Naga mesh shader WGSL writer #8481

Slightlyclueless · 2025-11-05T23:01:15Z

Connections
Based off of #8370

Description

Adds a WGSL writer for mesh shader functionality.

Testing
TODO

Squash or Rebase?
Squash

Checklist

Run cargo fmt.
Run taplo format.
Run cargo clippy --tests. If applicable, add:
- --target wasm32-unknown-unknown
Run cargo xtask test to run tests.
If this contains user-facing changes, add a CHANGELOG.md entry.

Slightlyclueless · 2025-11-05T23:11:46Z

All changes are in naga/src/back/wgsl/writer.rs and naga/src/back/wgsl/to_wgsl.rs. Code is ugly and I absolutely want to clean it up before I mark as ready to merge

Dunno what's up with the PR history lol.

inner-daemons · 2025-11-06T04:39:30Z

First comments

Logic to detect mesh shader logic is incorrect. You also need to check for mesh shader builtins, per primitive attributes anywhere, task payload variables. This is because files might contain types/data that isn't used in any shaders
The generated task payload variable doesn't have a storage class
I don't understand why the shader stage, workgroup size, task payload come in different orders between mesh and task shaders
Mesh thing doesn't have the variable it needs (@mesh(variable))
Per primitive isn't written correctly
You will want to make sure that the snapshot is correctly written

Slightlyclueless · 2025-11-06T18:51:30Z

Logic to detect mesh shader logic is incorrect. You also need to check for mesh shader builtins, per primitive attributes anywhere, task payload variables. This is because files might contain types/data that isn't used in any shaders

I though so, but wasn't sure. Will change,

The generated task payload variable doesn't have a storage class

Ah yep, not sure how I missed that one.

I don't understand why the shader stage, workgroup size, task payload come in different orders between mesh and task shaders

Because I am a clutz, and didn't sanity check that after going "Yay it no longer crashes".

Mesh thing doesn't have the variable it needs (@mesh(variable))

Whoops, that's what I get for not looking at the correct spec document.

Per primitive isn't written correctly

Not sure what you mean by this?

You will want to make sure that the snapshot is correctly written

Again, not sure what you mean. I may just be being a bit thick though

inner-daemons · 2025-11-06T21:46:03Z

For the snapshot, you will want to go to naga/tests/in/wgsl/mesh-shader.toml and change to this line
targets = "IR | ANALYSIS | WGSL"

Then run cargo xtask test snapshot before every commit. Snapshots are how WGPU tracks how shaders are written between commits, and how we check that naga continues to work on many shaders.

inner-daemons

Generally pretty good, some comments. Also please format your code and make sure that CI passes. At the bottom of the github page it'll say stuff like "CI / Test Linux x86_64 (pull_request) Failing after 4m", you can usually figure out the error by reading through these logs

inner-daemons · 2025-11-17T03:19:15Z

naga/src/back/wgsl/writer.rs

+                    if shader_stage == ShaderStage::Mesh {
+                        write!(
+                            self.out,
+                            "@{stage_str}({})",


Suggested change

"@{stage_str}({})",

"@{stage_str}({}) ",

inner-daemons · 2025-11-17T03:23:24Z

naga/tests/in/wgsl/mesh-shader.toml

+
+[msl]
+fake_missing_bindings = true
+lang_version = [2, 4]
+spirv_cross_compatibility = false
+zero_initialize_workgroup_memory = false
+
+[hlsl]
+shader_model = "V6_5"
+fake_missing_bindings = true
+zero_initialize_workgroup_memory = true
+
+[spv]
+version = [1, 4]
+capabilities = ["MeshShadingEXT"]


Suggested change

[msl]

fake_missing_bindings = true

lang_version = [2, 4]

spirv_cross_compatibility = false

zero_initialize_workgroup_memory = false

[hlsl]

shader_model = "V6_5"

fake_missing_bindings = true

zero_initialize_workgroup_memory = true

[spv]

version = [1, 4]

capabilities = ["MeshShadingEXT"]

inner-daemons · 2025-11-17T03:24:08Z

naga/tests/in/wgsl/mesh-shader.toml

+# Stolen from ray-query.toml
+


Suggested change

# Stolen from ray-query.toml

inner-daemons · 2025-11-17T03:24:57Z

naga/tests/in/wgsl/mesh-shader.toml

+
 god_mode = true
-targets = "IR | ANALYSIS"
+targets = "IR | ANALYSIS | WGSL"


