...
- At the moment there are no time limits for the idle-cycle jobs
- Different codes work on different systems. Contact cpo for details. The exception is dacapo/jacapo, which does not suspend correctly
- We are using other people's computing resources, so it is important to try not to do things like use up so much memory that the machines will crash
- A sample command for running espresso on oak:
Code Block source /nfs/slac/g/suncatfs/sw/espv10/setupenv /usr/local/bin/bsub -a mympi -n 2 -q idle -R "select[oak] span[ptile=4]" -o junk.log ./H.py
Cluster | Submit Command | Queue Name(s) | Cores per Node | Memory per Node | Interconnect | Current User(s) |
---|---|---|---|---|---|---|
pinto | gpaw-pinto-ver-bsub | idle | 8 | 48GB | Infiniband | montoyjh/chuan |
oak | gpaw-oak-ver-bsub | idle | 4 | 16GB | 1Gb ethernet | montoyjh |
fell | esp-fell-ver-bsub | idle | 8 | 16GB | 1Gb ethernet | kingsley |
lcls | esp-lcls-ver-bsub | psfehidle/psnehidle | 12 | 24GB | Infiniband | montoyjh/leannedc |
clustered systems | standard commands | psanacs-idleq | 16 | 32GB | 1Gb ethernet | ctsai89 |