virtualenv/FFmpeg
1
0
Fork 0
mirror of https://github.com/FFmpeg/FFmpeg.git synced 2024-12-12 19:18:44 +02:00
Code Issues Releases Activity
8c8072c29c
FFmpeg/libavcodec/aac/aacdec_dsp_template.c

689 lines
30 KiB
C
Raw Normal View History

aacdec: template scalefactor dequantization separately
2024-03-13 22:53:49 +02:00
/*
* AAC decoder
* Copyright (c) 2005-2006 Oded Shimon ( ods15 ods15 dyndns org )
* Copyright (c) 2006-2007 Maxim Gavrilov ( maxim.gavrilov gmail com )
* Copyright (c) 2008-2013 Alex Converse <alex.converse@gmail.com>
*
* AAC LATM decoder
* Copyright (c) 2008-2010 Paul Kendall <paul@kcbbs.gen.nz>
* Copyright (c) 2010 Janne Grunau <janne-libav@jannau.net>
*
* AAC decoder fixed-point implementation
* Copyright (c) 2013
* MIPS Technologies, Inc., California.
*
* 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
*/
aacdec: move aacdec.h into libavcodec/aac
2024-03-19 21:19:06 +02:00
#include "aacdec.h"
aacdec: template TNS application separately
2024-03-13 23:20:59 +02:00
#include "libavcodec/lpc_functions.h"
aacdec: template scalefactor dequantization separately
2024-03-13 22:53:49 +02:00
#include "libavcodec/aactab.h"
/**
* Convert integer scalefactors to the decoder's native expected
* scalefactor values.
*/
static void AAC_RENAME(dequant_scalefactors)(SingleChannelElement *sce)
{
IndividualChannelStream *ics = &sce->ics;
const int *sfo = sce->sfo;
INTFLOAT *sf = sce->AAC_RENAME(sf);
aacdec: move from scalefactor ranged arrays to flat arrays AAC uses an unconventional system to send scalefactors (the volume+quantization value for each band). Each window is split into either 1 or 8 blocks (long vs short), and transformed separately from one another, with the coefficients for each being also completely independent. The scalefactors slightly increase from 64 (long) to 128 (short) to accomodate better per-block-per-band volume for each window. To reduce overhead, the codec signals scalefactor sizes in an obtuse way, where each group's scalefactor types are sent via a variable length decoding, with a range. But our decoder was written in a way where those ranges were carried through the entire decoder, and to actually read them you had to use the range. Instead of having a dedicated array with a range for each scalefactor, just let the decoder directly index each scalefactor. This also switches the form of quantized scalefactors to the format the spec uses, where for intensity stereo and regular, scalefactors are stored in a scalefactor - 100 form, rather than as-is. USAC gets rid of the complex scalefactor handling. This commit permits for code sharing between both.
2024-05-14 19:07:43 +02:00
int idx = 0;
for (int g = 0; g < ics->num_window_groups; g++) {
for (int sfb = 0; sfb < ics->max_sfb; sfb++, idx++) {
switch (sce->band_type[g*ics->max_sfb + sfb]) {
aacdec: template scalefactor dequantization separately
2024-03-13 22:53:49 +02:00
case ZERO_BT:
aacdec: move from scalefactor ranged arrays to flat arrays AAC uses an unconventional system to send scalefactors (the volume+quantization value for each band). Each window is split into either 1 or 8 blocks (long vs short), and transformed separately from one another, with the coefficients for each being also completely independent. The scalefactors slightly increase from 64 (long) to 128 (short) to accomodate better per-block-per-band volume for each window. To reduce overhead, the codec signals scalefactor sizes in an obtuse way, where each group's scalefactor types are sent via a variable length decoding, with a range. But our decoder was written in a way where those ranges were carried through the entire decoder, and to actually read them you had to use the range. Instead of having a dedicated array with a range for each scalefactor, just let the decoder directly index each scalefactor. This also switches the form of quantized scalefactors to the format the spec uses, where for intensity stereo and regular, scalefactors are stored in a scalefactor - 100 form, rather than as-is. USAC gets rid of the complex scalefactor handling. This commit permits for code sharing between both.
2024-05-14 19:07:43 +02:00
sf[idx] = FIXR(0.);
aacdec: template scalefactor dequantization separately
2024-03-13 22:53:49 +02:00
break;
case INTENSITY_BT: /* fallthrough */
case INTENSITY_BT2:
#if USE_FIXED
aacdec: move from scalefactor ranged arrays to flat arrays AAC uses an unconventional system to send scalefactors (the volume+quantization value for each band). Each window is split into either 1 or 8 blocks (long vs short), and transformed separately from one another, with the coefficients for each being also completely independent. The scalefactors slightly increase from 64 (long) to 128 (short) to accomodate better per-block-per-band volume for each window. To reduce overhead, the codec signals scalefactor sizes in an obtuse way, where each group's scalefactor types are sent via a variable length decoding, with a range. But our decoder was written in a way where those ranges were carried through the entire decoder, and to actually read them you had to use the range. Instead of having a dedicated array with a range for each scalefactor, just let the decoder directly index each scalefactor. This also switches the form of quantized scalefactors to the format the spec uses, where for intensity stereo and regular, scalefactors are stored in a scalefactor - 100 form, rather than as-is. USAC gets rid of the complex scalefactor handling. This commit permits for code sharing between both.
2024-05-14 19:07:43 +02:00
sf[idx] = 100 - (sfo[idx] + 100);
aacdec: template scalefactor dequantization separately
2024-03-13 22:53:49 +02:00
#else
aacdec: move from scalefactor ranged arrays to flat arrays AAC uses an unconventional system to send scalefactors (the volume+quantization value for each band). Each window is split into either 1 or 8 blocks (long vs short), and transformed separately from one another, with the coefficients for each being also completely independent. The scalefactors slightly increase from 64 (long) to 128 (short) to accomodate better per-block-per-band volume for each window. To reduce overhead, the codec signals scalefactor sizes in an obtuse way, where each group's scalefactor types are sent via a variable length decoding, with a range. But our decoder was written in a way where those ranges were carried through the entire decoder, and to actually read them you had to use the range. Instead of having a dedicated array with a range for each scalefactor, just let the decoder directly index each scalefactor. This also switches the form of quantized scalefactors to the format the spec uses, where for intensity stereo and regular, scalefactors are stored in a scalefactor - 100 form, rather than as-is. USAC gets rid of the complex scalefactor handling. This commit permits for code sharing between both.
