-
Notifications
You must be signed in to change notification settings - Fork 11.8k
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
[flang][OpenMP] Add frontend support for ompx_bare clause #111106
base: main
Are you sure you want to change the base?
Conversation
@llvm/pr-subscribers-mlir-llvm @llvm/pr-subscribers-mlir-openmp Author: Ivan R. Ivanov (ivanradanov) ChangesFull diff: https://github.com/llvm/llvm-project/pull/111106.diff 11 Files Affected:
diff --git a/flang/lib/Lower/OpenMP/ClauseProcessor.cpp b/flang/lib/Lower/OpenMP/ClauseProcessor.cpp
index a4d2524bccf5c3..c7f34b3d65f324 100644
--- a/flang/lib/Lower/OpenMP/ClauseProcessor.cpp
+++ b/flang/lib/Lower/OpenMP/ClauseProcessor.cpp
@@ -354,6 +354,10 @@ bool ClauseProcessor::processNowait(mlir::omp::NowaitClauseOps &result) const {
return markClauseOccurrence<omp::clause::Nowait>(result.nowait);
}
+bool ClauseProcessor::processBare(mlir::omp::BareClauseOps &result) const {
+ return markClauseOccurrence<omp::clause::OmpxBare>(result.bare);
+}
+
bool ClauseProcessor::processNumTeams(
lower::StatementContext &stmtCtx,
mlir::omp::NumTeamsClauseOps &result) const {
diff --git a/flang/lib/Lower/OpenMP/ClauseProcessor.h b/flang/lib/Lower/OpenMP/ClauseProcessor.h
index 0c8e7bd47ab5a6..a04731f94ede37 100644
--- a/flang/lib/Lower/OpenMP/ClauseProcessor.h
+++ b/flang/lib/Lower/OpenMP/ClauseProcessor.h
@@ -74,6 +74,7 @@ class ClauseProcessor {
bool processHint(mlir::omp::HintClauseOps &result) const;
bool processMergeable(mlir::omp::MergeableClauseOps &result) const;
bool processNowait(mlir::omp::NowaitClauseOps &result) const;
+ bool processBare(mlir::omp::BareClauseOps &result) const;
bool processNumTeams(lower::StatementContext &stmtCtx,
mlir::omp::NumTeamsClauseOps &result) const;
bool processNumThreads(lower::StatementContext &stmtCtx,
diff --git a/flang/lib/Lower/OpenMP/Clauses.h b/flang/lib/Lower/OpenMP/Clauses.h
index 51bf0eab0f8d07..10bd1439563244 100644
--- a/flang/lib/Lower/OpenMP/Clauses.h
+++ b/flang/lib/Lower/OpenMP/Clauses.h
@@ -229,8 +229,8 @@ using NumTasks = tomp::clause::NumTasksT<TypeTy, IdTy, ExprTy>;
using NumTeams = tomp::clause::NumTeamsT<TypeTy, IdTy, ExprTy>;
using NumThreads = tomp::clause::NumThreadsT<TypeTy, IdTy, ExprTy>;
using OmpxAttribute = tomp::clause::OmpxAttributeT<TypeTy, IdTy, ExprTy>;
-using OmpxBare = tomp::clause::OmpxBareT<TypeTy, IdTy, ExprTy>;
using OmpxDynCgroupMem = tomp::clause::OmpxDynCgroupMemT<TypeTy, IdTy, ExprTy>;
+using OmpxBare = tomp::clause::OmpxBareT<TypeTy, IdTy, ExprTy>;
using Ordered = tomp::clause::OrderedT<TypeTy, IdTy, ExprTy>;
using Order = tomp::clause::OrderT<TypeTy, IdTy, ExprTy>;
using Partial = tomp::clause::PartialT<TypeTy, IdTy, ExprTy>;
diff --git a/flang/lib/Lower/OpenMP/OpenMP.cpp b/flang/lib/Lower/OpenMP/OpenMP.cpp
index 60c83586e468b6..b8b1a0ba2a69b7 100644
--- a/flang/lib/Lower/OpenMP/OpenMP.cpp
+++ b/flang/lib/Lower/OpenMP/OpenMP.cpp
@@ -1180,6 +1180,7 @@ static void genTargetClauses(
cp.processNowait(clauseOps);
cp.processThreadLimit(stmtCtx, clauseOps);
+ cp.processBare(clauseOps);
cp.processTODO<clause::Allocate, clause::Defaultmap, clause::Firstprivate,
clause::InReduction, clause::UsesAllocators>(
@@ -2764,6 +2765,7 @@ static void genOMP(lower::AbstractConverter &converter, lower::SymMap &symTable,
!std::holds_alternative<clause::ThreadLimit>(clause.u) &&
!std::holds_alternative<clause::Threads>(clause.u) &&
!std::holds_alternative<clause::UseDeviceAddr>(clause.u) &&
+ !std::holds_alternative<clause::OmpxBare>(clause.u) &&
!std::holds_alternative<clause::UseDevicePtr>(clause.u)) {
TODO(clauseLocation, "OpenMP Block construct clause");
}
diff --git a/flang/lib/Parser/openmp-parsers.cpp b/flang/lib/Parser/openmp-parsers.cpp
index cc2930cbd7ded5..f18a9929aa75ef 100644
