cuda - how does nvidia-smi work? -
what internal operation allows nvidia-smi fetch hardware level details? tool executes when process running on gpu device , gets utilization details, name , id of process etc. possible develop such tool @ user level? how nvml related?
nvidia-smi lean wrapper around nvml. can code nvml help of sdk contained in tesla deployment kit.
everything can done nvidia-smi can queried straight c library nvml. if need query info in app of sort, it's improve (and easier) write against nvml instead of parse nvidia-smi stdout (which format changed in past).
there python , perl bindings nvml library. can utilize higher level programming langugage c.
cuda gpgpu nvidia gpu-programming
No comments:
Post a Comment