Make instcombine a little more aggressive in combining vector shuffles.

Allow splats even if they don't match either of the original shuffles,
possibly due to undef entries in the shuffles masks.  Radar 8597790.
Also fix some 80-column violations.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@117719 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/lib/Transforms/InstCombine/InstCombineVectorOps.cpp b/lib/Transforms/InstCombine/InstCombineVectorOps.cpp
index a58124d7032e0c046a00e3f389146270ae641939..634add86a81cd785fd256fef1f7b4f463b81f5a8 100644 (file)
--- a/lib/Transforms/InstCombine/InstCombineVectorOps.cpp
+++ b/lib/Transforms/InstCombine/InstCombineVectorOps.cpp
@@ -515,37 +515,44 @@ Instruction *InstCombiner::visitShuffleVectorInst(ShuffleVectorInst &SVI) {
   // we are absolutely afraid of producing a shuffle mask not in the input
   // program, because the code gen may not be smart enough to turn a merged
   // shuffle into two specific shuffles: it may produce worse code.  As such,
-  // we only merge two shuffles if the result is one of the two input shuffle
-  // masks.  In this case, merging the shuffles just removes one instruction,
-  // which we know is safe.  This is good for things like turning:
-  // (splat(splat)) -> splat.
+  // we only merge two shuffles if the result is either a splat or one of the
+  // two input shuffle masks.  In this case, merging the shuffles just removes
+  // one instruction, which we know is safe.  This is good for things like
+  // turning: (splat(splat)) -> splat.
   if (ShuffleVectorInst *LHSSVI = dyn_cast<ShuffleVectorInst>(LHS)) {
     if (isa<UndefValue>(RHS)) {
       std::vector<unsigned> LHSMask = getShuffleMask(LHSSVI);
       
       if (LHSMask.size() == Mask.size()) {
         std::vector<unsigned> NewMask;
-        for (unsigned i = 0, e = Mask.size(); i != e; ++i)
-          if (Mask[i] >= e)
-            NewMask.push_back(2*e);
-          else
-            NewMask.push_back(LHSMask[Mask[i]]);
+        bool isSplat = true;
+        unsigned SplatElt = 2 * Mask.size(); // undef
+        for (unsigned i = 0, e = Mask.size(); i != e; ++i) {
+          unsigned MaskElt = 2 * e; // undef
+          if (Mask[i] < e)
+            MaskElt = LHSMask[Mask[i]];
+          // Check if this could still be a splat.
+          if (MaskElt < 2*e) {
+            if (SplatElt < 2*e && SplatElt != MaskElt)
+              isSplat = false;
+            SplatElt = MaskElt;
+          }
+          NewMask.push_back(MaskElt);
+        }
         
         // If the result mask is equal to the src shuffle or this
         // shuffle mask, do the replacement.
-        if (NewMask == LHSMask || NewMask == Mask) {
+        if (isSplat || NewMask == LHSMask || NewMask == Mask) {
           unsigned LHSInNElts =
           cast<VectorType>(LHSSVI->getOperand(0)->getType())->
           getNumElements();
           std::vector<Constant*> Elts;
+          const Type *Int32Ty = Type::getInt32Ty(SVI.getContext());
           for (unsigned i = 0, e = NewMask.size(); i != e; ++i) {
             if (NewMask[i] >= LHSInNElts*2) {
-              Elts.push_back(UndefValue::get(
-                                             Type::getInt32Ty(SVI.getContext())));
+              Elts.push_back(UndefValue::get(Int32Ty));
             } else {
-              Elts.push_back(ConstantInt::get(
-                                              Type::getInt32Ty(SVI.getContext()),
-                                              NewMask[i]));
+              Elts.push_back(ConstantInt::get(Int32Ty, NewMask[i]));
             }
           }
           return new ShuffleVectorInst(LHSSVI->getOperand(0),

diff --git a/test/Transforms/InstCombine/vec_shuffle.ll b/test/Transforms/InstCombine/vec_shuffle.ll
index 5132a8ff9bd14548fe920bc35d9abdc08fcea1fa..d4977fb807edf4fd789776f269bcf656b27ed9d3 100644 (file)
--- a/test/Transforms/InstCombine/vec_shuffle.ll
+++ b/test/Transforms/InstCombine/vec_shuffle.ll
@@ -87,3 +87,13 @@ define <4 x i8> @test9(<16 x i8> %tmp6) nounwind {
        %tmp9 = shufflevector <4 x i8> %tmp7, <4 x i8> undef, <4 x i32> < i32 3, i32 1, i32 2, i32 0 >          ; <<4 x i8>> [#uses=1]
        ret <4 x i8> %tmp9
 }
+
+; Redundant vector splats should be removed.  Radar 8597790.
+define <4 x i32> @test10(<4 x i32> %tmp5) nounwind {
+; CHECK: @test10
+; CHECK-NEXT: shufflevector
+; CHECK-NEXT: ret
+  %tmp6 = shufflevector <4 x i32> %tmp5, <4 x i32> undef, <4 x i32> <i32 1, i32 undef, i32 undef, i32 undef>
+  %tmp7 = shufflevector <4 x i32> %tmp6, <4 x i32> undef, <4 x i32> zeroinitializer
+  ret <4 x i32> %tmp7
+}