[clang][OpenMP] Add IRBuilder support for taskgroup

[shraiysh]

Summary

This patch makes use of OMPIRBuilder support for codegen of taskgroup
construct in clang.

Depends on D128203

[Meinersbur]

clang/lib/CodeGen/CGStmtOpenMP.cpp

5215

Instead of S.clauses().size() == 0, could we make it more explicit and extendable by adding a function like isSupportedbyOpenMPIRBuilder(S)?

[shraiysh]

Moved checking to isSupportedByOpenMPIRBuilder.

clang/lib/CodeGen/CGStmtOpenMP.cpp

5215

Done.

[Meinersbur]

LGTM

clang/lib/CodeGen/CGStmtOpenMP.cpp

5207–5209

simplification

[shraiysh]

Rebase with main

clang/lib/CodeGen/CGStmtOpenMP.cpp

 
 void CodeGenFunction::EmitOMPTaskwaitDirective(const OMPTaskwaitDirective &S) {
   OMPTaskDataTy Data;
   // Build list of dependences
   buildDependences(S, Data);
   CGM.getOpenMPRuntime().emitTaskwaitCall(*this, S.getBeginLoc(), Data);
 }
 
+bool isSupportedByOpenMPIRBuilder(const OMPTaskgroupDirective &T) {
+  if (T.clauses().size() > 0)
+    return false;
+  return true;
+}
+
 void CodeGenFunction::EmitOMPTaskgroupDirective(
     const OMPTaskgroupDirective &S) {
+  OMPLexicalScope Scope(*this, S, OMPD_unknown);
+  if (CGM.getLangOpts().OpenMPIRBuilder && isSupportedByOpenMPIRBuilder(S)) {
+    llvm::OpenMPIRBuilder &OMPBuilder = CGM.getOpenMPRuntime().getOMPBuilder();
+    using InsertPointTy = llvm::OpenMPIRBuilder::InsertPointTy;
+    InsertPointTy AllocaIP(AllocaInsertPt->getParent(),
+                           AllocaInsertPt->getIterator());
+
+    auto BodyGenCB = [&, this](InsertPointTy AllocaIP,
+                               InsertPointTy CodeGenIP) {
+      Builder.restoreIP(CodeGenIP);
+      EmitStmt(S.getInnermostCapturedStmt()->getCapturedStmt());
+    };
+    CodeGenFunction::CGCapturedStmtInfo CapStmtInfo;
+    if (!CapturedStmtInfo)
+      CapturedStmtInfo = &CapStmtInfo;
+    Builder.restoreIP(OMPBuilder.createTaskgroup(Builder, AllocaIP, BodyGenCB));
+    return;
+  }
   auto &&CodeGen = [&S](CodeGenFunction &CGF, PrePostActionTy &Action) {
     Action.Enter(CGF);
     if (const Expr *E = S.getReductionRef()) {
       SmallVector<const Expr *, 4> LHSs;
       SmallVector<const Expr *, 4> RHSs;
       OMPTaskDataTy Data;
       for (const auto *C : S.getClausesOfKind<OMPTaskReductionClause>()) {
         Data.ReductionVars.append(C->varlist_begin(), C->varlist_end());
if (const Expr *E = S.getReductionRef()) {
                                                            LHSs, RHSs, Data);
       const auto *VD = cast<VarDecl>(cast<DeclRefExpr>(E)->getDecl());
       CGF.EmitVarDecl(*VD);
       CGF.EmitStoreOfScalar(ReductionDesc, CGF.GetAddrOfLocalVar(VD),
                             /*Volatile=*/false, E->getType());
     }
     CGF.EmitStmt(S.getInnermostCapturedStmt()->getCapturedStmt());
   };
-  OMPLexicalScope Scope(*this, S, OMPD_unknown);
   CGM.getOpenMPRuntime().emitTaskgroupRegion(*this, CodeGen, S.getBeginLoc());
 }
 
