[mpich-discuss] Assertion error for bcast2 test (mpich2 1.3 with KNEM)

[Dave Goodell]

It's unlikely to solve your problem, but there is a slightly newer version available: http://runtime.bordeaux.inria.fr/knem/download/knem-0.9.3.tar.gz

Otherwise it will be a little while until I can get spare cycles for this, unfortunately.

-Dave

On Nov 10, 2010, at 3:36 PM CST, Miao wrote:

> Hi Dave,
> 
> Yes the one we used is the latest KNEM whose version is 0.9.2. Sorry that I forgot to mention it in the previous email.
> 
> Thanks,
> Miao
> 
> On Wed, Nov 10, 2010 at 11:51 AM, Dave Goodell <goodell at mcs.anl.gov> wrote:
> On Nov 4, 2010, at 9:17 PM CDT, Miao wrote:
> 
> > When I tried to run mpich2 1.3 nemesis with KNEM for the tests inside mpich2 source code, an assertion error for bcast2 test under test/mpi/coll came out as follow:
> >
> > $ ../../../bin/mpiexec -machinefile hosts -n 2 ./bcast2
> > Assertion failed in file mpid_nem_lmt_dma.c at line 474: ("should never be incomplete!", 0)
> > internal ABORT - process 1
> >
> > We are wondering if this is a bug within nemesis codes for KNEM and we'd like to report it to mpich2 group.
> 
> What version of knem are you using?  I just looked and noticed that the version we ship in the contrib directory of MPICH2 is very old (v0.5.0).  I've taken a note to update that, but in the interim you would be better off using the latest version from the knem website if you aren't already.
> 
> http://runtime.bordeaux.inria.fr/knem/
> 
> -Dave
> 
> _______________________________________________
> mpich-discuss mailing list
> mpich-discuss at mcs.anl.gov
> https://lists.mcs.anl.gov/mailman/listinfo/mpich-discuss
> 
> _______________________________________________
> mpich-discuss mailing list
> mpich-discuss at mcs.anl.gov
> https://lists.mcs.anl.gov/mailman/listinfo/mpich-discuss