--- a/flang/lib/Parser/openmp-parsers.cpp
+++ b/flang/lib/Parser/openmp-parsers.cpp
@@ -315,6 +315,7 @@ TYPE_PARSER(
parenthesized(scalarIntExpr))) ||
"NUM_THREADS" >> construct<OmpClause>(construct<OmpClause::NumThreads>(
parenthesized(scalarIntExpr))) ||
+ "OMPX_BARE" >> construct<OmpClause>(construct<OmpClause::OmpxBare>()) ||
"ORDER" >> construct<OmpClause>(construct<OmpClause::Order>(
parenthesized(Parser<OmpOrderClause>{}))) ||
"ORDERED" >> construct<OmpClause>(construct<OmpClause::Ordered>(
diff --git a/flang/test/Lower/OpenMP/KernelLanguage/bare-clause.f90 b/flang/test/Lower/OpenMP/KernelLanguage/bare-clause.f90
new file mode 100644
index 00000000000000..2a97a14516ec3a
--- /dev/null
+++ b/flang/test/Lower/OpenMP/KernelLanguage/bare-clause.f90
@@ -0,0 +1,10 @@
+! RUN: %flang_fc1 -emit-hlfir %openmp_flags -fopenmp-version=51 %s -o - | FileCheck %s
+
+program test
+ integer :: tmp
+ !$omp target teams ompx_bare num_teams(42) thread_limit(43)
+ tmp = 1
+ !$omp end target teams
+end program
+
+! CHECK: omp.target {{.*}} ompx_bare
diff --git a/llvm/include/llvm/Frontend/OpenMP/ConstructDecompositionT.h b/llvm/include/llvm/Frontend/OpenMP/ConstructDecompositionT.h
index 4bdfa1cf4c1490..7c0dbe1aae1221 100644
--- a/llvm/include/llvm/Frontend/OpenMP/ConstructDecompositionT.h
+++ b/llvm/include/llvm/Frontend/OpenMP/ConstructDecompositionT.h
@@ -236,6 +236,8 @@ struct ConstructDecompositionT {
const ClauseTy *);
bool applyClause(const tomp::clause::NowaitT<TypeTy, IdTy, ExprTy> &clause,
const ClauseTy *);
+ bool applyClause(const tomp::clause::OmpxBareT<TypeTy, IdTy, ExprTy> &clause,
+ const ClauseTy *);
bool
applyClause(const tomp::clause::OmpxAttributeT<TypeTy, IdTy, ExprTy> &clause,
const ClauseTy *);
@@ -1103,6 +1105,13 @@ bool ConstructDecompositionT<C, H>::applyClause(
return applyToOutermost(node);
}
+template <typename C, typename H>
+bool ConstructDecompositionT<C, H>::applyClause(
+ const tomp::clause::OmpxBareT<TypeTy, IdTy, ExprTy> &clause,
+ const ClauseTy *node) {
+ return applyToAll(node);
+}
+
template <typename C, typename H>
bool ConstructDecompositionT<C, H>::applyClause(
const tomp::clause::OmpxAttributeT<TypeTy, IdTy, ExprTy> &clause,
diff --git a/llvm/include/llvm/Frontend/OpenMP/OMP.td b/llvm/include/llvm/Frontend/OpenMP/OMP.td
index fcf087d1f9c6e4..ac5cf1babb4176 100644
--- a/llvm/include/llvm/Frontend/OpenMP/OMP.td
+++ b/llvm/include/llvm/Frontend/OpenMP/OMP.td
@@ -980,6 +980,7 @@ def OMP_Target : Directive<"target"> {
VersionedClause<OMPC_Device>,
VersionedClause<OMPC_If>,
VersionedClause<OMPC_NoWait>,
+ VersionedClause<OMPC_OMPX_Bare>,
VersionedClause<OMPC_OMPX_DynCGroupMem>,
VersionedClause<OMPC_ThreadLimit, 51>,
];
diff --git a/mlir/include/mlir/Dialect/OpenMP/OpenMPClauses.td b/mlir/include/mlir/Dialect/OpenMP/OpenMPClauses.td
index 886554f66afffc..fe84aa744b9dbb 100644
--- a/mlir/include/mlir/Dialect/OpenMP/OpenMPClauses.td
+++ b/mlir/include/mlir/Dialect/OpenMP/OpenMPClauses.td
@@ -1270,4 +1270,29 @@ class OpenMP_UseDevicePtrClauseSkip<
def OpenMP_UseDevicePtrClause : OpenMP_UseDevicePtrClauseSkip<>;
+//===----------------------------------------------------------------------===//
+// LLVM OpenMP extension `ompx_bare` clause
+//===----------------------------------------------------------------------===//
+
+class OpenMP_BareClauseSkip<
+ bit traits = false, bit arguments = false, bit assemblyFormat = false,
+ bit description = false, bit extraClassDeclaration = false
+ > : OpenMP_Clause<traits, arguments, assemblyFormat, description,
+ extraClassDeclaration> {
+ let arguments = (ins
+ UnitAttr:$bare
+ );
+
+ let optAssemblyFormat = [{
+ `ompx_bare` $bare
+ }];
+
+ let description = [{
+ ompx_bare allows `omp target teams` to be executed on a GPU with multi-dim
+ teams and threads.
