terminal protocol sdk internals performance
git license
tests glitch

2D software rasterizer tournament

This benchmark was done on a AMD Ryzen AI 7 445 core locked at 2.0Ghz

built using clang, cairo and qt6 from debian, blend2d statically linked. ctx dynamically linked.

Last updated:5.July 2026

Pixel formats

This shows performance for doing color fills of circles without and with alpha==1.0. The native formats of ctx, BGRA8, GRAYA8 GRAYAF RGBAF and CMYKAF are performing well, for some others like RGB565 shortcuts have been added to increase performance. The numbers and bar graphs are render calls per millisecond.

The data for the RGBA8 time-trial benchmark is a blend2d-bench dataset, the numbers in the tables below are render calls per milliseconds to render 256x256 bound shapes in the specified fill style - The data also can be visualized in more detail Blend2D's performance comparison page by using the custom dataset functionality.

	🥇	🥈	🥉
Blend2D	259	58	6	1
ctx	61	226	34	3
Qt6	1	27	154	142
cairo	3	13	130	178

The medal summary provides the overall data from the dataset at a glance. In the following summary we group the tests by how ctx is performing.

Heats won by ctx:

ctx in second place

ctx in third place

ctx in fourth place