Wgpu Users

271 Members
GPU architecture: https://docs.google.com/presentation/d/1qi2j-SZuzew7Rrf5VKEPDZAQQEitV40k9fKvwJNyicM/edit?usp=drivesdk17 Servers

Load older messages

17 Apr 2021
@groves:matrix.orggroves You might just need to install the wasm target as the error mentions, so something like rustup target add wasm32-unknown-unknown 03:25:46
@groves:matrix.orggroves We should add that step to the wiki 03:26:27
@blakej11:matrix.orgBlake Jonesyes that sure works a lot better, thank you!03:29:49
@blakej11:matrix.orgBlake Jones

... although now it's giving an error when trying to compile a dependent library:

   Compiling js-sys v0.3.49
error: expected `,`
   --> /Users/blakej/.cargo/registry/src/github.com-1ecc6299db9ec823/js-sys-0.3.49/src/lib.rs:639:39
639 |         #[wasm_bindgen(js_namespace = Atomics, catch)]
    |                                       ^^^^^^^

@blakej11:matrix.orgBlake Jones(it has over 100 of these within that library)03:32:18
@blakej11:matrix.orgBlake Jones that was while trying to compile the gecko branch of wgpu-rs, as suggested by the wiki; dunno if that advice is also outdated at this point 03:34:16
@blakej11:matrix.orgBlake Jones switching back to master works 03:34:25
@kvark:matrix.orgkvarkgecko branch needs to be moved, likely04:11:40
@blakej11:matrix.orgBlake Jonesyeah I'm having much more luck with the master branch04:12:05
@blakej11:matrix.orgBlake JonesI have a thing building, and although it's not working it's giving me stack traces in the javascript console that do look like the relevant bits of source04:12:42
@blakej11:matrix.orgBlake Jones(just using some of the examples from wgpu-rs)04:12:53
@blakej11:matrix.orgBlake JonesI've been trying in Chrome Canary; are they more likely to work in Mozilla Nightly?04:13:06
@kvark:matrix.orgkvarkof course, yes, the examples are more likely to run in Firefox04:13:27
@blakej11:matrix.orgBlake Jonescool, downloading that now04:13:41
@blakej11:matrix.orgBlake Jones(I'm so used to things saying "oh this only works in Chrome", it's nice to have the failure mode be the other way)04:14:21
@kvark:matrix.orgkvarkfor wgpu-rs, it's natural because it either goes into wgpu directly, or it goes into WASM, which, after going through Firefox JS bindings and C++ glue, also ends up in wgpu.04:18:59
@blakej11:matrix.orgBlake Jonespoked the right config switches, and got it working! This is amazing and magical04:21:38
@blakej11:matrix.orgBlake Jonesincredible work y'all have done04:21:43
@kvark:matrix.orgkvark nice! So I guess we should just move gecko to master at this point 04:22:03
@blakej11:matrix.orgBlake Jones yeah, this was just using the master branch 04:22:21
@blakej11:matrix.orgBlake Jones (also, if you're editing the wiki, I needed to use wasm-bindgen version 0.2.73 to get it working) 04:24:57
@blakej11:matrix.orgBlake Jonesis there a way to get the wasm display region to scale along with the size of the browser window? at least for me it's appearing as a fixed 1024x768 box that doesn't change when I resize the window06:34:49
In reply to @kvark:matrix.org
are you planning on using WGSL?
Hmm the biggest issue I have with transitioning is that I have a bunch of shader include files that I reuse almost everywhere 🤔
In reply to @blakej11:matrix.org
is there a way to get the wasm display region to scale along with the size of the browser window? at least for me it's appearing as a fixed 1024x768 box that doesn't change when I resize the window

If you’re using winit then I think it’s blocked by https://github.com/rust-windowing/winit/issues/1661 currently.

If you’re not using winit, you can resize it as you would normally in JS/CSS (there are a few ways to do it)

@vollautomat:matrix.orgvollautomat joined the room.13:49:31
Download image.png

I am writing a little path tracer with wgpu-rs and have encountered a weird behavior.
About my setup:

I have two Textures A and B (both used as StorageTexture).

Render loop:

  • Copy texture content B to A
  • Execute compute shader:
    • Read from Buffer A
    • Write to Buffer B
  • Execute render pass:
    • Present buffer B to screen

This works fine when I use 8bit color depth (rgba8 / Rgba8Unorm).
As soon as I switch to 16bit color depth (rgba16 / Rgba16Float) I get
weird purple artifacts in the rendered image.
When I resize the window (recreate the textures A and B) the artifacts change
and are sometimes even completely gone.

I encountered this issue on Linux (with Vulkan backend and X window manager).
I also tested it on MacOs (with Metal backend) but could not observe the issue there.

Has anybody encountered sometimes similar or has an idea what causes this?
Also: is this the right chat for this question or should I post it in the dev room?

@kvark:matrix.orgkvarkvollautomat: looks like we may be missing a barrier. Please file an issue and attach an API trace15:02:35
18 Apr 2021
@alex179ohm:matrix.orgalex179ohm joined the room.05:40:20

There are no newer messages yet.

Back to Room List