Re: [Bug-gnubg] Bitfields

[W.Stroop]

----- Original Message ----- 
From: "Jim Segrave" <address@hidden>
To: "W.Stroop" <address@hidden>
Cc: <address@hidden>; <address@hidden>
Sent: Monday, November 25, 2002 1:23 PM
Subject: Re: [Bug-gnubg] Bitfields


> On Mon 25 Nov 2002 (12:30 +0100), W.Stroop wrote:
> > 
> > ----- Original Message ----- 
> > From: "Jim Segrave" <address@hidden>
> > To: "W.Stroop" <address@hidden>
> > [snip]
> >  You get
> > >    sequences like:
> > >    if ( ! prc->fCubeful )
> > >     ...
> > >    compiling to:
> > >    testb  $0x1,0x78(%ebx)
> > >    jne    0x80791ad <RolloutGeneral+809>
> > 
> > That's a nice compiler .Unfortunenatly my mingw gcc produces this : 
> >        mov [ebp-10], 00401216
> >        mov eax, dword ptr [ebp-0C]
> >        mov dl, byte ptr [eax]
> >        and dl, 01
> >        test dl, dl
> >        jne 00401253
> > 

> Hmm - is this a gcc derivative? That's a very poor code
> generator. Leaving out the advantage of doing a testb instruction,
> there's no need to test dl after the 'and dl,01', as the z flag is

You could be right also. I  have a pentium 100 (dont laugh :)).
maybe gcc  generates better code for a more advanced processor .  

Rob