Hello Phil-- You report three problems
1) > > I would have posted this to the xplor-nih email list, but I cannot seem to > access it through the xplor-nih web page. My colleague, Dr. In-Ja Byeon, > suggested I email you direct with a sort of problem I am having with > xplor-nih. > To post on the mailing list, you must belong to the mailing list. I'm CCing this message to the list to make certain that it's actually working, as I haven't seen messages there for a while myself. 2) > The first issue is that running the xplor command using input file > redirection ( < input.py ) does not seem to produce any .sa output files. > This is counter to to older instructions that I used for my original PBS job > script. > ( eg. xplor -py -smp 16 -o anneal_py.out < anneal.py or xplor -py -parallel > -o anneal_py.out -machines xplor_machines.6071BFDD51DF41BB144C4A801790DE9E < > anneal.py ) > > Getting rid of the redirection ( < ) from the command line allows the jobs > to run to completion and create the sa files. I am not sure if this is a bug > or feature that I issued in one of the upgrades/ change-logs. I just wanted > to bring this to someone’s attention. Also, even though no sa files are > produced, there is no significant difference in the time to run the job with > or without input redirection. > I generally much prefer direct specification of the script name, rather than redirection, but when I tried it just now on my machine here, things worked. However, the output file names were .xplorInput_0.sa and .xplorInput_1.sa. It's really better not to use redirection. If you find public documentation suggesting using redirection, I'd appreciate it if you bring it to my attention, so I can get it fixed. 3) > Second, and this is the more vexing one, regardless of how many > processors I specify for an SMP or parallel run, the job becomes > time limited to two processes. Specifically the first two process > that write out to anneal_py.out and anneal_py.out.1 These first two > text files are about 10 times the size of the other files and the > two processes run for about 170 seconds (on the old hardware they > run for about 240 seconds). All of the other output files are much > smaller and only run for a small fraction of the time of the first > two. below you will see the last few lines of a job I ran > requesting 24 cores. Unless you've modified the anneal.py script, it's only calculating 2 structures, so it will not scale beyond 2 cores. If you change numberOfStructures to a larger number, you should find the other cores get busier. best regards-- Charles _______________________________________________ Xplor-nih mailing list Xplor-nih@cake.cit.nih.gov http://cake.cit.nih.gov/mailman/listinfo/xplor-nih
