Add support for transient textures on Vulkan and Metal #8247

opstic · 2025-09-19T08:12:16Z

Connections
Fixes #8239

Description
Exposes MTLStorageMode.memoryless for Metal and vk::MemoryPropertyFlags::LAZILY_ALLOCATED for Vulkan by adding TextureUsages::TRANSIENT.

Testing
Added corresponding TextureUsage flag to relevant examples.
Also added some render tests and validation testing.

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.

cwfitzgerald

Thanks for taking this on! It just so happened we just talked about wanting this in the WebGPU face to face, so super happy to have this! There's a few things that need to be added, but I don't think it will end up being too much work:

At the face to face, we've decided that transient textures wouldn't be a feature. It would be available everywhere, but a no-op on unsupported platforms. If we wanted to expose to the user if this is supported, we can do this in the AdapterInfo as a hint.
We need two pieces of validation:
a. If you add the transient attachment usage, the usage must be exactly RENDER_ATTACHMENT | TRANSIENT
b. When writing a renderpass to these attachments, the StoreOp must be discard.
With point 1 implemented, we should stick these flags on as many of the examples as relevant as we really want to encourage its use.
On VK, we need to add VK_MEMORY_PROPERTY_LAZILY_ALLOCATED_BIT to the memory so that it may be lazily allocated. I think you may be getting this already implicitly without explicitly asking for it, if you're seeing memory improvements.
We should add some tests
a. For all the validation I mentioned, these can be simple tests in wgpu-validation.
b. We should have at least one test in the wgpu-gpu tests which ensures we don't get validation errors from vk/metal.

Let me know if you have any questions or if any of this is unclear!

wgpu-hal/src/vulkan/device.rs

…Y_ALLOCATED` and applies it

opstic · 2025-09-20T07:09:13Z

Thanks for your input @cwfitzgerald!

For:

I added some validation and error types specific to the situations you mentioned, please check it again!
The most relevant looks to only be the msaa_line example, I looked through all of the other examples and it appears they don't use multisampling at all. I'll be adding it now.
Something I found was that none of the desktop GPUs I have supports memory types with LAZILY_ALLOCATED, the memory decrease must just have been run-to-run difference😭, I'll be proceeding by testing more carefully and on mobile from now on.

cwfitzgerald · 2025-09-20T07:51:15Z

Re 3: shadow can use it for the forward depth buffer, and I suspect there might be an extra example or two which only ever uses depth in a memoryless compatible way.

cwfitzgerald · 2025-09-20T07:52:34Z

Something I found was that none of the desktop GPUs I have supports memory types with LAZILY_ALLOCATED, the memory decrease must just have been run-to-run difference😭, I'll be proceeding by testing more carefully and on mobile from now on.

Yeah this is unfortunately a mobile only feature really, but hopefully the validation layers can save us on desktop. Apple is the main one that really benefits from it on "desktop".

opstic · 2025-09-20T07:56:35Z

Yeah this is unfortunately a mobile only feature really, but hopefully the validation layers can save us on desktop. Apple is the main one that really benefits from it on "desktop".

I did make it so that a if a memory type with the LAZILY_ALLOCATED bit was not found, it would just proceed normally with the TextureUsages::TRANSIENT being disregarded, just making sure that this behavior should be the intended one.

cwfitzgerald

Some comments and needs tests, but moving in the right direction!

wgpu-core/src/command/render.rs

wgpu-types/src/lib.rs

opstic · 2025-09-20T08:02:33Z

Re 3: shadow can use it for the forward depth buffer, and I suspect there might be an extra example or two which only ever uses depth in a memoryless compatible way.

Alright, I looked through it more carefully again, we should be able to apply TRANSIENT on msaa_line's resolve texture and shadow, skybox's depth texture. Adding them now.

Co-authored-by: Connor Fitzgerald <[email protected]>

Renders a single triangle with transient texture as target to make sure nothing is wrong

cwfitzgerald · 2025-09-21T20:33:08Z

Confirmed that this is working as expected on mac - will do another review.

cwfitzgerald · 2025-09-24T03:35:33Z

(I'm doing this to see how it does, I'll do a full human review tomorrow, no action is needed based on its comments).

Copilot

Pull Request Overview

This PR adds support for transient textures on Vulkan and Metal backends by introducing TextureUsages::TRANSIENT. This optimization allows textures to use MTLStorageMode.memoryless on Metal and VK_MEMORY_PROPERTY_FLAG_BITS_LAZILY_ALLOCATED on Vulkan, potentially reducing memory usage and bandwidth for textures that aren't reused between passes.

Key changes include:

Added TextureUsages::TRANSIENT flag with validation requiring StoreOp::Discard and incompatibility with other usages
Extended backend support detection via AdapterInfo.supports_transient
Updated examples to demonstrate transient texture usage for depth and MSAA textures

Reviewed Changes

Copilot reviewed 25 out of 25 changed files in this pull request and generated 3 comments.

