On Thu, May 05, 2005 at 11:17:21AM +0100, Angus Leeming wrote: > Martin Vermeer wrote:
... > >> So we did some good, just not enough? Are you able to give me any > >> pointers to the remaining sources? > > > > I doubled the blink rate (400 -> 200 msec) and traffic went from 450 to > > 900 B/s (estimated; firestarter's resolution is 0.1 kB/s). So I think it > > is really the cursor and nothing else. This is for a quiescent LyX > > containing one new, empty doc. > > Interesting. If you look at the source, you'll see that I recreate the > pixmaps only if the pixmap dimensions change. The rest is just copying the > contents of an existing pixmap to an area of another existing pixmap. Yes, that's how I read it too. > If you add some print statements you can check that the pixmaps are > created only occassionaly (the {no,v,h}cursor_pixmap_.resize(...) calls). > > If that hypothesis is correct, then the traffic is instructions to copy > one existing pixmap to an area of another existing pixmap and there's > nothing we can do about it. Plausible actually: 450 B/s at 400 ms corresponds to 200 B/blink or under 70 bytes per bitblit operation, where coordinates and height/width info must be transferred to the server. Our only hope would be a 'blink' attribute for pixels at the server, but AFAIK X doesn't do that... - Martin
pgp6Vfb1qECby.pgp
Description: PGP signature