Path: utzoo!utgpu!jarvis.csri.toronto.edu!rutgers!sun-barr!cs.utexas.edu!tut.cis.ohio-state.edu!ucbvax!SCFVM.BITNET!ZMLEB From: ZMLEB@SCFVM.BITNET (Lee Brotzman) Newsgroups: comp.lang.forth Subject: Re: ANSI X3J14 TC Feedback Message-ID: <8907061530.AA06615@jade.berkeley.edu> Date: 6 Jul 89 14:53:54 GMT References: Sender: daemon@ucbvax.BERKELEY.EDU Organization: The Internet Lines: 56 > >Before I saw your latest posting I was prepared to propose that it is better >to break hearts than to break code. I had been told that Cray One was 1's >complement but had heard that recent and new Cray's might be 2's complement. >If the latest Cray is 2's complement that settles it for me. If in 20 years >Univac and CDC have gotten along without Forth they can suffer a little >longer. > Funny. I always thought that the basic data type for Cray machines was floating point -- which is sign-and-magnitude, not 1's OR 2's complement. Oh well, that's confusion for you. One little nitpick about the above. If the Standard says that arithmetic is 2's complement, that doesn't mean that you can't have Forth on a 1's complement machine, it just means you can't have *Standard Forth*. There's a big difference. I don't have a Forth-83 Standard Forth on my PC because it doesn't use floored division (and also has a few other deviations from Gospel, as well), but I still have Forth. That brings me to the other topic: >The reward for doing a good job is another job. What do the TeleFolks think >about floored division? I think that its prescription was irresponsible. > I agree. At the time the Standard was drafted, all hardware division was what I believe they call "truncated" (?). This offended the mathematical sensibilities of the Forth Standards Team (Bob Berkey in particular) and they declared that Forth would use the "correct" floored division -- at a cost in performance. I have seen several disparate claims about exactly what that cost is, and since I am no expert I reserve my opinion. But it just seemed to fly in the face of the Forth philosophy to require a form of division that no computer architect had seen fit to implement in hardware. Here is a question: What have the Forth hardware architects (Computer Cowboys, Harris, Applied Physics Lab) implemented for division? Do they have floored or truncated division? A Standard, by definition, must define and stabilize "common practice". If floored division is not common practice even amongst the manufacturers of hardware implementations of Forth, then, in my humble opinion, the ANSI TC has no choice but to drop the requirement from the Standard, regardless of the mathematical arguments to the contrary. On the other side of the coin: will dropping the floored division requirement break existing code, and if so, how much? As I said, my Forth-83 system does not use floored division, so I would be unaffected. Even if ANSI Forth required floored division, I would be unaffected, because I would simply ignore that requirement. How many other Forth-83 systems have also ignored floored division? How much code has been written in those systems as opposed to "full" implementations of Forth-83? Those are questions that the TC must address. I don't know the answers, but hopefully some of you out there in Network land do. -- Lee Brotzman (FIGI-L Moderator) -- BITNET: ZMLEB@SCFVM SPAN: CHAMP::BROTZMAN -- Internet: zmleb@scfvm.gsfc.nasa.gov GEnie: L.BROTZMAN -- The government and my company don't know what I'm saying. -- Let's keep it that way. -- What do you call three lawyers up to their necks in quicksand? -- Not enough quicksand.