-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 ~ <mler...@umich.edu> wrote:
| 1) Does APBS take a lot of memory? | | APBS can certainly take a lot of memory. In addition, my plugin sets up a | grid that's a little finer than the APBS default (more grid points == | better calculation, but more memory). You can point my plugin at APBS's | version of psize.py for a grid setup that'll use less memory, and I'll add | some more options in the next version of the plugin that'll make it easier | to set up a coarser grid. Regarding this, I found, that 512 MB RAM is not enough to calculate APBS-electrostatics for a reasonably sized molecule. Small test-objects (BPTI, Peptides) work, but anything above 200 aa leads to ~ 60-80% memory consumption (includes swap, i.e. 1-1,2 GB in total) and apbs running at 1-2% CPU utilization for hours without any result. I will try to check on a machine with more memory. Is this intentionally so (i.e. is APBS getting extremely slow due to the ~ large amount of swap used), or is this some kind of bug (program hanging if machine runs out of mem)?. Are there benchmarks which give you a rough measure, what amount of memory to expect for which combination of box-size and grid-spacing? Another question: Has someone tried to compile APBS under Cygwin or any other Windows-environment (eeeeek, yes, but might be helpfull for people who don't have a local Linux-box). Or is there a chance to have some sort of remote-execution-facility (handing APBS-computation over to a remote machine, possibly one, which has access to the same filesystem)? - -- Bye, Marc Saric http://www.marcsaric.de -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.2.4 (GNU/Linux) Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org iD8DBQFA95JwvKxJUF29wRIRAru1AJ9AiyFfDYTD+zk29KTxa6aeRu4JzgCgg30p z/KKTCFUxZQinQWVitJbqA8= =sx2o -----END PGP SIGNATURE-----
