[VectorCombine] Fix for non-zero addrspace when creating vector load from scalar load

[bjope]

Summary

This is a fixup to commit 43bdac290663f4424f9fb, to make sure the
address space from the original load pointer is retained in the
vector pointer.

Resolves problem with

Assertion `castIsValid(op, S, Ty) && "Invalid cast!"' failed.

due to address space mismatch.

[spatel]

Thanks! LGTM - see inline for a minor mod.

llvm/lib/Transforms/Vectorize/VectorCombine.cpp

126–127

Let's put the address space into a local variable above this, so we don't have to repeat the 'get'. That will make future enhancements easier too.

llvm/lib/Transforms/Vectorize/VectorCombine.cpp

bool VectorCombine::vectorizeLoadInsert(Instruction &I) {
   // TODO: Allow insert/extract subvector if the type does not match.
   if (VectorTy != I.getType())
     return false;
   Align Alignment = Load->getAlign();
   const DataLayout &DL = I.getModule()->getDataLayout();
   if (!isSafeToLoadUnconditionally(PtrOp, VectorTy, Alignment, DL, Load, &DT))
     return false;
 
+  unsigned AS = Load->getPointerAddressSpace();
+
   // Original pattern: insertelt undef, load [free casts of] ScalarPtr, 0
-  int OldCost = TTI.getMemoryOpCost(Instruction::Load, ScalarTy, Alignment,
-                                    Load->getPointerAddressSpace());
+  int OldCost = TTI.getMemoryOpCost(Instruction::Load, ScalarTy, Alignment, AS);
   APInt DemandedElts = APInt::getOneBitSet(VecNumElts, 0);
   OldCost += TTI.getScalarizationOverhead(VectorTy, DemandedElts, true, false);
 
   // New pattern: load VecPtr
-  int NewCost = TTI.getMemoryOpCost(Instruction::Load, VectorTy, Alignment,
-                                    Load->getPointerAddressSpace());
+  int NewCost = TTI.getMemoryOpCost(Instruction::Load, VectorTy, Alignment, AS);
 
   // We can aggressively convert to the vector form because the backend can
   // invert this transform if it does not result in a performance win.
   if (OldCost < NewCost)
     return false;
 
   // It is safe and potentially profitable to load a vector directly:
   // inselt undef, load Scalar, 0 --> load VecPtr
   IRBuilder<> Builder(Load);
-  Value *CastedPtr = Builder.CreateBitCast(PtrOp, VectorTy->getPointerTo());
+  Value *CastedPtr = Builder.CreateBitCast(PtrOp, VectorTy->getPointerTo(AS));
   LoadInst *VecLd = Builder.CreateAlignedLoad(VectorTy, CastedPtr, Alignment);
   replaceValue(I, *VecLd);
   ++NumVecLoad;
   return true;
 }
 
 /// Determine which, if any, of the inputs should be replaced by a shuffle
 /// followed by extract from a different index.

llvm/test/Transforms/VectorCombine/X86/load.ll

 ; CHECK-NEXT:    ret <4 x float> [[R]]
 ;
   %gep = getelementptr inbounds <4 x float>, <4 x float>* %p, i64 0, i64 0
   %s = load float, float* %gep, align 16
   %r = insertelement <4 x float> undef, float %s, i64 0
   ret <4 x float> %r
 }
 
+; Should work with addrspace as well.
+
+define <4 x float> @gep00_load_f32_insert_v4f32_addrspace(<4 x float> addrspace(44)* align 16 dereferenceable(16) %p) {
+; CHECK-LABEL: @gep00_load_f32_insert_v4f32_addrspace(
+; CHECK-NEXT:    [[R:%.*]] = load <4 x float>, <4 x float> addrspace(44)* [[P:%.*]], align 16
+; CHECK-NEXT:    ret <4 x float> [[R]]
+;
+  %gep = getelementptr inbounds <4 x float>, <4 x float> addrspace(44)* %p, i64 0, i64 0
+  %s = load float, float addrspace(44)* %gep, align 16
+  %r = insertelement <4 x float> undef, float %s, i64 0
+  ret <4 x float> %r
+}
+
 ; If there are enough dereferenceable bytes, we can offset the vector load.
 
 define <8 x i16> @gep01_load_i16_insert_v8i16(<8 x i16>* align 16 dereferenceable(18) %p) {
 ; CHECK-LABEL: @gep01_load_i16_insert_v8i16(
 ; CHECK-NEXT:    [[GEP:%.*]] = getelementptr inbounds <8 x i16>, <8 x i16>* [[P:%.*]], i64 0, i64 1
 ; CHECK-NEXT:    [[TMP1:%.*]] = bitcast i16* [[GEP]] to <8 x i16>*
 ; CHECK-NEXT:    [[R:%.*]] = load <8 x i16>, <8 x i16>* [[TMP1]], align 2
 ; CHECK-NEXT:    ret <8 x i16> [[R]]