RepoMirrors/ffmpeg
1
0
Fork 0
mirror of https://git.ffmpeg.org/ffmpeg.git synced 2024-12-15 19:55:08 +00:00
Code Issues Projects Releases Wiki Activity
adf5dc90a9
ffmpeg/libavcodec/aacenc_ltp.c
Rostislav Pehlivanov d2ae5f77c6 aacenc: add SIMD optimizations for abs_pow34 and quantization 
Performance improvements:

quant_bands:
with:     681 decicycles in quant_bands, 8388453 runs,    155 skips
without: 1190 decicycles in quant_bands, 8388386 runs,    222 skips
Around 42% for the function

Twoloop coder:

abs_pow34:
with/without: 7.82s/8.17s
Around 4% for the entire encoder

Both:
with/without: 7.15s/8.17s
Around 12% for the entire encoder

Fast coder:

abs_pow34:
with/without: 3.40s/3.77s
Around 10% for the entire encoder

Both:
with/without: 3.02s/3.77s
Around 20% faster for the entire encoder

Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
Tested-by: Michael Niedermayer <michael@niedermayer.cc>
Reviewed-by: James Almer <jamrial@gmail.com>
2016-10-18 21:41:18 +01:00

237 lines
8.3 KiB
C
Raw Blame History

/*
* AAC encoder long term prediction extension
* Copyright (C) 2015 Rostislav Pehlivanov
*
* This file is part of FFmpeg.
*
* FFmpeg is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
* License as published by the Free Software Foundation; either
* version 2.1 of the License, or (at your option) any later version.
*
* FFmpeg is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
* License along with FFmpeg; if not, write to the Free Software
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/
/**
* @file
* AAC encoder long term prediction extension
* @author Rostislav Pehlivanov ( atomnuker gmail com )
*/
#include "aacenc_ltp.h"
#include "aacenc_quantization.h"
#include "aacenc_utils.h"
/**
* Encode LTP data.
*/
void ff_aac_encode_ltp_info(AACEncContext *s, SingleChannelElement *sce,
int common_window)
{
int i;
IndividualChannelStream *ics = &sce->ics;
if (s->profile != FF_PROFILE_AAC_LTP || !ics->predictor_present)
return;
if (common_window)
put_bits(&s->pb, 1, 0);
put_bits(&s->pb, 1, ics->ltp.present);
if (!ics->ltp.present)
return;
put_bits(&s->pb, 11, ics->ltp.lag);
put_bits(&s->pb, 3, ics->ltp.coef_idx);
for (i = 0; i < FFMIN(ics->max_sfb, MAX_LTP_LONG_SFB); i++)
put_bits(&s->pb, 1, ics->ltp.used[i]);
}
void ff_aac_ltp_insert_new_frame(AACEncContext *s)
{
int i, ch, tag, chans, cur_channel, start_ch = 0;
ChannelElement *cpe;
SingleChannelElement *sce;
for (i = 0; i < s->chan_map[0]; i++) {
cpe = &s->cpe[i];
tag = s->chan_map[i+1];
chans = tag == TYPE_CPE ? 2 : 1;
for (ch = 0; ch < chans; ch++) {
sce = &cpe->ch[ch];
cur_channel = start_ch + ch;
/* New sample + overlap */
memcpy(&sce->ltp_state[0], &sce->ltp_state[1024], 1024*sizeof(sce->ltp_state[0]));
memcpy(&sce->ltp_state[1024], &s->planar_samples[cur_channel][2048], 1024*sizeof(sce->ltp_state[0]));
memcpy(&sce->ltp_state[2048], &sce->ret_buf[0], 1024*sizeof(sce->ltp_state[0]));
sce->ics.ltp.lag = 0;
}
start_ch += chans;
}
}
static void get_lag(float *buf, const float *new, LongTermPrediction *ltp)
{
int i, j, lag, max_corr = 0;
float max_ratio;
for (i = 0; i < 2048; i++) {
float corr, s0 = 0.0f, s1 = 0.0f;
const int start = FFMAX(0, i - 1024);
for (j = start; j < 2048; j++) {
const int idx = j - i + 1024;
s0 += new[j]*buf[idx];
s1 += buf[idx]*buf[idx];
}
corr = s1 > 0.0f ? s0/sqrt(s1) : 0.0f;
if (corr > max_corr) {
max_corr = corr;
lag = i;
max_ratio = corr/(2048-start);
}
}
ltp->lag = FFMAX(av_clip_uintp2(lag, 11), 0);
ltp->coef_idx = quant_array_idx(max_ratio, ltp_coef, 8);
ltp->coef = ltp_coef[ltp->coef_idx];
}
static void generate_samples(float *buf, LongTermPrediction *ltp)
{
int i, samples_num = 2048;
if (!ltp->lag) {
ltp->present = 0;
return;
} else if (ltp->lag < 1024) {
samples_num = ltp->lag + 1024;
}
