[OpenACC][CIR] Implement Loop lowering of seq/auto/independent #138164
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
These just add a standard 'device_type' flag to the acc.loop, so implement that lowering. This also modifies the dialect to add helpers for these as well, to be consistent with the previous ones.
|
@llvm/pr-subscribers-mlir @llvm/pr-subscribers-clangir Author: Erich Keane (erichkeane) ChangesThese just add a standard 'device_type' flag to the acc.loop, so Full diff: https://github.com/llvm/llvm-project/pull/138164.diff 4 Files Affected:
diff --git a/clang/lib/CIR/CodeGen/CIRGenOpenACCClause.h b/clang/lib/CIR/CodeGen/CIRGenOpenACCClause.h
index b54682402d961..ff0bf6e7f55dd 100644
--- a/clang/lib/CIR/CodeGen/CIRGenOpenACCClause.h
+++ b/clang/lib/CIR/CodeGen/CIRGenOpenACCClause.h
@@ -147,13 +147,13 @@ class OpenACCClauseCIREmitter final
decodeDeviceType(clause.getArchitectures()[0].getIdentifierInfo()));
} else if constexpr (isOneOfTypes<OpTy, mlir::acc::ParallelOp,
mlir::acc::SerialOp, mlir::acc::KernelsOp,
- mlir::acc::DataOp>) {
+ mlir::acc::DataOp, mlir::acc::LoopOp>) {
// Nothing to do here, these constructs don't have any IR for these, as
// they just modify the other clauses IR. So setting of
// `lastDeviceTypeValues` (done above) is all we need.
} else {
// TODO: When we've implemented this for everything, switch this to an
- // unreachable. update, data, loop, routine, combined constructs remain.
+ // unreachable. update, data, routine, combined constructs remain.
return clauseNotImplemented(clause);
}
}
@@ -306,6 +306,36 @@ class OpenACCClauseCIREmitter final
llvm_unreachable("set, is only valid device_num constructs");
}
}
+
+ void VisitSeqClause(const OpenACCSeqClause &clause) {
+ if constexpr (isOneOfTypes<OpTy, mlir::acc::LoopOp>) {
+ operation.addSeq(builder.getContext(), lastDeviceTypeValues);
+ } else {
+ // TODO: When we've implemented this for everything, switch this to an
+ // unreachable. Routine, Combined constructs remain.
+ return clauseNotImplemented(clause);
+ }
+ }
+
+ void VisitAutoClause(const OpenACCAutoClause &clause) {
+ if constexpr (isOneOfTypes<OpTy, mlir::acc::LoopOp>) {
+ operation.addAuto(builder.getContext(), lastDeviceTypeValues);
+ } else {
+ // TODO: When we've implemented this for everything, switch this to an
+ // unreachable. Routine, Combined constructs remain.
+ return clauseNotImplemented(clause);
+ }
+ }
+
+ void VisitIndependentClause(const OpenACCIndependentClause &clause) {
+ if constexpr (isOneOfTypes<OpTy, mlir::acc::LoopOp>) {
+ operation.addIndependent(builder.getContext(), lastDeviceTypeValues);
+ } else {
+ // TODO: When we've implemented this for everything, switch this to an
+ // unreachable. Routine, Combined constructs remain.
