[FFmpeg-devel] Hacks on FATE

Ramiro Polla ramiro.polla
Sat Sep 26 17:05:57 CEST 2009


On Thu, Sep 24, 2009 at 5:04 PM, Reimar D?ffinger
<Reimar.Doeffinger at gmx.de> wrote:
> On Thu, Sep 24, 2009 at 11:42:30PM +0400, Michael Kostylev wrote:
>> > Michael Kostylev runs the machine, maybe he can comment...
>>
>> The issue was discussed on MPlayer-dev-eng:
>> http://lists.mplayerhq.hu/pipermail/mplayer-dev-eng/2009-July/061865.html
>>
>> On i386 (HAVE_7REGS && HAVE_TEN_OPERANDS) may be equal to zero
>> (for example, on my machine %ebp is not available), then
>> libavcodec/x86/mlpdsp.c can be built without -Wa,--divide.
>
> Hm. It seems to me that mlp assumes that the max samplerate is a multiple
> of 48000, thus I think the following patch improves the code and
> fixes this issue:
> Index: libavcodec/mlp.h
> ===================================================================
> --- libavcodec/mlp.h ? ?(revision 19461)
> +++ libavcodec/mlp.h ? ?(working copy)
> @@ -47,13 +47,15 @@
> ?*/
> ?#define MAX_SUBSTREAMS ? ? ?3
>
> +/** which multiple of 48000 the maximum sample rate is */
> +#define MAX_RATEFACTOR ? ? ?4
> ?/** maximum sample frequency seen in files */
> -#define MAX_SAMPLERATE ? ? ?192000
> +#define MAX_SAMPLERATE ? ? ?(MAX_RATEFACTOR * 48000)
>
> ?/** maximum number of audio samples within one access unit */
> -#define MAX_BLOCKSIZE ? ? ? (40 * (MAX_SAMPLERATE / 48000))
> +#define MAX_BLOCKSIZE ? ? ? (40 * MAX_RATEFACTOR)
> ?/** next power of two greater than MAX_BLOCKSIZE */
> -#define MAX_BLOCKSIZE_POW2 ?(64 * (MAX_SAMPLERATE / 48000))
> +#define MAX_BLOCKSIZE_POW2 ?(64 * MAX_RATEFACTOR)
>
> ?/** number of allowed filters */
> ?#define NUM_FILTERS ? ? ? ? 2

ok, can you suggest a commit message?

Ramiro Polla



More information about the ffmpeg-devel mailing list