Path: utzoo!censor!geac!torsqnt!news-server.csri.toronto.edu!cs.utexas.edu!wuarchive!rice!news
From: foo@erfordia.rice.edu (Mark Hall)
Newsgroups: comp.arch
Subject: Re: Register Count
Message-ID: <1991Jan14.233249.21161@rice.edu>
Date: 14 Jan 91 23:32:49 GMT
References: <25090@dime.cs.umass.edu> <1991Jan14.191057.14242@rice.edu> <25106@dime.cs.umass.edu>
Sender: news@rice.edu (News)
Organization: Rice University, Houston
Lines: 29

In article <25106@dime.cs.umass.edu> yodaiken@chelm.cs.umass.edu (victor yodaiken) writes:
)In article <1991Jan14.191057.14242@rice.edu> preston@ariel.rice.edu (Preston Briggs) writes:
)>I wrote:
)>>>
)>>>Optimality sounds like a fine reason to have lots of registers,
)>>>especially since optimal choices by the compiler are undecidable.
)>
)>and yodaiken@chelm.cs.umass.edu (victor yodaiken) writes:
)>>How's that? The compiler is allocating a finite set of resources. How
)>>can this problem be undecidable? Dificult, yes; Intractable, maybe; but
)>>I don't see where undecidability would come from. Did I miss something here?
)>
)>Because the path actually taken through the code can't, in general,
)>be known at compile-time.
)>
)>Preston Briggs
)
)Still don't see it.  The system state, i.e. contents of store and registers,
)appears to determine the path taken through a piece of code. 

   I think preston is referring to programs which read input in the 
 phrase "in general".
 
   Can you predict the state of registers for all possible 
 (possibly infinite) inputs streams? 

 - mark 
   former office-mate of preston   (how's THAT for name-dropping?)