I published a quick little crate for long-running “streaming” parallel tasks.
TLDR: I’m toying with writing a C standard library in Rust by porting musl-libc over function-by-function. The work is in progress at https://github.com/dikaiosune/rusl.
TLDR: Comparing cargo bench results to a slightly more robust method eliminates a lot of the noise, but there still appear to be a few performance regressions that both methods agree on. If someone has the statistics expertise to set me straight or help me take this further or both, please get in touch.