Can you apply this change to the other files here, like mesh-shader-lines.toml etc

inner-daemons · 2025-11-17T03:26:01Z

naga/tests/out/wgsl/glsl-inverse-polyfill.frag.wgsl

-   let det = (m[0][0] * adj[0][0] + m[0][1] * adj[1][0] + m[0][2] * adj[2][0] + m[0][3] * adj[3][0]);
-
-   return adj * (1 / det);
+fn _naga_inverse_4x4_f32(m: mat4x4<f32>) -> mat4x4<f32> {


Can you undo this? You can just run git checkout trunk -- naga/tests/out/wgsl/glsl-inverse-pollyfill.frag.wgsl

Unsure if this is due to changing the tabs to spaces/vice versa or due to line endings, I suspect the first

inner-daemons · 2025-11-17T03:35:14Z

naga/src/back/wgsl/writer.rs

+    fn write_attributes(
+        &mut self,
+        attributes: &[Attribute],
+        mesh_output_variable: Option<String>,


Lets not have this as another args, can you change the mesh stage attribute to being a special enum variant with this?

inner-daemons · 2025-11-17T03:37:32Z

naga/src/back/wgsl/writer.rs

+            if per_primitive {
+                vec![
+                    Attribute::PerPrimitive,
+                    Attribute::Location(location),
+                    Attribute::BlendSrc(blend_src),
+                    Attribute::Interpolate(interpolation, sampling),
+                ]
+            } else {
+                vec![
+                    Attribute::Location(location),
+                    Attribute::BlendSrc(blend_src),
+                    Attribute::Interpolate(interpolation, sampling),
+                ]
+            }


Lets try to just have one vector, and add an extra PerPrimitive element to it if needed

inner-daemons · 2025-11-17T03:39:48Z

naga/src/common/wgsl/to_wgsl.rs

            | Bi::LineIndices
-            | Bi::MeshTaskSize
-            | Bi::PointIndex
-            | Bi::VertexCount
-            | Bi::PrimitiveCount
-            | Bi::Vertices
-            | Bi::Primitives => return None,
+            | Bi::PointIndex => return None,


These should also be written, just like TriangleIndices

inner-daemons · 2025-11-17T03:40:27Z

naga/src/back/wgsl/writer.rs

+            if per_primitive {
+                vec![
+                    Attribute::PerPrimitive,
+                    Attribute::Location(location),
+                    Attribute::Interpolate(interpolation, sampling),
+                ]
+            } else {
+                vec![
+                    Attribute::Location(location),
+                    Attribute::Interpolate(interpolation, sampling),
+                ]
+            }


Lets try to just have one vector, and add an extra PerPrimitive element to it if needed

inner-daemons · 2025-11-17T03:42:27Z

naga/src/back/wgsl/writer.rs

+                    if ep.task_payload.is_some() {
+                        let payload_name = module.global_variables[ep.task_payload.unwrap()]
+                            .name
+                            .clone()
+                            .unwrap();
+                        vec![
+                            Attribute::Stage(ShaderStage::Mesh),
+                            Attribute::MeshTaskPayload(payload_name),
+                            Attribute::WorkGroupSize(ep.workgroup_size),
+                        ]
+                    } else {
+                        vec![
+                            Attribute::Stage(ShaderStage::Mesh),
+                            Attribute::WorkGroupSize(ep.workgroup_size),
+                        ]
+                    }


