home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.arch
- Path: sparky!uunet!charon.amdahl.com!pacbell.com!decwrl!sun-barr!ames!sgi!odin!mash.wpd.sgi.com!mash
- From: mash@mash.wpd.sgi.com (John R. Mashey)
- Subject: Re: RTX and SC32
- Message-ID: <1992Nov6.024502.10633@odin.corp.sgi.com>
- Sender: news@odin.corp.sgi.com (Net News)
- Nntp-Posting-Host: mash.wpd.sgi.com
- Organization: Silicon Graphics, Inc.
- References: <17131@mindlink.bc.ca> <1992Nov4.191038.12063@news.arc.nasa.gov> <CLIFFC.92Nov5101357@miranda.rice.edu>
- Date: Fri, 6 Nov 1992 02:45:02 GMT
- Lines: 45
-
- In article <CLIFFC.92Nov5101357@miranda.rice.edu>, cliffc@rice.edu (Cliff Click) writes:
-
- |>
- |> And now a soapbox:
- |>
- |> Compiler technolgy is driven by computer archetectures.
- |> If a big-name company produced a blazing stack machine and handed a few
- |> out to CS departments doing compiler research, compiler researchers
- |> (like myself, hint, hint) would find the moral equivalent of the
- |> "graph coloring register allocator" for stack machines.
-
- 1) Big-name companies have produced stack machines, including Burroughs
- B5000 & its A-series descendents, HP3000, and Tandem.
- 2) With the exception of the Burroughs, which lives on as the A-series,
- the others have gone away, or are in the process of being migrated elsewhere.
- 3) Compiler technology now drives computer architecture as much as the
- reverse; much of the current RISC generation was driven/designed by
- optimizing compiler people.
- 4) This is an architectural style that has been known & familiar for a long
- time ... nevertheless, it has failed to win many converts outside of a
- few niches, and some people who've had huge investments in it have chosen
- to switch.
-
- 5) For a general-purpose systems architecture, you need to do integer
- arithmetic, floating-point, virtual-memory-management, run "large" programs,
- and support a wide variety of common languages. It can be very
- misleading to compare transistor counts for implementations of
- architectures that have all of these things, with ones that don't have
- all of these. For example, it is quite easy to prove that PDP11s were
- "better" than VAXen (denser code & data fof same problems, after all);
- of course, life was really tough if you had more than 64KB of data.
-
- 6) In many cases, an architectural family that spans a wide range is
- less efficient than an architecture targeted to a narrow target. Sometimes
- the narrow one is enough better that it wins ... but other times, the
- broader ones are good enough, even at the narrow one's target, that the
- broader one wins on the basis of software commonality. After all,
- there are lots of flavors of X86s & 68Ks, and there are already RISCs
- whose family extends from $15 chips into high-end chips that power
- mainframe-like systems ....
-
- -john mashey DISCLAIMER: <generic disclaimer, I speak for me only, etc>
- UUCP: mash@sgi.com
- DDD: 415-390-3090
- USPS: Silicon Graphics 7U-005, 2011 N. Shoreline Blvd, Mountain View, CA 94039-7311
-