AMBER Archive (2002)

Subject: compiling MPICH

From: Thomas Steinbrecher (thomas.steinbrecher_at_physchem.uni-freiburg.de)
Date: Thu Jul 11 2002 - 07:46:12 CDT


Greetings AMBER users,

I have a question concerning MPICH:

I want to use AMBER with MPICH on a Linux cluster with Dual
Athlon processor nodes and a single Athlon processor
master.

It seems to me, that I cannot run sander with 2 processors
of the same computer

(using mpirun -np2 -nolocal with only the master and 1 dual
processor node in my machinefile).

it fails with:

P0_XXXX p4_error : 1

(XXXX is a 4 digit number different in several tries)

without the -nolocal option, it runs with 1 proc on the
master and 1 on the node

I compiled MPICH with the -dev=ch_p4 and -comm=shared
option. Was that right or do I have to recompile it with
different settings? What compilation options for MPICH
should I use?

Kind regards and sorry to trouble you again with newbie
questions,

Thomas