Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10 5/3/83; site hou5d.UUCP
Path: utzoo!linus!decvax!harpo!eagle!hou5h!hou5g!hou5f!hou5e!hou5d!mat
From: mat@hou5d.UUCP
Newsgroups: net.lang.c
Subject: Re: Must useless expressions be evaluated?
Message-ID: <702@hou5d.UUCP>
Date: Fri, 14-Oct-83 19:57:54 EDT
Article-I.D.: hou5d.702
Posted: Fri Oct 14 19:57:54 1983
Date-Received: Sun, 16-Oct-83 08:47:41 EDT
References: <317@ucbtopaz.CC.Berkeley.ARPA>
Organization: AT&T Information Systems Laboratories, Holmdel, NJ
Lines: 18


		I think useless expressions must be evaluated, simply
	because there is no syntactic criterion to determine the uselessness
	of any expression. For example,
			(i=j)+something()+k++;
	is not syntactically different from
			(i=j)+1+k++;


Not SYNTACTICALLY different, but SEMANTICALLY different.  And since the
compiler, at many levels from syntax description through parse trees
and code generation templates right out to the optimizer's linked lists
of statements or quads or basic blocks or whatever HAS got syntactic
information about live values and side effects available, it makes
sense to use that information.

					Mark Terribile
					Duke of deNet