[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Poor Sun timings, as competition...

  Date: Tue, 31 Jul 90 11:24 EDT
  From: Barry Margolin <barmar@think.com>
  Subject: Poor Sun timings, as competition...
  To: Hunter Barr <barr@BBN.COM>
  cc: slug@warbucks.ai.sri.com
      Date: Tue, 31 Jul 90 10:08:15 -0400
      From: barr@bbn.com (Hunter Barr)
         Date: Mon, 30 Jul 90 16:32 EDT
         From: Barry Margolin <barmar@think.com>
         The Compile System took about 14 minutes.  Compiling and loading it on a
         Sun 4/280 running Sun Common Lisp 3.0.1 takes about 4 CPU minutes with
         the developmentc compiler and 16 CPU minutes with the production
         compiler (wall times were about 5.5 minutes and 19 minutes).  And on the
         3630 at SLUG it took over 40 minutes.  So, if you're in a heavy
         edit-compile-run cycle the Sun wins.
      Excuse me, but aren't you still ending up with a comparison of
      Symbolics' obsolete model (3630) with Sun4s?  I think you intended
      this message to compare the Symbolics XL1200 with the Sun4.  
  The sentence immediately before the first one in your quote above was,
  "I brought a tape containing the source to our *Lisp Simulator, and
  compiled and loaded it on the XL1200 in the demo room."  The 14 minute
  time was on the XL1200.  It took 40 minutes on the 3630.
      Also, if you are most interested in "a heavy edit-compile-run cycle",
      then perhaps you could comment on how well the Sun development
      environment supports a programmer spoiled by Genera.  My opinion is
      that the Suns are quite usable, but still have a long way to go to be
      in the same league as Genera.
  Well, I don't care for the Sun as a development environment, but many of
  our developers get along pretty well.  Most of the Lispm users here had
  been using them since about release 4 or 5, so their work habits didn't
  include use of presentations; therefore, they don't miss the ability to
  click on everything in sight.  GNU Emacs is as good an editor as Zmacs,
  and we have meta-dot, c-sh-E, c-sh-C, c-sh-M, c-sh-A all implemented in
  GNU Emacs (by running Lisp in a subprocess (or on another network host)
  with its I/O connected to an Emacs buffer).  And running Lisp in an
  Emacs buffer also provides a history mechanism.

The way i read your timing results is that compiling and loading are
slow.  My guess is that his is mostly due to software than hardware.

Compiling and loading on a LISPM could be slow for at least the following

o I/O is slow.

o Compiler is slow.

o Loading and compiling do more on LISPM than on Suns, such as
  who-calls and meta-dot support.