案例库 · 工程与运营 · 技术决策 · 1977-1985
这条还没译成中文，下面是英文原文。
IEEE 754 made floating point deterministic, so one program ran the same everywhere
Designed for Intel's 8087 and standardized in 1985, IEEE 754 fixed bit layouts, rounding and special values, ending vendor-specific floating point.
Intel · IEEE
那一手
Before IEEE 754, every computer vendor implemented floating-point arithmetic differently; the same program could give different answers on different machines, and numerical software was not portable. William Kahan's status notes on the standard trace the fix back to 1977, during the design of the Intel 8087 that preceded IEEE 754.
The standard specifies what competitors left unspecified: the precise layout of bits in single and double precision, correct rounding, and defined behavior for exceptions, flags and modes, including special values like NaN and infinity. David Goldberg's ACM survey, hosted by Oracle, describes IEEE 754 as a binary standard that requires exact bit layouts rather than leaving encoding to each vendor.
IEEE 754 was approved in 1985, and because Intel's 8087 and every major CPU after it implemented the same rules, programs and algorithms could finally rely on identical behavior across platforms. Kahan's notes and Goldberg's survey became the canonical references for how the standard works.
为什么管用
- Defined bit layouts and rounding made results reproducible across vendors.
- Defined exceptional cases removed the undefined-behavior trap in numerical code.
- Intel shipped the first implementation, giving the standard a dominant early adopter.
可以搬走什么
A standard wins when it removes the risk users fear most: results that change when you change vendors. Specifying everything, even the edge cases others ignored, made numerical code portable.
后来呢
IEEE 754 became the floating-point foundation of nearly every processor, language and numerical library; later revisions (754-2008 and 754-2019) extended it to decimal and half-precision formats while keeping the same core design. Kahan's status lectures and Goldberg's survey remain the standard introductions.
资料来源
- Lecture Notes on the Status of IEEE Standard 754 for Binary Floating-Point Arithmetic (W. Kahan)
- What Every Computer Scientist Should Know About Floating-Point Arithmetic (D. Goldberg, ACM Computing Surveys 1991)
发现哪里写错了？告诉我们。