2024-05-14 19:07:43 +02:00
sf[idx] = ff_aac_pow2sf_tab[-sfo[idx] - 100 + POW_SF2_ZERO];
aacdec: template scalefactor dequantization separately
2024-03-13 22:53:49 +02:00
#endif /* USE_FIXED */
break;
case NOISE_BT:
#if USE_FIXED
aacdec: move from scalefactor ranged arrays to flat arrays AAC uses an unconventional system to send scalefactors (the volume+quantization value for each band). Each window is split into either 1 or 8 blocks (long vs short), and transformed separately from one another, with the coefficients for each being also completely independent. The scalefactors slightly increase from 64 (long) to 128 (short) to accomodate better per-block-per-band volume for each window. To reduce overhead, the codec signals scalefactor sizes in an obtuse way, where each group's scalefactor types are sent via a variable length decoding, with a range. But our decoder was written in a way where those ranges were carried through the entire decoder, and to actually read them you had to use the range. Instead of having a dedicated array with a range for each scalefactor, just let the decoder directly index each scalefactor. This also switches the form of quantized scalefactors to the format the spec uses, where for intensity stereo and regular, scalefactors are stored in a scalefactor - 100 form, rather than as-is. USAC gets rid of the complex scalefactor handling. This commit permits for code sharing between both.
2024-05-14 19:07:43 +02:00
sf[idx] = -(100 + sfo[idx]);
aacdec: template scalefactor dequantization separately
2024-03-13 22:53:49 +02:00
#else
aacdec: move from scalefactor ranged arrays to flat arrays AAC uses an unconventional system to send scalefactors (the volume+quantization value for each band). Each window is split into either 1 or 8 blocks (long vs short), and transformed separately from one another, with the coefficients for each being also completely independent. The scalefactors slightly increase from 64 (long) to 128 (short) to accomodate better per-block-per-band volume for each window. To reduce overhead, the codec signals scalefactor sizes in an obtuse way, where each group's scalefactor types are sent via a variable length decoding, with a range. But our decoder was written in a way where those ranges were carried through the entire decoder, and to actually read them you had to use the range. Instead of having a dedicated array with a range for each scalefactor, just let the decoder directly index each scalefactor. This also switches the form of quantized scalefactors to the format the spec uses, where for intensity stereo and regular, scalefactors are stored in a scalefactor - 100 form, rather than as-is. USAC gets rid of the complex scalefactor handling. This commit permits for code sharing between both.
2024-05-14 19:07:43 +02:00
sf[idx] = -ff_aac_pow2sf_tab[sfo[idx] + POW_SF2_ZERO];
aacdec: template scalefactor dequantization separately
2024-03-13 22:53:49 +02:00
#endif /* USE_FIXED */
break;
default:
#if USE_FIXED
aacdec: move from scalefactor ranged arrays to flat arrays AAC uses an unconventional system to send scalefactors (the volume+quantization value for each band). Each window is split into either 1 or 8 blocks (long vs short), and transformed separately from one another, with the coefficients for each being also completely independent. The scalefactors slightly increase from 64 (long) to 128 (short) to accomodate better per-block-per-band volume for each window. To reduce overhead, the codec signals scalefactor sizes in an obtuse way, where each group's scalefactor types are sent via a variable length decoding, with a range. But our decoder was written in a way where those ranges were carried through the entire decoder, and to actually read them you had to use the range. Instead of having a dedicated array with a range for each scalefactor, just let the decoder directly index each scalefactor. This also switches the form of quantized scalefactors to the format the spec uses, where for intensity stereo and regular, scalefactors are stored in a scalefactor - 100 form, rather than as-is. USAC gets rid of the complex scalefactor handling. This commit permits for code sharing between both.
2024-05-14 19:07:43 +02:00
sf[idx] = -sfo[idx] - 100;
aacdec: template scalefactor dequantization separately
2024-03-13 22:53:49 +02:00
#else
aacdec: move from scalefactor ranged arrays to flat arrays AAC uses an unconventional system to send scalefactors (the volume+quantization value for each band). Each window is split into either 1 or 8 blocks (long vs short), and transformed separately from one another, with the coefficients for each being also completely independent. The scalefactors slightly increase from 64 (long) to 128 (short) to accomodate better per-block-per-band volume for each window. To reduce overhead, the codec signals scalefactor sizes in an obtuse way, where each group's scalefactor types are sent via a variable length decoding, with a range. But our decoder was written in a way where those ranges were carried through the entire decoder, and to actually read them you had to use the range. Instead of having a dedicated array with a range for each scalefactor, just let the decoder directly index each scalefactor. This also switches the form of quantized scalefactors to the format the spec uses, where for intensity stereo and regular, scalefactors are stored in a scalefactor - 100 form, rather than as-is. USAC gets rid of the complex scalefactor handling. This commit permits for code sharing between both.
2024-05-14 19:07:43 +02:00
sf[idx] = -ff_aac_pow2sf_tab[sfo[idx] + POW_SF2_ZERO];
aacdec: template scalefactor dequantization separately
2024-03-13 22:53:49 +02:00
#endif /* USE_FIXED */
break;
}
}
}
}
aacdec: template mid/side stereo application separately
2024-03-13 22:59:35 +02:00
/**
* Mid/Side stereo decoding; reference: 4.6.8.1.3.
*/
static void AAC_RENAME(apply_mid_side_stereo)(AACDecContext *ac, ChannelElement *cpe)
{
const IndividualChannelStream *ics = &cpe->ch[0].ics;
INTFLOAT *ch0 = cpe->ch[0].AAC_RENAME(coeffs);
INTFLOAT *ch1 = cpe->ch[1].AAC_RENAME(coeffs);
const uint16_t *offsets = ics->swb_offset;
aacdec: move from scalefactor ranged arrays to flat arrays AAC uses an unconventional system to send scalefactors (the volume+quantization value for each band). Each window is split into either 1 or 8 blocks (long vs short), and transformed separately from one another, with the coefficients for each being also completely independent. The scalefactors slightly increase from 64 (long) to 128 (short) to accomodate better per-block-per-band volume for each window. To reduce overhead, the codec signals scalefactor sizes in an obtuse way, where each group's scalefactor types are sent via a variable length decoding, with a range. But our decoder was written in a way where those ranges were carried through the entire decoder, and to actually read them you had to use the range. Instead of having a dedicated array with a range for each scalefactor, just let the decoder directly index each scalefactor. This also switches the form of quantized scalefactors to the format the spec uses, where for intensity stereo and regular, scalefactors are stored in a scalefactor - 100 form, rather than as-is. USAC gets rid of the complex scalefactor handling. This commit permits for code sharing between both.
