But that is what happened [1]. The article, which is probably the most definitive investigation to date, has Greg Walsh claim they invented the fast inverse square root while working at Ardent, which Wikipedia claims was founded in 1985 [2], to meet product performance benchmarks.
The Green Hills Software C compiler had a fast square root algorithm of similar form:
(x >> 1) + {magic constant}
dating between 1983-1985 if I recall correctly. Also implemented to maximize floating point performance benchmarks, and again not drawn from academia. If my recollection is correct, that is one of the earliest known examples of the general technique and predates even the official IEEE 754 floating point specification which was not formally ratified until 1985 (but the standard was in development since 1977 and already de facto adopted by the time it was formally ratified).
The Green Hills Software C compiler had a fast square root algorithm of similar form:
dating between 1983-1985 if I recall correctly. Also implemented to maximize floating point performance benchmarks, and again not drawn from academia. If my recollection is correct, that is one of the earliest known examples of the general technique and predates even the official IEEE 754 floating point specification which was not formally ratified until 1985 (but the standard was in development since 1977 and already de facto adopted by the time it was formally ratified).[1] https://www.beyond3d.com/content/articles/15/
[2] https://en.wikipedia.org/wiki/Stardent_Inc.#Ardent_Computer_...