<mode=RANT>
As generale rule: please, I don't know how Blender works nor I have the time to google how to obtain anything out of it. When reporting a possible bug, always try to reduce the scene to the bare minimum. I don't need a pretty rendering, I need to be able to reproduce the problem. I look at numbers and it takes hours, trying to track something inside a 500MB scene and can easily become a nightmare. So post a cube with a hole when you can, not a 500MB scene.
</mode>
Sorry Dade. But i mainly posted Nakagin to make it as a free test Scene for everyone and to show new Development progress. Not only for bug tracking. anyway i understand the Big point make bug fix as efficient as possible. We should seriously care about that. We don't have infinite ressources to spent.
But Dade, a cube with hole doesn't show this problem
Don't know but my problem happen with HDRi. try.
Last edited by Sharlybg on Mon Feb 25, 2019 1:21 pm, edited 1 time in total.
Support LuxCoreRender project with salts and bounties
All jokes aside though, I keep comparing bidir vs path+pgi
And I constantly see that:
1. in more complex scenes(lots of objects, textures etc) pgi seem to give darker results than bidir (benchmark scene is one of them!)
2. in more simpler scenes, like for example a cube with a hole, pgi gives more light than bidir
lacilaci wrote: ↑Mon Feb 25, 2019 12:52 pm
1. in more complex scenes(lots of objects, textures etc) pgi seem to give darker results than bidir (benchmark scene is one of them!)
Isn't it the effect of caustics, that bidir can handle?
lacilaci wrote: ↑Mon Feb 25, 2019 12:52 pm
1. in more complex scenes(lots of objects, textures etc) pgi seem to give darker results than bidir (benchmark scene is one of them!)
Isn't it the effect of caustics, that bidir can handle?
No, scenes use clamping... Also I even made sure there are 100 path for all rays with PT while only 10+10 for bidir and nothing helps.
But in simle scenes pgi does even better job with filling the place with light... I suspect textures or materials somehow kill photons... Maybe..
EDIT: btw, like I already mentioned, setting normal angle to 90 degree will get results of PGI much closer to bidir, but bidir will still be brighter. (I'm not comparing to default reference, but bidir and/or pathtracing with 100 paths)
I have this idea.. maybe it's already working this way but I'd still ask.
Can we build photongi for all surfaces as if (glossiness threshold 0.0) for diffuse rays only?
But for glossy rays use threshold of 0.15 or 0.2... Does this make sense?
Or maybe use fresnel ior of glossy materials to blend between cached entry and pathtracing. So that reflections transitions to cache using objects IOR.
The brightness of the benchmark scene with PGI is like PT without PGI, but both darker than bidir, that's why I thought that the difference is the caustics.
lacilaci wrote: ↑Mon Feb 25, 2019 1:48 pm
I have this idea.. maybe it's already working this way but I'd still ask.
Can we build photongi for all surfaces as if (glossiness threshold 0.0) for diffuse rays only?
But for glossy rays use threshold of 0.15 or 0.2... Does this make sense?
It is possible but it will introduce more bias (i.e. difference from the reference rendering). I will check the outcome but I would not fixate too much in having every single part of the scene covered with cache entries: even a single (big) cache surface will terminate any path bouncing there. So even a single cache-enabled scene wall is like to be hit, first or later, by nearly all paths.
Just try to pick a normal scene and switch a wall from cache-enabled to cache-disabled, the difference in samples/sec will be hardly noticeable.
Support LuxCoreRender project with salts and bounties
