Luxrender gets MLT on GPU!
Re: Luxrender gets MLT on GPU!
This thread linked to is 1 year old, and was not updated since then...
I'm afraid that this is still a proof of concept without full integration into Lux :/
I'm afraid that this is still a proof of concept without full integration into Lux :/
polygonmanufaktur.de
Re: Luxrender gets MLT on GPU!
Yeah I just noticed how old that post was....
How about ERPT on GPU? Still much better than plain PT and quite MLT like
http://www.youtube.com/watch?v=c7wTaW46gzA
How about ERPT on GPU? Still much better than plain PT and quite MLT like
http://www.youtube.com/watch?v=c7wTaW46gzA
This video shows the real-time execution of my ERPT and PT implementations on the GPU. The Invisible Date scene shows the performance difference between PT and ERPT. Because all light in the room passes trough a small crack in the scene geometry, PT has a hard time finding light transport paths. ERPT does not have this problem. Whenever a light transport path is found, similar paths are easily obtained trough mutations. As a consequence, the overall variance of the ERPT rendering is much lower.
Re: Luxrender gets MLT on GPU!
Just for the case Glare is interested:
Energy Redistribution Path Tracing
Implementing Energy Redistribution Path Tracing
Some code fragments & interesting stuff (In the comments!)
Energy Redistribution Path Tracing
Implementing Energy Redistribution Path Tracing
Some code fragments & interesting stuff (In the comments!)
polygonmanufaktur.de
Re: Luxrender gets MLT on GPU!
[Citation needed]*CoolColJ wrote:How about ERPT on GPU? Still much better than plain PT and quite MLT like
* And YouTube propaganda is not a citation, see the recent "Infinite Detail / Voxels are the Future" controversy
Re: Luxrender gets MLT on GPU!
I saw the original voxel videos ages ago, how did it end up? They did look interesting...lycium wrote:[Citation needed]*CoolColJ wrote:How about ERPT on GPU? Still much better than plain PT and quite MLT like
* And YouTube propaganda is not a citation, see the recent "Infinite Detail / Voxels are the Future" controversy
Re: Luxrender gets MLT on GPU!
I don't know, but Battlefield 3 looks pretty good without voxels, don't you think? I don't see all the game developers scrambling to re-do everything with Future Magic Voxel Technology (TM), which in itself is a statement...
-
- Posts: 1828
- Joined: Mon Sep 04, 2006 3:33 pm
Re: Luxrender gets MLT on GPU!
Out of curiosity, what kind of new developments are out there in terms of those kinds of algorithms? I mean, MLT became huge and is now quite popular and I'm sure there are improvements, new algorithms, etc.
This ERPT stuff looks interesting (just from the pretty pictures and taglines) and I've heard other methods mentioned here and there (anyone remember Hungry Cat?), but I have no idea what people in that field are talking about these days.
Anything worthwhile? Anything that Glare is looking at?
This ERPT stuff looks interesting (just from the pretty pictures and taglines) and I've heard other methods mentioned here and there (anyone remember Hungry Cat?), but I have no idea what people in that field are talking about these days.
Anything worthwhile? Anything that Glare is looking at?
- Polinalkrimizei
- Posts: 647
- Joined: Sat May 02, 2009 6:59 am
Re: Luxrender gets MLT on GPU!
Also not that new, but I found it interesting:
http://www.eurogamer.net/articles/digit ... ted-detail
http://www.eurogamer.net/articles/digit ... ted-detail
Re: Luxrender gets MLT on GPU!
Well here is the guy's paper - plenty of GPU info, maybe you guys can make it work?lycium wrote:[Citation needed]*CoolColJ wrote:How about ERPT on GPU? Still much better than plain PT and quite MLT like
* And YouTube propaganda is not a citation, see the recent "Infinite Detail / Voxels are the Future" controversy
http://graphics.tudelft.nl/~dietger/thesis/thesis.pdf
The Energy Redistribution Path Tracing (ERPT) algorithm as proposed by Cline, is closely
related to the MLT algorithm from last section [11]. However, ERPT does not suffer from
startup bias. Also, ERPT does not require an accurate and unbiased estimate of the normal-
ization constant c, although it can increase performance. Furthermore, the mutation strategy
used in ERPT does not need to satisfy the ergodicity constraint.
