[Users] Carpet with Intel MPI on Stampede

[Ian Hinder]

On 19 Sep 2013, at 17:38, Roland Haas <roland.haas at physics.gatech.edu> wrote:

> Signed PGP part
> Hello Ian,
> 
> > Has anyone else seen these problems, and if so, do you have a 
> > workaround?  I'm using the standard simfactory optionlist with 512 
> > cores and num-threads = 8.
> I have seen these problems with SpEC as well (num-threads=1, about 500
> cores). Eg:
> 
> [170:c425-501] unexpected disconnect completion event from [26:c407-203]
> Assertion failed in file ../../dapl_conn_rc.c at line 1174: 0
> internal ABORT - process 170
> 
> I have no workaround unfortunately.
> 
> With SpEC I also saw ever increasing memory consumption when using
> mvapich (though as far as I could tell from mallinfo there was no sign
> of memory fragmenting the same way we had it in Carpet recently I think).

Thanks Roland.  I think I was also seeing increased memory consumption.  It's a shame because I think Intel MPI was maybe (from memory) 10% faster in overall run speed.

-- 
Ian Hinder
http://numrel.aei.mpg.de/people/hinder

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 203 bytes
Desc: Message signed with OpenPGP using GPGMail
Url : http://lists.einsteintoolkit.org/pipermail/users/attachments/20130919/7f792b02/attachment.bin