+ }];
+}
+
+def OpenMP_BareClause : OpenMP_BareClauseSkip<>;
+
#endif // OPENMP_CLAUSES
diff --git a/mlir/include/mlir/Dialect/OpenMP/OpenMPOps.td b/mlir/include/mlir/Dialect/OpenMP/OpenMPOps.td
index 66f63fc02fe2f3..ee8aeee1278b1c 100644
--- a/mlir/include/mlir/Dialect/OpenMP/OpenMPOps.td
+++ b/mlir/include/mlir/Dialect/OpenMP/OpenMPOps.td
@@ -1108,7 +1108,8 @@ def TargetOp : OpenMP_Op<"target", traits = [
OpenMP_AllocateClause, OpenMP_DependClause, OpenMP_DeviceClause,
OpenMP_HasDeviceAddrClause, OpenMP_IfClause, OpenMP_InReductionClause,
OpenMP_IsDevicePtrClause, OpenMP_MapClauseSkip<assemblyFormat = true>,
- OpenMP_NowaitClause, OpenMP_PrivateClause, OpenMP_ThreadLimitClause
+ OpenMP_NowaitClause, OpenMP_PrivateClause, OpenMP_ThreadLimitClause,
+ OpenMP_BareClause,
], singleRegion = true> {
let summary = "target construct";
let description = [{
diff --git a/mlir/lib/Dialect/OpenMP/IR/OpenMPDialect.cpp b/mlir/lib/Dialect/OpenMP/IR/OpenMPDialect.cpp
index d516c8d9e0be6c..a7c35a3bf3365b 100644
--- a/mlir/lib/Dialect/OpenMP/IR/OpenMPDialect.cpp
+++ b/mlir/lib/Dialect/OpenMP/IR/OpenMPDialect.cpp
@@ -1677,7 +1677,8 @@ void TargetOp::build(OpBuilder &builder, OperationState &state,
/*in_reduction_vars=*/{}, /*in_reduction_byref=*/nullptr,
/*in_reduction_syms=*/nullptr, clauses.isDevicePtrVars,
clauses.mapVars, clauses.nowait, clauses.privateVars,
- makeArrayAttr(ctx, clauses.privateSyms), clauses.threadLimit);
+ makeArrayAttr(ctx, clauses.privateSyms), clauses.threadLimit,
+ clauses.bare);
}
LogicalResult TargetOp::verify() {
|
@llvm/pr-subscribers-mlir Author: Ivan R. Ivanov (ivanradanov) ChangesFull diff: https://github.com/llvm/llvm-project/pull/111106.diff 11 Files Affected:
diff --git a/flang/lib/Lower/OpenMP/ClauseProcessor.cpp b/flang/lib/Lower/OpenMP/ClauseProcessor.cpp
index a4d2524bccf5c3..c7f34b3d65f324 100644
--- a/flang/lib/Lower/OpenMP/ClauseProcessor.cpp
+++ b/flang/lib/Lower/OpenMP/ClauseProcessor.cpp
@@ -354,6 +354,10 @@ bool ClauseProcessor::processNowait(mlir::omp::NowaitClauseOps &result) const {
return markClauseOccurrence<omp::clause::Nowait>(result.nowait);
}
+bool ClauseProcessor::processBare(mlir::omp::BareClauseOps &result) const {
+ return markClauseOccurrence<omp::clause::OmpxBare>(result.bare);
+}
+
bool ClauseProcessor::processNumTeams(
lower::StatementContext &stmtCtx,
mlir::omp::NumTeamsClauseOps &result) const {
diff --git a/flang/lib/Lower/OpenMP/ClauseProcessor.h b/flang/lib/Lower/OpenMP/ClauseProcessor.h
index 0c8e7bd47ab5a6..a04731f94ede37 100644
--- a/flang/lib/Lower/OpenMP/ClauseProcessor.h
+++ b/flang/lib/Lower/OpenMP/ClauseProcessor.h
@@ -74,6 +74,7 @@ class ClauseProcessor {
bool processHint(mlir::omp::HintClauseOps &result) const;
bool processMergeable(mlir::omp::MergeableClauseOps &result) const;
bool processNowait(mlir::omp::NowaitClauseOps &result) const;
+ bool processBare(mlir::omp::BareClauseOps &result) const;
bool processNumTeams(lower::StatementContext &stmtCtx,
mlir::omp::NumTeamsClauseOps &result) const;
bool processNumThreads(lower::StatementContext &stmtCtx,
diff --git a/flang/lib/Lower/OpenMP/Clauses.h b/flang/lib/Lower/OpenMP/Clauses.h
index 51bf0eab0f8d07..10bd1439563244 100644
--- a/flang/lib/Lower/OpenMP/Clauses.h
+++ b/flang/lib/Lower/OpenMP/Clauses.h
@@ -229,8 +229,8 @@ using NumTasks = tomp::clause::NumTasksT<TypeTy, IdTy, ExprTy>;
using NumTeams = tomp::clause::NumTeamsT<TypeTy, IdTy, ExprTy>;
using NumThreads = tomp::clause::NumThreadsT<TypeTy, IdTy, ExprTy>;
using OmpxAttribute = tomp::clause::OmpxAttributeT<TypeTy, IdTy, ExprTy>;
-using OmpxBare = tomp::clause::OmpxBareT<TypeTy, IdTy, ExprTy>;
using OmpxDynCgroupMem = tomp::clause::OmpxDynCgroupMemT<TypeTy, IdTy, ExprTy>;
+using OmpxBare = tomp::clause::OmpxBareT<TypeTy, IdTy, ExprTy>;
using Ordered = tomp::clause::OrderedT<TypeTy, IdTy, ExprTy>;
using Order = tomp::clause::OrderT<TypeTy, IdTy, ExprTy>;
using Partial = tomp::clause::PartialT<TypeTy, IdTy, ExprTy>;
diff --git a/flang/lib/Lower/OpenMP/OpenMP.cpp b/flang/lib/Lower/OpenMP/OpenMP.cpp
index 60c83586e468b6..b8b1a0ba2a69b7 100644
--- a/flang/lib/Lower/OpenMP/OpenMP.cpp
+++ b/flang/lib/Lower/OpenMP/OpenMP.cpp
@@ -1180,6 +1180,7 @@ static void genTargetClauses(
cp.processNowait(clauseOps);
cp.processThreadLimit(stmtCtx, clauseOps);
+ cp.processBare(clauseOps);
cp.processTODO<clause::Allocate, clause::Defaultmap, clause::Firstprivate,