 void CodeGenFunction::EmitOMPFlushDirective(const OMPFlushDirective &S) {
   llvm::AtomicOrdering AO = S.getSingleClause<OMPFlushClause>()
                                 ? llvm::AtomicOrdering::NotAtomic
                                 : llvm::AtomicOrdering::AcquireRelease;
   CGM.getOpenMPRuntime().emitFlush(

clang/test/OpenMP/taskgroup_codegen.cpp

 // NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py UTC_ARGS: --function-signature --include-generated-funcs --replace-value-regex "__omp_offloading_[0-9a-z]+_[0-9a-z]+" "reduction_size[.].+[.]" "pl_cond[.].+[.|,]" --prefix-filecheck-ir-name _
 // RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp -x c++ -triple x86_64-unknown-unknown -emit-llvm %s -fexceptions -fcxx-exceptions -o - | FileCheck %s --check-prefix=CHECK1
 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp -x c++ -std=c++11 -triple x86_64-unknown-unknown -fexceptions -fcxx-exceptions -emit-pch -o %t %s
 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp -x c++ -triple x86_64-unknown-unknown -fexceptions -fcxx-exceptions -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK1
 // RUN: %clang_cc1 -no-opaque-pointers -verify -triple x86_64-apple-darwin10 -fopenmp -fexceptions -fcxx-exceptions -debug-info-kind=line-tables-only -x c++ -emit-llvm %s -o - | FileCheck %s --check-prefix=DEBUG1
 
 // RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp-simd -x c++ -emit-llvm %s -fexceptions -fcxx-exceptions -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp-simd -x c++ -std=c++11 -triple x86_64-unknown-unknown -fexceptions -fcxx-exceptions -emit-pch -o %t %s
 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp-simd -x c++ -triple x86_64-unknown-unknown -fexceptions -fcxx-exceptions -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
 // RUN: %clang_cc1 -no-opaque-pointers -verify -triple x86_64-apple-darwin10 -fopenmp-simd -fexceptions -fcxx-exceptions -debug-info-kind=line-tables-only -x c++ -emit-llvm %s -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
+
+// RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp -fopenmp-enable-irbuilder -x c++ -triple x86_64-unknown-unknown -emit-llvm %s -fexceptions -fcxx-exceptions -o - | FileCheck %s --check-prefix=CHECK2
+
 // expected-no-diagnostics
 #ifndef HEADER
 #define HEADER
 
 void foo() { extern void mayThrow(); mayThrow(); }
 
 int main() {
   char a;

