On Thu, Aug 02, 2007 at 11:04:41AM +0200, Helge Hafting wrote: > Andre Poenitz wrote: > >On Tue, Jul 31, 2007 at 12:28:31PM +0200, Helge Hafting wrote: > >>[...] > >>Clearly, qt4 think it is clever to use DRI for some things when it > >>is available. It is probably faster than "standard X" when hardware > >>is used for DRI, but much slower when the software fallback is used. > >> > > > >Where in the Qt sources do these "clever things" happen? Grepping for > >"dri" does not yield anything interesting. > > I have no idea.
Yet you produce random accusations. > I arrived to the conclusion this way: > * Software DRI is known to be hopelessly slow, it exists mainly so > you can test the APIs without actual hardware. Or run some > low-impact stuff. Right. > * People observe that something in LyX is dead slow when > software DRI is in use, and speedier when it isn't. So the > slow software DRI obviously got used by running LyX. Or some funny interaction wich is triggered by what we do but nothing we are 'responsible' for. I sometimes start playing a DVD with ogle while running the flash plugin in Firefox. It usually works after retrying, and so far it always worked after ejecting and re-inserting the DVD. Now whom do I blame? Ogle? Firefox? Flash? Linux? Qt? LyX? I guess just blame LyX. > * As far as I know, the only way LyX uses the screen, is through > qt4. So qt4 must be the one to blame then. qt4 has support for > openGL it wouldn't be strange if they try to speed up some > operations by using DRI where they think it is noticeably faster. It would be quite strange. LyX only uses Qt widgets that are _not_ derived from QGLWidget. QGLWidget is the only thing in Qt that ever accesses OpenGL. You'll also notice that LyX is _not_ linked against libGL, which would be the case if we used the QtOpenGL module by accident. Andre'