clause::InReduction, clause::UsesAllocators>(
@@ -2764,6 +2765,7 @@ static void genOMP(lower::AbstractConverter &converter, lower::SymMap &symTable,
!std::holds_alternative<clause::ThreadLimit>(clause.u) &&
!std::holds_alternative<clause::Threads>(clause.u) &&
!std::holds_alternative<clause::UseDeviceAddr>(clause.u) &&
+ !std::holds_alternative<clause::OmpxBare>(clause.u) &&
!std::holds_alternative<clause::UseDevicePtr>(clause.u)) {
TODO(clauseLocation, "OpenMP Block construct clause");
}
diff --git a/flang/lib/Parser/openmp-parsers.cpp b/flang/lib/Parser/openmp-parsers.cpp
index cc2930cbd7ded5..f18a9929aa75ef 100644
--- a/flang/lib/Parser/openmp-parsers.cpp
+++ b/flang/lib/Parser/openmp-parsers.cpp
@@ -315,6 +315,7 @@ TYPE_PARSER(
parenthesized(scalarIntExpr))) ||
"NUM_THREADS" >> construct<OmpClause>(construct<OmpClause::NumThreads>(
parenthesized(scalarIntExpr))) ||
+ "OMPX_BARE" >> construct<OmpClause>(construct<OmpClause::OmpxBare>()) ||
"ORDER" >> construct<OmpClause>(construct<OmpClause::Order>(
parenthesized(Parser<OmpOrderClause>{}))) ||
"ORDERED" >> construct<OmpClause>(construct<OmpClause::Ordered>(
diff --git a/flang/test/Lower/OpenMP/KernelLanguage/bare-clause.f90 b/flang/test/Lower/OpenMP/KernelLanguage/bare-clause.f90
new file mode 100644
index 00000000000000..2a97a14516ec3a
--- /dev/null
+++ b/flang/test/Lower/OpenMP/KernelLanguage/bare-clause.f90
@@ -0,0 +1,10 @@
+! RUN: %flang_fc1 -emit-hlfir %openmp_flags -fopenmp-version=51 %s -o - | FileCheck %s
+
+program test
+ integer :: tmp
+ !$omp target teams ompx_bare num_teams(42) thread_limit(43)
+ tmp = 1
+ !$omp end target teams
+end program
+
+! CHECK: omp.target {{.*}} ompx_bare
diff --git a/llvm/include/llvm/Frontend/OpenMP/ConstructDecompositionT.h b/llvm/include/llvm/Frontend/OpenMP/ConstructDecompositionT.h
index 4bdfa1cf4c1490..7c0dbe1aae1221 100644
--- a/llvm/include/llvm/Frontend/OpenMP/ConstructDecompositionT.h
+++ b/llvm/include/llvm/Frontend/OpenMP/ConstructDecompositionT.h
@@ -236,6 +236,8 @@ struct ConstructDecompositionT {
const ClauseTy *);
bool applyClause(const tomp::clause::NowaitT<TypeTy, IdTy, ExprTy> &clause,
const ClauseTy *);
+ bool applyClause(const tomp::clause::OmpxBareT<TypeTy, IdTy, ExprTy> &clause,
+ const ClauseTy *);
bool
applyClause(const tomp::clause::OmpxAttributeT<TypeTy, IdTy, ExprTy> &clause,
const ClauseTy *);
@@ -1103,6 +1105,13 @@ bool ConstructDecompositionT<C, H>::applyClause(
return applyToOutermost(node);
}
+template <typename C, typename H>
+bool ConstructDecompositionT<C, H>::applyClause(
+ const tomp::clause::OmpxBareT<TypeTy, IdTy, ExprTy> &clause,
+ const ClauseTy *node) {
+ return applyToAll(node);
+}
+
template <typename C, typename H>
bool ConstructDecompositionT<C, H>::applyClause(
const tomp::clause::OmpxAttributeT<TypeTy, IdTy, ExprTy> &clause,
diff --git a/llvm/include/llvm/Frontend/OpenMP/OMP.td b/llvm/include/llvm/Frontend/OpenMP/OMP.td
index fcf087d1f9c6e4..ac5cf1babb4176 100644
--- a/llvm/include/llvm/Frontend/OpenMP/OMP.td
+++ b/llvm/include/llvm/Frontend/OpenMP/OMP.td
@@ -980,6 +980,7 @@ def OMP_Target : Directive<"target"> {
VersionedClause<OMPC_Device>,
VersionedClause<OMPC_If>,
VersionedClause<OMPC_NoWait>,
+ VersionedClause<OMPC_OMPX_Bare>,
VersionedClause<OMPC_OMPX_DynCGroupMem>,
VersionedClause<OMPC_ThreadLimit, 51>,
];
diff --git a/mlir/include/mlir/Dialect/OpenMP/OpenMPClauses.td b/mlir/include/mlir/Dialect/OpenMP/OpenMPClauses.td
index 886554f66afffc..fe84aa744b9dbb 100644
--- a/mlir/include/mlir/Dialect/OpenMP/OpenMPClauses.td
+++ b/mlir/include/mlir/Dialect/OpenMP/OpenMPClauses.td
@@ -1270,4 +1270,29 @@ class OpenMP_UseDevicePtrClauseSkip<
def OpenMP_UseDevicePtrClause : OpenMP_UseDevicePtrClauseSkip<>;
+//===----------------------------------------------------------------------===//
+// LLVM OpenMP extension `ompx_bare` clause
+//===----------------------------------------------------------------------===//
+
+class OpenMP_BareClauseSkip<
+ bit traits = false, bit arguments = false, bit assemblyFormat = false,
+ bit description = false, bit extraClassDeclaration = false
+ > : OpenMP_Clause<traits, arguments, assemblyFormat, description,
+ extraClassDeclaration> {
+ let arguments = (ins
+ UnitAttr:$bare
+ );
+
+ let optAssemblyFormat = [{
+ `ompx_bare` $bare
+ }];
+
+ let description = [{
+ ompx_bare allows `omp target teams` to be executed on a GPU with multi-dim
+ teams and threads.
