ChartGPU is an open-source WebGPU-based charting library that enables smooth panning and zooming even with 1 million data points.



Existing chart libraries have the drawback of not working smoothly as the number of data points increases. To address this issue, ChartGPU , a

WebGPU -based open source chart library, was developed.

GitHub - ChartGPU/ChartGPU: Beautiful, open source, WebGPU-based charting library
https://github.com/ChartGPU/ChartGPU

A chart library is a program package for displaying data as visual graphs or charts on websites and applications. There are many chart libraries available, but most of them become increasingly slow as the number of data points increases. Therefore, ChartGPU is a chart library built on WebGPU to process data using the GPU instead of the CPU.

ChartGPU can create line charts, area charts, bar charts, scatter plots, pie charts, candlestick charts, etc. For example, candlestick charts can render elements such as OHLC (Open, High, Low, Close) with different colors.



Supported browsers are Chrome 113 or later, Edge 113 or later, and Safari 18 or later. Firefox does not support WebGPU, so it is not supported at the time of writing.

There is also a ChartGPU demo page where you can see demos of graphs like pie charts and candlestick charts , as well as a demo of a graph with 1 million data points, demonstrating smooth operation.

Million Points - ChartGPU
https://chartgpu.github.io/ChartGPU/examples/million-points/



ChartGPU's creator, huntergemmer , wrote on Hacker News , a social message board, 'I developed ChartGPU because I've repeatedly run into the same wall. Chart libraries claim to be 'fast,' but they just can't keep up with more than 100,000 data points. Canvas2D is basically CPU-dependent, and even WebGL chart libraries perform most of their calculations on the CPU. ChartGPU uses WebGPU to move all calculations to the GPU. Largest Triangle Three Buckets (LTTB) downsampling is performed as a compute shader. Tooltips and hover hit tests are GPU-accelerated. Rendering uses instanced drawing. This resulted in a chart library that can pan and zoom smoothly at 60 fps even with 1 million data points.' He explained the reason for developing ChartGPU.

Related Posts:

in Software, Posted by logu_ii