[llvm] r275823 - [X86][AVX] Add target shuffle decode support for VBROADCAST

[Simon Pilgrim via llvm-commits]

Author: rksimon
Date: Mon Jul 18 12:32:59 2016
New Revision: 275823

URL: http://llvm.org/viewvc/llvm-project?rev=275823&view=rev
Log:
[X86][AVX] Add target shuffle decode support for VBROADCAST

Currently we only decode broadcasts from a vector of the same size.

Modified:
    llvm/trunk/lib/Target/X86/Utils/X86ShuffleDecode.cpp
    llvm/trunk/lib/Target/X86/Utils/X86ShuffleDecode.h
    llvm/trunk/lib/Target/X86/X86ISelLowering.cpp
    llvm/trunk/test/CodeGen/X86/vector-shuffle-combining-avx2.ll

Modified: llvm/trunk/lib/Target/X86/Utils/X86ShuffleDecode.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/Utils/X86ShuffleDecode.cpp?rev=275823&r1=275822&r2=275823&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/Utils/X86ShuffleDecode.cpp (original)
+++ llvm/trunk/lib/Target/X86/Utils/X86ShuffleDecode.cpp Mon Jul 18 12:32:59 2016
@@ -275,6 +275,12 @@ void DecodeUNPCKLMask(MVT VT, SmallVecto
   }
 }
 
+/// Decodes a broadcast of the first element of a vector.
+void DecodeVectorBroadcast(MVT DstVT, SmallVectorImpl<int> &ShuffleMask) {
+  unsigned NumElts = DstVT.getVectorNumElements();
+  ShuffleMask.append(NumElts, 0);
+}
+
 /// Decodes a broadcast of a subvector to a larger vector type.
 void DecodeSubVectorBroadcast(MVT DstVT, MVT SrcVT,
                               SmallVectorImpl<int> &ShuffleMask) {

Modified: llvm/trunk/lib/Target/X86/Utils/X86ShuffleDecode.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/Utils/X86ShuffleDecode.h?rev=275823&r1=275822&r2=275823&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/Utils/X86ShuffleDecode.h (original)
+++ llvm/trunk/lib/Target/X86/Utils/X86ShuffleDecode.h Mon Jul 18 12:32:59 2016
@@ -88,6 +88,9 @@ void DecodeUNPCKHMask(MVT VT, SmallVecto
 /// datatypes and vector widths.
 void DecodeUNPCKLMask(MVT VT, SmallVectorImpl<int> &ShuffleMask);
 
+/// Decodes a broadcast of the first element of a vector.
+void DecodeVectorBroadcast(MVT DstVT, SmallVectorImpl<int> &ShuffleMask);
+
 /// Decodes a broadcast of a subvector to a larger vector type.
 void DecodeSubVectorBroadcast(MVT DstVT, MVT SrcVT,
                               SmallVectorImpl<int> &ShuffleMask);

Modified: llvm/trunk/lib/Target/X86/X86ISelLowering.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86ISelLowering.cpp?rev=275823&r1=275822&r2=275823&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/X86ISelLowering.cpp (original)
+++ llvm/trunk/lib/Target/X86/X86ISelLowering.cpp Mon Jul 18 12:32:59 2016
@@ -3804,6 +3804,7 @@ static bool isTargetShuffle(unsigned Opc
   case X86ISD::MOVSD:
   case X86ISD::UNPCKL:
   case X86ISD::UNPCKH:
+  case X86ISD::VBROADCAST:
   case X86ISD::VPERMILPI:
   case X86ISD::VPERMILPV:
   case X86ISD::VPERM2X128:
@@ -4920,6 +4921,15 @@ static bool getTargetShuffleMask(SDNode
     DecodeZeroMoveLowMask(VT, Mask);
     IsUnary = true;
     break;
+  case X86ISD::VBROADCAST: {
+    // We only decode broadcasts of same-sized vectors at the moment.
+    if (N->getOperand(0).getValueType() == VT) {
+      DecodeVectorBroadcast(VT, Mask);
+      IsUnary = true;
+      break;
+    }
+    return false;
+  }
   case X86ISD::VPERMILPV: {
     IsUnary = true;
     SDValue MaskNode = N->getOperand(1);

Modified: llvm/trunk/test/CodeGen/X86/vector-shuffle-combining-avx2.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/vector-shuffle-combining-avx2.ll?rev=275823&r1=275822&r2=275823&view=diff
==============================================================================
--- llvm/trunk/test/CodeGen/X86/vector-shuffle-combining-avx2.ll (original)
+++ llvm/trunk/test/CodeGen/X86/vector-shuffle-combining-avx2.ll Mon Jul 18 12:32:59 2016
@@ -207,7 +207,6 @@ define <16 x i8> @combine_vpbroadcast_ps
 ; CHECK-LABEL: combine_vpbroadcast_pshufb_as_vpbroadcastb128:
 ; CHECK:       # BB#0:
 ; CHECK-NEXT:    vpbroadcastb %xmm0, %xmm0
-; CHECK-NEXT:    vpbroadcastb %xmm0, %xmm0
 ; CHECK-NEXT:    retq
   %1 = shufflevector <16 x i8> %a, <16 x i8> undef, <16 x i32> zeroinitializer
   %2 = call <16 x i8> @llvm.x86.ssse3.pshuf.b.128(<16 x i8> %1, <16 x i8> zeroinitializer)
@@ -218,8 +217,6 @@ define <32 x i8> @combine_vpbroadcast_ps
 ; CHECK-LABEL: combine_vpbroadcast_pshufb_as_vpbroadcastb256:
 ; CHECK:       # BB#0:
 ; CHECK-NEXT:    vpbroadcastb %xmm0, %ymm0
-; CHECK-NEXT:    vpxor %ymm1, %ymm1, %ymm1
-; CHECK-NEXT:    vpshufb %ymm1, %ymm0, %ymm0
 ; CHECK-NEXT:    retq
   %1 = shufflevector <32 x i8> %a, <32 x i8> undef, <32 x i32> zeroinitializer
   %2 = call <32 x i8> @llvm.x86.avx2.pshuf.b(<32 x i8> %1, <32 x i8> zeroinitializer)
@@ -230,7 +227,6 @@ define <4 x float> @combine_vpbroadcast_
 ; CHECK-LABEL: combine_vpbroadcast_pshufb_as_vpbroadcastss128:
 ; CHECK:       # BB#0:
 ; CHECK-NEXT:    vbroadcastss %xmm0, %xmm0
-; CHECK-NEXT:    vbroadcastss %xmm0, %xmm0
 ; CHECK-NEXT:    retq
   %1 = shufflevector <4 x float> %a, <4 x float> undef, <4 x i32> zeroinitializer
   %2 = bitcast <4 x float> %1 to <16 x i8>