# Ken Shirriff reverse-engineers Intel 8087 floating-point die

> **Key takeaway:** Ken Shirriff has published a detailed reverse-engineering analysis of how Intel's 8087 math coprocessor implemented floating-point addition and subtraction in hardware.

**Published:** 2026-10-22T11:00:30+00:00 
**Source:** Hackaday Hardware 
**Category:** open-hardware 
**Canonical URL:** https://fosswire.org/news/ken-shirriff-reverse-engineers-intel-8087-floating-point-die.html 

## Summary
Ken Shirriff has reverse-engineered the Intel 8087 die to reveal how the first mainstream x86 floating-point coprocessor implemented addition and subtraction in silicon. The 8087 delivered roughly 100x speedup over software emulation on the 8086 CPU, with much of the complexity hidden in shifter circuitry for exponent alignment and microcode handling of edge cases like infinities and overflows. Shirriff's analysis includes the actual microcode listings.

## How the 8087 accelerated floating-point math
- **Exponent alignment**: The 8087 uses extensive shifter circuitry to align floating-point mantissas by shifting one operand's bits according to the exponent difference
- **Speedup magnitude**: Basic floating-point addition ran approximately 100x faster than equivalent software routines on the 8086 CPU
- **Microcode implementation**: A significant portion of the operation sequencing is implemented in microcode, with Shirriff providing the actual listings
- **Edge case handling**: Dedicated silicon detects over- and underflows and handles special values like infinities—cases that are particularly expensive in software emulation
- **Historical significance**: The x87 ISA, despite later criticism, represented a major advance over pure integer-based software floating-point

## Why it matters
The 8087 established the hardware floating-point paradigm that made PC spreadsheets, CAD software, and early 3D games practical. Understanding its silicon-level implementation clarifies design tradeoffs that persist in modern FPUs.