Skip to content

[mlir][acc] Fix async only api on data entry operations #122818

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
Jan 14, 2025
Merged

[mlir][acc] Fix async only api on data entry operations #122818

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
10 changes: 8 additions & 2 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 @@ -445,7 +445,10 @@ class OpenACC_DataEntryOp<string mnemonic, string clause, string extraDescriptio
}
/// Return true if the op has the async attribute for the given device_type.
bool hasAsyncOnly(mlir::acc::DeviceType deviceType) {
for (auto attr : getAsyncOnlyAttr()) {
mlir::ArrayAttr asyncOnly = getAsyncOnlyAttr();
if (!asyncOnly)
return false;
for (auto attr : asyncOnly) {
auto deviceTypeAttr = mlir::dyn_cast<mlir::acc::DeviceTypeAttr>(attr);
if (deviceTypeAttr.getValue() == deviceType)
return true;
Expand Down Expand Up @@ -817,7 +820,10 @@ class OpenACC_DataExitOp<string mnemonic, string clause, string extraDescription
}
/// Return true if the op has the async attribute for the given device_type.
bool hasAsyncOnly(mlir::acc::DeviceType deviceType) {
for (auto attr : getAsyncOnlyAttr()) {
mlir::ArrayAttr asyncOnly = getAsyncOnlyAttr();
if (!asyncOnly)
return false;
for (auto attr : asyncOnly) {
auto deviceTypeAttr = mlir::dyn_cast<mlir::acc::DeviceTypeAttr>(attr);
if (deviceTypeAttr.getValue() == deviceType)
return true;
Expand Down
88 changes: 88 additions & 0 deletions mlir/unittests/Dialect/OpenACC/OpenACCOpsTest.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -77,6 +77,54 @@ TEST_F(OpenACCOpsTest, asyncOnlyTest) {
testAsyncOnly<SerialOp>(b, context, loc, dtypes);
}

template <typename Op>
void testAsyncOnlyDataEntry(OpBuilder &b, MLIRContext &context, Location loc,
llvm::SmallVector<DeviceType> &dtypes) {
auto memrefTy = MemRefType::get({}, b.getI32Type());
OwningOpRef<memref::AllocaOp> varPtrOp =
b.create<memref::AllocaOp>(loc, memrefTy);

TypedValue<PointerLikeType> varPtr =
cast<TypedValue<PointerLikeType>>(varPtrOp->getResult());
OwningOpRef<Op> op = b.create<Op>(loc, varPtr,
/*structured=*/true, /*implicit=*/true);

EXPECT_FALSE(op->hasAsyncOnly());
for (auto d : dtypes)
EXPECT_FALSE(op->hasAsyncOnly(d));

auto dtypeNone = DeviceTypeAttr::get(&context, DeviceType::None);
op->setAsyncOnlyAttr(b.getArrayAttr({dtypeNone}));
EXPECT_TRUE(op->hasAsyncOnly());
EXPECT_TRUE(op->hasAsyncOnly(DeviceType::None));
op->removeAsyncOnlyAttr();

auto dtypeHost = DeviceTypeAttr::get(&context, DeviceType::Host);
op->setAsyncOnlyAttr(b.getArrayAttr({dtypeHost}));
EXPECT_TRUE(op->hasAsyncOnly(DeviceType::Host));
EXPECT_FALSE(op->hasAsyncOnly());
op->removeAsyncOnlyAttr();

auto dtypeStar = DeviceTypeAttr::get(&context, DeviceType::Star);
op->setAsyncOnlyAttr(b.getArrayAttr({dtypeHost, dtypeStar}));
EXPECT_TRUE(op->hasAsyncOnly(DeviceType::Star));
EXPECT_TRUE(op->hasAsyncOnly(DeviceType::Host));
EXPECT_FALSE(op->hasAsyncOnly());

op->removeAsyncOnlyAttr();
}

TEST_F(OpenACCOpsTest, asyncOnlyTestDataEntry) {
testAsyncOnlyDataEntry<DevicePtrOp>(b, context, loc, dtypes);
testAsyncOnlyDataEntry<PresentOp>(b, context, loc, dtypes);
testAsyncOnlyDataEntry<CopyinOp>(b, context, loc, dtypes);
testAsyncOnlyDataEntry<CreateOp>(b, context, loc, dtypes);
testAsyncOnlyDataEntry<NoCreateOp>(b, context, loc, dtypes);
testAsyncOnlyDataEntry<AttachOp>(b, context, loc, dtypes);
testAsyncOnlyDataEntry<UpdateDeviceOp>(b, context, loc, dtypes);
testAsyncOnlyDataEntry<UseDeviceOp>(b, context, loc, dtypes);
}

template <typename Op>
void testAsyncValue(OpBuilder &b, MLIRContext &context, Location loc,
llvm::SmallVector<DeviceType> &dtypes) {
Expand Down Expand Up @@ -105,6 +153,46 @@ TEST_F(OpenACCOpsTest, asyncValueTest) {
testAsyncValue<SerialOp>(b, context, loc, dtypes);
}

template <typename Op>
void testAsyncValueDataEntry(OpBuilder &b, MLIRContext &context, Location loc,
llvm::SmallVector<DeviceType> &dtypes) {
auto memrefTy = MemRefType::get({}, b.getI32Type());
OwningOpRef<memref::AllocaOp> varPtrOp =
b.create<memref::AllocaOp>(loc, memrefTy);

TypedValue<PointerLikeType> varPtr =
cast<TypedValue<PointerLikeType>>(varPtrOp->getResult());
OwningOpRef<Op> op = b.create<Op>(loc, varPtr,
/*structured=*/true, /*implicit=*/true);

mlir::Value empty;
EXPECT_EQ(op->getAsyncValue(), empty);
for (auto d : dtypes)
EXPECT_EQ(op->getAsyncValue(d), empty);

OwningOpRef<arith::ConstantIndexOp> val =
b.create<arith::ConstantIndexOp>(loc, 1);
auto dtypeNvidia = DeviceTypeAttr::get(&context, DeviceType::Nvidia);
op->setAsyncOperandsDeviceTypeAttr(b.getArrayAttr({dtypeNvidia}));
op->getAsyncOperandsMutable().assign(val->getResult());
EXPECT_EQ(op->getAsyncValue(), empty);
EXPECT_EQ(op->getAsyncValue(DeviceType::Nvidia), val->getResult());

op->getAsyncOperandsMutable().clear();
op->removeAsyncOperandsDeviceTypeAttr();
}

TEST_F(OpenACCOpsTest, asyncValueTestDataEntry) {
testAsyncValueDataEntry<DevicePtrOp>(b, context, loc, dtypes);
testAsyncValueDataEntry<PresentOp>(b, context, loc, dtypes);
testAsyncValueDataEntry<CopyinOp>(b, context, loc, dtypes);
testAsyncValueDataEntry<CreateOp>(b, context, loc, dtypes);
testAsyncValueDataEntry<NoCreateOp>(b, context, loc, dtypes);
testAsyncValueDataEntry<AttachOp>(b, context, loc, dtypes);
testAsyncValueDataEntry<UpdateDeviceOp>(b, context, loc, dtypes);
testAsyncValueDataEntry<UseDeviceOp>(b, context, loc, dtypes);
}

template <typename Op>
void testNumGangsValues(OpBuilder &b, MLIRContext &context, Location loc,
llvm::SmallVector<DeviceType> &dtypes,
Expand Down
Loading