Path: utzoo!utgpu!news-server.csri.toronto.edu!rutgers!tut.cis.ohio-state.edu!purdue!mentor.cc.purdue.edu!l.cc.purdue.edu!cik
From: cik@l.cc.purdue.edu (Herman Rubin)
Newsgroups: comp.lang.misc
Subject: Re: Aggressive optimization
Message-ID: <2672@l.cc.purdue.edu>
Date: 23 Oct 90 10:01:06 GMT
References: <66071@lanl.gov> <29028:Oct1906:58:2890@kramden.acf.nyu.edu> <9239:Oct2003:53:1890@kramden.acf.nyu.edu>
Organization: Purdue University Statistics Department
Lines: 21

In article <9239:Oct2003:53:1890@kramden.acf.nyu.edu>, brnstnd@kramden.acf.nyu.edu (Dan Bernstein) writes:
			.........................

> . On the contrary: the compiler will always understand details of the
> machine language that are purposefully hidden from the language used by
> the programmer. The programmer and compiler optimize independently, in a
> sense, as long as the compiler doesn't try to do optimizations that it
> doesn't understand. So on a RISC machine I'll gladly have the compiler
> give me its 2x instruction scheduling benefit, no matter how much hand
> optimization time I put in.

I have no objection to the compiler helping to optimize, and even making 
suggestions to the programmer.  But I object to the first sentence; there
should be NO details of the machine language, instruction overlap, etc.,
hidden from the programmer.  The existence of a single machine instruction
can determine whether or not a block of code, or an algorithm, should even
be considered.
-- 
Herman Rubin, Dept. of Statistics, Purdue Univ., West Lafayette IN47907
Phone: (317)494-6054
hrubin@l.cc.purdue.edu (Internet, bitnet)	{purdue,pur-ee}!l.cc!cik(UUCP)