Does anyone knows how much memory
uses a display list which simply render
a N triangles with N normals ?
Does it use more mem than the simple data ?
Is it stored in the RAM or is in someway
sent to a graphic card ?
Thank you Fabio
Not much of an answer but:
it might use more memory (I don’t really know;
it might be stored on board the graphics card or it might be in RAM but either way
its way faster than immediate mode especially if you include your state changes in your display lists as well. In my experience (so far) its not the number of triangles so much that drops framerates but the number of state changes. But I am still learning.
“its not the number of triangles so much that drops framerates but the number of state changes.”
is it true? how are state changes like glEnable(GL_TEXTURE_2D) made? i always thought it’s just some sort of flag, and if it were, i dont think it would be costly.
maybe someone can clarify this a bit?
Well, if you think about it, saying glDisable(GL_TEXTURE_2D) will not just “set a flag”. The graphics chip, which was set up to suck texture data from RAM and interpolate it to the frame buffer now has to stop doing that and instead set itself up to interpolate the vertex RGB colors (or whatever). This “hiccups” the “flow” of the circuitry and causes performance drops.
Another thing which takes time is changing textures. If you are not sorting your polygons by texture, you may realize a speed gain by doing so (draw all polygons of the same texture at the same time, to minimize texture changes). The benefits of doing this vary by card, CPU, cache and memory system, though.
I wasn’t thinking of textures so much as translations and rotations (and push and pop matrices). My current project dropped framerates by over 50% when I introduced a dozen such in my rendering thread. The model itself only has about 3000 triangles but is complex with a lot of moving parts.