2024-05-14 19:07:43 +02:00
for (int g = 0; g < ics->num_window_groups; g++) {
aacdec: add a decoder for AAC USAC (xHE-AAC) This commit adds a decoder for the frequency-domain part of USAC. What works: - Mono - Stereo (no prediction) - Stereo (mid/side coding) - Stereo (complex prediction) What's left: - SBR - Speech coding Known issues: - Desync with certain sequences - Preroll crossover missing (shouldn't matter, bitrate adaptation only)
2024-05-16 11:36:12 +02:00
for (int sfb = 0; sfb < cpe->max_sfb_ste; sfb++) {
const int idx = g*cpe->max_sfb_ste + sfb;
aacdec: template mid/side stereo application separately
2024-03-13 22:59:35 +02:00
if (cpe->ms_mask[idx] &&
cpe->ch[0].band_type[idx] < NOISE_BT &&
cpe->ch[1].band_type[idx] < NOISE_BT) {
aacdec: move from scalefactor ranged arrays to flat arrays AAC uses an unconventional system to send scalefactors (the volume+quantization value for each band). Each window is split into either 1 or 8 blocks (long vs short), and transformed separately from one another, with the coefficients for each being also completely independent. The scalefactors slightly increase from 64 (long) to 128 (short) to accomodate better per-block-per-band volume for each window. To reduce overhead, the codec signals scalefactor sizes in an obtuse way, where each group's scalefactor types are sent via a variable length decoding, with a range. But our decoder was written in a way where those ranges were carried through the entire decoder, and to actually read them you had to use the range. Instead of having a dedicated array with a range for each scalefactor, just let the decoder directly index each scalefactor. This also switches the form of quantized scalefactors to the format the spec uses, where for intensity stereo and regular, scalefactors are stored in a scalefactor - 100 form, rather than as-is. USAC gets rid of the complex scalefactor handling. This commit permits for code sharing between both.
2024-05-14 19:07:43 +02:00
for (int group = 0; group < ics->group_len[g]; group++)
aacdec: template mid/side stereo application separately
2024-03-13 22:59:35 +02:00
#if USE_FIXED
aacdec: move from scalefactor ranged arrays to flat arrays AAC uses an unconventional system to send scalefactors (the volume+quantization value for each band). Each window is split into either 1 or 8 blocks (long vs short), and transformed separately from one another, with the coefficients for each being also completely independent. The scalefactors slightly increase from 64 (long) to 128 (short) to accomodate better per-block-per-band volume for each window. To reduce overhead, the codec signals scalefactor sizes in an obtuse way, where each group's scalefactor types are sent via a variable length decoding, with a range. But our decoder was written in a way where those ranges were carried through the entire decoder, and to actually read them you had to use the range. Instead of having a dedicated array with a range for each scalefactor, just let the decoder directly index each scalefactor. This also switches the form of quantized scalefactors to the format the spec uses, where for intensity stereo and regular, scalefactors are stored in a scalefactor - 100 form, rather than as-is. USAC gets rid of the complex scalefactor handling. This commit permits for code sharing between both.
2024-05-14 19:07:43 +02:00
ac->fdsp->butterflies_fixed(ch0 + group * 128 + offsets[sfb],
ch1 + group * 128 + offsets[sfb],
offsets[sfb+1] - offsets[sfb]);
aacdec: template mid/side stereo application separately
2024-03-13 22:59:35 +02:00
#else
aacdec: move from scalefactor ranged arrays to flat arrays AAC uses an unconventional system to send scalefactors (the volume+quantization value for each band). Each window is split into either 1 or 8 blocks (long vs short), and transformed separately from one another, with the coefficients for each being also completely independent. The scalefactors slightly increase from 64 (long) to 128 (short) to accomodate better per-block-per-band volume for each window. To reduce overhead, the codec signals scalefactor sizes in an obtuse way, where each group's scalefactor types are sent via a variable length decoding, with a range. But our decoder was written in a way where those ranges were carried through the entire decoder, and to actually read them you had to use the range. Instead of having a dedicated array with a range for each scalefactor, just let the decoder directly index each scalefactor. This also switches the form of quantized scalefactors to the format the spec uses, where for intensity stereo and regular, scalefactors are stored in a scalefactor - 100 form, rather than as-is. USAC gets rid of the complex scalefactor handling. This commit permits for code sharing between both.
2024-05-14 19:07:43 +02:00
ac->fdsp->butterflies_float(ch0 + group * 128 + offsets[sfb],
ch1 + group * 128 + offsets[sfb],
offsets[sfb+1] - offsets[sfb]);
aacdec: template mid/side stereo application separately
2024-03-13 22:59:35 +02:00
#endif /* USE_FIXED */
}
}
ch0 += ics->group_len[g] * 128;
ch1 += ics->group_len[g] * 128;
}
}
aacdec: template intensity stereo application separately
2024-03-13 23:01:44 +02:00
/**
* intensity stereo decoding; reference: 4.6.8.2.3
*
* @param ms_present Indicates mid/side stereo presence. [0] mask is all 0s;
* [1] mask is decoded from bitstream; [2] mask is all 1s;
* [3] reserved for scalable AAC
*/
static void AAC_RENAME(apply_intensity_stereo)(AACDecContext *ac,
ChannelElement *cpe, int ms_present)
{
const IndividualChannelStream *ics = &cpe->ch[1].ics;
SingleChannelElement *sce1 = &cpe->ch[1];
INTFLOAT *coef0 = cpe->ch[0].AAC_RENAME(coeffs), *coef1 = cpe->ch[1].AAC_RENAME(coeffs);
const uint16_t *offsets = ics->swb_offset;
int c;
INTFLOAT scale;
aacdec: move from scalefactor ranged arrays to flat arrays AAC uses an unconventional system to send scalefactors (the volume+quantization value for each band). Each window is split into either 1 or 8 blocks (long vs short), and transformed separately from one another, with the coefficients for each being also completely independent. The scalefactors slightly increase from 64 (long) to 128 (short) to accomodate better per-block-per-band volume for each window. To reduce overhead, the codec signals scalefactor sizes in an obtuse way, where each group's scalefactor types are sent via a variable length decoding, with a range. But our decoder was written in a way where those ranges were carried through the entire decoder, and to actually read them you had to use the range. Instead of having a dedicated array with a range for each scalefactor, just let the decoder directly index each scalefactor. This also switches the form of quantized scalefactors to the format the spec uses, where for intensity stereo and regular, scalefactors are stored in a scalefactor - 100 form, rather than as-is. USAC gets rid of the complex scalefactor handling. This commit permits for code sharing between both.
