[PATCH] D90101: [Clang][OpenMP][WIP] Avoid unnecessary privatization of mapper array when there is no user defined mapper
tianshilei1992 updated this revision to Diff 300738.
tianshilei1992 added a comment.
Fixed two test cases
Repository:
rG LLVM Github Monorepo
CHANGES SINCE LAST ACTION
https://reviews.llvm.org/D90101/new/
https://reviews.llvm.org/D90101
Files:
clang/lib/CodeGen/CGOpenMPRuntime.cpp
clang/lib/CodeGen/CGStmtOpenMP.cpp
clang/test/OpenMP/target_codegen.cpp
clang/test/OpenMP/target_enter_data_depend_codegen.cpp
clang/test/OpenMP/target_exit_data_depend_codegen.cpp
clang/test/OpenMP/target_parallel_depend_codegen.cpp
clang/test/OpenMP/target_parallel_for_codegen.cpp
clang/test/OpenMP/target_parallel_for_depend_codegen.cpp
clang/test/OpenMP/target_parallel_for_simd_depend_codegen.cpp
clang/test/OpenMP/target_simd_depend_codegen.cpp
clang/test/OpenMP/target_teams_codegen.cpp
clang/test/OpenMP/target_teams_depend_codegen.cpp
clang/test/OpenMP/target_teams_distribute_codegen.cpp
clang/test/OpenMP/target_teams_distribute_depend_codegen.cpp
clang/test/OpenMP/target_teams_distribute_parallel_for_depend_codegen.cpp
clang/test/OpenMP/target_teams_distribute_parallel_for_simd_depend_codegen.cpp
clang/test/OpenMP/target_teams_distribute_simd_codegen.cpp
clang/test/OpenMP/target_teams_distribute_simd_depend_codegen.cpp
clang/test/OpenMP/target_update_depend_codegen.cpp
Index: clang/test/OpenMP/target_update_depend_codegen.cpp
===
--- clang/test/OpenMP/target_update_depend_codegen.cpp
+++ clang/test/OpenMP/target_update_depend_codegen.cpp
@@ -64,7 +64,7 @@
// CK1: store i32 [[DEVICE]], i32* [[CAP_DEVICE]],
// CK1: [[DEV1:%.+]] = load i32, i32* %{{.+}}
// CK1: [[DEV2:%.+]] = sext i32 [[DEV1]] to i64
- // CK1: [[RES:%.+]] = call i8* @__kmpc_omp_target_task_alloc(%struct.ident_t* {{.+}}, i32 {{.+}}, i32 1, i[[sz]] {{72|40}}, i[[sz]] 4, i32 (i32, i8*)* bitcast (i32 (i32, %struct.kmp_task_t_with_privates*)* [[TASK_ENTRY0:@.+]] to i32 (i32, i8*)*), i64 [[DEV2]])
+ // CK1: [[RES:%.+]] = call i8* @__kmpc_omp_target_task_alloc(%struct.ident_t* {{.+}}, i32 {{.+}}, i32 1, i[[sz]] {{64|36}}, i[[sz]] 4, i32 (i32, i8*)* bitcast (i32 (i32, %struct.kmp_task_t_with_privates*)* [[TASK_ENTRY0:@.+]] to i32 (i32, i8*)*), i64 [[DEV2]])
// CK1: [[BC:%.+]] = bitcast i8* [[RES]] to %struct.kmp_task_t_with_privates*
// CK1: [[TASK_T:%.+]] = getelementptr inbounds %struct.kmp_task_t_with_privates, %struct.kmp_task_t_with_privates* [[BC]], i32 0, i32 0
// CK1: [[SHAREDS:%.+]] = getelementptr inbounds %struct.kmp_task_t, %struct.kmp_task_t* [[TASK_T]], i32 0, i32 0
@@ -130,7 +130,7 @@
// CK1: [[IF_BOOL:%.+]] = trunc i8 [[IF]] to i1
// CK1: [[IF:%.+]] = zext i1 [[IF_BOOL]] to i8
// CK1: store i8 [[IF]], i8* [[IF_DEVICE]],