+ return clauseNotImplemented(clause);
+ }
+ }
};
template <typename OpTy>
diff --git a/clang/test/CIR/CodeGenOpenACC/loop.cpp b/clang/test/CIR/CodeGenOpenACC/loop.cpp
index 792edfedaacc6..2757d935e1f76 100644
--- a/clang/test/CIR/CodeGenOpenACC/loop.cpp
+++ b/clang/test/CIR/CodeGenOpenACC/loop.cpp
@@ -30,4 +30,83 @@ extern "C" void acc_loop(int *A, int *B, int *C, int N) {
// CHECK-NEXT: } loc
// CHECK-NEXT: acc.yield
// CHECK-NEXT: } loc
+
+
+#pragma acc loop seq
+ for(unsigned I = 0; I < N; ++I);
+ // CHECK: acc.loop {
+ // CHECK: acc.yield
+ // CHECK-NEXT: } attributes {seq = [#acc.device_type<none>]} loc
+#pragma acc loop device_type(nvidia, radeon) seq
+ for(unsigned I = 0; I < N; ++I);
+ // CHECK: acc.loop {
+ // CHECK: acc.yield
+ // CHECK-NEXT: } attributes {seq = [#acc.device_type<nvidia>, #acc.device_type<radeon>]} loc
+#pragma acc loop device_type(radeon) seq
+ for(unsigned I = 0; I < N; ++I);
+ // CHECK: acc.loop {
+ // CHECK: acc.yield
+ // CHECK-NEXT: } attributes {seq = [#acc.device_type<radeon>]} loc
+#pragma acc loop seq device_type(nvidia, radeon)
+ for(unsigned I = 0; I < N; ++I);
+ // CHECK: acc.loop {
+ // CHECK: acc.yield
+ // CHECK-NEXT: } attributes {seq = [#acc.device_type<none>]} loc
+#pragma acc loop seq device_type(radeon)
+ for(unsigned I = 0; I < N; ++I);
+ // CHECK: acc.loop {
+ // CHECK: acc.yield
+ // CHECK-NEXT: } attributes {seq = [#acc.device_type<none>]} loc
+
+#pragma acc loop independent
+ for(unsigned I = 0; I < N; ++I);
+ // CHECK: acc.loop {
+ // CHECK: acc.yield
+ // CHECK-NEXT: } attributes {independent = [#acc.device_type<none>]} loc
+#pragma acc loop device_type(nvidia, radeon) independent
+ for(unsigned I = 0; I < N; ++I);
+ // CHECK: acc.loop {
+ // CHECK: acc.yield
+ // CHECK-NEXT: } attributes {independent = [#acc.device_type<nvidia>, #acc.device_type<radeon>]} loc
+#pragma acc loop device_type(radeon) independent
+ for(unsigned I = 0; I < N; ++I);
+ // CHECK: acc.loop {
+ // CHECK: acc.yield
+ // CHECK-NEXT: } attributes {independent = [#acc.device_type<radeon>]} loc
+#pragma acc loop independent device_type(nvidia, radeon)
+ for(unsigned I = 0; I < N; ++I);
+ // CHECK: acc.loop {
+ // CHECK: acc.yield
+ // CHECK-NEXT: } attributes {independent = [#acc.device_type<none>]} loc
+#pragma acc loop independent device_type(radeon)
+ for(unsigned I = 0; I < N; ++I);
+ // CHECK: acc.loop {
+ // CHECK: acc.yield
+ // CHECK-NEXT: } attributes {independent = [#acc.device_type<none>]} loc
+
+#pragma acc loop auto
+ for(unsigned I = 0; I < N; ++I);
+ // CHECK: acc.loop {
+ // CHECK: acc.yield
+ // CHECK-NEXT: } attributes {auto_ = [#acc.device_type<none>]} loc
+#pragma acc loop device_type(nvidia, radeon) auto
+ for(unsigned I = 0; I < N; ++I);
+ // CHECK: acc.loop {
+ // CHECK: acc.yield
+ // CHECK-NEXT: } attributes {auto_ = [#acc.device_type<nvidia>, #acc.device_type<radeon>]} loc
+#pragma acc loop device_type(radeon) auto
+ for(unsigned I = 0; I < N; ++I);
+ // CHECK: acc.loop {
+ // CHECK: acc.yield
+ // CHECK-NEXT: } attributes {auto_ = [#acc.device_type<radeon>]} loc
+#pragma acc loop auto device_type(nvidia, radeon)
+ for(unsigned I = 0; I < N; ++I);
+ // CHECK: acc.loop {
+ // CHECK: acc.yield
+ // CHECK-NEXT: } attributes {auto_ = [#acc.device_type<none>]} loc
+#pragma acc loop auto device_type(radeon)
+ for(unsigned I = 0; I < N; ++I);
+ // CHECK: acc.loop {
+ // CHECK: acc.yield
+ // CHECK-NEXT: } attributes {auto_ = [#acc.device_type<none>]} loc
}
diff --git a/mlir/include/mlir/Dialect/OpenACC/OpenACCOps.td b/mlir/include/mlir/Dialect/OpenACC/OpenACCOps.td
index 3ad8e4f9ccbeb..c3df064cf0ead 100644
--- a/mlir/include/mlir/Dialect/OpenACC/OpenACCOps.td
+++ b/mlir/include/mlir/Dialect/OpenACC/OpenACCOps.td
@@ -2198,6 +2198,14 @@ def OpenACC_LoopOp : OpenACC_Op<"loop",
/// Return the value of the worker clause for the given device_type
/// if present.
mlir::Value getGangValue(mlir::acc::GangArgType gangArgType, mlir::acc::DeviceType deviceType);
+
+ // Add an entry to the 'seq' attribute for each additional device types.