Show a summary per file

File	Description
wgpu-types/src/lib.rs	Added TRANSIENT texture usage flag, validation, and adapter support field
wgpu-hal/src/vulkan/	Implemented Vulkan lazily allocated memory support for transient textures
wgpu-hal/src/metal/	Added Metal memoryless storage mode support for transient textures
wgpu-core/src/	Added validation logic and error handling for transient texture usage
tests/	Added validation tests and GPU render tests for transient texture functionality
examples/	Updated depth and MSAA texture usage to include TRANSIENT flag

_{Tip: Customize your code reviews with copilot-instructions.md. Create the file or learn how to get started.}

wgpu-types/src/lib.rs

wgpu-hal/src/vulkan/device.rs

tests/tests/wgpu-validation/api/texture.rs

cwfitzgerald

Looking good, I thought of one more potential problem:

We need to validate out creating a texture with only TRANSIENT usage. I'm not sure what metal would do with this, but it can't be good 😆

One final thing, you should do a pass over the backends and make sure transient is advertised https://github.com/gfx-rs/wgpu/blob/trunk/wgpu-hal/src/lib.rs#L670-L673 wherever RENDER_ATTACHMENT is advertised.

Thanks for being so helpful with this process, we should be almost finished!!

tests/tests/wgpu-validation/api/texture.rs

wgpu-types/src/lib.rs

cwfitzgerald · 2025-09-25T03:46:40Z

wgpu-hal/src/noop/mod.rs

        driver: String::from("wgpu"),
        driver_info: String::new(),
        backend: wgt::Backend::Noop,
+        supports_transient: true,


Suggested change

supports_transient: true,

supports_transient: false,

See the change in the doc comment on this field for the doc comment

This needs to be updated here too

cwfitzgerald · 2025-09-25T03:49:11Z

wgpu-types/src/lib.rs

    /// Backend used for device
    pub backend: Backend,
+    /// Supports [`TextureUsages::TRANSIENT`]
+    pub supports_transient: bool,


Suggested change

pub supports_transient: bool,

pub native_transient_texture: bool,

Really not set on this name, but what I'm trying to avoid is the implication that TRANSIENT is not allowed

This still needs to be updated

Oops, been pretty busy recently. I should be able to finish the rest of the changes this weekend. Sorry for the wait.

No worries or rush! Just re-request a review from me whenever you're ready!

Co-authored-by: Connor Fitzgerald <[email protected]>

…'re using

JMS55 · 2025-10-02T15:43:38Z

What's the use case for this?

One thing I do in Bevy is use a dummy render target for a render pass, since I use storage atomics for my output. I have wanted render passes with no render attachments for this reason, so that we don't need to waste memory on a dummy render target. Would this be a suitable replacement?

cwfitzgerald · 2025-10-02T17:21:51Z

What's the use case for this?

Tilers where you need an attachment but never actually read from it from memory. You can avoid allocating that memory.

One thing I do in Bevy is use a dummy render target for a render pass, since I use storage atomics for my output. I have wanted render passes with no render attachments for this reason, so that we don't need to waste memory on a dummy render target. Would this be a suitable replacement?

This is better than nothing, but not really. It would only actually reduce memory on tilers.

opstic added 3 commits September 19, 2025 03:48

Preliminary implementation

56d648a

Merge branch 'gfx-rs:trunk' into transient

8c7dace

oops

598259c

opstic changed the title ~~Transient~~ Add support for transient textures on Vulkan and Metal Sep 19, 2025

opstic marked this pull request as draft September 19, 2025 08:27

Better implementation

9ec9242

opstic marked this pull request as ready for review September 19, 2025 09:03

Add entry in CHANGELOG.md

2185cd8

opstic force-pushed the transient branch from e4ca7ea to 2185cd8 Compare September 19, 2025 09:34

cwfitzgerald requested changes Sep 19, 2025

View reviewed changes

cwfitzgerald reviewed Sep 19, 2025

View reviewed changes

wgpu-hal/src/vulkan/device.rs Outdated Show resolved Hide resolved

opstic added 3 commits September 20, 2025 01:38

Remove Feature::TRANSIENT_ATTACHMENTS and add validation

625bf3b

Remove #[cfg(windows)] from find_memory_type_index

5644cd2

Tries to find a memory type that contains `MemoryPropertyFlags::LAZIL…

6c3d11b

…Y_ALLOCATED` and applies it

opstic force-pushed the transient branch from deaab97 to 6c3d11b Compare September 20, 2025 06:49

