Move view transformations to bevy_render by atlv24 · Pull Request #20313 · bevyengine/bevy

atlv24 · 2025-07-28T12:45:45Z

Objective

having view_transformations in bevy_pbr just because they need the bindings sucks, it hinders code reuse and means we often can't use these functions on custom view bindings.

Solution

Add binding-less version of view transformations and deprecate view_transformations.wgsl

Testing

3d_scene runs

The Plan:

this is a "trial run" first step: it includes no migration of the existing code-base, and does no breaking changes, its just a minimal example of what going in this direction will look like. If we like this direction i'll sink more time into this, writing a migration guide and some followup prs:

migrations for the codebase
removal of the deprecated functions (in 0.18, probably)
similar PRs for PBR functions

…ransformations.wgsl

github-actions · 2025-07-28T12:46:21Z

It looks like your PR is a breaking change, but you didn't provide a migration guide.

Please review the instructions for writing migration guides, then expand or revise the content in the migration guides directory to reflect your changes.

atlv24 · 2025-07-28T14:02:04Z

crates/bevy_render/src/view/view.wgsl

+fn depth_ndc_to_view_z(ndc_depth: f32, clip_from_view: mat4x4<f32>, view_from_clip: mat4x4<f32>) -> f32 {
+#ifdef VIEW_PROJECTION_PERSPECTIVE
+    return -perspective_camera_near(clip_from_view) / ndc_depth;
+#else ifdef VIEW_PROJECTION_ORTHOGRAPHIC
+    return -(clip_from_view[3][2] - ndc_depth) / clip_from_view[2][2];
+#else
+    let view_pos = view_from_clip * vec4(0.0, 0.0, ndc_depth, 1.0);
+    return view_pos.z / view_pos.w;
+#endif


note: the need for two matrices here smells funny to me. why do we need this? is there a way to write it in terms of one matrix?

It's because of optimisations. Using values from the forward transform (clip from view) to invert it.

atlv24 · 2025-07-28T14:03:54Z

crates/bevy_render/src/view/view.wgsl

+    let world_pos = world_from_view * vec4(view_pos, 1.0);
+    return world_pos.xyz;


note: most of these functions are the same thing with different names. "augment point with w = 1.0, mul, return xyz" or same thing but with w = 0.0 for directions. Do we want to unify them?

Not sure it's worth it. These functions are mostly so people don't need to understand exactly what the correct maths should be, I guess. Convenience. But extracting functions for... transform_position() and transform_direction() seems unnecessary to me.

tychedelia

This makes a lot of sense to me.

I think in the future, we may be able to better play with code generation, such that these functions link to a module that is provided at runtime by the WESL loader and contains the resources, but we can experiment with that later.

superdump

I'm fine with this but it's probably too late for 0.17.

superdump · 2025-07-30T14:24:45Z

Turns out it wasn't too late for 0.17.

# Objective - add migration guide for #20313

# Objective - split off bevy_shader from bevy_render to allow for defining shader libraries and materials for scenes without depending on a renderer ## Solution - a bunch of small refactor commits, then bevy_shader! (review commit by commit) ## Testing - please help with this, especially with spirv passthrough and decoupled_naga testing: i only ran 3d_scene at every step to make sure i didnt break anything. EDIT: this has been tested now, decoupled_naga works, and I also fixed a bug that was on main when wesl and decoupled_naga were both present it didnt compile, now it does # Future Work - Split ShaderCacheError out of PipelineCacheError (or just rename) (punted because breaking change) - Consider renaming PipelineCacheId to ShaderCacheId ? (punted because breaking change) - bevy_material - bevy_bsdf (bindingless pbr shader library) - move view.wgsl into bevy_camera after #20313 lands ?

# Objective - having view_transformations in bevy_pbr just because they need the bindings sucks, it hinders code reuse and means we often can't use these functions on custom view bindings. ## Solution - Add binding-less version of view transformations and deprecate view_transformations.wgsl ## Testing - 3d_scene runs # The Plan: this is a "trial run" first step: it includes no migration of the existing code-base, and does no breaking changes, its just a minimal example of what going in this direction will look like. If we like this direction i'll sink more time into this, writing a migration guide and some followup prs: - migrations for the codebase - removal of the deprecated functions (in 0.18, probably) - similar PRs for PBR functions

…0338) # Objective - add migration guide for bevyengine#20313

# Objective - split off bevy_shader from bevy_render to allow for defining shader libraries and materials for scenes without depending on a renderer ## Solution - a bunch of small refactor commits, then bevy_shader! (review commit by commit) ## Testing - please help with this, especially with spirv passthrough and decoupled_naga testing: i only ran 3d_scene at every step to make sure i didnt break anything. EDIT: this has been tested now, decoupled_naga works, and I also fixed a bug that was on main when wesl and decoupled_naga were both present it didnt compile, now it does # Future Work - Split ShaderCacheError out of PipelineCacheError (or just rename) (punted because breaking change) - Consider renaming PipelineCacheId to ShaderCacheId ? (punted because breaking change) - bevy_material - bevy_bsdf (bindingless pbr shader library) - move view.wgsl into bevy_camera after bevyengine#20313 lands ?