+ void addSeq(MLIRContext *, llvm::ArrayRef<DeviceType>);
+ // Add an entry to the 'independent' attribute for each additional device
+ // types.
+ void addIndependent(MLIRContext *, llvm::ArrayRef<DeviceType>);
+ // Add an entry to the 'auto' attribute for each additional device types.
+ void addAuto(MLIRContext *, llvm::ArrayRef<DeviceType>);
}];
let hasCustomAssemblyFormat = 1;
diff --git a/mlir/lib/Dialect/OpenACC/IR/OpenACC.cpp b/mlir/lib/Dialect/OpenACC/IR/OpenACC.cpp
index d23563f1f0fb0..39dbb0c92a309 100644
--- a/mlir/lib/Dialect/OpenACC/IR/OpenACC.cpp
+++ b/mlir/lib/Dialect/OpenACC/IR/OpenACC.cpp
@@ -2651,6 +2651,24 @@ void printLoopControl(OpAsmPrinter &p, Operation *op, Region ®ion,
p.printRegion(region, /*printEntryBlockArgs=*/false);
}
+void acc::LoopOp::addSeq(MLIRContext *context,
+ llvm::ArrayRef<DeviceType> effectiveDeviceTypes) {
+ setSeqAttr(addDeviceTypeAffectedOperandHelper(context, getSeqAttr(),
+ effectiveDeviceTypes));
+}
+
+void acc::LoopOp::addIndependent(
+ MLIRContext *context, llvm::ArrayRef<DeviceType> effectiveDeviceTypes) {
+ setIndependentAttr(addDeviceTypeAffectedOperandHelper(
+ context, getIndependentAttr(), effectiveDeviceTypes));
+}
+
+void acc::LoopOp::addAuto(MLIRContext *context,
+ llvm::ArrayRef<DeviceType> effectiveDeviceTypes) {
+ setAuto_Attr(addDeviceTypeAffectedOperandHelper(context, getAuto_Attr(),
+ effectiveDeviceTypes));
+}
+
//===----------------------------------------------------------------------===//
// DataOp
//===----------------------------------------------------------------------===//
|
|
@llvm/pr-subscribers-clang Author: Erich Keane (erichkeane) ChangesThese just add a standard 'device_type' flag to the acc.loop, so Full diff: https://github.com/llvm/llvm-project/pull/138164.diff 4 Files Affected:
diff --git a/clang/lib/CIR/CodeGen/CIRGenOpenACCClause.h b/clang/lib/CIR/CodeGen/CIRGenOpenACCClause.h
index b54682402d961..ff0bf6e7f55dd 100644
--- a/clang/lib/CIR/CodeGen/CIRGenOpenACCClause.h
+++ b/clang/lib/CIR/CodeGen/CIRGenOpenACCClause.h
@@ -147,13 +147,13 @@ class OpenACCClauseCIREmitter final
decodeDeviceType(clause.getArchitectures()[0].getIdentifierInfo()));
} else if constexpr (isOneOfTypes<OpTy, mlir::acc::ParallelOp,
mlir::acc::SerialOp, mlir::acc::KernelsOp,
- mlir::acc::DataOp>) {
+ mlir::acc::DataOp, mlir::acc::LoopOp>) {
// Nothing to do here, these constructs don't have any IR for these, as
// they just modify the other clauses IR. So setting of
// `lastDeviceTypeValues` (done above) is all we need.
} else {
// TODO: When we've implemented this for everything, switch this to an
- // unreachable. update, data, loop, routine, combined constructs remain.
