mirror of
https://github.com/openharmony/third_party_rust_memchr.git
synced 2026-07-01 08:14:31 -04:00
Andrew Gallant
4a09815c87
This represents an initial go at adding comprehensive criterion benchmarks. We hit every public API function, and vary benchmarks based on both corpus size and match frequency. The point of this exercise is to establish a baseline on the status quo, and make sure we don't introduce any major regressions after a refactoring toward vendor intrinsics.
This crate provides a safe interface libc's memchr and memrchr.
This crate also provides fallback implementations when either function is
unavailable.
Dual-licensed under MIT or the UNLICENSE.
Documentation
no_std
memchr links to the standard library by default, but you can disable the
use_std feature if you want to use it in a #![no_std] crate:
[dependencies]
memchr = { version = "2", default-features = false }
Performance
On my system (Linux/amd64), memchr is about an order of magnitude faster than
the more idiomatic haystack.iter().position(|&b| b == needle):
test iterator ... bench: 5,280 ns/iter (+/- 13) = 1893 MB/s
test iterator_reversed ... bench: 5,271 ns/iter (+/- 7) = 1897 MB/s
test libc_memchr ... bench: 202 ns/iter (+/- 0) = 49504 MB/s
test libc_memrchr ... bench: 197 ns/iter (+/- 1) = 50761 MB/s