PCF For DirectionalLight/SpotLight Shadows #8006

JMS55 · 2023-03-09T20:36:38Z

Objective

Improve antialiasing for non-point light shadow edges.
Very partially addresses Percentage Closer Filtering for shadows #3628.

Solution

Implements "The Witness"'s shadow map sampling technique.
- Ported from @superdump's old branch, all credit to them :)
Implements "Call of Duty: Advanced Warfare"'s stochastic shadow map sampling technique when the velocity prepass is enabled, for use with TAA.
- Uses interleaved gradient noise to generate a random angle, and then averages 8 samples in a spiral pattern, rotated by the random angle.
- I also tried spatiotemporal blue noise, but it was far too noisy to be filtered by TAA alone. In the future, we should try spatiotemporal blue noise + a specialized shadow denoiser such as https://gpuopen.com/fidelityfx-denoiser/#shadow. This approach would also be useful for hybrid rasterized applications with raytraced shadows.
- The COD presentation has an interesting temporal dithering of the noise for use with temporal supersampling that we should revisit when we get DLSS/FSR/other TSR.

Changelog

Added ShadowFilteringMethod. Improved directional light and spotlight shadow edges to be less aliased.

Migration Guide

Shadows cast by directional lights or spotlights now have smoother edges. To revert to the old behavior, add ShadowFilteringMethod::Hardware2x2 to your cameras.

…ic_dir_pcf

superdump

Basically ready to approve this, just a few nits/minor comments.

And we must make an issue and remember to implement Receiver Plane Depth Bias ( https://web.archive.org/web/20230309054654/http://developer.amd.com/wordpress/media/2012/10/Isidoro-ShadowMapping.pdf slides 36-38). We could additionally implement Adaptive Depth Bias for Shadows ( https://jcgt.org/published/0003/04/08/ ) with the tweaks from Adaptive Depth Bias for Soft Shadows ( https://w3-o.cs.hm.edu/users/nischwit/public_html/AdaptiveDepthBias_WSCG.pdf ) separately from this PR.

superdump · 2023-10-01T14:00:58Z

crates/bevy_pbr/src/render/shadow_sampling.wgsl

+    // NOTE: Due to non-uniform control flow above, we must use the level variant of the texture
+    // sampler to avoid use of implicit derivatives causing possible undefined behavior.


I think now this is split into its own function that this is no longer true. I have some memory of the control flow only applying within a function scope. Though I also remember thinking that was weird given it sounds like shader compilers generally inline everything? @cwfitzgerald - is the non-uniform control flow for the entire execution of the shader after entering the region of non-uniform control flow? Or is it scoped to functions? Or is the validation scoped to functions but the rule does apply?

cwfitzgerald confirmed that naga should emit validation errors if the texture sampling is done inside non-uniform control flow even if the code calls to another function.

Still, I think the Level variant makes sense regardless for shadow maps as they are not mipmapped and don't need gradients / mip level calculation to be done.

superdump · 2023-10-01T14:08:00Z

crates/bevy_pbr/src/render/shadow_sampling.wgsl

+}
+
+// https://web.archive.org/web/20230210095515/http://the-witness.net/news/2013/09/shadow-mapping-summary-part-1
+fn sample_shadow_map_castano_thirteen(light_local: vec2<f32>, depth: f32, array_index: i32) -> f32 {


This implements the 5x5 filter size but the implementation in MJP's has code for 3, 5, 7. We could add options for 3 and 7 later.

5x5 I found to give good results, didn't feel the need to do 3 or 7 atm, but yeah we can add it trivially in the future (at the cost of more pipeline key bits).

