Querying the set of active CUDA kernels on a GPU

Question

Is there a way to ask the GPU (or driver) to list the set of active (or dispatched or issued) CUDA kernels on a GPU, without attaching cuda-gdb to the owning CPU process and suspending it?

I'm imagining something like pstack, where the interface might look like:

> list-cuda-kernels $pid
gpu 0: kernel_foo
gpu 0: kernel_bar
gpu 1: kernel_baz

recent versions of nvidia-smi might do what you want on linux with a Tesla card. — talonmies
I'm using Fermi-based Tesla cards and as Eugene says, nvidia-smi only lists processes, not kernels. — Mr Fooz

Eugene Eugene · Accepted Answer · 2012-09-20T21:15:42

There is no tool or API to fetch list of the currently running kernels other then cuda-gdb (or any other CUDA debugger for that matter).

Querying the set of active CUDA kernels on a GPU

1 Answers