Setting default priority for matlabpool processes
10 visualizaciones (últimos 30 días)
Mostrar comentarios más antiguos
I ran my program, which has a resource heavy function call in a parfor loop, and I noticed my system started to hang frequently. I went into task manager and turned down the priority of every matlab process to alleviate the issue. I was wondering if there was a way to set the default process priority to a different (in my case lower) value. I would still like to do work while my code is running, since it will be going for at least a few hours, and if I need to run it again later, I don't want to have to set them one-by-one in taskman every time.
0 comentarios
Respuestas (3)
Matt Bigelow
el 7 de Mayo de 2018
I know this thread is old, but putting the following in a poolStartup.m script (see poolStartup - the file can also be on the MATLAB path) might be a convenient way to solve this:
if ispc
[~,~] = system(['wmic process where processid=' ...
num2str(feature('getpid')) ' call setpriority "below normal"']);
end
Using the getpid feature instead of putting in a process name ensures just your workers get a priority change and not your user session and not anyone else's MATLAB sessions or workers (if you're on a machine with multiple users and you have permission to change their process priorities).
0 comentarios
Edric Ellis
el 11 de Oct. de 2013
Editada: Edric Ellis
el 11 de Oct. de 2013
I assume you're running on windows - you could try using this file exchange submission: http://www.mathworks.com/matlabcentral/fileexchange/1841 and call it within an SPMD block.
JonasBGood
el 26 de Jun. de 2014
I stumbled upon this problem as well. For me (using win7) the system command wmic helps:
matlabpool open
cmd_str = 'wmic process where name="MATLAB.exe" CALL setpriority 64';
[~,~] = system(cmd_str);
This reduces the priority af ALL matlab instances on your local machine to 'low'. The number defines the priority:
idle: 64, below normal: 16384, normal: 32, above normal: 32768, high priority: 128, real time: 256
0 comentarios
Ver también
Categorías
Más información sobre Parallel Computing Fundamentals en Help Center y File Exchange.
Productos
Community Treasure Hunt
Find the treasures in MATLAB Central and discover how the community can help you!
Start Hunting!