Good question. I've thought about this for awhile already. I think the biggest hurdle is that I need access to Radeon ATIs and not nVidia. I'm also not sure how performance portable OpenMP or OpenACC is to the Radeon chipset (if at all). If I understand properly, most of the GPU code for bitcoin and other miners is written in openCL.
You are viewing a single comment's thread from:
Well, if you can do it on NVIDIA chips, that's OK too, I think.
Fundamentally different problem programming in CUDA vs. openCL. Different hardware and capabilities as well which will affect the design of the algorithm in order to achieve optimal performance.