Hello everyone,
does anything actually use the *STATUS* files which abinit creates? They used to be a problem on distributed file systems because of too frequent writing, then we reduced the frequency, but if they are not actually useful to anyone (or anything) maybe we could get rid of them altogether.
If they are purely for debugging purposes, it would be even better to have a default behavior of no status file, and then a possibility of full trace as in octopus, with a push_ and a pop_ function at the beginning and end of each subroutine (a good debugger will do the same job, of course), especially in parallel. We should scriptify the insertion of the functions - this is an eternal problem in octopus where someone is always forgetting a push or a pop and then everything breaks because the stack is too short or too long.
What do you think?
Matthieu
status files
status files
Matthieu Verstraete
University of Liege, Belgium
University of Liege, Belgium
Re: status files
I agree with Matthieu...
The proposal is to activate writing of STATUS files only in the "-enable-debug" case.
or a similar procedure...
Marc
The proposal is to activate writing of STATUS files only in the "-enable-debug" case.
or a similar procedure...
Marc
Marc Torrent
CEA - Bruyères-le-Chatel
France
CEA - Bruyères-le-Chatel
France
Re: status files
I would prefer an input variable, so you don't have to recompile, but this is incidental.
Matthieu
Matthieu
Matthieu Verstraete
University of Liege, Belgium
University of Liege, Belgium
Re: status files
Dear Matthieu, Marc,
The input variable already exists :
simply set the input variable istatr to a large value (i.e. 10000),
and forget about the STATUS file being written often...
I have no objection to have the default value of 49 to be changed, but I find that
it is a very nice tool to monitor the advancement of a job.
Perhaps, the default for parallel and sequential runs might be set to a different value...
Best regards,
Xavier
The input variable already exists :
simply set the input variable istatr to a large value (i.e. 10000),
and forget about the STATUS file being written often...
I have no objection to have the default value of 49 to be changed, but I find that
it is a very nice tool to monitor the advancement of a job.
Perhaps, the default for parallel and sequential runs might be set to a different value...
Best regards,
Xavier
Re: status files
I use the STATUS file quite frequently to monitor jobs, especially when I am running a new structure and/or new PAW sets that I am not quite sure about. I didn't know there was a variable to control its frequency, that's helpful!
Joe
Joe
Josef W. Zwanziger
Professor, Department of Chemistry
Canada Research Chair in NMR Studies of Materials
Dalhousie University
Halifax, NS B3H 4J3 Canada
jzwanzig@gmail.com
Professor, Department of Chemistry
Canada Research Chair in NMR Studies of Materials
Dalhousie University
Halifax, NS B3H 4J3 Canada
jzwanzig@gmail.com