[3,2,1,0,7,6,5,4,11,10,9,8,15,14,13,12]
-; CHECK-ZNVER4-NEXT:    retq
+; CHECK-LABEL: transform_VPERMILPSZrmk:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    kmovd %esi, %k1
+; CHECK-NEXT:    vpermilps {{.*#+}} zmm0 {%k1} = mem[3,2,1,0,7,6,5,4,11,10,9,8,15,14,13,12]
+; CHECK-NEXT:    retq
   %mask = bitcast i16 %mask_int to <16 x i1>
   %a = load <16 x float>, ptr %ap
   %shufp = shufflevector <16 x float> %a, <16 x float> poison, <16 x i32> <i32 3, i32 2, i32 1, i32 0, i32 7, i32 6, i32 5, i32 4, i32 11, i32 10, i32 9, i32 8, i32 15, i32 14, i32 13, i32 12>
   %res = select <16 x i1> %mask, <16 x float> %shufp, <16 x float> %b
   ret <16 x float> %res
 }
 
 define <8 x float> @transform_VPERMILPSYrmk(ptr %ap, <8 x float> %b, i8 %mask_int) nounwind {
-; CHECK-ICX-LABEL: transform_VPERMILPSYrmk:
-; CHECK-ICX:       # %bb.0:
-; CHECK-ICX-NEXT:    kmovd %esi, %k1
-; CHECK-ICX-NEXT:    vpshufd {{.*#+}} ymm0 {%k1} = mem[3,2,1,0,7,6,5,4]
-; CHECK-ICX-NEXT:    retq
+; CHECK-ICX-NO-BYPASS-DELAY-LABEL: transform_VPERMILPSYrmk:
+; CHECK-ICX-NO-BYPASS-DELAY:       # %bb.0:
+; CHECK-ICX-NO-BYPASS-DELAY-NEXT:    kmovd %esi, %k1
+; CHECK-ICX-NO-BYPASS-DELAY-NEXT:    vpshufd {{.*#+}} ymm0 {%k1} = mem[3,2,1,0,7,6,5,4]
+; CHECK-ICX-NO-BYPASS-DELAY-NEXT:    retq
+;
+; CHECK-ICX-BYPASS-DELAY-LABEL: transform_VPERMILPSYrmk:
+; CHECK-ICX-BYPASS-DELAY:       # %bb.0:
+; CHECK-ICX-BYPASS-DELAY-NEXT:    kmovd %esi, %k1
+; CHECK-ICX-BYPASS-DELAY-NEXT:    vpermilps {{.*#+}} ymm0 {%k1} = mem[3,2,1,0,7,6,5,4]
+; CHECK-ICX-BYPASS-DELAY-NEXT:    retq
 ;
 ; CHECK-V4-LABEL: transform_VPERMILPSYrmk:
 ; CHECK-V4:       # %bb.0:
 ; CHECK-V4-NEXT:    kmovd %esi, %k1
 ; CHECK-V4-NEXT:    vpermilps {{.*#+}} ymm0 {%k1} = mem[3,2,1,0,7,6,5,4]
 ; CHECK-V4-NEXT:    retq
 ;
 ; CHECK-AVX512-LABEL: transform_VPERMILPSYrmk:
; CHECK-ZNVER4-NEXT:    retq
   %mask = bitcast i8 %mask_int to <8 x i1>
   %a = load <8 x float>, ptr %ap
   %shufp = shufflevector <8 x float> %a, <8 x float> poison, <8 x i32> <i32 3, i32 2, i32 1, i32 0, i32 7, i32 6, i32 5, i32 4>
   %res = select <8 x i1> %mask, <8 x float> %shufp, <8 x float> %b
   ret <8 x float> %res
 }
 
