Vello CPU

CPU-based renderer

Vello CPU is a 2D graphics rendering engine written in Rust, for devices with no or underpowered GPUs.

We also develop Vello, which makes use of the GPU for 2D rendering and has higher performance than Vello CPU. Vello CPU is being developed as part of work to address shortcomings in Vello.

Usage

To use Vello CPU, you need to:

• Create a RenderContext, a 2D drawing context for a fixed-size scene area.
• For each object in your scene:
  • Set how the object will be painted, using set_paint.
  • Set the shape to be drawn for that object, using methods like fill_path, stroke_path, or glyph_run.
• Render it to an image using RenderContext::render.

use vello_cpu::{RenderContext, Resources, Pixmap};
use vello_cpu::{color::{palette::css, PremulRgba8}, kurbo::Rect};
let width = 10;
let height = 5;
let mut context = RenderContext::new(width, height);
let mut resources = Resources::new();
context.set_paint(css::MAGENTA);
context.fill_rect(&Rect::from_points((3., 1.), (7., 4.)));

let mut target = Pixmap::new(width, height);
// While calling `flush` is only strictly necessary if you are rendering using
// multiple threads, it is recommended to always do this.
context.flush();
context.render(&mut target, &mut resources);

let expected_render = b"\
    0000000000\
    0001111000\
    0001111000\
    0001111000\
    0000000000";
let magenta = css::MAGENTA.premultiply().to_rgba8();
let transparent = PremulRgba8 {r: 0, g: 0, b: 0, a: 0};
let mut result = Vec::new();
for pixel in target.data() {
    if *pixel == magenta {
        result.push(b'1');
    } else if *pixel == transparent {
        result.push(b'0');
    } else {
         panic!("Got unexpected pixel value {pixel:?}");
    }
}
assert_eq!(&result, expected_render);

See the examples for more complete demonstrations of Vello CPU's API.

Features

• std (enabled by default): Get floating point functions from the standard library (likely using your target's libc).
• libm: Use floating point implementations from libm.
• png(enabled by default): Allow loading Pixmaps from PNG images. Also required for rendering glyphs with an embedded PNG. Implies std.
• multithreading: Enable multi-threaded rendering. Implies std.
• text (enabled by default): Enables glyph rendering (glyph_run).
• u8_pipeline (enabled by default): Enable the u8 pipeline, for speed focused rendering using u8 math. The u8 pipeline will be used for OptimizeSpeed, if both pipelines are enabled. If you're using Vello CPU for application rendering, you should prefer this pipeline.
• f32_pipeline: Enable the f32 pipeline, which is slower but has more accurate results. This is espectially useful for rendering test snapshots. The f32 pipeline will be used for OptimizeQuality, if both pipelines are enabled.

At least one of std and libm is required; std overrides libm. At least one of u8_pipeline and f32_pipeline must be enabled. You might choose to disable one of these pipelines if your application won't use it, so as to reduce binary size.

Current state

Vello CPU is a solid CPU-only 2D renderer with broad, reliable feature support. It provides excellent performance across a wide range of workloads, with optimized SIMD implementations for all major architectures. The renderer is still under active development, however, and a few limitations remain:

• Complex filter graphs are currently not supported at all and will panic. In multi-threaded mode, even simple filters are currently unsupported.
• Parts of the API and its documentation are still suboptimal, for example the Resources lifecycle.
• Some exposed features remain experimental and are not recommended for use, including glyph caching. Experimental APIs are identified in their method documentation.
• There is still more room for performance improvements, in particular on x86 systems and also for multi-threaded rendering.

With that said, we are continuously improving Vello CPU and will address these and other limitations in future releases.

Performance

Performance benchmarks can be found here, As can be seen, Vello CPU achieves compelling performance on both, aarch64 and x86 platforms. We also have SIMD optimizations for WASM SIMD, meaning that you can expect good performance there as well.

Implementation

If you want to gain a better understanding of Vello CPU and the sparse strips paradigm, you can take a look at the accompanying master's thesis that was written on the topic. Note that parts of the descriptions might become outdated as the implementation changes, but it should give a good overview nevertheless.

Minimum supported Rust Version (MSRV)

This version of Vello CPU has been verified to compile with Rust 1.88 and later.

Future versions of Vello CPU might increase the Rust version requirement. It will not be treated as a breaking change and as such can even happen with small patch releases.

As time has passed, some of Vello CPU‘s dependencies could have released versions with a higher Rust requirement. If you encounter a compilation issue due to a dependency and don’t want to upgrade your Rust toolchain, then you could downgrade the dependency.

# Use the problematic dependency's name and version
cargo update -p package_name --precise 0.1.1

Community

Discussion of Vello CPU development happens in the Linebender Zulip, specifically the #vello channel. All public content can be read without logging in.

Contributions are welcome by pull request. The Rust code of conduct applies.

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.