Skip to content

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

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
amaiorano merged 2 commits into from
May 9, 2024
Merged

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

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
15e180b
Fix invalid module bitcode when indexing a swizzled bool vector
amaiorano May 6, 2024
e058de7
Address PR comments
amaiorano May 8, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
9 changes: 8 additions & 1 deletion tools/clang/lib/CodeGen/CGExpr.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -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,
Expand Down Expand Up @@ -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))
Comment thread
llvm-beanz marked this conversation as resolved.
if (llvm::MDNode *RangeInfo = getRangeForLoadFromType(Ty))
Load->setMetadata(llvm::LLVMContext::MD_range, RangeInfo);

Expand Down
5 changes: 5 additions & 0 deletions tools/clang/lib/CodeGen/CGExprCXX.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -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()));
Comment thread
llvm-beanz marked this conversation as resolved.
Builder.CreateStore(Elt, TmpEltGEP);
}
This = TmpThis;
Expand Down
30 changes: 30 additions & 0 deletions tools/clang/test/CodeGenDXIL/operators/swizzle/indexSwizzledBoolVec.hlsl
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -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$03@@B", i32 0, i32 2),
// CHECK-NEXT: [[VA2:%[a-z0-9\.]+]] = trunc i32 [[VA1]] to i1,
Comment thread
amaiorano marked this conversation as resolved.
// CHECK-NEXT: store i1 [[VA2]], i1* [[VA0]],
// CHECK-NEXT: [[VB0:%[a-z0-9\.]+]] = getelementptr <2 x i1>, <2 x i1>* [[TMP]], i32 0, i32 1,
// CHECK-NEXT: [[VB1:%[a-z0-9\.]+]] = load i32, i32* getelementptr inbounds (<4 x i32>, <4 x i32>* @"\01?v_bool4@?1??main@@YAXXZ@3V?$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();
}
}