February 2009 - FFmpeg-soc - lists.ffmpeg.org

[soc]: r4130 - amr/amrnbfloatdec.c
by kmalaussene 23 Feb '09

23 Feb '09

Author: kmalaussene Date: Mon Feb 23 02:24:24 2009 New Revision: 4130 Log: Add parameters to do_phase_dispersion in order to make the code more readable. Modified: amr/amrnbfloatdec.c Modified: amr/amrnbfloatdec.c ============================================================================== --- amr/amrnbfloatdec.c Mon Feb 23 01:39:40 2009 (r4129) +++ amr/amrnbfloatdec.c Mon Feb 23 02:24:24 2009 (r4130) @@ -699,7 +699,7 @@ static float fixed_gain_prediction(float * * @param p the context */ -void do_phase_dispersion(AMRContext *p) +void do_phase_dispersion(AMRContext *p, float *fixed_vector) { int ir_filter_strength; @@ -746,8 +746,8 @@ void do_phase_dispersion(AMRContext *p) // FIXME: extra memcpy (ff_celp_convolve_circf needs in and out not to overlap) float fixed_vector_temp[AMR_SUBFRAME_SIZE]; - memcpy(fixed_vector_temp, p->fixed_vector, sizeof(fixed_vector_temp)); - ff_celp_convolve_circf(p->fixed_vector, fixed_vector_temp, + memcpy(fixed_vector_temp, fixed_vector, sizeof(fixed_vector_temp)); + ff_celp_convolve_circf(fixed_vector, fixed_vector_temp, filters[ir_filter_strength], AMR_SUBFRAME_SIZE); } @@ -970,7 +970,7 @@ static int amrnb_decode_frame(AVCodecCon } } - do_phase_dispersion(p); + do_phase_dispersion(p, p->fixed_vector); /*** end of pre-processing ***/

1 0

[soc]: r4129 - amr/amrnbfloatdec.c
by kmalaussene 23 Feb '09

23 Feb '09

Author: kmalaussene Date: Mon Feb 23 01:39:40 2009 New Revision: 4129 Log: Remove unused variables. Modified: amr/amrnbfloatdec.c Modified: amr/amrnbfloatdec.c ============================================================================== --- amr/amrnbfloatdec.c Mon Feb 23 00:40:39 2009 (r4128) +++ amr/amrnbfloatdec.c Mon Feb 23 01:39:40 2009 (r4129) @@ -773,7 +773,7 @@ void do_phase_dispersion(AMRContext *p) static int synthesis(AMRContext *p, float *excitation, float *lpc, float *samples, uint8_t overflow) { - int i, j, overflow_temp = 0; + int i, overflow_temp = 0; // if an overflow has been detected, the pitch vector is scaled down by a // factor of 4

1 0

[soc]: r4128 - amr/amrnbfloatdec.c
by kmalaussene 22 Feb '09

22 Feb '09

Author: kmalaussene Date: Mon Feb 23 00:40:39 2009 New Revision: 4128 Log: Use ff_celp_lp_synthesis_filterf in synthesis. Modified: amr/amrnbfloatdec.c Modified: amr/amrnbfloatdec.c ============================================================================== --- amr/amrnbfloatdec.c Sun Feb 22 23:34:42 2009 (r4127) +++ amr/amrnbfloatdec.c Mon Feb 23 00:40:39 2009 (r4128) @@ -356,6 +356,9 @@ static void lsp2lpc(float *lsp, float *l lsp_double[i] = lsp[i]; ff_celp_lspf2lpc(lsp_double, lpc_coeffs); + + for(i=0; i<LP_FILTER_ORDER; i++) + lpc_coeffs[i] = -lpc_coeffs[i]; } /// @} @@ -800,11 +803,9 @@ static int synthesis(AMRContext *p, floa ff_apply_gain_ctrl(excitation, excitation, excitation_temp, AMR_SUBFRAME_SIZE); } + ff_celp_lp_synthesis_filterf(samples, lpc, excitation, AMR_SUBFRAME_SIZE, LP_FILTER_ORDER); + for(i=0; i<AMR_SUBFRAME_SIZE; i++) { - samples[i] = excitation[i]; - for(j=0; j<LP_FILTER_ORDER; j++) { - samples[i] -= lpc[j]*samples[i-j-1]; - } // detect overflow if(fabsf(samples[i])>1.0) { overflow_temp = 1;