+ // unreachable. update, data, routine, combined constructs remain.
return clauseNotImplemented(clause);
}
}
@@ -306,6 +306,36 @@ class OpenACCClauseCIREmitter final
llvm_unreachable("set, is only valid device_num constructs");
}
}
+
+ void VisitSeqClause(const OpenACCSeqClause &clause) {
+ if constexpr (isOneOfTypes<OpTy, mlir::acc::LoopOp>) {
+ operation.addSeq(builder.getContext(), lastDeviceTypeValues);
+ } else {
+ // TODO: When we've implemented this for everything, switch this to an
+ // unreachable. Routine, Combined constructs remain.
+ return clauseNotImplemented(clause);
+ }
+ }
+
+ void VisitAutoClause(const OpenACCAutoClause &clause) {
+ if constexpr (isOneOfTypes<OpTy, mlir::acc::LoopOp>) {
+ operation.addAuto(builder.getContext(), lastDeviceTypeValues);
+ } else {
+ // TODO: When we've implemented this for everything, switch this to an
+ // unreachable. Routine, Combined constructs remain.
+ return clauseNotImplemented(clause);
+ }
+ }
+
+ void VisitIndependentClause(const OpenACCIndependentClause &clause) {
+ if constexpr (isOneOfTypes<OpTy, mlir::acc::LoopOp>) {
+ operation.addIndependent(builder.getContext(), lastDeviceTypeValues);
+ } else {
+ // TODO: When we've implemented this for everything, switch this to an
+ // unreachable. Routine, Combined constructs remain.
+ return clauseNotImplemented(clause);
+ }
+ }
};
template <typename OpTy>
diff --git a/clang/test/CIR/CodeGenOpenACC/loop.cpp b/clang/test/CIR/CodeGenOpenACC/loop.cpp
index 792edfedaacc6..2757d935e1f76 100644
--- a/clang/test/CIR/CodeGenOpenACC/loop.cpp
+++ b/clang/test/CIR/CodeGenOpenACC/loop.cpp
@@ -30,4 +30,83 @@ extern "C" void acc_loop(int *A, int *B, int *C, int N) {
// CHECK-NEXT: } loc
// CHECK-NEXT: acc.yield
// CHECK-NEXT: } loc
+
+
+#pragma acc loop seq
+ for(unsigned I = 0; I < N; ++I);
+ // CHECK: acc.loop {
+ // CHECK: acc.yield
+ // CHECK-NEXT: } attributes {seq = [#acc.device_type<none>]} loc
+#pragma acc loop device_type(nvidia, radeon) seq
+ for(unsigned I = 0; I < N; ++I);
+ // CHECK: acc.loop {
+ // CHECK: acc.yield
+ // CHECK-NEXT: } attributes {seq = [#acc.device_type<nvidia>, #acc.device_type<radeon>]} loc
+#pragma acc loop device_type(radeon) seq
+ for(unsigned I = 0; I < N; ++I);
+ // CHECK: acc.loop {
+ // CHECK: acc.yield
+ // CHECK-NEXT: } attributes {seq = [#acc.device_type<radeon>]} loc
+#pragma acc loop seq device_type(nvidia, radeon)
+ for(unsigned I = 0; I < N; ++I);
+ // CHECK: acc.loop {
+ // CHECK: acc.yield
+ // CHECK-NEXT: } attributes {seq = [#acc.device_type<none>]} loc
+#pragma acc loop seq device_type(radeon)
+ for(unsigned I = 0; I < N; ++I);
+ // CHECK: acc.loop {
+ // CHECK: acc.yield
+ // CHECK-NEXT: } attributes {seq = [#acc.device_type<none>]} loc
+
+#pragma acc loop independent
+ for(unsigned I = 0; I < N; ++I);
+ // CHECK: acc.loop {
