From: Simon Pilgrim <llvm-dev@redking.me.uk>
Date: Fri, 21 Jun 2019 18:35:04 +0000 (+0000)
Subject: [X86][AVX] Combine INSERT_SUBVECTOR(SRC0, EXTRACT_SUBVECTOR(SRC1)) as shuffle
X-Git-Tag: android-x86-9.0-r1~1512
X-Git-Url: http://git.osdn.net/view?a=commitdiff_plain;h=59c7547c88a2017dddc7854028d9175a773a9df6;p=android-x86%2Fexternal-llvm.git

[X86][AVX] Combine INSERT_SUBVECTOR(SRC0, EXTRACT_SUBVECTOR(SRC1)) as shuffle

Subvector shuffling often ends up as insert/extract subvector.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@364090 91177308-0d34-0410-b5e6-96231b3b80d8
---

diff --git a/lib/Target/X86/X86ISelLowering.cpp b/lib/Target/X86/X86ISelLowering.cpp
index 30c05a3efb0..caa8c1d47a1 100644
--- a/lib/Target/X86/X86ISelLowering.cpp
+++ b/lib/Target/X86/X86ISelLowering.cpp
@@ -6706,7 +6706,6 @@ static bool getFauxShuffleMask(SDValue N, SmallVectorImpl<int> &Mask,
     return true;
   }
   case ISD::INSERT_SUBVECTOR: {
-    // Handle INSERT_SUBVECTOR(SRC0, SHUFFLE(SRC1)).
     SDValue Src = N.getOperand(0);
     SDValue Sub = N.getOperand(1);
     EVT SubVT = Sub.getValueType();
@@ -6714,12 +6713,26 @@ static bool getFauxShuffleMask(SDValue N, SmallVectorImpl<int> &Mask,
     if (!isa<ConstantSDNode>(N.getOperand(2)) ||
         !N->isOnlyUserOf(Sub.getNode()))
       return false;
+    int InsertIdx = N.getConstantOperandVal(2);
+    // Handle INSERT_SUBVECTOR(SRC0, EXTRACT_SUBVECTOR(SRC1)).
+    if (Sub.getOpcode() == ISD::EXTRACT_SUBVECTOR &&
+        Sub.getOperand(0).getValueType() == VT &&
+        isa<ConstantSDNode>(Sub.getOperand(1))) {
+      int ExtractIdx = Sub.getConstantOperandVal(1);
+      for (int i = 0; i != (int)NumElts; ++i)
+        Mask.push_back(i);
+      for (int i = 0; i != (int)NumSubElts; ++i)
+        Mask[InsertIdx + i] = NumElts + ExtractIdx + i;
+      Ops.push_back(Src);
+      Ops.push_back(Sub.getOperand(0));
+      return true;
+    }
+    // Handle INSERT_SUBVECTOR(SRC0, SHUFFLE(SRC1)).
     SmallVector<int, 64> SubMask;
     SmallVector<SDValue, 2> SubInputs;
     if (!resolveTargetShuffleInputs(peekThroughOneUseBitcasts(Sub), SubInputs,
                                     SubMask, DAG))
       return false;
-    int InsertIdx = N.getConstantOperandVal(2);
     if (SubMask.size() != NumSubElts) {
       assert(((SubMask.size() % NumSubElts) == 0 ||
               (NumSubElts % SubMask.size()) == 0) && "Illegal submask scale");
diff --git a/test/CodeGen/X86/var-permute-128.ll b/test/CodeGen/X86/var-permute-128.ll
index 4d48527f677..50cedfd2a29 100644
--- a/test/CodeGen/X86/var-permute-128.ll
+++ b/test/CodeGen/X86/var-permute-128.ll
@@ -1027,11 +1027,11 @@ define <16 x i8> @var_shuffle_v16i8_from_v32i8_v16i8(<32 x i8> %v, <16 x i8> %in
 ; AVX2-LABEL: var_shuffle_v16i8_from_v32i8_v16i8:
 ; AVX2:       # %bb.0:
