I''ve just implemented Compile vertex arrays but it doesn''t seem any faster than normals arrays, at all. Specifics: Drawing individual models of 200-600 polygons each, with one at 14000. Not textured, just lit and coloured. Total scene polygons : ~25000 at 32fps. System: Athlon 700, 384mb(133mhz) ram, Geforce2. Here is how I use CVA''s:

  

m_LockArrays(0, a_Mesh.Indices.size()-1);
glDrawElements(GL_TRIANGLES, a_Mesh.Indices.size(), GL_UNSIGNED_SHORT, a_Mesh.Indices.begin());
m_UnlockArrays();

  

At first I thought I might be fill limited as there is a fair amount of overdraw, however when I reduce the size from 800*600 to 100*100 the speed change is minimal (35fps in test above). I profiled the application, by far largest function time is the rendering function. Any idea''s? Many thanks Chris Brodie http:\\fourth.flipcode.com

also 14000 is prolly to many in one hit break that in two.
also m_LockArrays(0, a_Mesh.Indices.size()-1); is wrong its not the number of indices but the number of vertices.
but most of all for the example u gave i dont expect a speed up

try this

m_LockArrays(0, numverts);
glDrawElements(GL_TRIANGLES, a_Mesh.Indices.size(), GL_UNSIGNED_SHORT, a_Mesh.Indices.begin());
change texture

glDrawElements(GL_TRIANGLES, a_Mesh.Indices.size(), GL_UNSIGNED_SHORT, a_Mesh.Indices.begin());
m_UnlockArrays();

Thanks, I think I have a better understanding of this now... Though doing state changes betwen mesh rendering will require a number of engine changes...

Back to the drawing board...

Chris Brodie
http:\\fourth.flipcode.com