Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10.2 9/5/84; site uvacs.UUCP
Path: utzoo!watmath!clyde!bonnie!akgua!whuxlm!harpo!decvax!mcnc!ncsu!uvacs!mac
From: mac@uvacs.UUCP (Alex Colvin)
Newsgroups: net.lang
Subject: Re: Re: Recursion
Message-ID: <2367@uvacs.UUCP>
Date: Wed, 11-Sep-85 16:15:13 EDT
Article-I.D.: uvacs.2367
Posted: Wed Sep 11 16:15:13 1985
Date-Received: Sat, 14-Sep-85 05:52:51 EDT
References: <909@oddjob.UUCP> <163@ho95e.UUCP> <367@ttrdc.UUCP> <624@mmintl.UUCP> <712@gitpyr.UUCP> <448@sbcs.UUCP>
Organization: U.Va. CS dept.  Charlottesville, VA
Lines: 8

PL/I compilers descended from Multics PL/I do this kind of analysis.  PL/I
procedures are pretty messy, so there's good reason to optimize.

The transitive closure of a procedure call graph tells who calls whom.
This isn't too large, as PL/I has nested scopes, each of which tends to be
small.  External procedures are always considered potentially recursive.  A
procedure used for anything but a call (e.g. as a parameter or assigned to
a variable) are also potentially recursive. This turns out to do a good job.