 ; AVX2-NEXT:    # kill: def $xmm1 killed $xmm1 def $ymm1
-; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm2
-; AVX2-NEXT:    vpshufb %xmm1, %xmm2, %xmm2
+; AVX2-NEXT:    vpshufb %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[2,3,2,3]
 ; AVX2-NEXT:    vpshufb %xmm1, %xmm0, %xmm0
 ; AVX2-NEXT:    vpcmpgtb {{.*}}(%rip), %ymm1, %ymm1
-; AVX2-NEXT:    vpblendvb %ymm1, %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpblendvb %ymm1, %ymm0, %ymm2, %ymm0
 ; AVX2-NEXT:    # kill: def $xmm0 killed $xmm0 killed $ymm0
 ; AVX2-NEXT:    vzeroupper
 ; AVX2-NEXT:    retq
@@ -1039,11 +1039,11 @@ define <16 x i8> @var_shuffle_v16i8_from_v32i8_v16i8(<32 x i8> %v, <16 x i8> %in
 ; AVX512-LABEL: var_shuffle_v16i8_from_v32i8_v16i8:
 ; AVX512:       # %bb.0:
 ; AVX512-NEXT:    # kill: def $xmm1 killed $xmm1 def $ymm1
-; AVX512-NEXT:    vextracti128 $1, %ymm0, %xmm2
-; AVX512-NEXT:    vpshufb %xmm1, %xmm2, %xmm2
+; AVX512-NEXT:    vpshufb %xmm1, %xmm0, %xmm2
+; AVX512-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[2,3,2,3]
 ; AVX512-NEXT:    vpshufb %xmm1, %xmm0, %xmm0
 ; AVX512-NEXT:    vpcmpgtb {{.*}}(%rip), %ymm1, %ymm1
-; AVX512-NEXT:    vpblendvb %ymm1, %ymm2, %ymm0, %ymm0
+; AVX512-NEXT:    vpblendvb %ymm1, %ymm0, %ymm2, %ymm0
 ; AVX512-NEXT:    # kill: def $xmm0 killed $xmm0 killed $ymm0
 ; AVX512-NEXT:    vzeroupper
 ; AVX512-NEXT:    retq
@@ -1051,12 +1051,12 @@ define <16 x i8> @var_shuffle_v16i8_from_v32i8_v16i8(<32 x i8> %v, <16 x i8> %in
 ; AVX512VLBW-LABEL: var_shuffle_v16i8_from_v32i8_v16i8:
 ; AVX512VLBW:       # %bb.0:
 ; AVX512VLBW-NEXT:    # kill: def $xmm1 killed $xmm1 def $ymm1
-; AVX512VLBW-NEXT:    vextracti128 $1, %ymm0, %xmm2
-; AVX512VLBW-NEXT:    vpshufb %xmm1, %xmm2, %xmm2
+; AVX512VLBW-NEXT:    vpshufb %xmm1, %xmm0, %xmm2
+; AVX512VLBW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[2,3,2,3]
 ; AVX512VLBW-NEXT:    vpshufb %xmm1, %xmm0, %xmm0
 ; AVX512VLBW-NEXT:    vpcmpgtb {{.*}}(%rip), %ymm1, %k1
-; AVX512VLBW-NEXT:    vmovdqu8 %ymm2, %ymm0 {%k1}
-; AVX512VLBW-NEXT:    # kill: def $xmm0 killed $xmm0 killed $ymm0
+; AVX512VLBW-NEXT:    vmovdqu8 %ymm0, %ymm2 {%k1}
+; AVX512VLBW-NEXT:    vmovdqa %xmm2, %xmm0
 ; AVX512VLBW-NEXT:    vzeroupper
 ; AVX512VLBW-NEXT:    retq
 ;
diff --git a/test/CodeGen/X86/var-permute-256.ll b/test/CodeGen/X86/var-permute-256.ll
index 7e1780449fd..5d5ed467e85 100644
--- a/test/CodeGen/X86/var-permute-256.ll
+++ b/test/CodeGen/X86/var-permute-256.ll
@@ -170,39 +170,36 @@ define <16 x i16> @var_shuffle_v16i16(<16 x i16> %v, <16 x i16> %indices) nounwi
 ; AVX2:       # %bb.0:
 ; AVX2-NEXT:    vpmullw {{.*}}(%rip), %ymm1, %ymm1
 ; AVX2-NEXT:    vpaddw {{.*}}(%rip), %ymm1, %ymm1
-; AVX2-NEXT:    vinserti128 $1, %xmm0, %ymm0, %ymm2
+; AVX2-NEXT:    vpermq {{.*#+}} ymm2 = ymm0[2,3,2,3]
