To make a more “fair” comparison of gmpc vs. f.e. sonate or pympd. I did another memory profile, and let gmpc locally cache the whole playlist. (Normally gmpc only loads the songs visible).
The following profile is done on a gmpc (on a x86_64 bit) with all the plugins loaded (amazon cover fetcher, the wikipedia plugin, the lyric fetches, the artist image fetcher, the server statistics, xosd, qosd, and more), this uses around 4.5 mb of memory, then I execute a playlist search, that forces gmpc to fully load (and cache) the current playlist (of 14774 songs)
The results are still pretty good. One thing that might not show in the graphs is the time it takes from starting the program, and being able to use it (connected, and playlist is visible). Here gmpc is the absolute winner (because of the caching), sonate takes 14 seconds from starting it and being able to use it (this is on a fast 100mbit network), gmpc takes less then a second (and doesn’t slow down over wireless alot, where the other programs bottleneck on the bandwidth).
I will stop bragging now (sorry for it), but it does feel nice to see gmpc behave so well.