 // DEBUG1-NEXT:    ret void, !dbg [[DBG26:![0-9]+]]
 // DEBUG1:       terminate.lpad:
 // DEBUG1-NEXT:    [[TMP2:%.*]] = landingpad { i8*, i32 }
 // DEBUG1-NEXT:    catch i8* null, !dbg [[DBG25]]
 // DEBUG1-NEXT:    [[TMP3:%.*]] = extractvalue { i8*, i32 } [[TMP2]], 0, !dbg [[DBG25]]
 // DEBUG1-NEXT:    call void @__clang_call_terminate(i8* [[TMP3]]) #[[ATTR8]], !dbg [[DBG25]]
 // DEBUG1-NEXT:    unreachable, !dbg [[DBG25]]
 //
+
+// CHECK2-LABEL: define dso_local void @_Z3foov() #{{.+}} {
+// CHECK2:       entry:
+// CHECK2-NEXT:    call void @_Z8mayThrowv()
+// CHECK2-NEXT:    ret void
+
+// CHECK2-LABEL: define dso_local noundef i32 @main() #{{.+}} {
+// CHECK2:       entry:
+// CHECK2-NEXT:    %[[RETVAL:.+]] = alloca i32, align 4
+// CHECK2-NEXT:    %[[A:.+]] = alloca i8, align 1
+// CHECK2-NEXT:    store i32 0, i32* %[[RETVAL]], align 4
+// CHECK2-NEXT:    %[[OMP_THREAD_NUM:.+]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @{{.+}})
+// CHECK2-NEXT:    call void @__kmpc_taskgroup(%struct.ident_t* @{{.+}}, i32 %[[OMP_THREAD_NUM]])
+// CHECK2-NEXT:    store i8 2, i8* %[[A]], align 1
+// CHECK2-NEXT:    br label %taskgroup.exit
+
+// CHECK2:       taskgroup.exit:
+// CHECK2-NEXT:    call void @__kmpc_end_taskgroup(%struct.ident_t* @{{.+}}, i32 %[[OMP_THREAD_NUM]])
+// CHECK2-NEXT:    %[[OMP_THREAD_NUM2:.+]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @{{.+}})
+// CHECK2-NEXT:    call void @__kmpc_taskgroup(%struct.ident_t* @{{.+}}, i32 %[[OMP_THREAD_NUM2]])
+// CHECK2-NEXT:    call void @_Z3foov()
+// CHECK2-NEXT:    br label %taskgroup.exit2
+
+// CHECK2:       taskgroup.exit2:
+// CHECK2-NEXT:    call void @__kmpc_end_taskgroup(%struct.ident_t* @{{.+}}, i32 %[[OMP_THREAD_NUM2]])
+// CHECK2-NEXT:    %[[TMP:.+]] = load i8, i8* %[[A]], align 1
+// CHECK2-NEXT:    %[[CONV:.+]] = sext i8 %[[TMP]] to i32
+// CHECK2-NEXT:    ret i32 %[[CONV]]
+
+// CHECK2-LABEL: define dso_local void @_Z18parallel_taskgroupv() #{{.+}} {
+// CHECK2:       entry:
+// CHECK2-NEXT:    %[[OMP_THREAD_NUM:.+]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @{{.+}})
+// CHECK2-NEXT:    br label %omp_parallel
+
+// CHECK2:       omp_parallel:
+// CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @{{.+}}, i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @_Z18parallel_taskgroupv..omp_par to void (i32*, i32*, ...)*))
+// CHECK2-NEXT:    br label %omp.par.outlined.exit
+
+// CHECK2:       omp.par.outlined.exit:
+// CHECK2-NEXT:    br label %omp.par.exit.split
+
+// CHECK2:       omp.par.exit.split:
+// CHECK2-NEXT:    ret void
+
+// CHECK2-LABEL: define internal void @_Z18parallel_taskgroupv..omp_par(i32* noalias %{{.+}}, i32* noalias %{{.+}}) #{{.+}} {
+// CHECK2:       omp.par.entry:
+// CHECK2:         br label %omp.par.region
+
+// CHECK2:       omp.par.region:
+// CHECK2-NEXT:    %[[OMP_THREAD_NUM:.+]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @{{.+}})
+// CHECK2-NEXT:    call void @__kmpc_taskgroup(%struct.ident_t* @{{.+}}, i32 %[[OMP_THREAD_NUM]])
+// CHECK2-NEXT:    call void @_Z3foov()
+// CHECK2-NEXT:    br label %taskgroup.exit
+
+// CHECK2:       taskgroup.exit:
+// CHECK2-NEXT:    call void @__kmpc_end_taskgroup(%struct.ident_t* @{{.+}}, i32 %[[OMP_THREAD_NUM]])
+// CHECK2-NEXT:    br label %omp.par.region.parallel.after
+
+// CHECK2:       omp.par.region.parallel.after:
+// CHECK2-NEXT:    br label %omp.par.pre_finalize
+
+// CHECK2:       omp.par.pre_finalize:
+// CHECK2-NEXT:    br label %omp.par.outlined.exit.exitStub
+
+// CHECK2:       omp.par.outlined.exit.exitStub:
+// CHECK2-NEXT:    ret void