# Objective - Fix chrome rendering bug: <img width="1755" height="985" alt="image" src="https://github.com/user-attachments/assets/6293a960-b8fa-460e-9d71-e7f2a9fb2d20" /> ## Solution - Revert part of #20313 - I don't know why this fixes it, but #20960 also ran into it ## Testing -

# Objective This PR is to add SSAO support on WebGPU. ## Solution Add r16float detect and fallback to r32float if not supported. FYI the initial ssao PR [here](#7402). ## Testing - Did you test these changes? If so, how? Yes, here is the command to test the ssao example `RUSTFLAGS="--cfg getrandom_backend=\"wasm_js\"" cargo run --example ssao --target wasm32-unknown-unknown --features webgpu` and then http://localhost:1334, make sure it supports [WebGPU](https://caniuse.com/webgpu) - Are there any parts that need more testing? N/A - How can other people (reviewers) test your changes? Is there anything specific they need to know? 1. For the R32Float fallback, only the 1st commit is needed. However, I ran into some strange flickering issue on Chrome (and Chrome canary too), so I added a "detect_r16float_support" flag and a temp fix in the 2nd and 3rd commit, which are intended to be reverted when merging. 2. with the 1st commit, Safari works fine, but Chrome has the flickering, and I fixed it by partially reverting a shader change from #20313, which is very strange. It is probably caused by some shader optimization issue in Chrome, and I have no clue why, as the change is just moving the calculation to a different function. Here is the flickering on Mac Chrome(140.0.7339.133 (Official Build) (arm64)) <img width="353" height="327" alt="chrome_ssao" src="https://github.com/user-attachments/assets/0666673f-508e-4b57-a152-19327ffdb6f3" /> - If relevant, what platforms did you test these changes on, and are there any important ones you can't test? Tested on native(mac) and WebGPU. To test on native, set "detect_r16float_support" to false to force R32Float. I couldn't make the WebGPU ssao example running on Windows 11 Chrome or Firefox, there is some error in 'taa_pipeline' . ---

Add binding-less version of view transformations and deprecate view_t…

acf6815

…ransformations.wgsl

atlv24 added A-Rendering Drawing game state to the screen M-Migration-Guide A breaking change to Bevy's public API that needs to be noted in a migration guide S-Needs-Review Needs reviewer attention (from anyone!) to move forward labels Jul 28, 2025

github-project-automation bot added this to Rendering (Old) Jul 28, 2025

atlv24 added the D-Shaders This code uses GPU shader languages label Jul 28, 2025

atlv24 commented Jul 28, 2025

View reviewed changes

atlv24 requested review from IceSentry, robtfm, superdump and tychedelia July 29, 2025 02:31

tychedelia approved these changes Jul 29, 2025

View reviewed changes

atlv24 mentioned this pull request Jul 30, 2025

bevy_shader #20330

Merged

superdump approved these changes Jul 30, 2025

View reviewed changes

superdump added this pull request to the merge queue Jul 30, 2025

Merged via the queue into bevyengine:main with commit e08c78b Jul 30, 2025
41 checks passed

github-project-automation bot moved this to Done in Rendering (Old) Jul 30, 2025

atlv24 mentioned this pull request Jul 30, 2025

Add migration guide for view transformation deprecation #20338

Merged

github-merge-queue bot pushed a commit that referenced this pull request Jul 30, 2025

Add migration guide for view transformation deprecation (#20338)

f19aa8d

# Objective - add migration guide for #20313

tychedelia pushed a commit to tychedelia/bevy that referenced this pull request Jul 31, 2025

Add migration guide for view transformation deprecation (bevyengine#2…

d1c2165

…0338) # Objective - add migration guide for bevyengine#20313

atlv24 deleted the ad/view-binding-yeet-1 branch August 1, 2025 01:04

diyu-motif mentioned this pull request Sep 11, 2025

Added SSAO support on WebGPU, with R32Float fallback. #20960

Merged

atlv24 mentioned this pull request Sep 24, 2025

Fix chrome rendering bug #21202

Merged

alice-i-cecile mentioned this pull request Sep 26, 2025

Investigate why moving view transformations broke rendering on Chrome #21221

Open

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

Move view transformations to bevy_render#20313

Move view transformations to bevy_render#20313
superdump merged 1 commit intobevyengine:mainfrom
atlv24:ad/view-binding-yeet-1

atlv24 commented Jul 28, 2025 •

edited

Loading

Uh oh!

github-actions bot commented Jul 28, 2025

Uh oh!

atlv24 Jul 28, 2025

Uh oh!

superdump Jul 30, 2025

Uh oh!

atlv24 Jul 28, 2025

Uh oh!

superdump Jul 30, 2025

Uh oh!

tychedelia left a comment

Uh oh!

superdump left a comment

Uh oh!

superdump commented Jul 30, 2025

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

		let world_pos = world_from_view * vec4(view_pos, 1.0);
		return world_pos.xyz;

Uh oh!

Conversation

atlv24 commented Jul 28, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Objective

Solution

Testing

The Plan:

Uh oh!

github-actions bot commented Jul 28, 2025

Uh oh!

atlv24 Jul 28, 2025

Choose a reason for hiding this comment

Uh oh!

superdump Jul 30, 2025

Choose a reason for hiding this comment

Uh oh!

atlv24 Jul 28, 2025

Choose a reason for hiding this comment

Uh oh!

superdump Jul 30, 2025

Choose a reason for hiding this comment

Uh oh!

tychedelia left a comment

Choose a reason for hiding this comment

Uh oh!

superdump left a comment

Choose a reason for hiding this comment

Uh oh!

superdump commented Jul 30, 2025

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

atlv24 commented Jul 28, 2025 •

edited

Loading