Very interesting indeed Tristam, thanks for that.
On Nov 6, 7:47 pm, "Tristam MacDonald" <[EMAIL PROTECTED]> wrote: > On Thu, Nov 6, 2008 at 2:33 PM, Jonathan Hartley <[EMAIL PROTECTED]>wrote: > > > > > Thanks for all the great info, people. > > > While we're on the subject, just in case I've totally misunderstood: > > > Am I right in thinking that within a single batch, I can't draw one > > set of vertices with one modelview matrix setting, and a different (or > > even the same) set of vertices with another modelview setting? > > In general, that is correct. However, there are several ways to get around > this restriction, using a technique know as 'instancing'. > > Instancing is not as widely supported by OpenGL as by DirectX, primarily > because OpenGL draw calls are very lightweight. However, if you are drawing > many models with very few triangles, instancing may provided a decent > performance boost. > > On (very) recent hardware, you have access to the GL_ARB_draw_instanced > extension, which adds full instancing to > OpenGL:http://opengl.org/registry/specs/ARB/draw_instanced.txt > > On older hardware, you can fake instancing using shader uniform variables, > as described by > NVidia:http://http.download.nvidia.com/developer/SDK/Individual_Samples/DEMO... > > Thanks for any confirmation (or otherwise) of my understanding. > > > > > On Nov 6, 2:14 pm, "Tristam MacDonald" <[EMAIL PROTECTED]> wrote: > > > On Thu, Nov 6, 2008 at 12:08 AM, josch <[EMAIL PROTECTED]> wrote: > > > > > a guy called __doc__ advised me to replace my GL_QUADS by GL_TRIANGLES > > > > in my tile engine a few days ago because of performance isues. I did > > > > that and on my "modern" intel X3100 hardware i had EXACTLY the same > > > > framerates with triangles as with quads. > > > > My X3100 seems to defy several basic rules of optimisation - I think > > > integrated hardware just has very different performance characteristics. > > On > > > my old ATI X1600, index triangle strips were definitely the fastest, > > > followed very closely by indexed triangle lists. I found the difference > > too > > > small to measure if you are pushing less than 1,000 triangles per draw > > call > > > - so it may not be worth the bother to change. --~--~---------~--~----~------------~-------~--~----~ You received this message because you are subscribed to the Google Groups "pyglet-users" group. To post to this group, send email to [email protected] To unsubscribe from this group, send email to [EMAIL PROTECTED] For more options, visit this group at http://groups.google.com/group/pyglet-users?hl=en -~----------~----~----~----~------~----~------~--~---
