-
Notifications
You must be signed in to change notification settings - Fork 14.3k
[OpenMP] codegen support for masked combined construct parallel masked taskloop #121741
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
Conversation
@llvm/pr-subscribers-clang-codegen Author: CHANDRA GHALE (chandraghale) ChangesAdded codegen support for combined masked constructs Parallel masked taskloop. Full diff: https://github.com/llvm/llvm-project/pull/121741.diff 4 Files Affected:
diff --git a/clang/lib/CodeGen/CGStmt.cpp b/clang/lib/CodeGen/CGStmt.cpp
index 3974739d2abb47..06c434992ccba9 100644
--- a/clang/lib/CodeGen/CGStmt.cpp
+++ b/clang/lib/CodeGen/CGStmt.cpp
@@ -345,7 +345,8 @@ void CodeGenFunction::EmitStmt(const Stmt *S, ArrayRef<const Attr *> Attrs) {
cast<OMPParallelMasterTaskLoopDirective>(*S));
break;
case Stmt::OMPParallelMaskedTaskLoopDirectiveClass:
- llvm_unreachable("parallel masked taskloop directive not supported yet.");
+ EmitOMPParallelMaskedTaskLoopDirective(
+ cast<OMPParallelMaskedTaskLoopDirective>(*S));
break;
case Stmt::OMPParallelMasterTaskLoopSimdDirectiveClass:
EmitOMPParallelMasterTaskLoopSimdDirective(
diff --git a/clang/lib/CodeGen/CGStmtOpenMP.cpp b/clang/lib/CodeGen/CGStmtOpenMP.cpp
index 6cb37b20b7aeee..e45a5ea60e13cd 100644
--- a/clang/lib/CodeGen/CGStmtOpenMP.cpp
+++ b/clang/lib/CodeGen/CGStmtOpenMP.cpp
@@ -8024,6 +8024,24 @@ void CodeGenFunction::EmitOMPParallelMasterTaskLoopDirective(
emitEmptyBoundParameters);
}
+void CodeGenFunction::EmitOMPParallelMaskedTaskLoopDirective(
+ const OMPParallelMaskedTaskLoopDirective &S) {
+ auto &&CodeGen = [this, &S](CodeGenFunction &CGF, PrePostActionTy &Action) {
+ auto &&TaskLoopCodeGen = [&S](CodeGenFunction &CGF,
+ PrePostActionTy &Action) {
+ Action.Enter(CGF);
+ CGF.EmitOMPTaskLoopBasedDirective(S);
+ };
+ OMPLexicalScope Scope(CGF, S, OMPD_parallel, /*EmitPreInitStmt=*/false);
+ CGM.getOpenMPRuntime().emitMaskedRegion(CGF, TaskLoopCodeGen,
+ S.getBeginLoc());
+ };
+ auto LPCRegion =
+ CGOpenMPRuntime::LastprivateConditionalRAII::disable(*this, S);
+ emitCommonOMPParallelDirective(*this, S, OMPD_masked_taskloop, CodeGen,
+ emitEmptyBoundParameters);
+}
+
void CodeGenFunction::EmitOMPParallelMasterTaskLoopSimdDirective(
const OMPParallelMasterTaskLoopSimdDirective &S) {
auto &&CodeGen = [this, &S](CodeGenFunction &CGF, PrePostActionTy &Action) {
diff --git a/clang/lib/CodeGen/CodeGenFunction.h b/clang/lib/CodeGen/CodeGenFunction.h
index 1a5c42f8f974d0..1e8beca5513f05 100644
--- a/clang/lib/CodeGen/CodeGenFunction.h
+++ b/clang/lib/CodeGen/CodeGenFunction.h
@@ -3867,6 +3867,8 @@ class CodeGenFunction : public CodeGenTypeCache {
