OpenMP® Forum

That was what I expected but not what I hoped. If we remove the taskgroup, the code I have performs better if spawning the tasks inside a omp do (they are dependency-free, or well, dependency-free of any dependency structure recognised as of OpenMP 5.0 TR) by a reasonable amount. I guess the best op...

Hi, I have a question about how taskgroup behaves with multiple threads spawning tasks, as I'm struggling slightly with the description in the standard. The standard (OpenMP 5.0 TR) says: The taskgroup construct specifies a wait on completion of child tasks of the current task and their descendent t...

Naming task regions: With the new OMPT interface with OpenMP, it should be possible to develop tools to visualise the computation according to what thread is doing what work (which can be identified from things like the ompt_record_task_create_s.codeptr_ra for tasks for example). It may be useful f...

Hi, I'm having an issue with a piece of OpenMP code, where it works with one compiler but not another, and I'm struggling to understand if what I'm doing is unsupported according to the MPI Standard, or if I've made an error. The issue involves task's spawning tasks if some criteria is met, inside o...

Ok thanks. I'll have another think on how to do this then

Hi again. I'm confused as to if I understand how openMP (inout) dependencies on array sections work correctly. My understand would be as follows: If I create a task as follows: #pragma omp depend(inout: a[32:32]) task_function(a); and later: #pragma omp depend(inout: a[0:64]) The latter task should ...

Thanks for the detailed reponses Mark, that makes sense now.

This code probably doesn't do what you expect, because tasks generated by different threads are not sibling tasks. To make all the tasks sibling tasks, you would need to generate them in a master (or single) construct. If you do this, though, it likely results in less parallelism than is potentiall...

Hi, I've been working with tasks a bit and was just wanting to make sure my understanding of how race conditions, tasks and locks work was correct. The first is to do with the depend clause, and just to confirm what I understand the standard to mean - any tasks which have a dependency based on any v...