 ; AVX2-NEXT:    vpshufb %ymm1, %ymm2, %ymm2
-; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm3
-; AVX2-NEXT:    vpblendd {{.*#+}} ymm0 = ymm3[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-NEXT:    vinserti128 $1, %xmm0, %ymm0, %ymm0
 ; AVX2-NEXT:    vpshufb %ymm1, %ymm0, %ymm0
 ; AVX2-NEXT:    vpcmpgtb {{.*}}(%rip), %ymm1, %ymm1
-; AVX2-NEXT:    vpblendvb %ymm1, %ymm0, %ymm2, %ymm0
+; AVX2-NEXT:    vpblendvb %ymm1, %ymm2, %ymm0, %ymm0
 ; AVX2-NEXT:    retq
 ;
 ; AVX512-LABEL: var_shuffle_v16i16:
 ; AVX512:       # %bb.0:
 ; AVX512-NEXT:    vpmullw {{.*}}(%rip), %ymm1, %ymm1
 ; AVX512-NEXT:    vpaddw {{.*}}(%rip), %ymm1, %ymm1
-; AVX512-NEXT:    vinserti128 $1, %xmm0, %ymm0, %ymm2
+; AVX512-NEXT:    vpermq {{.*#+}} ymm2 = ymm0[2,3,2,3]
 ; AVX512-NEXT:    vpshufb %ymm1, %ymm2, %ymm2
-; AVX512-NEXT:    vextracti128 $1, %ymm0, %xmm3
-; AVX512-NEXT:    vpblendd {{.*#+}} ymm0 = ymm3[0,1,2,3],ymm0[4,5,6,7]
+; AVX512-NEXT:    vinserti128 $1, %xmm0, %ymm0, %ymm0
 ; AVX512-NEXT:    vpshufb %ymm1, %ymm0, %ymm0
 ; AVX512-NEXT:    vpcmpgtb {{.*}}(%rip), %ymm1, %ymm1
-; AVX512-NEXT:    vpblendvb %ymm1, %ymm0, %ymm2, %ymm0
+; AVX512-NEXT:    vpblendvb %ymm1, %ymm2, %ymm0, %ymm0
 ; AVX512-NEXT:    retq
 ;
 ; AVX512VLDQ-LABEL: var_shuffle_v16i16:
 ; AVX512VLDQ:       # %bb.0:
 ; AVX512VLDQ-NEXT:    vpmullw {{.*}}(%rip), %ymm1, %ymm1
 ; AVX512VLDQ-NEXT:    vpaddw {{.*}}(%rip), %ymm1, %ymm1
-; AVX512VLDQ-NEXT:    vinserti128 $1, %xmm0, %ymm0, %ymm2
+; AVX512VLDQ-NEXT:    vpermq {{.*#+}} ymm2 = ymm0[2,3,2,3]
 ; AVX512VLDQ-NEXT:    vpshufb %ymm1, %ymm2, %ymm2
-; AVX512VLDQ-NEXT:    vextracti128 $1, %ymm0, %xmm3
-; AVX512VLDQ-NEXT:    vpblendd {{.*#+}} ymm0 = ymm3[0,1,2,3],ymm0[4,5,6,7]
+; AVX512VLDQ-NEXT:    vinserti128 $1, %xmm0, %ymm0, %ymm0
 ; AVX512VLDQ-NEXT:    vpshufb %ymm1, %ymm0, %ymm0
 ; AVX512VLDQ-NEXT:    vpcmpgtb {{.*}}(%rip), %ymm1, %ymm1
-; AVX512VLDQ-NEXT:    vpblendvb %ymm1, %ymm0, %ymm2, %ymm0
+; AVX512VLDQ-NEXT:    vpblendvb %ymm1, %ymm2, %ymm0, %ymm0
 ; AVX512VLDQ-NEXT:    retq
 ;
 ; AVX512VLBW-LABEL: var_shuffle_v16i16:
@@ -293,45 +290,42 @@ define <32 x i8> @var_shuffle_v32i8(<32 x i8> %v, <32 x i8> %indices) nounwind {
 ;
 ; AVX2-LABEL: var_shuffle_v32i8:
 ; AVX2:       # %bb.0:
-; AVX2-NEXT:    vinserti128 $1, %xmm0, %ymm0, %ymm2
+; AVX2-NEXT:    vpermq {{.*#+}} ymm2 = ymm0[2,3,2,3]
 ; AVX2-NEXT:    vpshufb %ymm1, %ymm2, %ymm2
-; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm3
-; AVX2-NEXT:    vpblendd {{.*#+}} ymm0 = ymm3[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-NEXT:    vinserti128 $1, %xmm0, %ymm0, %ymm0
 ; AVX2-NEXT:    vpshufb %ymm1, %ymm0, %ymm0
 ; AVX2-NEXT:    vpcmpgtb {{.*}}(%rip), %ymm1, %ymm1
-; AVX2-NEXT:    vpblendvb %ymm1, %ymm0, %ymm2, %ymm0
+; AVX2-NEXT:    vpblendvb %ymm1, %ymm2, %ymm0, %ymm0
 ; AVX2-NEXT:    retq
