[OPENMP] Codegen for 'omp master' directive

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);
}

Differential Revision: http://reviews.llvm.org/D6473

llvm-svn: 223342

clang/lib/CodeGen/CGOpenMPRuntime.cpp

@@ -371,6 +371,22 @@ CGOpenMPRuntime::CreateRuntimeFunction(OpenMPRTLFunction Function) {
     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;
 }
@@ -631,6 +647,57 @@ void CGOpenMPRuntime::EmitOMPCriticalRegion(
   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.
+    // OpenMP [1.2.2 OpenMP Language Terminology]
+    // For C/C++, an executable statement, possibly compound, with a single
+    // entry at the top and a single exit at the bottom, or an OpenMP construct.
+    // * Access to the structured block must not be the result of a branch.
+    // * The point of exit cannot be a branch out of the structured block.
+    // * The point of entry must not be a call to setjmp().
+    // * longjmp() and throw() must not violate the entry/exit criteria.
+    // * An expression statement, iteration statement, selection statement, or
+    // try block is considered to be a structured block if the corresponding
+    // compound statement obtained by enclosing it in { and } would be a
+    // structured block.
+    // It is analyzed in Sema, so we can just call __kmpc_end_master() on
+    // fallthrough rather than pushing a normal cleanup for it.
+    RTLFn = CreateRuntimeFunction(OMPRTL__kmpc_end_master);
+    CGF.EmitRuntimeCall(RTLFn, Args);
+  });
+}
+
 void CGOpenMPRuntime::EmitOMPBarrierCall(CodeGenFunction &CGF,
                                          SourceLocation Loc,
                                          OpenMPLocationFlags Flags) {

clang/lib/CodeGen/CGOpenMPRuntime.h

@@ -96,7 +96,11 @@ class CGOpenMPRuntime {
     // 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,
   };
 
   CodeGenModule &CGM;
@@ -287,6 +291,13 @@ class CGOpenMPRuntime {
                                      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.
   ///

clang/lib/CodeGen/CGStmtOpenMP.cpp

@@ -490,8 +490,13 @@ 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) {

clang/test/OpenMP/master_codegen.cpp

@@ -0,0 +1,46 @@
+// 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