07-22-2020, 03:46 PM
Thanks @philrandal and @hestehandler
I think my problem was simply a misunderstanding of how the input_cache feature would work. Whatever I was expecting aside, here's what seems to happen.
With input_cache enabled, I load a list of tracks into the queue. When I begin playing a track the mem usage begins to climb, apparently because MPD is loading in the entire track. The mem usage more or less levels off at a higher value.
Then I begin playing the next track. The mem usage begins climbing again until it more or less levels off at a higher value.
This continues as I play more tracks.
This is contrary to the situation with no input_cache. The mem usage stays more or less constant.
Since this is a 512MB RPi3A+ I'm experimenting on, I chose "128 MB" (also tried "128MB") as the cache value. Unfortunately, such ruses as mpd -V don't tell me whether the input_cache is (or will be) enabled and what the parsed value is. I haven't yet loaded enough tracks to hit the cache size limit.
Regards,
Kent
I think my problem was simply a misunderstanding of how the input_cache feature would work. Whatever I was expecting aside, here's what seems to happen.
With input_cache enabled, I load a list of tracks into the queue. When I begin playing a track the mem usage begins to climb, apparently because MPD is loading in the entire track. The mem usage more or less levels off at a higher value.
Then I begin playing the next track. The mem usage begins climbing again until it more or less levels off at a higher value.
This continues as I play more tracks.
This is contrary to the situation with no input_cache. The mem usage stays more or less constant.
Since this is a 512MB RPi3A+ I'm experimenting on, I chose "128 MB" (also tried "128MB") as the cache value. Unfortunately, such ruses as mpd -V don't tell me whether the input_cache is (or will be) enabled and what the parsed value is. I haven't yet loaded enough tracks to hit the cache size limit.
Regards,
Kent