[OPENMP] Fix for PR33922: New ident_t flags for

__kmpc_for_static_fini(). Added special flags for calls of __kmpc_for_static_fini(), like previous ly for __kmpc_for_static_init(). Added flag OMP_IDENT_WORK_DISTRIBUTE for distribute cnstruct, OMP_IDENT_WORK_SECTIONS for sections-based constructs and OMP_IDENT_WORK_LOOP for loop-based constructs in location flags. llvm-svn: 312642
2017-09-06 16:17:35 +00:00 · 2017-09-06 16:17:35 +00:00 · f43f714213
parent 818d50a93d
commit f43f714213
7 changed files with 31 additions and 18 deletions
--- a/clang/lib/CodeGen/CGOpenMPRuntime.cpp
+++ b/clang/lib/CodeGen/CGOpenMPRuntime.cpp
@ -3059,11 +3059,19 @@ void CGOpenMPRuntime::emitDistributeStaticInit(
 }

 void CGOpenMPRuntime::emitForStaticFinish(CodeGenFunction &CGF,
-                                          SourceLocation Loc) {
+                                          SourceLocation Loc,
+                                          OpenMPDirectiveKind DKind) {
  if (!CGF.HaveInsertPoint())
    return;
  // Call __kmpc_for_static_fini(ident_t *loc, kmp_int32 tid);
-  llvm::Value *Args[] = {emitUpdateLocation(CGF, Loc), getThreadID(CGF, Loc)};
+  llvm::Value *Args[] = {
+      emitUpdateLocation(CGF, Loc,
+                         isOpenMPDistributeDirective(DKind)
+                             ? OMP_IDENT_WORK_DISTRIBUTE
+                             : isOpenMPLoopDirective(DKind)
+                                   ? OMP_IDENT_WORK_LOOP
+                                   : OMP_IDENT_WORK_SECTIONS),
+      getThreadID(CGF, Loc)};
  CGF.EmitRuntimeCall(createRuntimeFunction(OMPRTL__kmpc_for_static_fini),
                      Args);
 }
--- a/clang/lib/CodeGen/CGOpenMPRuntime.h
+++ b/clang/lib/CodeGen/CGOpenMPRuntime.h
@ -883,8 +883,10 @@ public:
  ///
  /// \param CGF Reference to current CodeGenFunction.
  /// \param Loc Clang source location.
+  /// \param DKind Kind of the directive for which the static finish is emitted.
  ///
-  virtual void emitForStaticFinish(CodeGenFunction &CGF, SourceLocation Loc);
+  virtual void emitForStaticFinish(CodeGenFunction &CGF, SourceLocation Loc,
+                                   OpenMPDirectiveKind DKind);

