Skip to content

[flang][OpenMP] Use an enum for DoConcurrentConversionPass::mapTo #96

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.

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
Jun 4, 2024
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
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
7 changes: 2 additions & 5 deletions flang/include/flang/Frontend/CodeGenOptions.h
Original file line number Diff line number Diff line change
Expand Up @@ -15,6 +15,7 @@
#ifndef FORTRAN_FRONTEND_CODEGENOPTIONS_H
#define FORTRAN_FRONTEND_CODEGENOPTIONS_H

#include "flang/Optimizer/Transforms/Utils.h"
#include "llvm/Frontend/Debug/Options.h"
#include "llvm/Frontend/Driver/CodeGenOptions.h"
#include "llvm/Support/CodeGen.h"
Expand Down Expand Up @@ -131,11 +132,7 @@ class CodeGenOptions : public CodeGenOptionsBase {

/// Optionally map `do concurrent` loops to OpenMP. This is only valid of
/// OpenMP is enabled.
enum class DoConcurrentMappingKind {
DCMK_None, // Do not lower `do concurrent` to OpenMP.
DCMK_Host, // Lower to run in parallel on the CPU.
DCMK_Device // Lower to run in parallel on the GPU.
};
using DoConcurrentMappingKind = fir::omp::DoConcurrentMappingKind;

// Define accessors/mutators for code generation options of enumeration type.
#define CODEGENOPT(Name, Bits, Default)
Expand Down
2 changes: 2 additions & 0 deletions flang/include/flang/Optimizer/Transforms/Passes.h
Original file line number Diff line number Diff line change
Expand Up @@ -10,10 +10,12 @@
#define FORTRAN_OPTIMIZER_TRANSFORMS_PASSES_H

#include "flang/Optimizer/Dialect/FIROps.h"
#include "flang/Optimizer/Transforms/Utils.h"
#include "mlir/Dialect/LLVMIR/LLVMAttrs.h"
#include "mlir/Dialect/OpenMP/OpenMPDialect.h"
#include "mlir/Pass/Pass.h"
#include "mlir/Pass/PassRegistry.h"

#include <memory>

namespace mlir {
Expand Down
18 changes: 13 additions & 5 deletions flang/include/flang/Optimizer/Transforms/Passes.td
Original file line number Diff line number Diff line change
Expand Up @@ -15,6 +15,7 @@
#define FLANG_OPTIMIZER_TRANSFORMS_PASSES

include "mlir/Pass/PassBase.td"
include "mlir/IR/EnumAttr.td"

def AbstractResultOpt
: Pass<"abstract-result"> {
Expand Down Expand Up @@ -413,16 +414,23 @@ def DoConcurrentConversionPass : Pass<"fopenmp-do-concurrent-conversion", "mlir:

Still to TODO:
- More extensive testing.
- Mapping to `target teams distribute parallel do`.
- Allowing the user to control mapping behavior: either to the host or
target.
}];

let dependentDialects = ["mlir::omp::OpenMPDialect"];

let options = [
Option<"mapTo", "map-to", "std::string", "",
"Try to map `do concurrent` loops to OpenMP (on host or device)">,
Option<"mapTo", "map-to",
"fir::omp::DoConcurrentMappingKind",
/*default=*/"fir::omp::DoConcurrentMappingKind::DCMK_None",
"Try to map `do concurrent` loops to OpenMP (on host or device)",
[{::llvm::cl::values(
clEnumValN(fir::omp::DoConcurrentMappingKind::DCMK_None,
"none", "Do not lower `do concurrent` to OpenMP"),
clEnumValN(fir::omp::DoConcurrentMappingKind::DCMK_Host,
"host", "Lower to run in parallel on the CPU"),
clEnumValN(fir::omp::DoConcurrentMappingKind::DCMK_Device,
"device", "Lower to run in parallel on the GPU")
)}]>,
];
}

Expand Down
12 changes: 12 additions & 0 deletions flang/include/flang/Optimizer/Transforms/Utils.h
Original file line number Diff line number Diff line change
Expand Up @@ -13,8 +13,13 @@
#ifndef FORTRAN_OPTIMIZER_TRANSFORMS_UTILS_H
#define FORTRAN_OPTIMIZER_TRANSFORMS_UTILS_H

#include "mlir/IR/Location.h"
#include "mlir/IR/Value.h"

namespace fir {

class FirOpBuilder;

using MinlocBodyOpGeneratorTy = llvm::function_ref<mlir::Value(
fir::FirOpBuilder &, mlir::Location, const mlir::Type &, mlir::Value,
mlir::Value, mlir::Value, const llvm::SmallVectorImpl<mlir::Value> &)>;
Expand All @@ -33,6 +38,13 @@ void genMinMaxlocReductionLoop(fir::FirOpBuilder &builder, mlir::Value array,
mlir::Type maskElemType, mlir::Value resultArr,
bool maskMayBeLogicalScalar);

namespace omp {
enum class DoConcurrentMappingKind {
DCMK_None, // Do not lower `do concurrent` to OpenMP.
DCMK_Host, // Lower to run in parallel on the CPU.
DCMK_Device // Lower to run in parallel on the GPU.
};
}
} // namespace fir

#endif // FORTRAN_OPTIMIZER_TRANSFORMS_UTILS_H
9 changes: 6 additions & 3 deletions flang/lib/Optimizer/Transforms/DoConcurrentConversion.cpp
Original file line number Diff line number Diff line change
Expand Up @@ -581,15 +581,17 @@ class DoConcurrentConversionPass

auto *context = &getContext();

if (mapTo != "host" && mapTo != "device") {
if (mapTo != fir::omp::DoConcurrentMappingKind::DCMK_Host &&
mapTo != fir::omp::DoConcurrentMappingKind::DCMK_Device) {
mlir::emitWarning(mlir::UnknownLoc::get(context),
"DoConcurrentConversionPass: invalid `map-to` value. "
"Valid values are: `host` or `device`");
return;
}

mlir::RewritePatternSet patterns(context);
patterns.insert<DoConcurrentConversion>(context, mapTo == "device");
patterns.insert<DoConcurrentConversion>(
context, mapTo == fir::omp::DoConcurrentMappingKind::DCMK_Device);
mlir::ConversionTarget target(*context);
target.addLegalDialect<fir::FIROpsDialect, hlfir::hlfirDialect,
mlir::arith::ArithDialect, mlir::func::FuncDialect,
Expand All @@ -611,7 +613,8 @@ class DoConcurrentConversionPass
std::unique_ptr<mlir::Pass>
fir::createDoConcurrentConversionPass(bool mapToDevice) {
DoConcurrentConversionPassOptions options;
options.mapTo = mapToDevice ? "device" : "host";
options.mapTo = mapToDevice ? fir::omp::DoConcurrentMappingKind::DCMK_Device
: fir::omp::DoConcurrentMappingKind::DCMK_Host;

return std::make_unique<DoConcurrentConversionPass>(options);
}