[OPENMP] Codegen for 'omp master' directive

[ABataev]

Summary

Patch adds 2 library functions to OpenMPRuntime class - int32 kmpc_master(ident_t *, int32 gtid) and void kmpc_end_master(ident_t *, int32 gtid);
For 'omp master' directive the next code is generated:

if (__kmpc_master(loc, gtid)) {
  <Associated structured block>;
  __kmpc_end_master(log, gtid);
}

[rjmccall]

Generally looks good. A few comments.

lib/CodeGen/CGOpenMPRuntime.cpp

685

Please add a comment here that it's illegal to branch in or out of the master block — which I assume you're enforcing in Sema somehow — so you can just call __kmpc_end_master on fallthrough rather than pushing a normal cleanup for it.

lib/CodeGen/CGOpenMPRuntime.h

101

Please leave a trailing comma on the last enumerator. Makes a minor improvement to diffs.

[ABataev]

Thanks for the review John! I really appreciate it!!!

lib/CodeGen/CGOpenMPRuntime.cpp

685

Yes, it is prohibited by the OpenMP spec and enforced in Sema. I'll add a comment.

lib/CodeGen/CGOpenMPRuntime.h

101

Ok

[ABataev]

Update after review

[rjmccall]

Looks good, thanks!

[ABataev]

Closed by commit rL223342 (authored by @ABataev).

lib/CodeGen/CGOpenMPRuntime.h

enum OpenMPRTLFunction {
     OMPRTL__kmpc_serialized_parallel,
     // Call to void __kmpc_end_serialized_parallel(ident_t *loc, kmp_int32
     // global_tid);
     OMPRTL__kmpc_end_serialized_parallel,
     // Call to void __kmpc_push_num_threads(ident_t *loc, kmp_int32 global_tid,
     // kmp_int32 num_threads);
     OMPRTL__kmpc_push_num_threads,
     // Call to void __kmpc_flush(ident_t *loc, ...);
-    OMPRTL__kmpc_flush
+    OMPRTL__kmpc_flush,
+    // Call to kmp_int32 __kmpc_master(ident_t *, kmp_int32 global_tid);
+    OMPRTL__kmpc_master,
+    // Call to void __kmpc_end_master(ident_t *, kmp_int32 global_tid);
+    OMPRTL__kmpc_end_master
   };
 
 private:
   CodeGenModule &CGM;
   /// \brief Default const ident_t object used for initialization of all other
   /// ident_t objects.
   llvm::Constant *DefaultOpenMPPSource;
   /// \brief Map of flags and corresponding default locations.
public:
   /// \param CriticalName Name of the critical region.
   /// \param CriticalOpGen Generator for the statement associated with the given
   /// critical region.
   virtual void EmitOMPCriticalRegion(CodeGenFunction &CGF,
                                      StringRef CriticalName,
                                      const std::function<void()> &CriticalOpGen,
                                      SourceLocation Loc);
 
+  /// \brief Emits a master region.
+  /// \param MasterOpGen Generator for the statement associated with the given
+  /// master region.
+  virtual void EmitOMPMasterRegion(CodeGenFunction &CGF,
+                                   const std::function<void()> &MasterOpGen,
+                                   SourceLocation Loc);
+
   /// \brief Emits a barrier for OpenMP threads.
   /// \param Flags Flags for the barrier.
   ///
   virtual void EmitOMPBarrierCall(CodeGenFunction &CGF, SourceLocation Loc,
                                   OpenMPLocationFlags Flags);
 
   /// \brief Emits call to void __kmpc_push_num_threads(ident_t *loc, kmp_int32
   /// global_tid, kmp_int32 num_threads) to generate code for 'num_threads'

