Path: csiph.com!newsfeed.hal-mli.net!feeder3.hal-mli.net!newsfeed.hal-mli.net!feeder1.hal-mli.net!eternal-september.org!feeder.eternal-september.org!border3.nntp.ams.giganews.com!Xl.tags.giganews.com!border1.nntp.ams.giganews.com!nntp.giganews.com!local2.nntp.ams.giganews.com!nntp.brightview.co.uk!news.brightview.co.uk.POSTED!not-for-mail
NNTP-Posting-Date: Wed, 26 Sep 2012 11:51:10 -0500
From: Gareth Owen <gwowen@gmail.com>
Newsgroups: comp.lang.c
Subject: Re: Swapping logic
References: <49f36538-d5ca-411c-bc58-1f17b55911b6@s8g2000vbc.googlegroups.com> <k3ndov$832$1@dont-email.me> <k3njsc$gjs$1@dont-email.me> <slrn3vfsk65443.b0j.andrews@sdf.lonestar.org> <87ehlpnxot.fsf@gmail.com> <YsGdnYAVmcHbA__NnZ2dnUVZ7vGdnZ2d@lyse.net>
Date: Wed, 26 Sep 2012 17:51:15 +0100
Message-ID: <87y5jwlocc.fsf@gmail.com>
User-Agent: Gnus/5.13 (Gnus v5.13) Emacs/23.1 (gnu/linux)
Cancel-Lock: sha1:ENN+DCInw7bKg+V8b0aV22HYjlg=
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Lines: 31
X-Usenet-Provider: http://www.giganews.com
X-Trace: sv3-VJXpcV6+w/DEXtTjj/90a10HUVvkQhQ1porjPemeJTJWEy+vtihaXLFJSQvkFIrO4VA30Y+MOh4e/iG!osUpa2QoMXNNM3lAtK+V/YnIJqREKJuruzfNeo2+zSRGGdzkRKuFhQYYAaScc8Q6XRbS
X-Abuse-and-DMCA-Info: Please be sure to forward a copy of ALL headers
X-Abuse-and-DMCA-Info: Otherwise we will be unable to process your complaint properly
X-Postfilter: 1.3.40
X-Original-Bytes: 2457
Xref: csiph.com comp.lang.c:26713

David Brown <david@westcontrol.removethisbit.com> writes:

> When you have particular specialised needs, you have to work very
> closely with the compiler - playing with command line switches and
> code layout, and analysing the exact generated assembly code.  For
> small embedded systems, or very time-critical code, many rules no
> longer apply

Right.  I agree completely. When Eric said "Don't do this", he
implicitly meant "Don't do this unless you really, really know why you".

Writing fancy-dan XOR exchange routines without first proving they're
necessary is classic premature optimisation, and we all know what
premature optimisation is.

> But for general programming - as done by the majority of programmers
> for the majority of the time - rules like this apply.  

Right. I agree completely. 

<RANT type=OT>

(And I speak as someone who's spent the day
trying to squeeze individual clock cycles out of an ARM7 routine because
the proprietary compiler does not optimise well -- the kind of day when
you're profiling branch probabilities to determine which branch is the
hot path, but have no way to communicate this to the compiler, which has
decided that no matter what you it's going to guess wrong :( ).  

I  freaking hate *compiler name elided*
</RANT>