Xref: utzoo comp.sys.intel:456 comp.sys.ibm.pc:16077 comp.lang.pascal:911
Path: utzoo!attcan!uunet!mcvax!diku!dkuug!ambush!leif
From: leif@ambush.UUCP (Leif Andrew Rump)
Newsgroups: comp.sys.intel,comp.sys.ibm.pc,comp.lang.pascal
Subject: How well is 80x87 used in different implementations?
Keywords: 80x87, 8087, 80287 80387, math processor, intel
Message-ID: <774@ambush.UUCP>
Date: 1 Jun 88 10:55:46 GMT
Organization: AmbraSoft A/S (Denmark)
Lines: 19

Don't think - or else you end like me asking questions! And my next
question is: Which routine is the fastest when you have a 80x87?

      a)    a := 22/7;                  b)   a := 22/7
            a := a + 1;                      b := 0;
            b := 0;                          for c := 1 to 1000 do
            for c := 1 to 1000 do              b := b + c;
              b := b + c;                    a := a + 1

It depends of the implementation of course, but does anybody out there
know a compiler/library that make b) faster than a) because it dosn't
wait for the calculation of the variabel a to finish before it has to
use it (a := a + 1)?

   Leif Andrew Rump, Ambrasoft A/S, Roejelskaer 15, DK-2840 Holte (Denmark)
 UUCP: leif@ambush.dk, phone: +45 2424 111; ABC BBS: +45 68 00 544 (2:505/38)

     "...with a liquid which was almost, but not quite, entirely unlike tea."
                                   - The Hitch hicker's Guide to the Galaxy