GitHub - linebender/vello: A GPU compute-centric 2D renderer.

Update 2020-05-22: A new section on forward progress has been added, and the discussion of synchronized shuffles has been improved.

Today's The Fast and the Curious post covers the launch of Skia's new rasterization backend, Graphite, in Chrome on Apple Silicon Macs. Graphite is instrumental in helpi…

Xilem 2024 plans

Vello is a new GPU accelerated renderer for 2D graphics that relies heavily on compute shaders for its operation. (It was formerly known as piet-gpu, but we renamed it recently because it is no longer based on the [Piet] render context abstraction, and has been substantially rewritten). As such, it depends heavily on having good infrastructure for writing and running compute shaders, consistent with its goals of running portably and reliably across a wide range of GPU hardware. Unfortunately, there hasn’t been a great off-the-shelf solution for that, so we’ve spent a good part of the last couple years building our own hand-rolled GPU abstraction layer, piet-gpu-hal.

At WWDC, Apple introduced Metal shader converter, a tool for converting shaders from DXIL (the main compilation target of HLSL in DirectX12) to Metal. While it is no doubt useful for reducing the cost of porting games from DirectX to Metal, I feel it does not move us any closer to a world of robust GPU infrastructure, and in many ways just adds more underspecified layers of complexity.

The GPU in your computer is about 10 to 100 times more powerful than the CPU, depending on workload. For real-time graphics rendering and machine learning, you are enjoying that power, and doing those workloads on a CPU is not viable. Why aren’t we exploiting that power for other workloads? What prevents a GPU from being a more general purpose computer?

This document tries to establish patterns among a list of crates and Github projects using the Vello renderer.