[llvm] r179291 - Fix for wrong instcombine on vector insert/extract

[Benjamin Kramer]

Author: d0k
Date: Thu Apr 11 10:10:09 2013
New Revision: 179291

URL: http://llvm.org/viewvc/llvm-project?rev=179291&view=rev
Log:
Fix for wrong instcombine on vector insert/extract

When trying to collapse sequences of insertelement/extractelement
instructions into single shuffle instructions, there is one specific
case where the Instruction Combiner wrongly updates the resulting
Mask of shuffle indexes.

The problem is in function CollectShuffleElments.

If we have a sequence of insert/extract element instructions
like the one below:

  %tmp1 = extractelement <4 x float> %LHS, i32 0
  %tmp2 = insertelement <4 x float> %RHS, float %tmp1, i32 1
  %tmp3 = extractelement <4 x float> %RHS, i32 2
  %tmp4 = insertelement <4 x float> %tmp2, float %tmp3, i32 3

Where:
  . %RHS will have a mask of [4,5,6,7]
  . %LHS will have a mask of [0,1,2,3]

The Mask of shuffle indexes is wrongly computed to [4,1,6,7]
instead of [4,0,6,7].
When analyzing %tmp2 in order to compute the Mask for the
resulting shuffle instruction, the algorithm forgets to update
the mask index at position 1 with the index associated to the
element extracted from %LHS by instruction %tmp1.

Patch by Andrea DiBiagio!

Modified:
    llvm/trunk/lib/Transforms/InstCombine/InstCombineVectorOps.cpp
    llvm/trunk/test/Transforms/InstCombine/vec_shuffle.ll

Modified: llvm/trunk/lib/Transforms/InstCombine/InstCombineVectorOps.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/InstCombine/InstCombineVectorOps.cpp?rev=179291&r1=179290&r2=179291&view=diff
==============================================================================
--- llvm/trunk/lib/Transforms/InstCombine/InstCombineVectorOps.cpp (original)
+++ llvm/trunk/lib/Transforms/InstCombine/InstCombineVectorOps.cpp Thu Apr 11 10:10:09 2013
@@ -336,6 +336,10 @@ static Value *CollectShuffleElements(Val
 
         if (VecOp == RHS) {
           Value *V = CollectShuffleElements(EI->getOperand(0), Mask, RHS);
+          // Update Mask to reflect that `ScalarOp' has been inserted at
+          // position `InsertedIdx' within the vector returned by IEI.
+          Mask[InsertedIdx % NumElts] = Mask[ExtractedIdx];
+
           // Everything but the extracted element is replaced with the RHS.
           for (unsigned i = 0; i != NumElts; ++i) {
             if (i != InsertedIdx)

Modified: llvm/trunk/test/Transforms/InstCombine/vec_shuffle.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Transforms/InstCombine/vec_shuffle.ll?rev=179291&r1=179290&r2=179291&view=diff
==============================================================================
--- llvm/trunk/test/Transforms/InstCombine/vec_shuffle.ll (original)
+++ llvm/trunk/test/Transforms/InstCombine/vec_shuffle.ll Thu Apr 11 10:10:09 2013
@@ -196,3 +196,30 @@ define <4 x i16> @test13e(<4 x i16> %lhs
            <4 x i16> %lhs, <4 x i16> %rhs
   ret <4 x i16> %A
 }
+
+; Check that sequences of insert/extract element are
+; collapsed into shuffle instruction with correct shuffle indexes.
+
+define <4 x float> @test14a(<4 x float> %LHS, <4 x float> %RHS) {
+; CHECK: @test14a
+; CHECK-NEXT: shufflevector <4 x float> %LHS, <4 x float> %RHS, <4 x i32> <i32 4, i32 0, i32 6, i32 6>
+; CHECK-NEXT: ret <4 x float> %tmp4
+        %tmp1 = extractelement <4 x float> %LHS, i32 0
+        %tmp2 = insertelement <4 x float> %RHS, float %tmp1, i32 1
+        %tmp3 = extractelement <4 x float> %RHS, i32 2
+        %tmp4 = insertelement <4 x float> %tmp2, float %tmp3, i32 3
+        ret <4 x float> %tmp4
+}
+
+define <4 x float> @test14b(<4 x float> %LHS, <4 x float> %RHS) {
+; CHECK: @test14b
+; CHECK-NEXT: shufflevector <4 x float> %LHS, <4 x float> %RHS, <4 x i32> <i32 4, i32 3, i32 6, i32 6>
+; CHECK-NEXT: ret <4 x float> %tmp5
+        %tmp0 = extractelement <4 x float> %LHS, i32 3
+        %tmp1 = insertelement <4 x float> %RHS, float %tmp0, i32 0
+        %tmp2 = extractelement <4 x float> %tmp1, i32 0
+        %tmp3 = insertelement <4 x float> %RHS, float %tmp2, i32 1
+        %tmp4 = extractelement <4 x float> %RHS, i32 2
+        %tmp5 = insertelement <4 x float> %tmp3, float %tmp4, i32 3
+        ret <4 x float> %tmp5
+}