However it must be clear that it is pretty much "Path+GPU+PhotonGI" Vs "BiDir+CPU+Guided sampling" (i.e. guided sampling make most sense to accelerate BIDir and if it is worth doing, it is quite debatable compared "Path+GPU+PhotonGI").
So do you mean there is no gain in speed by adding guiding method on top of PhotonGI ?
I don't think pathguiding is doable on gpu, there is also no photongi for bidir.
So I guess what Dade suggests is that here are two different ways to optimize rendering on luxcore
for gpu = pathtracing+photongi
for cpu = bidir+pathguiding which might still be inferior to pathtracing on gpu in terms of performance
I also don't think this new env.map caching will work on bidir(?)
I think the best all around solution would be focus on pathOcl+photonGI + some fast way to do caustics so that bidir becomes obsolete.
...However, to stay on topic. Direct light sampling and shadows from lights are much more worse than ecycles or octane and I hope this new envcache could help with coming up with a solution.
We can have all them nice gi and denoising if the shadows are super noisy.
lacilaci wrote: ↑Fri May 24, 2019 11:03 am
I think the best all around solution would be focus on pathOcl+photonGI + some fast way to do caustics so that bidir becomes obsolete.
I understand, that you prefer speed over accuracy, but please take into consideration, that many people uses LuxCore for optical study and experimenting, so I hope BiDir won't be disappear from LuxCore, at least not before it has some real unbiased alternative.
From experience, IMHO, BiDir Spectral engines are the future of quality renderings.
For fast GPU renderings applying Unity or Unreal is advised. With latest versions they had advanced to RT GI, better AA with hardly any visually noticeable difference between those and this latest biased GI LuxCore2.2beta. Except more issues and inconsistencies, lack of features in Lux. Then there are EEVEE, E-Cycles, CryEngine, RedShift, ...
Competing for speed makes Lux loose quality and interest along the runway which i see is slowly getting it behind all others.
kintuX wrote: ↑Sat May 25, 2019 9:20 am
From experience, IMHO, BiDir Spectral engines are the future of quality renderings.
For fast GPU renderings applying Unity or Unreal is advised. With latest versions they had advanced to RT GI, better AA with hardly any visually noticeable difference between those and this latest biased GI LuxCore2.2beta. Except more issues and inconsistencies, lack of features in Lux. Then there are EEVEE, E-Cycles, CryEngine, RedShift, ...
Competing for speed makes Lux loose quality and interest along the runway which i see is slowly getting it behind all others.
Why would you need to loose quality, have you seen the new corona version?
On demand caustics and dispersion per light and some crazy light solver that handles large amounts of lights easily and that all runs butter smooth on cpu with handful of cores only
And who would loose interest? I would really like to hear who would loose interest in luxcore if it - didn't remove, but stopped adding to it's bidir component... I really want to hear about those losess..
Dade wrote: ↑Mon May 20, 2019 7:30 pm
This method is very similar to a guided method. It "guides" only for env. light sources but it could "guides" for any kind of incoming light (direct and indirect).
Dade wrote: ↑Thu May 23, 2019 5:40 pm
However it must be clear that it is pretty much "Path+GPU+PhotonGI" Vs "BiDir+CPU+Guided sampling" (i.e. guided sampling make most sense to accelerate BIDir and if it is worth doing, it is quite debatable compared "Path+GPU+PhotonGI").
The problem with BiDir is not so much lack of speed but missing SDS paths which is contrary to the LuxCore statement:
"LuxCoreRender is built on physically based equations that model the transportation of light. This allows it to accurately capture a wide range of phenomena which most other rendering programs are simply unable to reproduce."
Would guided sampling help BiDir resolve SDS-paths from intersectible lights faster?
Path + caustic cache is no replacement because it is biased and has problems with photon distribution and resolution.
BiDirVM does not work reliably.
So there is currently no engine in LuxCore which is unbiased and "path complete" and anything helping with that is most welcome.
Dade wrote: ↑Mon May 20, 2019 7:30 pm
This method is very similar to a guided method. It "guides" only for env. light sources but it could "guides" for any kind of incoming light (direct and indirect).
Dade wrote: ↑Thu May 23, 2019 5:40 pm
However it must be clear that it is pretty much "Path+GPU+PhotonGI" Vs "BiDir+CPU+Guided sampling" (i.e. guided sampling make most sense to accelerate BIDir and if it is worth doing, it is quite debatable compared "Path+GPU+PhotonGI").
The problem with BiDir is not so much lack of speed...
I wonder what is bigger problem od bidir than the fact that you would have to spend ages to get a single frame that is pretty much indifferent to pathtracing done on gpu within minutes...
lacilaci wrote: ↑Sat May 25, 2019 12:44 pm
I wonder what is bigger problem od bidir than the fact that you would have to spend ages to get a single frame that is pretty much indifferent to pathtracing done on gpu within minutes...
The answer was there, but you elegantly left it off from the quote:
epilectrolytics wrote:The problem with BiDir is not so much lack of speed but missing SDS paths which is contrary to the LuxCore statement
lacilaci wrote: ↑Sat May 25, 2019 12:44 pm
I wonder what is bigger problem od bidir than the fact that you would have to spend ages to get a single frame that is pretty much indifferent to pathtracing done on gpu within minutes...
The answer was there, but you elegantly left it off from the quote:
epilectrolytics wrote:The problem with BiDir is not so much lack of speed but missing SDS paths which is contrary to the LuxCore statement
No, bidir is shit in terms of performance. being able to render sds isn't going to change a thing. It is always going to be inferior unless something like RTX on gpus will turn into BiDirCores on gpu...
Look, I respect, that you are a professional, who wants fast results due to productivity.
Why can't you accept, that there are other people with other viewpoints?
Simplistically, you want LuxCore to do what Cycles can do, but a little faster and/or more conveniently. For my part, I hope you will get what you want.
On the other hand I started to use Lux for those things that would be impossible to do with Cycles, and I really hope that these wont be cut off.
For me the question is: which is the better, if you can only do it slowly, or if you can't do it at all?
