Awesome

Parser Benchmark for Oxc, Swc and Biome

The purpose of this benchmark is for people who wants to evaluate and compare the performance characteristics of these parsers.

The numbers indicate that Oxc is at least 3 times faster than Swc and 5 times faster than Biome.

CPU

Codspeed Measurement

Codspeed measures performance by cpu instructions.

Mac i7 6 cores

Mac mini M2 8 cores

cal.com.tsx

	oxc	swc	biome
no-drop	`4.0 ms` (1.00x)	`14.0 ms` (3.50x)	`18.7 ms` (4.68x)
parallel	`7.4 ms` (1.00x)	`26.4 ms` (3.56x)	`38.7 ms` (5.21x)
single-thread	`4.0 ms` (1.00x)	`14.9 ms` (3.73x)	`20.1 ms` (5.04x)

typescript.js

	oxc	swc	biome
no-drop	`30.3 ms` (1.00x)	`100.3 ms` (3.31x)	`149.7 ms` (4.94x)
parallel	`52.0 ms` (1.00x)	`164.5 ms` (3.17x)	`296.5 ms` (5.71x)
single-thread	`29.8 ms` (1.00x)	`108.0 ms` (3.62x)	`159.5 ms` (5.34x)

MacBook Pro M3 Max

cal.com.tsx

	oxc	swc	biome
no-drop	`3.4 ms` (1.00x)	`13.4 ms` (3.99x)	`16.7 ms` (4.97x)
parallel	`5.8 ms` (1.00x)	`23.8 ms` (4.14x)	`30.1 ms` (5.23x)
single-thread	`3.4 ms` (1.00x)	`14.4 ms` (4.28x)	`18.3 ms` (5.42x)

typescript.js

	oxc	swc	biome
no-drop	`26.3 ms` (1.00x)	`84.1 ms` (3.20x)	`130.1 ms` (4.94x)
parallel	`36.1 ms` (1.00x)	`126.3 ms` (3.50x)	`225.9 ms` (6.26x)
single-thread	`26.4 ms` (1.00x)	`91.0 ms` (3.45x)	`139.3 ms` (5.28x)

Run benchmark locally

Run the following command on your machine for replication.

cargo bench

Generate the table

pnpm i
pnpm run table

Maximum Resident Set Size

./memory.sh

./files/cal.com.tsx
oxc   11.5 mb (1.00x)
swc   16.6 mb (1.44x)
biome 22.5 mb (1.95x)

./files/typescript.js
oxc    68.8 mb (1.00x)
swc    92.0 mb (1.34x)
biome 117.4 mb (1.70x)

Setup

Uses mimalloc as the global allocator
Uses the following release profile

[profile.release]
opt-level     = 3
lto           = "fat"
codegen-units = 1
strip         = "symbols"
debug         = false
panic         = "abort"

single-thread

This is the standard benchmark run in a single thread.

group.bench_with_input(id, &source, |b, source| {
    b.iter(|| Self::parse(source))
});

no-drop

This uses the iter_with_large_drop function, which does not take AST drop time into account. Notice there is only a 0.3ms difference for oxc, but 7ms difference for swc.

AST drop time can become a bottleneck in applications such as as bundler, where there are a few thousands of files need to be parsed.

group.bench_with_input(id, &source, |b, source| {
    b.iter_with_large_drop(|| Self::parse(source))
});

parallel

This benchmark uses the total number of physical cores as the total number of files to parse per bench iteration. For example it parses 6 files in parallel on my Mac i7 6 cores.

This can indicate the existence of global resource contention.

let cpus = num_cpus::get_physical();
group.bench_with_input(id, &source, |b, source| {
    b.iter(|| {
        (0..cpus).into_par_iter().for_each(|_| {
            Self::parse(source);
        });
    })
});