Hi Raghav,

Thank you for using Argobots!

Yes, in this case, please use ABT_thread. I understand that the naming is extremely confusing, but ABT_thread (=user-level threads) works as a task in typical HPC fine-grained task-parallel programs. Other libraries used for task parallel programs also use "threads" for task implementations (Qthreads, MassiveThreads, ...). You can consider ABT_task (=tasklet) as a special user-level thread that cannot yield. The Argobots paper (Argobots: A Lightweight Low-Level Threading and Tasking Framework) describes the details.

If you have any further questions, please feel free to ask here!

Thanks,

Shintaro

On Thu, Oct 7, 2021 at 11:28 AM Raghav Gupta via discuss <discuss@lists.argobots.org> wrote:

Pasting the code snippet link again without any formatting - https://gist.github.com/raghav0307/9252fc7b568bc35baab29bdc9da70618

On Thu, Oct 7, 2021 at 11:55 PM Raghav Gupta <raghav18076@iiitd.ac.in> wrote:
Hi,

I am trying to use argobots for writing a recursive task parallel program.

I have taken a very simple recursive task program example, where I am creating a parent argobot task which creates two child argobots tasks and then waits for them by calling ABT_task_free(). When I run this program with one xstream I find that it doesn't terminate. (Code snippet - https://gist.github.com/raghav0307/9252fc7b568bc35baab29bdc9da70618 )

My understanding is that this is because the parent task never yields the xstream to the child tasks, and hence the program gets stuck.

This program terminates when I use argobot threads instead of argobot tasks. Could someone point me to why this is not working with argobot tasks?

Best,
Raghav

_______________________________________________
discuss mailing list
discuss@lists.argobots.org
https://lists.argobots.org/mailman/listinfo/discuss