Subject: [linux-audio-dev] academic Q. on nice settings
From: D. Stimits (stimits_AT_idcomm.com)
Date: Wed May 15 2002 - 09:59:50 EEST
This is just an academic question. Let's say you are doing something
that processes audio and/or video data, either for display and use, or
to burn it to a cd or file, and the overall process will run for several
hours. Assume it is a single cpu box, without preemption or low latency
patches.
If the app is reniced to a low priority of 10, versus a high priority of
-1, but system load is such that the actual cpu time *on average*
remains the same, how will the timing and quality of this vary? I
imagine that there would be a statistical variation where the time spent
in any contiguous moment of working on the data would be more consistent
under higher priority, and maybe have increased variability of the
lengths of time spent away from processing this data. People often talk
about gaps in audio, maximum latency, average latency, so on, and it is
the characteristic of latency and buffer sizes/underruns/overruns that I
am curious about. Conjecture is fine, I'm just looking for any insight
someone here might have.
D. Stimits, stimits_AT_idcomm.com
This archive was generated by hypermail 2b28 : Wed May 15 2002 - 09:52:32 EEST