Perhaps consider using SIMD?

[antonmarsden]

Just a thought :)
https://software.intel.com/en-us/articles/fast-computation-of-adler32-checksums

It appears that there's already an SIMD implementation on this fork: https://github.com/myfreeweb/adler32-simd-rs

[remram44]

I'm taking his benchmark d28c892 😉

I don't know if SIMD is crucial, since I'm seeing speeds well over 1GB/s without, but if there is interest and the implementation is good why not.

[antonmarsden]

You make a fair point. The current implementation is probably adequate for now, and improvements would only really benefit applications doing a lot of [de]compression (file/image processing).

[antonmarsden]

I'm going to experiment with an implementation of this - looks like adler32 is still being used by miniz_oxide, which is a pure Rust [de]compressor, which is used by flate2. adler32 is also still used by the "older" compression libraries.