EmitOMPMasterTaskLoopSimdDirective(const OMPMasterTaskLoopSimdDirective &S);
void EmitOMPParallelMasterTaskLoopDirective(
const OMPParallelMasterTaskLoopDirective &S);
+ void EmitOMPParallelMaskedTaskLoopDirective(
+ const OMPParallelMaskedTaskLoopDirective &S);
void EmitOMPParallelMasterTaskLoopSimdDirective(
const OMPParallelMasterTaskLoopSimdDirective &S);
void EmitOMPDistributeDirective(const OMPDistributeDirective &S);
diff --git a/clang/test/OpenMP/parallel_masked_taskloop_codegen.c b/clang/test/OpenMP/parallel_masked_taskloop_codegen.c
new file mode 100644
index 00000000000000..ed89ef92bb768d
--- /dev/null
+++ b/clang/test/OpenMP/parallel_masked_taskloop_codegen.c
@@ -0,0 +1,62 @@
+// NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py UTC_ARGS: --include-generated-funcs --prefix-filecheck-ir-name _ --version 5
+// RUN: %clang_cc1 -triple x86_64-unknown-unknown -fopenmp -fopenmp-version=52 -x c -emit-llvm %s -o - | FileCheck %s
+// expected-no-diagnostics
+#define N 100
+void parallel_masked_taskloop(){
+ #pragma omp parallel masked taskloop
+ for( int i = 0; i < N; i++)
+ ;
+
+}
+
+int main()
+{
+ parallel_masked_taskloop();
+}
+// CHECK-LABEL: define dso_local void @parallel_masked_taskloop(
+// CHECK-SAME: ) #[[ATTR0:[0-9]+]] {
+// CHECK-NEXT: [[ENTRY:.*:]]
+// CHECK-NEXT: call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB1:[0-9]+]], i32 0, ptr @parallel_masked_taskloop.omp_outlined)
+// CHECK-NEXT: ret void
+//
+//
+// CHECK-LABEL: define internal void @parallel_masked_taskloop.omp_outlined(
+// CHECK-SAME: ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1:[0-9]+]] {
+// CHECK-NEXT: [[ENTRY:.*:]]
+// CHECK-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
+// CHECK-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
+// CHECK-NEXT: [[AGG_CAPTURED:%.*]] = alloca [[STRUCT_ANON:%.*]], align 1
+// CHECK-NEXT: [[TMP:%.*]] = alloca i32, align 4
+// CHECK-NEXT: store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
+// CHECK-NEXT: store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
+// CHECK-NEXT: [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
+// CHECK-NEXT: [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
+// CHECK-NEXT: [[TMP2:%.*]] = call i32 @__kmpc_masked(ptr @[[GLOB1]], i32 [[TMP1]], i32 0)
+// CHECK-NEXT: [[TMP3:%.*]] = icmp ne i32 [[TMP2]], 0
+// CHECK-NEXT: br i1 [[TMP3]], label %[[OMP_IF_THEN:.*]], label %[[OMP_IF_END:.*]]
+// CHECK: [[OMP_IF_THEN]]:
+// CHECK-NEXT: call void @__kmpc_taskgroup(ptr @[[GLOB1]], i32 [[TMP1]])
+// CHECK-NEXT: [[TMP4:%.*]] = call ptr @__kmpc_omp_task_alloc(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i64 80, i64 0, ptr @.omp_task_entry.)
