Re: Myrinet error - help intrepreting

Hi Jeff,

On Tue, 2003-08-05 at 05:41, Jeff Layton wrote:
>    I'm running a Scalapack code on a cluster with Myrinet.
> I'm using the Intel compilers and MPICH-GM (version
> 1.2.5..10-1). When I'm runing I'm code, I see the following
> in the output file:
> 
> [0]: alloc failed, not enough memory (Fatal Error)
> Context: <(gmpi_allocate_packet) malloc: send buf>
> 
> I'm not sure what this error message is telling me. Are my
> message sizes too large?

No, it means that MPICH-GM tried to malloc some memory (not much, less
than 16 KB) to buffer a packet to send, and it failed. It seems that you
are running out of memory on the machine. If it's not due to normal
memory usage with Scalapack, it may be due to a slow or non-responsive
node, triggering buffering on the send side (this buffering is limited
to 8 MB).

Are you sure you have enough place for the Scalapack workspaces ?

Patrick
-- 

Patrick Geoffray
Myricom, Inc.
http://www.myri.com