In a cpp header I want to have the following work:

inline constexpr auto gravitational_constant = math::softfloat::float64_t(6.674e-11);

// This will work when it's an integer value
inline constexpr auto gravitational_constant = math::softfloat::float64_t(6);

Just can't use constexpr when it's a double or float.

Is it possible to make this work as an constexpr?

constexpr-ness is either broken or needs improvement as current runs on Godbolt are not compiling with C++20 constexpr.

The intent of this issue is to repair anything (if anything is actually wrong) and get a good constexprcheck running on Godbolt.

soft_double needs configurable I/O streaming to print values to streams.

Figure out how to handle global interactions with float, double and long double. Use Boost strategy as a guide (but only allow interactions with built-in(s) having size 8 bytes or fewer.

Try and use GCC 13 locally and in CI when available.

The functions floor and ceil have been included. Create tests for these functions.

Add more tests and depth of CI

Consider adding the functions log, exp and pow.

Without implementing a fully compliant collection of <cmath> functions for soft_double, it can be helpful to include log, exp and pow, as these are relatively easy to write and can be used in a myriad of mathematical calculations.

Add VS2022 support in the VS workspace project, in CI and also add VS2022 ASAN run.

We are already back to C++14, so with a few more constexpr functions, then C++11 is within reach.

Add vc 14.3 CI job using existing vs2022 workspace

The circular functions sin and cos need better argument scaling and reconstruction. These can be done on the interval 0...pi/4 and reconstructed using known methods.

CI would be improved if some embedded AVR jobs were added

Something seems unreliable with MinGW host running on windows-latest in CI. Try to find and fix this. Some entry like SEH error during compilation. The executable is not created.

Add apple-clang-gcc builds to CI

The C files can be converted to heade-only design.

Add more examples to round out the scope and depth of the examples.

The point of this issue is to optimize regarding space/speed the implementations of elementary transcendental functions such as power functions exp, log, trigonometric functions sin, cos, etc.

Add support for an outer namespace to shield the namespace math from the global namespace scope. Examples of this are done in the wide-integer and wide-decimal projects

Consider in CI adding a Boost.Math test such as computing a cube root with boost::math::cbrt.hpp.

Use the new reduced dependencies of standalond math.

The specializaion of 'numeric_limits` can be refined. Ensure that the member values agree with the implementation, as some of these are unclear at the moment (such as infinities, NaNs, etc.).

Make better namespace names for the soft_double type.

Add some trigonometric functions including sine, cosing, tangent. Consider maybe also adding hyperbolic trigonometric functions.

I'm trying to make a deterministic physics engine built on soft double but I would need these trig functions.

atan, acos, asin, atan2

I could probably implement them but if someone has them done already that would be great.