for (i = 0; i < samples_num; i++)
buf[i] = ltp->coef*buf[i + 2048 - ltp->lag];
memset(&buf[i], 0, (2048 - i)*sizeof(float));
}
/**
* Process LTP parameters
* @see Patent WO2006070265A1
*/
void ff_aac_update_ltp(AACEncContext *s, SingleChannelElement *sce)
{
float *pred_signal = &sce->ltp_state[0];
const float *samples = &s->planar_samples[s->cur_channel][1024];
if (s->profile != FF_PROFILE_AAC_LTP)
return;
/* Calculate lag */
get_lag(pred_signal, samples, &sce->ics.ltp);
generate_samples(pred_signal, &sce->ics.ltp);
}
void ff_aac_adjust_common_ltp(AACEncContext *s, ChannelElement *cpe)
{
int sfb, count = 0;
SingleChannelElement *sce0 = &cpe->ch[0];
SingleChannelElement *sce1 = &cpe->ch[1];
if (!cpe->common_window ||
sce0->ics.window_sequence[0] == EIGHT_SHORT_SEQUENCE ||
sce1->ics.window_sequence[0] == EIGHT_SHORT_SEQUENCE) {
sce0->ics.ltp.present = 0;
return;
}
for (sfb = 0; sfb < FFMIN(sce0->ics.max_sfb, MAX_LTP_LONG_SFB); sfb++) {
int sum = sce0->ics.ltp.used[sfb] + sce1->ics.ltp.used[sfb];
if (sum != 2) {
sce0->ics.ltp.used[sfb] = 0;
} else if (sum == 2) {
count++;
}
}
sce0->ics.ltp.present = !!count;
sce0->ics.predictor_present = !!count;
}
/**
* Mark LTP sfb's
*/
void ff_aac_search_for_ltp(AACEncContext *s, SingleChannelElement *sce,
int common_window)
{
int w, g, w2, i, start = 0, count = 0;
int saved_bits = -(15 + FFMIN(sce->ics.max_sfb, MAX_LTP_LONG_SFB));
float *C34 = &s->scoefs[128*0], *PCD = &s->scoefs[128*1];
float *PCD34 = &s->scoefs[128*2];
const int max_ltp = FFMIN(sce->ics.max_sfb, MAX_LTP_LONG_SFB);
if (sce->ics.window_sequence[0] == EIGHT_SHORT_SEQUENCE) {
if (sce->ics.ltp.lag) {
memset(&sce->ltp_state[0], 0, 3072*sizeof(sce->ltp_state[0]));
memset(&sce->ics.ltp, 0, sizeof(LongTermPrediction));
}
return;
}
if (!sce->ics.ltp.lag || s->lambda > 120.0f)
return;
for (w = 0; w < sce->ics.num_windows; w += sce->ics.group_len[w]) {
start = 0;
for (g = 0; g < sce->ics.num_swb; g++) {
int bits1 = 0, bits2 = 0;
float dist1 = 0.0f, dist2 = 0.0f;
if (w*16+g > max_ltp) {
start += sce->ics.swb_sizes[g];
continue;
}
for (w2 = 0; w2 < sce->ics.group_len[w]; w2++) {
int bits_tmp1, bits_tmp2;
FFPsyBand *band = &s->psy.ch[s->cur_channel].psy_bands[(w+w2)*16+g];
for (i = 0; i < sce->ics.swb_sizes[g]; i++)
PCD[i] = sce->coeffs[start+(w+w2)*128+i] - sce->lcoeffs[start+(w+w2)*128+i];
s->abs_pow34(C34, &sce->coeffs[start+(w+w2)*128], sce->ics.swb_sizes[g]);
s->abs_pow34(PCD34, PCD, sce->ics.swb_sizes[g]);
dist1 += quantize_band_cost(s, &sce->coeffs[start+(w+w2)*128], C34, sce->ics.swb_sizes[g],
sce->sf_idx[(w+w2)*16+g], sce->band_type[(w+w2)*16+g],
s->lambda/band->threshold, INFINITY, &bits_tmp1, NULL, 0);
dist2 += quantize_band_cost(s, PCD, PCD34, sce->ics.swb_sizes[g],
sce->sf_idx[(w+w2)*16+g],
sce->band_type[(w+w2)*16+g],
s->lambda/band->threshold, INFINITY, &bits_tmp2, NULL, 0);
bits1 += bits_tmp1;
bits2 += bits_tmp2;
}
if (dist2 < dist1 && bits2 < bits1) {
for (w2 = 0; w2 < sce->ics.group_len[w]; w2++)
for (i = 0; i < sce->ics.swb_sizes[g]; i++)
sce->coeffs[start+(w+w2)*128+i] -= sce->lcoeffs[start+(w+w2)*128+i];
sce->ics.ltp.used[w*16+g] = 1;
saved_bits += bits1 - bits2;
count++;
}
start += sce->ics.swb_sizes[g];
}
}
sce->ics.ltp.present = !!count && (saved_bits >= 0);
sce->ics.predictor_present = !!sce->ics.ltp.present;
/* Reset any marked sfbs */
if (!sce->ics.ltp.present && !!count) {
for (w = 0; w < sce->ics.num_windows; w += sce->ics.group_len[w]) {
start = 0;
for (g = 0; g < sce->ics.num_swb; g++) {
if (sce->ics.ltp.used[w*16+g]) {
for (w2 = 0; w2 < sce->ics.group_len[w]; w2++) {
for (i = 0; i < sce->ics.swb_sizes[g]; i++) {
sce->coeffs[start+(w+w2)*128+i] += sce->lcoeffs[start+(w+w2)*128+i];
}
}
}
start += sce->ics.swb_sizes[g];
}
}
}
}
Reference in New Issue View Git Blame Copy Permalink