[X86][SSE] Combine shuffles to MOVSS/MOVSD whatever the domain.

We already do the same thing in shuffle lowering; but don't do it if we have SSE41 (PBLEND) instead.

llvm-svn: 289937

diff --git a/llvm/lib/Target/X86/X86ISelLowering.cpp b/llvm/lib/Target/X86/X86ISelLowering.cpp
index 2b496b0..1de40c1 100644 (file)
--- a/llvm/lib/Target/X86/X86ISelLowering.cpp
+++ b/llvm/lib/Target/X86/X86ISelLowering.cpp
@@ -26072,17 +26072,17 @@ static bool matchBinaryVectorShuffle(MVT MaskVT, ArrayRef<int> Mask,
       ShuffleVT = MVT::v4f32;
       return true;
     }
-    if (isTargetShuffleEquivalent(Mask, {0, 3}) && FloatDomain) {
-      if (Subtarget.hasSSE2()) {
-        std::swap(V1, V2);
-        Shuffle = X86ISD::MOVSD;
-        ShuffleVT = MVT::v2f64;
-        return true;
-      }
+    if (isTargetShuffleEquivalent(Mask, {0, 3}) && Subtarget.hasSSE2() &&
+        (FloatDomain || !Subtarget.hasSSE41())) {
+      std::swap(V1, V2);
+      Shuffle = X86ISD::MOVSD;
+      ShuffleVT = MaskVT;
+      return true;
     }
-    if (isTargetShuffleEquivalent(Mask, {4, 1, 2, 3}) && FloatDomain) {
+    if (isTargetShuffleEquivalent(Mask, {4, 1, 2, 3}) &&
+        (FloatDomain || !Subtarget.hasSSE41())) {
       Shuffle = X86ISD::MOVSS;
-      ShuffleVT = MVT::v4f32;
+      ShuffleVT = MaskVT;
       return true;
     }
   }

diff --git a/llvm/test/CodeGen/X86/vector-shuffle-combining.ll b/llvm/test/CodeGen/X86/vector-shuffle-combining.ll
index 88f1202..a65d830 100644 (file)
--- a/llvm/test/CodeGen/X86/vector-shuffle-combining.ll
+++ b/llvm/test/CodeGen/X86/vector-shuffle-combining.ll
@@ -1804,9 +1804,7 @@ define <4 x float> @combine_test4b(<4 x float> %a, <4 x float> %b) {
 define <4 x i8> @combine_test1c(<4 x i8>* %a, <4 x i8>* %b) {
 ; SSE2-LABEL: combine_test1c:
 ; SSE2:       # BB#0:
-; SSE2-NEXT:    movd {{.*#+}} xmm1 = mem[0],zero,zero,zero
-; SSE2-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3],xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7]
-; SSE2-NEXT:    punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; SSE2-NEXT:    movss {{.*#+}} xmm1 = mem[0],zero,zero,zero
 ; SSE2-NEXT:    movd {{.*#+}} xmm0 = mem[0],zero,zero,zero
 ; SSE2-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
 ; SSE2-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3]
@@ -1815,9 +1813,7 @@ define <4 x i8> @combine_test1c(<4 x i8>* %a, <4 x i8>* %b) {
 ;
 ; SSSE3-LABEL: combine_test1c:
 ; SSSE3:       # BB#0:
-; SSSE3-NEXT:    movd {{.*#+}} xmm1 = mem[0],zero,zero,zero
-; SSSE3-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3],xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7]
-; SSSE3-NEXT:    punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; SSSE3-NEXT:    movss {{.*#+}} xmm1 = mem[0],zero,zero,zero
 ; SSSE3-NEXT:    movd {{.*#+}} xmm0 = mem[0],zero,zero,zero
 ; SSSE3-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
 ; SSSE3-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3]