Or some view flags or something.

crates/bevy_pbr/src/render/shadow_sampling.wgsl

crates/bevy_pbr/src/render/shadows.wgsl

superdump · 2023-10-01T14:18:00Z

crates/bevy_pbr/src/render/shadows.wgsl

-    // NOTE: Due to non-uniform control flow above, we must use the level variant of the texture
-    // sampler to avoid use of implicit derivatives causing possible undefined behavior.
-#ifdef NO_ARRAY_TEXTURES_SUPPORT
-    return textureSampleCompareLevel(


See my other comment on this in the place where the code was moved. I want to double-check with cwfitzgerald. :)

crates/bevy_pbr/src/render/shadow_sampling.wgsl

Co-authored-by: Robert Swain <[email protected]>

crates/bevy_pbr/src/render/shadow_sampling.wgsl

superdump · 2023-10-02T16:43:02Z

crates/bevy_pbr/src/render/shadows.wgsl

-    // NOTE: Due to non-uniform control flow above, we must use the level variant of the texture
-    // sampler to avoid use of implicit derivatives causing possible undefined behavior.
-#ifdef NO_ARRAY_TEXTURES_SUPPORT
-    return textureSampleCompareLevel(


As mentioned, bother blocks are needed and both should be textureSampleCompareLevel.

…ic_dir_pcf

@superdump

# Objective - Improve antialiasing for non-point light shadow edges. - Very partially addresses bevyengine#3628. ## Solution - Implements "The Witness"'s shadow map sampling technique. - Ported from @superdump's old branch, all credit to them :) - Implements "Call of Duty: Advanced Warfare"'s stochastic shadow map sampling technique when the velocity prepass is enabled, for use with TAA. - Uses interleaved gradient noise to generate a random angle, and then averages 8 samples in a spiral pattern, rotated by the random angle. - I also tried spatiotemporal blue noise, but it was far too noisy to be filtered by TAA alone. In the future, we should try spatiotemporal blue noise + a specialized shadow denoiser such as https://gpuopen.com/fidelityfx-denoiser/#shadow. This approach would also be useful for hybrid rasterized applications with raytraced shadows. - The COD presentation has an interesting temporal dithering of the noise for use with temporal supersampling that we should revisit when we get DLSS/FSR/other TSR. --- ## Changelog * Added `ShadowFilteringMethod`. Improved directional light and spotlight shadow edges to be less aliased. ## Migration Guide * Shadows cast by directional lights or spotlights now have smoother edges. To revert to the old behavior, add `ShadowFilteringMethod::Hardware2x2` to your cameras. --------- Co-authored-by: IceSentry <[email protected]> Co-authored-by: Daniel Chia <[email protected]> Co-authored-by: robtfm <[email protected]> Co-authored-by: Brandon Dyer <[email protected]> Co-authored-by: Edgar Geier <[email protected]> Co-authored-by: Robert Swain <[email protected]> Co-authored-by: Elabajaba <[email protected]> Co-authored-by: IceSentry <[email protected]>

@superdump

# Objective - Improve antialiasing for non-point light shadow edges. - Very partially addresses bevyengine#3628. ## Solution - Implements "The Witness"'s shadow map sampling technique. - Ported from @superdump's old branch, all credit to them :) - Implements "Call of Duty: Advanced Warfare"'s stochastic shadow map sampling technique when the velocity prepass is enabled, for use with TAA. - Uses interleaved gradient noise to generate a random angle, and then averages 8 samples in a spiral pattern, rotated by the random angle. - I also tried spatiotemporal blue noise, but it was far too noisy to be filtered by TAA alone. In the future, we should try spatiotemporal blue noise + a specialized shadow denoiser such as https://gpuopen.com/fidelityfx-denoiser/#shadow. This approach would also be useful for hybrid rasterized applications with raytraced shadows. - The COD presentation has an interesting temporal dithering of the noise for use with temporal supersampling that we should revisit when we get DLSS/FSR/other TSR. --- ## Changelog * Added `ShadowFilteringMethod`. Improved directional light and spotlight shadow edges to be less aliased. ## Migration Guide * Shadows cast by directional lights or spotlights now have smoother edges. To revert to the old behavior, add `ShadowFilteringMethod::Hardware2x2` to your cameras. --------- Co-authored-by: IceSentry <[email protected]> Co-authored-by: Daniel Chia <[email protected]> Co-authored-by: robtfm <[email protected]> Co-authored-by: Brandon Dyer <[email protected]> Co-authored-by: Edgar Geier <[email protected]> Co-authored-by: Robert Swain <[email protected]> Co-authored-by: Elabajaba <[email protected]> Co-authored-by: IceSentry <[email protected]>

