Newsgroups: rec.arts.int-fiction
Path: news.duke.edu!newsgate.duke.edu!nntp-out.monmouth.com!newspeer.monmouth.com!feed2.onemain.com!feed1.onemain.com!news-spur1.maxwell.syr.edu!news.maxwell.syr.edu!news-out.uswest.net!uunet!chi.uu.net!nyc.uu.net!world!meta
From: meta@pobox.com (mathew)
Subject: Re: [Inform] Using Inform to port BASIC programs
User-Agent: MacSOUP/2.4.6
Sender: news@world.std.com (Mr Usenet Himself)
Message-ID: <1ejdtah.f1j3o7m7k3vyN%meta@pobox.com>
Date: Wed, 1 Nov 2000 01:02:02 GMT
References: <39FF1F9E.939B4628@hotmail.com> <8tn9t6$5f3$1@eastnews1.east.sun.com> <39ff37b2.52600300@news.demon.nl> <39ff4f62.148037085@goliath2.usenet-access.com>
Nntp-Posting-Host: ppp0b145.std.com
Organization: Evil Geniuses For A Better Tomorrow
Lines: 26
Xref: news.duke.edu rec.arts.int-fiction:80215

Carl Muckenhoupt <carl@wurb.com> wrote:
> Hardly.  It's simply sound advice, analogous to "If you want wacky
> string manipulation outside of an IF context, use perl" or "If you
> want your code to run very very fast, use assembly language".

I hate to be picky, as it's rather outside the scope of this newsgroup,
but the latter is very very oversimplistic advice, and unsound in many
circumstances.

A more accurate piece of advice would be "If you want your code to run
very very fast, profile it and write the critical bottleneck routines in
assembly language -- but only if you're a true expert, or you're using a
very simple CPU architecture, or you can't get a good optimizing
compiler."

An optimizing compiler will do better than an average human programmer
at writing assembler.  On architectures where there are complex and
subtle behaviors such as delayed and speculative branch, heavy
pipelining and asynchronous parallel instruction execution, compilers
will do better than almost any human programmer for any non-trivial
piece of code.


mathew
--
No taxation without representation!
