I once met a computer - in his eighties in the 1970s. He'd worked for the Admiralty during the war as a general purpose computer helping with ship design, navigation tables and the like. He knew that many "wrinkles"...
At Continental Gummiwerke in Hannover they have a "Data Processing Department" that was founded in the 1880s. All data processing (and at Conti the term includes things like language translation) has been centralised for well over a century - there's a photograph of rows of desks with mechanical multipliers on them. It isn't that long ago - late 1970s - that a presentation of the Hitachi S6 (Itel AS/6, NAS AS/7000) included an in-depth discussion of how the thing multiplied - it used the /168 HSA algorithm with some tweaks. One big difference was that it also used that algorithm on halfword arithmetic. That led to other weirdness. My COBOL is now rusty, but IIRC there are two ways of handling an array or table - indexes and subscripts? One was computed when set, the other when used, and ANSI COBOL used halfwords. I think it was indexes that were computed when used. Anyway, you could write a simple COBOL programme for the S6 that would outperform a (same MIPS, nominally) /168 by a factor of three. I've never trusted MIPS or benchmarks since. -- Phil Payne http://www.isham-research.co.uk +44 7833 654 800 ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