Lets try to just have one vector, and add an extra TaskPayload element to it if needed

inner-daemons · 2025-11-17T03:47:52Z

Also feel free to mark this as no longer a draft

And just to manage expectations, this probably won't be merged until at least the 26th, as I don't have the power to do that so it'll have to have a second review (hopefully with fewer comments)

inner-daemons and others added 30 commits August 14, 2025 12:53

Initial commit

1c90d19

Other initial changes

8c3e550

Updated shader snapshots

85bbc5a

Added new HLSL limitation

ccf8467

Moved error to global variable error

e55c02f

Merge branch 'trunk' into mesh-shading/naga-ir

f3a31a4

Added docs to per_primitive

0f6da75

Added a little bit more docs here and there in IR

3017214

Merge branch 'trunk' into mesh-shading/naga-ir

19b55b5

Adding validation to ensure that task shaders have a task payload

198437b

Updated spec to reflect the change to payload variables

64000e4

Merge branch 'trunk' into mesh-shading/naga-ir

0575e98

Updated the mesh shading spec because it was goofy

b572ec7

Merge branch 'trunk' into mesh-shading/naga-ir

34d0411

Merge branch 'trunk' into mesh-shading/naga-ir

02664e4

some doc tweaks

7bec4dd

Tried to clarify docs a little

2fcb853

Merge branch 'trunk' into mesh-shading/naga-ir

3009b5a

Tried to update spec

8bfe106

Removed a warning

6ccaeec

Addressed comment about docs mistake

5b7ba11

Merge branch 'trunk' into mesh-shading/naga-ir

29c6972

Merge branch 'trunk' into mesh-shading/naga-ir

63fa8b5

Merge branch 'trunk' into mesh-shading/naga-ir

26c8681

Merge branch 'trunk' into mesh-shading/naga-ir

d9cac9c

Merge branch 'trunk' into mesh-shading/naga-ir

c112cb4

Merge branch 'trunk' into mesh-shading/naga-ir

e1ff67d

Merge branch 'trunk' into mesh-shading/naga-ir

64644f7

Merge branch 'trunk' into mesh-shading/naga-ir

739948b

Merge branch 'trunk' into mesh-shading/naga-ir

7ca25a4

inner-daemons and others added 8 commits November 1, 2025 01:30

Added changelog entry

64798dd

Merge branch 'trunk' into mesh-shading/wgsl-parse

c8dccd3

Made parser respect enable extension

bd923cd

Updated mesh shader spec

d95070a

Cleaned up the mesh shader analyzer function

ace7e17

Merge branch 'trunk' into mesh-shading/wgsl-parse

f55dbd9

Initial Commit

5bd8385

cargo fmt

3de9940

Slightlyclueless added 3 commits November 5, 2025 23:18

Yeet unused import

8fd04d4

I Forgor

3b7ad04

Fix merge conflicts

74ba2fc

inner-daemons self-assigned this Nov 6, 2025

inner-daemons mentioned this pull request Nov 6, 2025

Mesh Shaders Tracking Issue #7197

Open

36 tasks

Slightlyclueless added 2 commits November 6, 2025 19:13

Fix per primitive stuff

d69df5e

Add task payload storage class

b797538

Change snapshot

e5438e3

inner-daemons mentioned this pull request Nov 10, 2025

WGSL parsing for mesh shaders #8370

Merged

6 tasks

Slightlyclueless added 2 commits November 16, 2025 23:31

Write the mesh output variable, in quite possibly the laziest way

694a188

Correct feature detection. I hope...

c131238

inner-daemons self-requested a review November 16, 2025 23:48

Slightlyclueless added 3 commits November 16, 2025 23:58

Merge branch 'trunk' into naga-mesh-wgpu-writer

250ae47

setting the variable got dropped in merge conflicts...

9d78bc5

Commit snapshots

03445d2

inner-daemons suggested changes Nov 17, 2025

View reviewed changes

inner-daemons assigned cwfitzgerald Nov 17, 2025

Naga mesh shader WGSL writer #8481

Are you sure you want to change the base?

Naga mesh shader WGSL writer #8481

Conversation

Slightlyclueless commented Nov 5, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Slightlyclueless commented Nov 5, 2025

Uh oh!

inner-daemons commented Nov 6, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Slightlyclueless commented Nov 6, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

inner-daemons commented Nov 6, 2025

Uh oh!

inner-daemons left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

inner-daemons commented Nov 17, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

Slightlyclueless commented Nov 5, 2025 •

edited

Loading

inner-daemons commented Nov 6, 2025 •

edited

Loading

Slightlyclueless commented Nov 6, 2025 •

edited

Loading

inner-daemons commented Nov 17, 2025 •

edited

Loading