Cycles: path guiding #92571

New Issue

Status

Required for Blender 3.4

Release notes
Manual documentation
Update to latest OpenPGL release
Investigate volume direct light correctness, see TODO in shade_volume.h
Tests
- Enable once deterministic sampling is added
- More complete regression .blends

Other

Closure support
- Diffuse BSDF
- Translucent BSDF
- Specular BSDF
- BSSRDF
- Volume Phases
Performance
- Reduce deterministic training overhead?
- Learn guiding vs. BSDF sampling probability?
MNEE compatibility
GPU support

Background

For more efficient rendering of difficult light setups, including caustics to some extent.
https://cgg.mff.cuni.cz/~jirka/path-guiding-in-production/2019/index.htm

Based on integrating Intel Open Path Guiding Library:
https://github.com/OpenPathGuidingLibrary/openpgl

**Status** Required for Blender 3.4 - [x] Release notes - [x] Manual documentation - [x] Update to latest OpenPGL release - [x] Investigate volume direct light correctness, see TODO in `shade_volume.h` - [ ] Tests - [ ] Enable once deterministic sampling is added - [ ] More complete regression .blends Other - [ ] Closure support - [x] Diffuse BSDF - [ ] Translucent BSDF - [x] Specular BSDF - [ ] BSSRDF - [x] Volume Phases - [ ] Performance - [ ] Reduce deterministic training overhead? - [ ] Learn guiding vs. BSDF sampling probability? - [ ] MNEE compatibility - [ ] GPU support **Background** For more efficient rendering of difficult light setups, including caustics to some extent. https://cgg.mff.cuni.cz/~jirka/path-guiding-in-production/2019/index.htm Based on integrating Intel Open Path Guiding Library: https://github.com/OpenPathGuidingLibrary/openpgl

Changed status from 'Needs Triage' to: 'Confirmed'

Added subscriber: @brecht

Added subscriber: @Alaska

Added subscriber: @SteffenD

Added subscriber: @Emi_Martinez

Added subscriber: @Eary

Added subscriber: @GeorgiaPacific

Added subscriber: @Roggii-4

Added subscriber: @sbchild

Added subscriber: @lsscpp

Added subscriber: @DiegoCortes

A question, what do you think of this: https://graphics.cg.uni-saarland.de/publications/rath-2020-siggraph-guiding.html
I talked about this on the devtalk thread some time ago: https://devtalk.blender.org/t/cycles-x-feedback/18598/405
It is said that it only adds three line of code to the Müller paper and it improves the result so much.
I am not sure how true this "three lines of code" statement is, but if it's really that easy to add on top of Müller et al. maybe this is something that can be considered?

But if this is actually not the case then nevermind.

A question, what do you think of this: https://graphics.cg.uni-saarland.de/publications/rath-2020-siggraph-guiding.html I talked about this on the devtalk thread some time ago: https://devtalk.blender.org/t/cycles-x-feedback/18598/405 It is said that it only adds three line of code to the Müller paper and it improves the result so much. I am not sure how true this "three lines of code" statement is, but if it's really that easy to add on top of Müller et al. maybe this is something that can be considered? But if this is actually not the case then nevermind.

Added subscriber: @tpet

Added subscriber: @Rowan-Ibbeken

Any updates on this, such as a rough timescale? The lack of caustics is one of the only things keeping Cycles from matching Corona and Luxcore for pure realism.

Added subscriber: @Garek

In #92571#1261165, @Rowan-Ibbeken wrote:
Any updates on this, such as a rough timescale? The lack of caustics is one of the only things keeping Cycles from matching Corona and Luxcore for pure realism.

There are no time estimates still, see recent comment from William:
https://code.blender.org/2021/11/cycles-x-project-update/#comment-126279

> In #92571#1261165, @Rowan-Ibbeken wrote: > Any updates on this, such as a rough timescale? The lack of caustics is one of the only things keeping Cycles from matching Corona and Luxcore for pure realism. There are no time estimates still, see recent comment from William: https://code.blender.org/2021/11/cycles-x-project-update/#comment-126279

In #92571#1261207, @Garek wrote:

In #92571#1261165, @Rowan-Ibbeken wrote:
Any updates on this, such as a rough timescale? The lack of caustics is one of the only things keeping Cycles from matching Corona and Luxcore for pure realism.

There are no time estimates still, see recent comment from William:
https://code.blender.org/2021/11/cycles-x-project-update/#comment-126279

Ah I see, we'll just have to be patient!

> In #92571#1261207, @Garek wrote: >> In #92571#1261165, @Rowan-Ibbeken wrote: >> Any updates on this, such as a rough timescale? The lack of caustics is one of the only things keeping Cycles from matching Corona and Luxcore for pure realism. > > There are no time estimates still, see recent comment from William: > https://code.blender.org/2021/11/cycles-x-project-update/#comment-126279 Ah I see, we'll just have to be patient!

Added subscriber: @makizar

Added subscriber: @johannes.wilde

Added subscriber: @xavierh

Added subscriber: @xrogueleaderx

Added subscriber: @Dangry

Added subscriber: @thomasmcs

Added subscriber: @TheRedWaxPolice

This issue was referenced by 66a863e301

This issue was referenced by 66a863e30197990e20a1e9a45923af6331935b21

This issue was referenced by bd249eb4f3

This issue was referenced by bd249eb4f37e65b2f4b3deef775bb136a2c234d1

This issue was referenced by 75a6d3abf7

This issue was referenced by 75a6d3abf75f3082adf5240ae34973844c0d9a09

This issue was referenced by 6d19da0b2d

This issue was referenced by 6d19da0b2d468f099e0c1f56392ab8a1750d114f

Added subscriber: @Yuro

Added subscriber: @AlexeyAdamitsky

Added subscriber: @costa

Added subscriber: @Max-49

Added subscriber: @kisvenus

Added subscriber: @Siah-2

Added subscriber: @skyscapeparadise

Added subscriber: @YesungYoon

Added subscriber: @RossoFiore

Added subscriber: @ThomasDinges

Added subscriber: @MoonMoon

Added subscriber: @Artwithblender10

Added subscriber: @JanErik

Download

What's New

Blender Studio

Manual

Developers Blog

Documentation

Benchmark

Blender Conference

Development Fund

One-time Donations

Cycles: path guiding #92571