Doubt about MPI_Comm_shrink

[Lucas Baptista De Moraes]

Hello everyone, I'm getting started with MPI and ULFM studies, and I've a doubt of using MPI_Comm_shrink. If I raise a failure in some rank, like rank 0, and call MPI_Comm_shrink, my code will work, but without the output of rank 0. Is there a way of getting the output from this rank or it's impossible?

int main (int argc, char **argv)

{

MPI_Init (NULL, NULL);

int siterank;

MPI_Comm siteparent, sitemachine, sitecomm, ft_sitecomm;

MPI_Comm_get_parent (&siteparent);

siteparent = sitecomm;

sitecomm = MPI_COMM_WORLD;

MPI_Comm_rank (sitecomm, &siterank);

MPI_Comm_set_errhandler (sitecomm, MPI_ERRORS_RETURN);

if (siterank==0) raise (SIGKILL);

MPIX_Comm_shrink (sitecomm, &ft_sitecomm);

printf ("Hey there, I'm Site %d, and I'm working! :) \n\n", siterank);

MPI_Comm_spawn ("machine", MPI_ARGV_NULL, 3, MPI_INFO_NULL, 0, ft_sitecomm,

&sitemachine, MPI_ERRCODES_IGNORE);

MPI_Finalize ();

return 0;

}

[George Bosilca]

Lucas,

First: MPIX_Comm_shrink is a collective function it must therefore be called by all alive processes in sitecomm (and not only by some of them).

Second: In your code you kill your process 0 (raise(SIGKILL) is as if you were doing a "kill -9" so your process is killed by the MPI daemon). This means that process 0 will not execute anything after the kill call, so it will never get to the printf. Thus will should never see the printf from 0 (if you do that would be worrisome and definitively incorrect from ULFM perspective).

George.

--
You received this message because you are subscribed to the Google Groups "User Level Fault Mitigation" group.
To unsubscribe from this group and stop receiving emails from it, send an email to ulfm+uns...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/ulfm/72f8ae90-0d81-4774-8de5-7b9fa9ca4137%40googlegroups.com.