Shader Question

[lonnietc]

I am learning and experimenting with Shaders and actually came across your Rust-GPU-ShaderToys project: https://github.com/Rust-GPU/rust-gpu-shadertoys for which I have been able to compile and run on my Windows 11 (x64) testbed.

I am seeking to develop a type of 3D voxel SDF compression engine that would be written heavily for shaders but I am not clear on some things.

Does Rust-GPU just allow for writing the shader and then compiling it to run, or can you dynamically modify the shader from Rust. Not sure if this is clear, but the idea is that I need to be able to dynamically add/remove and modify shaders to a running shader, if that is even possible.

Any thoughts?

Is there a mailing list or Discord channel as well?

Thanks and have a great day

[schell]

Hi @lonnietc :)

Currently Rust-GPU only supports compiling your Rust shaders, there's no support for runtime modification. As far as I know, modifying shaders always requires "recompiling" them, but what counts as "recompiling" is different depending on the runtime.

Do you have a platform or runtime in mind that provides this feature?

[Firestar99]

To elaborate on "dynamically modify the shader" a bit more:

• We are a shader compiler, we just turn your rust code into spirv binaries, what you do with them is up to you and what API you're using.
• Most APIs (wgpu, vulkan) you create a "Pipeline" using shaders loaded from the "spirv blob", and then use it to submit your draws or dispatches. Which allows you to just create a new pipeline to replace the existing one, and render your next frame or whatever with the new pipeline.
  • We have a watch API that notifies you when source code changed, so you can recreate your pipelines. It works like an std::sync::mpsc::Reciever with recv() and try_recv() functions.
  • Compiling at runtime means that the device needs to have rust and rust-gpu dependencies installed. Whereas compiling in a build script and include_bytes!() the spirv blob allows you to run anywhere you can compile to.
• You can also use "specialization constants" (in Vulkan, I know wgpu works on support, unsure about it's state) to specify some constants within your shader when you create the pipeline, at runtime, instead of at compile time.

[lonnietc]

Morning All,

Truly appreciate the discussion and clarification on these questions as I am still in the learning curve for a number of areas but things are crystalizing more each day.

I am also impressed a lot with the Rust-GPU-ShaderToys project and will try to use and learn from it as well. I must say that it seems to handle many concurrent running shaders extremely well as I ran, as a quick test, 3 instances of it and estimated that there were something like 84 shaders running at one time and it looks like it could go further as well. Nice.

I think that we can close this ticket now and have a great day, my friends.

[Firestar99]

As a side note, just last week I started on a new template you can use with cargo generate --git https://github.com/Rust-GPU/rust-gpu-template

And there is an unofficial discord: https://discord.gg/8XAZaAZB