Add q16-fixed-exp: a Q16.16 fixed-point exp() library for Cortex-M (C + ASM)

[OscarJ12]

This library implements fixed_mul(), fixed_exp() (for x ≥ 0) and fixed_exp_signed() (full signed support) in Q16.16 format, entirely in integer math—no FPU or divisions. It uses:

Binary decomposition of the fractional part (11-bit resolution) for sub-0.1 % accuracy on positive exponents

LUT + one-point linear interpolation for negative exponents (≤ 0.1 % error over [0,10])

A tiny lookup table (~324 B) and just one extra multiply per call for negatives

Both portable C and highly optimized ARM Cortex-M3 assembly

Based on O. W. Jackson’s TechRxiv paper “A Fixed-Point Binary Decomposition Method for Efficient Exponential Approximation in Embedded Systems” (May 2025):
https://www.techrxiv.org/users/921611/articles/1293706-a-fixed-point-binary-decomposition-method-for-efficient-exponential-approximation-in-embedded-systems

GitHub: https://github.com/you/q16-fixed-exp