[esp-r] Re: problem with parallel implementation including CFD

Andrew Cowie cn06arc at leeds.ac.uk
Sun Feb 3 20:59:28 GMT 2013


Aizaz,

Interesting, I seem to have solved the problem by a somewhat cheeky hack. I have added an extra command line option/argument pair to the bps module that is appended onto the temp_DFS file if given; I simply use the parallel worker index as the argument. Seems to have done the trick, if I still encounter problems though I will give your solution a try.

Cheers,
Andy

----- Reply message -----
From: "Aizaz Samuel" <aizaz.a.samuel at strath.ac.uk>
To: "esp-r at lists.strath.ac.uk" <esp-r at lists.strath.ac.uk>
Subject: [esp-r] Re: problem with parallel implementation including CFD
Date: Sun, Feb 3, 2013 19:53



Andrew,

I had a similar problem when trying to run multiple instances on an HPC. The models did not have CFD active though. Don't know how applicable that would be to your case but I overcame the problem by running the different instances from different directories (which were copies of each other).

Regards,
Aizaz

Dr Aizaz Samuel
Energy Systems Research Unit
Department of Mechanical and Aerospace Engineering
University of Strathclyde, Temporary Address:
University Centre (above Sports Centre)
347 Cathedral Street
GLASGOW, UK
G1 2TB
e: samuel at esru.strath.ac.uk<mailto:samuel at esru.strath.ac.uk>
t: +44 141 548 5765
f: +44 141 552 5105
w: www.strath.ac.uk/esru<http://www.strath.ac.uk/esru>

[cid:part3.03010902.00050309 at strath.ac.uk]



On 01/02/13 17:45, Andrew Cowie wrote:

Hi all,

I'm running multiple instances of ESP-r in parallel on a HPC cluster. When CFD is included in the analyses it always crashes with a Fortran runtime error: end of file (always at line 152 of cfutil.F).  I'm in the process of tracing the error now, i just wondered if anyone more familiar with the workings of the DFS module knows of a reason why this might be happening, for example a temporary file that might be being written to at the same time by the multiple parallel instances of ESP-r.  It only seems to happen when it is run in parallel, and does not seem to happen every time, which is what leads me to believe that this may be the problem.

Cheers,
Andy Cowie

_______________________________________________
esp-r mailing list
esp-r at lists.strath.ac.uk<mailto:esp-r at lists.strath.ac.uk>
http://lists.strath.ac.uk/mailman/listinfo/esp-r


-------------- next part --------------
A non-text attachment was scrubbed...
Name: THE_Uni_of_the_Year_Logo.jpg
Type: image/jpeg
Size: 46666 bytes
Desc: THE_Uni_of_the_Year_Logo.jpg
Url : http://lists.strath.ac.uk/archives/esp-r/attachments/20130203/4d7609cd/attachment-0001.jpg 


More information about the esp-r mailing list