Responses to Perkin Elmer Unix query (60 lines)

[utzoo!decvax!ucbvax!unix-wizards]

>From dave at UCLA-Security Sat Dec 26 22:32:09 1981
Here are the three responses I got to my query regarding Unix on a
Perkin Elmer 3200 series computer.  Since there was some interest, I am
redistributing them to the list.

From: lee at UTEXAS-11 (Bill Lee)

  Several of us from UT were up at the Dallas PE sales office several
  months back and Wollongong was demoing their Unix. This was just
  before PE had announced the availability of Unix for their machines.
  We had a couple of hours of hands on time and it looked like a real
  Unix (it should be). However, the performance wasn't very good. This
  was explained away as being a configuration problem, i.e. they must
  have specified available memory wrong or something. Maybe but it was
  pretty slow. Slower running 2 or 3 users than our 11/70 is with 6
  users. Even doing a man <foo> would produce several very noticeable
  pauses (about every 20 lines or so) even if I was the only one
  actually running anything. I believe that this was on a 3220. The
  other thing was that we managed to crash the system twice without
  trying. The first time was a C program that looped on doing a fork and
  a wait. The same program does not crash our 11/70. The next time we
  tried it from the shell. In a shell while loop we ran
  /usr/games/cooky. This also put the machine in the weeds.  This also
  irritated the guys from Wallongong because they we trying to demo to
  business types when this happened. We couldn't reproduce the Shell
  loop crash but it swamped the CPU when we ran it again. They claimed
  that they were coming out with an optimizer that will make C programs
  run much faster. My recomendation is to run some real loads on the
  machine that you are considering before buying and see if you can
  really get the performance you want.

From: ucbvax!chico!duke!unc!smb
In-real-life: Steven M. Bellovin

  Jim Ellis, Lynn TennEyck, and I ran some evaluations and benchmarks on
  PE UNIX this past spring.  Basically, it's straight V7; they've
  resisted the temptation to make "improvements".  The benchmarks showed
  it inferior to 4BSD, but the optimizer on the C compiler was broken
  that day, and we *had* recompiled the kernel.  Overall, I'd say it was
  a nice system, but needed more work to improve reliability and
  performance, and to remove a few warts.  The most notable wart was
  that the maximum stack size is set statically at link time.

From: ucbvax!chico!duke!jte

  1) I believe I have convinced them to modify the compiler to be
     smarter about how much stack space to allocate, and to give a
     reasonable run-time msg when one runs out.

  2) The F77 compiler also had problems (notably complex variables) but
     they like the idea of putting their Fortran up under unix. That
     should be an attraction if it is ever done.

  3) The machine has auto-reboot hardware which they do not take
     advantage of since it is a V7 system. On the other hand, they have
     ported UCB's vi and csh. I don't know if I was able to convince
     them to make use of the auto-reboot hardware or not. They haven't
     done it yet.

Thank you for your responses.
Dave