[FFmpeg-devel] [PATCH] Common ACELP routines (3/3) - LPC decoding routines

[Vladimir Voroshilov]

Michael Niedermayer wrote: 
> On Fri, Apr 25, 2008 at 10:24:25AM +0700, Vladimir Voroshilov wrote:
> > Michael Niedermayer wrote: 
> > > On Fri, Apr 25, 2008 at 08:26:35AM +0700, Vladimir Voroshilov wrote:
> > > [...]
> > > > +void ff_acelp_reorder_lsf(int16_t* lsfq, int lsfq_min_distance, int lsfq_min, int lsfq_max)
> > > > +{
> > > > +    int i, j;
> > > > +
> > > > +    /* sort lsfq in ascending order. float bubble agorithm, O(n)*/
> > > 
> > > > +    for(i=9; i<9; i++)
> > > 
> > > i do not think that will work
> > 
> > It passed all tests what means that data were already ordered.
> > Anyway since there is nothing in spec about LSFs ordering I'd
> > prefer keep sorting there.
> [...]
> 
> > +/**
> > + * \brief decodes polynomial coefficients from LSP
> > + * \param f [out] (Q24) decoded polynomial coefficients
> > + * \param lsp (Q15) LSP coefficients
> > + */
> > +static void lsp2poly(int* f, const int16_t* lsp)
> > +{
> > +    int i, j;
> > +
> > +    f[0] = 0x1000000;          // 1.0 in Q24
> > +    f[1] = -lsp[0] << 10;      // *2 and Q15 -> Q24
> > +
> > +    for(i=2; i<=5; i++)
> > +    {
> > +        f[i] = f[i-2];
> > +
> > +        for(j=i; j>1; j--)
> > +#ifdef G729_BITEXACT
> > +            f[j] -= (mul_32_16(f[j-1]>>1, lsp[2*i-2])<<2) - f[j-2];
> > +#else
> > +            f[j] -= (mul_32_16(f[j-1], lsp[2*i-2])<<1) - f[j-2];
> 
> mul_32_16 does >>15, you do <<1 afterwards, thats redundant and looses
> precission, you could as well >> by 14

I've found that i can use existing MULH and MULL (and replaced it
everywhere in local tree already).
I've defined FRAC_BITS=15 before including mathops.h and
called MULL everywhere except this place.
I'm trying to eliminate extra shifts here by reducing fixed point base
(looks like two least significant bits of f[i] are always zero, but I'm not
sure).

> 
> 
> [...]
> > +/**
> > + * \brief Convert LSF to LSP
> > + * \param lsp [out] LSP coefficients (-(1<<15) <= lsp < (1<<15))
> > + * \param lsf LSF coefficients (0 <= lsf < PI * (1<<13))
> > + *
> > + * \remark It is safe to pass the same array in lsf and lsp parameters
> > + */
> > +void ff_acelp_lsp2lpc(int16_t* lp, const int16_t* lsp);
> 
> the comment and the function names contradict each other lsp2lpc vs. lsf2lsp
> 
> 
> > +
> > +/**
> > + * \brief LSP to LP conversion (3.2.6 of G.729)
> > + * \param lp [out] (Q12) decoded LP coefficients
> > + * \param lsp (Q15) LSP coefficients
> > + */
> > +void ff_acelp_lsf2lsp(int16_t *lsp, const int16_t *lsf);

> same issue

Fixed in local tree.

 
> also please change the Q12/15 notation to some normal fixed point one with
> min/one/max

Like this?:

/**
 * \param param1 bla bla bla
...
 * \return some value
 *
 *  allowed values for param1:
 *   min: -0x10000
 *   max:  0x10000
 *   one   0x01000
 */

Or something different?
Give me sample please.

-- 
Regards,
Vladimir Voroshilov mailto:voroshil at gmail.com
Omsk State University
JID: voroshil at jabber.ru
ICQ: 95587719