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: Fri, 14 Apr 2000 10:22:46 +0400
Organization: MTU-Intel ISP
Lines: 38
Message-ID: <38F6B936.3D092A46@mtu-net.ru>
References: <Pine DOT SUN DOT 3 DOT 91 DOT 1000413112851 DOT 28180G-100000 AT is> <38F6137B DOT 47481761 AT mtu-net DOT ru> <38f6342c DOT 52524603 AT news DOT freeserve DOT net> <38F637C7 DOT 4F4ECB6 AT mtu-net DOT ru> <38F6A957 DOT 90A167DE AT is DOT elta DOT co DOT il>
NNTP-Posting-Host: ppp97-38.dialup.mtu-net.ru
Mime-Version: 1.0
Content-Type: text/plain; charset=koi8-r
Content-Transfer-Encoding: 7bit
X-Trace: gavrilo.mtu.ru 955693397 6856 212.188.97.38 (14 Apr 2000 06:23:17 GMT)
X-Complaints-To: usenet-abuse AT mtu DOT ru
NNTP-Posting-Date: 14 Apr 2000 06:23:17 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

OKay, but is the ``: "g" (value)'' invalid thing unsupported by GCC/AS? 
If not, what are the actual limitations/restrictions to it?

I've never seen such ones. Just only a limit of up to 6 arguments, if all they
are put to the conventional registers (EAX,EBX,ECX,EDX,ESI,EDI).

Is there something extra I should know for using :"g"() correctly?

Why GCC/AS accept *different rules* of inline ASM under different optimization
switches?

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

Eli Zaretskii wrote:
> 
> "Alexei A. Frounze" wrote:
> >
> > Steamer wrote:
> > > The bugs in your code are in the interface between the C and the inline
> > > assembly.  They may not cause a problem if you don't use -O2, but they
> > > surely will as soon as the optimizer starts trying to rearrange things,
> > > remove redundant instructions, avoid unnecessary memory access, etc.
> >
> > Thus design of the optimizer is buggy as well. :))
> 
> No, the optimizer is okay, and the interface between GCC and Gas is okay as
> well.  You just need to put into the inline assembly the necessary magic
> (constraints and clobbered registers) to let GCC know enough about the inline
> fragment so that it could Do The Right Thing.
> 
> However, if you intend purposefully to fight GCC and deliberately not abide
> by the rules of correct inline assembly, don't be surprised if GCC and Gas
> will fight back and produce invalid code... ;-)