[FFmpeg-devel] [PATCH] Add a time_base field to AVFilterPad.

[Stefano Sabatini]

On date Wednesday 2010-10-06 02:44:51 +0200, Michael Niedermayer encoded:
> On Wed, Oct 06, 2010 at 02:00:53AM +0200, Stefano Sabatini wrote:
> > On date Tuesday 2010-10-05 13:48:41 +0200, Michael Niedermayer encoded:
> > > On Tue, Oct 05, 2010 at 11:16:51AM +0200, Stefano Sabatini wrote:
> > > > On date Monday 2010-10-04 00:12:15 +0200, Michael Niedermayer encoded:
> > > > > On Sun, Oct 03, 2010 at 11:23:49PM +0200, Stefano Sabatini wrote:
> > > > > > On date Sunday 2010-10-03 21:07:55 +0200, Michael Niedermayer encoded:
> > > > [...]
> > > > > > static int config_output(AVFilterLink *outlink)
> > > > > > {
> > > > > >     AVFilterContext *ctx = outlink->src;
> > > > > > 
> > > > > >     /* choose the output timebase */
> > > > > >     AVRational tb1 = ctx->input_pads[MAIN   ].time_base;
> > > > > >     AVRational tb2 = ctx->input_pads[OVERLAY].time_base;
> > > > > >     AVRational tb;
> > > > > >     int exact = 1;
> > > > > > 
> > > > > >     exact *= av_reduce(&tb1.num, &tb1.den, tb1.num, tb1.den, INT32_MAX);
> > > > > >     exact *= av_reduce(&tb2.num, &tb2.den, tb2.num, tb2.den, INT32_MAX);
> > > > > >     exact *= av_reduce(&tb.num, &tb.den,
> > > > > >                        (int64_t)tb1.num * tb2.num, (int64_t)tb1.den * tb2.den, INT_MAX);
> > > > > 
> > > > > your code makes no sense
> > > > > with a/b and c/d
> > > > > 
> > > > > you need something like:
> > > > > reduce(gcd(ad,cb)/bd)
> > > > 
> > > > AVRational = {int, int}
> > > >
> > > 
> > > > int*int may overflow int64_t,
> > > 
> > > not on any supported platform nor would it work if such platform was supported
> > > because that case would then inevtitable fail on all other platforms
> > > 
> > > 
> > > > so I convert {int,int} to {int32_t,int32_t}.
> > > 
> > > I couldnt fit 2 apples in the box so i cut them in half and put 2 halfs in
> > > that solves the problem that i cannot put more than 1 whole apple in
> > > 
> > > 
> > > > 
> > > > Finally I do int32_t*int32_t in the third reduce, so I know that it
> > > > can't overflow int64_t:
> > > >    exact *= av_reduce(&tb.num, &tb.den,
> > > >                       (int64_t)tb1.num * tb2.num, (int64_t)tb1.den * tb2.den, INT_MAX);
> > > 
> > > your code makes absolutely no sense
> > > 
> > > try
> > > tb1=1/6
> > > tb2=1/15
> > > correct result=1/30
> > > your code produces 1/90 this is wrong its not overflow you apply the wrong
> > > math to it like doing a multiplication where you would need divission
> > 
> > Uhm.. yes I'm (I was?) stupid, give me some more time to meditate on
> > this, hope the math is right now, a possibly more correct code should
> > be:
> > 
> > static int config_output(AVFilterLink *outlink)
> > {
> >     AVFilterContext *ctx = outlink->src;
> > 
> >     // common timebase computation:
> >     // tb1 = n1 / d1
> >     // tb2 = n2 / d2
> >     // tb = gcd(n1, n2) / lcm(d1, d2) =
> >     //      gcd(n1, n2) * gcd(d1, d2) / d1*d2
> 
> tb1=5/5
> tb2=1/1
> 
> gcd(5, 1) * gcd(5, 1) / 5*1

Reducing tb1 and tb2 should solve the issue in this case.

    int exact;
    // common timebase computation:
    // tb1 = n1 / d1
    // tb2 = n2 / d2
    // tb = gcd(n1, n2) / lcm(d1, d2) =
    //      gcd(n1, n2) * gcd(d1, d2) / d1*d2
    AVRational *tb1 = &ctx->input_pads[MAIN   ].time_base;
    AVRational *tb2 = &ctx->input_pads[OVERLAY].time_base;
    AVRational tb;
    av_reduce(&tb1->num, &tb1->den, tb1->num, tb1->den, INT_MAX);
    av_reduce(&tb2->num, &tb2->den, tb2->num, tb2->den, INT_MAX);
    exact = av_reduce(&tb.num, &tb.den,
                      (int64_t)av_gcd(tb1->num, tb2->num) * av_gcd(tb1->den, tb2->den),
                      (int64_t)tb1->den * tb2->den, INT_MAX);
    av_log(ctx, AV_LOG_INFO,
           "main_tb:%d/%d overlay_tb:%d/%d -> tb:%d/%d exact:%d\n",
           tb1->num, tb1->den, tb2->num, tb2->den, tb.num, tb.den, exact);
    *tb1 = *tb2 = tb;
    if (!exact)
        av_log(ctx, AV_LOG_WARNING,
               "Timestamp conversion inexact, timestamp information loss may occurr\n");

Regards.
-- 
FFmpeg = Fantastic and Foolish Muttering Power Erroneous Gadget