Vello Documentation

Vello Sparse Strips is an experimental implementation of the Vello renderer designed to improve device compatibility and performance stability. It aims to run on GPUs without compute shader support (using only fragment and vertex shaders) and handle constrained memory conditions. It is based on the 'sparse rendering approach' which uses efficient tiling, sorting, and sparse strip allocation for both CPU and hybrid CPU/GPU workloads.

Warning: This implementation is not yet suitable for production use.

The vello_example_scenes crate is a collection of scene implementations designed for use with Vello Sparse Strips examples. It provides various rendering scenarios including:

• Basic shapes and paths
• Text rendering
• Image handling
• Gradients and blending
• SVG rendering
• Clipping operations
• COLR emoji stress testing

Vello Hybrid is a hybrid CPU/GPU renderer for 2D vector graphics. It balances performance by using the CPU for path processing and initial geometry setup, while leveraging the GPU for fast rendering, compositing, and blending.

The architecture consists of two primary components:

• Scene: Manages the render context and performs path processing on the CPU.
• Renderer or WebGlRenderer: Handles GPU resource management and executes draw operations.

Unlike many GPU 2D renderers that rely on fixed-function MSAA (Multi-Sample Anti-Aliasing), Vello uses a compute-centric approach where all pixels are generated by code.

Key Quality Features:

• Exact-area Anti-aliasing: Uses an approach similar to libart to prevent the stepping or graininess often seen with low-setting MSAA.
• Addressing Conflation Artifacts: The architecture allows for swapping soft-alpha compositing with supersampling to eliminate seams when compositing antialiased elements.
• Gamma Correctness: The vision includes addressing gamma-correct rendering and tone mapping to prevent text and hairline strokes from appearing weak or spindly.
• Subpixel Rendering: Potential support for per-channel alpha compositing to enable high-quality RGB subpixel rendering for text.

The vello_common crate provides shared data structures (paths, tiles, strips), geometry processing utilities, and common logic for rendering stages. It is designed as a foundation for vello_cpu and vello_hybrid to minimize duplication.

Note: This crate is not intended to be used on its own. You should instead use one of the renderers that depend on it. Currently, vello_cpu is the primary published renderer using this crate. If you require high-performance GPU-based 2D rendering, use the main vello crate (which does not use vello_common).

Vello aims to support a modern 2D imaging model (similar to PDF, SVG, and HTML Canvas) while leveraging GPU capabilities for features that are difficult to implement efficiently in standard APIs.

Planned/Explored Extensions:

• Custom Shaders/Effects: Using WebGPU as a portable way to provide extension points (similar to Direct2D's Custom Effects) for things like watercolor brush strokes.
• HDR Support: Implementing color transformations and tone mapping within the compositing pipeline.
• Efficient Blurs: Implementing specialized shaders for common UI elements, such as highly optimized shaders for blurred rounded rectangles, to avoid the overhead of large temporary buffer allocations.

Vello envisions offloading variable font rendering to the GPU. Because variable font technology relies on multiplying vectors of "deltas" with basis functions, it is highly suited for massively parallel GPU architectures.

Potential Benefits:

• Efficient Animation: Smoothly animating variable font axes.
• Layout Optimization: Adjusting font axes dynamically to improve paragraph justification or support complex scripts (e.g., Arabic kashida) without performance penalties.