http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55251
Bug #: 55251
Summary: inconsistent OpenMP tasks scheduling
Classification: Unclassified
Product: gcc
Version: unknown
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: libgomp
AssignedTo: unassig...@gcc.gnu.org
ReportedBy: alfredo.butt...@gmail.com
I have problems when executing a standard OpenMP tasks example from the OpenMP
3.1 API document (Example A.15.2c):
struct node {
struct node *left;
struct node *right;
};
extern void process(struct node *);
void postorder_traverse( struct node *p ) {
if (p->left)
#pragma omp task
// p is firstprivate by default
postorder_traverse(p->left);
if (p->right)
#pragma omp task
// p is firstprivate by default
postorder_traverse(p->right);
#pragma omp taskwait
process(p);
}
int main(){
...
#pragma omp parallel
{
#pragma omp master
{
postorder_traverse(root);
}
}
...
The problem is related to the scheduling of tasks. I want to count the number
of tasks executed by each OpenMP thread to check whether the workload
distribution is balanced. Now, if I use two threads then the tasks are pretty
much equally distributed. If I use 3 or 4 threads, the master thread always
executes only one task (corresponding to the root, topmost node of the tree).
With more than 4 threads the behavior is inconsistent in the sense that
sometimes the tasks are well distributed, sometimes the master does only one
task. Is this a bug? This happens with different versions of gcc from 4.4 up to
4.7 but not with the intel compiler.
Thanks
Alfredo