1 0

[soc]: r4127 - amr/amrnbfloatdec.c
by kmalaussene 22 Feb '09

22 Feb '09

Author: kmalaussene Date: Sun Feb 22 23:34:42 2009 New Revision: 4127 Log: Fix inclusion in Doxy group. Modified: amr/amrnbfloatdec.c Modified: amr/amrnbfloatdec.c ============================================================================== --- amr/amrnbfloatdec.c Sun Feb 22 23:33:12 2009 (r4126) +++ amr/amrnbfloatdec.c Sun Feb 22 23:34:42 2009 (r4127) @@ -691,8 +691,6 @@ static float fixed_gain_prediction(float /// @defgroup amr_pre_processing pre-processing functions /// @{ -/// @} - /** * adaptive phase dispersion; forming of total excitation * @@ -754,6 +752,8 @@ void do_phase_dispersion(AMRContext *p) p->prev_ir_filter_strength = ir_filter_strength; } +/// @} + /// @defgroup amr_synthesis synthesis functions /// @{

1 0

[soc]: r4126 - in amr: amr-ffmpeg.diff amrnbfloatdec.c
by kmalaussene 22 Feb '09

22 Feb '09

Author: kmalaussene Date: Sun Feb 22 23:33:12 2009 New Revision: 4126 Log: Use ff_celp_convolve_circf in celp_filtes.[ch] instead of convolve_circ. Modified: amr/amr-ffmpeg.diff amr/amrnbfloatdec.c Modified: amr/amr-ffmpeg.diff ============================================================================== --- amr/amr-ffmpeg.diff Sun Feb 22 22:41:33 2009 (r4125) +++ amr/amr-ffmpeg.diff Sun Feb 22 23:33:12 2009 (r4126) @@ -136,3 +136,67 @@ Index: libavcodec/qcelp_lsp.c { lpc[i] *= bandwith_expansion_coeff; bandwith_expansion_coeff *= QCELP_BANDWITH_EXPANSION_COEFF; +Index: libavcodec/celp_filters.c +=================================================================== +--- libavcodec/celp_filters.c (revision 17468) ++++ libavcodec/celp_filters.c (working copy) +@@ -85,6 +85,31 @@ + return 0; + } + ++void ff_celp_convolve_circf( ++ float* fc_out, ++ const float* fc_in, ++ const float* filter, ++ int len) ++{ ++ int i, k; ++ ++ memset(fc_out, 0, len * sizeof(float)); ++ ++ /* Since there are few pulses over an entire subframe (i.e. almost ++ all fc_in[i] are zero) it is faster to loop over fc_in first. */ ++ for(i=0; i<len; i++) ++ { ++ if(fc_in[i]) ++ { ++ for(k=0; k<i; k++) ++ fc_out[k] += fc_in[i] * filter[len + k - i]; ++ ++ for(k=i; k<len; k++) ++ fc_out[k] += fc_in[i] * filter[ k - i]; ++ } ++ } ++} ++ + void ff_celp_lp_synthesis_filterf( + float *out, + const float* filter_coeffs, +Index: libavcodec/celp_filters.h +=================================================================== +--- libavcodec/celp_filters.h (revision 17468) ++++ libavcodec/celp_filters.h (working copy) +@@ -70,6 +70,23 @@ + int rounder); + + /** ++ * Circularly convolve fixed vector with a phase dispersion impulse ++ * response filter (D.6.2 of G.729 and 6.1.5 of AMR). ++ * @param fc_out vector with filter applied ++ * @param fc_in source vector ++ * @param filter phase filter coefficients ++ * ++ * fc_out[n] = sum(i,0,len-1){ fc_in[i] * filter[(len + n - i)%len] } ++ * ++ * \note fc_in and fc_out should not overlap! ++ */ ++void ff_celp_convolve_circf( ++ float* fc_out, ++ const float* fc_in, ++ const float* filter, ++ int len); ++ ++/** + * LP synthesis filter. + * @param out [out] pointer to output buffer + * - the array out[-filter_length, -1] must Modified: amr/amrnbfloatdec.c ============================================================================== --- amr/amrnbfloatdec.c Sun Feb 22 22:41:33 2009 (r4125) +++ amr/amrnbfloatdec.c Sun Feb 22 23:33:12 2009 (r4126) @@ -36,6 +36,7 @@ #include "libavutil/common.h" #include "internal.h" #include "celp_math.h" +#include "celp_filters.h" #include "amrnbfloatdata.h" void ff_celp_lspf2lpc(const double *lspf, float *lpc); @@ -690,42 +691,6 @@ static float fixed_gain_prediction(float /// @defgroup amr_pre_processing pre-processing functions /// @{ -/** - * Circularly convolve the fixed vector with a phase dispersion impulse response - * filter. - * - * @param fixed_vector pointer to the fixed vector - * @param ir_filter pointer to the impulse response filter - */ - -static void convolve_circ(float *fixed_vector, const float *ir_filter) -{ - int i, j, k; - int npulses = 0, pulse_positions[AMR_SUBFRAME_SIZE]; - float fixed_vector_temp[AMR_SUBFRAME_SIZE]; - - memcpy(fixed_vector_temp, fixed_vector, AMR_SUBFRAME_SIZE*sizeof(float)); - memset(fixed_vector, 0, AMR_SUBFRAME_SIZE*sizeof(float)); - - // Find non-zero pulses (most are zero) - for(i=0; i<AMR_SUBFRAME_SIZE; i++) { - if(fixed_vector_temp[i]) { - pulse_positions[npulses] = i; - npulses++; - } - } - - for(i=0; i<npulses; i++) { - k = 0; - for(j=pulse_positions[i]; j<AMR_SUBFRAME_SIZE; j++) { - fixed_vector[j] += fixed_vector_temp[pulse_positions[i]]*ir_filter[k++]; - } - for(j=0; j<pulse_positions[i]; j++) { - fixed_vector[j] += fixed_vector_temp[pulse_positions[i]]*ir_filter[k++]; - } - } -} - /// @} /** @@ -777,7 +742,12 @@ void do_phase_dispersion(AMRContext *p) const float **filters = p->cur_frame_mode == MODE_795 ? ir_filters_lookup_MODE_795 : ir_filters_lookup; // circularly convolve the fixed vector with the impulse response - convolve_circ(p->fixed_vector, filters[ir_filter_strength]); + // FIXME: extra memcpy (ff_celp_convolve_circf needs in and out not to overlap) + float fixed_vector_temp[AMR_SUBFRAME_SIZE]; + + memcpy(fixed_vector_temp, p->fixed_vector, sizeof(fixed_vector_temp)); + ff_celp_convolve_circf(p->fixed_vector, fixed_vector_temp, + filters[ir_filter_strength], AMR_SUBFRAME_SIZE); } // update ir filter strength history