2024-05-14 19:07:43 +02:00
for (int g = 0; g < ics->num_window_groups; g++) {
for (int sfb = 0; sfb < ics->max_sfb; sfb++) {
const int idx = g*ics->max_sfb + sfb;
aacdec: template intensity stereo application separately
2024-03-13 23:01:44 +02:00
if (sce1->band_type[idx] == INTENSITY_BT ||
sce1->band_type[idx] == INTENSITY_BT2) {
aacdec: move from scalefactor ranged arrays to flat arrays AAC uses an unconventional system to send scalefactors (the volume+quantization value for each band). Each window is split into either 1 or 8 blocks (long vs short), and transformed separately from one another, with the coefficients for each being also completely independent. The scalefactors slightly increase from 64 (long) to 128 (short) to accomodate better per-block-per-band volume for each window. To reduce overhead, the codec signals scalefactor sizes in an obtuse way, where each group's scalefactor types are sent via a variable length decoding, with a range. But our decoder was written in a way where those ranges were carried through the entire decoder, and to actually read them you had to use the range. Instead of having a dedicated array with a range for each scalefactor, just let the decoder directly index each scalefactor. This also switches the form of quantized scalefactors to the format the spec uses, where for intensity stereo and regular, scalefactors are stored in a scalefactor - 100 form, rather than as-is. USAC gets rid of the complex scalefactor handling. This commit permits for code sharing between both.
2024-05-14 19:07:43 +02:00
c = -1 + 2 * (sce1->band_type[idx] - 14);
if (ms_present)
c *= 1 - 2 * cpe->ms_mask[idx];
scale = c * sce1->AAC_RENAME(sf)[idx];
for (int group = 0; group < ics->group_len[g]; group++)
aacdec: template intensity stereo application separately
2024-03-13 23:01:44 +02:00
#if USE_FIXED
aacdec: move from scalefactor ranged arrays to flat arrays AAC uses an unconventional system to send scalefactors (the volume+quantization value for each band). Each window is split into either 1 or 8 blocks (long vs short), and transformed separately from one another, with the coefficients for each being also completely independent. The scalefactors slightly increase from 64 (long) to 128 (short) to accomodate better per-block-per-band volume for each window. To reduce overhead, the codec signals scalefactor sizes in an obtuse way, where each group's scalefactor types are sent via a variable length decoding, with a range. But our decoder was written in a way where those ranges were carried through the entire decoder, and to actually read them you had to use the range. Instead of having a dedicated array with a range for each scalefactor, just let the decoder directly index each scalefactor. This also switches the form of quantized scalefactors to the format the spec uses, where for intensity stereo and regular, scalefactors are stored in a scalefactor - 100 form, rather than as-is. USAC gets rid of the complex scalefactor handling. This commit permits for code sharing between both.
2024-05-14 19:07:43 +02:00
subband_scale(coef1 + group * 128 + offsets[sfb],
coef0 + group * 128 + offsets[sfb],
scale,
23,
offsets[sfb + 1] - offsets[sfb], ac->avctx);
aacdec: template intensity stereo application separately
2024-03-13 23:01:44 +02:00
#else
aacdec: move from scalefactor ranged arrays to flat arrays AAC uses an unconventional system to send scalefactors (the volume+quantization value for each band). Each window is split into either 1 or 8 blocks (long vs short), and transformed separately from one another, with the coefficients for each being also completely independent. The scalefactors slightly increase from 64 (long) to 128 (short) to accomodate better per-block-per-band volume for each window. To reduce overhead, the codec signals scalefactor sizes in an obtuse way, where each group's scalefactor types are sent via a variable length decoding, with a range. But our decoder was written in a way where those ranges were carried through the entire decoder, and to actually read them you had to use the range. Instead of having a dedicated array with a range for each scalefactor, just let the decoder directly index each scalefactor. This also switches the form of quantized scalefactors to the format the spec uses, where for intensity stereo and regular, scalefactors are stored in a scalefactor - 100 form, rather than as-is. USAC gets rid of the complex scalefactor handling. This commit permits for code sharing between both.
2024-05-14 19:07:43 +02:00
ac->fdsp->vector_fmul_scalar(coef1 + group * 128 + offsets[sfb],
coef0 + group * 128 + offsets[sfb],
scale,
offsets[sfb + 1] - offsets[sfb]);
aacdec: template intensity stereo application separately
2024-03-13 23:01:44 +02:00
#endif /* USE_FIXED */
}
}
coef0 += ics->group_len[g] * 128;
coef1 += ics->group_len[g] * 128;
}
}
aacdec: template TNS application separately
2024-03-13 23:20:59 +02:00
/**
* Decode Temporal Noise Shaping filter coefficients and apply all-pole filters; reference: 4.6.9.3.
*
* @param decode 1 if tool is used normally, 0 if tool is used in LTP.
* @param coef spectral coefficients
*/
static void AAC_RENAME(apply_tns)(void *_coef_param, TemporalNoiseShaping *tns,
IndividualChannelStream *ics, int decode)
{
const int mmm = FFMIN(ics->tns_max_bands, ics->max_sfb);
int w, filt, m, i;
int bottom, top, order, start, end, size, inc;
INTFLOAT *coef_param = _coef_param;
INTFLOAT lpc[TNS_MAX_ORDER];
INTFLOAT tmp[TNS_MAX_ORDER+1];
UINTFLOAT *coef = coef_param;
if(!mmm)
return;
for (w = 0; w < ics->num_windows; w++) {
bottom = ics->num_swb;
for (filt = 0; filt < tns->n_filt[w]; filt++) {
top = bottom;
bottom = FFMAX(0, top - tns->length[w][filt]);
order = tns->order[w][filt];
if (order == 0)
continue;
// tns_decode_coef
compute_lpc_coefs(tns->AAC_RENAME(coef)[w][filt], order, lpc, 0, 0, 0);
start = ics->swb_offset[FFMIN(bottom, mmm)];
end = ics->swb_offset[FFMIN( top, mmm)];
if ((size = end - start) <= 0)
continue;
if (tns->direction[w][filt]) {
inc = -1;
start = end - 1;
} else {
inc = 1;
}
start += w * 128;
if (decode) {
// ar filter
for (m = 0; m < size; m++, start += inc)
for (i = 1; i <= FFMIN(m, order); i++)
coef[start] -= AAC_MUL26((INTFLOAT)coef[start - i * inc], lpc[i - 1]);
} else {
// ma filter
for (m = 0; m < size; m++, start += inc) {
tmp[0] = coef[start];
for (i = 1; i <= FFMIN(m, order); i++)
coef[start] += AAC_MUL26(tmp[i], lpc[i - 1]);
for (i = order; i > 0; i--)
tmp[i] = tmp[i - 1];
}
}
}
}
}
aacdec: template LTP windowing separately The function is called only internally in DSP, so we do not need to expose it. apply_ltp on MIPS uses this function, but due to the function being just a glue function with no real optimizations, duplicate it there.
