# Parallel Table and CUDA

I have a large matrix computation and I use Parallel table, which launches 4 Kernels. I do not have a GPU. Will having a GPU help me in Launching more Kernels and solve the problem faster? All my computation is Linear Algebra. I have looked at CUDALink help in mathematica and was still not sure whether Parallel Table is an inbuilt function or not?

• Yes... it will. – David G. Stork Mar 30 '17 at 23:20
• It depends on the number of steps you process per matrix. The crux is the time you will need to copy the data from the "CPU" to the "GPU" and back. That takes the most time in GPU operation. You may also use "zero copy" memory asignment to avoid copying from CPU memory to GPU. memory (direct access by GPU to CPU memory). But that I have never made it, so I can not comment on the speed of this. Btw. there are also packages available for CUDA that provide linear algebra functions for CUDA. – Eisbär Mar 31 '17 at 8:26

ParallelTable uses CPU cores so having a GPU will not speed it up. However, you can use the GPU by writing your matrix operations with the CUDA Linear Algebra functions found in the CUDALink Guide.