parfor doesn't continue to run other jobs when only one worker is working

7 views (last 30 days)
I have a parfor loop that is supposed to run 7 jobs with 4 workers. Each job executes a .exe program in their individual folders. It successfully completed 5 of the jobs. The 7th job's exe is hanging due to whatever reason, but I also noticed the 6th job wasn't even started after a long time. Why wouldn't Matlab allocate the 6th job to a worker to start it in this case?
The code is simply:
parfor i = 1:size(rotFold,1)
[~,~] = system(['cd /d ',rotFold{i},' & EPSC.exe']);
end
Thanks!
Jesse

Accepted Answer

Edric Ellis
Edric Ellis on 26 May 2020
parfor sends loop iterations to workers in groups known as "subranges" for efficiency. Also, recent versions of Parallel Computing Toolbox "eagerly" send subranges to workers slightly ahead of when the workers need them. Both of these things taken together can mean that if a single iteration of your loop hangs in this way, then other iterations might never start executing.

More Answers (0)

Categories

Find more on MATLAB Parallel Server in Help Center and File Exchange

Products


Release

R2019a

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!