Execute several statements in parallel without parfor

Hi All!

I can't seem to find a way to tell Matlab to execute multiple lines of code in parallel that I don't necessarily want to put into a parfor loop. Is there a way to do this? I did look around to try to find this, but maybe I'm just not searching with the right keywords...

I do have the parallel computing toolbox by the way.

Thanks! -Mike

Here's a code example:

% setup variables

% Now want to execute the following lines in parallel:
% Fresnel sine and cosine integrals
  ca1 = mfun('FresnelC', alpha1);
  sa1 = mfun('FresnelS', alpha1);
  ca2 = mfun('FresnelC', alpha2);
  sa2 = mfun('FresnelS', alpha2);
  cb1 = mfun('FresnelC', beta1);
  sb1 = mfun('FresnelS', beta1);
  cb2 = mfun('FresnelC', beta2);
  sb2 = mfun('FresnelS', beta2);

% Now end parallel execution, and use computed values in other calculations

Answer 1

[Yoav Livneh]

You can perform several tasks in parallel using spmd (Single Program Multiple Data). For example:

% Gather all inputs into one variable
Input = [alpha1; alpha1; alpha2; alpha2; beta1; beta1; beta2; beta2];
Method = {'FresnelC'; FresnelS'; 'FresnelC'; FresnelS'; 'FresnelC'; FresnelS'; 'FresnelC'; FresnelS'};
% Open matlabpool. Requires 8 workers
matlabpool open 8
% Perform all 8 calculation together
spmd
  %%%mfun cannot be a nested function, has to be an m file %%%
  Results = mfun(Method{labindex}, Input(labindex));
end
% Export results. Assume mun output is a scalar
Results = [Results{:}];
matlabpool close

spmd has some limitations, you should read the help file on it. The main idea is that you run your code on every wirker once. So if you have less than 8 workers you won't be able to run all 8 lines at the same time. Try experimenting with it, to see how it works. Good luck!

Answer 2

[Konrad Malkowski]

Another options of executing code in parallel, without using parfor is to use the distributed jobs API.

sched = findResource('scheduler','type','local');
job = createJob(sched);
input = [alpha1; alpha1; alpha2; alpha2; beta1; beta1; beta2; beta2];
method = {'FresnelC'; 'FresnelS'; 'FresnelC'; 'FresnelS'; 'FresnelC'; 'FresnelS'; 'FresnelC'; 'FresnelS'};
for index = 1:numel(input)
   task = createTask(job, @mfun, 1, {input(index), method{index}})
end
submit(job)
wait(job)
answers = job.getAllOutputArguments()