Skip to content

[acc] Add firstprivate/private/reduction to acc.kernels #170387

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
SusanTan merged 3 commits into from
Dec 3, 2025
Merged

[acc] Add firstprivate/private/reduction to acc.kernels #170387

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
77e571e
add firstprivate/private to acc kernel
SusanTan Dec 2, 2025
93b2030
add reduction
SusanTan Dec 2, 2025
ab6ab93
add split
SusanTan Dec 3, 2025
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
8 changes: 3 additions & 5 deletions flang/lib/Lower/OpenACC.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -3024,11 +3024,9 @@ static Op createComputeOp(
}
addOperand(operands, operandSegments, ifCond);
addOperand(operands, operandSegments, selfCond);
if constexpr (!std::is_same_v<Op, mlir::acc::KernelsOp>) {
addOperands(operands, operandSegments, reductionOperands);
addOperands(operands, operandSegments, privateOperands);
addOperands(operands, operandSegments, firstprivateOperands);
}
addOperands(operands, operandSegments, reductionOperands);
addOperands(operands, operandSegments, privateOperands);
addOperands(operands, operandSegments, firstprivateOperands);
addOperands(operands, operandSegments, dataClauseOperands);

Op computeOp;
Expand Down
28 changes: 21 additions & 7 deletions mlir/include/mlir/Dialect/OpenACC/OpenACCOps.td
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2002,8 +2002,7 @@ def OpenACC_KernelsOp : OpenACC_Op<"kernels",
corresponding `device_type` attributes must be modified as well.
}];

let arguments = (ins
Variadic<IntOrIndex>:$asyncOperands,
let arguments = (ins Variadic<IntOrIndex>:$asyncOperands,
OptionalAttr<DeviceTypeArrayAttr>:$asyncOperandsDeviceType,
OptionalAttr<DeviceTypeArrayAttr>:$asyncOnly,
Variadic<IntOrIndex>:$waitOperands,
Expand All @@ -2018,12 +2017,12 @@ def OpenACC_KernelsOp : OpenACC_Op<"kernels",
OptionalAttr<DeviceTypeArrayAttr>:$numWorkersDeviceType,
Variadic<IntOrIndex>:$vectorLength,
OptionalAttr<DeviceTypeArrayAttr>:$vectorLengthDeviceType,
Optional<I1>:$ifCond,
Optional<I1>:$selfCond,
UnitAttr:$selfAttr,
Optional<I1>:$ifCond, Optional<I1>:$selfCond, UnitAttr:$selfAttr,
Variadic<OpenACC_AnyPointerOrMappableType>:$reductionOperands,
Variadic<OpenACC_AnyPointerOrMappableType>:$privateOperands,
Variadic<OpenACC_AnyPointerOrMappableType>:$firstprivateOperands,
Variadic<OpenACC_AnyPointerOrMappableType>:$dataClauseOperands,
OptionalAttr<DefaultValueAttr>:$defaultAttr,
UnitAttr:$combined);
OptionalAttr<DefaultValueAttr>:$defaultAttr, UnitAttr:$combined);

let regions = (region AnyRegion:$region);

Expand Down Expand Up @@ -2111,6 +2110,18 @@ def OpenACC_KernelsOp : OpenACC_Op<"kernels",
/// types.
void addWaitOperands(MLIRContext *, bool hasDevnum, mlir::ValueRange,
llvm::ArrayRef<DeviceType>);

/// Adds a private clause variable to this operation, including its recipe.
void addPrivatization(MLIRContext *, mlir::acc::PrivateOp op,
mlir::acc::PrivateRecipeOp recipe);
/// Adds a firstprivate clause variable to this operation, including its
/// recipe.
void addFirstPrivatization(MLIRContext *, mlir::acc::FirstprivateOp op,
mlir::acc::FirstprivateRecipeOp recipe);
/// Adds a reduction clause variable to this operation, including its
/// recipe.
void addReduction(MLIRContext *, mlir::acc::ReductionOp op,
mlir::acc::ReductionRecipeOp recipe);
}];

