[FFmpeg-devel] [PATCH] [3/3] SIPR 16k decoder

[Måns Rullgård]

"Ronald S. Bultje" <rsbultje at gmail.com> writes:

> Hi,
>
> On Wed, Jan 13, 2010 at 11:15 PM, Vitor Sessak <vitor1001 at gmail.com> wrote:
>> And now the decoder code.
>
> +static void lsf2lsp(const float *lsf, double *lsp)
> +{
> +    int i;
> +
> +    for (i = 0; i < LP_FILTER_ORDER_16k; i++)
> +        lsp[i] = cosf(lsf[i]);
> +}
>
> So, in light of all recent discussion on types for whatnot, let's settle this.
>
> What type should LSFs, LSPs and LPCs be? I mean, all decoders do this.
> We can easily share it as long as we agree on this silly little thing.

Is there any particular reason for using double there?
Single-precision float is usually much faster, especially if it can be
simded.

BTW, watch out for cosf() missing on the usual broken systems.

-- 
M?ns Rullg?rd
mans at mansr.com