Skip to content

[flang][cuda] Additional update to ExternalNameConversion #119276

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 2 commits into from
Dec 10, 2024
Merged

[flang][cuda] Additional update to ExternalNameConversion #119276

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
a136c79
[flang][cuda] Additional update to ExternalNameConversion
clementval Dec 9, 2024
3c336e3
clang-format
clementval Dec 9, 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
59 changes: 34 additions & 25 deletions flang/lib/Optimizer/Transforms/ExternalNameConversion.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -60,23 +60,30 @@ void ExternalNameConversionPass::runOnOperation() {

llvm::DenseMap<mlir::StringAttr, mlir::FlatSymbolRefAttr> remappings;

auto processFctOrGlobal = [&](mlir::Operation &funcOrGlobal) {
auto symName = funcOrGlobal.getAttrOfType<mlir::StringAttr>(
mlir::SymbolTable::getSymbolAttrName());
auto deconstructedName = fir::NameUniquer::deconstruct(symName);
if (fir::NameUniquer::isExternalFacingUniquedName(deconstructedName)) {
auto newName = mangleExternalName(deconstructedName, appendUnderscoreOpt);
auto newAttr = mlir::StringAttr::get(context, newName);
mlir::SymbolTable::setSymbolName(&funcOrGlobal, newAttr);
auto newSymRef = mlir::FlatSymbolRefAttr::get(newAttr);
remappings.try_emplace(symName, newSymRef);
if (llvm::isa<mlir::func::FuncOp>(funcOrGlobal))
funcOrGlobal.setAttr(fir::getInternalFuncNameAttrName(), symName);
}
};

auto renameFuncOrGlobalInModule = [&](mlir::Operation *module) {
for (auto &funcOrGlobal : module->getRegion(0).front()) {
if (llvm::isa<mlir::func::FuncOp>(funcOrGlobal) ||
llvm::isa<fir::GlobalOp>(funcOrGlobal)) {
auto symName = funcOrGlobal.getAttrOfType<mlir::StringAttr>(
mlir::SymbolTable::getSymbolAttrName());
auto deconstructedName = fir::NameUniquer::deconstruct(symName);
if (fir::NameUniquer::isExternalFacingUniquedName(deconstructedName)) {
auto newName =
mangleExternalName(deconstructedName, appendUnderscoreOpt);
auto newAttr = mlir::StringAttr::get(context, newName);
mlir::SymbolTable::setSymbolName(&funcOrGlobal, newAttr);
auto newSymRef = mlir::FlatSymbolRefAttr::get(newAttr);
remappings.try_emplace(symName, newSymRef);
if (llvm::isa<mlir::func::FuncOp>(funcOrGlobal))
funcOrGlobal.setAttr(fir::getInternalFuncNameAttrName(), symName);
}
for (auto &op : module->getRegion(0).front()) {
if (mlir::isa<mlir::func::FuncOp, fir::GlobalOp>(op)) {
processFctOrGlobal(op);
} else if (auto gpuMod = mlir::dyn_cast<mlir::gpu::GPUModuleOp>(op)) {
for (auto &gpuOp : gpuMod.getBodyRegion().front())
if (mlir::isa<mlir::func::FuncOp, fir::GlobalOp,
mlir::gpu::GPUFuncOp>(gpuOp))
processFctOrGlobal(gpuOp);
}
}
};
Expand All @@ -85,23 +92,25 @@ void ExternalNameConversionPass::runOnOperation() {
// globals.
renameFuncOrGlobalInModule(op);

// Do the same in GPU modules.
if (auto mod = mlir::dyn_cast_or_null<mlir::ModuleOp>(*op))
for (auto gpuMod : mod.getOps<mlir::gpu::GPUModuleOp>())
renameFuncOrGlobalInModule(gpuMod);

if (remappings.empty())
return;

// Update all uses of the functions and globals that have been renamed.
op.walk([&remappings](mlir::Operation *nestedOp) {
llvm::SmallVector<std::pair<mlir::StringAttr, mlir::SymbolRefAttr>> updates;
for (const mlir::NamedAttribute &attr : nestedOp->getAttrDictionary())
if (auto symRef = llvm::dyn_cast<mlir::SymbolRefAttr>(attr.getValue()))
if (auto remap = remappings.find(symRef.getRootReference());
remap != remappings.end())
if (auto symRef = llvm::dyn_cast<mlir::SymbolRefAttr>(attr.getValue())) {
if (auto remap = remappings.find(symRef.getLeafReference());
remap != remappings.end()) {
mlir::SymbolRefAttr symAttr = mlir::FlatSymbolRefAttr(remap->second);
if (mlir::isa<mlir::gpu::LaunchFuncOp>(nestedOp))
symAttr = mlir::SymbolRefAttr::get(
symRef.getRootReference(),
{mlir::FlatSymbolRefAttr(remap->second)});
updates.emplace_back(std::pair<mlir::StringAttr, mlir::SymbolRefAttr>{
attr.getName(), mlir::SymbolRefAttr(remap->second)});
attr.getName(), symAttr});
}
}
for (auto update : updates)
nestedOp->setAttr(update.first, update.second);
});
Expand Down
16 changes: 14 additions & 2 deletions flang/test/Fir/CUDA/cuda-extranal-mangling.mlir
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,13 +1,25 @@
// RUN: fir-opt --split-input-file --external-name-interop %s | FileCheck %s

module @mod attributes {gpu.container_module} {

gpu.module @cuda_device_mod {
gpu.func @_QPfoo() {
gpu.func @_QPfoo() kernel {
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Is there a type in the filename ?

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Yes! Good catch! It was from a previous commit. I'll address that

fir.call @_QPthreadfence() fastmath<contract> : () -> ()
gpu.return
}
func.func private @_QPthreadfence() attributes {cuf.proc_attr = #cuf.cuda_proc<device>}
}

// CHECK-LABEL: gpu.func @_QPfoo
func.func @test() -> () {
%0 = llvm.mlir.constant(0 : i64) : i64
%1 = llvm.mlir.constant(0 : i32) : i32
gpu.launch_func @cuda_device_mod::@_QPfoo blocks in (%0, %0, %0) threads in (%0, %0, %0) : i64 dynamic_shared_memory_size %1
return
}

// CHECK-LABEL: gpu.func @foo_()
// CHECK: fir.call @threadfence_()
// CHECK: func.func private @threadfence_()
// CHECK: gpu.launch_func @cuda_device_mod::@foo_

}
Loading