Re: NAMD on Linux cluster with Myrinet

From: Alberto Santana (asantana_at_uprm.edu)
Date: Thu Oct 14 2004 - 11:26:08 CDT

I followed Florian's suggestion and built charm++ with:

build charm++ mpi-linux icc gm2 -O2 -L/usr/rels/mpich/lib/shared
-I/usr/rels/mpich/include -L/usr/rels/mpich/lib
-L/opt/gm/lib -I/opt/gm/include -lmpich -lgm -DCMK_OPTIMIZE=1

 and added -lpmpich and -lpthread before compiling namd2, otherwise it won't
compile.

The charm++ test goes well. When running the BPTI files used to do the
benchmark it aborts with
...
Info: 2643 DEGREES OF FREEDOM
Info: 449 HYDROGEN GROUPS
Info: TOTAL MASS = 6446.48 amu
Info: TOTAL CHARGE = 6 e
Info: *****************************
Info: Entering startup phase 0 with 43082 kB of memory in use.
Info: Entering startup phase 1 with 43082 kB of memory in use.
Info: Entering startup phase 2 with 43272 kB of memory in use.
Info: Entering startup phase 3 with 43278 kB of memory in use.
FATAL ERROR: Bug in PatchMap::initialize - negative grid dimension.
FATAL ERROR: Bug in PatchMap::initialize - negative grid dimension.
[0] MPI Abort by user Aborting program !
[0] Aborting program!

Note that I used -O2 as suggested by Gengbin
(http://www.ks.uiuc.edu/Research/namd/mailing_list/namd-l/0693.html)

Alberto

On Monday 11 October 2004 08:49 pm, you wrote:
> Hello Alberto,
>
> i recently (two days ago) compiled NAMD 2.5 on my linux cluster with
> myrinet. in order for charm to work with myrinet i compiled it with
>
> ./build charm++ mpi-linux gm2 -O3 -I$MPICH/include -L$MPICH/lib
> -L$GM/lib -lmpich -lgm -DCMK_OPTIMIZE=1
>
> note the gm2 flag for gm version 2.xx. as i have athlon cpus i didn't
> use icc but gcc3.3 instead. you must have mpich-gm >1.2.5..12 installed
> in oder for everything to work (at least with the mpi-linux version).
> mpich-gm should be compiled with the same compiler used for building charm.
>
> Regards
>
> Florian Barth

This archive was generated by hypermail 2.1.6 : Wed Feb 29 2012 - 15:38:55 CST