let assemblyFormat = [{
Expand All @@ -2119,17 +2130,20 @@ def OpenACC_KernelsOp : OpenACC_Op<"kernels",
`dataOperands` `(` $dataClauseOperands `:` type($dataClauseOperands) `)`
| `async` `` custom<DeviceTypeOperandsWithKeywordOnly>($asyncOperands,
type($asyncOperands), $asyncOperandsDeviceType, $asyncOnly)
| `firstprivate` `(` $firstprivateOperands `:` type($firstprivateOperands) `)`
| `num_gangs` `(` custom<NumGangs>($numGangs,
type($numGangs), $numGangsDeviceType, $numGangsSegments) `)`
| `num_workers` `(` custom<DeviceTypeOperands>($numWorkers,
type($numWorkers), $numWorkersDeviceType) `)`
| `private` `(` $privateOperands `:` type($privateOperands) `)`
Copy link
Contributor

@razvanlupusoru razvanlupusoru Dec 2, 2025

Choose a reason for hiding this comment

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

Should reduction also be added for completeness/consistency?

Copy link
Contributor Author

@SusanTan SusanTan Dec 2, 2025

Choose a reason for hiding this comment

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

was considering that... will add it!

| `vector_length` `(` custom<DeviceTypeOperands>($vectorLength,
type($vectorLength), $vectorLengthDeviceType) `)`
| `wait` `` custom<WaitClause>($waitOperands, type($waitOperands),
$waitOperandsDeviceType, $waitOperandsSegments, $hasWaitDevnum,
$waitOnly)
| `self` `(` $selfCond `)`
| `if` `(` $ifCond `)`
| `reduction` `(` $reductionOperands `:` type($reductionOperands) `)`
)
$region attr-dict-with-keyword
}];
Expand Down
21 changes: 21 additions & 0 deletions mlir/lib/Dialect/OpenACC/IR/OpenACC.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2675,6 +2675,27 @@ LogicalResult acc::KernelsOp::verify() {
return checkDataOperands<acc::KernelsOp>(*this, getDataClauseOperands());
}

void acc::KernelsOp::addPrivatization(MLIRContext *context,
mlir::acc::PrivateOp op,
mlir::acc::PrivateRecipeOp recipe) {
op.setRecipeAttr(mlir::SymbolRefAttr::get(context, recipe.getSymName()));
getPrivateOperandsMutable().append(op.getResult());
}

void acc::KernelsOp::addFirstPrivatization(
MLIRContext *context, mlir::acc::FirstprivateOp op,
mlir::acc::FirstprivateRecipeOp recipe) {
op.setRecipeAttr(mlir::SymbolRefAttr::get(context, recipe.getSymName()));
getFirstprivateOperandsMutable().append(op.getResult());
}

void acc::KernelsOp::addReduction(MLIRContext *context,
mlir::acc::ReductionOp op,
mlir::acc::ReductionRecipeOp recipe) {
op.setRecipeAttr(mlir::SymbolRefAttr::get(context, recipe.getSymName()));
getReductionOperandsMutable().append(op.getResult());
}

void acc::KernelsOp::addNumWorkersOperand(
MLIRContext *context, mlir::Value newValue,
llvm::ArrayRef<DeviceType> effectiveDeviceTypes) {
Expand Down
82 changes: 82 additions & 0 deletions mlir/test/Dialect/OpenACC/ops.mlir
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -731,6 +731,59 @@ func.func @testserialop(%a: memref<10xf32>, %b: memref<10xf32>, %c: memref<10x10

// -----

// Test acc.kernels with private and firstprivate operands, similar to acc.serial.

acc.private.recipe @privatization_memref_10_f32 : memref<10xf32> init {
^bb0(%arg0: memref<10xf32>):
%0 = memref.alloc() : memref<10xf32>
acc.yield %0 : memref<10xf32>
} destroy {
^bb0(%arg0: memref<10xf32>):
memref.dealloc %arg0 : memref<10xf32>
acc.terminator
}

acc.private.recipe @privatization_memref_10_10_f32 : memref<10x10xf32> init {
^bb0(%arg0: memref<10x10xf32>):
%1 = memref.alloc() : memref<10x10xf32>
acc.yield %1 : memref<10x10xf32>
} destroy {
^bb0(%arg0: memref<10x10xf32>):
memref.dealloc %arg0 : memref<10x10xf32>
acc.terminator
}

