Add tests with i8, and add nuw flag to GEP

Author: Icohedron

llvm/lib/Target/DirectX/DXILLegalizePass.cpp

@@ -98,9 +98,9 @@ static void fixI8UseChain(Instruction &I,
       ElementType = AI->getAllocatedType();
     if (auto *GEP = dyn_cast<GetElementPtrInst>(NewOperands[0])) {
       ElementType = GEP->getSourceElementType();
-      if (ElementType->isArrayTy())
-        ElementType = ElementType->getArrayElementType();
     }
+    if (ElementType->isArrayTy())
+      ElementType = ElementType->getArrayElementType();
     LoadInst *NewLoad = Builder.CreateLoad(ElementType, NewOperands[0]);
     ReplacedValues[Load] = NewLoad;
     ToRemove.push_back(Load);
@@ -593,7 +593,8 @@ legalizeLoadStoreOnArrayAllocas(Instruction &I,
 
   IRBuilder<> Builder(&I);
   Value *Zero = Builder.getInt32(0);
-  Value *GEP = Builder.CreateInBoundsGEP(Ty, AllocaPtrOp, {Zero, Zero});
+  Value *GEP = Builder.CreateGEP(Ty, AllocaPtrOp, {Zero, Zero}, "",
+                                 GEPNoWrapFlags::all());
 
   Value *NewLoadStore = nullptr;
   if (auto *LI = dyn_cast<LoadInst>(&I))

llvm/test/CodeGen/DirectX/legalize-i8.ll

@@ -127,10 +127,9 @@ define i32 @i8_geps_index0() {
   ; CHECK:         [[LOAD:%.*]] = load i32, ptr [[GEP]], align 4
   ; CHECK-NEXT:    ret i32 [[LOAD]]
   %1 = alloca [2 x i32], align 8
-  %2 = getelementptr inbounds nuw i8, ptr %1, i32 0
-  %3 = load i8, ptr %2
-  %4 = sext i8 %3 to i32
-  ret i32 %4
+  %2 = load i8, ptr %1
+  %3 = sext i8 %2 to i32
+  ret i32 %3
 }
 
 define i32 @i8_geps_index1() {
@@ -149,11 +148,14 @@ define i32 @i8_geps_index1() {
 define i32 @i8_gep_store() {
   ; CHECK-LABEL: define i32 @i8_gep_store(
   ; CHECK-NEXT:    [[ALLOCA:%.*]] = alloca [2 x i32], align 8
+  ; CHECK-NEXT:    [[GEP:%.*]] = getelementptr inbounds nuw [2 x i32], ptr [[ALLOCA]], i32 0, i32 0
+  ; CHECK-NEXT:    store i32 0, ptr [[GEP]], align 4
   ; CHECK-NEXT:    [[GEP:%.*]] = getelementptr inbounds nuw [2 x i32], ptr [[ALLOCA]], i32 0, i32 1
   ; CHECK-NEXT:    store i32 1, ptr [[GEP]], align 4
   ; CHECK:         [[LOAD:%.*]] = load i32, ptr [[GEP]], align 4
   ; CHECK-NEXT:    ret i32 [[LOAD]]
   %1 = alloca [2 x i32], align 8
+  store i8 0, ptr %1
   %2 = getelementptr inbounds nuw i8, ptr %1, i32 4
   store i8 1, ptr %2
   %3 = load i8, ptr %2

llvm/test/CodeGen/DirectX/legalize-load-store-array-alloca.ll

@@ -3,7 +3,7 @@
 define float @load() {
 ; CHECK-LABEL: define float @load
 ; CHECK-NEXT:    [[ALLOCA:%.*]] = alloca [2 x float], align 4
-; CHECK-NEXT:    [[GEP:%.*]] = getelementptr inbounds [2 x float], ptr [[ALLOCA]], i32 0, i32 0
+; CHECK-NEXT:    [[GEP:%.*]] = getelementptr inbounds nuw [2 x float], ptr [[ALLOCA]], i32 0, i32 0
 ; CHECK-NEXT:    [[LOAD:%.*]] = load float, ptr [[GEP]], align 4
 ; CHECK-NEXT:    ret float [[LOAD]]
   %a = alloca [2 x float], align 4
@@ -14,7 +14,7 @@ define float @load() {
 define void @store() {
 ; CHECK-LABEL: define void @store
 ; CHECK-NEXT:    [[ALLOCA:%.*]] = alloca [3 x i32], align 4
-; CHECK-NEXT:    [[GEP:%.*]] = getelementptr inbounds [3 x i32], ptr [[ALLOCA]], i32 0, i32 0
+; CHECK-NEXT:    [[GEP:%.*]] = getelementptr inbounds nuw [3 x i32], ptr [[ALLOCA]], i32 0, i32 0
 ; CHECK-NEXT:    store i32 0, ptr [[GEP]], align 4
 ; CHECK-NEXT:    ret void
   %a = alloca [3 x i32], align 4