2024-03-16 02:28:41 +02:00
/**
* Apply windowing and MDCT to obtain the spectral
* coefficient from the predicted sample by LTP.
*/
static inline void AAC_RENAME(windowing_and_mdct_ltp)(AACDecContext *ac,
INTFLOAT *out, INTFLOAT *in,
IndividualChannelStream *ics)
{
const INTFLOAT *lwindow = ics->use_kb_window[0] ? AAC_RENAME2(aac_kbd_long_1024) : AAC_RENAME2(sine_1024);
const INTFLOAT *swindow = ics->use_kb_window[0] ? AAC_RENAME2(aac_kbd_short_128) : AAC_RENAME2(sine_128);
const INTFLOAT *lwindow_prev = ics->use_kb_window[1] ? AAC_RENAME2(aac_kbd_long_1024) : AAC_RENAME2(sine_1024);
const INTFLOAT *swindow_prev = ics->use_kb_window[1] ? AAC_RENAME2(aac_kbd_short_128) : AAC_RENAME2(sine_128);
if (ics->window_sequence[0] != LONG_STOP_SEQUENCE) {
ac->fdsp->vector_fmul(in, in, lwindow_prev, 1024);
} else {
memset(in, 0, 448 * sizeof(*in));
ac->fdsp->vector_fmul(in + 448, in + 448, swindow_prev, 128);
}
if (ics->window_sequence[0] != LONG_START_SEQUENCE) {
ac->fdsp->vector_fmul_reverse(in + 1024, in + 1024, lwindow, 1024);
} else {
ac->fdsp->vector_fmul_reverse(in + 1024 + 448, in + 1024 + 448, swindow, 128);
memset(in + 1024 + 576, 0, 448 * sizeof(*in));
}
ac->mdct_ltp_fn(ac->mdct_ltp, out, in, sizeof(INTFLOAT));
}
aacdec: template LTP application separately
2024-03-14 05:52:28 +02:00
/**
* Apply the long term prediction
*/
static void AAC_RENAME(apply_ltp)(AACDecContext *ac, SingleChannelElement *sce)
{
const LongTermPrediction *ltp = &sce->ics.ltp;
const uint16_t *offsets = sce->ics.swb_offset;
int i, sfb;
if (sce->ics.window_sequence[0] != EIGHT_SHORT_SEQUENCE) {
INTFLOAT *predTime = sce->AAC_RENAME(output);
INTFLOAT *predFreq = ac->AAC_RENAME(buf_mdct);
int16_t num_samples = 2048;
if (ltp->lag < 1024)
num_samples = ltp->lag + 1024;
for (i = 0; i < num_samples; i++)
predTime[i] = AAC_MUL30(sce->AAC_RENAME(ltp_state)[i + 2048 - ltp->lag], ltp->AAC_RENAME(coef));
memset(&predTime[i], 0, (2048 - i) * sizeof(*predTime));
aacdec: template LTP windowing separately The function is called only internally in DSP, so we do not need to expose it. apply_ltp on MIPS uses this function, but due to the function being just a glue function with no real optimizations, duplicate it there.
2024-03-16 02:28:41 +02:00
AAC_RENAME(windowing_and_mdct_ltp)(ac, predFreq, predTime, &sce->ics);
aacdec: template LTP application separately
2024-03-14 05:52:28 +02:00
if (sce->tns.present)
AAC_RENAME(apply_tns)(predFreq, &sce->tns, &sce->ics, 0);
for (sfb = 0; sfb < FFMIN(sce->ics.max_sfb, MAX_LTP_LONG_SFB); sfb++)
if (ltp->used[sfb])
for (i = offsets[sfb]; i < offsets[sfb + 1]; i++)
sce->AAC_RENAME(coeffs)[i] += (UINTFLOAT)predFreq[i];
}
}
/**
* Update the LTP buffer for next frame
*/
static void AAC_RENAME(update_ltp)(AACDecContext *ac, SingleChannelElement *sce)
{
IndividualChannelStream *ics = &sce->ics;
INTFLOAT *saved = sce->AAC_RENAME(saved);
INTFLOAT *saved_ltp = sce->AAC_RENAME(coeffs);
const INTFLOAT *lwindow = ics->use_kb_window[0] ? AAC_RENAME2(aac_kbd_long_1024) : AAC_RENAME2(sine_1024);
const INTFLOAT *swindow = ics->use_kb_window[0] ? AAC_RENAME2(aac_kbd_short_128) : AAC_RENAME2(sine_128);
int i;
if (ics->window_sequence[0] == EIGHT_SHORT_SEQUENCE) {
memcpy(saved_ltp, saved, 512 * sizeof(*saved_ltp));
memset(saved_ltp + 576, 0, 448 * sizeof(*saved_ltp));
ac->fdsp->vector_fmul_reverse(saved_ltp + 448, ac->AAC_RENAME(buf_mdct) + 960, &swindow[64], 64);
for (i = 0; i < 64; i++)
saved_ltp[i + 512] = AAC_MUL31(ac->AAC_RENAME(buf_mdct)[1023 - i], swindow[63 - i]);
} else if (1 && ics->window_sequence[0] == LONG_START_SEQUENCE) {
memcpy(saved_ltp, ac->AAC_RENAME(buf_mdct) + 512, 448 * sizeof(*saved_ltp));
memset(saved_ltp + 576, 0, 448 * sizeof(*saved_ltp));
ac->fdsp->vector_fmul_reverse(saved_ltp + 448, ac->AAC_RENAME(buf_mdct) + 960, &swindow[64], 64);
for (i = 0; i < 64; i++)
saved_ltp[i + 512] = AAC_MUL31(ac->AAC_RENAME(buf_mdct)[1023 - i], swindow[63 - i]);
} else if (1) { // LONG_STOP or ONLY_LONG
ac->fdsp->vector_fmul_reverse(saved_ltp, ac->AAC_RENAME(buf_mdct) + 512, &lwindow[512], 512);
for (i = 0; i < 512; i++)
saved_ltp[i + 512] = AAC_MUL31(ac->AAC_RENAME(buf_mdct)[1023 - i], lwindow[511 - i]);
}
memcpy(sce->AAC_RENAME(ltp_state), sce->AAC_RENAME(ltp_state)+1024,
1024 * sizeof(*sce->AAC_RENAME(ltp_state)));
memcpy(sce->AAC_RENAME(ltp_state) + 1024, sce->AAC_RENAME(output),
1024 * sizeof(*sce->AAC_RENAME(ltp_state)));
memcpy(sce->AAC_RENAME(ltp_state) + 2048, saved_ltp,
1024 * sizeof(*sce->AAC_RENAME(ltp_state)));
}
aacdec: template windowing and transforms separately
2024-03-16 01:14:32 +02:00
/**
* Conduct IMDCT and windowing.
