Is it possible to locally distribute an array for GPU processing?

I am running into issues with larger gpuArrays due to insufficient memory on my GPU. My idea was to partition the array as one would do with parallel computing. Unfortunately, I have not found a way to do this. My next idea would be to implement algorithms in a way that manually splits arrays (or minimizes memory usage).

[Matt J]

Not clear how partitioning would help you, unless you have multiple GPUs that you're trying to partition across. But anyway, no, I don't believe there is a partitioning mechanism for gpuArray.

Answer 1

[Joss Knight]

If you have multiple GPUs you can perform independent computation on each one using a parallel pool, see here for instance: http://blogs.mathworks.com/loren/2013/06/24/running-monte-carlo-simulations-on-multiple-gpus/