Fix invalid module bitcode when indexing a swizzled bool vector (#6582)

When indexing a swizzled bool vector, some HLSL-specific code in
EmitCXXMemberOrOperatorMemberCallExpr kicks in to handle the
HLSLVecType. In this case, we’re dealing with an ExtVectorElt because of
the swizzle, so this function creates a GEP, Load, and Store on the
vector. However, boolean scalars are returned as type i11 while the
store is storing to a bool, which is an i32, so we need to insert a cast
before the store.

tools/clang/lib/CodeGen/CGExpr.cpp

@@ -1137,6 +1137,12 @@ llvm::MDNode *CodeGenFunction::getRangeForLoadFromType(QualType Ty) {
   return MDHelper.createRange(Min, End);
 }
 
+static bool ShouldEmitRangeMD(llvm::Value *Value, QualType Ty) {
+  if (hasBooleanRepresentation(Ty))
+    return cast<llvm::IntegerType>(Value->getType())->getBitWidth() != 1;
+  return true;
+}
+
 llvm::Value *CodeGenFunction::EmitLoadOfScalar(llvm::Value *Addr, bool Volatile,
                                                unsigned Alignment, QualType Ty,
                                                SourceLocation Loc,
@@ -1236,7 +1242,8 @@ llvm::Value *CodeGenFunction::EmitLoadOfScalar(llvm::Value *Addr, bool Volatile,
       EmitCheck(std::make_pair(Check, Kind), "load_invalid_value", StaticArgs,
                 EmitCheckValue(Load));
     }
-  } else if (CGM.getCodeGenOpts().OptimizationLevel > 0)
+  } else if (CGM.getCodeGenOpts().OptimizationLevel > 0 &&
+             ShouldEmitRangeMD(Load, Ty))
     if (llvm::MDNode *RangeInfo = getRangeForLoadFromType(Ty))
       Load->setMetadata(llvm::LLVMContext::MD_range, RangeInfo);
 

tools/clang/lib/CodeGen/CGExprCXX.cpp

@@ -235,12 +235,17 @@ RValue CodeGenFunction::EmitCXXMemberOrOperatorMemberCallExpr(
 
           llvm::Constant *zero = Builder.getInt32(0);
           llvm::Value *TmpThis = CreateTempAlloca(Ty);
+          QualType ElTy = hlsl::GetElementTypeOrType(Base->getType());
+          bool IsBool = ElTy->isSpecificBuiltinType(BuiltinType::Bool);
           for (unsigned i = 0; i < Ty->getVectorNumElements(); i++) {
             llvm::Value *EltIdx = Elts->getAggregateElement(i);
             llvm::Value *EltGEP = Builder.CreateGEP(This, {zero, EltIdx});
             llvm::Value *TmpEltGEP =
                 Builder.CreateGEP(TmpThis, {zero, Builder.getInt32(i)});
             llvm::Value *Elt = Builder.CreateLoad(EltGEP);
+            if (IsBool)
+              Elt = Builder.CreateTrunc(
+                  Elt, llvm::Type::getInt1Ty(getLLVMContext()));
             Builder.CreateStore(Elt, TmpEltGEP);
           }
           This = TmpThis;

tools/clang/test/CodeGenDXIL/operators/swizzle/indexSwizzledBoolVec.hlsl

@@ -0,0 +1,30 @@
+// Test indexing a swizzled bool vector
+// RUN: %dxc -fcgl -T cs_6_0 %s | FileCheck %s
+
+// This was asserting in Instructions.cpp with:
+// void llvm::StoreInst::AssertOK(): Assertion `getOperand(0)->getType() == cast<PointerType>(getOperand(1)->getType())->getElementType() && "Ptr must be a pointer to Val type!"' failed.
+
+// Make sure load of i32 gets truncated to i1 when indexing bool vectors
+// CHECK:       [[TMP:%[a-z0-9\.]+]] = alloca <2 x i1>
+// CHECK:       [[VA0:%[a-z0-9\.]+]] = getelementptr <2 x i1>, <2 x i1>* [[TMP]], i32 0, i32 0,
+// CHECK-NEXT:  [[VA1:%[a-z0-9\.]+]] = load i32, i32* getelementptr inbounds (<4 x i32>, <4 x i32>* @"\01?v_bool4@?1??main@@YAXXZ@3V?$vector@_N$03@@B", i32 0, i32 2),
+// CHECK-NEXT:  [[VA2:%[a-z0-9\.]+]] = trunc i32 [[VA1]] to i1,
+// CHECK-NEXT:  store i1 [[VA2]], i1* [[VA0]],
+// CHECK-NEXT:  [[VB0:%[a-z0-9\.]+]] = getelementptr <2 x i1>, <2 x i1>* [[TMP]], i32 0, i32 1,
+// CHECK-NEXT:  [[VB1:%[a-z0-9\.]+]] = load i32, i32* getelementptr inbounds (<4 x i32>, <4 x i32>* @"\01?v_bool4@?1??main@@YAXXZ@3V?$vector@_N$03@@B", i32 0, i32 3),
+// CHECK-NEXT:  [[VB2:%[a-z0-9\.]+]] = trunc i32 [[VB1]] to i1,
+// CHECK-NEXT:  store i1 [[VB2]], i1* [[VB0]],
+
+
+cbuffer cbuffer_tint_symbol_3 : register(b0) {
+    uint4 global_uint4[1];
+};
+
+[numthreads(1, 1, 1)]
+void main() {
+    const bool4 v_bool4 = bool4(true, true, true, true);
+    const uint gx = global_uint4[0].x;
+    if (v_bool4.zw[gx] == 0) {
+        GroupMemoryBarrierWithGroupSync();
+    }
+}