Preact Benchmarks

A collection of benchmarks for Preact and its ecosystem of libraries.

The apps directory contains a bunch of apps that can be benchmarked. Each directory under an apps contains HTML files that are the benchmarks for that app. Inside an apps folders are implementations of that app using different Preact libraries (e.g. class components, hooks, compat, signals). The dependencies directory contains different versions of Preact and its ecosystem libraries. The cli directory contains a command line interface to run benchmarks. You can specify using options to the cli for this repository which implementation and dependency version to use for a benchmark.

Getting started

1. Clone this repository
2. Run pnpm install
3. Run pnpm start to start the benchmark server

Use the benchmark server to development/modify a benchmark implementation. The server will automatically reload the benchmark when you make changes files.

Running benchmarks

1. Run pnpm bench to run a benchmark
2. Follow the prompts to select which implementation and dependency versions to compare

Run pnpm bench --help to see all available options.

$ pnpm bench --help

  Description
    Run the given benchmark using the specified implementation with the specified dependencies.
    If no benchmark file, no dependencies, or no implementations are specified, will prompt for one.

  Usage
    $ preact-bench bench [benchmark_file] [options]

  Options
    --interactive        Prompt for options. Defaults to true of no benchmark file,
		                     dependencies, or implementations are specified  (default false)
    -d, --dependency     What group of dependencies (comma-delimited) and version to
		                     use for a run of the benchmark (package@version)  (default latest)
    -i, --impl           What implementation of the benchmark to run  (default preact-class)
    -n, --sample-size    Minimum number of times to run each benchmark  (default 25)
    -h, --horizon        The degrees of difference to try and resolve when auto-sampling
		                     ("N%" or "Nms", comma-delimited)  (default 5%)
    -t, --timeout        Maximum number of minutes to spend auto-sampling  (default 1)
    --trace              Enable performance tracing (Chrome only)  (default false)
    --debug              Enable debug logging  (default false)
    -b, --browser        Which browser to run the benchmarks in: chrome, chrome-headless,
		                     firefox, firefox-headless, safari, edge  (default chrome-headless)
    -p, --port           What port to run the benchmark server on  (default 5173)
    -h, --help           Displays this message

  Examples
    $ preact-bench bench
    $ preact-bench bench apps/todo/todo.html
    $ preact-bench bench apps/todo/todo.html -d preact@local -d preact@latest
    $ preact-bench bench apps/todo/todo.html -d preact@local -d preact@main -i preact-hooks
    $ preact-bench bench apps/todo/todo.html -d preact@local,signals@local -d preact@main,signals@local -i preact-signals -n 2 -t 0
    $ preact-bench bench apps/todo/todo.html -d preact@local -d preact@main --trace

Benchmarking within another repository

This repository is intended to be included as a submodule in another repository. This allows you to run benchmarks against local changes in that repository. The dev script in this repository starts a benchmarking dev server that is useful when benchmarking changes in another repository.

$ pnpm dev --help

Description
    Run a dev server to interactively run a benchmark while developing changes

  Usage
    $ preact-bench dev [benchmark_file] [options]

  Options
    --interactive        Prompt for options  (default false)
    -d, --dependency     What group of dependencies (comma-delimited) and version to
		                     use for a run of the benchmark (package@version)  (default latest)
    -i, --impl           What implementation of the benchmark to run  (default preact-class)
    -n, --sample-size    Minimum number of times to run each benchmark  (default 25)
    -h, --horizon        The degrees of difference to try and resolve when auto-sampling
		                     ("N%" or "Nms", comma-delimited)  (default 5%)
    -t, --timeout        Maximum number of minutes to spend auto-sampling  (default 1)
    --trace              Enable performance tracing (Chrome only)  (default false)
    --debug              Enable debug logging  (default false)
    -b, --browser        Which browser to run the benchmarks in: chrome, chrome-headless,
		                     firefox, firefox-headless, safari, edge  (default chrome-headless)
    -p, --port           What port to run the benchmark server on  (default 5173)
    -h, --help           Displays this message

  Examples
    $ preact-bench dev apps/todo/todo.html -d preact@local -d preact@main -i preact-hooks
    $ preact-bench dev apps/todo/todo.html -d preact@local -d preact@local-pinned -i preact-hooks

This command shares the same options as the bench command. Once you start the server you can press b⏎ to re-build your local Preact repository (or whatever repository this is within) and re-run the configured benchmarks.

$ pnpm dev apps/many-updates/many-updates.html -i preact -d preact@local -d preact@local-pinned -n 2 -t 0

> @preact/benchmarks@0.0.1 dev /Users/andre_wiggins/github/preactjs/preact-v10/benchmarks
> node cli/bin/preact-bench.js dev "apps/many-updates/many-updates.html" "-i" "preact" "-d" "preact@local" "-d" "preact@local-pinned" "-n" "2" "-t" "0"

  ➜  Local:   http://localhost:5173/
  ➜  Network: use --host to expose
  ➜  press p + enter Pin current local changes into local-pinned
  ➜  press b + enter run Benchmarks
  ➜  press h + enter show help

You can also press the p⏎ key to build your local repos changes and copy them into the relevant local-pinned directory. This command is useful when you want to compare different local changes against each other.