[TTI] Fix cast cost on vector types.

- Only split vector types when both src and dst types are splittable.

llvm/include/llvm/CodeGen/BasicTTIImpl.h

@@ -781,9 +781,10 @@ class BasicTTIImplBase : public TargetTransformInfoImplCRTPBase<T> {
       // cost of the split itself. Count that as 1, to be consistent with
       // TLI->getTypeLegalizationCost().
       if ((TLI->getTypeAction(Src->getContext(), TLI->getValueType(DL, Src)) ==
-           TargetLowering::TypeSplitVector) ||
-          (TLI->getTypeAction(Dst->getContext(), TLI->getValueType(DL, Dst)) ==
-           TargetLowering::TypeSplitVector)) {
+               TargetLowering::TypeSplitVector ||
+           TLI->getTypeAction(Dst->getContext(), TLI->getValueType(DL, Dst)) ==
+               TargetLowering::TypeSplitVector) &&
+          Src->getVectorNumElements() > 1 && Dst->getVectorNumElements() > 1) {
         Type *SplitDst = VectorType::get(Dst->getVectorElementType(),
                                          Dst->getVectorNumElements() / 2);
         Type *SplitSrc = VectorType::get(Src->getVectorElementType(),

llvm/test/Transforms/LICM/AMDGPU/bitcast.ll

@@ -0,0 +1,21 @@
+; RUN: opt -licm -mtriple=amdgcn -S -o - %s | FileCheck %s
+
+; CHECK-LABEL: foo
+; CHECK: ret
+define void @foo(i8* %d, <1 x i32>* %s, i32 %idx) {
+entry:
+  br label %for.body
+
+for.body:
+  %v0 = load <1 x i32>, <1 x i32>* %s
+  %v1 = bitcast <1 x i32> %v0 to <4 x i8>
+  br label %for.cond
+
+for.cond:
+  %e0 = extractelement <4 x i8> %v1, i32 %idx
+  store i8 %e0, i8* %d
+  br i1 false, label %for.exit, label %for.body
+
+for.exit:
+  ret void
+}

llvm/test/Transforms/LICM/AMDGPU/lit.local.cfg

@@ -0,0 +1,2 @@
+if not 'AMDGPU' in config.root.targets:
+    config.unsupported = True