- // CK1: [[RES:%.+]] = call i8* @__kmpc_omp_task_alloc(%struct.ident_t* {{.+}}, i32 {{.+}}, i32 1, i[[sz]] {{72|40}}, i[[sz]] 1, i32 (i32, i8*)* bitcast (i32 (i32, %struct.kmp_task_t_with_privates{{.+}}*)* [[TASK_ENTRY2:@.+]] to i32 (i32, i8*)*))
+ // CK1: [[RES:%.+]] = call i8* @__kmpc_omp_task_alloc(%struct.ident_t* {{.+}}, i32 {{.+}}, i32 1, i[[sz]] {{64|36}}, i[[sz]] 1, i32 (i32, i8*)* bitcast (i32 (i32, %struct.kmp_task_t_with_privates{{.+}}*)* [[TASK_ENTRY2:@.+]] to i32 (i32, i8*)*))
// CK1: [[RES_BC:%.+]] = bitcast i8* [[RES]] to %struct.kmp_task_t_with_privates{{.+}}*
// CK1: [[TASK_T:%.+]] = getelementptr inbounds %struct.kmp_task_t_with_privates{{.+}}, %struct.kmp_task_t_with_privates{{.+}}* [[RES_BC]], i32 0, i32 0
// CK1: [[SHAREDS:%.+]] = getelementptr inbounds %struct.kmp_task_t, %struct.kmp_task_t* [[TASK_T]], i32 0, i32 0
@@ -213,7 +213,7 @@
// CK1: [[GEPBP0:%.+]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[BP]], i32 0, i32 0
// CK1: [[GEPP0:%.+]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[P]], i32 0, i32 0
// CK1: [[GEPS0:%.+]] = getelementptr inbounds [1 x i64], [1 x i64]* [[S]], i32 0, i32 0
- // CK1: [[RES:%.+]] = call i8* @__kmpc_omp_task_alloc(%struct.ident_t* {{.+}}, i32 {{.+}}, i32 1, i[[sz]] {{72|40}}, i[[sz]] 1, i32 (i32, i8*)* bitcast (i32 (i32, %struct.kmp_task_t_with_privates{{.+}}*)* [[TASK_ENTRY3:@.+]] to i32 (i32, i8*)*))
+ // CK1: [[RES:%.+]] = call i8* @__kmpc_omp_task_alloc(%struct.ident_t* {{.+}}, i32 {{.+}}, i32 1, i[[sz]] {{64|36}}, i[[sz]] 1, i32 (i32, i8*)* bitcast (i32 (i32, %struct.kmp_task_t_with_privates{{.+}}*)* [[TASK_ENTRY3:@.+]] to i32 (i32, i8*)*))
// CK1: [[RES_BC:%.+]] = bitcast i8* [[RES]] to %struct.kmp_task_t_with_privates{{.+}}*
// CK1: [[TASK_T:%.+]] = getelementptr inbounds %struct.kmp_task_t_with_privates{{.+}}, %struct.kmp_task_t_with_privates{{.+}}* [[RES_BC]], i32 0, i32 0
// CK1: [[PRIVS:%.+]] = getelementptr inbounds %struct.kmp_task_t_with_privates{{.+}}, %struct.kmp_task_t_with_privates{{.+}}* [[RES_BC]], i32 0, i32 1
@@ -298,7 +298,7 @@
// CK1: store double* %{{.+}}, double** [[P1_BC]],
// CK1: [[GEPBP0:%.+]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[BP]], i32 0, i32 0
//
[PATCH] D90101: [Clang][OpenMP][WIP] Avoid unnecessary privatization of mapper array when there is no user defined mapper
jdoerfert accepted this revision. jdoerfert added a comment. This revision is now accepted and ready to land. 4 tests seem to fail still, otherwise this LGTM. Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D90101/new/ https://reviews.llvm.org/D90101 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[PATCH] D90101: [Clang][OpenMP][WIP] Avoid unnecessary privatization of mapper array when there is no user defined mapper
tianshilei1992 updated this revision to Diff 300517.
tianshilei1992 added a comment.
