Path: utzoo!attcan!uunet!husc6!uwvax!oddjob!mimsy!chris
From: chris@mimsy.UUCP (Chris Torek)
Newsgroups: comp.misc
Subject: Re: Basics of Program Design
Message-ID: <12298@mimsy.UUCP>
Date: 3 Jul 88 06:41:08 GMT
References: <901@td2cad.intel.com> <3061@rpp386.UUCP> <395@proxftl.UUCP>
Distribution: na
Organization: U of Maryland, Dept. of Computer Science, Coll. Pk., MD 20742
Lines: 31

In article <395@proxftl.UUCP> bill@proxftl.UUCP (T. William Wells) answers
Bob Webber with:
[tail-recursive strlen deleted]
>... Merely asserting that "any decent compiler would..." is not relevant.
>This just asserts that YOU think that a compiler that does not deal with
>tail recursion is not decent.  In the real world, most compilers do NOT
>deal with tail recursion, it not often being very useful to do this for
>C programs.

Actually, there are (only) two reasons `real world' C compilers do nothing
about tail recursion:  Hysterical Raisins, and debugging.

top:
	since C compilers do not eliminate tail recursion, programmers
	avoided it; since programmers avoided it, it does not occur
	often; since it does not occur often, compilers that eliminate
	it do work that rarely pays off; since the work rarely pays
	off, those compilers are not as fast as others; since the
	compilers are not as fast, they are not considered as good;
	since they are not considered as good, they are not used; since
	they are not used, C compilers do not eliminate tail recursion;
	goto top.

:-)

Debugging programs in which tail recursion has been optimised away is
often a confusing experience.  Combined with the Raisins, this may
make enough of an argument to keep compiler writers from doing more work.
-- 
In-Real-Life: Chris Torek, Univ of MD Comp Sci Dept (+1 301 454 7163)
Domain:	chris@mimsy.umd.edu	Path:	uunet!mimsy!chris