From mboxrd@z Thu Jan 1 00:00:00 1970 X-Spam-Checker-Version: SpamAssassin 3.4.4 (2020-01-24) on polar.synack.me X-Spam-Level: X-Spam-Status: No, score=-1.3 required=5.0 tests=BAYES_00,INVALID_MSGID autolearn=no autolearn_force=no version=3.4.4 X-Google-Language: ENGLISH,ASCII-7-bit X-Google-Thread: 103376,fedc2d05e82c9174 X-Google-Attributes: gid103376,public From: "Robert I. Eachus" Subject: Re: Calculating SQRT in ADA Date: 1999/04/02 Message-ID: <3705597C.BBD4AA2A@mitre.org>#1/1 X-Deja-AN: 462113201 Content-Transfer-Encoding: 7bit References: <7dbv6t$4u5$1@nnrp1.dejanews.com> <19990324201959.00800.00000708@ngol04.aol.com> <7dei9a$dvo$1@nnrp1.dejanews.com> <7dhjhi$27a$1@nnrp1.dejanews.com> <36FFF83A.BE789C93@mitre.org> <7dq51l$29h$1@nnrp1.dejanews.com> X-Accept-Language: en Content-Type: text/plain; charset=us-ascii Organization: The MITRE Corporation Mime-Version: 1.0 Newsgroups: comp.lang.ada Date: 1999-04-02T00:00:00+00:00 List-Id: robert_dewar@my-dejanews.com wrote: > It would be nice to see chapter and verse cited for this > (perhaps this horrible behavioor comes from the Honeywell > compiler that you always like to cite even though it never > saw the light of day :-) The Ada/SIL compiler was both used internally at Honeywell and for training military personnel. But you are right that it was never sold commercially. > In fact the IBM trigonometric routines were excellent on > both the 7094 and the 360. Indeed it is from Hirondo Kuki > (I hope I remember the spelling right, this is from a long > time ago on the 7094) that I learned about how math library > routines were written. Yes, the math libraries on the 7094 were excellent. The original libraries on the 360 series were not, but IBM improved them rapidly. However, for the Apollo program, NASA continued to use 4 7094s that were supposed to have been replaced by 360's: IBM could get the math right, or the necessary speed but not both. (To be fair, it was both the shorter word lenght and the trucaction semantics on the 360 that made it unsuitable for this application. IBM did finally come up with the 360/91 that could make the grade.) However, the first IBM PCs had an abominable math library for BASIC. (Can you say Microsoft kiddies? I knew you could.) On the other hand, on the 68000 family, Microsoft BASIC used the excellent MOtorola libraries. > And please note, this discussion is about elementary > functions, your reference to URAND, though it reflects > your own particular interest in random number generation, > is not germane to the thread on square root! Remember the Savage benchmark? It is not surprising to find no cumulative error with it nowadays. But even ten years ago, it wasn't that unusual to be much more concerned about the error introduced than about the speed performance.