Hi Molpro-Users,
The problems that I have been having running the included test set have
surfaced again, this time with the precompiled binary. When I attempt to run
the binary on some test jobs (namely bccd_opt, but others occasionally as well,
with less reproducibility), the output terminates with the lines
0.524 MB (compressed) written to integral file ( 86.2%)
Node minimum: 0.524 MB, node maximum: 0.524 MB
NUMBER OF SORTED TWO-ELECTRON INTEGRALS: 34949. BUFFER LENGTH: 32768
NUMBER OF SEGMENTS: 1 SEGMENT LENGTH: 34949 RECORD LENGTH: 524288
Memory used in sort: 0.59 MW
This is the same point where the self-compiled (IFC 8.0) version (using the
standard Fedora Core 2 release BLAS and LAPACK libraries) failed. When the job
fails, it does not terminate. Instead, the process proceeds with 99.9% CPU
usage and %1.0 Mem usage. The process at this point cannot be stopped by the
kill command (with or without -9) and the machine must be hard-rebooted in
order to stop the process.
All of this happens on a single node running Fedora Core 2 with 2 Xeon Cpus
(athough molpro is run serially). All I know about the Disk at this point is
that it is SCSI.
Curiously, I can run the entire test set without problems using my desktop
machine which is a single P4 running Fedora Core 2. This machine runs the
preinstalled binaries, as well as self-compiled versions using IFC 8.0 and 7.0.
Can someone tell me what it is that molpro might be doing at the point where it
stops working? It sounds to me like this may be a problem with the disk, since
the lines that should follow (as given by the successful run on my desktop) are:
SORT1 READ 62160. AND WROTE 33713. INTEGRALS IN 1 RECORDS. CPU
TIME: 0.01 SEC, REAL TIME: 0.07 SEC
SORT2 READ 33713. AND WROTE 34949. INTEGRALS IN 2 RECORDS. CPU
TIME: 0.01 SEC, REAL TIME: 0.04 SEC
Is there a system log that is present on a FC2 machine that I can check to
verify the nature of the problem?
Cheers,
Seth
ccmsccmsccmsccmsccmsccmsccmsccmsccmsccmsccmsccmsccmsccmsccmsccms
Dr Seth Olsen, PhD
Postdoctoral Fellow, Computational Systems Biology Group
Centre for Computational Molecular Science
Chemistry Building,
The University of Queensland
Qld 4072, Brisbane, Australia
tel (617) 33653732
fax (617) 33654623
email: [EMAIL PROTECTED]
Web: www.ccms.uq.edu.au
ccmsccmsccmsccmsccmsccmsccmsccmsccmsccmsccmsccmsccmsccmsccmsccms
