On Thu, Jul 15, 2004 at 05:45:36PM +0200, Denis Oliver Kropp wrote: > Quoting Ville Syrj�l�: > > On Thu, Jul 15, 2004 at 11:43:28AM +0100, Torgeir Veimo wrote: > > > On Thu, 2004-07-15 at 13:39 +0300, Ville Syrj�l� wrote: > > > > On Sat, Jun 26, 2004 at 01:17:54AM +0300, Ville Syrj�l� wrote: > > > > > I have some mga driver stuff pending (I ripped some fixed dmatmp stuff > > > > > from Mesa cvs) so I think I should update the tarball. > > > > > > > > Ok I just uploaded my latest stuff. I'm pretty sure the code was working > > > > correctly before but now I'm getting segfaults in mallopt() in libc when I > > > > quit the DirectFBGL examples. On the other hand Mesa demos via freeglut > > > > don't segfault. Quake3 hangs when I quit it or try to change the "system" > > > > options. > > > > > > So you actually run quake3 successfully under directfb with hardware > > > acceleration? > > > > Yes. Dok came up with the hack originally. It creates a new surface for > > the OpenGL app which is quite ok for fullscreen DGA apps but it's really > > bad for non-DGA apps since input is handled by the original X window and > > the graphics are drawn in the other surface. ThinkTanks is one such app. > > Nevertheless I've played it and other games. I'm pretty sure I got ut2k3 > > to work also but that was through SDL. > > Cool, do you have some benchmarks with ut2k3 running on Windows/GL, XFree/GL > and DirectFBGL?
No. It was hitting a lot of sw fallbacks. The only way I could play it was by hacking around them in the driver. And even then it was quite slow (especailly the outdoor maps). Projective texturing was causing a lot of sw fallback and the rest came from the texcombine extensions. There is an option in the came which should turn off ptex but even with that option ptex was still used it in some places :( -- Ville Syrj�l� [EMAIL PROTECTED] http://www.sci.fi/~syrjala/
