LuxCoreRender v2.3 features poll

Discussion related to the Engine functionality, implementations and API.

Select the most wanted features

CUDA/Optix 7 with RTX hardware support (aka fu@k small and/or quality-of-life features :mrgreen: )
7
3%
Extend our SIGGRAPH 2012 Disney material to include SIGGRAPH 2015 new features (https://github.com/LuxCoreRender/LuxCore/issues/91)
18
9%
Support for multiple UVs/Color/Alpha sets for each mesh vertex (https://github.com/LuxCoreRender/LuxCore/issues/134)
10
5%
Support Corona-like hybrid glass (https://github.com/LuxCoreRender/LuxCore/issues/233)
13
6%
Reduce the number of OpenCL kernel compilations required
18
9%
Triplanar mapping with blending (https://github.com/LuxCoreRender/LuxCore/issues/153)
8
4%
Add an image plugin to write over the image (https://github.com/LuxCoreRender/LuxCore/issues/45)
0
No votes
Procedural stochastic texturing (https://github.com/LuxCoreRender/LuxCore/issues/228)
4
2%
Porting of procedural Cycles/Eevee textures to LuxCore
7
3%
Micropolygon displacement
25
12%
Bevel shader
11
5%
Dirt shader
6
3%
Light linking (https://github.com/LuxCoreRender/LuxCore/issues/147)
3
1%
Coating film interference (https://github.com/LuxCoreRender/LuxCore/issues/175)
6
3%
Rayswitch (https://github.com/LuxCoreRender/LuxCore/issues/148)
6
3%
Geopattern (https://cgpress.org/archives/fstorm-tea ... ttern.html)
4
2%
New complete Bidirectional path tracing with vertex merging (BIDIRCPUVM)
11
5%
Principled hair BSDF shader and dedicated hair geometry
7
3%
Bokeh texture / Optical vignetting support (http://bertrand-benoit.com/blog/wonky-bokeh/)
3
1%
Fixed and faster env. visibility cache
3
1%
Custom color management inside Blender to match LuxCoreUI (and network renderings)
2
1%
Custom split normals support in Blender
5
2%
An option to store compressed image maps in GPU memory
3
1%
Support for 3D LUT (.cube)
1
0%
PhotonGI support for light groups
4
2%
Path guiding for Bidirectional path tracing
15
7%
Add procedural textures specifically tailored for heterogeneous volumes (for instance clouds, fog, sand clouds, fog, sedimenting heterogeneous mixtures, fruit juice, milk, turbid lakes)
7
3%
 
Total votes: 207

User avatar
Dade
Developer
Developer
Posts: 3241
Joined: Mon Dec 04, 2017 8:36 pm
Location: Italy

LuxCoreRender v2.3 features poll

Post by Dade » Thu Oct 03, 2019 1:42 pm

As discussed here, select the 6 most wanted features for you.

v2.3 will be mostly focused on small and/or quality-of-life features. It supposed to be released in few months, not a year ... but given the long list of features, as usually happen, they are likely to require a couple of versions (i.e. v2.3 - v2.4).

A special word about the very first option, CUDA/Optix 7 with RTX hardware support: this feature requires a lot of work and it is likely to polarize all our resources dedicated to core development for some months. So vote for it only if you are ready to not see very much else happen for some time.

Note: v2.3 already includes the support for the new re-worked PhotonGI caustic cache (mostly useful for rendering SDS paths).
Support LuxCoreRender project with salts and bounties

User avatar
Sharlybg
Donor
Donor
Posts: 1521
Joined: Mon Dec 04, 2017 10:11 pm
Location: Ivory Coast

Re: LuxCoreRender v2.3 features poll

Post by Sharlybg » Thu Oct 03, 2019 4:02 pm

I think the two bidir related feature can be merged. Why not ?
Support LuxCoreRender project with salts and bounties

Portfolio : https://www.behance.net/DRAVIA

User avatar
Dade
Developer
Developer
Posts: 3241
Joined: Mon Dec 04, 2017 8:36 pm
Location: Italy

Re: LuxCoreRender v2.3 features poll

Post by Dade » Thu Oct 03, 2019 4:15 pm

Sharlybg wrote:
Thu Oct 03, 2019 4:02 pm
I think the two bidir related feature can be merged. Why not ?
Because it is 2 times more time consuming to develop both than one of them and one is not related to the other (i.e. they don't overlap).
Support LuxCoreRender project with salts and bounties

