About two years ago I adapted the Daniel Lemire simdjson algorithm to build a csv memory index. There are some nifty tricks that he employed to max out the processing capacity on a vectorized/SIMD platform.
After watching some of the presentations at RustNation UK, I was inspired to "just post it" to get the ball rolling. The code works on an x86 platform. There are a few tasks to get it to work on multiple platforms and otherwise improve the api. This might include using the Rust simd crate depending on how much progress it's made since I last checked.
If anyone is interested in practicing some low-level Rust and having the work be available for everyone to share, please let me know: firstname.lastname@example.org.