 ;
 ; AVX512-LABEL: var_shuffle_v32i8:
 ; AVX512:       # %bb.0:
-; AVX512-NEXT:    vinserti128 $1, %xmm0, %ymm0, %ymm2
+; AVX512-NEXT:    vpermq {{.*#+}} ymm2 = ymm0[2,3,2,3]
 ; AVX512-NEXT:    vpshufb %ymm1, %ymm2, %ymm2
-; AVX512-NEXT:    vextracti128 $1, %ymm0, %xmm3
-; AVX512-NEXT:    vpblendd {{.*#+}} ymm0 = ymm3[0,1,2,3],ymm0[4,5,6,7]
+; AVX512-NEXT:    vinserti128 $1, %xmm0, %ymm0, %ymm0
 ; AVX512-NEXT:    vpshufb %ymm1, %ymm0, %ymm0
 ; AVX512-NEXT:    vpcmpgtb {{.*}}(%rip), %ymm1, %ymm1
-; AVX512-NEXT:    vpblendvb %ymm1, %ymm0, %ymm2, %ymm0
+; AVX512-NEXT:    vpblendvb %ymm1, %ymm2, %ymm0, %ymm0
 ; AVX512-NEXT:    retq
 ;
 ; AVX512VLDQ-LABEL: var_shuffle_v32i8:
 ; AVX512VLDQ:       # %bb.0:
-; AVX512VLDQ-NEXT:    vinserti128 $1, %xmm0, %ymm0, %ymm2
+; AVX512VLDQ-NEXT:    vpermq {{.*#+}} ymm2 = ymm0[2,3,2,3]
 ; AVX512VLDQ-NEXT:    vpshufb %ymm1, %ymm2, %ymm2
-; AVX512VLDQ-NEXT:    vextracti128 $1, %ymm0, %xmm3
-; AVX512VLDQ-NEXT:    vpblendd {{.*#+}} ymm0 = ymm3[0,1,2,3],ymm0[4,5,6,7]
+; AVX512VLDQ-NEXT:    vinserti128 $1, %xmm0, %ymm0, %ymm0
 ; AVX512VLDQ-NEXT:    vpshufb %ymm1, %ymm0, %ymm0
 ; AVX512VLDQ-NEXT:    vpcmpgtb {{.*}}(%rip), %ymm1, %ymm1
-; AVX512VLDQ-NEXT:    vpblendvb %ymm1, %ymm0, %ymm2, %ymm0
+; AVX512VLDQ-NEXT:    vpblendvb %ymm1, %ymm2, %ymm0, %ymm0
 ; AVX512VLDQ-NEXT:    retq
 ;
 ; AVX512VLBW-LABEL: var_shuffle_v32i8:
 ; AVX512VLBW:       # %bb.0:
-; AVX512VLBW-NEXT:    vextracti128 $1, %ymm0, %xmm2
-; AVX512VLBW-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm0[4,5,6,7]
-; AVX512VLBW-NEXT:    vinserti128 $1, %xmm0, %ymm0, %ymm0
-; AVX512VLBW-NEXT:    vpshufb %ymm1, %ymm0, %ymm0
+; AVX512VLBW-NEXT:    vinserti128 $1, %xmm0, %ymm0, %ymm2
+; AVX512VLBW-NEXT:    vpshufb %ymm1, %ymm2, %ymm2
+; AVX512VLBW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[2,3,2,3]
 ; AVX512VLBW-NEXT:    vpcmpgtb {{.*}}(%rip), %ymm1, %k1
-; AVX512VLBW-NEXT:    vpshufb %ymm1, %ymm2, %ymm0 {%k1}
+; AVX512VLBW-NEXT:    vpshufb %ymm1, %ymm0, %ymm2 {%k1}
+; AVX512VLBW-NEXT:    vmovdqa %ymm2, %ymm0
 ; AVX512VLBW-NEXT:    retq
 ;
 ; VLVBMI-LABEL: var_shuffle_v32i8:
diff --git a/test/CodeGen/X86/vector-shuffle-256-v4.ll b/test/CodeGen/X86/vector-shuffle-256-v4.ll
index 099aad76ba7..33513fef481 100644
--- a/test/CodeGen/X86/vector-shuffle-256-v4.ll
+++ b/test/CodeGen/X86/vector-shuffle-256-v4.ll
@@ -447,11 +447,23 @@ define <4 x double> @shuffle_v4f64_0167(<4 x double> %a, <4 x double> %b) {
 }
 
 define <4 x double> @shuffle_v4f64_1054(<4 x double> %a, <4 x double> %b) {
-; ALL-LABEL: shuffle_v4f64_1054:
-; ALL:       # %bb.0:
-; ALL-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
-; ALL-NEXT:    vpermilpd {{.*#+}} ymm0 = ymm0[1,0,3,2]
-; ALL-NEXT:    retq
+; AVX1OR2-LABEL: shuffle_v4f64_1054:
+; AVX1OR2:       # %bb.0:
