From mboxrd@z Thu Jan  1 00:00:00 1970
X-Spam-Checker-Version: SpamAssassin 3.4.4 (2020-01-24) on polar.synack.me
X-Spam-Level: 
X-Spam-Status: No, score=-1.9 required=5.0 tests=BAYES_00 autolearn=ham
	autolearn_force=no version=3.4.4
X-Google-Language: ENGLISH,ASCII-7-bit
X-Google-Thread: 103376,ac78eacbae7081a6
X-Google-Attributes: gid103376,public
X-Google-ArrivalTime: 2000-09-06 06:51:33 PST
Message-ID: <39B64BC6.DF9ECCCC@maths.unine.ch>
Date: Wed, 06 Sep 2000 15:51:01 +0200
From: Gautier de Montmollin <gautier.demontmollin@maths.unine.ch>
Organization: Maths - Uni =?iso-8859-1?Q?Neuch=E2tel?=
X-Mailer: Mozilla 4.75 (Macintosh; U; PPC)
X-Accept-Language: en
MIME-Version: 1.0
Newsgroups: comp.lang.ada
Subject: Re: asm in ada
References: <Ihps5.149$xr6.264251@nnrp6.proxad.net>
 <fTit5.192$N13.24942@eagle.america.net> <8p5e1h$ai5$1@nnrp1.deja.com>
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
NNTP-Posting-Host: mac13-32.unine.ch
X-Trace: 6 Sep 2000 15:51:02 +0100, mac13-32.unine.ch
Path: 
 supernews.google.com!sn-xit-02!sn-east!supernews.com!news-feed.riddles.org.uk!newsfeed.direct.ca!look.ca!sunqbc.risq.qc.ca!news.imp.ch!news.ifi.unizh.ch!news-zh.switch.ch!sitelnet.unine.ch!mac13-32.unine.ch
Xref: supernews.google.com comp.lang.ada:522
Date: 2000-09-06T15:51:01+02:00
List-Id: <comp.lang.ada>

Robert Dewar:

> It is often better to either try to avoid the use of ASM
> completely (it would be interesting to know why you are using
> it, so often there are ways to do what you want in Ada after
> all), or to write a small assembler routine (which surprisingly
> is likely to be more portable than writing asm inserts).

Apart of hardware-oriented instructions, asm code seems to me
useless if the compiler provides advanced optimizations, like GNAT.
It is in general useless to try to outperform it. It's better
to see the feedback given by produced asm ("gcc -S") and
execution speed, for choosing the most convenient data structures,
"to inline or not to inline", etc. . A good surprise among others
is that passing and using an elegant unconstrained vector seems
faster than passing N numbers as parameters - surely it is more
"speaking" to the optimizer. With 3.13 it is surely even nicer...

As for ObjectAda, despite its wonderful programming environment,
I fear that it doesn't optimize at all (slow code and no option
beyond suppress_all). Unless I'm wrong ?...

In the Borland Pascal days, "to optimize" meant "to write
inline assembler code for that"...

______________________________________________________
Gautier  --  http://members.xoom.com/gdemont/gsoft.htm