  /// Call __kmpc_dispatch_next(
  ///          ident_t *loc, kmp_int32 tid, kmp_int32 *p_lastiter,
--- a/clang/lib/CodeGen/CGStmtOpenMP.cpp
+++ b/clang/lib/CodeGen/CGStmtOpenMP.cpp
@ -1692,7 +1692,8 @@ void CodeGenFunction::EmitOMPOuterLoop(
  // Tell the runtime we are done.
  auto &&CodeGen = [DynamicOrOrdered, &S](CodeGenFunction &CGF) {
    if (!DynamicOrOrdered)
-      CGF.CGM.getOpenMPRuntime().emitForStaticFinish(CGF, S.getLocEnd());
+      CGF.CGM.getOpenMPRuntime().emitForStaticFinish(CGF, S.getLocEnd(),
+                                                     S.getDirectiveKind());
  };
  OMPCancelStack.emitExit(*this, S.getDirectiveKind(), CodeGen);
 }
@ -2256,7 +2257,8 @@ bool CodeGenFunction::EmitOMPWorksharingLoop(
        EmitBlock(LoopExit.getBlock());
        // Tell the runtime we are done.
        auto &&CodeGen = [&S](CodeGenFunction &CGF) {
-          CGF.CGM.getOpenMPRuntime().emitForStaticFinish(CGF, S.getLocEnd());
+          CGF.CGM.getOpenMPRuntime().emitForStaticFinish(CGF, S.getLocEnd(),
+                                                         S.getDirectiveKind());
        };
        OMPCancelStack.emitExit(*this, S.getDirectiveKind(), CodeGen);
      } else {
@ -2487,7 +2489,8 @@ void CodeGenFunction::EmitSections(const OMPExecutableDirective &S) {
                         [](CodeGenFunction &) {});
    // Tell the runtime we are done.
    auto &&CodeGen = [&S](CodeGenFunction &CGF) {
-      CGF.CGM.getOpenMPRuntime().emitForStaticFinish(CGF, S.getLocEnd());
+      CGF.CGM.getOpenMPRuntime().emitForStaticFinish(CGF, S.getLocEnd(),
+                                                     S.getDirectiveKind());
    };
    CGF.OMPCancelStack.emitExit(CGF, S.getDirectiveKind(), CodeGen);
    CGF.EmitOMPReductionClauseFinal(S, /*ReductionKind=*/OMPD_parallel);
@ -3120,7 +3123,7 @@ void CodeGenFunction::EmitOMPDistributeLoop(const OMPLoopDirective &S,
                         [](CodeGenFunction &) {});
        EmitBlock(LoopExit.getBlock());
        // Tell the runtime we are done.
-        RT.emitForStaticFinish(*this, S.getLocStart());
+        RT.emitForStaticFinish(*this, S.getLocStart(), S.getDirectiveKind());
      } else {
        // Emit the outer loop, which requests its work chunk [LB..UB] from
        // runtime and runs the inner loop to process it.
--- a/clang/test/OpenMP/distribute_codegen.cpp
+++ b/clang/test/OpenMP/distribute_codegen.cpp
@ -83,7 +83,7 @@ void without_schedule_clause(float *a, float *b, float *c, float *d) {
 // CHECK:  [[BBINNEND]]:
 // CHECK:  br label %[[LPEXIT:.+]]
 // CHECK:  [[LPEXIT]]:
-// CHECK:  call void @__kmpc_for_static_fini(%ident_t* [[DEF_LOC_0]], i32 [[GBL_TIDV]])
+// CHECK:  call void @__kmpc_for_static_fini(%ident_t* [[DEF_LOC_DISTRIBUTE_0]], i32 [[GBL_TIDV]])
 // CHECK:  ret void


@ -145,7 +145,7 @@ void static_not_chunked(float *a, float *b, float *c, float *d) {
 // CHECK:  [[BBINNEND]]:
 // CHECK:  br label %[[LPEXIT:.+]]
 // CHECK:  [[LPEXIT]]:
-// CHECK:  call void @__kmpc_for_static_fini(%ident_t* [[DEF_LOC_0]], i32 [[GBL_TIDV]])
+// CHECK:  call void @__kmpc_for_static_fini(%ident_t* [[DEF_LOC_DISTRIBUTE_0]], i32 [[GBL_TIDV]])
 // CHECK:  ret void


@ -207,7 +207,7 @@ void static_chunked(float *a, float *b, float *c, float *d) {
 // CHECK:  [[BBINNEND]]:
 // CHECK:  br label %[[LPEXIT:.+]]
 // CHECK:  [[LPEXIT]]:
-// CHECK:  call void @__kmpc_for_static_fini(%ident_t* [[DEF_LOC_0]], i32 [[GBL_TIDV]])
+// CHECK:  call void @__kmpc_for_static_fini(%ident_t* [[DEF_LOC_DISTRIBUTE_0]], i32 [[GBL_TIDV]])
 // CHECK:  ret void

 // CHECK-LABEL: test_precond
--- a/clang/test/OpenMP/for_codegen.cpp
+++ b/clang/test/OpenMP/for_codegen.cpp
@ -52,7 +52,7 @@ void without_schedule_clause(float *a, float *b, float *c, float *d) {
 // CHECK-NEXT: br label %{{.+}}
  }
 // CHECK: [[LOOP1_END]]
-// CHECK: call void @__kmpc_for_static_fini([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]])
+// CHECK: call void @__kmpc_for_static_fini([[IDENT_T_TY]]* [[LOOP_LOC]], i32 [[GTID]])
 // CHECK-NOT: __kmpc_barrier
 // CHECK: ret void
 }
@ -93,7 +93,7 @@ void static_not_chunked(float *a, float *b, float *c, float *d) {
 // CHECK-NEXT: br label %{{.+}}
  }
 // CHECK: [[LOOP1_END]]
-// CHECK: call void @__kmpc_for_static_fini([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]])
+// CHECK: call void @__kmpc_for_static_fini([[IDENT_T_TY]]* [[LOOP_LOC]], i32 [[GTID]])
 // CHECK: call {{.+}} @__kmpc_barrier([[IDENT_T_TY]]* [[IMPLICIT_BARRIER_LOC]], i32 [[GTID]])
 // CHECK: ret void
 }
@ -153,7 +153,7 @@ void static_chunked(float *a, float *b, float *c, float *d) {
 // CHECK-NEXT: store i32 [[ADD_UB]], i32* [[OMP_UB]]