lib/CodeGen/CGOpenMPRuntime.cpp

CGOpenMPRuntime::CreateRuntimeFunction(OpenMPRTLFunction Function) {
   case OMPRTL__kmpc_flush: {
     // Build void __kmpc_flush(ident_t *loc, ...);
     llvm::Type *TypeParams[] = {getIdentTyPointerTy()};
     llvm::FunctionType *FnTy =
         llvm::FunctionType::get(CGM.VoidTy, TypeParams, /*isVarArg*/ true);
     RTLFn = CGM.CreateRuntimeFunction(FnTy, "__kmpc_flush");
     break;
   }
+  case OMPRTL__kmpc_master: {
+    // Build kmp_int32 __kmpc_master(ident_t *loc, kmp_int32 global_tid);
+    llvm::Type *TypeParams[] = {getIdentTyPointerTy(), CGM.Int32Ty};
+    llvm::FunctionType *FnTy =
+        llvm::FunctionType::get(CGM.Int32Ty, TypeParams, /*isVarArg=*/false);
+    RTLFn = CGM.CreateRuntimeFunction(FnTy, /*Name=*/"__kmpc_master");
+    break;
+  }
+  case OMPRTL__kmpc_end_master: {
+    // Build void __kmpc_end_master(ident_t *loc, kmp_int32 global_tid);
+    llvm::Type *TypeParams[] = {getIdentTyPointerTy(), CGM.Int32Ty};
+    llvm::FunctionType *FnTy =
+        llvm::FunctionType::get(CGM.VoidTy, TypeParams, /*isVarArg=*/false);
+    RTLFn = CGM.CreateRuntimeFunction(FnTy, /*Name=*/"__kmpc_end_master");
+    break;
+  }
   }
   return RTLFn;
 }
 
 llvm::Constant *
 CGOpenMPRuntime::getOrCreateThreadPrivateCache(const VarDecl *VD) {
   // Lookup the entry, lazily creating it if necessary.
   return GetOrCreateInternalVariable(CGM.Int8PtrPtrTy,
void CGOpenMPRuntime::EmitOMPCriticalRegion(
   auto RTLFn = CreateRuntimeFunction(CGOpenMPRuntime::OMPRTL__kmpc_critical);
   CGF.EmitRuntimeCall(RTLFn, Args);
   CriticalOpGen();
   // Build a call to __kmpc_end_critical
   RTLFn = CreateRuntimeFunction(CGOpenMPRuntime::OMPRTL__kmpc_end_critical);
   CGF.EmitRuntimeCall(RTLFn, Args);
 }
 
+static void EmitOMPIfStmt(CodeGenFunction &CGF, llvm::Value *IfCond,
+                          const std::function<void()> &BodyOpGen) {
+  llvm::Value *CallBool = CGF.EmitScalarConversion(
+      IfCond,
+      CGF.getContext().getIntTypeForBitwidth(/*DestWidth=*/32, /*Signed=*/true),
+      CGF.getContext().BoolTy);
+
+  auto *ThenBlock = CGF.createBasicBlock("omp_if.then");
+  auto *ContBlock = CGF.createBasicBlock("omp_if.end");
+  // Generate the branch (If-stmt)
+  CGF.Builder.CreateCondBr(CallBool, ThenBlock, ContBlock);
+  CGF.EmitBlock(ThenBlock);
+  BodyOpGen();
+  // Emit the rest of bblocks/branches
+  CGF.EmitBranch(ContBlock);
+  CGF.EmitBlock(ContBlock, true);
+}
+
+void CGOpenMPRuntime::EmitOMPMasterRegion(
+    CodeGenFunction &CGF, const std::function<void()> &MasterOpGen,
+    SourceLocation Loc) {
+  // if(__kmpc_master(ident_t *, gtid)) {
+  //   MasterOpGen();
+  //   __kmpc_end_master(ident_t *, gtid);
+  // }
+  // Prepare arguments and build a call to __kmpc_master
+  llvm::Value *Args[] = {EmitOpenMPUpdateLocation(CGF, Loc),
+                         GetOpenMPThreadID(CGF, Loc)};
+  auto RTLFn = CreateRuntimeFunction(OMPRTL__kmpc_master);
+  auto *IsMaster = CGF.EmitRuntimeCall(RTLFn, Args);
+  EmitOMPIfStmt(CGF, IsMaster, [&]() -> void {
+    MasterOpGen();
+    // Build a call to __kmpc_end_master
+    RTLFn = CreateRuntimeFunction(OMPRTL__kmpc_end_master);
+    CGF.EmitRuntimeCall(RTLFn, Args);
+  });
+}
+
 void CGOpenMPRuntime::EmitOMPBarrierCall(CodeGenFunction &CGF,
                                          SourceLocation Loc,
                                          OpenMPLocationFlags Flags) {
   // Build call __kmpc_barrier(loc, thread_id)
   llvm::Value *Args[] = {EmitOpenMPUpdateLocation(CGF, Loc, Flags),
                          GetOpenMPThreadID(CGF, Loc)};
   auto RTLFn = CreateRuntimeFunction(CGOpenMPRuntime::OMPRTL__kmpc_barrier);
   CGF.EmitRuntimeCall(RTLFn, Args);

lib/CodeGen/CGStmtOpenMP.cpp

 void CodeGenFunction::EmitOMPSectionDirective(const OMPSectionDirective &) {
   llvm_unreachable("CodeGen for 'omp section' is not supported yet.");
 }
 
 void CodeGenFunction::EmitOMPSingleDirective(const OMPSingleDirective &) {
   llvm_unreachable("CodeGen for 'omp single' is not supported yet.");
 }
 
