[FFmpeg-devel] [PATCH] change AVCodecContext.use_lpc to lpc_type and lpc_passes

[Justin Ruggles]

Hi,

New patch attached.

Michael Niedermayer wrote:

> On Wed, Jul 07, 2010 at 10:01:43PM -0400, Justin Ruggles wrote:
>> Hi,
>>
>> This patch adds 2 new fields to AVCodecContext, lpc_type and lpc_passes,
>> and deprecates use_lpc.  Separating the 2 makes more sense logically and
>> allows for adding of other LPC algorithms.  It also gives the user an
>> explicit option to not use LPC prediction.
>>
>> The generic LPC code and the FLAC and ALAC encoders are modified to use
>> the new fields.  FLAC currently uses the use_lpc field, so I've added
>> some backwards compatibility code.  The ALAC encoder does not, so it
>> adds functionality in addition to using the new fields.  I can separate
>> that into its own patch if desired.
> [...]
>> Index: libavcodec/options.c
>> ===================================================================
>> --- libavcodec/options.c	(revision 24100)
>> +++ libavcodec/options.c	(working copy)
>> @@ -382,7 +382,9 @@
>>  {"ivlc", "intra vlc table", 0, FF_OPT_TYPE_CONST, CODEC_FLAG2_INTRA_VLC, INT_MIN, INT_MAX, V|E, "flags2"},
>>  {"b_sensitivity", "adjusts sensitivity of b_frame_strategy 1", OFFSET(b_sensitivity), FF_OPT_TYPE_INT, 40, 1, INT_MAX, V|E},
>>  {"compression_level", NULL, OFFSET(compression_level), FF_OPT_TYPE_INT, FF_COMPRESSION_DEFAULT, INT_MIN, INT_MAX, V|A|E},
>> +#if LIBAVCODEC_VERSION_MAJOR < 53
>>  {"use_lpc", "sets whether to use LPC mode (FLAC)", OFFSET(use_lpc), FF_OPT_TYPE_INT, -1, INT_MIN, INT_MAX, A|E},
>> +#endif
>>  {"lpc_coeff_precision", "LPC coefficient precision (FLAC)", OFFSET(lpc_coeff_precision), FF_OPT_TYPE_INT, DEFAULT, 0, INT_MAX, A|E},
>>  {"min_prediction_order", NULL, OFFSET(min_prediction_order), FF_OPT_TYPE_INT, -1, INT_MIN, INT_MAX, A|E},
>>  {"max_prediction_order", NULL, OFFSET(max_prediction_order), FF_OPT_TYPE_INT, -1, INT_MIN, INT_MAX, A|E},
>> @@ -417,6 +419,12 @@
>>  {"intra_refresh", "use periodic insertion of intra blocks instead of keyframes", 0, FF_OPT_TYPE_CONST, CODEC_FLAG2_INTRA_REFRESH, INT_MIN, INT_MAX, V|E, "flags2"},
>>  {"crf_max", "in crf mode, prevents vbv from lowering quality beyond this point", OFFSET(crf_max), FF_OPT_TYPE_FLOAT, DEFAULT, 0, 51, V|E},
>>  {"log_level_offset", "set the log level offset", OFFSET(log_level_offset), FF_OPT_TYPE_INT, 0, INT_MIN, INT_MAX },
> 
>> +{"lpc_type", "specify LPC algorithm", OFFSET(lpc_type), FF_OPT_TYPE_INT, FF_LPC_TYPE_DEFAULT, INT_MIN, INT_MAX, A|E},
> 
> min/max dont look correct

fixed.


> [...]
> 
>> @@ -2672,6 +2675,25 @@
>>      float crf_max;
>>  
>>      int log_level_offset;
>> +
>> +    /**
>> +     * Determines which LPC analysis algorithm to use.
>> +     * - encoding: Set by user
>> +     * - decoding: unused
>> +     */
>> +    int lpc_type;
> 
>> +#define FF_LPC_TYPE_DEFAULT   -1
>> +#define FF_LPC_TYPE_NONE       0
>> +#define FF_LPC_TYPE_FIXED      1
>> +#define FF_LPC_TYPE_LEVINSON   2
>> +#define FF_LPC_TYPE_CHOLESKY   3
> 
> this possibly should be an enum

fixed.

> except above iam ok with the changes to files i maintain

And I maintain lpc.c and flacenc.c.

I decided to leave out the more complex changes to the ALAC encoder.  I
will send a separate patch later.

-Justin

-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: lpc_type_and_passes_2.patch
URL: <http://lists.mplayerhq.hu/pipermail/ffmpeg-devel/attachments/20100710/f05cd362/attachment.asc>