Rust fast &[u8] to integer parser

Got the idea from here (source).

Faster on x8664 (uses SIMD, SSE4.1, AVX2), but can be used even if you don't have x8664 SIMD capable cpu (and it will be still faster than str::parse).

Supports negative values and validates the input.

Supported output types: u8, i8, u16, i16, u32, i32, u64, i64, u128, i128, usize, isize.

Has good test coverage, and can be considered safe.

To enable SIMD it needs the target-feature or target-cpu flags set, or it will fallback to non-SIMD functions. To do it you can copy the ./.cargo/config.toml in your project, or you can use one of the following environment variables:

• RUSTFLAGS="-C target-feature=+sse2,+sse3,+sse4.1,+ssse3,+avx,+avx2"

• RUSTFLAGS="-C target-cpu=native"

For Windows PowerShell you can set it with $Env:RUSTFLAGS='-C target-feature=+sse2,+sse3,+sse4.1,+ssse3,+avx,+avx2'

By default the target-feature is set in ./.cargo/config.toml, but seems like it works only inside this project.

If you have &str then use .as_bytes()

Supports no_std with --no-default-features

Maybe todo: There is a posibility to parse base 16 input, by first finding (a,b,c,d,e,f) and then increment them by 9

Examples

``` let val: u64 = atoisimd::parse("1234".asbytes()).unwrap(); asserteq!(val, 1234u64);

asserteq!(atoisimd::parse::("-2345".asbytes()).unwrap(), -2345i64); ```

Benchmarks

You can run cargo bench on your machine (or individually with cargo bench -- "parse u64")

Results

More information you can find here.