Fixed some test cases
Repository:
rG LLVM Github Monorepo
CHANGES SINCE LAST ACTION
https://reviews.llvm.org/D90101/new/
https://reviews.llvm.org/D90101
Files:
clang/lib/CodeGen/CGOpenMPRuntime.cpp
clang/lib/CodeGen/CGStmtOpenMP.cpp
clang/test/OpenMP/target_codegen.cpp
clang/test/OpenMP/target_enter_data_depend_codegen.cpp
clang/test/OpenMP/target_exit_data_depend_codegen.cpp
clang/test/OpenMP/target_parallel_depend_codegen.cpp
clang/test/OpenMP/target_parallel_for_codegen.cpp
clang/test/OpenMP/target_parallel_for_depend_codegen.cpp
clang/test/OpenMP/target_parallel_for_simd_depend_codegen.cpp
clang/test/OpenMP/target_simd_depend_codegen.cpp
clang/test/OpenMP/target_teams_codegen.cpp
clang/test/OpenMP/target_teams_depend_codegen.cpp
clang/test/OpenMP/target_teams_distribute_codegen.cpp
clang/test/OpenMP/target_teams_distribute_depend_codegen.cpp
clang/test/OpenMP/target_teams_distribute_parallel_for_depend_codegen.cpp
clang/test/OpenMP/target_teams_distribute_parallel_for_simd_depend_codegen.cpp
clang/test/OpenMP/target_teams_distribute_simd_codegen.cpp
clang/test/OpenMP/target_teams_distribute_simd_depend_codegen.cpp
clang/test/OpenMP/target_update_depend_codegen.cpp
Index: clang/test/OpenMP/target_update_depend_codegen.cpp
===
--- clang/test/OpenMP/target_update_depend_codegen.cpp
+++ clang/test/OpenMP/target_update_depend_codegen.cpp
@@ -64,7 +64,7 @@
// CK1: store i32 [[DEVICE]], i32* [[CAP_DEVICE]],
// CK1: [[DEV1:%.+]] = load i32, i32* %{{.+}}
// CK1: [[DEV2:%.+]] = sext i32 [[DEV1]] to i64
- // CK1: [[RES:%.+]] = call i8* @__kmpc_omp_target_task_alloc(%struct.ident_t* {{.+}}, i32 {{.+}}, i32 1, i[[sz]] {{72|40}}, i[[sz]] 4, i32 (i32, i8*)* bitcast (i32 (i32, %struct.kmp_task_t_with_privates*)* [[TASK_ENTRY0:@.+]] to i32 (i32, i8*)*), i64 [[DEV2]])
+ // CK1: [[RES:%.+]] = call i8* @__kmpc_omp_target_task_alloc(%struct.ident_t* {{.+}}, i32 {{.+}}, i32 1, i[[sz]] {{64|36}}, i[[sz]] 4, i32 (i32, i8*)* bitcast (i32 (i32, %struct.kmp_task_t_with_privates*)* [[TASK_ENTRY0:@.+]] to i32 (i32, i8*)*), i64 [[DEV2]])
// CK1: [[BC:%.+]] = bitcast i8* [[RES]] to %struct.kmp_task_t_with_privates*
// CK1: [[TASK_T:%.+]] = getelementptr inbounds %struct.kmp_task_t_with_privates, %struct.kmp_task_t_with_privates* [[BC]], i32 0, i32 0
// CK1: [[SHAREDS:%.+]] = getelementptr inbounds %struct.kmp_task_t, %struct.kmp_task_t* [[TASK_T]], i32 0, i32 0
@@ -130,7 +130,7 @@
// CK1: [[IF_BOOL:%.+]] = trunc i8 [[IF]] to i1
// CK1: [[IF:%.+]] = zext i1 [[IF_BOOL]] to i8
// CK1: store i8 [[IF]], i8* [[IF_DEVICE]],
- // CK1: [[RES:%.+]] = call i8* @__kmpc_omp_task_alloc(%struct.ident_t* {{.+}}, i32 {{.+}}, i32 1, i[[sz]] {{72|40}}, i[[sz]] 1, i32 (i32, i8*)* bitcast (i32 (i32, %struct.kmp_task_t_with_privates{{.+}}*)* [[TASK_ENTRY2:@.+]] to i32 (i32, i8*)*))
+ // CK1: [[RES:%.+]] = call i8* @__kmpc_omp_task_alloc(%struct.ident_t* {{.+}}, i32 {{.+}}, i32 1, i[[sz]] {{64|36}}, i[[sz]] 1, i32 (i32, i8*)* bitcast (i32 (i32, %struct.kmp_task_t_with_privates{{.+}}*)* [[TASK_ENTRY2:@.+]] to i32 (i32, i8*)*))
// CK1: [[RES_BC:%.+]] = bitcast i8* [[RES]] to %struct.kmp_task_t_with_privates{{.+}}*
