[FFmpeg-devel] [PATCH] AAC decoder

[Michael Niedermayer]

On Sun, Jun 01, 2008 at 09:28:27PM +0100, Robert Swain wrote:
> 2008/4/2 Michael Niedermayer <michaelni at gmx.at>:
> > On Tue, Apr 01, 2008 at 04:56:48PM +0200, Andreas ?man wrote:
> >> +    }
> >> +
> >> +    dsputil_init(&ac->dsp, avccontext);
> >> +
> >> +    /* Initialize RNG dither */
> >> +    av_init_random(0x1f2e3d4c, &ac->random_state);
> >> +
> >> +    // -1024 - compensate wrong IMDCT method
> >> +    // 32768 - values in AAC build for ready float->int 16 bit audio, using
> >> +    // BIAS method instead needs values -1<x<1
> >> +    for (i = 0; i < 256; i++)
> >> +        ac->intensity_tab[i] = pow(0.5, (i - 100) / 4.);
> >> +    for (i = 0; i < sizeof(ac->ivquant_tab)/sizeof(ac->ivquant_tab[0]); i++)
> >> +        ac->ivquant_tab[i] = pow(i, 4./3);
> >> +
> >> +    if(ac->dsp.float_to_int16 == ff_float_to_int16_c) {
> >> +        ac->add_bias = 385.0f;
> >> +        ac->sf_scale = 1. / (-1024. * 32768.);
> >> +    } else {
> >> +        ac->add_bias = 0.0f;
> >> +        ac->sf_scale = 1. / -1024.;
> >> +    }
> >> +    for (i = 0; i < 256; i++)
> >> +        ac->pow2sf_tab[i] = pow(2, (i - 100)/4.) * ac->sf_scale;
> >> +
> >> +    if(init_vlc(&ac->mainvlc, 7, sizeof(code)/sizeof(code[0]),
> >> +            bits, sizeof(bits[0]), sizeof(bits[0]),
> >> +            code, sizeof(code[0]), sizeof(code[0]),
> >> +            0) < 0)
> >> +        return -1;
> >> +
> >> +#ifdef AAC_SSR
> >> +    if (ac->audioObjectType == AOT_AAC_SSR) {
> >> +        ff_mdct_init(&ac->mdct, 9, 1);
> >> +        ff_mdct_init(&ac->mdct_small, 6, 1);
> >> +        // windows init
> >> +        ff_kbd_window_init(ac->kbd_long_1024, 4.0, 256);
> >> +        ff_kbd_window_init(ac->kbd_short_128, 6.0, 32);
> >> +        sine_window_init(ac->sine_long_1024, 512);
> >> +        sine_window_init(ac->sine_short_128, 64);
> >> +        ssr_context_init(&ac->ssrctx);
> >> +    } else {
> >> +#endif /* AAC_SSR */
> >> +        ff_mdct_init(&ac->mdct, 11, 1);
> >> +        ff_mdct_init(&ac->mdct_small, 8, 1);
> >> +        // windows init
> >> +        ff_kbd_window_init(ac->kbd_long_1024, 4.0, 1024);
> >> +        ff_kbd_window_init(ac->kbd_short_128, 6.0, 128);
> >> +        sine_window_init(ac->sine_long_1024, 2048);
> >> +        sine_window_init(ac->sine_short_128, 256);
> >
> > There are 4 different sine windows, why are they duplicated in each context?
> > They should be in static arrays generated at runtime or static const tables
> > hardcoded depening on ENABLE_HARDCODED_TABLES. This applies likely to other
> > tables as well.
> 
> One could do the same for the tables of Kaiser-Bessel Derived window
> coefficients as well. I have done this for the 'main' AAC (i.e. not
> the SSR case) but I'm not sure whether to put the effort into doing
> the same for SSR and handling the different arrays if we are going to
> rip out the SSR and LTP code anyway. What are your demands? :)

I do not demand ENABLE_HARDCODED_TABLES for all tables no, its nice to
have if someone likes to work on that ...
What i do "demand" is that constant tables should not be in the context
and that const tables which are identical between codecs should be shared.
sin, cos, pow2 really are quite common i doubt aac is the only thing using
them ...

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

Complexity theory is the science of finding the exact solution to an
approximation. Benchmarking OTOH is finding an approximation of the exact
-------------- 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/20080601/836e03cd/attachment.pgp>