So breath of the wild packs roughness into that blue channel. The logic here is they don't use AO, and unless a metalness map is needed, they just use two maps... albedo and normal (normal X, normal Y, roughness). Okay...
-
Pokaż ten wątek
-
This looks like this.pic.twitter.com/XOWNepsqmm
1 odpowiedź 0 podanych dalej 9 polubionychPokaż ten wątek -
So if you are using PBR, you still need an Ambient Occlusion channel. So rather than pack roughness into the normal blue channel, it makes sense to put the AO channel in there- because material variants will change the albedo, metal and roughness but never the ao.
1 odpowiedź 0 podanych dalej 12 polubionychPokaż ten wątek -
So, if we do that, we can pack roughness into the alpha of the albedo- store this as linear, and do the gamma for the albedo in the shader. That means only two texture samples for the majority of environmental maps- wood, dirt, mud, stone, leaves etc.
1 odpowiedź 0 podanych dalej 7 polubionychPokaż ten wątek -
So thats a huge improvement when you are always hitting your 16 sample limit on blended materials. However, here is where my crazy train is headed... what if we can cram metallic into that too?
2 odpowiedzi 0 podanych dalej 8 polubionychPokaż ten wątek -
PBR materials are either metallic or dielectric. The metallic map is therefore mostly black and white, with the grey pixels used for anti-aliasing and dialetric grime on metal surfaces. What if instead of using a whole channel for that, we use part of the roughness?
2 odpowiedzi 2 podane dalej 10 polubionychPokaż ten wątek -
So, say of 256 bit, 16 is enough for metal. If you remap roughness to 240 bits, you can cram metal into the 16 above that. In other words, deducting 240 and saturating the roughness channel gives you metal.
3 odpowiedzi 0 podanych dalej 6 polubionychPokaż ten wątek -
So that's just two maps. Albedo (roughness/metallic) Normal (normal x, normal y, ambient occlusion) Thoughts?
3 odpowiedzi 1 podany dalej 8 polubionychPokaż ten wątek -
W odpowiedzi do @delaneykingrox
Yeah this is something I've thought about a lot, but for the projects the savings hasn't been worth the compression trade-off. On LBP we packed normal as DXT1/5, or optional height.
1 odpowiedź 0 podanych dalej 1 polubiony -
W odpowiedzi do @calvinatorr
That's my concern. Did you play with numbers? Would the state change speed improvement be a trade off?
2 odpowiedzi 0 podanych dalej 0 polubionych
The state change you might see a tiny benefit but it really is going to depend. In general less parameters/less SRVs you need to bind on each context change the better
-
-
Dziękujemy. Twitter skorzysta z tych informacji, aby Twoja oś czasu bardziej Ci odpowiadała. CofnijCofnij
-
Wydaje się, że ładowanie zajmuje dużo czasu.
Twitter jest przeciążony lub wystąpił chwilowy problem. Spróbuj ponownie lub sprawdź status Twittera, aby uzyskać więcej informacji.