+ }];
+}
+
+def OpenMP_BareClause : OpenMP_BareClauseSkip<>;
+
#endif // OPENMP_CLAUSES
diff --git a/mlir/include/mlir/Dialect/OpenMP/OpenMPOps.td b/mlir/include/mlir/Dialect/OpenMP/OpenMPOps.td
index 66f63fc02fe2f3..ee8aeee1278b1c 100644
--- a/mlir/include/mlir/Dialect/OpenMP/OpenMPOps.td
+++ b/mlir/include/mlir/Dialect/OpenMP/OpenMPOps.td
@@ -1108,7 +1108,8 @@ def TargetOp : OpenMP_Op<"target", traits = [
OpenMP_AllocateClause, OpenMP_DependClause, OpenMP_DeviceClause,
OpenMP_HasDeviceAddrClause, OpenMP_IfClause, OpenMP_InReductionClause,
OpenMP_IsDevicePtrClause, OpenMP_MapClauseSkip<assemblyFormat = true>,
- OpenMP_NowaitClause, OpenMP_PrivateClause, OpenMP_ThreadLimitClause
+ OpenMP_NowaitClause, OpenMP_PrivateClause, OpenMP_ThreadLimitClause,
+ OpenMP_BareClause,
], singleRegion = true> {
let summary = "target construct";
let description = [{
diff --git a/mlir/lib/Dialect/OpenMP/IR/OpenMPDialect.cpp b/mlir/lib/Dialect/OpenMP/IR/OpenMPDialect.cpp
index d516c8d9e0be6c..a7c35a3bf3365b 100644
--- a/mlir/lib/Dialect/OpenMP/IR/OpenMPDialect.cpp
+++ b/mlir/lib/Dialect/OpenMP/IR/OpenMPDialect.cpp
@@ -1677,7 +1677,8 @@ void TargetOp::build(OpBuilder &builder, OperationState &state,
/*in_reduction_vars=*/{}, /*in_reduction_byref=*/nullptr,
/*in_reduction_syms=*/nullptr, clauses.isDevicePtrVars,
clauses.mapVars, clauses.nowait, clauses.privateVars,
- makeArrayAttr(ctx, clauses.privateSyms), clauses.threadLimit);
+ makeArrayAttr(ctx, clauses.privateSyms), clauses.threadLimit,
+ clauses.bare);
}
LogicalResult TargetOp::verify() {
|
@llvm/pr-subscribers-flang-openmp Author: Ivan R. Ivanov (ivanradanov) ChangesFull diff: https://github.com/llvm/llvm-project/pull/111106.diff 11 Files Affected:
diff --git a/flang/lib/Lower/OpenMP/ClauseProcessor.cpp b/flang/lib/Lower/OpenMP/ClauseProcessor.cpp
index a4d2524bccf5c3..c7f34b3d65f324 100644
--- a/flang/lib/Lower/OpenMP/ClauseProcessor.cpp
+++ b/flang/lib/Lower/OpenMP/ClauseProcessor.cpp
@@ -354,6 +354,10 @@ bool ClauseProcessor::processNowait(mlir::omp::NowaitClauseOps &result) const {
return markClauseOccurrence<omp::clause::Nowait>(result.nowait);
}
+bool ClauseProcessor::processBare(mlir::omp::BareClauseOps &result) const {
+ return markClauseOccurrence<omp::clause::OmpxBare>(result.bare);
+}
+
bool ClauseProcessor::processNumTeams(
lower::StatementContext &stmtCtx,
mlir::omp::NumTeamsClauseOps &result) const {
diff --git a/flang/lib/Lower/OpenMP/ClauseProcessor.h b/flang/lib/Lower/OpenMP/ClauseProcessor.h
index 0c8e7bd47ab5a6..a04731f94ede37 100644
--- a/flang/lib/Lower/OpenMP/ClauseProcessor.h
+++ b/flang/lib/Lower/OpenMP/ClauseProcessor.h
@@ -74,6 +74,7 @@ class ClauseProcessor {
bool processHint(mlir::omp::HintClauseOps &result) const;
bool processMergeable(mlir::omp::MergeableClauseOps &result) const;
bool processNowait(mlir::omp::NowaitClauseOps &result) const;
+ bool processBare(mlir::omp::BareClauseOps &result) const;
