The Intel 8087 chip: floating-point addition and subtraction
Summary
This article reverse-engineers the Intel 8087 floating-point unit, focusing on the addition and subtraction microcode, and explains how 80-bit temporary real numbers are processed. It covers the datapath, the 69-bit integer adder, the shifter, exponent handling, and numerous special cases (infinities, NaN, denormals, zeros), tying these details to historical context and IEEE-754 lineage.