MaeritTips

• work in project dir (there is a link named work)
• number of procesors to use -1
• list of all available CPUs $LL_MACHINE_LIST
• que: qsub -np 20
• qsub_debug -np 4
• comandos

cancel	Cancels a job
class
_get_machine_list
hold
q	returns information about jobs in the job queue.
status
submit
w

llclass
       class   wall clock          cpu class
        name         time         time description
---
------------ ------------ ------------ debug            00:10:00     00:10:00 application testing interactive      02:00:00     01:00:00 interactive tasks (gui, compilation, ...) large            72:00:00     72:00:00 Large 512 Tasks (256 nodes) medium           72:00:00     72:00:00 Medium 256 Tasks (128 nodes) sequential       24:00:00     24:00:00 Sequential 1 Node small            72:00:00     72:00:00 Small 128 tasks (64 nodes)

example of cmd file

#!/bin/bash

#@ job_type = parallel
#@ class = debug
#@ initialdir = /home/AB0001/u25959269/work/Ribosome/ProjMatch/test1/Iter_1/
#@ output = OUTPUT/example.$(schedd_hostname).$(jobid).$(stepid).out
#@ error = OUTPUT/example.$(schedd_hostname).$(jobid).$(stepid).err
#@ restart = no
#@ requirements = (Feature == "myrinet")
#@ total_tasks = 4
#@ blocking = unlimited
#@ queue

Define execution

Environment export MP_EUILIB=gm export OBJECT_MODE=32 export MP_RSH=ssh

#Conseguimos la Machine_list del loadleveler NPROCS=cat $LL_MACHINE_LIST | wc -l
mpirun -np $NPROCS -machinefile $LL_MACHINE_LIST  /home/AB0001/u25959269/xmipp/bin/xmipp_mpi_angular_projection_matching -i /gpfs/projects/AB0001001/Ribosome/ProjMatch/test1/noefg_head.sel -vol ../Iter_1/Iter_1_reference_volume.vol -o proj_match -sam 20 -max_shift 1000 -Ri 0 -Ro 28 -output_refs -output_classes -ang_search 1000

http://www.hlrn.de/doc_colony/loadl/commands

Page with workload

http://www.cesvima.upm.es/informacion/magerit

-- Main.RobertoMarabini - 17 Oct 2007