bool processNumTeams(lower::StatementContext &stmtCtx,
mlir::omp::NumTeamsClauseOps &result) const;
bool processNumThreads(lower::StatementContext &stmtCtx,
diff --git a/flang/lib/Lower/OpenMP/Clauses.h b/flang/lib/Lower/OpenMP/Clauses.h
index 51bf0eab0f8d07..10bd1439563244 100644
--- a/flang/lib/Lower/OpenMP/Clauses.h
+++ b/flang/lib/Lower/OpenMP/Clauses.h
@@ -229,8 +229,8 @@ using NumTasks = tomp::clause::NumTasksT<TypeTy, IdTy, ExprTy>;
using NumTeams = tomp::clause::NumTeamsT<TypeTy, IdTy, ExprTy>;
using NumThreads = tomp::clause::NumThreadsT<TypeTy, IdTy, ExprTy>;
using OmpxAttribute = tomp::clause::OmpxAttributeT<TypeTy, IdTy, ExprTy>;
-using OmpxBare = tomp::clause::OmpxBareT<TypeTy, IdTy, ExprTy>;
using OmpxDynCgroupMem = tomp::clause::OmpxDynCgroupMemT<TypeTy, IdTy, ExprTy>;
+using OmpxBare = tomp::clause::OmpxBareT<TypeTy, IdTy, ExprTy>;
using Ordered = tomp::clause::OrderedT<TypeTy, IdTy, ExprTy>;
using Order = tomp::clause::OrderT<TypeTy, IdTy, ExprTy>;
using Partial = tomp::clause::PartialT<TypeTy, IdTy, ExprTy>;
diff --git a/flang/lib/Lower/OpenMP/OpenMP.cpp b/flang/lib/Lower/OpenMP/OpenMP.cpp
index 60c83586e468b6..b8b1a0ba2a69b7 100644
--- a/flang/lib/Lower/OpenMP/OpenMP.cpp
+++ b/flang/lib/Lower/OpenMP/OpenMP.cpp
@@ -1180,6 +1180,7 @@ static void genTargetClauses(
cp.processNowait(clauseOps);
cp.processThreadLimit(stmtCtx, clauseOps);
+ cp.processBare(clauseOps);
cp.processTODO<clause::Allocate, clause::Defaultmap, clause::Firstprivate,
clause::InReduction, clause::UsesAllocators>(
@@ -2764,6 +2765,7 @@ static void genOMP(lower::AbstractConverter &converter, lower::SymMap &symTable,
!std::holds_alternative<clause::ThreadLimit>(clause.u) &&
!std::holds_alternative<clause::Threads>(clause.u) &&
!std::holds_alternative<clause::UseDeviceAddr>(clause.u) &&
+ !std::holds_alternative<clause::OmpxBare>(clause.u) &&
!std::holds_alternative<clause::UseDevicePtr>(clause.u)) {
TODO(clauseLocation, "OpenMP Block construct clause");
}
diff --git a/flang/lib/Parser/openmp-parsers.cpp b/flang/lib/Parser/openmp-parsers.cpp
index cc2930cbd7ded5..f18a9929aa75ef 100644
--- a/flang/lib/Parser/openmp-parsers.cpp
+++ b/flang/lib/Parser/openmp-parsers.cpp
@@ -315,6 +315,7 @@ TYPE_PARSER(
parenthesized(scalarIntExpr))) ||
"NUM_THREADS" >> construct<OmpClause>(construct<OmpClause::NumThreads>(
parenthesized(scalarIntExpr))) ||
+ "OMPX_BARE" >> construct<OmpClause>(construct<OmpClause::OmpxBare>()) ||
"ORDER" >> construct<OmpClause>(construct<OmpClause::Order>(
parenthesized(Parser<OmpOrderClause>{}))) ||
"ORDERED" >> construct<OmpClause>(construct<OmpClause::Ordered>(
diff --git a/flang/test/Lower/OpenMP/KernelLanguage/bare-clause.f90 b/flang/test/Lower/OpenMP/KernelLanguage/bare-clause.f90
new file mode 100644
index 00000000000000..2a97a14516ec3a
--- /dev/null
+++ b/flang/test/Lower/OpenMP/KernelLanguage/bare-clause.f90
@@ -0,0 +1,10 @@
+! RUN: %flang_fc1 -emit-hlfir %openmp_flags -fopenmp-version=51 %s -o - | FileCheck %s