*/
static void AAC_RENAME(imdct_and_windowing)(AACDecContext *ac, SingleChannelElement *sce)
{
IndividualChannelStream *ics = &sce->ics;
INTFLOAT *in = sce->AAC_RENAME(coeffs);
INTFLOAT *out = sce->AAC_RENAME(output);
INTFLOAT *saved = sce->AAC_RENAME(saved);
const INTFLOAT *swindow = ics->use_kb_window[0] ? AAC_RENAME2(aac_kbd_short_128) : AAC_RENAME2(sine_128);
const INTFLOAT *lwindow_prev = ics->use_kb_window[1] ? AAC_RENAME2(aac_kbd_long_1024) : AAC_RENAME2(sine_1024);
const INTFLOAT *swindow_prev = ics->use_kb_window[1] ? AAC_RENAME2(aac_kbd_short_128) : AAC_RENAME2(sine_128);
INTFLOAT *buf = ac->AAC_RENAME(buf_mdct);
INTFLOAT *temp = ac->AAC_RENAME(temp);
int i;
// imdct
if (ics->window_sequence[0] == EIGHT_SHORT_SEQUENCE) {
for (i = 0; i < 1024; i += 128)
ac->mdct128_fn(ac->mdct128, buf + i, in + i, sizeof(INTFLOAT));
} else {
ac->mdct1024_fn(ac->mdct1024, buf, in, sizeof(INTFLOAT));
}
/* window overlapping
* NOTE: To simplify the overlapping code, all 'meaningless' short to long
* and long to short transitions are considered to be short to short
* transitions. This leaves just two cases (long to long and short to short)
* with a little special sauce for EIGHT_SHORT_SEQUENCE.
*/
if ((ics->window_sequence[1] == ONLY_LONG_SEQUENCE || ics->window_sequence[1] == LONG_STOP_SEQUENCE) &&
(ics->window_sequence[0] == ONLY_LONG_SEQUENCE || ics->window_sequence[0] == LONG_START_SEQUENCE)) {
ac->fdsp->vector_fmul_window( out, saved, buf, lwindow_prev, 512);
} else {
memcpy( out, saved, 448 * sizeof(*out));
if (ics->window_sequence[0] == EIGHT_SHORT_SEQUENCE) {
ac->fdsp->vector_fmul_window(out + 448 + 0*128, saved + 448, buf + 0*128, swindow_prev, 64);
ac->fdsp->vector_fmul_window(out + 448 + 1*128, buf + 0*128 + 64, buf + 1*128, swindow, 64);
ac->fdsp->vector_fmul_window(out + 448 + 2*128, buf + 1*128 + 64, buf + 2*128, swindow, 64);
ac->fdsp->vector_fmul_window(out + 448 + 3*128, buf + 2*128 + 64, buf + 3*128, swindow, 64);
ac->fdsp->vector_fmul_window(temp, buf + 3*128 + 64, buf + 4*128, swindow, 64);
memcpy( out + 448 + 4*128, temp, 64 * sizeof(*out));
} else {
ac->fdsp->vector_fmul_window(out + 448, saved + 448, buf, swindow_prev, 64);
memcpy( out + 576, buf + 64, 448 * sizeof(*out));
}
}
// buffer update
if (ics->window_sequence[0] == EIGHT_SHORT_SEQUENCE) {
memcpy( saved, temp + 64, 64 * sizeof(*saved));
ac->fdsp->vector_fmul_window(saved + 64, buf + 4*128 + 64, buf + 5*128, swindow, 64);
ac->fdsp->vector_fmul_window(saved + 192, buf + 5*128 + 64, buf + 6*128, swindow, 64);
ac->fdsp->vector_fmul_window(saved + 320, buf + 6*128 + 64, buf + 7*128, swindow, 64);
memcpy( saved + 448, buf + 7*128 + 64, 64 * sizeof(*saved));
} else if (ics->window_sequence[0] == LONG_START_SEQUENCE) {
memcpy( saved, buf + 512, 448 * sizeof(*saved));
memcpy( saved + 448, buf + 7*128 + 64, 64 * sizeof(*saved));
} else { // LONG_STOP or ONLY_LONG
memcpy( saved, buf + 512, 512 * sizeof(*saved));
}
}
aacdec_dsp: implement 768-point transform and windowing Required for USAC
2024-03-25 07:13:38 +02:00
/**
* Conduct IMDCT and windowing for 768-point frames.
*/
static void AAC_RENAME(imdct_and_windowing_768)(AACDecContext *ac, SingleChannelElement *sce)
{
IndividualChannelStream *ics = &sce->ics;
INTFLOAT *in = sce->AAC_RENAME(coeffs);
INTFLOAT *out = sce->AAC_RENAME(output);
INTFLOAT *saved = sce->AAC_RENAME(saved);
const INTFLOAT *swindow = ics->use_kb_window[0] ? AAC_RENAME(aac_kbd_short_96) : AAC_RENAME(sine_96);
const INTFLOAT *lwindow_prev = ics->use_kb_window[1] ? AAC_RENAME(aac_kbd_long_768) : AAC_RENAME(sine_768);
const INTFLOAT *swindow_prev = ics->use_kb_window[1] ? AAC_RENAME(aac_kbd_short_96) : AAC_RENAME(sine_96);
INTFLOAT *buf = ac->AAC_RENAME(buf_mdct);
INTFLOAT *temp = ac->AAC_RENAME(temp);
int i;
// imdct
if (ics->window_sequence[0] == EIGHT_SHORT_SEQUENCE) {
for (i = 0; i < 8; i++)
ac->mdct96_fn(ac->mdct96, buf + i * 96, in + i * 96, sizeof(INTFLOAT));
} else {
ac->mdct768_fn(ac->mdct768, buf, in, sizeof(INTFLOAT));
}
/* window overlapping
* NOTE: To simplify the overlapping code, all 'meaningless' short to long
* and long to short transitions are considered to be short to short
* transitions. This leaves just two cases (long to long and short to short)
* with a little special sauce for EIGHT_SHORT_SEQUENCE.
