Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10.1 6/24/83; site decvax.UUCP
Path: utzoo!watmath!clyde!burl!hou3c!hocda!houxm!houxz!vax135!floyd!harpo!decvax!minow
From: minow@decvax.UUCP (Martin Minow)
Newsgroups: net.lang.c
Subject: So maybe you can optimize C after all
Message-ID: <5@decvax.UUCP>
Date: Tue, 29-May-84 23:47:39 EDT
Article-I.D.: decvax.5
Posted: Tue May 29 23:47:39 1984
Date-Received: Fri, 1-Jun-84 08:13:19 EDT
Organization: DEC UNIX Engineering Group
Lines: 18

In a posting a few months ago, I noted that the Draft Ansi C Standard
was very restrictive as to how C may be optimized.  It precluded
common sub-expression recognition, for example.

The latest draft of the standard, received late last week, had a
paragraph that explained that the standard described an "abstract
machine" in which processing takes place so fast as to render
optimizations irrelevant.  It than suggests that a compiler for
a real machine must give the same results as one for the abstract
machine.  If I understand this correctly -- and, please note that
I am quoting from memory -- it would seem to indicate that C
compilers can optimize the code and still conform to the Standard.

Please accept my apologies for any confusion I may have caused.

Martin Minow
decvax!minow