Path: utzoo!utgpu!utstat!jarvis.csri.toronto.edu!mailrus!tut.cis.ohio-state.edu!ucbvax!hplabs!hpda!hpcuhb!hpindda!carlson
From: carlson@hpindda.HP.COM (Bob Carlson)
Newsgroups: comp.sys.mac
Subject: Re: Re: New Mac Rumours (really language-based chips)
Message-ID: <40260006@hpindda.HP.COM>
Date: 27 Feb 89 19:00:31 GMT
References: <5360035@hpccc.HP.COM>
Organization: HP Information Networks, Cupertino, CA
Lines: 24

/ hpindda:comp.sys.mac / dorourke@polyslo.CalPoly.EDU (Sir Hoyt) / 10:16 am  Feb 24, 1989 /
In article <70755@ti-csl.csc.ti.com> holland@m2.UUCP (Fred Hollander) writes:
>>This is very interesting.  I don't have the strongest background in
>>hardware architecture, but, could you please explain how a processor
>>could be optimized for a specific high level language?

>  If you want to see a good example of a processor being optimized for
>a particular language Unisys {really Burroghs} A-Series arch. is
>optimized for Algol.  The way they do this is by providing Machine
>language instructions that implement whole/parts of constructs present
>in the target high level language.  

Burroughs loved this and carried to extremes.  The Burroughs medium
systems were optimized for COBOL.  Most COBOL lines translated into
a single line of code.  Some skeletons lurk in this closet.  Rumour has
it that Knuth designed some of the more esoteric instructions under a
consulting contract.  Also, an architect (Mike Mahon) on this system 
eventually became the chief architect of the HPPA RISC-like system.  
Quite a switch going from COBOL instructions to RISC.  Burrough's small
systems, the 1700, had dynamicly changable control store.  Each running
program could be running a different instruction set!  Cute idea, but it
barked and had to be housebroken.

Cheers, Bob