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 1 commit
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
73 changes: 73 additions & 0 deletions llvm/lib/Target/DirectX/DXILDataScalarization.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -300,11 +300,84 @@ bool DataScalarizerVisitor::visitExtractElementInst(ExtractElementInst &EEI) {
return replaceDynamicExtractElementInst(EEI);
}

static void buildConstExprGEPChain(GetElementPtrInst &GEPI, Value *CurrentPtr,
SmallVector<ConstantExpr *, 4> &GEPChain,
IRBuilder<> &Builder) {
// Process the rest of the chain in reverse order (skipping the innermost)
for (int I = GEPChain.size() - 2; I >= 0; I--) {
ConstantExpr *CE = GEPChain[I];
GetElementPtrInst *GEPInst =
cast<GetElementPtrInst>(CE->getAsInstruction());
GEPInst->insertBefore(GEPI.getIterator());
SmallVector<Value *, MaxVecSize> CurrIndices(GEPInst->indices());

// Create a new GEP instruction
Type *SourceTy = GEPInst->getSourceElementType();
CurrentPtr =
Builder.CreateGEP(SourceTy, CurrentPtr, CurrIndices, GEPInst->getName(),
GEPInst->getNoWrapFlags());

// If this is the outermost GEP, update the main GEPI
if (I == 0) {
GEPI.setOperand(GEPI.getPointerOperandIndex(), CurrentPtr);
}

// Clean up the temporary instruction
GEPInst->eraseFromParent();
}
}

bool DataScalarizerVisitor::visitGetElementPtrInst(GetElementPtrInst &GEPI) {
Value *PtrOperand = GEPI.getPointerOperand();
Type *OrigGEPType = GEPI.getSourceElementType();
Type *NewGEPType = OrigGEPType;
bool NeedsTransform = false;
// Check if the pointer operand is a ConstantExpr GEP
if (auto *PtrOpGEPCE = dyn_cast<ConstantExpr>(PtrOperand);
PtrOpGEPCE && PtrOpGEPCE->getOpcode() == Instruction::GetElementPtr) {

// Collect all nested GEPs in the chain
SmallVector<ConstantExpr *, 4> GEPChain;
Value *BasePointer = PtrOpGEPCE->getOperand(0);
GEPChain.push_back(PtrOpGEPCE);

// Walk up the chain to find all nested GEPs and the base pointer
while (auto *NextGEP = dyn_cast<ConstantExpr>(BasePointer)) {
if (NextGEP->getOpcode() != Instruction::GetElementPtr)
break;

GEPChain.push_back(NextGEP);
BasePointer = NextGEP->getOperand(0);
}

// Check if the base pointer is a global that needs replacement
if (GlobalVariable *NewGlobal = lookupReplacementGlobal(BasePointer)) {
IRBuilder<> Builder(&GEPI);

// Create a new GEP for the innermost GEP (last in the chain)
ConstantExpr *InnerGEPCE = GEPChain.back();
GetElementPtrInst *InnerGEP =
cast<GetElementPtrInst>(InnerGEPCE->getAsInstruction());
InnerGEP->insertBefore(GEPI.getIterator());

SmallVector<Value *, MaxVecSize> Indices(InnerGEP->indices());
Type *NewGEPType = NewGlobal->getValueType();
Value *NewInnerGEP =
Builder.CreateGEP(NewGEPType, NewGlobal, Indices, InnerGEP->getName(),
InnerGEP->getNoWrapFlags());

// If there's only one GEP in the chain, update the main GEPI directly
if (GEPChain.size() == 1)
GEPI.setOperand(GEPI.getPointerOperandIndex(), NewInnerGEP);
else
// For multiple GEPs, we need to create a chain of GEPs
buildConstExprGEPChain(GEPI, NewInnerGEP, GEPChain, Builder);

// Clean up the innermost GEP
InnerGEP->eraseFromParent();
return true;
}
}

if (GlobalVariable *NewGlobal = lookupReplacementGlobal(PtrOperand)) {
NewGEPType = NewGlobal->getValueType();
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
}
Loading