[clang] 30f9a4f - [OpenMP] codegen support for masked combined construct parallel masked taskloop simd. (#121746)
via cfe-commits
cfe-commits at lists.llvm.org
Tue Jan 14 04:56:49 PST 2025
Author: CHANDRA GHALE
Date: 2025-01-14T18:26:46+05:30
New Revision: 30f9a4f75412850d603fde29b59d27dd9d31f380
URL: https://github.com/llvm/llvm-project/commit/30f9a4f75412850d603fde29b59d27dd9d31f380
DIFF: https://github.com/llvm/llvm-project/commit/30f9a4f75412850d603fde29b59d27dd9d31f380.diff
LOG: [OpenMP] codegen support for masked combined construct parallel masked taskloop simd. (#121746)
Added codegen support for combined masked constructs `Parallel masked
taskloop simd`.
Added implementation for `EmitOMPParallelMaskedTaskLoopSimdDirective`.
Co-authored-by: Chandra Ghale <ghale at pe31.hpc.amslabs.hpecorp.net>
Added:
clang/test/OpenMP/parallel_masked_taskloop_simd_codegen.c
Modified:
clang/lib/CodeGen/CGStmt.cpp
clang/lib/CodeGen/CGStmtOpenMP.cpp
clang/lib/CodeGen/CodeGenFunction.h
Removed:
################################################################################
diff --git a/clang/lib/CodeGen/CGStmt.cpp b/clang/lib/CodeGen/CGStmt.cpp
index 4ba8ee1ca17d4e..10d44e9c90b27e 100644
--- a/clang/lib/CodeGen/CGStmt.cpp
+++ b/clang/lib/CodeGen/CGStmt.cpp
@@ -355,8 +355,8 @@ void CodeGenFunction::EmitStmt(const Stmt *S, ArrayRef<const Attr *> Attrs) {
cast<OMPParallelMasterTaskLoopSimdDirective>(*S));
break;
case Stmt::OMPParallelMaskedTaskLoopSimdDirectiveClass:
- llvm_unreachable(
- "parallel masked taskloop simd directive not supported yet.");
+ EmitOMPParallelMaskedTaskLoopSimdDirective(
+ cast<OMPParallelMaskedTaskLoopSimdDirective>(*S));
break;
case Stmt::OMPDistributeDirectiveClass:
EmitOMPDistributeDirective(cast<OMPDistributeDirective>(*S));
diff --git a/clang/lib/CodeGen/CGStmtOpenMP.cpp b/clang/lib/CodeGen/CGStmtOpenMP.cpp
index 2b4ca65e169a6e..8e694b95dc7e73 100644
--- a/clang/lib/CodeGen/CGStmtOpenMP.cpp
+++ b/clang/lib/CodeGen/CGStmtOpenMP.cpp
@@ -8072,6 +8072,24 @@ void CodeGenFunction::EmitOMPParallelMasterTaskLoopSimdDirective(
emitEmptyBoundParameters);
}
+void CodeGenFunction::EmitOMPParallelMaskedTaskLoopSimdDirective(
+ const OMPParallelMaskedTaskLoopSimdDirective &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_simd, CodeGen,
+ emitEmptyBoundParameters);
+}
+
// Generate the instructions for '#pragma omp target update' directive.
void CodeGenFunction::EmitOMPTargetUpdateDirective(
const OMPTargetUpdateDirective &S) {
diff --git a/clang/lib/CodeGen/CodeGenFunction.h b/clang/lib/CodeGen/CodeGenFunction.h
index b115c15bf01a95..ba8ed040477cc9 100644
--- a/clang/lib/CodeGen/CodeGenFunction.h
+++ b/clang/lib/CodeGen/CodeGenFunction.h
@@ -3885,6 +3885,8 @@ class CodeGenFunction : public CodeGenTypeCache {
const OMPParallelMaskedTaskLoopDirective &S);
void EmitOMPParallelMasterTaskLoopSimdDirective(
const OMPParallelMasterTaskLoopSimdDirective &S);
+ void EmitOMPParallelMaskedTaskLoopSimdDirective(
+ const OMPParallelMaskedTaskLoopSimdDirective &S);
void EmitOMPDistributeDirective(const OMPDistributeDirective &S);
void EmitOMPDistributeParallelForDirective(
const OMPDistributeParallelForDirective &S);
diff --git a/clang/test/OpenMP/parallel_masked_taskloop_simd_codegen.c b/clang/test/OpenMP/parallel_masked_taskloop_simd_codegen.c
new file mode 100644
index 00000000000000..0b51b302f9fcda
--- /dev/null
+++ b/clang/test/OpenMP/parallel_masked_taskloop_simd_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_simd(){
+ #pragma omp parallel masked taskloop simd
+ for( int i = 0; i < N; i++)
+ ;
+
+}
+
+int main()
+{
+ parallel_masked_taskloop_simd();
+}
+// CHECK-LABEL: define dso_local void @parallel_masked_taskloop_simd(
+// 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_simd.omp_outlined)
+// CHECK-NEXT: ret void
+//
+//
+// CHECK-LABEL: define internal void @parallel_masked_taskloop_simd.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_simd()
+// CHECK-NEXT: ret i32 0
+
More information about the cfe-commits
mailing list