+// CHECK-NEXT: [[TMP5:%.*]] = getelementptr inbounds nuw [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], ptr [[TMP4]], i32 0, i32 0
+// CHECK-NEXT: [[TMP6:%.*]] = getelementptr inbounds nuw [[STRUCT_KMP_TASK_T:%.*]], ptr [[TMP5]], i32 0, i32 5
+// CHECK-NEXT: store i64 0, ptr [[TMP6]], align 8
+// CHECK-NEXT: [[TMP7:%.*]] = getelementptr inbounds nuw [[STRUCT_KMP_TASK_T]], ptr [[TMP5]], i32 0, i32 6
+// CHECK-NEXT: store i64 99, ptr [[TMP7]], align 8
+// CHECK-NEXT: [[TMP8:%.*]] = getelementptr inbounds nuw [[STRUCT_KMP_TASK_T]], ptr [[TMP5]], i32 0, i32 7
+// CHECK-NEXT: store i64 1, ptr [[TMP8]], align 8
+// CHECK-NEXT: [[TMP9:%.*]] = getelementptr inbounds nuw [[STRUCT_KMP_TASK_T]], ptr [[TMP5]], i32 0, i32 9
+// CHECK-NEXT: call void @llvm.memset.p0.i64(ptr align 8 [[TMP9]], i8 0, i64 8, i1 false)
+// CHECK-NEXT: [[TMP10:%.*]] = load i64, ptr [[TMP8]], align 8
+// CHECK-NEXT: call void @__kmpc_taskloop(ptr @[[GLOB1]], i32 [[TMP1]], ptr [[TMP4]], i32 1, ptr [[TMP6]], ptr [[TMP7]], i64 [[TMP10]], i32 1, i32 0, i64 0, ptr null)
+// CHECK-NEXT: call void @__kmpc_end_taskgroup(ptr @[[GLOB1]], i32 [[TMP1]])
+// CHECK-NEXT: call void @__kmpc_end_masked(ptr @[[GLOB1]], i32 [[TMP1]])
+// CHECK-NEXT: br label %[[OMP_IF_END]]
+// CHECK: [[OMP_IF_END]]:
+// CHECK-NEXT: ret void
+//
+// CHECK-LABEL: define dso_local i32 @main(
+// CHECK-SAME: ) #[[ATTR0]] {
+// CHECK-NEXT: [[ENTRY:.*:]]
+// CHECK-NEXT: call void @parallel_masked_taskloop()
+// CHECK-NEXT: ret i32 0
+
|
@llvm/pr-subscribers-clang Author: CHANDRA GHALE (chandraghale) ChangesAdded codegen support for combined masked constructs Parallel masked taskloop. Full diff: https://github.com/llvm/llvm-project/pull/121741.diff 4 Files Affected:
diff --git a/clang/lib/CodeGen/CGStmt.cpp b/clang/lib/CodeGen/CGStmt.cpp
index 3974739d2abb47..06c434992ccba9 100644
--- a/clang/lib/CodeGen/CGStmt.cpp
+++ b/clang/lib/CodeGen/CGStmt.cpp
@@ -345,7 +345,8 @@ void CodeGenFunction::EmitStmt(const Stmt *S, ArrayRef<const Attr *> Attrs) {
cast<OMPParallelMasterTaskLoopDirective>(*S));
break;
case Stmt::OMPParallelMaskedTaskLoopDirectiveClass:
- llvm_unreachable("parallel masked taskloop directive not supported yet.");
+ EmitOMPParallelMaskedTaskLoopDirective(
+ cast<OMPParallelMaskedTaskLoopDirective>(*S));
break;
case Stmt::OMPParallelMasterTaskLoopSimdDirectiveClass:
EmitOMPParallelMasterTaskLoopSimdDirective(
diff --git a/clang/lib/CodeGen/CGStmtOpenMP.cpp b/clang/lib/CodeGen/CGStmtOpenMP.cpp
index 6cb37b20b7aeee..e45a5ea60e13cd 100644
--- a/clang/lib/CodeGen/CGStmtOpenMP.cpp
+++ b/clang/lib/CodeGen/CGStmtOpenMP.cpp
@@ -8024,6 +8024,24 @@ void CodeGenFunction::EmitOMPParallelMasterTaskLoopDirective(
emitEmptyBoundParameters);
}
+void CodeGenFunction::EmitOMPParallelMaskedTaskLoopDirective(
+ const OMPParallelMaskedTaskLoopDirective &S) {
+ auto &&CodeGen = [this, &S](CodeGenFunction &CGF, PrePostActionTy &Action) {
+ auto &&TaskLoopCodeGen = [&S](CodeGenFunction &CGF,
+ PrePostActionTy &Action) {
+ Action.Enter(CGF);
+ CGF.EmitOMPTaskLoopBasedDirective(S);
+ };
+ OMPLexicalScope Scope(CGF, S, OMPD_parallel, /*EmitPreInitStmt=*/false);
+ CGM.getOpenMPRuntime().emitMaskedRegion(CGF, TaskLoopCodeGen,
+ S.getBeginLoc());
+ };
+ auto LPCRegion =
+ CGOpenMPRuntime::LastprivateConditionalRAII::disable(*this, S);
