Path: utzoo!attcan!uunet!aplcen!samsung!brutus.cs.uiuc.edu!apple!motcsd!hpda!hpcuhc!edwardm
From: edwardm@hpcuhc.HP.COM (Edward McClanahan)
Newsgroups: comp.sys.m88k
Subject: Re: Register Allocation (was Re: Info about 88open & standards)
Message-ID: <100050001@hpcuhc.HP.COM>
Date: 21 Nov 89 18:47:18 GMT
References: <1989Nov16.212149.9770@paris.ics.uci.edu>
Organization: Hewlett Packard, Cupertino
Lines: 37

Ron Guilmette *debates*:

> If I'm wrong that "callee-saves" always produces better performance
> (with smaller executables to boot) then such tests could also have another
> significant benefit.  They could shut me up, and I'm sure everybody
> would be in favor of that! :-)

I offer a simple example...

  Caller uses 2 registers...

  Callee uses 30 registers...

Using the Caller-saved convention, every time a call is made, 2 registers
are *saved*.

Using the Callee-saved convention, every time a call is made, 30 registers
are *saved*.

Similarly, consider the opposite situation (where the caller uses 30 registers
and the callee uses only 2).  In this case, it would be better to have used
the callee-saved convention.

Which of the above is more common?  This is the very question that the OCS
(and HP and MIPS and ...) had to consider.

Your arguments about other *special conditions* (such as a callee only needing
particular registers in a infrequently executed path, etc...) are all valid,
but what is the frequency of these conditions?

=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=

  Edward McClanahan
  Hewlett Packard Company
  Mail Stop 47UE              -or-     edwardm%hpda@hplabs.hp.com
  19447 Pruneridge Avenue
  Cupertino, CA  95014                 Phone: (408)447-5651