1 0

[soc]: r4125 - amr/amrnbfloatdec.c
by kmalaussene 22 Feb '09

22 Feb '09

Author: kmalaussene Date: Sun Feb 22 22:41:33 2009 New Revision: 4125 Log: Fix previous commit, use >= instead of <=. Modified: amr/amrnbfloatdec.c Modified: amr/amrnbfloatdec.c ============================================================================== --- amr/amrnbfloatdec.c Sun Feb 22 22:39:46 2009 (r4124) +++ amr/amrnbfloatdec.c Sun Feb 22 22:41:33 2009 (r4125) @@ -940,10 +940,10 @@ static int amrnb_decode_frame(AVCodecCon if(p->cur_frame_mode == MODE_122 || p->cur_frame_mode == MODE_795) { p->pitch_gain[4] = qua_gain_pit [amr_subframe->p_gain]; p->fixed_gain_factor = qua_gain_code[amr_subframe->fixed_gain]; - }else if(p->cur_frame_mode <= MODE_67) { + }else if(p->cur_frame_mode >= MODE_67) { p->pitch_gain[4] = gains_high[amr_subframe->p_gain][0]; p->fixed_gain_factor = gains_high[amr_subframe->p_gain][1]; - }else if(p->cur_frame_mode <= MODE_515) { + }else if(p->cur_frame_mode >= MODE_515) { p->pitch_gain[4] = gains_low[amr_subframe->p_gain][0]; p->fixed_gain_factor = gains_low[amr_subframe->p_gain][1]; }else {

