Silver: (bug 315913 comment 11) I wonder how much the atom service costs us here... might be worth a followup patch/bug to check a) how often getCellProperties is called, and b) whether caching all the atoms in a JS object helps significantly over getAtom (it should avoid the costly XPC transition, at least). Neil: (bug 315913 comment 12) Basically, on every paint or hit-test of a cell. It would rock if someone could profile this. I'm not likely to have time for that soon, and the current speed is pretty darn good for me (certainly better than the RDF view we used to have), but if we can get even better, we definitely should.
Created attachment 339860 [details] [diff] [review] Patch Patch to cache atoms. We don't really need to set client.atomSvc now, and not doing it probably helps in never leaking crap. So I figured that'd be a bright idea. :-) http://spreadsheets.google.com/pub?key=pA2pwiKbiGdGnxkdPUUshTg Has some profiling data. This should get us at least a 30% speed gain, apparently, at a minimal cost (<1ms startup time).
Checking in mozilla/extensions/irc/xul/content/static.js; /cvsroot/mozilla/extensions/irc/xul/content/static.js,v <-- static.js new revision: 1.286; previous revision: 1.285 done