+ // CHECK: acc.yield
+ // CHECK-NEXT: } attributes {independent = [#acc.device_type<none>]} loc
+#pragma acc loop device_type(nvidia, radeon) independent
+ for(unsigned I = 0; I < N; ++I);
+ // CHECK: acc.loop {
+ // CHECK: acc.yield
+ // CHECK-NEXT: } attributes {independent = [#acc.device_type<nvidia>, #acc.device_type<radeon>]} loc
+#pragma acc loop device_type(radeon) independent
+ for(unsigned I = 0; I < N; ++I);
+ // CHECK: acc.loop {
+ // CHECK: acc.yield
+ // CHECK-NEXT: } attributes {independent = [#acc.device_type<radeon>]} loc
+#pragma acc loop independent device_type(nvidia, radeon)
+ for(unsigned I = 0; I < N; ++I);
+ // CHECK: acc.loop {
+ // CHECK: acc.yield
+ // CHECK-NEXT: } attributes {independent = [#acc.device_type<none>]} loc
+#pragma acc loop independent device_type(radeon)
+ for(unsigned I = 0; I < N; ++I);
+ // CHECK: acc.loop {
+ // CHECK: acc.yield
+ // CHECK-NEXT: } attributes {independent = [#acc.device_type<none>]} loc
+
+#pragma acc loop auto
+ for(unsigned I = 0; I < N; ++I);
+ // CHECK: acc.loop {
+ // CHECK: acc.yield
+ // CHECK-NEXT: } attributes {auto_ = [#acc.device_type<none>]} loc
+#pragma acc loop device_type(nvidia, radeon) auto
+ for(unsigned I = 0; I < N; ++I);
+ // CHECK: acc.loop {
+ // CHECK: acc.yield
+ // CHECK-NEXT: } attributes {auto_ = [#acc.device_type<nvidia>, #acc.device_type<radeon>]} loc
+#pragma acc loop device_type(radeon) auto
+ for(unsigned I = 0; I < N; ++I);
+ // CHECK: acc.loop {
+ // CHECK: acc.yield
+ // CHECK-NEXT: } attributes {auto_ = [#acc.device_type<radeon>]} loc
+#pragma acc loop auto device_type(nvidia, radeon)
+ for(unsigned I = 0; I < N; ++I);
+ // CHECK: acc.loop {
+ // CHECK: acc.yield
+ // CHECK-NEXT: } attributes {auto_ = [#acc.device_type<none>]} loc
+#pragma acc loop auto device_type(radeon)
+ for(unsigned I = 0; I < N; ++I);
+ // CHECK: acc.loop {
+ // CHECK: acc.yield
+ // CHECK-NEXT: } attributes {auto_ = [#acc.device_type<none>]} loc
}
diff --git a/mlir/include/mlir/Dialect/OpenACC/OpenACCOps.td b/mlir/include/mlir/Dialect/OpenACC/OpenACCOps.td
index 3ad8e4f9ccbeb..c3df064cf0ead 100644
--- a/mlir/include/mlir/Dialect/OpenACC/OpenACCOps.td
+++ b/mlir/include/mlir/Dialect/OpenACC/OpenACCOps.td
@@ -2198,6 +2198,14 @@ def OpenACC_LoopOp : OpenACC_Op<"loop",
/// Return the value of the worker clause for the given device_type
/// if present.
mlir::Value getGangValue(mlir::acc::GangArgType gangArgType, mlir::acc::DeviceType deviceType);
+
+ // Add an entry to the 'seq' attribute for each additional device types.
+ void addSeq(MLIRContext *, llvm::ArrayRef<DeviceType>);
+ // Add an entry to the 'independent' attribute for each additional device
+ // types.
+ void addIndependent(MLIRContext *, llvm::ArrayRef<DeviceType>);
+ // Add an entry to the 'auto' attribute for each additional device types.
