cuda-gdb: single GPU controlled by separate thread unsupported?

Accelerated Computing CUDA CUDA Programming and Performance
1

I noticed that in the known issues for cuda-gdb, it says that “Multi-GPU applications are not supported”. Does this also mean that single-GPU runs where the CUDA runtime API calls are made in a thread other than the main program thread are also unsupported?

I was trying out cuda-gdb with HOOMD and breakpoints set inside kernels wouldn’t break, while in small standalone programs, everything seemed to work fine.

2

hmmmm, unsure. I’m going to go with “probably” but I’ll double check to make sure.