*/
if ((ics->window_sequence[1] == ONLY_LONG_SEQUENCE || ics->window_sequence[1] == LONG_STOP_SEQUENCE) &&
(ics->window_sequence[0] == ONLY_LONG_SEQUENCE || ics->window_sequence[0] == LONG_START_SEQUENCE)) {
ac->fdsp->vector_fmul_window( out, saved, buf, lwindow_prev, 384);
} else {
memcpy( out, saved, 336 * sizeof(*out));
if (ics->window_sequence[0] == EIGHT_SHORT_SEQUENCE) {
ac->fdsp->vector_fmul_window(out + 336 + 0*96, saved + 336, buf + 0*96, swindow_prev, 48);
ac->fdsp->vector_fmul_window(out + 336 + 1*96, buf + 0*96 + 48, buf + 1*96, swindow, 48);
ac->fdsp->vector_fmul_window(out + 336 + 2*96, buf + 1*96 + 48, buf + 2*96, swindow, 48);
ac->fdsp->vector_fmul_window(out + 336 + 3*96, buf + 2*96 + 48, buf + 3*96, swindow, 48);
ac->fdsp->vector_fmul_window(temp, buf + 3*96 + 48, buf + 4*96, swindow, 48);
memcpy( out + 336 + 4*96, temp, 48 * sizeof(*out));
} else {
ac->fdsp->vector_fmul_window(out + 336, saved + 336, buf, swindow_prev, 48);
memcpy( out + 432, buf + 48, 336 * sizeof(*out));
}
}
// buffer update
if (ics->window_sequence[0] == EIGHT_SHORT_SEQUENCE) {
memcpy( saved, temp + 48, 48 * sizeof(*saved));
ac->fdsp->vector_fmul_window(saved + 48, buf + 4*96 + 48, buf + 5*96, swindow, 48);
ac->fdsp->vector_fmul_window(saved + 144, buf + 5*96 + 48, buf + 6*96, swindow, 48);
ac->fdsp->vector_fmul_window(saved + 240, buf + 6*96 + 48, buf + 7*96, swindow, 48);
memcpy( saved + 336, buf + 7*96 + 48, 48 * sizeof(*saved));
} else if (ics->window_sequence[0] == LONG_START_SEQUENCE) {
memcpy( saved, buf + 384, 336 * sizeof(*saved));
memcpy( saved + 336, buf + 7*96 + 48, 48 * sizeof(*saved));
} else { // LONG_STOP or ONLY_LONG
memcpy( saved, buf + 384, 384 * sizeof(*saved));
}
}
aacdec: template windowing and transforms separately
2024-03-16 01:14:32 +02:00
/**
* Conduct IMDCT and windowing.
*/
static void AAC_RENAME(imdct_and_windowing_960)(AACDecContext *ac, SingleChannelElement *sce)
{
IndividualChannelStream *ics = &sce->ics;
INTFLOAT *in = sce->AAC_RENAME(coeffs);
INTFLOAT *out = sce->AAC_RENAME(output);
INTFLOAT *saved = sce->AAC_RENAME(saved);
const INTFLOAT *swindow = ics->use_kb_window[0] ? AAC_RENAME(aac_kbd_short_120) : AAC_RENAME(sine_120);
const INTFLOAT *lwindow_prev = ics->use_kb_window[1] ? AAC_RENAME(aac_kbd_long_960) : AAC_RENAME(sine_960);
const INTFLOAT *swindow_prev = ics->use_kb_window[1] ? AAC_RENAME(aac_kbd_short_120) : AAC_RENAME(sine_120);
INTFLOAT *buf = ac->AAC_RENAME(buf_mdct);
INTFLOAT *temp = ac->AAC_RENAME(temp);
int i;
// imdct
if (ics->window_sequence[0] == EIGHT_SHORT_SEQUENCE) {
for (i = 0; i < 8; i++)
ac->mdct120_fn(ac->mdct120, buf + i * 120, in + i * 128, sizeof(INTFLOAT));
} else {
ac->mdct960_fn(ac->mdct960, buf, in, sizeof(INTFLOAT));
}
/* window overlapping
* NOTE: To simplify the overlapping code, all 'meaningless' short to long
* and long to short transitions are considered to be short to short
* transitions. This leaves just two cases (long to long and short to short)
* with a little special sauce for EIGHT_SHORT_SEQUENCE.
*/
if ((ics->window_sequence[1] == ONLY_LONG_SEQUENCE || ics->window_sequence[1] == LONG_STOP_SEQUENCE) &&
(ics->window_sequence[0] == ONLY_LONG_SEQUENCE || ics->window_sequence[0] == LONG_START_SEQUENCE)) {
ac->fdsp->vector_fmul_window( out, saved, buf, lwindow_prev, 480);
} else {
memcpy( out, saved, 420 * sizeof(*out));
if (ics->window_sequence[0] == EIGHT_SHORT_SEQUENCE) {
ac->fdsp->vector_fmul_window(out + 420 + 0*120, saved + 420, buf + 0*120, swindow_prev, 60);
ac->fdsp->vector_fmul_window(out + 420 + 1*120, buf + 0*120 + 60, buf + 1*120, swindow, 60);
ac->fdsp->vector_fmul_window(out + 420 + 2*120, buf + 1*120 + 60, buf + 2*120, swindow, 60);
ac->fdsp->vector_fmul_window(out + 420 + 3*120, buf + 2*120 + 60, buf + 3*120, swindow, 60);
ac->fdsp->vector_fmul_window(temp, buf + 3*120 + 60, buf + 4*120, swindow, 60);
memcpy( out + 420 + 4*120, temp, 60 * sizeof(*out));
} else {
ac->fdsp->vector_fmul_window(out + 420, saved + 420, buf, swindow_prev, 60);
memcpy( out + 540, buf + 60, 420 * sizeof(*out));
}
}
// buffer update
if (ics->window_sequence[0] == EIGHT_SHORT_SEQUENCE) {
memcpy( saved, temp + 60, 60 * sizeof(*saved));
ac->fdsp->vector_fmul_window(saved + 60, buf + 4*120 + 60, buf + 5*120, swindow, 60);
ac->fdsp->vector_fmul_window(saved + 180, buf + 5*120 + 60, buf + 6*120, swindow, 60);
ac->fdsp->vector_fmul_window(saved + 300, buf + 6*120 + 60, buf + 7*120, swindow, 60);
memcpy( saved + 420, buf + 7*120 + 60, 60 * sizeof(*saved));
} else if (ics->window_sequence[0] == LONG_START_SEQUENCE) {
memcpy( saved, buf + 480, 420 * sizeof(*saved));
memcpy( saved + 420, buf + 7*120 + 60, 60 * sizeof(*saved));
} else { // LONG_STOP or ONLY_LONG
memcpy( saved, buf + 480, 480 * sizeof(*saved));
}
}
aacdec_dsp: implement 768-point transform and windowing Required for USAC
