Rewrite the ash runner #419

Firestar99 · 2025-09-30T14:20:41Z

Requires #442

Followup from #418

I've noticed that running the ash runner with debug layers enabled (--debug-layers) prints tons of validation errors, all of which should be considered programming errors and sort of equivalent to a runtime panic. Instead of trying to fix what is there, I decided to completely rewrite the ash runner instead, and update it to the modern Vulkan 1.3 way you'd render things.

use Vulkan 1.3
use DynamicRendering feature replacing Framebuffers
extract MyDevice struct containing vk initialization
decouple rendering from swapchain management
redo swapchain sync, fixes validation errors, intentionally kept basic
make rendering lazily recreate its pipeline

MacOS: removes the ash-molten dependency, which means we now require MoltenVK (or the Vulkan SDK) to be installed on MacOS for the ash runner to work. The wgpu runner still works as normal, with the metal backend.

Reviews

Please test this on your machine with validation layers turned on, if possible (requires the Vulkan SDK). Report back if it runs and on which OS and graphics card.

git checkout simplify-ash-runner2
cargo run --bin example-runner-ash -- --debug-layer --shader=sky
cargo run --bin example-runner-ash -- --debug-layer --shader=mouse

Note:

--shader=sky: holding arrow keys should change brightness
--shader=mouse: should be animated
--debug-layer requires vulkan sdk

examples/runners/ash/src/main.rs

nnethercote · 2025-09-30T22:21:36Z

Thanks for cleaning this up, I don't understand everything about this code but the parts I do understand look good. I can see you've generally improved the error handling, fixed a bunch of "FIXME" comments, etc.

It would be nice to have some commit messages that are more than a single line. The "ash runner: rewrite" commit in particular changes a lot of code with zero explanation.

MarijnS95 · 2025-10-02T10:38:59Z

examples/runners/ash/src/device.rs

+    let message_id_name = if callback_data.p_message_id_name.is_null() {
+        Cow::from("")
+    } else {
+        unsafe { CStr::from_ptr(callback_data.p_message_id_name).to_string_lossy() }
+    };


https://docs.rs/ash/0.38.0/ash/vk/struct.DebugUtilsMessengerCallbackDataEXT.html#method.message_id_name_as_c_str would this help?

Thanks, I very much would! I think that section is just copied from the old runner :D

eddyb · 2025-10-13T14:15:22Z

examples/runners/wgpu/src/lib.rs

-        // Hack: spirv_builder builds into a custom directory if running under cargo, to not
-        // deadlock, and the default target directory if not. However, packages like `proc-macro2`
-        // have different configurations when being built here vs. when building
-        // rustc_codegen_spirv normally, so we *want* to build into a separate target directory, to
-        // not have to rebuild half the crate graph every time we run. So, pretend we're running
-        // under cargo by setting these environment variables.
-        unsafe {
-            std::env::set_var("OUT_DIR", env!("OUT_DIR"));
-            std::env::set_var("PROFILE", env!("PROFILE"));
-        }


Have you checked that running the wgpu runner, after this change, doesn't result in spurious rebuilds every time? (it might only happen on debug or release, probably only on release if I had to guess)

AFAIK, without the OUT_DIR and PROFILE env vars, spirv-builder will not pass its custom --target-dir to Cargo, and therefore the same top-level target dir will get used for both the runner and the shaders, which can cause rebuilds due to build deps/proc macros (when their Cargo feature sets differ etc.).

But it's possible there could be other reasons why this isn't a problem anymore.

As of d416b65 from the original difftest PR #216, the env vars have been replaced with cargo_metadata to ensure we always grab the correct target dir. But they haven't propagated to all examples...

See diff: d416b65#diff-84c98121dbc8466f5e6e1503a3771f2ec4f8f37fa048ebf01c201b1b5cd1b191L935-L962

Oh, that's great, glad it's been fixed, would it be too much to ask for a separate PR for at least updating the wgpu example, as a nicer place to "document" the fact that this change happened?

