[comp.arch] Design philosophy

daveb@geac.UUCP (David Collier-Brown) (08/02/88)

From article <1988Aug1.062659.25971@utzoo.uucp>, by henry@utzoo.uucp (Henry Spencer):
> To sort of paraphrase a comment the Mips people have made about memory:
> think twice before building specialized hardware to do something a
> general-purpose CPU can do, because many people are putting enormous
> resources into making the g-p CPUs better and faster, and they may well
> catch up with you -- probably sooner than you think.  Exploiting mass-
> market products can work better than trying to compete with them.

  I have to both agree and disagree with this...

  I've been burned by the mass market catching up with the
performance of special-purpose processors we'd designed, so I can
strongly agree with the dictum above.  If you want your company to
go out of business, just **try** to compete with a silicon foundry.
  One well-known example is the Lisp machine, which got caught up to
in several areas (but not all areas!) as soon as they were released.
A less well-know example (on the net) is vCISC technology, which
suffers from very high per-generation design costs

  On the other hand, some functions that a cpu does are "simple" (in
some logical sense), and can usefully be encapsulated in an
interface, communicated with according to a fixed protocol, and
shoved off into a device.  If the device is too slow, shove in next
weeks' low-cost processor.  The week after, shove in the next
incarnation of the special device...

  I see these two approaches as much more of a business-decision
problem than an architectural issue: I posted last week the
special->general->special cycle that Foley & Van Dam noticed in
graphics devices, and suspect that it applies to many more cases
than just graphics.

> -- 
> MSDOS is not dead, it just     |     Henry Spencer at U of Toronto Zoology
> smells that way.               | uunet!mnetor!utzoo!henry henry@zoo.toronto.edu

 Yea, verily!

--dave
ps: I'd love to have a board which emulates an X25 line on one side
    and contains a processor I can program to drive a line of
    poll-and-select terminals on the other side.  Now **there's** a
    business decision to unmake at high speed!
-- 
 David Collier-Brown.  {mnetor yunexus utgpu}!geac!daveb
 Geac Computers Ltd.,  |  Computer science loses its
 350 Steelcase Road,   |  memory, if not its mind,
 Markham, Ontario.     |  every six months.

z8my@vax5.CIT.CORNELL.EDU (07/10/89)

It seems to me, and has been pointed out before, that Herman Rubin of Purdue
wants a language, or a language construct to allow him to include non-port-
able commands that compile to some specific, machine-dependent code.

Unfortunately for him, nearly everyone else in comp.arch (I don't read
cop.lang.misc) feels that portability is more important.

So can we end this endless argument?

Sam Paik
d65y@vax5.cit.cornell.edu