Skip to content

[DirectX] Support ConstExpr GEPs #150082

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
merged 4 commits into from
Jul 23, 2025
Merged

[DirectX] Support ConstExpr GEPs #150082

Show file tree
Hide file tree
Changes from 3 commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
ebfd7d0
[DirectX] Support ConstExpr GEPs
farzonl Jul 15, 2025
2c0e73c
address pr comment
farzonl Jul 23, 2025
d2aab98
fix typo indicies need to be on GOp not GEPI
farzonl Jul 23, 2025
50ac604
remove comment to address pr comment
farzonl Jul 23, 2025
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
46 changes: 29 additions & 17 deletions llvm/lib/Target/DirectX/DXILDataScalarization.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -301,41 +301,53 @@ bool DataScalarizerVisitor::visitExtractElementInst(ExtractElementInst &EEI) {
}

bool DataScalarizerVisitor::visitGetElementPtrInst(GetElementPtrInst &GEPI) {
Value *PtrOperand = GEPI.getPointerOperand();
Type *OrigGEPType = GEPI.getSourceElementType();
Type *NewGEPType = OrigGEPType;
GEPOperator *GOp = cast<GEPOperator>(&GEPI);
Value *PtrOperand = GOp->getPointerOperand();
Type *NewGEPType = GOp->getSourceElementType();
bool NeedsTransform = false;

// Unwrap GEP ConstantExprs to find the base operand and element type
while (auto *CE = dyn_cast<ConstantExpr>(PtrOperand)) {
if (auto *GEPCE = dyn_cast<GEPOperator>(CE)) {
GOp = GEPCE;
PtrOperand = GEPCE->getPointerOperand();
NewGEPType = GEPCE->getSourceElementType();
} else
break;
}

if (GlobalVariable *NewGlobal = lookupReplacementGlobal(PtrOperand)) {
NewGEPType = NewGlobal->getValueType();
PtrOperand = NewGlobal;
NeedsTransform = true;
} else if (AllocaInst *Alloca = dyn_cast<AllocaInst>(PtrOperand)) {
Type *AllocatedType = Alloca->getAllocatedType();
// Only transform if the allocated type is an array
if (AllocatedType != OrigGEPType && isa<ArrayType>(AllocatedType)) {
if (isa<ArrayType>(AllocatedType) &&
AllocatedType != GOp->getResultElementType()) {
NewGEPType = AllocatedType;
NeedsTransform = true;
}
}

// Scalar geps should remain scalars geps. The dxil-flatten-arrays pass will
// convert these scalar geps into flattened array geps
if (!isa<ArrayType>(OrigGEPType))
NewGEPType = OrigGEPType;

// Note: We bail if this isn't a gep touched via alloca or global
// transformations
if (!NeedsTransform)
return false;

IRBuilder<> Builder(&GEPI);
SmallVector<Value *, MaxVecSize> Indices(GEPI.indices());
// Keep scalar GEPs scalar; dxil-flatten-arrays will do flattening later
if (!isa<ArrayType>(GOp->getSourceElementType()))
NewGEPType = GOp->getSourceElementType();

IRBuilder<> Builder(&GEPI);
SmallVector<Value *, MaxVecSize> Indices(GOp->indices());
Value *NewGEP = Builder.CreateGEP(NewGEPType, PtrOperand, Indices,
GEPI.getName(), GEPI.getNoWrapFlags());
GEPI.replaceAllUsesWith(NewGEP);
GEPI.eraseFromParent();
GOp->getName(), GOp->getNoWrapFlags());

GOp->replaceAllUsesWith(NewGEP);

if (auto *CE = dyn_cast<ConstantExpr>(GOp))
CE->destroyConstant();
else if (auto *OldGEPI = dyn_cast<GetElementPtrInst>(GOp))
OldGEPI->eraseFromParent(); // This will always be true in visit* context

return true;
}

Expand Down
80 changes: 80 additions & 0 deletions llvm/test/CodeGen/DirectX/bugfix_150050_data_scalarize_const_gep.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,80 @@
; RUN: opt -S -passes='dxil-data-scalarization' -mtriple=dxil-pc-shadermodel6.4-library %s | FileCheck %s --check-prefixes=SCHECK,CHECK
; RUN: opt -S -passes='dxil-data-scalarization,function(scalarizer<load-store>),dxil-flatten-arrays' -mtriple=dxil-pc-shadermodel6.4-library %s | FileCheck %s --check-prefixes=FCHECK,CHECK

