🐉 Rust GPU

This is a very early stage project to make Rust a first-class language and ecosystem for building GPU code 🚀🚧

Current Status

Compiling a hello world triangle vulkan shader works, and a significant portion of the core library also compiles. However, many things aren't implemented yet: for example, branching isn't supported yet! That means that while being technically usable, this project is far from being production-ready.

Background

Historically in games GPU programming has been done through writing either HLSL, or to a lesser extent GLSL. These are simple programming languages that have evolved along with rendering APIs over the years. However, as game engines have evolved, these languages have failed to provide mechanisms for dealing with large codebases, and have generally stayed behind the curve compared to other programming languages.

In part this is because it's a niche language for a niche market, and in part this has been because the industry as a whole has sunk quite a lot of time and effort into the status quo. While over-all better alternatives to both languages exist, none of them are in a place to replace HLSL or GLSL. Either because they're vendor locked, or because they don't suppor the traditional graphics pipeline. Examples of this include CUDA and OpenCL. And while attempts have been made to create language in this space, none of them have gained any notable traction in the gamedev community.

Our hope with this project is that we push the industry forward by bringing an existing low-level, safe and high performance language to the GPU; that language being Rust. And with it come some additional benefits that can't be overlooked: a package/module system that's one of the industry's best, built in safety against race-conditions or out of bounds memory access etc.

Why Embark?

At Embark, we've been building our in-house engine from the ground up in Rust. We have previous in-house experience developing RLSL (a first prototype of this idea) and we have some of the world's best rendering engineers that are familiar with the problems in current shading languages. So, we're in a unique place to solve this problem. We want to streamline our own internal development, facilitate code-sharing between GPU and CPU, but most importantly: to enable our users to very rapidly build great looking experiences.

If we do this project right, one wouldn't necessarily need a team of rendering engineers to build a good looking game, instead one would simply use a few of the existing open-source crates that provide the graphical effects needed to create the experience you're after. Instead of sharing snippets of TAA code on forum posts, one could simply add the right crate(s).

Project scope

This project will involve a few things if we want to get the experience right, and it's quite broad. Initial stages will involve mostly just setting up the backend, however the project will be broader then that.

• Currently only SPIR-V support is planned (the open compiler target for Vulkan) but it's not unlikely that in future versions this will / should support DXIL (the target for DirectX) or WHLSL (the WebGPU shading language that's bijective with SPIR-V)
• The first and primary target will be Vulkan shaders (vertex, fragment, etc.). However, Vulkan compute (what spir-v calls GLCompute) is also on the roadmap, as well as OpenCL compute (what spir-v calls Kernel).
• Implement a rustc compiler backend, plugging in via -Z codegen-backend. This is the same mechanism that rustc_codegen_cranelift and rustc_codegen_gcc use.
• crates.io support to be able to publish SPIR-V crates
• An Embark-provided rendering / framegraph abstraction to take advantage of this and to make it easy for users to re-use rendering effects.

An in-depth exploration of our roadmap and milestones can be found here.

Process

We use this repo as a monorepo for everything related to the project: crates, tools, shaders, examples, tests, and design documents. This way, we can use issues and PRs covering everything in the same place, cross-reference stuff within the repo, as well as with other GitHub repos (rspirv/Rust/Ark).

We meet weekly over a discord call to discuss design and triage issues. Each meeting has an issue with agenda, links and minutes.

We have a #rust-gpu Discord channel for fast discussion and collaboration.

Getting started

There are a few different components to this repo:

• rfcs for in-depth discussion and specs.
• rustc_codegen_spirv for the compiler itself.
• rspirv-linker for the linker (used by the compiler).
• spirv-std for GPU intrinsics, types, and other library items used by GPU crates.
• spirv-builder for a convenient way of building a GPU crate in a CPU build.rs file.

To get started, first, we need to install some prerequisites. Nightly Rust is required for now. You may use the provided setup.sh/setup.bat scripts for this, or, just manually do what's in the scripts, which is:

rustup install nightly
rustup +nightly component add rust-src rustc-dev llvm-tools-preview

SPIRV-Tools is also required to be installed to your PATH (for now, eventually we will automatically build and link it instead of calling executables)

Note the rust-toolchain file in this repository that specifies nightly, this is equivalent to passing cargo +nightly build, without having to type that out. (If you run that rustup component add in this directory, you don't need to pass +nightly either)

Next, look at the examples folder. There are two projects here: examples/example-shader and examples/example-runner. The example-shader project is a "GPU crate", one that will be compiled to a spir-v module. The example-runner project is a normal, CPU crate that uses vulkan to consume the example-shader spir-v module to display a "hello world" triangle.

Run the example!

cargo run --bin example-runner

This will build rustc_codegen_spirv, the compiler, then use that compiler to build example-shader into a spir-v module, then finally, build a vulkan sample app (taken from ash's examples) using the built spir-v module to display a triangle in a window.

All of this is orchestrated by the spirv-builder crate, which is used in example-runner's build.rs file. Please look at that file, as well as both example projects in general, to see how to set up your own shaders!

Be aware that this project is in a very early phase - if the above doesn't work, please file an issue!

Getting started, for power users who don't want to use spirv-builder.

If you would like to build the compiler, rustc_codegen_spirv is the relevant folder. Install the prerequisites, as above, then, cd rustc_codegen_spirv && cargo build. This produces an .so file, located at ./target/debug/librustc_codegen_spirv.so (or .dll/.dylib depending on your platform).

This file is a dynamically loaded backend for rustc - you may tell rustc to use it as a backend through the -Z codegen-backend=... flag. To pass this to rustc through cargo, set the environment variable RUSTFLAGS="-Z codegen-backend=$PATH_TO_FILE".

Then, when building a GPU crate, we need to configure some flags when we call cargo. First, we need to build libcore ourselves - we obviously have no spir-v libcore installed on our system! Use the flag -Z build-std=core. Then, we need to tell rustc to generate spir-v instead of x86 code: --target spirv-unknown-unknown.

Overall, building your own spir-v crate looks like:

export RUSTFLAGS="-Zcodegen-backend=$THIS_REPO/target/debug/librustc_codegen_spirv.so"
cargo build -Z build-std=core --target spirv-unknown-unknown --release

(with an appropriate path for $THIS_REPO, and replacing export with set if you're on windows as well as the proper dll name)

This will produce a target/spirv-unknown-unknown/release/crate_name.spv file.

To create a GPU crate, look at the examples/example-shader crate. In short, reference the spirv-std crate, and use intrinsics defined there to create your shader.

This is all a little convoluted, hence the spirv-builder crate handles a lot of this.

Contributing

We welcome community contributions to this project.

Please read our Contributor Guide for more information on how to get started.

License

Licensed under either of

• Apache License, Version 2.0, (LICENSE-APACHE or http://www.apache.org/licenses/LICENSE-2.0)
• MIT license (LICENSE-MIT or http://opensource.org/licenses/MIT)

at your option.

Contribution

Unless you explicitly state otherwise, any contribution intentionally submitted for inclusion in the work by you, as defined in the Apache-2.0 license, shall be dual licensed as above, without any additional terms or conditions.