 define <4 x float> @transform_VPERMILPSrmk(ptr %ap, <4 x float> %b, i4 %mask_int) nounwind {
-; CHECK-ICX-LABEL: transform_VPERMILPSrmk:
-; CHECK-ICX:       # %bb.0:
-; CHECK-ICX-NEXT:    kmovd %esi, %k1
-; CHECK-ICX-NEXT:    vpshufd {{.*#+}} xmm0 {%k1} = mem[3,2,1,0]
-; CHECK-ICX-NEXT:    retq
+; CHECK-ICX-NO-BYPASS-DELAY-LABEL: transform_VPERMILPSrmk:
+; CHECK-ICX-NO-BYPASS-DELAY:       # %bb.0:
+; CHECK-ICX-NO-BYPASS-DELAY-NEXT:    kmovd %esi, %k1
+; CHECK-ICX-NO-BYPASS-DELAY-NEXT:    vpshufd {{.*#+}} xmm0 {%k1} = mem[3,2,1,0]
+; CHECK-ICX-NO-BYPASS-DELAY-NEXT:    retq
+;
+; CHECK-ICX-BYPASS-DELAY-LABEL: transform_VPERMILPSrmk:
+; CHECK-ICX-BYPASS-DELAY:       # %bb.0:
+; CHECK-ICX-BYPASS-DELAY-NEXT:    kmovd %esi, %k1
+; CHECK-ICX-BYPASS-DELAY-NEXT:    vpermilps {{.*#+}} xmm0 {%k1} = mem[3,2,1,0]
+; CHECK-ICX-BYPASS-DELAY-NEXT:    retq
 ;
 ; CHECK-V4-LABEL: transform_VPERMILPSrmk:
 ; CHECK-V4:       # %bb.0:
 ; CHECK-V4-NEXT:    kmovd %esi, %k1
 ; CHECK-V4-NEXT:    vpermilps {{.*#+}} xmm0 {%k1} = mem[3,2,1,0]
 ; CHECK-V4-NEXT:    retq
 ;
 ; CHECK-AVX512-LABEL: transform_VPERMILPSrmk:
 ; CHECK-AVX512:       # %bb.0:
 ; CHECK-AVX512-NEXT:    kmovd %esi, %k1
 ; CHECK-AVX512-NEXT:    vpermilps {{.*#+}} xmm0 {%k1} = mem[3,2,1,0]
 ; CHECK-AVX512-NEXT:    retq
 ;
 ; CHECK-ZNVER4-LABEL: transform_VPERMILPSrmk:
 ; CHECK-ZNVER4:       # %bb.0:
 ; CHECK-ZNVER4-NEXT:    kmovd %esi, %k1
 ; CHECK-ZNVER4-NEXT:    vpermilps {{.*#+}} xmm0 {%k1} = mem[3,2,1,0]
 ; CHECK-ZNVER4-NEXT:    retq
   %mask = bitcast i4 %mask_int to <4 x i1>
   %a = load <4 x float>, ptr %ap
   %shufp = shufflevector <4 x float> %a, <4 x float> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
   %res = select <4 x i1> %mask, <4 x float> %shufp, <4 x float> %b
   ret <4 x float> %res
 }
+;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line:
+; CHECK-ICX: {{.*}}

llvm/test/CodeGen/X86/tuning-shuffle-permilps.ll

 ;
 ; CHECK-AVX1-DELAY-LABEL: transform_VPERMILPSYrm:
 ; CHECK-AVX1-DELAY:       # %bb.0:
 ; CHECK-AVX1-DELAY-NEXT:    vpermilps {{.*#+}} ymm0 = mem[3,2,1,0,7,6,5,4]
 ; CHECK-AVX1-DELAY-NEXT:    retq
 ;
 ; CHECK-AVX2-LABEL: transform_VPERMILPSYrm:
 ; CHECK-AVX2:       # %bb.0:
-; CHECK-AVX2-NEXT:    vpshufd {{.*#+}} ymm0 = mem[3,2,1,0,7,6,5,4]
+; CHECK-AVX2-NEXT:    vpermilps {{.*#+}} ymm0 = mem[3,2,1,0,7,6,5,4]
 ; CHECK-AVX2-NEXT:    retq
 ;
 ; CHECK-AVX2-DELAY-LABEL: transform_VPERMILPSYrm:
 ; CHECK-AVX2-DELAY:       # %bb.0:
 ; CHECK-AVX2-DELAY-NEXT:    vpermilps {{.*#+}} ymm0 = mem[3,2,1,0,7,6,5,4]
 ; CHECK-AVX2-DELAY-NEXT:    retq
 ;
-; CHECK-ICX-LABEL: transform_VPERMILPSYrm:
-; CHECK-ICX:       # %bb.0:
-; CHECK-ICX-NEXT:    vpshufd {{.*#+}} ymm0 = mem[3,2,1,0,7,6,5,4]
-; CHECK-ICX-NEXT:    retq
+; CHECK-ICX-NO-BYPASS-DELAY-LABEL: transform_VPERMILPSYrm:
+; CHECK-ICX-NO-BYPASS-DELAY:       # %bb.0:
+; CHECK-ICX-NO-BYPASS-DELAY-NEXT:    vpshufd {{.*#+}} ymm0 = mem[3,2,1,0,7,6,5,4]
+; CHECK-ICX-NO-BYPASS-DELAY-NEXT:    retq
+;
+; CHECK-ICX-BYPASS-DELAY-LABEL: transform_VPERMILPSYrm:
+; CHECK-ICX-BYPASS-DELAY:       # %bb.0:
+; CHECK-ICX-BYPASS-DELAY-NEXT:    vpermilps {{.*#+}} ymm0 = mem[3,2,1,0,7,6,5,4]
+; CHECK-ICX-BYPASS-DELAY-NEXT:    retq
 ;
 ; CHECK-SNB-LABEL: transform_VPERMILPSYrm:
 ; CHECK-SNB:       # %bb.0:
 ; CHECK-SNB-NEXT:    vpermilps {{.*#+}} ymm0 = mem[3,2,1,0,7,6,5,4]
 ; CHECK-SNB-NEXT:    retq
   %a = load <8 x float>, ptr %ap
   %shufp = shufflevector <8 x float> %a, <8 x float> poison, <8 x i32> <i32 3, i32 2, i32 1, i32 0, i32 7, i32 6, i32 5, i32 4>
   ret <8 x float> %shufp
 }
 
