Version 2 (modified by raasch, 14 years ago) (diff)

--

Hints for using the SGI-Altix at HLRN

  • How to use the totalview-debugger on hicegate and bicegate

Starting from Rev 223, PALM allows using of the totalview-debugger on hice and bice. The following gives a brief instruction:

1.) Add an additional block "lcsgih parallel debug" in the mrun configuration file (.mrun.config):

%totalview true lcsgih parallel debug

Use mpt-MPI-version (mvapich should also work):

%mpilib mpt lcsgih parallel debug %compiler_name ifort lcsgih parallel debug

and include library option "-lmpi" within the %lopts settings.

2.) Copy .mrun.config into directory ~/palm/current_version on hicegate. Also copy parameter-file and other files required for the run to the respective subdirectories under ~/palm/current_version (e.g. JOBS/...).

3.) Log in on hicegate (it is essential to use "-X" as ssh-option !!) and execuute following commands (for a debug-run with 4 cores on one node):

qsub -I -X -l nodes=1:ppn=4 -l walltime=1000 -q testq

# starts a so-called interactive job

module load mpt # or other mpi-version module load netcdf module load totalview

mrun -d ....

# usual mrun-call, options as required by user, but WITHOUT option -h # values given for -X and -T options must match the qsub settings, # e.g. in this case "-X4 -T4"

After short time, the totalview-window should open (if testq is filled with other jobs, you may have to wait for a longer time).

4.) After closing the totalview window, leave the interactive job with "exit".

The following links lead to the totalview-online-documentation:

The ISV online documentation server

https://www.hlrn.de/doc/

now also serves the documentation for TotalView? 8.6.1-1 (see the Online Documentation Overview

https://www.hlrn.de/home/view/System/DocumentationOverview

for access information).

Please also follow the second link, because only this provides you with access information (username, password) for the online documentation!