A few days ago I submitted the question whether in
compiling openmpi for amber9 (dual opteron machine
with Linux Debian amd64 etch) with intel compilers I
should take into account that intel compilers are
already providing runtime to a QM code (NWChem 5.0
which has its own - not MPI - parallelization). No
answer was taken for "no risk".
The runtime shows the location of intel:
locate libimf.so
/opt/intel/fce/9.1.036/lib/libimf.so
/opt/intel/cce/9.1.042/lib/libimf.so
Now I have started trying to install a MPI library:
cd /usr/local
tar xvf openmpi-1.2.1.tar
cd /usr/local/openmpi-1.2.1
export FC=/opt/intelcce/9.1.042/bin/icc
I did nothing more because
set | grep FC
reported nothing (I expected the environment variable
FC), while
set |grep PWD
reports the following:
OLDPWD=/opt/intel/cce/9.1.042
PWD=/opt/intel/cce/9.1.042/bin
Now "setenv"
reports
PWD=/opt/intel/cce/9.1.042/bin
Unfortunately I did not check setenv before the above
"export".
The command "locate" reports libimf.so located as
above, before trying to install the MPI library.
Therefore, I assume that the correct runtime is still
provided to the QM code. Though, I am confused by the
above "reports" and would be grateful for your advice
before risking to mess my system.
Thanks
francesco pietra
____________________________________________________________________________________
Get the Yahoo! toolbar and be alerted to new email wherever you're surfing.
http://new.toolbar.yahoo.com/toolbar/features/mail/index.php
-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber.scripps.edu
To unsubscribe, send "unsubscribe amber" to majordomo.scripps.edu
Received on Sun Jun 17 2007 - 06:07:24 PDT