@superdump

# Objective - Improve antialiasing for non-point light shadow edges. - Very partially addresses bevyengine#3628. ## Solution - Implements "The Witness"'s shadow map sampling technique. - Ported from @superdump's old branch, all credit to them :) - Implements "Call of Duty: Advanced Warfare"'s stochastic shadow map sampling technique when the velocity prepass is enabled, for use with TAA. - Uses interleaved gradient noise to generate a random angle, and then averages 8 samples in a spiral pattern, rotated by the random angle. - I also tried spatiotemporal blue noise, but it was far too noisy to be filtered by TAA alone. In the future, we should try spatiotemporal blue noise + a specialized shadow denoiser such as https://gpuopen.com/fidelityfx-denoiser/#shadow. This approach would also be useful for hybrid rasterized applications with raytraced shadows. - The COD presentation has an interesting temporal dithering of the noise for use with temporal supersampling that we should revisit when we get DLSS/FSR/other TSR. --- ## Changelog * Added `ShadowFilteringMethod`. Improved directional light and spotlight shadow edges to be less aliased. ## Migration Guide * Shadows cast by directional lights or spotlights now have smoother edges. To revert to the old behavior, add `ShadowFilteringMethod::Hardware2x2` to your cameras. --------- Co-authored-by: IceSentry <[email protected]> Co-authored-by: Daniel Chia <[email protected]> Co-authored-by: robtfm <[email protected]> Co-authored-by: Brandon Dyer <[email protected]> Co-authored-by: Edgar Geier <[email protected]> Co-authored-by: Robert Swain <[email protected]> Co-authored-by: Elabajaba <[email protected]> Co-authored-by: IceSentry <[email protected]>

JMS55 and others added 30 commits January 1, 2023 15:35

Extract constants

c157cae

Remove newline

b6acf22

fix unwrap on empty frag shader in prepass

715ca65

docs

35093ed

Merge commit '35093ed2e81b737fadfa5b50c32a6e49245acdf0' into gtao

5c9bb20

Fix merge

826b19a

WIP FSR2 plugin

060c52d

Further plugin setup

14efad2

WIP: FSR2Plugin

77ca0c8

FSR2 jitter (breaks TAA for now)

8f5fda4

Fix shader ifdefs

1bb796f

Extract projection

63003f7

Misc

0b757e0

Fix dep path

f48bb83

Add neccesary texture usages

1d48c29

Misc

3d5c81c

Add device field

ff68974

Cleanup jitter

e9942d7

Error checking

0e38fb4

Store device in context

b35586f

WIP: Update TAA impl

cada547

Fixes

88112be

Move TAA to core pipeline

2cc10d5

Merge commit 'cab065bad4302e4df427ffbafcdb112e755321ce' into taa-only

bc87f4c

Misc

bf89ceb

Merge fixes

a545e2a

Documentation

c44c33a

Misc formatting

1ddfa48

Appease clippy

e76cd0e

Add plugin note to docs

bf36eb0

JMS55 added 4 commits September 14, 2023 09:57

Document numbers

4d86b6d

Merge commit '8192ac6f1e572601b4ada08fa2f82114e07f8aec' into stochast…

2867d90

…ic_dir_pcf

Merge commit '8ace2ff9e361dd7ef1bc620b84674def0cb56454' into stochast…

c26f643

…ic_dir_pcf

Merge commit '20ed3e0e765405ee334a44ce3c49c72bd29ab753' into stochast…

8234d7d

…ic_dir_pcf

Elabajaba approved these changes Sep 30, 2023

View reviewed changes

alice-i-cecile added the S-Ready-For-Final-Review This PR has been approved by the community. It's ready for a maintainer to consider merging it label Sep 30, 2023

superdump requested changes Oct 1, 2023

View reviewed changes

superdump reviewed Oct 1, 2023

View reviewed changes

crates/bevy_pbr/src/render/shadow_sampling.wgsl Outdated Show resolved Hide resolved

