Skip to content

[flang][cuda] Lower syncwarp to NVVM intrinsic #126164

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
Feb 7, 2025
Merged

[flang][cuda] Lower syncwarp to NVVM intrinsic #126164

Show file tree
Hide file tree
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
1 change: 1 addition & 0 deletions flang/include/flang/Optimizer/Builder/IntrinsicCall.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -406,6 +406,7 @@ struct IntrinsicLibrary {
mlir::Value genSyncThreadsAnd(mlir::Type, llvm::ArrayRef<mlir::Value>);
mlir::Value genSyncThreadsCount(mlir::Type, llvm::ArrayRef<mlir::Value>);
mlir::Value genSyncThreadsOr(mlir::Type, llvm::ArrayRef<mlir::Value>);
void genSyncWarp(llvm::ArrayRef<fir::ExtendedValue>);
fir::ExtendedValue genSystem(std::optional<mlir::Type>,
mlir::ArrayRef<fir::ExtendedValue> args);
void genSystemClock(llvm::ArrayRef<fir::ExtendedValue>);
Expand Down
13 changes: 13 additions & 0 deletions flang/lib/Optimizer/Builder/IntrinsicCall.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -680,6 +680,7 @@ static constexpr IntrinsicHandler handlers[]{
{"syncthreads_and", &I::genSyncThreadsAnd, {}, /*isElemental=*/false},
{"syncthreads_count", &I::genSyncThreadsCount, {}, /*isElemental=*/false},
{"syncthreads_or", &I::genSyncThreadsOr, {}, /*isElemental=*/false},
{"syncwarp", &I::genSyncWarp, {}, /*isElemental=*/false},
{"system",
&I::genSystem,
{{{"command", asBox}, {"exitstat", asBox, handleDynamicOptional}}},
Expand Down Expand Up @@ -7704,6 +7705,18 @@ IntrinsicLibrary::genSyncThreadsOr(mlir::Type resultType,
return builder.create<fir::CallOp>(loc, funcOp, args).getResult(0);
}

// SYNCWARP
void IntrinsicLibrary::genSyncWarp(llvm::ArrayRef<fir::ExtendedValue> args) {
assert(args.size() == 1);
constexpr llvm::StringLiteral funcName = "llvm.nvvm.bar.warp.sync";
mlir::Value mask = fir::getBase(args[0]);
mlir::FunctionType funcType =
mlir::FunctionType::get(builder.getContext(), {mask.getType()}, {});
auto funcOp = builder.createFunction(loc, funcName, funcType);
llvm::SmallVector<mlir::Value> argsList{mask};
builder.create<fir::CallOp>(loc, funcOp, argsList);
}

// SYSTEM
fir::ExtendedValue
IntrinsicLibrary::genSystem(std::optional<mlir::Type> resultType,
Expand Down
2 changes: 1 addition & 1 deletion flang/module/cudadevice.f90
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -49,7 +49,7 @@ attributes(device) integer function syncthreads_or(value)
public :: syncthreads_or

interface
attributes(device) subroutine syncwarp(mask) bind(c, name='__syncwarp')
attributes(device) subroutine syncwarp(mask)
integer, value :: mask
end subroutine
end interface
Expand Down
6 changes: 3 additions & 3 deletions flang/test/Lower/CUDA/cuda-device-proc.cuf
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -47,7 +47,7 @@ end

! CHECK-LABEL: func.func @_QPdevsub() attributes {cuf.proc_attr = #cuf.cuda_proc<global>}
! CHECK: fir.call @llvm.nvvm.barrier0() fastmath<contract> : () -> ()
! CHECK: fir.call @__syncwarp(%{{.*}}) proc_attrs<bind_c> fastmath<contract> : (i32) -> ()
! CHECK: fir.call @llvm.nvvm.bar.warp.sync(%c1{{.*}}) fastmath<contract> : (i32) -> ()
! CHECK: fir.call @llvm.nvvm.membar.gl() fastmath<contract> : () -> ()
! CHECK: fir.call @llvm.nvvm.membar.cta() fastmath<contract> : () -> ()
! CHECK: fir.call @llvm.nvvm.membar.sys() fastmath<contract> : () -> ()
Expand Down Expand Up @@ -102,13 +102,13 @@ end
! CHECK-LABEL: func.func @_QPhost1()
! CHECK: cuf.kernel
! CHECK: fir.call @llvm.nvvm.barrier0() fastmath<contract> : () -> ()
! CHECK: fir.call @__syncwarp(%c1{{.*}}) proc_attrs<bind_c> fastmath<contract> : (i32) -> ()
! CHECK: fir.call @llvm.nvvm.bar.warp.sync(%c1{{.*}}) fastmath<contract> : (i32) -> ()
! CHECK: fir.call @llvm.nvvm.barrier0.and(%c1{{.*}}) fastmath<contract> : (i32) -> i32
! CHECK: fir.call @llvm.nvvm.barrier0.popc(%c1{{.*}}) fastmath<contract> : (i32) -> i32
! CHECK: fir.call @llvm.nvvm.barrier0.or(%c1{{.*}}) fastmath<contract> : (i32) -> i32

! CHECK: func.func private @llvm.nvvm.barrier0()
! CHECK: func.func private @__syncwarp(i32) attributes {cuf.proc_attr = #cuf.cuda_proc<device>, fir.bindc_name = "__syncwarp", fir.proc_attrs = #fir.proc_attrs<bind_c>}
! CHECK: func.func private @llvm.nvvm.bar.warp.sync(i32)
! CHECK: func.func private @llvm.nvvm.membar.gl()
! CHECK: func.func private @llvm.nvvm.membar.cta()
! CHECK: func.func private @llvm.nvvm.membar.sys()
Expand Down