 define <4 x float> @transform_VPERMILPSrm(ptr %ap) nounwind {
 ; CHECK-AVX1-LABEL: transform_VPERMILPSrm:
 ; CHECK-AVX1:       # %bb.0:
-; CHECK-AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = mem[3,2,1,0]
+; CHECK-AVX1-NEXT:    vpermilps {{.*#+}} xmm0 = mem[3,2,1,0]
 ; CHECK-AVX1-NEXT:    retq
 ;
 ; CHECK-AVX1-DELAY-LABEL: transform_VPERMILPSrm:
 ; CHECK-AVX1-DELAY:       # %bb.0:
 ; CHECK-AVX1-DELAY-NEXT:    vpermilps {{.*#+}} xmm0 = mem[3,2,1,0]
 ; CHECK-AVX1-DELAY-NEXT:    retq
 ;
 ; CHECK-AVX2-LABEL: transform_VPERMILPSrm:
 ; CHECK-AVX2:       # %bb.0:
-; CHECK-AVX2-NEXT:    vpshufd {{.*#+}} xmm0 = mem[3,2,1,0]
+; CHECK-AVX2-NEXT:    vpermilps {{.*#+}} xmm0 = mem[3,2,1,0]
 ; CHECK-AVX2-NEXT:    retq
 ;
 ; CHECK-AVX2-DELAY-LABEL: transform_VPERMILPSrm:
 ; CHECK-AVX2-DELAY:       # %bb.0:
 ; CHECK-AVX2-DELAY-NEXT:    vpermilps {{.*#+}} xmm0 = mem[3,2,1,0]
 ; CHECK-AVX2-DELAY-NEXT:    retq
 ;
-; CHECK-ICX-LABEL: transform_VPERMILPSrm:
-; CHECK-ICX:       # %bb.0:
-; CHECK-ICX-NEXT:    vpshufd {{.*#+}} xmm0 = mem[3,2,1,0]
-; CHECK-ICX-NEXT:    retq
-;
-; CHECK-SNB-LABEL: transform_VPERMILPSrm:
-; CHECK-SNB:       # %bb.0:
-; CHECK-SNB-NEXT:    vpshufd {{.*#+}} xmm0 = mem[3,2,1,0]
-; CHECK-SNB-NEXT:    retq
+; CHECK-ICX-NO-BYPASS-DELAY-LABEL: transform_VPERMILPSrm:
+; CHECK-ICX-NO-BYPASS-DELAY:       # %bb.0:
+; CHECK-ICX-NO-BYPASS-DELAY-NEXT:    vpshufd {{.*#+}} xmm0 = mem[3,2,1,0]
+; CHECK-ICX-NO-BYPASS-DELAY-NEXT:    retq
+;
+; CHECK-ICX-BYPASS-DELAY-LABEL: transform_VPERMILPSrm:
+; CHECK-ICX-BYPASS-DELAY:       # %bb.0:
+; CHECK-ICX-BYPASS-DELAY-NEXT:    vpermilps {{.*#+}} xmm0 = mem[3,2,1,0]
+; CHECK-ICX-BYPASS-DELAY-NEXT:    retq
+;
+; CHECK-SNB-NO-BYPASS-DELAY-LABEL: transform_VPERMILPSrm:
+; CHECK-SNB-NO-BYPASS-DELAY:       # %bb.0:
+; CHECK-SNB-NO-BYPASS-DELAY-NEXT:    vpshufd {{.*#+}} xmm0 = mem[3,2,1,0]
+; CHECK-SNB-NO-BYPASS-DELAY-NEXT:    retq
+;
+; CHECK-SNB-BYPASS-DELAY-LABEL: transform_VPERMILPSrm:
+; CHECK-SNB-BYPASS-DELAY:       # %bb.0:
+; CHECK-SNB-BYPASS-DELAY-NEXT:    vpermilps {{.*#+}} xmm0 = mem[3,2,1,0]
+; CHECK-SNB-BYPASS-DELAY-NEXT:    retq
   %a = load <4 x float>, ptr %ap
   %shufp = shufflevector <4 x float> %a, <4 x float> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
   ret <4 x float> %shufp
 }
 ;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line:
-; CHECK-ICX-BYPASS-DELAY: {{.*}}
-; CHECK-ICX-NO-BYPASS-DELAY: {{.*}}
-; CHECK-SNB-BYPASS-DELAY: {{.*}}
-; CHECK-SNB-NO-BYPASS-DELAY: {{.*}}
+; CHECK-ICX: {{.*}}