Hopefully we get it soon, but we don’t want NVIDIA to rush out a buggy debugger, now do we ;) CUDA 2.0 was a little behind schedule, so maybe the debugger is too.
In other words, once you spent 2 days looking for a wrong memory access that would be obvious, and that you find it out within 30s using gdb, you just can’t live without it anymore.
Actually, profiling and debugging is a real issue with CUDA (among many others), for instance i don’t know if any of you have tried valgrind on CUBLAS… i don’t know if it is possible that we ever get the proper hooks to get a decent support of valgind without nvidia to disclose too much information External Media
Oh no, I don’t want to rush them at all. Just got triggered by another website stating there was a debugger. So when I checked the NVIDIA page, I got curious about the status :)