MATLAB: Troubleshooting Matlab Parallel computing bottlenecks

parallel computingtroubleshooting

Hello,

can you please tell me if there is some function in Matlab or anything else, that would help me to find bottlenecks of running my code in parallel? Running my code on 1 core takes 2,5 minutes per loop, but on 2 cores it takes almost 4 minutes and running on 5 cores it is 7 minutes. In last option it makes 5 loops during those 7 minutes (because of 1 loop on each core), but in my opinion it has to be slowed by some bottleneck, because it takes really too much time compared to normal 1 core iteration. It is still faster running like this, but I was expecting it to be much better 😀

Thanks 🙂

Best Answer

Take a look at the parallel profiler.

Related Solutions

MATLAB: Troubleshooting tips

I tried it, and it doesn't complain if you provide values for all the constants and make sure that V has two elements, for example:

d1 = rand; y = rand; n2 = rand; k2 = rand; n0 = rand; d2 = rand; n3 = rand; k3 = rand; R2 = rand;
F = @(V) ...
F(rand(2,1))

MATLAB: Error code troubleshooting

Those would not be the lines that have the error messages. Those would, though, define Fdx, Fdy, and Fdz to be vectors. Then you use them as full vectors in (e.g.)

X(k) = X(k-1)+(Vx(k-1)*t)-(Fdx*t^2);

so the right hand side is going to give a vector result. A vector result cannot be stored into a single array element.

I don't know. Perhaps you want

X(k) = X(k-1)+(Vx(k-1)*t)-(Fdx(k-1)*t^2);

And perhaps you don't. You might be needing to total across all the Fdx entries or something like that (I do not happen to know the math behind the problem you are trying to solve.)

Best Answer

Related Solutions

MATLAB: Troubleshooting tips

MATLAB: Error code troubleshooting

Related Question