How tp specify GPU Id in conditional breakpoints

Hi, all

My PC have 2 GPUs. The debugged program uses them both simultaneously (in different threads). I want to inspect a kernel running in only GPU #0 under certain conditions. Therefore, I should specify Card Id in the conditional breakpoint expression. But I cannot find a way to do it. Please, help!


AFAIK, you cannot get the GPU id in cuda kernel, the workaround solution I use is add one argument which contains GPU id in cuda kernel and set the GPU id in CPU side.