+; AVX1OR2-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; AVX1OR2-NEXT:    vpermilpd {{.*#+}} ymm0 = ymm0[1,0,3,2]
+; AVX1OR2-NEXT:    retq
+;
+; AVX512VL-SLOW-LABEL: shuffle_v4f64_1054:
+; AVX512VL-SLOW:       # %bb.0:
+; AVX512VL-SLOW-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; AVX512VL-SLOW-NEXT:    vpermilpd {{.*#+}} ymm0 = ymm0[1,0,3,2]
+; AVX512VL-SLOW-NEXT:    retq
+;
+; AVX512VL-FAST-LABEL: shuffle_v4f64_1054:
+; AVX512VL-FAST:       # %bb.0:
+; AVX512VL-FAST-NEXT:    vmovapd {{.*#+}} ymm2 = [1,0,5,4]
+; AVX512VL-FAST-NEXT:    vpermt2pd %ymm1, %ymm2, %ymm0
+; AVX512VL-FAST-NEXT:    retq
   %shuffle = shufflevector <4 x double> %a, <4 x double> %b, <4 x i32> <i32 1, i32 0, i32 5, i32 4>
   ret <4 x double> %shuffle
 }
@@ -990,19 +1002,11 @@ define <4 x i64> @shuffle_v4i64_0142(<4 x i64> %a, <4 x i64> %b) {
 ; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
 ; AVX2-NEXT:    retq
 ;
-; AVX512VL-SLOW-LABEL: shuffle_v4i64_0142:
-; AVX512VL-SLOW:       # %bb.0:
-; AVX512VL-SLOW-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
-; AVX512VL-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,1,2,2]
-; AVX512VL-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
-; AVX512VL-SLOW-NEXT:    retq
-;
-; AVX512VL-FAST-LABEL: shuffle_v4i64_0142:
-; AVX512VL-FAST:       # %bb.0:
-; AVX512VL-FAST-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm1
-; AVX512VL-FAST-NEXT:    vmovdqa {{.*#+}} ymm2 = [0,1,6,2]
-; AVX512VL-FAST-NEXT:    vpermt2q %ymm1, %ymm2, %ymm0
-; AVX512VL-FAST-NEXT:    retq
+; AVX512VL-LABEL: shuffle_v4i64_0142:
+; AVX512VL:       # %bb.0:
+; AVX512VL-NEXT:    vmovdqa {{.*#+}} ymm2 = [0,1,4,2]
+; AVX512VL-NEXT:    vpermt2q %ymm1, %ymm2, %ymm0
+; AVX512VL-NEXT:    retq
   %shuffle = shufflevector <4 x i64> %a, <4 x i64> %b, <4 x i32> <i32 0, i32 1, i32 4, i32 2>
   ret <4 x i64> %shuffle
 }
@@ -1198,11 +1202,17 @@ define <4 x i64> @shuffle_v4i64_1054(<4 x i64> %a, <4 x i64> %b) {
 ; AVX2-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[2,3,0,1,6,7,4,5]
 ; AVX2-NEXT:    retq
 ;
-; AVX512VL-LABEL: shuffle_v4i64_1054:
-; AVX512VL:       # %bb.0:
-; AVX512VL-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
-; AVX512VL-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[2,3,0,1,6,7,4,5]
-; AVX512VL-NEXT:    retq
+; AVX512VL-SLOW-LABEL: shuffle_v4i64_1054:
+; AVX512VL-SLOW:       # %bb.0:
+; AVX512VL-SLOW-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; AVX512VL-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[2,3,0,1,6,7,4,5]
+; AVX512VL-SLOW-NEXT:    retq
+;
+; AVX512VL-FAST-LABEL: shuffle_v4i64_1054:
+; AVX512VL-FAST:       # %bb.0:
+; AVX512VL-FAST-NEXT:    vmovdqa {{.*#+}} ymm2 = [1,0,5,4]
+; AVX512VL-FAST-NEXT:    vpermt2q %ymm1, %ymm2, %ymm0
+; AVX512VL-FAST-NEXT:    retq
