Is it possible to assign different kernels to different multiprocessors (on the same graphic card) and run different kernel functions simultaneously with CUDA 2.3? I am looking for a way to run slightly different applications accessing the same data reside on the global data at the same time. Searching through the forum, I found a post around Oct, 2007 which indicating this is not possible - http://forums.nvidia.com/lofiversion/index.php?t47547.html
I would like to know, is this possible now with CUDA 2.3? If so, how to do it?