[HDRP][Path Tracing] Slight robustness improvement for sample rescaling #5800
Conversation
|
Hi! This comment will help you figure out which jobs to run before merging your PR. The suggestions are dynamic based on what files you have changed. HDRP Depending on the scope of your PR, you may need to run more jobs than what has been suggested. Please speak to your lead or a Graphics SDET (#devs-graphics-automation) if you are unsure. |
|
It appears that you made a non-draft PR! |
There was a problem hiding this comment.
For the fix in the PR, couln't find any issues after clamping the value ✔️
Using this PR to give feedback on top of theses fixes.
Theses improvements helps a lot what happens with volumetric fog in raster !
There is still some discrepencies, not linked to specific issues but just unsuported features of the volumetric fog override by the pathtracer.
What I would like is to include those in the documentation. (Adding a Volumetric Fog section, under sub surface scattering section for example, here)
Here's a list of what is currently not supported.
Some are linked to how they are rendered in raster so would be expected (like the quality stuff) but others are just unsupported but could make sense in pathtracing pipeline:
- Density volume, now called Local Volumetric Fog (plays a large role is discrepencies in project where there is one setup)
- Shadow dimmer in Volumetric options in lights is ignored
- Ambient light probe dimmer
- Volumetric fog distance
- Slice deformation uniformity
- Denoising mode
- All the quality controls
- Directionnal light only checkbox
- Anisotropy
|
@sebastienlagarde, @remi-chapelain, I'll add a couple of these missing features that are easy to add (light and shadow dimmers, directional-only mode) to this PR, do not merge it just yet. |
|
I added support for
The PR is now ready for review. |
There was a problem hiding this comment.
Thanks for the additions, theses are great to better match what happens in raster ! 🥳
I tested and couldn't find any issues ✔️
However, would be great to just add a section in the doc of the pathtracer just to note the rest of the things unsupported for now :
- Local volumetric fog (aka old density volume)
- Ambient light probe dimmer
- Volumetric fog distance
- Anisotropy
- And raster quality settings (denoising, quality control, slice deformation uniformity.. etc)
|
@remi-chapelain, the doc will be updated in another PR. |
* Add missing DisallowMultipleComponent attribute in AdditionalData components #5859 * [HDRP][Path Tracing] Slight robustness improvement for sample rescaling #5800 * recaptured reflection probes, updated normals on 2nd room's floor (#5873) * Hair Documentation Pass (#5868) * Hair Documentation Pass * Add the images * Apply various feedback to the documentation * Fix small typo * Update the OSX reference image (forgot to add it previous PR, causing failure) and rerun Yamato * Update the tooltips * [HDRP] Remove Fake GTAO bounce on the occlusion when used for direct lighting #5836 * Update 5014_VolumetricCloudsBanding.png * Update 2307_Shadow_VeryHigh.png Co-authored-by: Arttu Peltonen <77337829+arttu-peltonen@users.noreply.github.com> Co-authored-by: Emmanuel Turquin <emmanuel@turquin.org> Co-authored-by: pierre-unity <39901544+pierre-unity@users.noreply.github.com> Co-authored-by: John Parsaie <johnpa@unity3d.com>
3 participants
When rescaling a sample for later reuse (in path tracing), there could be situations where, due to numerical imprecision, the sample value would reach 1.0, potentially causing issues in subsequent computations (NaNs, zeros, etc.). Such issues were spotted for instance in volumetric scattering. This PR addresses that problem.