Hi
I am running Amber8 on IBM machine.
The spec of the machine is
OS: SuSE Linux Enterprise Server 8 for ia32/kernel 2.4.21-138
File system: ext3, GPFS for Linux 2.3
Scheduler: Load Leveler 3.2.1
Compiler: IBM XLC 7.0, XLF 9.1
Parallel Lib. : MPICH_GM 1.2.6..14a
CPU: PowerPC970, 2.2 GHz 2-way
Memory: 2G
Network link: 1 Myrinet/1 Gb Ethernet
Number of nodes: 480
I succesfully compiled and have no problem running AMBER8 on this machine (well.. most of times). But when I run sander (REM, to be more specific) with large number of nodes (say 128 nodes), some of the nodes (usually 1 or 2 nodes) crash during the computation. The crash pattern is not consistant and sometime there is no crach and I can finish the job. ( I use 32 replicas which means each replica runs with parallel) The admin says the machine looks perfectly OK and he suspects there must be some subtle issues on Amber8 on this machine. Admin says there is no crach log. ( And there is no crash log from Amber8 side either.) The number of protein residues I am running is about 75 with GBSA. The admin suspect there could be some memory leak or something.
I know very few place use this machine set up but any of your experience on this matter will be highly appreciated.
Thank you very much.
Soo
-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber.scripps.edu
To unsubscribe, send "unsubscribe amber" to majordomo.scripps.edu
Received on Mon Aug 01 2005 - 03:53:00 PDT
