From: "Alexei A. Frounze" <alex DOT fru AT mtu-net DOT ru>
Newsgroups: comp.os.msdos.djgpp
Subject: Re: inefficiency of GCC output code & -O problem
Date: Wed, 12 Apr 2000 00:01:35 +0400
Organization: MTU-Intel ISP
Lines: 44
Message-ID: <38F3849F.D8E341B4@mtu-net.ru>
References: <38F20E7A DOT 3330E9A4 AT mtu-net DOT ru> <38F2250B DOT 1DC270D5 AT maths DOT unine DOT ch> <38F22D66 DOT 42D78282 AT mtu-net DOT ru> <38F2EBD4 DOT 68D0C6DE AT maths DOT unine DOT ch>
NNTP-Posting-Host: ppp104-221.dialup.mtu-net.ru
Mime-Version: 1.0
Content-Type: text/plain; charset=koi8-r
Content-Transfer-Encoding: 7bit
X-Trace: gavrilo.mtu.ru 955483625 76853 212.188.104.221 (11 Apr 2000 20:07:05 GMT)
X-Complaints-To: usenet-abuse AT mtu DOT ru
NNTP-Posting-Date: 11 Apr 2000 20:07:05 GMT
X-Mailer: Mozilla 4.61 [en] (Win95; I)
X-Accept-Language: en,ru
To: djgpp AT delorie DOT com
DJ-Gateway: from newsgroup comp.os.msdos.djgpp
Reply-To: djgpp AT delorie DOT com

Gautier wrote:
> 
> "Alexei A. Frounze":
> 
> > Well, but it work with registers even when value needs to be zeroed, shifted and
> > so on. It loads registers in very simple situations, where possible to adjust
> > some bytes and we're done.
> 
> What appears to me is that among a set of time-optimising variants, GCC doesn't
> seem to choose the most space-optimising one. But maybe the values of registers
> are re-used later in your code ? Some <32 bit manipulations are said to slow
> down a lot the Pentium-class processors.

No, they are not used. At least these registers are filled in with other data
immediately after that. So... :)

> 
> > > On Intel x86s there is not much to do - there are so few registers - but
> > > anyway GCC is very smart at register mapping !
> 
> > Not a few. That's enough.
> 
> Well... but _compared_ to other processors the choice of registers is very poor.

The most awful problem is insufficiency of RAM and stack. Intel C51
microcontroller has 8 (one-byte) registers. That's enough. But it has only 128
bytes of RAM... :)

> > What's GNAT?
> 
> The GNU Ada95. You can see Ada95 as a standardized Turbo Pascal with some
> enhancements. Since type conversions are explicit (unlike TP for the 8/16/32
> bit integers), you detect the "ugly problems" at source :-).

Pascal is enough for me. I'll probably program in Delphi (visual version of
object pascal for Win32) for the 1st time. Then I'll try C++ for the same tasks.

bye.
Alexei A. Frounze
-----------------------------------------
Homepage: http://alexfru.chat.ru
Mirror:   http://members.xoom.com/alexfru