[FFmpeg-devel] [RFC] use ff_avc_find_startcode in ff_find_start_code

[Michael Niedermayer]

On Tue, Feb 19, 2008 at 05:13:44PM +0100, Reimar D?ffinger wrote:
> Hello,
> On Tue, Feb 19, 2008 at 04:41:35PM +0100, Michael Niedermayer wrote:
> > > 844 dezicycles in A2, 63644 runs, 1892 skips
> >                                     ^^^^
> > please do something about the skips!
> 
> That was because I ran it on the wrong file :-(
> 
> 
> > [...]
> > > And yes, I remembered to disable speedstep this time.
> > 
> > Thanks you
> > Also please dont run anything else (like mp3 players) while testing
> > and provide benchmark scores for 3 runs so we know how much randomnes
> > is in there.
> 
> Ok, now something hopefully correct, all kinds of combinations tested.
> For the matrix test file things are within measurement errors for
> overall time.
> The higher limit to skip values seems to have pushed it to an advantage
> for my code, but I am quite suspicious about the values.
> A clear difference for the GXF file, though I admit it is an extremely
> stupidly encoded file, so the values there are not worht too much
> either.

Could you post the asm code gcc generates for you for this function?
mine generates:
.L670:
        movzbl  -1(%ebx), %eax
        cmpb    $1, %al
        jbe     .L651
        addl    $3, %ebx
        cmpl    %ebx, %ebp
        ja      .L670


and for the new code:
.L10:
        addl    $4, %ecx
.LBE4:
        cmpl    %ebx, %edi
        jbe     .L8
.LVL16:
.L19:
.LBB5:
        movl    -4(%ecx), %eax
.LVL17:
.LBE5:
        leal    -4(%ecx), %esi
.LVL18:
.LBB6:
        movl    %ecx, %ebx
.LVL19:
        leal    -16843009(%eax), %edx
        notl    %eax
        andl    %eax, %edx
        andl    $-2139062144, %edx
        je      .L10


Also what cpu and gcc version do you have?

[...]
-- 
Michael     GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB

It is dangerous to be right in matters on which the established authorities
are wrong. -- Voltaire
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: Digital signature
URL: <http://lists.mplayerhq.hu/pipermail/ffmpeg-devel/attachments/20080219/038c6c77/attachment.pgp>