-void CodeGenFunction::EmitOMPMasterDirective(const OMPMasterDirective &) {
-  llvm_unreachable("CodeGen for 'omp master' is not supported yet.");
+void CodeGenFunction::EmitOMPMasterDirective(const OMPMasterDirective &S) {
+  CGM.getOpenMPRuntime().EmitOMPMasterRegion(
+      *this, [&]() -> void {
+    RunCleanupsScope Scope(*this);
+    EmitStmt(cast<CapturedStmt>(S.getAssociatedStmt())->getCapturedStmt());
+    EnsureInsertPoint();
+  }, S.getLocStart());
 }
 
 void CodeGenFunction::EmitOMPCriticalDirective(const OMPCriticalDirective &S) {
   CGM.getOpenMPRuntime().EmitOMPCriticalRegion(
       *this, S.getDirectiveName().getAsString(), [&]() -> void {
     RunCleanupsScope Scope(*this);
     EmitStmt(cast<CapturedStmt>(S.getAssociatedStmt())->getCapturedStmt());
     EnsureInsertPoint();

test/OpenMP/master_codegen.cpp

+// RUN: %clang_cc1 -verify -fopenmp=libiomp5 -x c++ -emit-llvm %s -fexceptions -fcxx-exceptions -o - | FileCheck %s
+// RUN: %clang_cc1 -fopenmp=libiomp5 -x c++ -std=c++11 -triple x86_64-unknown-unknown -fexceptions -fcxx-exceptions -emit-pch -o %t %s
+// RUN: %clang_cc1 -fopenmp=libiomp5 -x c++ -triple x86_64-unknown-unknown -fexceptions -fcxx-exceptions -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s
+// expected-no-diagnostics
+
+#ifndef HEADER
+#define HEADER
+
+// CHECK:       [[IDENT_T_TY:%.+]] = type { i32, i32, i32, i32, i8* }
+
+// CHECK:       define void [[FOO:@.+]]()
+
+void foo() {}
+
+// CHECK-LABEL: @main
+int main() {
+  // CHECK:       [[A_ADDR:%.+]] = alloca i8
+  char a;
+
+// CHECK:       [[GTID:%.+]] = call i32 @__kmpc_global_thread_num([[IDENT_T_TY]]* [[DEFAULT_LOC:@.+]])
+// CHECK:       [[RES:%.+]] = call i32 @__kmpc_master([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]])
+// CHECK-NEXT:  [[IS_MASTER:%.+]] = icmp ne i32 [[RES]], 0
+// CHECK-NEXT:  br i1 [[IS_MASTER]], label {{%?}}[[THEN:.+]], label {{%?}}[[EXIT:.+]]
+// CHECK:       [[THEN]]
+// CHECK-NEXT:  store i8 2, i8* [[A_ADDR]]
+// CHECK-NEXT:  call void @__kmpc_end_master([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]])
+// CHECK-NEXT:  br label {{%?}}[[EXIT]]
+// CHECK:       [[EXIT]]
+#pragma omp master
+  a = 2;
+// CHECK:       [[RES:%.+]] = call i32 @__kmpc_master([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]])
+// CHECK-NEXT:  [[IS_MASTER:%.+]] = icmp ne i32 [[RES]], 0
+// CHECK-NEXT:  br i1 [[IS_MASTER]], label {{%?}}[[THEN:.+]], label {{%?}}[[EXIT:.+]]
+// CHECK:       [[THEN]]
+// CHECK-NEXT:  call void [[FOO]]()
+// CHECK-NEXT:  call void @__kmpc_end_master([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]])
+// CHECK-NEXT:  br label {{%?}}[[EXIT]]
+// CHECK:       [[EXIT]]
+#pragma omp master
+  foo();
+// CHECK-NOT:   call i32 @__kmpc_master
+// CHECK-NOT:   call void @__kmpc_end_master
+  return a;
+}
+
+#endif