[X86][AVX] Add support for shuffle decoding of vperm2f128/vperm2i128 with zero'd...

The vperm2f128/vperm2i128 shuffle mask decoding was not attempting to deal with shuffles that give zero lanes. This patch fixes this so that the assembly printer can provide shuffle comments.

As this decoder is also used in X86ISelLowering for shuffle combining, I've added an early-out to match existing behaviour. The hope is that we can add zero support in the future, this would allow other ops' decodes (e.g. insertps) to be combined as well.

Differential Revision: http://reviews.llvm.org/D10593

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@241516 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/lib/Target/X86/Utils/X86ShuffleDecode.cpp b/lib/Target/X86/Utils/X86ShuffleDecode.cpp
index 9777c0d85e9bd3a27ffdff355456c3110695b6c3..b1c69e779f3fdf8c224c8fd0ff2b9c966ac1ab3d 100644 (file)
--- a/lib/Target/X86/Utils/X86ShuffleDecode.cpp
+++ b/lib/Target/X86/Utils/X86ShuffleDecode.cpp
@@ -255,15 +255,13 @@ void DecodeUNPCKLMask(MVT VT, SmallVectorImpl<int> &ShuffleMask) {
 
 void DecodeVPERM2X128Mask(MVT VT, unsigned Imm,
                           SmallVectorImpl<int> &ShuffleMask) {
-  if (Imm & 0x88)
-    return; // Not a shuffle
-
   unsigned HalfSize = VT.getVectorNumElements() / 2;
 
   for (unsigned l = 0; l != 2; ++l) {
-    unsigned HalfBegin = ((Imm >> (l * 4)) & 0x3) * HalfSize;
+    unsigned HalfMask = Imm >> (l * 4);
+    unsigned HalfBegin = (HalfMask & 0x3) * HalfSize;
     for (unsigned i = HalfBegin, e = HalfBegin + HalfSize; i != e; ++i)
-      ShuffleMask.push_back(i);
+      ShuffleMask.push_back(HalfMask & 8 ? SM_SentinelZero : i);
   }
 }
 

diff --git a/lib/Target/X86/X86ISelLowering.cpp b/lib/Target/X86/X86ISelLowering.cpp
index a92ab5ae2a07e9a9fa97717b555682d6e66c451d..5ad31b742a6d7eb10254f199dd4d570dd4119ae8 100644 (file)
--- a/lib/Target/X86/X86ISelLowering.cpp
+++ b/lib/Target/X86/X86ISelLowering.cpp
@@ -4390,6 +4390,7 @@ static SDValue getShuffleVectorZeroOrUndef(SDValue V2, unsigned Idx,
 /// IsUnary to true if only uses one source. Note that this will set IsUnary for
 /// shuffles which use a single input multiple times, and in those cases it will
 /// adjust the mask to only have indices within that single input.
+/// FIXME: Add support for Decode*Mask functions that return SM_SentinelZero.
 static bool getTargetShuffleMask(SDNode *N, MVT VT,
                                  SmallVectorImpl<int> &Mask, bool &IsUnary) {
   unsigned NumElems = VT.getVectorNumElements();
@@ -4519,6 +4520,10 @@ static bool getTargetShuffleMask(SDNode *N, MVT VT,
     ImmN = N->getOperand(N->getNumOperands()-1);
     DecodeVPERM2X128Mask(VT, cast<ConstantSDNode>(ImmN)->getZExtValue(), Mask);
     if (Mask.empty()) return false;
+    // Mask only contains negative index if an element is zero.
+    if (std::any_of(Mask.begin(), Mask.end(), 
+                    [](int M){ return M == SM_SentinelZero; }))
+      return false;
     break;
   case X86ISD::MOVSLDUP:
     DecodeMOVSLDUPMask(VT, Mask);

