Daniel Lemire
Daniel Lemire
@ScottBailey What do you think of @wolfpld's solution?
> I feel like this is unexpected behavior? Here is how I view things. We have two code paths. One that were the cache location is set. This code path...
(I also have no power here. I am just a user.) > If we force a cache path we reduce functionality for others. Concretely, what would be lost? > But...
@ScottBailey (So we are clear, I have no power here and I am not claiming you are wrong.) I think that the patch we are discussing is equivalent to making...
We could start with scalar implementations and work our way up. Importantly, we need reasonable benchmarks.
I suspect that the spec is https://datatracker.ietf.org/doc/html/rfc4648
@Demonese That's a good comment and maybe one could first propose a specification and submit it for discussion. This could be reviewed by the Node.js folks and others.
I worked on the *decoding* and the code is in production. - Daniel Lemire, "Fast decoding of base32 strings," in Daniel Lemire's blog, July 20, 2023, https://lemire.me/blog/2023/07/20/fast-decoding-of-base32-strings/. - https://github.com/NLnetLabs/simdzone/blob/main/src/westmere/base32.h -...
@Cirnoo The simdjson library validates what you parse, so it will happily accept bad JSON as long as it does not need to access the bad JSON... ```C++ #include "simdjson.h"...