acc.firstprivate.recipe @firstprivatization_memref_10xf32 : memref<10xf32> init {
^bb0(%arg0: memref<10xf32>):
%2 = memref.alloca() : memref<10xf32>
acc.yield %2 : memref<10xf32>
} copy {
^bb0(%arg0: memref<10xf32>, %arg1: memref<10xf32>):
memref.copy %arg0, %arg1 : memref<10xf32> to memref<10xf32>
acc.terminator
} destroy {
^bb0(%arg0: memref<10xf32>):
acc.terminator
}

func.func @testkernelspriv(%a: memref<10xf32>, %b: memref<10xf32>, %c: memref<10x10xf32>) -> () {
%priv_a = acc.private varPtr(%a : memref<10xf32>) recipe(@privatization_memref_10_f32) -> memref<10xf32>
%priv_c = acc.private varPtr(%c : memref<10x10xf32>) recipe(@privatization_memref_10_10_f32) -> memref<10x10xf32>
%firstp = acc.firstprivate varPtr(%b : memref<10xf32>) varType(tensor<10xf32>) recipe(@firstprivatization_memref_10xf32) -> memref<10xf32>
acc.kernels firstprivate(%firstp : memref<10xf32>) private(%priv_a, %priv_c : memref<10xf32>, memref<10x10xf32>) {
}
return
}

// CHECK-LABEL: func.func @testkernelspriv(
// CHECK: %[[PRIV_A:.*]] = acc.private varPtr(%{{.*}} : memref<10xf32>) recipe(@privatization_memref_10_f32) -> memref<10xf32>
// CHECK: %[[PRIV_C:.*]] = acc.private varPtr(%{{.*}} : memref<10x10xf32>) recipe(@privatization_memref_10_10_f32) -> memref<10x10xf32>
// CHECK: %[[FIRSTP:.*]] = acc.firstprivate varPtr(%{{.*}} : memref<10xf32>) varType(tensor<10xf32>) recipe(@firstprivatization_memref_10xf32) -> memref<10xf32>
// CHECK: acc.kernels firstprivate(%[[FIRSTP]] : memref<10xf32>) private(%[[PRIV_A]], %[[PRIV_C]] : memref<10xf32>, memref<10x10xf32>) {
// CHECK-NEXT: }

// -----

func.func @testdataop(%a: memref<f32>, %b: memref<f32>, %c: memref<f32>) -> () {
%ifCond = arith.constant true

Expand Down Expand Up @@ -1602,6 +1655,35 @@ func.func @acc_reduc_test(%a : memref<i64>) -> () {

// -----

acc.reduction.recipe @reduction_add_memref_i64 : memref<i64> reduction_operator <add> init {
^bb0(%arg0: memref<i64>):
%c0_i64 = arith.constant 0 : i64
%alloca = memref.alloca() : memref<i64>
memref.store %c0_i64, %alloca[] : memref<i64>
acc.yield %alloca : memref<i64>
} combiner {
^bb0(%arg0: memref<i64>, %arg1: memref<i64>):
%0 = memref.load %arg0[] : memref<i64>
%1 = memref.load %arg1[] : memref<i64>
%2 = arith.addi %0, %1 : i64
memref.store %2, %arg0[] : memref<i64>
acc.yield %arg0 : memref<i64>
}

func.func @acc_kernels_reduc_test(%a : memref<i64>) -> () {
%reduction_a = acc.reduction varPtr(%a : memref<i64>) recipe(@reduction_add_memref_i64) -> memref<i64>
acc.kernels reduction(%reduction_a : memref<i64>) {
}
return
}

// CHECK-LABEL: func.func @acc_kernels_reduc_test(
// CHECK-SAME: %[[ARG0:.*]]: memref<i64>)
// CHECK: %[[REDUCTION_A:.*]] = acc.reduction varPtr(%[[ARG0]] : memref<i64>) recipe(@reduction_add_memref_i64) -> memref<i64>
// CHECK-NEXT: acc.kernels reduction(%[[REDUCTION_A]] : memref<i64>)

// -----

func.func @testdeclareop(%a: memref<f32>, %b: memref<f32>, %c: memref<f32>) -> () {
%0 = acc.copyin varPtr(%a : memref<f32>) -> memref<f32>
// copyin(zero)
Expand Down
Loading