0

I was wondering if setting up realtime priority would help whisper.cpp to run faster, but not only it doesn't run any faster, it runs substantially slower.

I'm running it using sudo chrt -r 99 ... from an ssh connection. This process runs in 4 threads (the CPU has 6 cores).

Why would a process (ML inference) run slower in realtime?

Improve this question
1
  • The threads running in real-time could be starving system threads that are trying to do I/O. It's possible that you have two "free" cores, but every IO may require a chain of threads to do work. If those 4 threads run any threads themselves then those subsequent threads could combine to starve the system of time. Related: unix.stackexchange.com/questions/48519/…
    – Mokubai ♦
    Commented Oct 18, 2023 at 12:58

0

Reset to default

You must log in to answer this question.

Browse other questions tagged .