+
+program test
+ integer :: tmp
+ !$omp target teams ompx_bare num_teams(42) thread_limit(43)
+ tmp = 1
+ !$omp end target teams
+end program
+
+! CHECK: omp.target {{.*}} ompx_bare
diff --git a/llvm/include/llvm/Frontend/OpenMP/ConstructDecompositionT.h b/llvm/include/llvm/Frontend/OpenMP/ConstructDecompositionT.h
index 4bdfa1cf4c1490..7c0dbe1aae1221 100644
--- a/llvm/include/llvm/Frontend/OpenMP/ConstructDecompositionT.h
+++ b/llvm/include/llvm/Frontend/OpenMP/ConstructDecompositionT.h
@@ -236,6 +236,8 @@ struct ConstructDecompositionT {
const ClauseTy *);
bool applyClause(const tomp::clause::NowaitT<TypeTy, IdTy, ExprTy> &clause,
const ClauseTy *);
+ bool applyClause(const tomp::clause::OmpxBareT<TypeTy, IdTy, ExprTy> &clause,
+ const ClauseTy *);
bool
applyClause(const tomp::clause::OmpxAttributeT<TypeTy, IdTy, ExprTy> &clause,
const ClauseTy *);
@@ -1103,6 +1105,13 @@ bool ConstructDecompositionT<C, H>::applyClause(
return applyToOutermost(node);
}
+template <typename C, typename H>
+bool ConstructDecompositionT<C, H>::applyClause(
+ const tomp::clause::OmpxBareT<TypeTy, IdTy, ExprTy> &clause,
+ const ClauseTy *node) {
+ return applyToAll(node);
+}
+
template <typename C, typename H>
bool ConstructDecompositionT<C, H>::applyClause(
const tomp::clause::OmpxAttributeT<TypeTy, IdTy, ExprTy> &clause,
diff --git a/llvm/include/llvm/Frontend/OpenMP/OMP.td b/llvm/include/llvm/Frontend/OpenMP/OMP.td
index fcf087d1f9c6e4..ac5cf1babb4176 100644
--- a/llvm/include/llvm/Frontend/OpenMP/OMP.td
+++ b/llvm/include/llvm/Frontend/OpenMP/OMP.td
@@ -980,6 +980,7 @@ def OMP_Target : Directive<"target"> {
VersionedClause<OMPC_Device>,
VersionedClause<OMPC_If>,
VersionedClause<OMPC_NoWait>,
+ VersionedClause<OMPC_OMPX_Bare>,
VersionedClause<OMPC_OMPX_DynCGroupMem>,
VersionedClause<OMPC_ThreadLimit, 51>,
];
diff --git a/mlir/include/mlir/Dialect/OpenMP/OpenMPClauses.td b/mlir/include/mlir/Dialect/OpenMP/OpenMPClauses.td
index 886554f66afffc..fe84aa744b9dbb 100644
--- a/mlir/include/mlir/Dialect/OpenMP/OpenMPClauses.td
+++ b/mlir/include/mlir/Dialect/OpenMP/OpenMPClauses.td
@@ -1270,4 +1270,29 @@ class OpenMP_UseDevicePtrClauseSkip<
def OpenMP_UseDevicePtrClause : OpenMP_UseDevicePtrClauseSkip<>;
+//===----------------------------------------------------------------------===//
+// LLVM OpenMP extension `ompx_bare` clause
+//===----------------------------------------------------------------------===//
+
+class OpenMP_BareClauseSkip<
+ bit traits = false, bit arguments = false, bit assemblyFormat = false,
+ bit description = false, bit extraClassDeclaration = false
+ > : OpenMP_Clause<traits, arguments, assemblyFormat, description,
+ extraClassDeclaration> {
+ let arguments = (ins
+ UnitAttr:$bare
+ );
+
+ let optAssemblyFormat = [{
+ `ompx_bare` $bare
+ }];
+
+ let description = [{
+ ompx_bare allows `omp target teams` to be executed on a GPU with multi-dim
+ teams and threads.
