Using parfor

Hi I successfully implemented a parfor loop in Matlab. However on my computer which has 4 cores I am not getting the improvement in computation time I expected to get. When I go to the task manager and check the CPU usage it rarely goes over 25% which is the same as before the parfor loop What could be the problem?

Answer 1

[Thomas]

What kind of problem have you set up using parfor.. a general fact about the Parallel Computing Toolbox is that the parallel tasks have to be pretty intensive/expensive in order for parallelization be effective..

If your computation is trivial then you might not get the speedup you expect.. Not a lot more can be said without looking at what code you have parallelized.

Also, is your machine quad core or a two core hyperthreaded which would show up as 4 virtual cores?

[Peter]

Hi thanks for your reply,

My processor is an Intel(R) Core(TM)2 Quad CPU Q8200 @2.33GHz.

Yes the tasks are pretty heavy.I use bintprog with a hundreds or thousands of variables that take a while to compute and other computationally expensive stuff.

I define the parfor this way on MATLAB 7.5.0(R2007b):

parfor (r=1:m_dash,4)

... end

Is that the correct way?

Answer 2

[Edric Ellis]

One thing to check - have you opened MATLABPOOL before issuing PARFOR? You need to do this in each MATLAB session to make the extra workers available. I.e.

   matlabpool open local 4