// CK1: [[TASK_T:%.+]] = getelementptr inbounds %struct.kmp_task_t_with_privates{{.+}}, %struct.kmp_task_t_with_privates{{.+}}* [[RES_BC]], i32 0, i32 0
// CK1: [[SHAREDS:%.+]] = getelementptr inbounds %struct.kmp_task_t, %struct.kmp_task_t* [[TASK_T]], i32 0, i32 0
@@ -213,7 +213,7 @@
// CK1: [[GEPBP0:%.+]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[BP]], i32 0, i32 0
// CK1: [[GEPP0:%.+]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[P]], i32 0, i32 0
// CK1: [[GEPS0:%.+]] = getelementptr inbounds [1 x i64], [1 x i64]* [[S]], i32 0, i32 0
- // CK1: [[RES:%.+]] = call i8* @__kmpc_omp_task_alloc(%struct.ident_t* {{.+}}, i32 {{.+}}, i32 1, i[[sz]] {{72|40}}, i[[sz]] 1, i32 (i32, i8*)* bitcast (i32 (i32, %struct.kmp_task_t_with_privates{{.+}}*)* [[TASK_ENTRY3:@.+]] to i32 (i32, i8*)*))
+ // CK1: [[RES:%.+]] = call i8* @__kmpc_omp_task_alloc(%struct.ident_t* {{.+}}, i32 {{.+}}, i32 1, i[[sz]] {{64|36}}, i[[sz]] 1, i32 (i32, i8*)* bitcast (i32 (i32, %struct.kmp_task_t_with_privates{{.+}}*)* [[TASK_ENTRY3:@.+]] to i32 (i32, i8*)*))
// CK1: [[RES_BC:%.+]] = bitcast i8* [[RES]] to %struct.kmp_task_t_with_privates{{.+}}*
// CK1: [[TASK_T:%.+]] = getelementptr inbounds %struct.kmp_task_t_with_privates{{.+}}, %struct.kmp_task_t_with_privates{{.+}}* [[RES_BC]], i32 0, i32 0
// CK1: [[PRIVS:%.+]] = getelementptr inbounds %struct.kmp_task_t_with_privates{{.+}}, %struct.kmp_task_t_with_privates{{.+}}* [[RES_BC]], i32 0, i32 1
@@ -298,7 +298,7 @@
// CK1: store double* %{{.+}}, double** [[P1_BC]],
// CK1: [[GEPBP0:%.+]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[BP]], i32 0, i32 0
//
[PATCH] D90101: [Clang][OpenMP][WIP] Avoid unnecessary privatization of mapper array when there is no user defined mapper
tianshilei1992 updated this revision to Diff 300509.
tianshilei1992 added a comment.
Fixed an issue that caused compiler crash
Repository:
rG LLVM Github Monorepo
CHANGES SINCE LAST ACTION
https://reviews.llvm.org/D90101/new/
https://reviews.llvm.org/D90101
Files:
clang/lib/CodeGen/CGOpenMPRuntime.cpp
clang/lib/CodeGen/CGStmtOpenMP.cpp
Index: clang/lib/CodeGen/CGStmtOpenMP.cpp
===
--- clang/lib/CodeGen/CGStmtOpenMP.cpp
+++ clang/lib/CodeGen/CGStmtOpenMP.cpp
@@ -4210,16 +4210,21 @@
/*IndexTypeQuals=*/0);
SVD = createImplicitFirstprivateForType(getContext(), Data, SizesType, CD,
S.getBeginLoc());
-MVD = createImplicitFirstprivateForType(
-getContext(), Data, BaseAndPointerAndMapperType, CD, S.getBeginLoc());
TargetScope.addPrivate(
BPVD, []() { return InputInfo.BasePointersArray; });
TargetScope.addPrivate(PVD,
[]() { return InputInfo.PointersArray; });
TargetScope.addPrivate(SVD,
[]() { return InputInfo.SizesArray; });
-TargetScope.addPrivate(MVD,
- []() { return InputInfo.MappersArray; });
+// If there is no user-defined mapper, the mapper array will be nullptr. In
+// this case, we don't need to privatize it.
+if (!dyn_cast_or_null(
+InputInfo.MappersArray.getPointer())) {
+ MVD = createImplicitFirstprivateForType(
+ getContext(), Data, BaseAndPointerAndMapperType, CD, S.getBeginLoc());
+ TargetScope.addPrivate(MVD,
+ []() { return InputInfo.MappersArray; });
+}
}
(void)TargetScope.Privatize();
// Build list of dependences.
