0026ed2f9c
When an omp for loop is canceled the constructed objects are being destructed
twice.
It looks like the desired code is:
{
Obj o;
If (cancelled) branch-through-cleanups to cancel.exit.
}
[cleanups]
cancel.exit:
__kmpc_for_static_fini
br cancel.cont (*)
cancel.cont:
__kmpc_barrier
return
The problem seems to be the branch to cancel.cont is currently also going
through the cleanups calling them again. This change just does a direct branch
instead.
Patch By: michael.p.rice@intel.com
Differential Revision: https://reviews.llvm.org/D35854
llvm-svn: 309288
47 lines
1.3 KiB
C++
47 lines
1.3 KiB
C++
// RUN: %clang_cc1 -std=c++11 -fopenmp -fopenmp-version=45 -triple x86_64-apple-darwin13.4.0 -emit-llvm -o - %s | FileCheck %s
|
|
|
|
//CHECK: call i32 @__kmpc_cancel
|
|
//CHECK: br {{.*}}label %[[EXIT:[^,].+]], label %[[CONTINUE:.+]]
|
|
//CHECK: [[EXIT]]:
|
|
//CHECK: store i32 [[EXIT_SLOT:[0-9]+]]
|
|
//CHECK: br label %[[CLEANUP:.+]]
|
|
//CHECK: [[CONTINUE]]:
|
|
//CHECK: store i32 [[CONT_SLOT:[0-9]+]],
|
|
//CHECK: br label %[[CLEANUP]]
|
|
//CHECK: [[CLEANUP]]:
|
|
//CHECK-NEXT: call void @_ZN3ObjD1Ev
|
|
//CHECK: switch i32{{.*}}, label %[[UNREACHABLE:.+]] [
|
|
//CHECK: i32 [[CONT_SLOT]], label %[[CLEANUPCONT:.+]]
|
|
//CHECK: i32 [[EXIT_SLOT]], label %[[CANCELEXIT:.+]]
|
|
//CHECK-NEXT: ]
|
|
//CHECK: [[CLEANUPCONT]]:
|
|
//CHECK: br label %[[CANCELCONT:.+]]
|
|
//CHECK: [[CANCELCONT]]:
|
|
//CHECK-NEXT: call void @__kmpc_barrier(
|
|
//CHECK-NEXT: ret void
|
|
//CHECK: [[UNREACHABLE]]:
|
|
//CHECK-NEXT: unreachable
|
|
//CHECK-NEXT: }
|
|
|
|
struct Obj {
|
|
int a; Obj(); Obj(const Obj& r) = delete; Obj &operator=(const Obj& r);
|
|
~Obj();
|
|
};
|
|
|
|
void foo() {
|
|
int i,count = 0;
|
|
Obj obj;
|
|
|
|
#pragma omp parallel private(i) num_threads(1)
|
|
{
|
|
#pragma omp for reduction(+:count) lastprivate(obj)
|
|
for (i=0; i<1000; i++) {
|
|
if(i==100) {
|
|
obj.a = 100;
|
|
#pragma omp cancel for
|
|
}
|
|
count++;
|
|
}
|
|
}
|
|
}
|