 // CHECK: [[O_LOOP1_END]]
-// CHECK: call void @__kmpc_for_static_fini([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]])
+// CHECK: call void @__kmpc_for_static_fini([[IDENT_T_TY]]* [[LOOP_LOC]], i32 [[GTID]])
 // CHECK: call {{.+}} @__kmpc_barrier([[IDENT_T_TY]]* [[IMPLICIT_BARRIER_LOC]], i32 [[GTID]])
 // CHECK: ret void
 }
--- a/clang/test/OpenMP/parallel_for_codegen.cpp
+++ b/clang/test/OpenMP/parallel_for_codegen.cpp
@ -21,7 +21,7 @@ void with_var_schedule() {
 // CHECK: [[CHUNK_VAL:%.+]] = load i8, i8* [[CHUNK]],
 // CHECK: [[CHUNK_SIZE:%.+]] = sext i8 [[CHUNK_VAL]] to i64
 // CHECK: call void @__kmpc_for_static_init_8u([[IDENT_T_TY]]* [[LOOP_LOC]], i32 [[GTID:%[^,]+]], i32 33, i32* [[IS_LAST:%[^,]+]], i64* [[OMP_LB:%[^,]+]], i64* [[OMP_UB:%[^,]+]], i64* [[OMP_ST:%[^,]+]], i64 1, i64 [[CHUNK_SIZE]])
-// CHECK: call void @__kmpc_for_static_fini([[IDENT_T_TY]]* [[DEFAULT_LOC:@[^,]+]], i32 [[GTID]])
+// CHECK: call void @__kmpc_for_static_fini([[IDENT_T_TY]]* [[LOOP_LOC]], i32 [[GTID]])
 #pragma omp parallel for schedule(static, char(a))
  for (unsigned long long i = 1; i < 2; ++i) {
  }
@ -64,7 +64,7 @@ void without_schedule_clause(float *a, float *b, float *c, float *d) {
 // CHECK-NEXT: br label %{{.+}}
  }
 // CHECK: [[LOOP1_END]]
-// CHECK: call void @__kmpc_for_static_fini([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]])
+// CHECK: call void @__kmpc_for_static_fini([[IDENT_T_TY]]* [[LOOP_LOC]], i32 [[GTID]])
 // CHECK: ret void
 }

@ -105,7 +105,7 @@ void static_not_chunked(float *a, float *b, float *c, float *d) {
 // CHECK-NEXT: br label %{{.+}}
  }
 // CHECK: [[LOOP1_END]]
-// CHECK: call void @__kmpc_for_static_fini([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]])
+// CHECK: call void @__kmpc_for_static_fini([[IDENT_T_TY]]* [[LOOP_LOC]], i32 [[GTID]])
 // CHECK: ret void
 }

@ -165,7 +165,7 @@ void static_chunked(float *a, float *b, float *c, float *d) {
 // CHECK-NEXT: store i32 [[ADD_UB]], i32* [[OMP_UB]]

 // CHECK: [[O_LOOP1_END]]
-// CHECK: call void @__kmpc_for_static_fini([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]])
+// CHECK: call void @__kmpc_for_static_fini([[IDENT_T_TY]]* [[LOOP_LOC]], i32 [[GTID]])
 // CHECK: ret void
 }

--- a/clang/test/OpenMP/sections_codegen.cpp
+++ b/clang/test/OpenMP/sections_codegen.cpp
@ -70,7 +70,7 @@ int main() {
 // CHECK-NEXT: br label %[[INNER_FOR_COND]]
 // CHECK:      [[INNER_LOOP_END]]
  }
-// CHECK:      call void @__kmpc_for_static_fini(%{{.+}}* @{{.+}}, i32 [[GTID]])
+// CHECK:      call void @__kmpc_for_static_fini(%{{.+}}* [[SECTIONS_LOC]], i32 [[GTID]])
 // CHECK:      call void @__kmpc_barrier(%{{.+}}* [[IMPLICIT_BARRIER_SECTIONS_LOC]],
 #pragma omp sections nowait
  {