Hi, On Dec 17, 5:45 pm, Matthew Tedder <[email protected]> wrote: > Any comments/recommendations?
Instead of CUDA, I'd recommend OpenCL, as it is an open standard that NVIDIA and others support: http://www.nvidia.com/object/cuda_opencl_new.html > I am exploring whether or not it might be practical to tie in some > CUDA and forking. I don't know very much about how V8 works; but from what I've read about V8 and threads, this might not work well as you'll have only one instance of the V8 engine running at a time. But, perhaps this isn't an issue, as the threads are in CUDA's C dialect and are on the GPU and not the CPU. If you know more than I, then you probably have a much better idea about this. :) > Forking might be along the lines of this, for example: > > do together { .. } > > Whereby each statement within those braces would be forked onto its > own process. and the "do together" statement would be finished when > all of the processes within have returned. This sounds really interesting. I'm not sure how to do it though. > CUDA integration is quite different. Honestly, still working that out > in my mind. > > My purpose is for fluid dynamics processes for robotics applications. > C and C++ are efficient but development times could be greatly reduced > by employing v8. > > Matthew For some time, I've thought OpenCL bindings to V8 as done for OpenGL by V8-GL and WebGL would be awesome. For example, see V8-GL's code: https://github.com/philogb/v8-gl I've used V8-GL to just play around with primatives, and I think it's fun to work with. I know that OpenCL is shown on Wikipedia as a competitor to CUDA, but I don't know enough about either to compare them. However, CUDA seems to be owned by NVIDIA while OpenCL is an open standard (that NVIDIA helps develop). If CUDA's what you want, then I think it would probably be a great choice. I hope this gives you some ideas, pikpik -- v8-users mailing list [email protected] http://groups.google.com/group/v8-users
