Strange bug, again with catalyst 12.4 and a HD 7970 and on a linux 64-bit machine.
If i run the attached programme with a local work size < 64, it throws a floating point exception. See also the backtrace, as it is. It works fine if i use the CPU device instead of the Tahiti device, or if the size >= 64.
It seems this is normally ignored in C, but in Java(JOCL) it causes an abort, for no apparently adequate reason.
I've been pulling my hair out for a couple of days with various bugs in my or others code exposed by the new driver/device, but I think this is one is at least not mine (the reason it's 42 is i was playing with a port of clsurf).
Thanks for reporting this. I'll let the QA team know. An AMD person will follow up with you by email on this.