Statistical non-parametric tests on the GPU

Hi, I want to do statistical non-parametric tests on the GPU, has anyone done this before? I googled it but could not find any good reference.

I’m no expert of non-parametric tests but as I understand it, it is necessary to do something like 5000 permutations to get a good result, and this can take a while on a normal CPU. I’ve implemented the GLM in CUDA and can calculate the t-test value, now I want to extend this to non-parametric tests.