Xref: utzoo comp.sys.dec:936 comp.arch:7887 Path: utzoo!attcan!uunet!lll-winken!ames!haven!purdue!decwrl!vixie From: vixie@decwrl.dec.com (Paul A Vixie) Newsgroups: comp.sys.dec,comp.arch Subject: Re: DECstation 3100 info. Message-ID: Date: 17 Jan 89 10:27:00 GMT References: <979@isieng.UUCP> <85330@sun.uucp> <13508@jumbo.dec.com> <1989Jan17.042349.5379@utzoo.uucp> Sender: vixie@decwrl.dec.com Organization: DEC Western Research Lab Lines: 21 In-reply-to: henry@utzoo.uucp's message of 17 Jan 89 04:23:49 GMT [Spencer] # Well, "fully" if you discount the mishmash of four or five different # floating-point formats, not all of which exist on all VAXen last I heard... There have been rules for subsetting Vaxen since before there were Vaxen. (I know because I just went looking for "trap type 9" and I ran across a prehistoric version of the Vax Arch Ref Man in the DECWRL library). Some implementations leave out some floating point types; others leave out the POLY and EDITPC and sometimes MOVC. These you do in software; if the hardware can do it (either by basic design or because of an accelerator), you don't see the traps and your code runs faster. I'll admit, it's not the way I'd design something today. But in 1970, they had some design ideas that they couldn't implement in hardware "yet", and some other ideas that they knew would be hard to put into a low-end model later on. Looking back at it, it seems like they did okay. -- Paul Vixie Work: vixie@decwrl.dec.com decwrl!vixie +1 415 853 6600 Play: paul@vixie.sf.ca.us vixie!paul +1 415 864 7013