I have an implementation of Finite Strip Method, but sometimes my computer get slow to solve some problem with a lot of degrees of freedom. Since I have a GPU, GeForce® GTX 1060 6Gb, I would like to use its cores to run my code in parallel. Does anyone know any source of information that could explain how to implement parallel pool using the GPU cores?
Hi Joao,
The MATLAB "answers" and "file exchange" are an excellent place to start for MATLAB-related questions.
I recently setup my computer with a NVIDIA GTX1070 for GPU based image analysis. It is setup in Linux and I am mainly using Python wrappers and packages to do this. What a great improvement for my workflow! I stopped using MATLAB several years ago in favour of the open-source Python.
Not sure where you are at in your setup, but: First steps are that the hardware is installed correctly and the driver is installed (the CUDA package from NVIDIA). Next, is your code formatted to run in parallel? i.e. "vectorized"? It may be possible to use built-in MATLAB commands directly. Also, consider profiling your code to identify the intensive functions and offload those to the GPU through MATLAB.
Ryan.
You can find tips on how to use your GPU and GPU implementations of many algorithms in this book: Book GPU Programming in MATLAB
Either you can directly use built-in functions enhanced to take advantage of GPU hardware or you can write your own vectorised function using 'arrayfun' or 'bsxfun' wrappers. However, using GPU in MatLab is not as easy as using parfor loop.
- Badges
- Science topic
- Similar topics
- Biological Science
- Kinesiology
- Running