r366483 - [OPENMP]Fix sharing of threadprivate variables with TLS support.

Alexey Bataev via cfe-commits cfe-commits at lists.llvm.org
Thu Jul 18 12:40:24 PDT 2019


Author: abataev
Date: Thu Jul 18 12:40:24 2019
New Revision: 366483

URL: http://llvm.org/viewvc/llvm-project?rev=366483&view=rev
Log:
[OPENMP]Fix sharing of threadprivate variables with TLS support.

If the threadprivate variable is used in the copyin clause on inner
parallel directive with TLS support, we capture this variable in all
outer OpenMP scopes. It leads to the fact that in all scopes we're
working with the original variable, not the threadprivate copies.

Modified:
    cfe/trunk/lib/Sema/SemaOpenMP.cpp
    cfe/trunk/test/OpenMP/parallel_copyin_codegen.cpp

Modified: cfe/trunk/lib/Sema/SemaOpenMP.cpp
URL: http://llvm.org/viewvc/llvm-project/cfe/trunk/lib/Sema/SemaOpenMP.cpp?rev=366483&r1=366482&r2=366483&view=diff
==============================================================================
--- cfe/trunk/lib/Sema/SemaOpenMP.cpp (original)
+++ cfe/trunk/lib/Sema/SemaOpenMP.cpp Thu Jul 18 12:40:24 2019
@@ -1882,6 +1882,13 @@ bool Sema::isOpenMPPrivateDecl(const Val
         !isOpenMPSimdDirective(DSAStack->getCurrentDirective()))
       return true;
   }
+  if (const auto *VD = dyn_cast<VarDecl>(D)) {
+    if (DSAStack->isThreadPrivate(const_cast<VarDecl *>(VD)) &&
+        DSAStack->isForceVarCapturing() &&
+        !DSAStack->hasExplicitDSA(
+            D, [](OpenMPClauseKind K) { return K == OMPC_copyin; }, Level))
+      return true;
+  }
   return DSAStack->hasExplicitDSA(
              D, [](OpenMPClauseKind K) { return K == OMPC_private; }, Level) ||
          (DSAStack->isClauseParsingMode() &&

Modified: cfe/trunk/test/OpenMP/parallel_copyin_codegen.cpp
URL: http://llvm.org/viewvc/llvm-project/cfe/trunk/test/OpenMP/parallel_copyin_codegen.cpp?rev=366483&r1=366482&r2=366483&view=diff
==============================================================================
--- cfe/trunk/test/OpenMP/parallel_copyin_codegen.cpp (original)
+++ cfe/trunk/test/OpenMP/parallel_copyin_codegen.cpp Thu Jul 18 12:40:24 2019
@@ -19,6 +19,7 @@
 // RUN: %clang_cc1 -verify -fopenmp -x c++ -std=c++11 -DLAMBDA -triple x86_64-linux -emit-llvm %s -o - | FileCheck -check-prefix=TLS-LAMBDA %s
 // RUN: %clang_cc1 -verify -fopenmp -x c++ -fblocks -DBLOCKS -triple x86_64-linux -emit-llvm %s -o - | FileCheck -check-prefix=TLS-BLOCKS %s
 // RUN: %clang_cc1 -verify -fopenmp -x c++ -std=c++11 -DARRAY -triple x86_64-linux-gnu -emit-llvm %s -o - | FileCheck -check-prefix=TLS-ARRAY %s
+// RUN: %clang_cc1 -verify -fopenmp -x c++ -DNESTED -triple x86_64-linux -emit-llvm %s -o - | FileCheck %s -check-prefix=NESTED
 
 // RUN: %clang_cc1 -verify -fopenmp-simd -x c++ -triple x86_64-linux -emit-llvm %s -o - | FileCheck --check-prefix SIMD-ONLY1 %s
 // RUN: %clang_cc1 -fopenmp-simd -x c++ -std=c++11 -triple x86_64-linux -emit-pch -o %t %s
@@ -28,7 +29,7 @@
 // RUN: %clang_cc1 -verify -fopenmp-simd -x c++ -std=c++11 -DARRAY -triple x86_64-linux-gnu -emit-llvm %s -o - | FileCheck --check-prefix SIMD-ONLY1 %s
 // SIMD-ONLY1-NOT: {{__kmpc|__tgt}}
 // expected-no-diagnostics
-#ifndef ARRAY
+#if !defined(ARRAY) && !defined(NESTED)
 #ifndef HEADER
 #define HEADER
 
@@ -493,7 +494,7 @@ int main() {
 // TLS-CHECK: ret void
 
 #endif
-#else
+#elif defined(ARRAY)
 // ARRAY-LABEL: array_func
 // TLS-ARRAY-LABEL: array_func
 
@@ -522,6 +523,24 @@ void array_func() {
 #pragma omp parallel copyin(a, s)
   ;
 }
-#endif
-
+#elif defined(NESTED)
+int t;
+#pragma omp threadprivate(t)
+// NESTED: foo
+void foo() {
+  // NESTED: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @{{.+}}, i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* [[OUTLINED:@.+]] to void (i32*, i32*, ...)*))
+#pragma omp parallel
+#pragma omp parallel copyin(t)
+  ++t;
+}
+// NESTED: define {{.*}}void [[OUTLINED]](
+// NESTED: [[T:%.+]] = call i32* [[THRP_T:@.+]]()
+// NESTED:  call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @{{.+}}, i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*)* [[OUTLINED1:@.+]] to void (i32*, i32*, ...)*), i32* [[T]])
+
+// NESTED: define {{.*}}void [[OUTLINED1]](
+// NESTED: [[T_MASTER:%.+]] = load i32*, i32** %
+// NESTED: [[T:%.+]] = call i32* [[THRP_T]]()
+// NESTED: [[T_MASTER_VAL:%.+]] = load i32, i32* [[T_MASTER]],
+// NESTED: store i32 [[T_MASTER_VAL]], i32* [[T]],
+#endif // NESTED
 




More information about the cfe-commits mailing list