Heinz N. Gies
Heinz N. Gies
`s/scheduled//` not much scheduled about it as far as I can tell :( but no it wasn't, it was about a different time it was down
I looked at this and int looks very similar to the SWAR parsing already employed in simd-json: https://github.com/simd-lite/simd-json/blob/main/src/numberparse.rs#L136 am I missing something
here is running the benchmark on my test bench: ``` $ RUSTFLAGS='-C codegen-units=1 -C target-cpu=native' cargo +nightly bench running 13 tests test bench_naive_bytes ... bench: 8 ns/iter (+/- 0) =...
I've looked a bit more at the code and as to where it's fast, it assumes a perfect 16 byte wide number, any real algorithm would need to also determine...
c2, c4 and c8 are really important measurements since for parsing JSON numbers are rarely ever exactly 16 digits long. So a solution that is really fast w/ 16 digits...
really nice writeup 👍
HI! Frist a bit explenation, the reason why we use `&mut [u8]` or `str` is that we do use a form of in situ parsing instead of allocating memory for...
Ja, I'm not worried :) the simdjson code is beautiful so it is always a pleasure to port :D just juggling the usual 10000 things to find the time 😂
I got no idea how that'd work in rust but that just means we got to learn something new :D
Thanks @CJP10 go for it :)