diff --git a/test/CodeGen/X86/avx-vperm2x128.ll b/test/CodeGen/X86/avx-vperm2x128.ll
index 74d20f348b5298c9c79cc33ab07ec4062db33b95..4e43f6f5192105d3fbea76b72aa594f4a249326b 100644 (file)
--- a/test/CodeGen/X86/avx-vperm2x128.ll
+++ b/test/CodeGen/X86/avx-vperm2x128.ll
@@ -269,7 +269,7 @@ entry:
 define <4 x double> @vperm2z_0x08(<4 x double> %a) {
 ; ALL-LABEL: vperm2z_0x08:
 ; ALL:       # BB#0:
-; ALL-NEXT:    vperm2f128 $40, %ymm0, %ymm0, %ymm0
+; ALL-NEXT:    vperm2f128 {{.*#+}} ymm0 = zero,zero,ymm0[0,1]
 ; ALL-NEXT:    retq
   %s = shufflevector <4 x double> %a, <4 x double> <double 0.0, double 0.0, double undef, double undef>, <4 x i32> <i32 4, i32 5, i32 0, i32 1>
   ret <4 x double> %s
@@ -279,7 +279,7 @@ define <4 x double> @vperm2z_0x18(<4 x double> %a) {
 ; ALL-LABEL: vperm2z_0x18:
 ; ALL:       # BB#0:
 ; ALL-NEXT:    vxorpd %ymm1, %ymm1, %ymm1
-; ALL-NEXT:    vblendpd $12, %ymm0, %ymm1, %ymm0
+; ALL-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3]
 ; ALL-NEXT:    retq
   %s = shufflevector <4 x double> %a, <4 x double> <double 0.0, double 0.0, double undef, double undef>, <4 x i32> <i32 4, i32 5, i32 2, i32 3>
   ret <4 x double> %s
@@ -288,7 +288,7 @@ define <4 x double> @vperm2z_0x18(<4 x double> %a) {
 define <4 x double> @vperm2z_0x28(<4 x double> %a) {
 ; ALL-LABEL: vperm2z_0x28:
 ; ALL:       # BB#0:
-; ALL-NEXT:    vperm2f128 $40, %ymm0, %ymm0, %ymm0
+; ALL-NEXT:    vperm2f128 {{.*#+}} ymm0 = zero,zero,ymm0[0,1]
 ; ALL-NEXT:    retq
   %s = shufflevector <4 x double> <double 0.0, double 0.0, double undef, double undef>, <4 x double> %a, <4 x i32> <i32 0, i32 1, i32 4, i32 5>
   ret <4 x double> %s
@@ -298,7 +298,7 @@ define <4 x double> @vperm2z_0x38(<4 x double> %a) {
 ; ALL-LABEL: vperm2z_0x38:
 ; ALL:       # BB#0:
 ; ALL-NEXT:    vxorpd %ymm1, %ymm1, %ymm1
-; ALL-NEXT:    vblendpd $12, %ymm0, %ymm1, %ymm0
+; ALL-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3]
 ; ALL-NEXT:    retq
   %s = shufflevector <4 x double> <double 0.0, double 0.0, double undef, double undef>, <4 x double> %a, <4 x i32> <i32 0, i32 1, i32 6, i32 7>
   ret <4 x double> %s
@@ -307,7 +307,7 @@ define <4 x double> @vperm2z_0x38(<4 x double> %a) {
 define <4 x double> @vperm2z_0x80(<4 x double> %a) {
 ; ALL-LABEL: vperm2z_0x80:
 ; ALL:       # BB#0:
-; ALL-NEXT:    vperm2f128 $128, %ymm0, %ymm0, %ymm0
+; ALL-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[0,1],zero,zero
 ; ALL-NEXT:    retq
   %s = shufflevector <4 x double> %a, <4 x double> <double 0.0, double 0.0, double undef, double undef>, <4 x i32> <i32 0, i32 1, i32 4, i32 5>
   ret <4 x double> %s
@@ -316,7 +316,7 @@ define <4 x double> @vperm2z_0x80(<4 x double> %a) {
 define <4 x double> @vperm2z_0x81(<4 x double> %a) {
 ; ALL-LABEL: vperm2z_0x81:
 ; ALL:       # BB#0:
-; ALL-NEXT:    vperm2f128 $129, %ymm0, %ymm0, %ymm0
+; ALL-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],zero,zero
 ; ALL-NEXT:    retq
   %s = shufflevector <4 x double> %a, <4 x double> <double 0.0, double 0.0, double undef, double undef>, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
   ret <4 x double> %s
@@ -325,7 +325,7 @@ define <4 x double> @vperm2z_0x81(<4 x double> %a) {
 define <4 x double> @vperm2z_0x82(<4 x double> %a) {
 ; ALL-LABEL: vperm2z_0x82:
 ; ALL:       # BB#0:
-; ALL-NEXT:    vperm2f128 $128, %ymm0, %ymm0, %ymm0
+; ALL-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[0,1],zero,zero
 ; ALL-NEXT:    retq
   %s = shufflevector <4 x double> <double 0.0, double 0.0, double undef, double undef>, <4 x double> %a, <4 x i32> <i32 4, i32 5, i32 0, i32 1>
   ret <4 x double> %s
@@ -334,7 +334,7 @@ define <4 x double> @vperm2z_0x82(<4 x double> %a) {
 define <4 x double> @vperm2z_0x83(<4 x double> %a) {
 ; ALL-LABEL: vperm2z_0x83:
 ; ALL:       # BB#0:
-; ALL-NEXT:    vperm2f128 $129, %ymm0, %ymm0, %ymm0
+; ALL-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],zero,zero
 ; ALL-NEXT:    retq
   %s = shufflevector <4 x double> <double 0.0, double 0.0, double undef, double undef>, <4 x double> %a, <4 x i32> <i32 6, i32 7, i32 0, i32 1>
   ret <4 x double> %s
@@ -345,8 +345,8 @@ define <4 x double> @vperm2z_0x83(<4 x double> %a) {
 define <4 x i64> @vperm2z_int_0x83(<4 x i64> %a, <4 x i64> %b) {
 ; ALL-LABEL: vperm2z_int_0x83:
 ; ALL:       # BB#0:
-; AVX1:    vperm2f128 $129, %ymm0, %ymm0, %ymm0
-; AVX2:    vperm2i128 $129, %ymm0, %ymm0, %ymm0
+; AVX1:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],zero,zero
+; AVX2:    vperm2i128 {{.*#+}} ymm0 = ymm0[2,3],zero,zero
   %s = shufflevector <4 x i64> <i64 0, i64 0, i64 undef, i64 undef>, <4 x i64> %a, <4 x i32> <i32 6, i32 7, i32 0, i32 1>
   %c = add <4 x i64> %b, %s
   ret <4 x i64> %c