Reputation: 6623
I coded a program to create a color lookup table. I did it in CUDA and OpenCL, from my point of view both programs are pretty much the same, i.e. use the same amount of constant memory, global memory, same loops and branching code, etc.
I measure the running time and CUDA performed slightly better than OpenCL. My question is if using CUDA+NvidiaGPU is faster than OpenCL+NvidiaGPU because CUDA is the native way of programming such GPU?
Could you share some links to info related on this topic?
Upvotes: 4
Views: 7422
Reputation: 1638
OpenCL and CUDA are equally fast if they are tweaked correctly for the target architecture. However, tweaking may negatively impact portability.
Links:
Upvotes: 3