Yea I should have probably split that into it's own PR to begin with... sure let's split c1d2b48 into it's own PR

I even mentioned it in the commit:

ash & wgpu runner: remove build script forwarding PROFILE env var

no longer necessary since d416b65

Pulled out the commit to #442, ~~you can now independently merge this PR from the new one~~

My bad, I found some extra stuff that #442 should probably remove, which caused conflicts and this PR now depends on that one.

no longer necessary since d416b65

* use Vulkan 1.3 * use DynamicRendering feature replacing Framebuffers * extract `MyDevice` struct containing vk initialization * decouple rendering from swapchain management * redo swapchain sync, fixes validation errors, intentionally kept basic * make rendering lazily recreate its pipeline

Firestar99 requested review from LegNeato, eddyb and schell as code owners September 30, 2025 14:20

Firestar99 force-pushed the simplify-ash-runner2 branch from 19159e1 to 0e10f07 Compare September 30, 2025 15:07

nnethercote reviewed Sep 30, 2025

View reviewed changes

examples/runners/ash/src/main.rs Outdated Show resolved Hide resolved

nnethercote reviewed Sep 30, 2025

View reviewed changes

examples/runners/ash/src/main.rs Outdated Show resolved Hide resolved

nnethercote reviewed Sep 30, 2025

View reviewed changes

examples/runners/ash/src/main.rs Outdated Show resolved Hide resolved

Firestar99 force-pushed the simplify-ash-runner2 branch from 0e10f07 to def7735 Compare October 1, 2025 10:23

Firestar99 mentioned this pull request Oct 1, 2025

ash runner: update ash-molten to my fork #421

Closed

Firestar99 force-pushed the simplify-ash-runner2 branch from 6597b17 to 3007d25 Compare October 1, 2025 13:11

MarijnS95 reviewed Oct 2, 2025

View reviewed changes

Firestar99 force-pushed the simplify-ash-runner2 branch from 2c6bab2 to fb61b79 Compare October 2, 2025 12:08

eddyb reviewed Oct 13, 2025

View reviewed changes

Firestar99 mentioned this pull request Oct 16, 2025

accept absolute target_dir_path and properly resolve relative ones with cargo_metadata #442

Open

Firestar99 force-pushed the simplify-ash-runner2 branch 2 times, most recently from d3c8635 to 5d13b21 Compare October 16, 2025 16:01

Firestar99 added 12 commits October 16, 2025 18:07

remove example build scripts forwarding PROFILE env var

51798a4

no longer necessary since d416b65

ash runner: make debug msgs easier to read

c00816c

ash runner: constant window updates, don't wait for new events

3a75912

ash runner: remove multimodule, simplify pipeline creation

51b5247

ash runner: use bytemuck

d73ab4d

ash runner: general code cleanup

4d01d44

ash runner: split it into different modules

94c1464

ash runner: fix lints

652b547

ash runner: remove ash-molten, require vulkan sdk on macos

a2fca2f

ash runner: switch Vulkan Memory Model from extension to core feature

0295dba

ash runner: cleanup debug callback

c9c689f

Firestar99 force-pushed the simplify-ash-runner2 branch from 5d13b21 to c9c689f Compare October 16, 2025 16:09

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

Rewrite the ash runner #419

Rewrite the ash runner #419

Uh oh!

Firestar99 commented Sep 30, 2025 •

edited

Loading

Uh oh!

Uh oh!

Uh oh!

Uh oh!

nnethercote commented Sep 30, 2025

Uh oh!

MarijnS95 Oct 2, 2025

Uh oh!

Firestar99 Oct 2, 2025

Uh oh!

eddyb Oct 13, 2025

Uh oh!

Firestar99 Oct 13, 2025 •

edited

Loading

Uh oh!

eddyb Oct 15, 2025

Uh oh!

Firestar99 Oct 16, 2025

Uh oh!

Firestar99 Oct 16, 2025 •

edited

Loading

Uh oh!

Firestar99 Oct 16, 2025

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

Rewrite the ash runner #419

Are you sure you want to change the base?

Rewrite the ash runner #419

Uh oh!

Conversation

Firestar99 commented Sep 30, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Requires #442

Reviews

Uh oh!

Uh oh!

Uh oh!

Uh oh!

nnethercote commented Sep 30, 2025

Uh oh!

MarijnS95 Oct 2, 2025

Choose a reason for hiding this comment

Uh oh!

Firestar99 Oct 2, 2025

Choose a reason for hiding this comment

Uh oh!

eddyb Oct 13, 2025

Choose a reason for hiding this comment

Uh oh!

Firestar99 Oct 13, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

eddyb Oct 15, 2025

Choose a reason for hiding this comment

Uh oh!

Firestar99 Oct 16, 2025

Choose a reason for hiding this comment

Uh oh!

Firestar99 Oct 16, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Firestar99 Oct 16, 2025

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

Firestar99 commented Sep 30, 2025 •

edited

Loading

Firestar99 Oct 13, 2025 •

edited

Loading

Firestar99 Oct 16, 2025 •

edited

Loading