[llvm] r212181 - X86: When combining shuffles just remove shuffles that are completely redundant.

[Benjamin Kramer]

Author: d0k
Date: Wed Jul  2 10:09:44 2014
New Revision: 212181

URL: http://llvm.org/viewvc/llvm-project?rev=212181&view=rev
Log:
X86: When combining shuffles just remove shuffles that are completely redundant.

CombineTo doesn't allow replacing a node with itself so this would crash if the
combined shuffle is the same as the input shuffle.

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

Modified: llvm/trunk/lib/Target/X86/X86ISelLowering.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86ISelLowering.cpp?rev=212181&r1=212180&r2=212181&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/X86ISelLowering.cpp (original)
+++ llvm/trunk/lib/Target/X86/X86ISelLowering.cpp Wed Jul  2 10:09:44 2014
@@ -18442,6 +18442,13 @@ static bool combineRedundantDWordShuffle
   V = DAG.getNode(X86ISD::PSHUFD, DL, MVT::v4i32, V.getOperand(0),
                   getV4X86ShuffleImm8ForMask(Mask, DAG));
 
+  // It is possible that one of the combinable shuffles was completely absorbed
+  // by the other, just replace it and revisit all users in that case.
+  if (Old.getNode() == V.getNode()) {
+    DCI.CombineTo(N.getNode(), N.getOperand(0), /*AddTo=*/true);
+    return true;
+  }
+
   // Replace N with its operand as we're going to combine that shuffle away.
   DAG.ReplaceAllUsesWith(N, N.getOperand(0));
 

Modified: llvm/trunk/test/CodeGen/X86/vector-shuffle-combining.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/vector-shuffle-combining.ll?rev=212181&r1=212180&r2=212181&view=diff
==============================================================================
--- llvm/trunk/test/CodeGen/X86/vector-shuffle-combining.ll (original)
+++ llvm/trunk/test/CodeGen/X86/vector-shuffle-combining.ll Wed Jul  2 10:09:44 2014
@@ -66,6 +66,16 @@ define <4 x i32> @combine_pshufd5(<4 x i
   ret <4 x i32> %d
 }
 
+define <4 x i32> @combine_pshufd6(<4 x i32> %a) {
+; CHECK-SSE2-LABEL: @combine_pshufd6
+; CHECK-SSE2:       # BB#0:
+; CHECK-SSE2-NEXT:    pshufd $0
+; CHECK-SSE2-NEXT:    retq
+  %b = call <4 x i32> @llvm.x86.sse2.pshuf.d(<4 x i32> %a, i8 0)
+  %c = call <4 x i32> @llvm.x86.sse2.pshuf.d(<4 x i32> %b, i8 8)
+  ret <4 x i32> %c
+}
+
 define <8 x i16> @combine_pshuflw1(<8 x i16> %a) {
 ; CHECK-SSE2-LABEL: @combine_pshuflw1
 ; CHECK-SSE2:       # BB#0: