Floating-Point Builtin Functions

Floating-Point Builtin Functions

The alternative to directly accessing the bits as an unsigned integer is to use the existing C++ functions. There are various existing functions for bitwise manipulation of floating-point numbers, in two categories: standard C++ library functions and compiler-specific intrinsics.

C++ has standard functions for the manipulation of floating-point numbers, and their bitwise representations.

• std::signbit — Portably test the sign bit of a floating-point number.
• std::copysign — Portably copies the sign bit from one float, merging it with the value of another (i.e. another's exponent and mantissa).

There are also various compiler-specific “intrinsics” or “builtins” to manipulate floating-point numbers. For Microsoft Visual Studio C++, these are in <intrin.h> and there are also versions for GCC and other compilers.

• frexp — Get the mantissa and exponent.
• ldexp — Bitshifting by an integer shift-count.
• scalbn — Also integer bitshift on a float.
• logb — Extracts the exponent.
• ilogb — Extracts the exponent to integer.
• modf — Splits into whole and fractional parts.
• fma — Fused multiply add on float (Microsoft intrinsic)
• remainder — Get fractional part of floating-point (Microsoft intrinsic)
• _fcvt — Low-level convert float to string (Microsoft intrinsic)

For many of the listed functions, there are additional versions for different floating-point data types, such as float, double and long double. For example, “frexp” will split a double type into its significand (fractional part) and exponent integer, but there's also “frexpf” for 32-bit float types, and “frexpl” for long double types.

• Next: • Up: Table of Contents • Buy: Generative AI in C++: Coding Transformers and LLMs

The new AI programming book by Aussie AI co-founders: AI coding in C++ Transformer engine speedups LLM models Phone and desktop AI Code examples Research citations Get your copy from Amazon: Generative AI in C++