Lecture by ptexArticle, This article is between writing notes and not writing, but there are some highlights (oneCodeA copy of data) is good.
An article about NV before ptex has been introduced. In short, texture is broken into a large texture and used in Atlas, which can greatly save space:
Mipmap computing code:
This is a record, but let's try again:
Efficiency
In actual use of ptex, it is much more complicated to read textures than to read textures directly using tex2d. In this article, we will also show the mipmap to a texture. Therefore, we need to solve the problem of mipmap and trilinear filter by ourselves, anisotropic filter, Handling Inconsistent resolution issues and so on.
As a result, the data in the text is normal. Texture sample is 1/6 faster than ptex, but the author says there is still a lot of room for optimization (alas, can you not optimize the data ?)
Then memory consumption is an advantage, and we also saw % 7 waste on % 37 (% 37 should be a relatively general or relatively poor art level, there will be no such big gap in reality, but the advantage is definitely there ).