1 0

[soc]: r4124 - amr/amrnbfloatdec.c
by kmalaussene 22 Feb '09

22 Feb '09

Author: kmalaussene Date: Sun Feb 22 22:39:46 2009 New Revision: 4124 Log: Replace multiple mode == || by 1 mode <=. Modified: amr/amrnbfloatdec.c Modified: amr/amrnbfloatdec.c ============================================================================== --- amr/amrnbfloatdec.c Sun Feb 22 20:55:14 2009 (r4123) +++ amr/amrnbfloatdec.c Sun Feb 22 22:39:46 2009 (r4124) @@ -940,11 +940,10 @@ static int amrnb_decode_frame(AVCodecCon if(p->cur_frame_mode == MODE_122 || p->cur_frame_mode == MODE_795) { p->pitch_gain[4] = qua_gain_pit [amr_subframe->p_gain]; p->fixed_gain_factor = qua_gain_code[amr_subframe->fixed_gain]; - }else if(p->cur_frame_mode == MODE_67 || p->cur_frame_mode == MODE_74 || - p->cur_frame_mode == MODE_102) { + }else if(p->cur_frame_mode <= MODE_67) { p->pitch_gain[4] = gains_high[amr_subframe->p_gain][0]; p->fixed_gain_factor = gains_high[amr_subframe->p_gain][1]; - }else if(p->cur_frame_mode == MODE_515 || p->cur_frame_mode == MODE_59) { + }else if(p->cur_frame_mode <= MODE_515) { p->pitch_gain[4] = gains_low[amr_subframe->p_gain][0]; p->fixed_gain_factor = gains_low[amr_subframe->p_gain][1]; }else {

1 0

[soc]: r4123 - amr/amrnbfloatdec.c
by kmalaussene 22 Feb '09

22 Feb '09

Author: kmalaussene Date: Sun Feb 22 20:55:14 2009 New Revision: 4123 Log: Reindent after last commit. Modified: amr/amrnbfloatdec.c Modified: amr/amrnbfloatdec.c ============================================================================== --- amr/amrnbfloatdec.c Sun Feb 22 20:54:41 2009 (r4122) +++ amr/amrnbfloatdec.c Sun Feb 22 20:55:14 2009 (r4123) @@ -429,7 +429,7 @@ static void decode_pitch_lag(int *lag_in if(mode == MODE_795) { *lag_int += av_clip(prev_lag_int - 10, PITCH_LAG_MIN, PITCH_LAG_MAX - 19); }else { - *lag_int += av_clip(prev_lag_int - 5, PITCH_LAG_MIN, PITCH_LAG_MAX - 9); + *lag_int += av_clip(prev_lag_int - 5, PITCH_LAG_MIN, PITCH_LAG_MAX - 9); } } }

1 0

[soc]: r4122 - amr/amrnbfloatdec.c
by kmalaussene 22 Feb '09

22 Feb '09

Author: kmalaussene Date: Sun Feb 22 20:54:41 2009 New Revision: 4122 Log: Fix decoding of pitch lag for subframe 2 & 4 in MODE_795. Modified: amr/amrnbfloatdec.c Modified: amr/amrnbfloatdec.c ============================================================================== --- amr/amrnbfloatdec.c Sun Feb 22 17:42:51 2009 (r4121) +++ amr/amrnbfloatdec.c Sun Feb 22 20:54:41 2009 (r4122) @@ -426,7 +426,11 @@ static void decode_pitch_lag(int *lag_in // 10923>>15 is approximately 1/3 *lag_int = ( ((pitch_index + 2)*10923)>>15 ) - 1; *lag_frac = pitch_index - *lag_int * 3 - 2; + if(mode == MODE_795) { + *lag_int += av_clip(prev_lag_int - 10, PITCH_LAG_MIN, PITCH_LAG_MAX - 19); + }else { *lag_int += av_clip(prev_lag_int - 5, PITCH_LAG_MIN, PITCH_LAG_MAX - 9); + } } } }

1 0

[soc]: r4121 - libavfilter/diffs/02_ffmpeg_filters.diff
by stefano 22 Feb '09

22 Feb '09

Author: stefano Date: Sun Feb 22 17:42:51 2009 New Revision: 4121 Log: Apply whitespaces cosmetics/vertical align for improving readability. Modified: libavfilter/diffs/02_ffmpeg_filters.diff Modified: libavfilter/diffs/02_ffmpeg_filters.diff ============================================================================== --- libavfilter/diffs/02_ffmpeg_filters.diff Sun Feb 22 11:42:29 2009 (r4120) +++ libavfilter/diffs/02_ffmpeg_filters.diff Sun Feb 22 17:42:51 2009 (r4121) @@ -164,9 +164,9 @@ Index: ffmpeg.c + (frame_padtop + frame_padbottom))) { + char crop_args[255]; + AVFilterContext *filt_scale; -+ snprintf(crop_args, 255, "%d:%d", codec->width - -+ (frame_padleft + frame_padright), -+ codec->height - (frame_padtop + frame_padbottom)); ++ snprintf(crop_args, 255, "%d:%d", ++ codec->width - (frame_padleft + frame_padright), ++ codec->height - (frame_padtop + frame_padbottom)); + filt_scale = avfilter_open(avfilter_get_by_name("scale"), NULL); + if (!filt_scale) + return -1;

1 0