On a related note; I made this little snippet of CUDA code, to prototype using greenlets in the place of threads, and it works great:
http://pastebin.com/mGGWn28A I imagine it would work just as well for openCL; rather than having a thread which switches at unpredictable times, greenlets yield at specified times, which allows one to overcome any such issues in a rather simple manner, while at the same time being far lower in overhead than threads. The downside is that one needs explicit statements to yield in end user code; but then again, some wise man once said there is something to be said for explicit over implicit.
_______________________________________________ PyOpenCL mailing list [email protected] http://lists.tiker.net/listinfo/pyopencl
