P1467R9: Extended floating-point types and standard names #105196

Activity

1. ldionne

   added

   c++23

   libc++libc++ C++ Standard Library. Not GNU libstdc++. Not libc++abi.libc++ C++ Standard Library. Not GNU libstdc++. Not libc++abi.

   wg21-paperIssues representing the implementation of a WG21 paperIssues representing the implementation of a WG21 paper

   on Aug 20, 2024
2. ldionne

   added this to libc++ Standards Conformanceon Aug 20, 2024
3. ldionne

   moved this to Todo in libc++ Standards Conformanceon Aug 21, 2024
4. philnik777

   marked [libc++] Add missing C++23 <stdfloat> header to libc++ #165005 as a duplicate of this issue on Oct 27, 2025
5. piotrrak commented on Jan 13, 2026

   

   piotrrak

   on Jan 13, 2026

   More actions

   @philnik777 Hi I've just started working on this issue for both clang and libc++ parts, could you please assign me?

   ❤️React with ❤️2Reacted by Jan Piotrowicz and A. Jiang
6. piotrrak commented on Jan 16, 2026

   

   piotrrak

   on Jan 16, 2026

   More actions

   @philnik777 Hi I've just started working on this issue for both clang and libc++ parts, could you please assign me?

   Sorry about that, just found out that compiler project I've just joined was cancelled, thus I won't be able to work on that full time, as I've planned.
7. johnplatts commented on Feb 10, 2026

   

   johnplatts

   on Feb 10, 2026

   More actions

   clang/lib/Frontend/InitPreprocessor.cpp needs to be updated to define the __STDCPP_FLOAT16_T__ macro in C++23 or later modes if _Float16 is available.

   In addition, the std::float16_t typedef needs to be added to <stdfloat> header if the __STDCPP_FLOAT16_T__ macro is defined.

   libstdc++ actually defines std::float16_t as a typedef for _Float16 in C++23 or later mode if the __STDCPP_FLOAT16_T__ macro is defined.

   Also need to implement std::numeric_limits<_Float16>, std::from_chars(const char*, const char*, _Float16&, std::chars_format), std::to_chars(char*, char*, _Float16, std::chars_format), std::to_chars(char*, char*, _Float16, std::chars_format, int), and the functions in <cmath> for _Float16 in C++23 and later modes if the _Float16 type is available.

   Also need to add support for std::formatter<_Float16, CharT> in C++23 and later modes.
8. rybkr commented on Mar 24, 2026

   

   rybkr

   on Mar 24, 2026

   Last edited by rybkr

   More actions

   @philnik777 I'm interested in taking a libc++-only part of this, at least initially:

   • Add a <stdfloat> header in libc++ that exposes the aliases when the corresponding __STDCPP_FLOAT* macros are available
   • Remove <stdfloat> from headers_not_available in header_information.py and regenerate module inputs
   • Add some libc++ tests for header availability

   I would not touch Clang predefined macros or __cpp_lib_stdfloat in this patch. Is this a reasonable decomposition?
9. frederick-vs-ja commented on Mar 27, 2026

   

   frederick-vs-ja

   on Mar 27, 2026

   Contributor

   More actions

   Also need to implement std::numeric_limits<_Float16>, std::from_chars(const char*, const char*, _Float16&, std::chars_format), std::to_chars(char*, char*, _Float16, std::chars_format), std::to_chars(char*, char*, _Float16, std::chars_format, int), and the functions in <cmath> for _Float16 in C++23 and later modes if the _Float16 type is available.

   Also need to add support for std::formatter<_Float16, CharT> in C++23 and later modes.

   I think it's conforming to backport all additions except for new typedef names.