# William Kahan

> 1933– · Mathematician, Computer Scientist
>
> **Recorded contribution:** IEEE 754 floating-point standard; "Father of floating point"; Kahan summation algorithm

## How to use this dossier

Read for a causal chain, not a hero story: inherited problem → contribution → mechanism → downstream capability → limit. Then close the page and complete the reconstruction exercise from memory.

## 1. Historical orientation

William Kahan (born 1933) is a numerical analyst who was the principal architect and forceful advocate behind IEEE 754 floating-point arithmetic. Before standardization, machines differed in formats, rounding, exceptional values, and underflow behavior, making numerical software unpredictable. IEEE 754 defined binary formats, correctly rounded basic operations, infinities, NaNs, signed zero, subnormals, and exception behavior. Kahan also devised compensated summation, which tracks low-order information lost during repeated addition. The standard does not make floating-point identical to real arithmetic; it makes finite approximations and exceptional behavior sufficiently specified that programmers can analyze them.

## 2. The problem inherited

Incompatible and poorly specified floating-point systems produced silent overflow, destructive underflow, and nonportable numerical results.

## 3. The central contribution

Kahan drove a standard arithmetic contract that balances range, precision, gradual underflow, rounding, and diagnosable exceptions across hardware and software.

## 4. Reconstruct the mechanism

1. Encode a finite number with sign, biased exponent, and significand, reserving exponent patterns for zero, infinity, and NaN.
2. Compute an exact conceptual result for a basic operation and round it according to a specified mode, normally nearest-even.
3. Represent tiny values with subnormals so underflow loses precision gradually instead of jumping directly to zero.
4. Propagate exceptional values and set status flags so invalid, overflow, underflow, and inexact events can be detected.

## 5. What changed downstream

- IEEE 754 made numerical behavior far more portable across processors and languages.
- NaNs, infinities, and subnormals enabled continued computation with explicit exceptional semantics.
- Kahan summation became a standard example of algorithm design that respects finite precision.

## 6. Attribution, limits, and uncertainty

- IEEE 754 was negotiated by a committee of hardware and numerical experts; Kahan was pivotal but not alone.
- Compiler reassociation, extended precision, fused operations, and parallel reductions can still change results.
- NaN propagation and exception flags do not replace domain-specific error analysis or validated algorithms.

## 7. Reconstruction lab

Decode several 32-bit floating-point patterns, including a subnormal, infinity, and NaN. Sum one million small values after one large value using naïve and Kahan summation, then explain the rounding error from the significand alignment.

## 8. Evidence trail

- [William Kahan](https://people.eecs.berkeley.edu/~wkahan/) — University of California, Berkeley
- [IEEE Standard for Floating-Point Arithmetic](https://standards.ieee.org/ieee/754/6210/) — IEEE Standards Association

---

*Research checked 2026-08-09. Dates, roles, and claims about living people are historical snapshots. Linked sources remain the authority; this dossier is original instructional synthesis.*