+ }];
+}
+
+def OpenMP_BareClause : OpenMP_BareClauseSkip<>;
+
#endif // OPENMP_CLAUSES
diff --git a/mlir/include/mlir/Dialect/OpenMP/OpenMPOps.td b/mlir/include/mlir/Dialect/OpenMP/OpenMPOps.td
index 66f63fc02fe2f3..ee8aeee1278b1c 100644
--- a/mlir/include/mlir/Dialect/OpenMP/OpenMPOps.td
+++ b/mlir/include/mlir/Dialect/OpenMP/OpenMPOps.td
@@ -1108,7 +1108,8 @@ def TargetOp : OpenMP_Op<"target", traits = [
OpenMP_AllocateClause, OpenMP_DependClause, OpenMP_DeviceClause,
OpenMP_HasDeviceAddrClause, OpenMP_IfClause, OpenMP_InReductionClause,
OpenMP_IsDevicePtrClause, OpenMP_MapClauseSkip<assemblyFormat = true>,
- OpenMP_NowaitClause, OpenMP_PrivateClause, OpenMP_ThreadLimitClause
+ OpenMP_NowaitClause, OpenMP_PrivateClause, OpenMP_ThreadLimitClause,
+ OpenMP_BareClause,
], singleRegion = true> {
let summary = "target construct";
let description = [{
diff --git a/mlir/lib/Dialect/OpenMP/IR/OpenMPDialect.cpp b/mlir/lib/Dialect/OpenMP/IR/OpenMPDialect.cpp
index d516c8d9e0be6c..a7c35a3bf3365b 100644
--- a/mlir/lib/Dialect/OpenMP/IR/OpenMPDialect.cpp
+++ b/mlir/lib/Dialect/OpenMP/IR/OpenMPDialect.cpp
@@ -1677,7 +1677,8 @@ void TargetOp::build(OpBuilder &builder, OperationState &state,
/*in_reduction_vars=*/{}, /*in_reduction_byref=*/nullptr,
/*in_reduction_syms=*/nullptr, clauses.isDevicePtrVars,
clauses.mapVars, clauses.nowait, clauses.privateVars,
- makeArrayAttr(ctx, clauses.privateSyms), clauses.threadLimit);
+ makeArrayAttr(ctx, clauses.privateSyms), clauses.threadLimit,
+ clauses.bare);
}
LogicalResult TargetOp::verify() {
|
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
LGTM
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Please add a TODO message in Lowering if there is no translation from OpenMP dialect to LLVM IR. Otherwise this will manifest as a crash.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
The general approach to add the clause to the dialect looks good to me. One question I have is whether ompx_bare
is only specifically supported for the combined target teams
construct or if it can appear on a standalone target
. If that's the case, I guess a semantics check for it would also be in order.
At the moment, the num_teams_upper
and thread_limit
arguments are defined as optional. I'm guessing they should become variadic and add some MLIR verification code to ensure the ompx_bare
clause, num_teams
and thread_limit
are always in sync. Is that the plan or did you have something else in mind?
I actually tried to implement the semantic check for ompx_bare a while back before I realized that the parser doesn't handle it. I might still have that code lying around somewhere. We should include the corresponding check in clang as well. I'll try to add a commit to this PR with these changes. |
I don't have permission to push to Ivan's fork, so I pushed it to my fork. It's the most recent commit: |
3d9064b
to
0638f66
Compare
@jdoerfert Could you have a quick look at the clang-side change if that is acceptable? |
62e67d2
to
bde9b8c
Compare
No description provided.