I'm currently working on a module for dynamic texture advection - continuous alpha blending of current image in framebuffer with noise textures. I'm using an OffScreenCanvas3D to capture the screen image and find the memory blips to be quite insignificant but the FPS to be not so good.
For screen capture, what's the difference between using an OffScreenCanvas3D and using a Raster in normal Canvas3D to achieve the same purpose ? With the Raster I can get the z-buffer values too it appears, but that aside is there any other difference in efficiency ?
I tried grep-ing the native source to find instances of glCopyTexSubImage use, but couldn't find any except for those in glext.h. Isn't that gl function more efficient than glReadPixels for screen capture (atleast according an nVidia article !) ? If so, any chance of exposing it if it ain't there already.