Floating-Point Addition and Subtraction on Intel’s 8087 hackaday.com

Ken Shirriff reverse-engineered the Intel 8087’s floating-point addition and subtraction circuits and microcode, using the chip’s die to explain how they handle shifting, overflow, underflow, and edge cases. The 8087 could perform these operations roughly 100 times faster than the 8086, making floating-point calculations more practical in PC software.