JMS55 and others added 3 commits October 1, 2023 14:17

Update crates/bevy_pbr/src/render/shadows.wgsl

0e60a90

Co-authored-by: Robert Swain <[email protected]>

Update crates/bevy_pbr/src/render/shadow_sampling.wgsl

600b3a8

Co-authored-by: Robert Swain <[email protected]>

Change sample_shadow_map_hardware

24eab26

JMS55 requested a review from superdump October 1, 2023 21:23

superdump requested changes Oct 2, 2023

View reviewed changes

JMS55 added 2 commits October 5, 2023 12:46

Fix sample_shadow_map_hardware()

e24e1c5

Merge commit 'a9622408665662eff9ddae83d913c6cfa2fa61d2' into stochast…

394fdba

…ic_dir_pcf

superdump approved these changes Oct 5, 2023

View reviewed changes

superdump enabled auto-merge October 5, 2023 19:51

alice-i-cecile disabled auto-merge October 6, 2023 21:33

alice-i-cecile enabled auto-merge October 6, 2023 21:34

alice-i-cecile disabled auto-merge October 7, 2023 16:56

Merge commit '154a49044514fb21b0f83f4f077d76380e12a8a8' into stochast…

4f9dff2

…ic_dir_pcf

alice-i-cecile enabled auto-merge October 7, 2023 17:01

alice-i-cecile added this pull request to the merge queue Oct 7, 2023

Merged via the queue into bevyengine:main with commit 1f95a48 Oct 7, 2023
21 checks passed

DGriffin91 mentioned this pull request Oct 8, 2023

StandardMaterial doesn't work in webgl2 on Windows with DX12 #10048

Closed

cart mentioned this pull request Oct 13, 2023

News: Release 0.12 bevyengine/bevy-website#754

Merged

43 tasks

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

PCF For DirectionalLight/SpotLight Shadows #8006

PCF For DirectionalLight/SpotLight Shadows #8006

JMS55 commented Mar 9, 2023 •

edited

Loading

superdump left a comment •

edited

Loading

superdump Oct 1, 2023

superdump Oct 1, 2023

superdump Oct 1, 2023

JMS55 Oct 1, 2023

superdump Oct 2, 2023

superdump Oct 1, 2023

superdump Oct 2, 2023

		// NOTE: Due to non-uniform control flow above, we must use the level variant of the texture
		// sampler to avoid use of implicit derivatives causing possible undefined behavior.

PCF For DirectionalLight/SpotLight Shadows #8006

PCF For DirectionalLight/SpotLight Shadows #8006

Conversation

JMS55 commented Mar 9, 2023 • edited Loading

Objective

Solution

Changelog

Migration Guide

superdump left a comment • edited Loading

Choose a reason for hiding this comment

superdump Oct 1, 2023

Choose a reason for hiding this comment

superdump Oct 1, 2023

Choose a reason for hiding this comment

superdump Oct 1, 2023

Choose a reason for hiding this comment

JMS55 Oct 1, 2023

Choose a reason for hiding this comment

superdump Oct 2, 2023

Choose a reason for hiding this comment

superdump Oct 1, 2023

Choose a reason for hiding this comment

superdump Oct 2, 2023

Choose a reason for hiding this comment

JMS55 commented Mar 9, 2023 •

edited

Loading

superdump left a comment •

edited

Loading