I want to implement rendering to texture (for NV cards). But pbuffers+glCopyTexSubImage seems to be not very fast, because glCopyTexSubImage call
copies image from pbuffer to texture. Can I do it faster?

Please excuse my poor English.


Try rendering to backbuffer, and glCopyTexSubImage. No wglMakeCurrent calls then…???

Which driver are you using? I got serious slow down going from 12.00 to ELSA 12.40 on a dynamic cube map program. Will hopefully investigate tonight.


glCopyTexSubImage2D() from an RGB8 pbuffer or back buffer should be fast – when copying to an RGB8 texture. You can fall off the fast path, but if the formats match (and you’re using Rel 10 drivers), then it should be very fast.



My demo runs pretty fast (40 frames per second on gf256 sdr with 256x256 cubemap generated every frame). But I use glCopyTexImage, this function does copy pixels. Can I implement direct rendering to texture ?

u can’t render render direct to texture yet! i believe ive heard talk of an extension in the future that might supply this functionality. a cube map (where u update the same textures often ) is better served by copytexsubimage + NOT copytexImage. theres also a pdf on nvidia’s developer relations that goes a bit into all this.