+ void addAuto(MLIRContext *, llvm::ArrayRef<DeviceType>);
}];
let hasCustomAssemblyFormat = 1;
diff --git a/mlir/lib/Dialect/OpenACC/IR/OpenACC.cpp b/mlir/lib/Dialect/OpenACC/IR/OpenACC.cpp
index d23563f1f0fb0..39dbb0c92a309 100644
--- a/mlir/lib/Dialect/OpenACC/IR/OpenACC.cpp
+++ b/mlir/lib/Dialect/OpenACC/IR/OpenACC.cpp
@@ -2651,6 +2651,24 @@ void printLoopControl(OpAsmPrinter &p, Operation *op, Region ®ion,
p.printRegion(region, /*printEntryBlockArgs=*/false);
}
+void acc::LoopOp::addSeq(MLIRContext *context,
+ llvm::ArrayRef<DeviceType> effectiveDeviceTypes) {
+ setSeqAttr(addDeviceTypeAffectedOperandHelper(context, getSeqAttr(),
+ effectiveDeviceTypes));
+}
+
+void acc::LoopOp::addIndependent(
+ MLIRContext *context, llvm::ArrayRef<DeviceType> effectiveDeviceTypes) {
+ setIndependentAttr(addDeviceTypeAffectedOperandHelper(
+ context, getIndependentAttr(), effectiveDeviceTypes));
+}
+
+void acc::LoopOp::addAuto(MLIRContext *context,
+ llvm::ArrayRef<DeviceType> effectiveDeviceTypes) {
+ setAuto_Attr(addDeviceTypeAffectedOperandHelper(context, getAuto_Attr(),
+ effectiveDeviceTypes));
+}
+
//===----------------------------------------------------------------------===//
// DataOp
//===----------------------------------------------------------------------===//
|
|
@llvm/pr-subscribers-mlir-openacc Author: Erich Keane (erichkeane) ChangesThese just add a standard 'device_type' flag to the acc.loop, so Full diff: https://github.com/llvm/llvm-project/pull/138164.diff 4 Files Affected:
diff --git a/clang/lib/CIR/CodeGen/CIRGenOpenACCClause.h b/clang/lib/CIR/CodeGen/CIRGenOpenACCClause.h
index b54682402d961..ff0bf6e7f55dd 100644
--- a/clang/lib/CIR/CodeGen/CIRGenOpenACCClause.h
+++ b/clang/lib/CIR/CodeGen/CIRGenOpenACCClause.h
@@ -147,13 +147,13 @@ class OpenACCClauseCIREmitter final
decodeDeviceType(clause.getArchitectures()[0].getIdentifierInfo()));
} else if constexpr (isOneOfTypes<OpTy, mlir::acc::ParallelOp,
mlir::acc::SerialOp, mlir::acc::KernelsOp,
- mlir::acc::DataOp>) {
+ mlir::acc::DataOp, mlir::acc::LoopOp>) {
// Nothing to do here, these constructs don't have any IR for these, as
// they just modify the other clauses IR. So setting of
// `lastDeviceTypeValues` (done above) is all we need.
} else {
// TODO: When we've implemented this for everything, switch this to an
- // unreachable. update, data, loop, routine, combined constructs remain.
+ // unreachable. update, data, routine, combined constructs remain.
return clauseNotImplemented(clause);
}
}
@@ -306,6 +306,36 @@ class OpenACCClauseCIREmitter final
llvm_unreachable("set, is only valid device_num constructs");
}
}
+
+ void VisitSeqClause(const OpenACCSeqClause &clause) {
+ if constexpr (isOneOfTypes<OpTy, mlir::acc::LoopOp>) {
+ operation.addSeq(builder.getContext(), lastDeviceTypeValues);
+ } else {
+ // TODO: When we've implemented this for everything, switch this to an
+ // unreachable. Routine, Combined constructs remain.
+ return clauseNotImplemented(clause);
+ }
+ }
+
+ void VisitAutoClause(const OpenACCAutoClause &clause) {
+ if constexpr (isOneOfTypes<OpTy, mlir::acc::LoopOp>) {
+ operation.addAuto(builder.getContext(), lastDeviceTypeValues);
+ } else {
+ // TODO: When we've implemented this for everything, switch this to an
+ // unreachable. Routine, Combined constructs remain.
+ return clauseNotImplemented(clause);
+ }
+ }
+
+ void VisitIndependentClause(const OpenACCIndependentClause &clause) {
+ if constexpr (isOneOfTypes<OpTy, mlir::acc::LoopOp>) {
+ operation.addIndependent(builder.getContext(), lastDeviceTypeValues);
+ } else {
+ // TODO: When we've implemented this for everything, switch this to an
+ // unreachable. Routine, Combined constructs remain.
