https://gcc.gnu.org/bugzilla/show_bug.cgi?id=118038
--- Comment #9 from Gero <gero.peterhoff at gmx dot net> --- Hi Andrew, I would like to provide future function (https://en.cppreference.com/w/c/experimental/fpext1) in C++ and/or improve existing functions that are not up to standard. Leterem also includes your quadmath::nanq, which does not support payload (https://gcc.gnu.org/bugzilla/show_bug.cgi?id=113259) All these functions must of course work with known FP types. But my demand goes further --> they must be constexpr, if it is possible. First of all I want to provide constexpr equivalents to (S)NAN(X) - of course with different nomenclature in C++/as a template. You provide several builin functions, but they are not fully usable (e.g. constexpr vs not constexpr) and of course they are limited to gcc - but I need a compiler independent solution. To achieve these goals I need various bit operations on FP types. This can be accomplished by bit-casting the FP types to INT types, performing the operations, and bit-casting to FP again. I also need various bit masks, which of course must also be available constexpr. So: bit_cast should only do exactly what it is supposed to do: bit-cast. No more, no less. And it must be value-independent! thx Gero
