Benchmarks

These are timings of prefix for Strauss PHP and Strauss-RS on four projects of different sizes. They were taken on one machine, so treat them as a guide to the size of the difference. Your results will vary with your hardware, your project and your disk.

Results

Each time is the median of several runs: seven for the small and medium projects, three for the large and extra large ones.

Project	Workers	Strauss PHP	Strauss-RS	Speedup	PHP peak memory	RS peak memory
small	1	625.4 ms	22.8 ms	27.4x	70 MiB	9.5 MiB
small	12	831.3 ms	9.1 ms	91.0x	411 MiB	11.7 MiB
medium	1	1.11 s	46.5 ms	24.0x	83 MiB	13.1 MiB
medium	12	1.38 s	16.2 ms	85.5x	599 MiB	19.5 MiB
large	1	18.72 s	761.3 ms	24.6x	259 MiB	81.6 MiB
large	12	16.84 s	188.7 ms	89.3x	1,291 MiB	97.7 MiB
xlarge	1	42.26 s	1.50 s	28.2x	990 MiB	336.2 MiB
xlarge	12	55.05 s	395.6 ms	139.1x	5,556 MiB	405.4 MiB

• Small: monolog
• Medium: guzzle + flysystem + finder
• Large: laravel/framework
• X-Large: aws-sdk-php

How they were run

• Wall time for prefix command
• Median of 5 runs after 1 warm-up, from a clean output directory each time.
• Both tools were given the worker count explicitly (1 against 1, 12 against 12).
• Memory is peak total across all processes, sampled on Linux.
• Strauss PHP 0.31.1 on PHP 8.4.23 (static CLI build, no OPcache, no memory limit).
• Strauss-RS release build.
• Intel i5-10600 (12 threads), 32 GB RAM.
• All projects with locked dependencies.

What to keep in mind

• These measure the prefix step. A Composer install and your own build steps take the same time with either tool.

How to reproduce benchmarks

Clone the repo from GitHub, then run the following commands

cargo run --release --features benchmark -- benchmark setup
cargo run --release --features benchmark -- benchmark run

Setup downloads the pinned Strauss PHP 0.31.1 phar and runs composer install for each project. run has these options:

Option	Description
--project <NAME>	Only run this project. Can be given more than once. By default every project in benchmark.json is run.
--tool <rs|php>	Only run this tool. Can be given more than once. By default both are run.
--runs <N>	How many timed runs of each. The default is 5.
--warmup <N>	How many runs to throw away first. The default is 1.
--workers <N>	How many workers in the runs with many. Must be 2 or more. The default is one per CPU.
--php <PATH>	The PHP binary that runs Strauss PHP and checks the output. The default is php from your PATH.
--json <FILE>	Write the results, and what they were measured on, to this JSON file.
--keep	Leave the last output in place, instead of deleting it.
--no-check	Don't parse the output with PHP.
--no-tree-memory	Don't take the extra run that adds up the memory of every process.
--allow-debug	Run even though this is a debug build. The times mean nothing in a debug build.