Re: [mpg123-users] High memory usage

[Thomas O. <tho...@or...>]

Am Tue, 24 Mar 2015 13:14:57 +0200
schrieb Aristos Vasiliou <ar...@ar...>: 

> - I am using the latest version of Raspbian
> - I am working with the mpg123 binary
> - Just one instance of mpg123 is running (to my knowledge)
> - This is the command I am using: /usr/bin/mpg123 -Z -q -f 57236 
> /root/media/* &
> - I will be able to provide time data in a few hours as I don't have 
> access on the pi right now.

OK, I'll wait for the data for more judgement. So you just start mpg123
with a long list of file name arguments, to play randomly. OK, I'll
compare on my x86 box.

shell$ find /data/thorma/var/music/sampler -name '*.mp3' | wc -l
856

856 tracks should be good. 2380 22968

shell$ /usr/bin/time /usr/bin/mpg123 -Z -q -f 57236 /data/thorma/var/music/sampler/*/*.mp3
^C^C^C^C^C^C0.89user 0.16system 2:12.00elapsed 0%CPU (0avgtext+0avgdata 9536maxresident)k
8112inputs+0outputs (0major+698minor)pagefaults 0swaps

Time says 9536 K maxresident. Monitoring with top and /proc/$pid/stat
says 22968 K virtual memory (which is still a lot, but should mostly
belong to libasound) and 2380 K resident. So this should actually burn
below 3 M of RAM for you. There is some dependency on playlist length.
With 5429 tracks, the resident memory usage here is near 12 M.
Perhaps one can optimize that, as the raw playlist (the file paths) is
only about half a MiB in this case. The increase of about 9 MiB does
not look that efficient, but it is still far away from consuming all
memory of an RPi.


Alrighty then,

Thomas