Hi all, a quick question. I have a working Mathematica code that runs 8 kernels using ParallelDo. Now, to speed things up I have been trying, without success, to have each of these kernels compute the values of a function (it is the same function for all the kernels) for 1000 different values of the independent variables and return these values in an array called results. I have tried Needs["CUDALink`"] from inside and outside the ParallelDo instruction, and ParallelNeeds["CUDALink`"] from outside ParallelDo. In any case, it seems that Mathematica does not recognize any CUDA instruction. I was wondering if anyone has tried runing CUDA code from within ParallelDo. If you have or know of an example please let me know. Thanks.
Gus