@@ -4269,8 +4274,10 @@
CGF.GetAddrOfLocalVar(PVD), /*Index=*/0);
InputInfo.SizesArray = CGF.Builder.CreateConstArrayGEP(
CGF.GetAddrOfLocalVar(SVD), /*Index=*/0);
- InputInfo.MappersArray = CGF.Builder.CreateConstArrayGEP(
- CGF.GetAddrOfLocalVar(MVD), /*Index=*/0);
+ // If MVD is nullptr, the mapper array is not privatized
+ if (MVD)
+InputInfo.MappersArray = CGF.Builder.CreateConstArrayGEP(
+CGF.GetAddrOfLocalVar(MVD), /*Index=*/0);
}
Action.Enter(CGF);
Index: clang/lib/CodeGen/CGOpenMPRuntime.cpp
===
--- clang/lib/CodeGen/CGOpenMPRuntime.cpp
+++ clang/lib/CodeGen/CGOpenMPRuntime.cpp
@@ -8870,10 +8870,8 @@
/// Additional arguments for emitOffloadingArraysArgument function.
struct ArgumentsOptions {
bool ForEndCall = false;
- bool IsTask = false;
ArgumentsOptions() = default;
- ArgumentsOptions(bool ForEndCall, bool IsTask)
- : ForEndCall(ForEndCall), IsTask(IsTask) {}
+ ArgumentsOptions(bool ForEndCall) : ForEndCall(ForEndCall) {}
};
} // namespace
@@ -8909,9 +8907,9 @@
: Info.MapTypesArray,
/*Idx0=*/0,
/*Idx1=*/0);
-// Always emit the mapper array address in case of a target task for
-// privatization.
-if (!Options.IsTask && !Info.HasMapper)
+// If there is no user-defined mapper, set the mapper array to nullptr to
+// avoid an unnecessary data privatization
+if (!Info.HasMapper)
MappersArrayArg = llvm::ConstantPointerNull::get(CGM.VoidPtrPtrTy);
else
MappersArrayArg =
@@ -9664,11 +9662,9 @@
TargetDataInfo Info;
// Fill up the arrays and create the arguments.
emitOffloadingArrays(CGF, CombinedInfo, Info);
-bool HasDependClauses = D.hasClausesOfKind();
-emitOffloadingArraysArgument(CGF, Info.BasePointersArray,
- Info.PointersArray, Info.SizesArray,
- Info.MapTypesArray, Info.MappersArray, Info,
- {/*ForEndTask=*/false, HasDependClauses});
+emitOffloadingArraysArgument(
+CGF, Info.BasePointersArray, Info.PointersArray, Info.SizesArray,
+Info.MapTypesArray, Info.MappersArray, Info, {/*ForEndTask=*/false});
InputInfo.NumberOfTargetItems = Info.NumberOfPtrs;
InputInfo.BasePointersArray =
Address(Info.BasePointersArray, CGM.getPointerAlign());
@@ -10319,8 +10315,7 @@
llvm::Value *MappersArrayArg = nullptr;
emitOffloadingArraysArgument(CGF, BasePointersArrayArg, PointersArrayArg,
SizesArrayArg, MapTypesArrayArg,
- MappersArrayArg, Info,
- {/*ForEndCall=*/true, /*IsTask=*/false});
+ MappersArrayArg, Info, {/*ForEndCall=*/true});
// Emit device ID if any.
llvm::Value *DeviceID = nullptr;
@@ -10519,10 +10514,9 @@
// Fill up the arrays and create the arguments.
emitOffloadingArrays(CGF, CombinedInfo, Info);
bool HasDependClauses = D.hasClausesOfKind();
-
[PATCH] D90101: [Clang][OpenMP][WIP] Avoid unnecessary privatization of mapper array when there is no user defined mapper
tianshilei1992 created this revision.
Herald added subscribers: cfe-commits, guansong, yaxunl.
Herald added a project: clang.
tianshilei1992 requested review of this revision.
Herald added a reviewer: jdoerfert.
Herald added a subscriber: sstefan1.
In current implementation, if it requires an outer task, the mapper array will
be privatized no matter whether it has mapper. In fact, when there is no
mapper, the mapper array only contains number of nullptr. In the libomptarget,
the use of mapper array is `if (mappers_array && mappers_array[i])`, which
means we can directly set mapper array to nullptr if there is no mapper. This
can avoid unnecessary data copy.
In this patch, the data privatization will not be emitted if the mapper array
is nullptr. When it comes to the emit of task body, the nullptr will be used
directly.