   %shuffle = shufflevector <4 x i64> %a, <4 x i64> %b, <4 x i32> <i32 1, i32 0, i32 5, i32 4>
   ret <4 x i64> %shuffle
 }
diff --git a/test/CodeGen/X86/vector-shuffle-256-v8.ll b/test/CodeGen/X86/vector-shuffle-256-v8.ll
index 3f9d9ade272..9969167a10d 100644
--- a/test/CodeGen/X86/vector-shuffle-256-v8.ll
+++ b/test/CodeGen/X86/vector-shuffle-256-v8.ll
@@ -935,11 +935,23 @@ define <8 x float> @shuffle_v8f32_76543210(<8 x float> %a, <8 x float> %b) {
 }
 
 define <8 x float> @shuffle_v8f32_3210ba98(<8 x float> %a, <8 x float> %b) {
-; ALL-LABEL: shuffle_v8f32_3210ba98:
-; ALL:       # %bb.0:
-; ALL-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
-; ALL-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
-; ALL-NEXT:    retq
+; AVX1OR2-LABEL: shuffle_v8f32_3210ba98:
+; AVX1OR2:       # %bb.0:
+; AVX1OR2-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; AVX1OR2-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
+; AVX1OR2-NEXT:    retq
+;
+; AVX512VL-SLOW-LABEL: shuffle_v8f32_3210ba98:
+; AVX512VL-SLOW:       # %bb.0:
+; AVX512VL-SLOW-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; AVX512VL-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
+; AVX512VL-SLOW-NEXT:    retq
+;
+; AVX512VL-FAST-LABEL: shuffle_v8f32_3210ba98:
+; AVX512VL-FAST:       # %bb.0:
+; AVX512VL-FAST-NEXT:    vmovaps {{.*#+}} ymm2 = [3,2,1,0,11,10,9,8]
+; AVX512VL-FAST-NEXT:    vpermt2ps %ymm1, %ymm2, %ymm0
+; AVX512VL-FAST-NEXT:    retq
   %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 3, i32 2, i32 1, i32 0, i32 11, i32 10, i32 9, i32 8>
   ret <8 x float> %shuffle
 }
@@ -1064,11 +1076,24 @@ define <8 x float> @shuffle_v8f32_ba987654(<8 x float> %a, <8 x float> %b) {
 }
 
 define <8 x float> @shuffle_v8f32_ba983210(<8 x float> %a, <8 x float> %b) {
-; ALL-LABEL: shuffle_v8f32_ba983210:
-; ALL:       # %bb.0:
-; ALL-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
-; ALL-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
-; ALL-NEXT:    retq
+; AVX1OR2-LABEL: shuffle_v8f32_ba983210:
+; AVX1OR2:       # %bb.0:
+; AVX1OR2-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1OR2-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
+; AVX1OR2-NEXT:    retq
+;
+; AVX512VL-SLOW-LABEL: shuffle_v8f32_ba983210:
+; AVX512VL-SLOW:       # %bb.0:
+; AVX512VL-SLOW-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX512VL-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
+; AVX512VL-SLOW-NEXT:    retq
+;
+; AVX512VL-FAST-LABEL: shuffle_v8f32_ba983210:
+; AVX512VL-FAST:       # %bb.0:
+; AVX512VL-FAST-NEXT:    vmovaps {{.*#+}} ymm2 = [3,2,1,0,11,10,9,8]
+; AVX512VL-FAST-NEXT:    vpermi2ps %ymm0, %ymm1, %ymm2
+; AVX512VL-FAST-NEXT:    vmovaps %ymm2, %ymm0
+; AVX512VL-FAST-NEXT:    retq
   %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 11, i32 10, i32 9, i32 8, i32 3, i32 2, i32 1, i32 0>
