On Wed, Aug 01, 2007 at 11:58:27AM +1000, Darren Freeman wrote: > On Tue, 2007-07-31 at 00:00 +0300, Martin Vermeer wrote: > > ...and more importantly, you just confirmed my diagnosis of > > the problem. There is something in qt, or LyX's use of qt, > > that responds very badly to DRI both being turned on and > > practically unusable. > > Could all who are interested in this bug/thread please consider the > following "bug" which I think is related to why LyX is much slower on > figure floats: > > http://bugzilla.lyx.org/show_bug.cgi?id=4002 > > "The figure float box is roughly the full width of the window, but in > fact it is > changing a small amount with every keypress. This causes the figures to move > because they are in a centred paragraph."
This looks like a separate bug. Yes, I have noticed this too over a slow line. If the graphic is re-placed, it is re-drawn, pixel by pixel. This is unfortunate. Perhaps we should get rid of the centering on-screen. . > I'm not keeping up with what X with and without DRI are doing. With DRI, > perhaps it's > supposed to locally blit from one place to another. However, who knows how > good or bad the s/w > implementation is, it might have been faster to send it again from the source > buffer. > > Could the 16/24 bit issue be related to a format conversion between LyX/Qt4 > and the X display, which > is skipped in one case but not the other? No, my understanding is that (for some arches, like i810) hardware accel only works / is supported by X for 16 bbp. > Have fun, > Darren More fun than we bargained for ;-/ - Martin
