Hi,
First let me apologize for this late reply. Somehow I skipped over this
mail. Ok now to answer your mail.
Unfortunately there is no MPI portable way to do this. Even after
defining your own error handler other than MPI_ERRORS_ABORT its not
trivial to discover and print out which process failed and then to map
this process id to specific rank in a specific communicator. Note all
this can be done with some effort for LAM however I emphasize again there
is *no* MPI portable way of doing it.
Best,
Manish Chablani
------------------------------------------------------
Graduate Student, CS Department, Indiana University.
http://www.cs.indiana.edu/~mchablan
LAM/MPI Developer
Make today a LAM/MPI day !!!
http://www.lam-mpi.org
------------------------------------------------------
On Mon, 31 Mar 2003, Konstantin Skaburskas wrote:
> Hi,
>
> How can I get rank of a died at collective operation process? (Is this a
> realization specific procedure?)
>
> E.g., at MPI level using MPI_ERROR_STRING() it is possible to map an error
> code returned by failed MPI operation to a string. But in my case
> (MPI_[ALL]REDUCE()) this string say nothing about rank of a died process -
> "MPI_Error_string: process in local group is dead".
>
> What should I do to get rank of this process?
>
> Thank you in advance.
>
> Regards,
> Konstantin Skaburskas
>
> _______________________________________________
> This list is archived at http://www.lam-mpi.org/MailArchives/lam/
>
|
