...glTexSubImage bug?

to have a procedural skydome texture above my head, i use perlin noise and glTexSubImage2D() function.

when i update the whole texture, everything works good, but is a bit slow.

to make it faster, i use the texture as a wrappable buffer, and i update only a single line of the texture while i modify the texture matrix accordingly to make the texture scroll.

well, if i use the generic MS implementation, it works.

if i use the nvidia ICD driver (tested on a tnt and nv10) it don’t work: it seems to don’t accept lines != 0 into the offset param.

have i found another nvidia ICD bug?
what are your experiences?


glTexSubImage2D seems to be really slow??
is this the only way to modify | replace a texture???

I am searching for a best direct way to access to my texture buffers…


texture update can be done with a bunch of functions, but from tests i’ve made, subimage updates are the fastest.

glTexSubImage() is not that slow: it depends on what is the size of the texture.
for a RGB 2D fire procedural texture of 64x64 pixels is fast enough

for my skydome the texture is 512x512 alpha, and is working very fast, (bug apart)
since i update a bunch of texture lines every once.

unfortunately, opengl doesn’t support direct memory access


[This message has been edited by dmy (edited 05-03-2000).]