User avatar
Dade
Developer
Developer
Posts: 3241
Joined: Mon Dec 04, 2017 8:36 pm
Location: Italy

Re: LuxCoreRender v2.3 features poll

Post by Dade » Sun Oct 06, 2019 4:32 pm

And the winners are:

1) Micropolygon displacement (https://github.com/LuxCoreRender/LuxCore/issues/241)

2) Extend our SIGGRAPH 2012 Disney material to include SIGGRAPH 2015 new features (https://github.com/LuxCoreRender/LuxCore/issues/91)

3) Reduce the number of OpenCL kernel compilations required (https://github.com/LuxCoreRender/LuxCore/issues/242)

4) Path guiding for Bidirectional path tracing

5) Support Corona-like hybrid glass (https://github.com/LuxCoreRender/LuxCore/issues/233)

6) New complete Bidirectional path tracing with vertex merging (BIDIRCPUVM) (https://github.com/LuxCoreRender/LuxCore/issues/243)

7) Bevel shader (https://github.com/LuxCoreRender/LuxCore/issues/132)

8) Support for multiple UVs/Color/Alpha sets for each mesh vertex (https://github.com/LuxCoreRender/LuxCore/issues/134)

9) Porting of procedural Cycles/Eevee textures to LuxCore (https://github.com/LuxCoreRender/LuxCore/issues/244)

10) CUDA/Optix 7 with RTX hardware support (aka fu@k small and/or quality-of-life features :mrgreen: ) (https://github.com/LuxCoreRender/LuxCore/issues/245)

(correct ?)

A couple of comments:

#3 can be done up to the point where recompilation is required only for editing of recursive materials/textures. The last step will be done with the materials/textures rework.

#4, nowadays I consider classic BiDir with MIS a very bad idea performance wise. Neither I consider guided ray tracing a particular well woriking idea in practice: I implemented guided path tracing in old LuxRender and results weren't good. As far as I know, neither the authors 's paper (some Corona developer) are so found of it because it is not used in Corona too.
Short version: I'm not going to work on it, if someone want, be my guest.

#7 may not be not exactly replicable as it works in Cycles but there are a couple of other options (especially after having implemented #1).

#9 I suggest to suspend this until the materials/textures rework has been done: there is the high risk of having to redo a lot of work a that point.

#10 I'm even a bit relief of the obtained quite low priority. May be I will work on it on an experimental branch, on the long run but not going to happen soon.

P.S. indeed, not all 10 will be in v2.3 or it will be require 6 months (I want to release another version before the end of the year).
Support LuxCoreRender project with salts and bounties

User avatar
B.Y.O.B.
Developer
Developer
Posts: 2997
Joined: Mon Dec 04, 2017 10:08 pm
Location: Germany
Contact:

Re: LuxCoreRender v2.3 features poll

Post by B.Y.O.B. » Sun Oct 06, 2019 4:55 pm

Dade wrote:
Sun Oct 06, 2019 4:32 pm
#9 I suggest to suspend this until the materials/textures rework has been done: there is the high risk of having to redo a lot of work a that point.
Ok.
Dade wrote:
Sun Oct 06, 2019 4:32 pm
#10 I'm even a bit relief of the obtained quite low priority. May be I will work on it on an experimental branch, on the long run but not going to happen soon.
By the way, would the features available in CUDA make it possible to have Bidir on GPU with less work than on OpenCL?
Support LuxCoreRender project with salts and bounties

User avatar
epilectrolytics
Donor
Donor
Posts: 570
Joined: Thu Oct 04, 2018 6:06 am

