[clang] aedb30f - [OpenMP] codegen support for masked combined construct parallel masked taskloop (#121741)
via cfe-commits
cfe-commits at lists.llvm.org
Thu Jan 9 03:08:40 PST 2025
Author: CHANDRA GHALE
Date: 2025-01-09T16:38:36+05:30
New Revision: aedb30fdc7cf83290b0f0d4246cf94fd62be749b
URL: https://github.com/llvm/llvm-project/commit/aedb30fdc7cf83290b0f0d4246cf94fd62be749b
DIFF: https://github.com/llvm/llvm-project/commit/aedb30fdc7cf83290b0f0d4246cf94fd62be749b.diff
LOG: [OpenMP] codegen support for masked combined construct parallel masked taskloop (#121741)
Added codegen support for combined masked constructs Parallel masked
taskloop.
Added implementation for EmitOMPParallelMaskedTaskLoopDirective.
---------
Co-authored-by: Chandra Ghale <ghale at pe31.hpc.amslabs.hpecorp.net>
Added:
clang/test/OpenMP/parallel_masked_taskloop_codegen.c
Modified:
clang/docs/OpenMPSupport.rst
clang/docs/ReleaseNotes.rst
clang/lib/CodeGen/CGStmt.cpp
clang/lib/CodeGen/CGStmtOpenMP.cpp
clang/lib/CodeGen/CodeGenFunction.h
Removed:
################################################################################
diff --git a/clang/docs/OpenMPSupport.rst b/clang/docs/OpenMPSupport.rst
index 481362dba3f51d..6703e1f55d8f19 100644
--- a/clang/docs/OpenMPSupport.rst
+++ b/clang/docs/OpenMPSupport.rst
@@ -320,7 +320,9 @@ implementation.
+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
| misc | nothing directive | :good:`done` | D123286 |
+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| misc | masked construct and related combined constructs | :part:`worked on` | D99995, D100514 |
+| misc | masked construct and related combined constructs | :good:`done` | D99995, D100514, PR-121741(parallel_masked_taskloop) |
+| | | | PR-121746(parallel_masked_task_loop_simd),PR-121914(masked_taskloop) |
+| | | | PR-121916(masked_taskloop_simd) |
+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
| misc | default(firstprivate) & default(private) | :good:`done` | D75591 (firstprivate), D125912 (private) |
+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
diff --git a/clang/docs/ReleaseNotes.rst b/clang/docs/ReleaseNotes.rst
index ba522e310c9e44..0ea39ccfc62670 100644
--- a/clang/docs/ReleaseNotes.rst
+++ b/clang/docs/ReleaseNotes.rst
@@ -1303,6 +1303,8 @@ OpenMP Support
- Changed the OpenMP DeviceRTL to use 'generic' IR. The
``LIBOMPTARGET_DEVICE_ARCHITECTURES`` CMake argument is now unused and will
always build support for AMDGPU and NVPTX targets.
+- Added support for combined masked constructs 'omp parallel masked taskloop',
+ 'omp parallel masked taskloop simd','omp masked taskloop' and 'omp masked taskloop simd' directive.
Improvements
^^^^^^^^^^^^
diff --git a/clang/lib/CodeGen/CGStmt.cpp b/clang/lib/CodeGen/CGStmt.cpp
index e9a8500cc19933..c87ec899798aa2 100644
--- a/clang/lib/CodeGen/CGStmt.cpp
+++ b/clang/lib/CodeGen/CGStmt.cpp
@@ -346,7 +346,8 @@ 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 f36e09bae81044..4e7b0dcd31ecf9 100644
--- a/clang/lib/CodeGen/CGStmtOpenMP.cpp
+++ b/clang/lib/CodeGen/CGStmtOpenMP.cpp
@@ -8012,6 +8012,24 @@ 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 e2dc0b1e381684..457e1477bb2eed 100644
--- a/clang/lib/CodeGen/CodeGenFunction.h
+++ b/clang/lib/CodeGen/CodeGenFunction.h
@@ -3874,6 +3874,8 @@ 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
+
More information about the cfe-commits
mailing list