Merge branch 'trunk' into transient

a16ec0c

opstic added 2 commits September 20, 2025 03:48

cargo fmt

aafa05e

Add check in Metal for MTLStorageMode::Memoryless support

703bf26

cwfitzgerald requested changes Sep 20, 2025

View reviewed changes

wgpu-core/src/command/render.rs Outdated Show resolved Hide resolved

wgpu-types/src/lib.rs Outdated Show resolved Hide resolved

wgpu-types/src/lib.rs Outdated Show resolved Hide resolved

opstic and others added 4 commits September 20, 2025 04:03

Apply TextureUsages::TRANSIENT to relevant examples

02201f4

Improve documentation by @cwfitzgerald

c1c04ec

Co-authored-by: Connor Fitzgerald <[email protected]>

Improve error message by @cwfitzgerald

77bf9af

Co-authored-by: Connor Fitzgerald <[email protected]>

Move TextureUses::TRANSIENT to 1 << 12

faea1c3

opstic force-pushed the transient branch from 57a50b0 to fc854d7 Compare September 20, 2025 08:21

Improve TextureUsages docs

56b3956

opstic added 4 commits September 21, 2025 04:53

Formatting

1a09f3c

Fix WebGPU compile

0e72950

Add simple transient test case

3d063d7

Renders a single triangle with transient texture as target to make sure nothing is wrong

Actually use the transient flag in test

69d8110

opstic force-pushed the transient branch from 76e9122 to 69d8110 Compare September 21, 2025 11:15

Merge branch 'trunk' into transient

5b0eb92

opstic requested a review from cwfitzgerald September 21, 2025 11:16

Well that was a nice demonstration of validation

4598ebd

Update CHANGELOG.md

dfb60de

cwfitzgerald requested a review from Copilot September 24, 2025 03:35

Copilot AI reviewed Sep 24, 2025

View reviewed changes

wgpu-types/src/lib.rs Outdated Show resolved Hide resolved

wgpu-hal/src/vulkan/device.rs Show resolved Hide resolved

tests/tests/wgpu-validation/api/texture.rs Show resolved Hide resolved

opstic added 2 commits September 23, 2025 23:46

Use proper formatting for documentation

8182e41

Merge branch 'trunk' into transient

0124879

cwfitzgerald self-assigned this Sep 24, 2025

Merge branch 'trunk' into transient

7af040e

cwfitzgerald requested changes Sep 25, 2025

View reviewed changes

opstic and others added 10 commits September 25, 2025 00:29

Improve validation usage filtering by @cwfitzgerald

b6e05b1

Co-authored-by: Connor Fitzgerald <[email protected]>

Improve documentation by @cwfitzgerald

1332b31

Co-authored-by: Connor Fitzgerald <[email protected]>

Improve documentation by @cwfitzgerald

88b5cd4

Co-authored-by: Connor Fitzgerald <[email protected]>

Limit supports_memoryless_storage to Apple Silicon

99ff738

Merge remote-tracking branch 'fork/transient' into transient

126e45b

cargo fmt

305a8fa

Move TextureUsages::TRANSIENT in guaranteed_format_features

a3d8e86

Merge branch 'trunk' into transient

584f4f7

Remove unnecessary mut

193a138

MTLGPUFamily::Metal4 doesn't exist yet in the version on metal we…

440964f

…'re using

	pub supports_transient: bool,
	pub native_transient_texture: bool,

Add support for transient textures on Vulkan and Metal #8247

Are you sure you want to change the base?

Add support for transient textures on Vulkan and Metal #8247

Uh oh!

Conversation

opstic commented Sep 19, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

cwfitzgerald left a comment • 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!

Uh oh!

opstic commented Sep 20, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

cwfitzgerald commented Sep 20, 2025

Uh oh!

cwfitzgerald commented Sep 20, 2025

Uh oh!

opstic commented Sep 20, 2025

Uh oh!

cwfitzgerald left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

opstic commented Sep 20, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

cwfitzgerald commented Sep 21, 2025

Uh oh!

cwfitzgerald commented Sep 24, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull Request Overview

Reviewed Changes

Uh oh!

Uh oh!

Uh oh!

Uh oh!

cwfitzgerald left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

cwfitzgerald Sep 25, 2025

Choose a reason for hiding this comment

Uh oh!

cwfitzgerald Oct 1, 2025

Choose a reason for hiding this comment

Uh oh!

cwfitzgerald Sep 25, 2025

Choose a reason for hiding this comment

Uh oh!

cwfitzgerald Oct 1, 2025

Choose a reason for hiding this comment

Uh oh!

opstic Oct 1, 2025

Choose a reason for hiding this comment

Uh oh!

cwfitzgerald Oct 1, 2025

Choose a reason for hiding this comment

Uh oh!

JMS55 commented Oct 2, 2025

Uh oh!

cwfitzgerald commented Oct 2, 2025

Uh oh!

Uh oh!

opstic commented Sep 19, 2025 •

edited

Loading

cwfitzgerald left a comment •

edited

Loading

opstic commented Sep 20, 2025 •

edited

Loading

opstic commented Sep 20, 2025 •

edited

Loading

cwfitzgerald commented Sep 24, 2025 •

edited

Loading