When using parallel computing toolbox can I just replace all my "for" loops with "parfor" ?

I'm running an intensive calculation and want to minimise my computation time.

Answer 1

[Walter Roberson]

No, there are a lot of restrictions on what can be inside a parfor

https://www.mathworks.com/help/distcomp/parfor.html#bvjbrbx-1

https://www.mathworks.com/help/distcomp/convert-for-loops-into-parfor-loops.html

https://www.mathworks.com/help/distcomp/nested-parfor-loops-and-for-loops.html#bvjfi19-1

https://www.mathworks.com/help/distcomp/troubleshoot-variables-in-parfor-loops.html

[matlabkid602]

How do I use my GPU on matlab?

[Walter Roberson]

https://www.mathworks.com/discovery/matlab-gpu.html

If you have a NVIDIA graphics GPU, you can probably only get decent performance for single precision, as double precision on those is either 1/32 or 1/24 of the single precision performance until you get to relatively high end.

[matlabkid602]

okay I have a 1080ti with compute capability of 6.1

[Walter Roberson]

The 1080ti has FP64 performance of about 332, which is 1/32 of the FP32 rate; https://en.wikipedia.org/wiki/GeForce_10_series . That is not bad compared to the other graphics cards, but it is far far below the FP64 performance available in some of the Tesla series; https://en.wikipedia.org/wiki/List_of_Nvidia_graphics_processing_units#Tesla or the Quadro GP100 (about 5300)

[matlabkid602]

Yes thanks for the information, the 1080ti is a gaming card.