@aTile = hidden addrspace(3) global [10 x [10 x <4 x i32>]] zeroinitializer, align 16
@bTile = hidden addrspace(3) global [10 x [10 x i32]] zeroinitializer, align 16
@cTile = internal global [2 x [2 x <2 x i32>]] zeroinitializer, align 16
@dTile = internal global [2 x [2 x [2 x <2 x i32>]]] zeroinitializer, align 16

define void @CSMain() {
; CHECK-LABEL: define void @CSMain() {
; CHECK-NEXT: [[ENTRY:.*:]]
; CHECK-NEXT: [[AFRAGPACKED_I_SCALARIZE:%.*]] = alloca [4 x i32], align 16
;
; SCHECK-NEXT: [[TMP0:%.*]] = getelementptr inbounds [10 x <4 x i32>], ptr addrspace(3) getelementptr inbounds ([10 x [10 x [4 x i32]]], ptr addrspace(3) @aTile.scalarized, i32 0, i32 1), i32 0, i32 2
; SCHECK-NEXT: [[TMP1:%.*]] = load <4 x i32>, ptr addrspace(3) [[TMP0]], align 16
; SCHECK-NEXT: store <4 x i32> [[TMP1]], ptr [[AFRAGPACKED_I_SCALARIZE]], align 16
;
; FCHECK-NEXT: [[AFRAGPACKED_I_SCALARIZE_I14:%.*]] = getelementptr [4 x i32], ptr [[AFRAGPACKED_I_SCALARIZE]], i32 0, i32 1
; FCHECK-NEXT: [[AFRAGPACKED_I_SCALARIZE_I25:%.*]] = getelementptr [4 x i32], ptr [[AFRAGPACKED_I_SCALARIZE]], i32 0, i32 2
; FCHECK-NEXT: [[AFRAGPACKED_I_SCALARIZE_I36:%.*]] = getelementptr [4 x i32], ptr [[AFRAGPACKED_I_SCALARIZE]], i32 0, i32 3
; FCHECK-NEXT: [[DOTI07:%.*]] = load i32, ptr addrspace(3) getelementptr inbounds ([400 x i32], ptr addrspace(3) @aTile.scalarized.1dim, i32 0, i32 48), align 16
; FCHECK-NEXT: [[DOTI119:%.*]] = load i32, ptr addrspace(3) getelementptr ([400 x i32], ptr addrspace(3) @aTile.scalarized.1dim, i32 0, i32 49), align 4
; FCHECK-NEXT: [[DOTI2211:%.*]] = load i32, ptr addrspace(3) getelementptr ([400 x i32], ptr addrspace(3) @aTile.scalarized.1dim, i32 0, i32 50), align 8
; FCHECK-NEXT: [[DOTI3313:%.*]] = load i32, ptr addrspace(3) getelementptr ([400 x i32], ptr addrspace(3) @aTile.scalarized.1dim, i32 0, i32 51), align 4
; FCHECK-NEXT: store i32 [[DOTI07]], ptr [[AFRAGPACKED_I_SCALARIZE]], align 16
; FCHECK-NEXT: store i32 [[DOTI119]], ptr [[AFRAGPACKED_I_SCALARIZE_I14]], align 4
; FCHECK-NEXT: store i32 [[DOTI2211]], ptr [[AFRAGPACKED_I_SCALARIZE_I25]], align 8
; FCHECK-NEXT: store i32 [[DOTI3313]], ptr [[AFRAGPACKED_I_SCALARIZE_I36]], align 4
;
; CHECK-NEXT: ret void
entry:
%aFragPacked.i = alloca <4 x i32>, align 16
%0 = load <4 x i32>, ptr addrspace(3) getelementptr inbounds ([10 x <4 x i32>], ptr addrspace(3) getelementptr inbounds ([10 x [10 x <4 x i32>]], ptr addrspace(3) @aTile, i32 0, i32 1), i32 0, i32 2), align 16
store <4 x i32> %0, ptr %aFragPacked.i, align 16
ret void
}

