[mpich-discuss] Dismatch in the data size field of ch3 packet header between sender and receiver

[Dave Goodell]

The log files are for internal debugging, and may not be reliably reporting message sizes, or may not be reporting all received messages.  Especially if you are only using the ch3_msg class.  Our debug messages are unfortunately not entirely consistent about which class and level they use...

Probably the only way to check is to inspect the code to manually ensure that all packet send/recv actions are appropriately and consistently logged.

-Dave

On Jun 18, 2010, at 2:06 PM CDT, Bowen Zhou wrote:

> Hi all,
> 
> Please forgive my ignorance if the following question is silly and
> meaningless in your opinion. I would be very thankful if you could shed
> a light on it.
> 
> I am running NASPar benchmark on MPICH2 with the ch3 message debugging
> enabled. By adding up the data size of sent and received packets in the
> log respectively, I found the total amount of data sent from all nodes
> is greater than the total amount of data received by all nodes. IMHO,
> either I have misconfiguration in enabling the ch3 debugging log or the
> ch3 message debugging does not cover all received packets. And maybe
> something else. I hope you MPICH gurus could give some hints.
> 
> My one-line command for running NASPar on MPICH2 as follows:
> mpiexec -n $nprocs $app -mpich-dbg=file -mpich-dbg-class=ch3_msg
> 
> And my configure for the MPICH2 I used is as follows:
> ./configure --enable-sharedlibs=gcc --enable-g=log
> 
> And I am willing to provide other info upon request.
> 
> TIA,
> 
> Bowen
> 
> _______________________________________________
> mpich-discuss mailing list
> mpich-discuss at mcs.anl.gov
> https://lists.mcs.anl.gov/mailman/listinfo/mpich-discuss