Skip to content

[AMDGPU] Use removeFnAttrFromReachable in lower-module-lds pass. #92686

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 1 commit into from
May 20, 2024
Merged

[AMDGPU] Use removeFnAttrFromReachable in lower-module-lds pass. #92686

Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
44 changes: 1 addition & 43 deletions llvm/lib/Target/AMDGPU/AMDGPULowerModuleLDSPass.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -862,48 +862,6 @@ class AMDGPULowerModuleLDS {
return N;
}

/// Strip "amdgpu-no-lds-kernel-id" from any functions where we may have
/// introduced its use. If AMDGPUAttributor ran prior to the pass, we inferred
/// the lack of llvm.amdgcn.lds.kernel.id calls.
void removeNoLdsKernelIdFromReachable(CallGraph &CG, Function *KernelRoot) {
KernelRoot->removeFnAttr("amdgpu-no-lds-kernel-id");

SmallVector<Function *> WorkList({CG[KernelRoot]->getFunction()});
SmallPtrSet<Function *, 8> Visited;
bool SeenUnknownCall = false;

while (!WorkList.empty()) {
Function *F = WorkList.pop_back_val();

for (auto &CallRecord : *CG[F]) {
if (!CallRecord.second)
continue;

Function *Callee = CallRecord.second->getFunction();
if (!Callee) {
if (!SeenUnknownCall) {
SeenUnknownCall = true;

// If we see any indirect calls, assume nothing about potential
// targets.
// TODO: This could be refined to possible LDS global users.
for (auto &ExternalCallRecord : *CG.getExternalCallingNode()) {
Function *PotentialCallee =
ExternalCallRecord.second->getFunction();
assert(PotentialCallee);
if (!isKernelLDS(PotentialCallee))
PotentialCallee->removeFnAttr("amdgpu-no-lds-kernel-id");
}
}
} else {
Callee->removeFnAttr("amdgpu-no-lds-kernel-id");
if (Visited.insert(Callee).second)
WorkList.push_back(Callee);
}
}
}
}

DenseMap<Function *, GlobalVariable *> lowerDynamicLDSVariables(
Module &M, LDSUsesInfoTy &LDSUsesInfo,
DenseSet<Function *> const &KernelsThatIndirectlyAllocateDynamicLDS,
Expand Down Expand Up @@ -1059,7 +1017,7 @@ class AMDGPULowerModuleLDS {
//
// TODO: We could filter out subgraphs that do not access LDS globals.
for (Function *F : KernelsThatAllocateTableLDS)
removeNoLdsKernelIdFromReachable(CG, F);
removeFnAttrFromReachable(CG, F, "amdgpu-no-lds-kernel-id");
}

DenseMap<Function *, GlobalVariable *> KernelToCreatedDynamicLDS =
Expand Down
Loading