Thanks It’s all per-pixel too, the vertex program is pretty much only transforming vertices, the fragment programs do all the work.
Yeah, that seam is caused by unmatched texture coords and the vertices have to be duplicated, and unfortuantly no one has been able to tell me a way around it. And I’m already welding the model when it’s loaded up (it goes away if I ignore the tex coords when welding, but that screws up the tex coords).
Originally posted by pocketmoon:
When my long awaited GFFX turns up I’ll gladly test the app, and also try CopyTex vs CopyPixels for rendering a float buffer
Ok, I’ve updated it (same link as above), so now it doesn’t require support for ARB_make_current_read.
If the extension isn’t found, the final combine (post processing, gamma correction, exposure) is done using the frame buffer instead of on the float buffer (you probably wouldn’t notice a difference… Other than the screen shots having the FPS overlay).
It should work on NV30 emulation now.
pokes nVidia Where’s my make current read?!
Edit:
BTW, the reason I really didn’t didn’t want to do it that way, is because it destroys the HDR screen shots (clamps the output to [0.0,1.0] instead of float output).
[This message has been edited by NitroGL (edited 02-06-2003).]
It’s all per-pixel too, the vertex program is pretty much only transforming vertices, the fragment programs do all the work.