+ return clauseNotImplemented(clause);
+ }
+ }
};
template <typename OpTy>
diff --git a/clang/test/CIR/CodeGenOpenACC/loop.cpp b/clang/test/CIR/CodeGenOpenACC/loop.cpp
index 792edfedaacc6..2757d935e1f76 100644
--- a/clang/test/CIR/CodeGenOpenACC/loop.cpp
+++ b/clang/test/CIR/CodeGenOpenACC/loop.cpp
@@ -30,4 +30,83 @@ extern "C" void acc_loop(int *A, int *B, int *C, int N) {
// CHECK-NEXT: } loc
// CHECK-NEXT: acc.yield
// CHECK-NEXT: } loc
+
+
+#pragma acc loop seq
+ for(unsigned I = 0; I < N; ++I);
+ // CHECK: acc.loop {
+ // CHECK: acc.yield
+ // CHECK-NEXT: } attributes {seq = [#acc.device_type<none>]} loc
+#pragma acc loop device_type(nvidia, radeon) seq
+ for(unsigned I = 0; I < N; ++I);
+ // CHECK: acc.loop {
+ // CHECK: acc.yield
+ // CHECK-NEXT: } attributes {seq = [#acc.device_type<nvidia>, #acc.device_type<radeon>]} loc
+#pragma acc loop device_type(radeon) seq
+ for(unsigned I = 0; I < N; ++I);
+ // CHECK: acc.loop {
+ // CHECK: acc.yield
+ // CHECK-NEXT: } attributes {seq = [#acc.device_type<radeon>]} loc
+#pragma acc loop seq device_type(nvidia, radeon)
+ for(unsigned I = 0; I < N; ++I);
+ // CHECK: acc.loop {
+ // CHECK: acc.yield
+ // CHECK-NEXT: } attributes {seq = [#acc.device_type<none>]} loc
+#pragma acc loop seq device_type(radeon)
+ for(unsigned I = 0; I < N; ++I);
+ // CHECK: acc.loop {
+ // CHECK: acc.yield
+ // CHECK-NEXT: } attributes {seq = [#acc.device_type<none>]} loc
+
+#pragma acc loop independent
+ for(unsigned I = 0; I < N; ++I);
+ // CHECK: acc.loop {
+ // CHECK: acc.yield
+ // CHECK-NEXT: } attributes {independent = [#acc.device_type<none>]} loc
+#pragma acc loop device_type(nvidia, radeon) independent
+ for(unsigned I = 0; I < N; ++I);
+ // CHECK: acc.loop {
+ // CHECK: acc.yield
+ // CHECK-NEXT: } attributes {independent = [#acc.device_type<nvidia>, #acc.device_type<radeon>]} loc
+#pragma acc loop device_type(radeon) independent
+ for(unsigned I = 0; I < N; ++I);
+ // CHECK: acc.loop {
+ // CHECK: acc.yield
+ // CHECK-NEXT: } attributes {independent = [#acc.device_type<radeon>]} loc
+#pragma acc loop independent device_type(nvidia, radeon)
+ for(unsigned I = 0; I < N; ++I);
+ // CHECK: acc.loop {
+ // CHECK: acc.yield
+ // CHECK-NEXT: } attributes {independent = [#acc.device_type<none>]} loc
+#pragma acc loop independent device_type(radeon)
+ for(unsigned I = 0; I < N; ++I);
+ // CHECK: acc.loop {
+ // CHECK: acc.yield
+ // CHECK-NEXT: } attributes {independent = [#acc.device_type<none>]} loc
+
+#pragma acc loop auto
+ for(unsigned I = 0; I < N; ++I);
+ // CHECK: acc.loop {
+ // CHECK: acc.yield
+ // CHECK-NEXT: } attributes {auto_ = [#acc.device_type<none>]} loc
+#pragma acc loop device_type(nvidia, radeon) auto
+ for(unsigned I = 0; I < N; ++I);
+ // CHECK: acc.loop {
+ // CHECK: acc.yield
+ // CHECK-NEXT: } attributes {auto_ = [#acc.device_type<nvidia>, #acc.device_type<radeon>]} loc
+#pragma acc loop device_type(radeon) auto
+ for(unsigned I = 0; I < N; ++I);
+ // CHECK: acc.loop {
+ // CHECK: acc.yield
+ // CHECK-NEXT: } attributes {auto_ = [#acc.device_type<radeon>]} loc
+#pragma acc loop auto device_type(nvidia, radeon)
+ for(unsigned I = 0; I < N; ++I);
+ // CHECK: acc.loop {
+ // CHECK: acc.yield
+ // CHECK-NEXT: } attributes {auto_ = [#acc.device_type<none>]} loc
+#pragma acc loop auto device_type(radeon)
+ for(unsigned I = 0; I < N; ++I);
+ // CHECK: acc.loop {
+ // CHECK: acc.yield
+ // CHECK-NEXT: } attributes {auto_ = [#acc.device_type<none>]} loc
}
diff --git a/mlir/include/mlir/Dialect/OpenACC/OpenACCOps.td b/mlir/include/mlir/Dialect/OpenACC/OpenACCOps.td
index 3ad8e4f9ccbeb..c3df064cf0ead 100644
--- a/mlir/include/mlir/Dialect/OpenACC/OpenACCOps.td
+++ b/mlir/include/mlir/Dialect/OpenACC/OpenACCOps.td
@@ -2198,6 +2198,14 @@ def OpenACC_LoopOp : OpenACC_Op<"loop",
/// Return the value of the worker clause for the given device_type
/// if present.
