Path: utzoo!attcan!ncrcan!scocan!seanf
From: seanf@sco.COM (Sean Fagan)
Newsgroups: comp.lang.misc
Subject: Re: Algol, and language design
Message-ID: <1990Jul28.185054.2595@sco.COM>
Date: 28 Jul 90 22:50:54 GMT
References: <25630@cs.yale.edu> <58091@lanl.gov> <1990Jul26.024449.1777@esegue.segue.boston.ma.us> <2406@l.cc.purdue.edu> <1990Jul27.010930.12560@lth.se>
Reply-To: seanf@sco.COM (Sean Fagan)
Organization: The Santa Cruz Operation, Inc.
Lines: 30

In article <1990Jul27.010930.12560@lth.se> bengtl@maths.lth.se (Bengt Larsson) writes:
>Yes, of course some architectures made (make??) it somewhat difficult 
>to have recursive procedures (the ones that stored the return adress in
>the first word of the subroutine comes to mind).

"somewhat more difficult."  That's *all*.

The CDC Cyber 170 machines (of which most people know I am a big fan 8-))
have no stack, and, on a subroutine call, write an instruction into the
first word of the subroutine ("jump <retaddr>," basicly).

Yet:  Pascal was developed on this machine (literally!  Wirth used a Cyber
for the first implementation of Pascal).  Algol and Simula run quite well on
it.  So does PL/I (although the compiler eats up all of memory 8-)).  There
are even a couple of C compilers out there, I've been told.

No, it's not as fast.  The only reason you don't really want to do it on a
Cyber is memory space:  stacks can eat up *lots* of memory, and 170's only
have 128Kwords of memory.  On most "modern" machines, this is less a problem
(but, then, most of them have a stack).

Incidently:  Crays stuff the return address into a register, which your
routine can then save if it wants to (leaf routines, of course, don't need
to save it).

-- 
Sean Eric Fagan  | "let's face it, finding yourself dead is one 
seanf@sco.COM    |   of life's more difficult moments."
uunet!sco!seanf  |   -- Mark Leeper, reviewing _Ghost_
(408) 458-1422   | Any opinions expressed are my own, not my employers'.