Quote:Original post by kovacsp If I have redundant render states.. then who is going to drop them? d3d? the driver? Does it worth checking with getrenderstate befor setrenderstate? Or I could have a "mirror" too.
You should drop them. D3D will unless you are running a pure device. The driver may, or may not. Your application knows what states are used, how and why so it will be much more efficient for you to not send redundent state.
If I have a mipmap chain of textures loaded. Then I use that texture. In this case, will all of the levels be loaded in the videomem, or just one (or a few) that is actually being used. In other words.. will the videomem usage increase or decrease?
Thanks for the help, Peter
------------------------------------------------------------Neo, the Matrix should be 16-byte aligned for better performance!
Currently the entire texture (all mips) must be transferre to the card at once. In DirectX Next, individual memory pages will be transferred as needed.
If a texture must be loaded into videomem because it is used, then the entire texture will be loaded, right? That means that even if I don't care about continous loading through AGP, I still need free videomem that can hold my _biggest_ texture? I mean, is it possible for the card to load just portions of the texture?
And the other, similar one: When using a cube map for environment mapping.. will all 6 faces be loaded, or only the ones that it really needs?
Thanks again, Peter
------------------------------------------------------------Neo, the Matrix should be 16-byte aligned for better performance!
Yes an entire texture must be loaded into video memory at once (as my last reply said). As for cube maps, I'm not 100% certain, but I imagine that currently all 6 faces, and all their mips must be sent to the card at once. For a definitive answer to obscure questions like these you should ask these questions on the Microsoft DirectX Mailing List where MS can give you extremely detailed answers on the inner workings of DX.