mlir::Value getGangValue(mlir::acc::GangArgType gangArgType, mlir::acc::DeviceType deviceType);
+
+ // Add an entry to the 'seq' attribute for each additional device types.
+ void addSeq(MLIRContext *, llvm::ArrayRef<DeviceType>);
+ // Add an entry to the 'independent' attribute for each additional device
+ // types.
+ void addIndependent(MLIRContext *, llvm::ArrayRef<DeviceType>);
+ // Add an entry to the 'auto' attribute for each additional device types.
+ void addAuto(MLIRContext *, llvm::ArrayRef<DeviceType>);
}];
let hasCustomAssemblyFormat = 1;
diff --git a/mlir/lib/Dialect/OpenACC/IR/OpenACC.cpp b/mlir/lib/Dialect/OpenACC/IR/OpenACC.cpp
index d23563f1f0fb0..39dbb0c92a309 100644
--- a/mlir/lib/Dialect/OpenACC/IR/OpenACC.cpp
+++ b/mlir/lib/Dialect/OpenACC/IR/OpenACC.cpp
@@ -2651,6 +2651,24 @@ void printLoopControl(OpAsmPrinter &p, Operation *op, Region ®ion,
p.printRegion(region, /*printEntryBlockArgs=*/false);
}
+void acc::LoopOp::addSeq(MLIRContext *context,
+ llvm::ArrayRef<DeviceType> effectiveDeviceTypes) {
+ setSeqAttr(addDeviceTypeAffectedOperandHelper(context, getSeqAttr(),
+ effectiveDeviceTypes));
+}
+
+void acc::LoopOp::addIndependent(
+ MLIRContext *context, llvm::ArrayRef<DeviceType> effectiveDeviceTypes) {
+ setIndependentAttr(addDeviceTypeAffectedOperandHelper(
+ context, getIndependentAttr(), effectiveDeviceTypes));
+}
+
+void acc::LoopOp::addAuto(MLIRContext *context,
+ llvm::ArrayRef<DeviceType> effectiveDeviceTypes) {
+ setAuto_Attr(addDeviceTypeAffectedOperandHelper(context, getAuto_Attr(),
+ effectiveDeviceTypes));
+}
+
//===----------------------------------------------------------------------===//
// DataOp
//===----------------------------------------------------------------------===//
|
razvanlupusoru
approved these changes
May 1, 2025
IanWood1
pushed a commit
to IanWood1/llvm-project
that referenced
this pull request
May 6, 2025
…138164) These just add a standard 'device_type' flag to the acc.loop, so implement that lowering. This also modifies the dialect to add helpers for these as well, to be consistent with the previous ones.
IanWood1
pushed a commit
to IanWood1/llvm-project
that referenced
this pull request
May 6, 2025
…138164) These just add a standard 'device_type' flag to the acc.loop, so implement that lowering. This also modifies the dialect to add helpers for these as well, to be consistent with the previous ones.
IanWood1
pushed a commit
to IanWood1/llvm-project
that referenced
this pull request
May 6, 2025
…138164) These just add a standard 'device_type' flag to the acc.loop, so implement that lowering. This also modifies the dialect to add helpers for these as well, to be consistent with the previous ones.
GeorgeARM
pushed a commit
to GeorgeARM/llvm-project
that referenced
this pull request
May 7, 2025
…138164) These just add a standard 'device_type' flag to the acc.loop, so implement that lowering. This also modifies the dialect to add helpers for these as well, to be consistent with the previous ones.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
These just add a standard 'device_type' flag to the acc.loop, so
implement that lowering. This also modifies the dialect to add helpers
for these as well, to be consistent with the previous ones.