[libcxx-commits] [PATCH] D70631: Microsoft's floating-point to_chars powered by Ryu and Ryu Printf

Tue Aug 31 11:29:02 PDT 2021

ldionne added a comment.

In D70631#2973662 <https://reviews.llvm.org/D70631#2973662>, @jorgbrown wrote:

> In D70631#2973239 <https://reviews.llvm.org/D70631#2973239>, @ldionne wrote:
>
>> In D70631#2577688 <https://reviews.llvm.org/D70631#2577688>, @Mordante wrote:
>>
>>> @ldionne should we enabling floating point support by a CMake configuration option which is ON by default. The tables aren't small and it could add bloat for embedded devices.
>>
>> Hmm, for now I'd make them available everywhere, and in the future, I'd add a knob if that is deemed useful. I believe that most embedded implementations will dead strip everything away if they don't use it anyways, so that might never be an issue.
>>
>> Once this is shipped, we'll only be missing `from_chars`, right? We should be able to use https://github.com/fastfloat/fast_float to do that. We can ask if Daniel Lemire would relicense his code.
>
> I'm pretty sure I can get the Abseil team to license absl::from_chars (https://abseil.io/about/design/charconv), if desired.

That would be nice.

What would you folks think about hiding all this implementation inside the compiled library? Thinking about it, we don't really gain anything from having it in the headers, and if we hide it in the compiled library (and only export the few `to_chars` symbols we need), we can modify the implementation at will later without breaking ABI.

Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D70631/new/

https://reviews.llvm.org/D70631