+ emitCommonOMPParallelDirective(*this, S, OMPD_masked_taskloop, CodeGen,
+ emitEmptyBoundParameters);
+}
+
void CodeGenFunction::EmitOMPParallelMasterTaskLoopSimdDirective(
const OMPParallelMasterTaskLoopSimdDirective &S) {
auto &&CodeGen = [this, &S](CodeGenFunction &CGF, PrePostActionTy &Action) {
diff --git a/clang/lib/CodeGen/CodeGenFunction.h b/clang/lib/CodeGen/CodeGenFunction.h
index 1a5c42f8f974d0..1e8beca5513f05 100644
--- a/clang/lib/CodeGen/CodeGenFunction.h
+++ b/clang/lib/CodeGen/CodeGenFunction.h
@@ -3867,6 +3867,8 @@ class CodeGenFunction : public CodeGenTypeCache {
EmitOMPMasterTaskLoopSimdDirective(const OMPMasterTaskLoopSimdDirective &S);
void EmitOMPParallelMasterTaskLoopDirective(
const OMPParallelMasterTaskLoopDirective &S);
+ void EmitOMPParallelMaskedTaskLoopDirective(
+ const OMPParallelMaskedTaskLoopDirective &S);
void EmitOMPParallelMasterTaskLoopSimdDirective(
const OMPParallelMasterTaskLoopSimdDirective &S);
void EmitOMPDistributeDirective(const OMPDistributeDirective &S);
diff --git a/clang/test/OpenMP/parallel_masked_taskloop_codegen.c b/clang/test/OpenMP/parallel_masked_taskloop_codegen.c
new file mode 100644
index 00000000000000..ed89ef92bb768d
--- /dev/null
+++ b/clang/test/OpenMP/parallel_masked_taskloop_codegen.c
@@ -0,0 +1,62 @@
+// NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py UTC_ARGS: --include-generated-funcs --prefix-filecheck-ir-name _ --version 5
+// RUN: %clang_cc1 -triple x86_64-unknown-unknown -fopenmp -fopenmp-version=52 -x c -emit-llvm %s -o - | FileCheck %s
+// expected-no-diagnostics
+#define N 100
+void parallel_masked_taskloop(){
+ #pragma omp parallel masked taskloop
+ for( int i = 0; i < N; i++)
+ ;
+
+}
+
+int main()
+{
+ parallel_masked_taskloop();
+}
+// CHECK-LABEL: define dso_local void @parallel_masked_taskloop(
+// CHECK-SAME: ) #[[ATTR0:[0-9]+]] {
+// CHECK-NEXT: [[ENTRY:.*:]]
+// CHECK-NEXT: call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB1:[0-9]+]], i32 0, ptr @parallel_masked_taskloop.omp_outlined)
+// CHECK-NEXT: ret void
+//
+//
+// CHECK-LABEL: define internal void @parallel_masked_taskloop.omp_outlined(
+// CHECK-SAME: ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1:[0-9]+]] {
+// CHECK-NEXT: [[ENTRY:.*:]]
+// CHECK-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
+// CHECK-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
+// CHECK-NEXT: [[AGG_CAPTURED:%.*]] = alloca [[STRUCT_ANON:%.*]], align 1
+// CHECK-NEXT: [[TMP:%.*]] = alloca i32, align 4
+// CHECK-NEXT: store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
+// CHECK-NEXT: store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
+// CHECK-NEXT: [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
+// CHECK-NEXT: [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
+// CHECK-NEXT: [[TMP2:%.*]] = call i32 @__kmpc_masked(ptr @[[GLOB1]], i32 [[TMP1]], i32 0)
+// CHECK-NEXT: [[TMP3:%.*]] = icmp ne i32 [[TMP2]], 0
+// CHECK-NEXT: br i1 [[TMP3]], label %[[OMP_IF_THEN:.*]], label %[[OMP_IF_END:.*]]
+// CHECK: [[OMP_IF_THEN]]:
+// CHECK-NEXT: call void @__kmpc_taskgroup(ptr @[[GLOB1]], i32 [[TMP1]])
+// CHECK-NEXT: [[TMP4:%.*]] = call ptr @__kmpc_omp_task_alloc(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i64 80, i64 0, ptr @.omp_task_entry.)