define void @Main() {
; CHECK-LABEL: define void @Main() {
; CHECK-NEXT: [[ENTRY:.*:]]
; CHECK-NEXT: [[BFRAGPACKED_I:%.*]] = alloca i32, align 16
;
; SCHECK-NEXT: [[TMP0:%.*]] = getelementptr inbounds [10 x i32], ptr addrspace(3) getelementptr inbounds ([10 x [10 x i32]], ptr addrspace(3) @bTile, i32 0, i32 1), i32 0, i32 1
; SCHECK-NEXT: [[TMP1:%.*]] = load i32, ptr addrspace(3) [[TMP0]], align 16
; SCHECK-NEXT: store i32 [[TMP1]], ptr [[BFRAGPACKED_I]], align 16
;
; FCHECK-NEXT: [[TMP0:%.*]] = load i32, ptr addrspace(3) getelementptr inbounds ([100 x i32], ptr addrspace(3) @bTile.1dim, i32 0, i32 11), align 16
; FCHECK-NEXT: store i32 [[TMP0]], ptr [[BFRAGPACKED_I]], align 16
;
; CHECK-NEXT: ret void
entry:
%bFragPacked.i = alloca i32, align 16
%0 = load i32, ptr addrspace(3) getelementptr inbounds ([10 x i32], ptr addrspace(3) getelementptr inbounds ([10 x [10 x i32]], ptr addrspace(3) @bTile, i32 0, i32 1), i32 0, i32 1), align 16
store i32 %0, ptr %bFragPacked.i, align 16
ret void
}

define void @global_nested_geps_3d() {
; CHECK-LABEL: define void @global_nested_geps_3d() {
; SCHECK-NEXT: [[TMP1:%.*]] = getelementptr inbounds <2 x i32>, ptr getelementptr inbounds ([2 x <2 x i32>], ptr getelementptr inbounds ([2 x [2 x [2 x i32]]], ptr @cTile.scalarized, i32 0, i32 1), i32 0, i32 1), i32 0, i32 1
; SCHECK-NEXT: [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 4
;
; FCHECK-NEXT: [[TMP1:%.*]] = load i32, ptr getelementptr inbounds ([8 x i32], ptr @cTile.scalarized.1dim, i32 0, i32 7), align 4
;
; CHECK-NEXT: ret void
%1 = load i32, i32* getelementptr inbounds (<2 x i32>, <2 x i32>* getelementptr inbounds ([2 x <2 x i32>], [2 x <2 x i32>]* getelementptr inbounds ([2 x [2 x <2 x i32>]], [2 x [2 x <2 x i32>]]* @cTile, i32 0, i32 1), i32 0, i32 1), i32 0, i32 1), align 4
ret void
}

define void @global_nested_geps_4d() {
; CHECK-LABEL: define void @global_nested_geps_4d() {
; SCHECK-NEXT: [[TMP1:%.*]] = getelementptr inbounds <2 x i32>, ptr getelementptr inbounds ([2 x <2 x i32>], ptr getelementptr inbounds ([2 x [2 x <2 x i32>]], ptr getelementptr inbounds ([2 x [2 x [2 x [2 x i32]]]], ptr @dTile.scalarized, i32 0, i32 1), i32 0, i32 1), i32 0, i32 1), i32 0, i32 1
; SCHECK-NEXT: [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 4
;
; FCHECK-NEXT: [[TMP1:%.*]] = load i32, ptr getelementptr inbounds ([16 x i32], ptr @dTile.scalarized.1dim, i32 0, i32 15), align 4
;
; CHECK-NEXT: ret void
%1 = load i32, i32* getelementptr inbounds (<2 x i32>, <2 x i32>* getelementptr inbounds ([2 x <2 x i32>], [2 x <2 x i32>]* getelementptr inbounds ([2 x [2 x <2 x i32>]], [2 x [2 x <2 x i32>]]* getelementptr inbounds ([2 x [2 x [2 x <2 x i32>]]], [2 x [2 x [2 x <2 x i32>]]]* @dTile, i32 0, i32 1), i32 0, i32 1), i32 0, i32 1), i32 0, i32 1), align 4
ret void
}