Workers aborted when saving surf-figures to disk using parfor

This code

parfor ii = 1:5
    surf(rand(500))
    saveas(gcf, ['test_' num2str(ii)], 'eps');
end

causes me the error

>> parfortest
Warning: A worker aborted during execution of the parfor
loop. The parfor loop will now run again on the remaining
workers. 
> In parallel_function at 589
  In parfortest at 1 
Error using parallel_function (line 597)
All workers aborted during execution of the parfor loop.

Error in parfortest (line 1)
parfor ii = 1:5

The client lost connection to lab 2. This might be due to
network problems, or the interactive communicating job
might have errored.

The error occurs only for surfaces above a certain dimension (e.g. surf(rand(5)) is fine). Other 3D-plot as trisurf or plot3 are affected as well while 2D-plots like plot or hist seem to be no problem.

Running this code with R2013b on Mac OS 10.9.3 on a Intel Core 2 Duo

Answer 1

[Edric Ellis]

Unfortunately this is likely to be caused by a known bug, the report is here: http://www.mathworks.co.uk/support/bugreports/253188 (I know this refers to 'pcolor' - it's likely that surf has the same problem). Unfortunately, the workaround is not applicable in your case because workers always run '-nodisplay'.

Note that the bug should be fixed in R2014b.

Answer 2

[Corbin Wilhelmi]

In order to avoid duplicate posts, I have run across this issue as well using 2014b which, according to Edric's link, the bug was reportedly fixed. I am running a simulation with ~140K iterations. The program will go through a couple hundred and then give a very similar error to the one above. Below is psuedocode for what I am doing. Please note that many are embedded functions and I will try to explain as simply as I can...

-------

Load various .mat files parpool(5) %Create parallel pool on 5 cores.

parfor 1:length(someVariable)

[out1 out2] = function1(in1, in2, in3)

end %End parfor loop

-------

FUNCTION 1 EXPLAINED:

figure; axis([0 1 0 1 0 1]) %Open figure and set axis for three dimensions

set(gcf,'Visible','off') %Hide figure, similar to -nosplash

variable = function2(in1) %This function loads a .mat file, does nothing else.

for i = 1:size(someVariable,2) %Size of some variable across columns

     Do these operations

end

for i = 1:size(someVariable,1) %Size of some variable across rows

     fill3(attributes) %Creates 3 dimensional solid figures

end

function3(in1, in2) %Saves the figure, nothing more

end

%End of Function1

-------

Because of using eval(['...']) commands, those have to be in separate functions. I don't know if this is just a work around but it seems to do the job for me. If anyone would like to try and duplicate this error please send me your email and I will send you the pertinent files. For completeness sake: I am running MATLAB 2014b on 64bit Ubuntu (14 I think) with hex core i7, 64gb ram.

Am I looking at a similar error as above but using fill3 instead of pcolor as stated in the error report?

Thanks in advance for your help and consideration.

Best

Corbin

[rico]

Unfortunately, I can't help with your problem. But it might be useful to know that for me, the bug was actually fixed in R2014b. Since I updated, the error does no longer occur.

[Corbin Wilhelmi]

I think it might be a new bug that is related to the one you came across, I'm emailing Edric the information and files today.

Answer 3

[Ao]

I also encountered the same error when using Matlab r2014a. However, if I chose to save the figures as .fig instead of .eps or other file types, it is indeed workable. I am still searching for a solution of how to save the pictures as the type I want.

If you have already solved the problem, please kindly let me know the solution. Thank you very much!

[Edric Ellis]

If you are able to use R2014b, the bug should be fixed there.

[Ao]

Thank you very much! I have upgraded to R2014b, and now it is workable.