Try the benchmarks app online.

To run the benchmarks locally:

yarn benchmarks
open ./packages/benchmarks/dist/index.html

Develop against these benchmarks:

yarn compile --watch
yarn benchmarks --watch

These benchmarks are approximations of extreme cases that libraries may encounter. Their purpose is to provide an early-warning signal for performance regressions. Each test report includes the mean and standard deviation of the timings, and approximations of the time spent in scripting (S) and layout (L).

The components used in the render benchmarks are simple enough to be implemented by multiple UI or style libraries. The benchmark implementations and the features of the style libraries are only approximately equivalent in functionality.

No benchmark will run for more than 20 seconds.

These cases look at the performance of mounting and rendering large trees of elements that use static styles.

This case looks at the performance of repeated style updates to a large mounted tree. Some libraries choose to inject new styles for each "dynamic style", whereas others choose to use inline styles. Libraries without built-in support for dynamic styles (i.e., they rely on user-authored inline styles) are not included.

MacBook Pro (13-inch, Early 2011); 2.3 GHz Intel Core i5; 8 GB 1333 MHz DDR3 RAM. Google Chrome 63.

Typical render timings: mean ± standard deviations.

Moto G4 (Android 7); Octa-core (4x1.5 GHz & 4x1.2 Ghz); 2 GB RAM. Google Chrome 63.

Typical render timings: mean ± standard deviations.