Skip to content

[flang][cuda] Adapt ExternalNameConversion to work in gpu module #117039

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
Nov 20, 2024
Merged

[flang][cuda] Adapt ExternalNameConversion to work in gpu module #117039

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
fe22808
[flang][cuda] Adapat ExternalNameConversion to work in gpu module
clementval Nov 20, 2024
84f0341
Remove braces
clementval Nov 20, 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
47 changes: 29 additions & 18 deletions flang/lib/Optimizer/Transforms/ExternalNameConversion.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -12,6 +12,7 @@
#include "flang/Optimizer/Dialect/FIROpsSupport.h"
#include "flang/Optimizer/Support/InternalNames.h"
#include "flang/Optimizer/Transforms/Passes.h"
#include "mlir/Dialect/GPU/IR/GPUDialect.h"
#include "mlir/IR/Attributes.h"
#include "mlir/IR/SymbolTable.h"
#include "mlir/Pass/Pass.h"
Expand Down Expand Up @@ -58,26 +59,36 @@ void ExternalNameConversionPass::runOnOperation() {
auto *context = &getContext();

llvm::DenseMap<mlir::StringAttr, mlir::FlatSymbolRefAttr> remappings;
// Update names of external Fortran functions and names of Common Block
// globals.
for (auto &funcOrGlobal : op->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);

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);
}
}
}
}
};

// Update names of external Fortran functions and names of Common Block
// 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;
Expand Down
13 changes: 13 additions & 0 deletions flang/test/Fir/CUDA/cuda-extranal-mangling.mlir
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,13 @@
// RUN: fir-opt --split-input-file --external-name-interop %s | FileCheck %s

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

// CHECK-LABEL: gpu.func @_QPfoo
// CHECK: fir.call @threadfence_()
// CHECK: func.func private @threadfence_()
Loading