+// CHECK-NEXT: [[TMP5:%.*]] = getelementptr inbounds nuw [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], ptr [[TMP4]], i32 0, i32 0
+// CHECK-NEXT: [[TMP6:%.*]] = getelementptr inbounds nuw [[STRUCT_KMP_TASK_T:%.*]], ptr [[TMP5]], i32 0, i32 5
+// CHECK-NEXT: store i64 0, ptr [[TMP6]], align 8
+// CHECK-NEXT: [[TMP7:%.*]] = getelementptr inbounds nuw [[STRUCT_KMP_TASK_T]], ptr [[TMP5]], i32 0, i32 6
+// CHECK-NEXT: store i64 99, ptr [[TMP7]], align 8
+// CHECK-NEXT: [[TMP8:%.*]] = getelementptr inbounds nuw [[STRUCT_KMP_TASK_T]], ptr [[TMP5]], i32 0, i32 7
+// CHECK-NEXT: store i64 1, ptr [[TMP8]], align 8
+// CHECK-NEXT: [[TMP9:%.*]] = getelementptr inbounds nuw [[STRUCT_KMP_TASK_T]], ptr [[TMP5]], i32 0, i32 9
+// CHECK-NEXT: call void @llvm.memset.p0.i64(ptr align 8 [[TMP9]], i8 0, i64 8, i1 false)
+// CHECK-NEXT: [[TMP10:%.*]] = load i64, ptr [[TMP8]], align 8
+// CHECK-NEXT: call void @__kmpc_taskloop(ptr @[[GLOB1]], i32 [[TMP1]], ptr [[TMP4]], i32 1, ptr [[TMP6]], ptr [[TMP7]], i64 [[TMP10]], i32 1, i32 0, i64 0, ptr null)
+// CHECK-NEXT: call void @__kmpc_end_taskgroup(ptr @[[GLOB1]], i32 [[TMP1]])
+// CHECK-NEXT: call void @__kmpc_end_masked(ptr @[[GLOB1]], i32 [[TMP1]])
+// CHECK-NEXT: br label %[[OMP_IF_END]]
+// CHECK: [[OMP_IF_END]]:
+// CHECK-NEXT: ret void
+//
+// CHECK-LABEL: define dso_local i32 @main(
+// CHECK-SAME: ) #[[ATTR0]] {
+// CHECK-NEXT: [[ENTRY:.*:]]
+// CHECK-NEXT: call void @parallel_masked_taskloop()
+// CHECK-NEXT: ret i32 0
+
|
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.
Looks like the old master directive implementation.
Update OpenMPSupport.rst |
@alexey-bataev Updated OpenMPSupport.rst . Updated the doc for all the related split-ed PRs ( PR-121746 , 121914 , 121916 ) for combined masked construct in this PR only to avoid merge conflict. |
Also update ReleaseNotes.rst here and in other patches |
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.
LG
@alexey-bataev release notes updated. Updated for other patches as well in this PR. |
Added codegen support for combined masked constructs Parallel masked taskloop.
Added implementation for EmitOMPParallelMaskedTaskLoopDirective.