Repository:
rG LLVM Github Monorepo
https://reviews.llvm.org/D90101
Files:
clang/lib/CodeGen/CGOpenMPRuntime.cpp
clang/lib/CodeGen/CGStmtOpenMP.cpp
Index: clang/lib/CodeGen/CGStmtOpenMP.cpp
===
--- clang/lib/CodeGen/CGStmtOpenMP.cpp
+++ clang/lib/CodeGen/CGStmtOpenMP.cpp
@@ -4210,16 +4210,21 @@
/*IndexTypeQuals=*/0);
SVD = createImplicitFirstprivateForType(getContext(), Data, SizesType, CD,
S.getBeginLoc());
-MVD = createImplicitFirstprivateForType(
-getContext(), Data, BaseAndPointerAndMapperType, CD, S.getBeginLoc());
TargetScope.addPrivate(
BPVD, []() { return InputInfo.BasePointersArray; });
TargetScope.addPrivate(PVD,
[]() { return InputInfo.PointersArray; });
TargetScope.addPrivate(SVD,
[]() { return InputInfo.SizesArray; });
-TargetScope.addPrivate(MVD,
- []() { return InputInfo.MappersArray; });
+// If there is no user-defined mapper, the mapper array will be nullptr. In
+// this case, we don't need to privatize it.
+if (!dyn_cast_or_null(
+InputInfo.MappersArray.getPointer())) {
+ MVD = createImplicitFirstprivateForType(
+ getContext(), Data, BaseAndPointerAndMapperType, CD, S.getBeginLoc());
+ TargetScope.addPrivate(MVD,
+ []() { return InputInfo.MappersArray; });
+}
}
(void)TargetScope.Privatize();
// Build list of dependences.
@@ -4269,8 +4274,10 @@
CGF.GetAddrOfLocalVar(PVD), /*Index=*/0);
InputInfo.SizesArray = CGF.Builder.CreateConstArrayGEP(
CGF.GetAddrOfLocalVar(SVD), /*Index=*/0);
- InputInfo.MappersArray = CGF.Builder.CreateConstArrayGEP(
- CGF.GetAddrOfLocalVar(MVD), /*Index=*/0);
+ // If MVD is nullptr, the mapper array is not privatized
+ if (!MVD)
+InputInfo.MappersArray = CGF.Builder.CreateConstArrayGEP(
+CGF.GetAddrOfLocalVar(MVD), /*Index=*/0);
}
Action.Enter(CGF);
Index: clang/lib/CodeGen/CGOpenMPRuntime.cpp
===
--- clang/lib/CodeGen/CGOpenMPRuntime.cpp
+++ clang/lib/CodeGen/CGOpenMPRuntime.cpp
@@ -8870,10 +8870,8 @@
/// Additional arguments for emitOffloadingArraysArgument function.
struct ArgumentsOptions {
bool ForEndCall = false;
- bool IsTask = false;
ArgumentsOptions() = default;
- ArgumentsOptions(bool ForEndCall, bool IsTask)
- : ForEndCall(ForEndCall), IsTask(IsTask) {}
+ ArgumentsOptions(bool ForEndCall) : ForEndCall(ForEndCall) {}
};
} // namespace
@@ -8909,9 +8907,9 @@
: Info.MapTypesArray,
/*Idx0=*/0,
/*Idx1=*/0);
-// Always emit the mapper array address in case of a target task for
-// privatization.
-if (!Options.IsTask && !Info.HasMapper)
+// If there is no user-defined mapper, set the mapper array to nullptr to
+// avoid an unnecessary data privatization
+if (!Info.HasMapper)
MappersArrayArg = llvm::ConstantPointerNull::get(CGM.VoidPtrPtrTy);
else
MappersArrayArg =
@@ -9664,11 +9662,9 @@
TargetDataInfo Info;
// Fill up the arrays and create the arguments.
emitOffloadingArrays(CGF, CombinedInfo, Info);
-bool HasDependClauses = D.hasClausesOfKind();
-emitOffloadingArraysArgument(CGF, Info.BasePointersArray,
- Info.PointersArray, Info.SizesArray,
- Info.MapTypesArray, Info.MappersArray, Info,
- {/*ForEndTask=*/false, HasDependClauses});
+emitOffloadingArraysArgument(
+CGF, Info.BasePointersArray, Info.PointersArray, Info.SizesArray,
+Info.MapTypesArray, Info.MappersArray, Info, {/*ForEndTask=*/false});
InputInfo.NumberOfTargetItems = Info.NumberOfPtrs;
InputInfo.BasePointersArray =
Address(Info.BasePointersArray, CGM.getPointerAlign());
@@ -10319,8 +10315,7 @@
