Deadlock for addition operator

[Giang Huynh]

Hello everyone,

I'm using addition operators on a variable called 'hat_solution' by

----------------

  hat_solution=0;
  hat_solution.add(1.0, old_solution,timestep, old_vel_solution);
  hat_solution.add(0.5*timestep*timestep*(1.0-2.0*beta),old_acc_solution );

----------------

, the above variables are trillinos block vectors, and initialized by

----------------

  // Actual solution at time step n
  old_solution.reinit(partition_relevant, mpi_com);
  old_vel_solution.reinit(partition_relevant, mpi_com);
  old_acc_solution.reinit(partition_relevant, mpi_com);

  // hat solution at time step n-1
  hat_solution.reinit(partition_relevant, mpi_com);

----------------

On my personal laptop, the code runs well with multiple cores, but there was a deadlock  when calling  add() function on a supercomputer. 

I really appreciate if anyone has some thoughts on this issue.

Thanks,

Giang

[Timo Heister]

Giang,

Parallel hangs and similar bugs are hard to debug and fix. One thing
that is helpful is to run each rank in a debugger (or attach after
running) and print a stack trace. This will tell you if some ranks are
stuck somewhere earlier. Are you sure all ranks made it to the .add
call?

It could be that you forgot a call to compress() after assembling into
the vector somewhere before this addition.

On Sun, May 28, 2023 at 11:25 PM Giang Huynh <dongg...@gmail.com> wrote:
>
> Hello everyone, I'm using addition operators on a variable called 'hat_solution' by ---------------- hat_solution=0; hat_solution. add(1. 0, old_solution,timestep, old_vel_solution); hat_solution. add(0. 5*timestep*timestep*(1. 0-2. 0*beta),old_acc_solution

> ZjQcmQRYFpfptBannerStart
> This Message Is From an External Sender
> Use caution when opening links or attachments if you do not recognize the sender.
>
> ZjQcmQRYFpfptBannerEnd

> --
> The deal.II project is located at http://www.dealii.org/
> For mailing list/forum options, see https://groups.google.com/d/forum/dealii?hl=en
> ---
> You received this message because you are subscribed to the Google Groups "deal.II User Group" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to dealii+un...@googlegroups.com.
> To view this discussion on the web visit https://groups.google.com/d/msgid/dealii/d36e19a4-b437-435d-afd9-9491b31aa845n%40googlegroups.com.



--
Timo Heister
http://www.math.clemson.edu/~heister/