Re: LuxCoreRender v2.3 features poll

Post by epilectrolytics » Sun Oct 06, 2019 5:51 pm

Dade wrote:
Sun Oct 06, 2019 4:32 pm
I implemented guided path tracing in old LuxRender and results weren't good. As far as I know, neither the authors 's paper (some Corona developer) are so found of it because it is not used in Corona too.
I don't know about Corona, but according to this very recent paper the Guys from Pixar, Renderman and Weta Digital are using it in their production renderers, though it looks like not with BiDir but normal path tracing implementation.
Is there nothing in this paper worth implementing in LuxCore?
It would be an unbiased acceleration strategy welcome to many Lux users no matter if path or BiDir.
Dade wrote:
Sun Oct 06, 2019 4:32 pm
#10 I'm even a bit relief of the obtained quite low priority. May be I will work on it on an experimental branch, on the long run but not going to happen soon.
I think it's wise to wait if AMD will introduce hardware raytrace acceleration with their next generation and on which API, maybe Vulkan turns out to be the way to go.
But as OpenCL is dead and slowly rotting away, sooner or later LuxCore will need a transition to newer and faster API(s)...


Happy Birthday and all the best wishes for further LuxCore development!
:D
MBPro 15" 16GB i7-4850HQ GT750M, MacOS 10.13.6 & Win10Pro PC 16GB Ryzen 2700X, 2 x RTX 2070

User avatar
epilectrolytics
Donor
Donor
Posts: 570
Joined: Thu Oct 04, 2018 6:06 am

Re: LuxCoreRender v2.3 features poll

Post by epilectrolytics » Sun Oct 06, 2019 5:57 pm

B.Y.O.B. wrote:
Sun Oct 06, 2019 4:55 pm
By the way, would the features available in CUDA make it possible to have Bidir on GPU with less work than on OpenCL?
I guess the bad efficiency of Metropolis sampling on GPU will harm BiDir no matter if on CUDA or OpenCL.
MBPro 15" 16GB i7-4850HQ GT750M, MacOS 10.13.6 & Win10Pro PC 16GB Ryzen 2700X, 2 x RTX 2070

User avatar
Dade
Developer
Developer
Posts: 3241
Joined: Mon Dec 04, 2017 8:36 pm
Location: Italy

Re: LuxCoreRender v2.3 features poll

Post by Dade » Sun Oct 06, 2019 6:14 pm

epilectrolytics wrote:
Sun Oct 06, 2019 5:57 pm
B.Y.O.B. wrote:
Sun Oct 06, 2019 4:55 pm
By the way, would the features available in CUDA make it possible to have Bidir on GPU with less work than on OpenCL?
I guess the bad efficiency of Metropolis sampling on GPU will harm BiDir no matter if on CUDA or OpenCL.
Yes, that is one of the many reasons. About CUDA, I don't want to maintain another code base (aside C++ and OpenCL C). So there are 2 options:

1) share the CPU C++ code with CUDA;

2) share the OpenCL C code with CUDA.

The second is relatively easy to achieve with few Macros, the code is also already written with GPUs in mind so it will work well there. The upcoming re-work to reduce OpenCL kernel re-compilations will also move in that directions. Once I can compile OpenCL C code with CUDA, having to write only the CPU code to transfer data, etc. is a relatively doable task.

P.S. Cycles shares the same code between C/OpenCL/CUDA (it is written in C, not C++).
Support LuxCoreRender project with salts and bounties

User avatar
Dade
Developer
Developer
Posts: 3241
Joined: Mon Dec 04, 2017 8:36 pm
Location: Italy

Re: LuxCoreRender v2.3 features poll

Post by Dade » Sun Oct 06, 2019 6:15 pm

epilectrolytics wrote:
Sun Oct 06, 2019 5:51 pm
Happy Birthday and all the best wishes for further LuxCore development!
:D
Thanks :D I'm starting to have an age where you have a grim look at birthdays :mrgreen:
Support LuxCoreRender project with salts and bounties

User avatar
Dade
Developer
Developer
Posts: 3241
Joined: Mon Dec 04, 2017 8:36 pm
Location: Italy

Re: LuxCoreRender v2.3 features poll

Post by Dade » Sun Oct 06, 2019 6:51 pm

The LuxCoreRender project has created (with its own funds) a bounty for each requested new v2.3 feature: viewtopic.php?f=9&t=316#p17250

Note: you can contribute and push higher the priority of a feature according your needs by donating on one of the above bounties (aka you can vote with your money too).

Note: In case you are interested to hunt one of the above bounties, you are welcome but coordinate your work with Project developers or your pull request may be not accepted (at least to avoid duplicate work).
Support LuxCoreRender project with salts and bounties

Post Reply