[MPICH] RE: tuning threshold for Allgather

[Rajeev Thakur]

It is not a typo. There could be a better number, but the cutoff point in
our experiments was around there. (It represents the total size of the
message gathered, not per-process size.)

Rajeev 

> -----Original Message-----
> From: amith rajith mamidala [mailto:mamidala at cse.ohio-state.edu] 
> Sent: Monday, April 16, 2007 12:42 PM
> To: mpich-discuss at mcs.anl.gov
> Cc: thakur at mcs.anl.gov; mvapich-core at cse.ohio-state.edu
> Subject: tuning threshold for Allgather
> 
> Hi all,
> 
> I would like to check with you if the following value for
> the parameter should be 8192 or is it tuned for 81920?
> This parameter is the same for both mpich1 and mpich2 distributions.
> 
> #define MPIR_ALLGATHER_SHORT_MSG 81920
> 
> Thanks,
> Amith
> 
>