Re: Luxrender gets MLT on GPU!
It's surely not the first time we've heard of it, and we don't rush to implement everything we read about
Wasn't ERPT supposed to save a certain "unbiased" renderer from its convergence problems, because the previous magic bullet (MLT) never happened? I think I heard some propaganda like that, don't remember the details...
I'm not questioning the work in the thesis, but I do want to point out that there are many good methods out there, and each would say they are the best thing ever
Wasn't ERPT supposed to save a certain "unbiased" renderer from its convergence problems, because the previous magic bullet (MLT) never happened? I think I heard some propaganda like that, don't remember the details...
I'm not questioning the work in the thesis, but I do want to point out that there are many good methods out there, and each would say they are the best thing ever
Re: Luxrender gets MLT on GPU!
The thing is ERPT works on a GPU, MLT is much harder to do I presumelycium wrote:It's surely not the first time we've heard of it, and we don't rush to implement everything we read about
Wasn't ERPT supposed to save a certain "unbiased" renderer from its convergence problems, because the previous magic bullet (MLT) never happened? I think I heard some propaganda like that, don't remember the details...
I'm not questioning the work in the thesis, but I do want to point out that there are many good methods out there, and each would say they are the best thing ever
And ERPT does things PT finds hard to do
Win, Win
Re: Luxrender gets MLT on GPU!
here is a video by the guy who wrote the paper comparing PT and ERPT on a GPU
The smoothness of ERPT is very nice
http://www.youtube.com/watch?v=d9X_PhFIL1o
The smoothness of ERPT is very nice
http://www.youtube.com/watch?v=d9X_PhFIL1o
Re: Luxrender gets MLT on GPU!
The point here is that I'm maybe not the only one who would like to see a new nice speed boost for Indigo, since speedups are awesome new features everybody loves!lycium wrote:It's surely not the first time we've heard of it, and we don't rush to implement everything we read about
The biggest reason why not everybody uses unbiased raytracers is speed of the engine, and not some misssing beginner Tutorials
Fast engines even can afford a unholy pricetag, like "VRED" or "Arnold Renderer"
polygonmanufaktur.de
- zeitmeister
- Posts: 2010
- Joined: Tue Apr 22, 2008 4:11 am
- Location: Limburg/Lahn, Germany
- Contact:
Luxrender gets MLT on GPU!
I know guys who earn their money with VRED. But from time to time they use Indigo, too.
So there might be potential
The main features I like on VRED:
- very fast render refresh
- very relieable and stable in network rendering
- fantastic Material Library for nearly everything (especially Automotive)
You choose a 3D model, you choose a backplate hdri, you choose the materials, AND IT almost FITS VERY CLOSE TO REALITY.
Never seen that before.
Hm.
So there might be potential
The main features I like on VRED:
- very fast render refresh
- very relieable and stable in network rendering
- fantastic Material Library for nearly everything (especially Automotive)
You choose a 3D model, you choose a backplate hdri, you choose the materials, AND IT almost FITS VERY CLOSE TO REALITY.
Never seen that before.
Hm.
Cheers, David
DAVIDGUDELIUS // 3D.PORTFOLIO
·
Indigo 4.4.15 | Indigo for C4D 4.4.13.1 | C4D R23 | Mac OS X 10.13.6 | Windows 10 Professional x64
DAVIDGUDELIUS // 3D.PORTFOLIO
·
Indigo 4.4.15 | Indigo for C4D 4.4.13.1 | C4D R23 | Mac OS X 10.13.6 | Windows 10 Professional x64
Who is online
Users browsing this forum: Majestic-12 [Bot] and 36 guests