[mpich-discuss] interrupt safety on MPI_Comm_create

[Dave Goodell]

It means that it's not safe to call this routine from an interrupt handler, which on a unix-ish system means a signal handler.  The basic idea is that if you are in the middle of this routine and a signal comes interrupts that call and causes you to call the routine again, you will have corruption, deadlocks, or other badness because of the way that the routine is designed.

Oversubscribing cores has nothing to do with this.

-Dave

On Jan 20, 2012, at 11:39 AM CST, Nichols A. Romero wrote:

> This documentation:
> http://www.mcs.anl.gov/research/projects/mpi/www/www3/MPI_Comm_create.html
> 
> states that MPI_Comm_create is not interrupt safe. 
> 
> Note that I am not a computer scientist. What would be a scenario where interrupt safety might be a concern? Could oversubscribing cores with MPI tasks cause
> problems?
> 
> 
> Nichols A. Romero, Ph.D.
> Argonne Leadership Computing Facility
> Argonne National Laboratory
> Building 240 Room 2-127
> 9700 South Cass Avenue
> Argonne, IL 60490
> (630) 252-3441
> 
> 
> _______________________________________________
> mpich-discuss mailing list     mpich-discuss at mcs.anl.gov
> To manage subscription options or unsubscribe:
> https://lists.mcs.anl.gov/mailman/listinfo/mpich-discuss