LuxCore v2.3 proposed features

Discussion related to the Engine functionality, implementations and API.
Post Reply
kintuX
Posts: 473
Joined: Wed Jan 10, 2018 2:37 am

Re: LuxCore v2.3 proposed features

Post by kintuX » Tue Oct 01, 2019 11:11 pm

Sharlybg wrote:
Tue Oct 01, 2019 11:46 am
but OpenCL is a dead dinosaur
As opencl is an open standard what is going to be set as a replacement worldwide not only for luxcore ?
There's Vulkan. And the "hybrid" thingie is crazy fast in RPR. No RTX needed. Like EEVEE with tracing ON :lol: seriously, no joke. They even stand ready for USD :geek:. Octane also going for it.

I think it ain't worth competing for milliseconds with the world and along loosing quality.
For example, newer Sobol is PITA to look at, doesn't clear as fast as it used to... just MHO.
So why not really working on better UX, faster flow, fixing bugs, getting that trailer out of the garage, getting along with amazing shaders, materials & tonemapper like fStormers (yes, Blender sucks in that regard and even RTX cores can't help with shading complexity)... And get things ready for long winter nights after solstice... keeping it in focus with 20/20 vision.
Why not? It's not like suddenly a mass with billions will start pouring in ;)

happyboy
Posts: 63
Joined: Sat Jun 22, 2019 5:39 am

Re: LuxCore v2.3 proposed features

Post by happyboy » Wed Oct 02, 2019 1:35 am

lacilaci wrote:
Tue Oct 01, 2019 7:22 am
The list goes on, it's just much greater flexibility.
Wow! Then that's definitely in my wishlist as well! Thank you!

happyboy
Posts: 63
Joined: Sat Jun 22, 2019 5:39 am

Re: LuxCore v2.3 proposed features

Post by happyboy » Wed Oct 02, 2019 3:34 am

B.Y.O.B. wrote:
Tue Oct 01, 2019 8:41 am
How is it exposed then? OpenGL?
All modern graphics API supports it: OpenGL, DX, Vulkan, Metal and even WebGL.
But the really annoying things that the supported formats for Desktop and mobile are almost mutually exclusive. For example, BC7 is a great format for PC, but mobiles don't support it; ASTC is great for (recent) mobiles, but PC doesn't support. For realtime applications, selecting compressed texture format is non-trivial because optimal choices depend on whether you're using alpha, whether it's used as normal mapping, whether it's HDR etc.

One way to make our life easier is to use supercompressed textures, like this:
https://github.com/binomialLLC/basis_universal

happyboy
Posts: 63
Joined: Sat Jun 22, 2019 5:39 am

Re: LuxCore v2.3 proposed features

Post by happyboy » Wed Oct 02, 2019 3:38 am

kintuX wrote:
Tue Oct 01, 2019 11:11 pm
There's Vulkan. And the "hybrid" thingie is crazy fast in RPR. No RTX needed. Like EEVEE with tracing ON :lol: seriously, no joke. They even stand ready for USD :geek:. Octane also going for it.
I'm also inclined to Vulkan (partially because I'm developing real-time 3D applcations), but as far as I know Vulkan raytracing is only implemented by NVidia so in order to support AMD cards, luxcore actually needs to have TWO vulkan backends: compute-based and VulkanRT-based?

User avatar
Dade
Developer
Developer
Posts: 3249
Joined: Mon Dec 04, 2017 8:36 pm
Location: Italy

Re: LuxCore v2.3 proposed features

Post by Dade » Wed Oct 02, 2019 8:02 am

Updated the first post list.
Support LuxCoreRender project with salts and bounties

User avatar
Sharlybg
Donor
Donor
Posts: 1525
Joined: Mon Dec 04, 2017 10:11 pm
Location: Ivory Coast

Re: LuxCore v2.3 proposed features

Post by Sharlybg » Wed Oct 02, 2019 9:39 am

Dade wrote:
Wed Oct 02, 2019 8:02 am
Updated the first post list.
New complete Bidirectional path tracing with vertex merging (BIDIRCPUVM)
If Bidir have to exist it need more love , a kind performance boost via online learning algo: guided bidir


Image
Support LuxCoreRender project with salts and bounties

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

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

Re: LuxCore v2.3 proposed features

Post by epilectrolytics » Wed Oct 02, 2019 10:54 am

Sharlybg wrote:
Wed Oct 02, 2019 9:39 am
If Bidir have to exist it need more love , a kind performance boost via online learning algo: guided bidir
I agree on path guiding for BiDir, perhaps instead of developing an online learning algorithm the already existing PGI photon cache could be used?
Edit: Thinking about it, using PGI makes no sense when BiDir already does light tracing with MIS, better to build the cache from there.

But this is a v2.4 topic IMO, the list is already way too long for v2.3.
Last edited by epilectrolytics on Wed Oct 02, 2019 2:58 pm, edited 1 time in total.
MBPro 15" 16GB i7-4850HQ GT750M, MacOS 10.13.6 & Win10Pro PC 16GB Ryzen 2700X, 2 x RTX 2070

User avatar
Sharlybg
Donor
Donor
Posts: 1525
Joined: Mon Dec 04, 2017 10:11 pm
Location: Ivory Coast

Re: LuxCore v2.3 proposed features

Post by Sharlybg » Wed Oct 02, 2019 11:14 am

I think we should now vote for features priority.
No matter the priority i think opencl compilation time is currently a serious issue compared to other renderer.
Support LuxCoreRender project with salts and bounties

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

User avatar
Dade
Developer
Developer
Posts: 3249
Joined: Mon Dec 04, 2017 8:36 pm
Location: Italy

Re: LuxCore v2.3 proposed features

Post by Dade » Wed Oct 02, 2019 2:34 pm

Updated the first post list.
Support LuxCoreRender project with salts and bounties

User avatar
Dade
Developer
Developer
Posts: 3249
Joined: Mon Dec 04, 2017 8:36 pm
Location: Italy

Re: LuxCore v2.3 proposed features

Post by Dade » Wed Oct 02, 2019 2:59 pm

It is possible to compile OpenCL C in NVIDIA PTX: https://stackoverflow.com/questions/879 ... o-ptx-code

PTX is the intermediate binary file format used for compiled CUDA GPU code. This would allow me to have to rewrite only the host code (i.e. the code running on the CPU), without touching the code running on the GPU. Than I would have to change the ray tracing related kernel to use Optix 7 in order to use RTX hardware.
Support LuxCoreRender project with salts and bounties

Post Reply