   ret <8 x float> %shuffle
 }
@@ -2240,11 +2265,23 @@ define <8 x i32> @shuffle_v8i32_76543210(<8 x i32> %a, <8 x i32> %b) {
 }
 
 define <8 x i32> @shuffle_v8i32_3210ba98(<8 x i32> %a, <8 x i32> %b) {
-; ALL-LABEL: shuffle_v8i32_3210ba98:
-; ALL:       # %bb.0:
-; ALL-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
-; ALL-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
-; ALL-NEXT:    retq
+; AVX1OR2-LABEL: shuffle_v8i32_3210ba98:
+; AVX1OR2:       # %bb.0:
+; AVX1OR2-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; AVX1OR2-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
+; AVX1OR2-NEXT:    retq
+;
+; AVX512VL-SLOW-LABEL: shuffle_v8i32_3210ba98:
+; AVX512VL-SLOW:       # %bb.0:
+; AVX512VL-SLOW-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; AVX512VL-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
+; AVX512VL-SLOW-NEXT:    retq
+;
+; AVX512VL-FAST-LABEL: shuffle_v8i32_3210ba98:
+; AVX512VL-FAST:       # %bb.0:
+; AVX512VL-FAST-NEXT:    vmovdqa {{.*#+}} ymm2 = [3,2,1,0,11,10,9,8]
+; AVX512VL-FAST-NEXT:    vpermt2d %ymm1, %ymm2, %ymm0
+; AVX512VL-FAST-NEXT:    retq
   %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 3, i32 2, i32 1, i32 0, i32 11, i32 10, i32 9, i32 8>
   ret <8 x i32> %shuffle
 }
diff --git a/test/CodeGen/X86/vector-shuffle-512-v8.ll b/test/CodeGen/X86/vector-shuffle-512-v8.ll
index dea5457baea..ff3a3106337 100644
--- a/test/CodeGen/X86/vector-shuffle-512-v8.ll
+++ b/test/CodeGen/X86/vector-shuffle-512-v8.ll
@@ -2252,13 +2252,21 @@ define <8 x double> @shuffle_v2f64_v8f64_01010101(<2 x double> %a) {
 
 ;FIXME: compressp
 define <4 x double> @test_v8f64_2346 (<8 x double> %v) {
-; ALL-LABEL: test_v8f64_2346:
-; ALL:       # %bb.0:
-; ALL-NEXT:    vextractf128 $1, %ymm0, %xmm1
-; ALL-NEXT:    vextractf64x4 $1, %zmm0, %ymm0
-; ALL-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,1,0,2]
-; ALL-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; ALL-NEXT:    ret{{[l|q]}}
+; AVX512F-LABEL: test_v8f64_2346:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vbroadcasti64x4 {{.*#+}} zmm1 = [2,3,4,6,2,3,4,6]
+; AVX512F-NEXT:    # zmm1 = mem[0,1,2,3,0,1,2,3]
+; AVX512F-NEXT:    vpermq %zmm0, %zmm1, %zmm0
+; AVX512F-NEXT:    # kill: def $ymm0 killed $ymm0 killed $zmm0
+; AVX512F-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_v8f64_2346:
+; AVX512F-32:       # %bb.0:
+; AVX512F-32-NEXT:    vbroadcasti64x4 {{.*#+}} zmm1 = [2,0,3,0,4,0,6,0,2,0,3,0,4,0,6,0]
+; AVX512F-32-NEXT:    # zmm1 = mem[0,1,2,3,0,1,2,3]
+; AVX512F-32-NEXT:    vpermq %zmm0, %zmm1, %zmm0
+; AVX512F-32-NEXT:    # kill: def $ymm0 killed $ymm0 killed $zmm0
+; AVX512F-32-NEXT:    retl
   %res = shufflevector <8 x double> %v, <8 x double> undef, <4 x i32> <i32 2, i32 3, i32 4, i32 6>
   ret <4 x double> %res
 }