TES Skyrim SE 0.310

[Oyama]

Boris

Can we use ' ENBParams01 ' in the bloom shader as we used " BloomParameters " in oldrim version ?
Or, in any case, are still " BloomParameters " and " TempParameters " valid as additional parameters in bloom or lens shader ?

Don't know if I'm clear here...

[Marty McFly]

Afaik EBloomAmount was never usable in enbbloom.fx. And as there are no ini controls over contrast and blue shirt (and no respective code in the example) there's no use for such data in enbbloom.fx. Temp parameters were only there to control bloom pass 1 and 2 which run multiple times even though the technique doesn't say so.

[ENBSeries]

Oyama
Yes, they are about same thing, just ENBParams contains more data in one variable.

Pfuscher
I can't get specular amount from anywhere, except alpha channel of normal or diffuse textures, which not fit to specular amount well. I don't know, this feature is useless on practice, like a wax.

[Oyama]

Marty and Boris
Thank you both

[Marty McFly]

Boris, is mipmapping supported for the textures in enbeffectprepass.fx and if not, can you enable it?

[ENBSeries]

No, mipmapping unsupported. But i don't know why you need mipmaps, drawing them complicated task for preset makers, by performance is same as to use available render targets. It can be useful only at read time to increase performance a little. If you have some idea when they really good and can't be "emulated", tell.

[Marty McFly]

Well, until your AO/IL solution comes to SSE, I'm porting my AO/IL to enbeffectprepass.fx. In fact I already did, it's just the one piece that makes it so fast: instead of having a fixed input texture size, it uses higher mipmaps for more distant samples, both depth and color. It's an idea I got from the Alchemy AO, they do the same. It's responsible for 99% of the performance of the algorithm, my AO on ReShade is very very fast and performance is not dependant on sample radius. But without the mipmap thing, using IL is quite the resource hog. If you now suspect that quality of the AO degrades when doing that, visual differences are almost not existant. I know you want to add your AO but in the meantime, this'll be a good alternative. Right now almost every preset author uses ReShade on top of ENB just for my AO and that's obviously not a viable solution.
If you don't believe me, try ReShade + my AO out on any game.

[ENBSeries]

Why i should not believe you, if precision parameter is lod factor of mipmap texture in my code? But it do have quality impact, you just don't know where to look at. Instead of implementing mipmap for this feature, maybe you should try to make atlas from same render target and read data from it. Or even unmap image to wavelet compression and then it will be faster to read two lods at once.

[Marty McFly]

I know that generating mipmaps costs fps. I know you think I'm stupid and from your point of view this may even be right. Interesting, didn't know that your AO works that way. I assume sampling quality is bit depth of render target?
On another note, wouldn't atlas still mean that the AO samples from the fullscreen texture? And this wavelet compression, how does that work?

Can't you just add that? I mean, if not specified, mipmaps won't be generated anyways so it'll change nothing for other users but for projects that need it, it'll be superb. I see that adding temporary render targets and different techniques gives a lot of freedom already, would be nice to give shader authors more freedom, as you said yourself, the SSE ENB won't come close to old Skyrim ENB. See what I did on GTASA with the projection matrices and all that stuff.

[ENBSeries]

It's great start to say that i think you are stupid...
SamplingQuality is amount of samples read in cycle.
You don't need full screen resolution for reading AO/IL data, it's wasting of performance. Run visualisation mode of SSAO/SSIL in old skyrim version of the mod and modify texture scaling from 1.0 till 0.75, you will have big problem to find difference. Effect is low frequency, blur prepass is better for it smoothness to remove all kind of aliasing, so resolution of input textures can be lower easily. So you can use extra space from render targets to draw other useful data.
https://upload.wikimedia.org/wikipedia/ ... nstein.png This is wavelet compression of image, it computes difference between nearest pixels compared to middle blurred (which is equal to mipmap lod). It do have compression artifacts, but they are minor for AO computation.
I don't know yet how to add mipmaps to render targets, at least it's better that you asked for automatic generation of them, so some hint in technique can do that. But how to create mipmapped targets, on the fly if some shader contains it? Need to think about it.