Hi,all.
As I read the <Real time rendering,3rd>,I get two confuse that: (and I assume the API is DX9)
1.While the hardware can generate mipmaps automatically,do I have the motivation to pre-generated it by DxTex or PhotoShop's plugin?
I heared that the auto way doesn't have a good visual and sometimes may cause some bad effects(and even maybe effect the performance?--I don't clear.).As opposed to,the pre-way can get best visual,but also suffer more memory consumed.So,what's the most recommended way is used nowadays?
2.In the RTR.3rd,it said the Unconstrained Anisotropic Filtering can replace the mipmap,and takes less memory and has better visual.Though it's a further improve way as it declares.So,I want to ask is it now be the mainstream?If not,is it because of the hardware capacity?
Thanks very much!
Harry
Some confuse about the mip-map and unconstrained anisotropic filtering?
You could easily pre-generate your textures using the DirectX Texture Tool. It creates mip-maps automatically as dx would.
There are pro and cons on pre-generating it
+less loading time(i think so, but I am not sure)
-more memory usage on the hard drive
I hope that helps you!
There are pro and cons on pre-generating it
+less loading time(i think so, but I am not sure)
-more memory usage on the hard drive
I hope that helps you!
You could easily pre-generate your textures using the DirectX Texture Tool. It creates mip-maps automatically as dx would.
There are pro and cons on pre-generating it
+less loading time(i think so, but I am not sure)
-more memory usage on the hard drive
Yep,it's helpful to me.Thanks!
And more if you can tell me is:
We can use the photoshop's powerful plugin to pre-generate a more perfect mipmap,so if compare to this approach,which way we usually better choose?And maybe the motivation I'm eager to get the answer is what's the mainstream approach this industry now using.
Automatic mipmap generation is really intended for GPU-writeable textures such as render targets, and not static textures. If you can, you probably want to generate the mipmaps yourself. If you do it yourself you can ensure proper handling of things like sRGB conversion, normal maps, and alpha-tested textures. You also can't use auto-generated mipmaps for compressed texture formats, which is a major downside.
Anisotropic filtering doesn't replace mipmapping, it works alongside it. The basic idea is that you retrieve prefiltered texels when you can, and take additional samples for high degrees of anisotropy. Technically you could eliminate the need for mipmaps by taking LOTS of samples to cover the entire sampled region, but this would be really expensive without much benefit.
Anisotropic filtering doesn't replace mipmapping, it works alongside it. The basic idea is that you retrieve prefiltered texels when you can, and take additional samples for high degrees of anisotropy. Technically you could eliminate the need for mipmaps by taking LOTS of samples to cover the entire sampled region, but this would be really expensive without much benefit.
Automatic mipmap generation is really intended for GPU-writeable textures such as render targets, and not static textures. If you can, you probably want to generate the mipmaps yourself. If you do it yourself you can ensure proper handling of things like sRGB conversion, normal maps, and alpha-tested textures. You also can't use auto-generated mipmaps for compressed texture formats, which is a major downside. Anisotropic filtering doesn't replace mipmapping, it works alongside it. The basic idea is that you retrieve prefiltered texels when you can, and take additional samples for high degrees of anisotropy. Technically you could eliminate the need for mipmaps by taking LOTS of samples to cover the entire sampled region, but this would be really expensive without much benefit.
Thanks very much!You make me a deeply understood about this topic.And I'm clear now.
This topic is closed to new replies.
Advertisement
Popular Topics
Advertisement