======How To Run Magma on the Cluster======

**Magma** is accessed by first loading the module

  module load chpc/math/magma/2.26

and then simply executing the command

  magma

However, Magma is licensed software and only licensed to run on the large memory nodes (usually called the "fat" nodes).

You will thus need to request a fat node using ''qsub'' for an interactive session //or// by submitting a job script.

=====Interactive Session=====

To request a login shell on a fat node use

  qsub -I -P PROJECTID -q bigmem -W group_list=bigmemq  -l select=1:ncpus=4:mpiprocs=4
  
where you must replace ''PROJECTID'' by //your// project name.

If you receive the error message:

  qsub: Bad GID for job execution

then you do not currently have access to the ''bigmem'' queue and must ask your PI to request that access through our Helpdesk.

Otherwise you should see a message similar to

  qsub: waiting for job 4068905.sched01 to start

Note: the job ID number ("4068905" in the above example) will be different for you.

Depending on how busy the ''bigmem'' queue is you could wait several minutes (or hours) for the session to start — which is why it is preferable to use a job script.

When the interactive session starts you will see that your command line prompt changes from ''login1'' to ''fat01'' (or other ''fat0''N).  Now you can load the module as above:

  module load chpc/math/magma/2.26

and start Magma with the command

  magma

After you have finished running Magma and exited back to the command shell (the ''fat0'' prompt) simply use the Ctrl-D key combination to end the interactive session.

====Example interactive session:====

<code bash>
fat04:~$ module load chpc/math/magma/2.26
fat04:~$ magma
Magma V2.26-10    Fri Mar 18 2022 09:59:31 on fat04    [Seed = 3853761563]
Type ? for help.  Type <Ctrl>-D to quit.
> 

Total time: 0.399 seconds, Total memory usage: 32.09MB
</code>

=====Job Script=====

Please see the [[quick:start|CHPC Quick Start Guide]] and [[guide:pbspro|PBSPro Scheduler Guide]] for information on job scripts and the scheduler used on the CHPC cluster.
 
===Video===

This video demonstrates job scripts for the ''bigmem'' queue and fat nodes:

  * [[https://www.youtube.com/watch?v=ZZc1Iic6bDc| Advanced Jobscripts Demo]]

====Example Job Script====

This example job script requests 8 cores on a fat node, and a maximum wall time (run time) of 1 hour and 30 minutes.

<code bash>
#!/bin/bash
#PBS -N magma-example
#PBS -l select=1:ncpus=8:mpiprocs=8
#PBS -l walltime=1:30:00
#PBS -P PROJECTID
#PBS -q bigmem
#PBS -W group_list=bigmemq


CWD=/mnt/lustre/users/$USER/workingdirectory
cd $CWD

module load  chpc/math/magma/2.26

nproc=`cat $PBS_NODEFILE | wc -l`

magma MAGMAPROGTORUN
</code>

**Note:**

  * Replace ''PROJID'' with //your// project name.
  * Always use a working sub-directory in your Lustre directory.
  * Replace ''workingdirectory'' with the path of your working directory on Lustre.
  * Replace ''MAGMAPROGTORUN'' with the appropriate options and arguments to the magma application.

Save the file, for example as ''magma1.pbs'' and then submit to the scheduler with

  qsub magma1.pbs

You can check the status of the queue with

  qstat -u $USER

which should display something like:

  Job id            Name             User              Time Use S Queue
  ----------------  ---------------- ----------------  -------- - -----  
  4088494.sched01   magma-example    username                 0 Q bigmem

The ''Q'' under ''S'' (for //State//) indicates that this job is queued and waiting to run.  A running job would look like:

  Job id            Name             User              Time Use S Queue
  ----------------  ---------------- ----------------  -------- - -----
  4088497.sched01   magma-example    username          00:38:21 R bigmem

with ''R'' indicating running state.

> ''username'' will be replaced by your cluster account user name.