I have been playing a bit with massif, and a little gnuplot script. Below you see a small analyse of gmpc (on opensuse 0.11 x86_64) connected to my mpd (With 23.000 items in the playlist). First you see gmpc start, connect, fill the playlist and load the taglists. After that you see a huge rise this is a database search for aa then a, this matches 24151 songs, after that I clear the search window, releasing the memory. After that I do search for “Clapton” in the metadata browser, this is slow (gmpc reports it takes 1.5 seconds). I am wondering if I should make the search use a lazy loading model, to reduce the memory usage, or is a 10mb extra acceptable?