William Kahan

William Kahan was the primary architect of the IEEE 754 standard for floating-point arithmetic, formally adopted in 1985, which defines how virtually all modern computers represent and manipulate non-integer numbers. His work introduced predictable rounding behavior, NaN (Not a Number), and denormalized numbers, ending an era where different machines used incompatible floating-point systems.

Kahan received the 1989 Turing Award for his fundamental contributions to numerical analysis, most notably his design of the IEEE 754 floating-point standard. The ACM citation highlighted how his work made floating-point arithmetic reliable and portable across different computer architectures, a transformation that enabled consistent scientific computing worldwide.

The Kahan summation algorithm, also called compensated summation, reduces the numerical error that accumulates when adding many floating-point numbers of varying magnitudes. It uses a running compensation variable to track and correct for rounding errors at each addition step, and it is widely used in scientific and statistical computing.

Kahan consulted for Intel in the late 1970s on the design of the 8087 floating-point coprocessor, which was one of the first hardware implementations of what would become IEEE 754. His work on the 8087 demonstrated that the standard could be efficiently realized in silicon, and the chip's design influenced floating-point hardware across the PC industry for decades.