Philipp Klaus Krause wrote: > Ian Romanick wrote: > >> That said, there were some patches floating around *years* ago that >> exposed larger point sizes. I could have sworn that those got >> committed. Does anyone know the status of those patches? > > AFAIR these patches were by Ronald Schneidegger. The first time they > appeared he thought that he would soon find aout how to do > GL_ARB_point_sprite and antialiased points, too and wanted to make a > patch for both. Not quite. I thought I could make ARB_point_sprite and ARB_point_parameters work. However, I couldn't get ARB_point_sprite to work correctly. ARB_point_parameters is easy but the patch I've got needs some cleanup (even more so considering the recent addition of hw-implemented ARB_vertex_program, where the point size stuff currently won't quite work neither). Antialiased points are a different story, the hardware simply lacks that capability. It either needs emulation with triangles (or another primitive), negating any advantage that points have in the first place, or emulation with textured rectangles and a hacked up alpha blend / alpha test (which is the solution fglrx does, but fails pretty miserably depending on what blend function you use, you can see that with mesa's pointsize demo if you play around with the options). Considering that both solutions are quite complex and the behaviour currently implemented is fully spec compliant I never implemented either solution.
> It seems he never found the time to do so. Since many > applications like games use mostly antialiased points he thought that > the patch wouldn't be that useful. I later mentioned on dri-devel that > some applications like the free modeller wings3d really need larger > aliased point sizes to be useable. I thought a patch to support larger > aliased point sizes got committed a year ago, but I nevr tested it since > at that time I was already using my current i915. Yes, large point sizes are indeed supported since quite some time, but as said only for aliased points (and only on r200, not r100-based radeons - r100 doesn't support the point sprite primitive which is needed for large points (despite I failed to actually use it for point sprites...)). I can't tell you off-hand what Mesa version you need, but this was added quite some time ago (drm version 1.13 is needed at least). Roland _______________________________________________ Mesa3d-dev mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/mesa3d-dev