2024-03-25 07:13:38 +02:00
aacdec: template windowing and transforms separately
2024-03-16 01:14:32 +02:00
static void AAC_RENAME(imdct_and_windowing_ld)(AACDecContext *ac, SingleChannelElement *sce)
{
IndividualChannelStream *ics = &sce->ics;
INTFLOAT *in = sce->AAC_RENAME(coeffs);
INTFLOAT *out = sce->AAC_RENAME(output);
INTFLOAT *saved = sce->AAC_RENAME(saved);
INTFLOAT *buf = ac->AAC_RENAME(buf_mdct);
// imdct
ac->mdct512_fn(ac->mdct512, buf, in, sizeof(INTFLOAT));
// window overlapping
if (ics->use_kb_window[1]) {
// AAC LD uses a low overlap sine window instead of a KBD window
memcpy(out, saved, 192 * sizeof(*out));
ac->fdsp->vector_fmul_window(out + 192, saved + 192, buf, AAC_RENAME2(sine_128), 64);
memcpy( out + 320, buf + 64, 192 * sizeof(*out));
} else {
ac->fdsp->vector_fmul_window(out, saved, buf, AAC_RENAME2(sine_512), 256);
}
// buffer update
memcpy(saved, buf + 256, 256 * sizeof(*saved));
}
static void AAC_RENAME(imdct_and_windowing_eld)(AACDecContext *ac, SingleChannelElement *sce)
{
UINTFLOAT *in = sce->AAC_RENAME(coeffs);
INTFLOAT *out = sce->AAC_RENAME(output);
INTFLOAT *saved = sce->AAC_RENAME(saved);
INTFLOAT *buf = ac->AAC_RENAME(buf_mdct);
int i;
const int n = ac->oc[1].m4ac.frame_length_short ? 480 : 512;
const int n2 = n >> 1;
const int n4 = n >> 2;
const INTFLOAT *const window = n == 480 ? AAC_RENAME(ff_aac_eld_window_480) :
AAC_RENAME(ff_aac_eld_window_512);
// Inverse transform, mapped to the conventional IMDCT by
// Chivukula, R.K.; Reznik, Y.A.; Devarajan, V.,
// "Efficient algorithms for MPEG-4 AAC-ELD, AAC-LD and AAC-LC filterbanks,"
// International Conference on Audio, Language and Image Processing, ICALIP 2008.
// URL: http://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=4590245&isnumber=4589950
for (i = 0; i < n2; i+=2) {
INTFLOAT temp;
temp = in[i ]; in[i ] = -in[n - 1 - i]; in[n - 1 - i] = temp;
temp = -in[i + 1]; in[i + 1] = in[n - 2 - i]; in[n - 2 - i] = temp;
}
if (n == 480)
ac->mdct480_fn(ac->mdct480, buf, in, sizeof(INTFLOAT));
else
ac->mdct512_fn(ac->mdct512, buf, in, sizeof(INTFLOAT));
for (i = 0; i < n; i+=2) {
buf[i + 0] = -(UINTFLOAT)(USE_FIXED + 1)*buf[i + 0];
buf[i + 1] = (UINTFLOAT)(USE_FIXED + 1)*buf[i + 1];
}
// Like with the regular IMDCT at this point we still have the middle half
// of a transform but with even symmetry on the left and odd symmetry on
// the right
// window overlapping
// The spec says to use samples [0..511] but the reference decoder uses
// samples [128..639].
for (i = n4; i < n2; i ++) {
out[i - n4] = AAC_MUL31( buf[ n2 - 1 - i] , window[i - n4]) +
AAC_MUL31( saved[ i + n2] , window[i + n - n4]) +
AAC_MUL31(-saved[n + n2 - 1 - i] , window[i + 2*n - n4]) +
AAC_MUL31(-saved[ 2*n + n2 + i] , window[i + 3*n - n4]);
}
for (i = 0; i < n2; i ++) {
out[n4 + i] = AAC_MUL31( buf[ i] , window[i + n2 - n4]) +
AAC_MUL31(-saved[ n - 1 - i] , window[i + n2 + n - n4]) +
AAC_MUL31(-saved[ n + i] , window[i + n2 + 2*n - n4]) +
AAC_MUL31( saved[2*n + n - 1 - i] , window[i + n2 + 3*n - n4]);
}
for (i = 0; i < n4; i ++) {
out[n2 + n4 + i] = AAC_MUL31( buf[ i + n2] , window[i + n - n4]) +
AAC_MUL31(-saved[n2 - 1 - i] , window[i + 2*n - n4]) +
AAC_MUL31(-saved[n + n2 + i] , window[i + 3*n - n4]);
}
// buffer update
memmove(saved + n, saved, 2 * n * sizeof(*saved));
memcpy( saved, buf, n * sizeof(*saved));
}
aacdec: move fixed-point clipping to a separate function
2024-03-16 05:29:07 +02:00
static void AAC_RENAME(clip_output)(AACDecContext *ac, ChannelElement *che,
int type, int samples)
{
#if USE_FIXED
/* preparation for resampler */
for (int j = 0; j < samples; j++){
che->ch[0].output_fixed[j] = (int32_t)av_clip64((int64_t)che->ch[0].output_fixed[j]*128,
INT32_MIN, INT32_MAX-0x8000)+0x8000;
if (type == TYPE_CPE || (type == TYPE_SCE && ac->oc[1].m4ac.ps == 1))
che->ch[1].output_fixed[j] = (int32_t)av_clip64((int64_t)che->ch[1].output_fixed[j]*128,
INT32_MIN, INT32_MAX-0x8000)+0x8000;
}
#endif
}
aacdec: move prediction to separate files
2024-03-16 06:16:50 +02:00
static inline void reset_all_predictors(PredictorState *ps)
{
int i;
for (i = 0; i < MAX_PREDICTORS; i++)
reset_predict_state(&ps[i]);
}
static inline void reset_predictor_group(PredictorState *ps, int group_num)
{
int i;
for (i = group_num - 1; i < MAX_PREDICTORS; i += 30)
reset_predict_state(&ps[i]);
}
/**
* Apply AAC-Main style frequency domain prediction.
*/
static void AAC_RENAME(apply_prediction)(AACDecContext *ac, SingleChannelElement *sce)
{
int sfb, k;
if (!sce->ics.predictor_initialized) {
reset_all_predictors(sce->AAC_RENAME(predictor_state));
sce->ics.predictor_initialized = 1;
}
if (sce->ics.window_sequence[0] != EIGHT_SHORT_SEQUENCE) {
for (sfb = 0;
sfb < ff_aac_pred_sfb_max[ac->oc[1].m4ac.sampling_index];
sfb++) {
for (k = sce->ics.swb_offset[sfb];
k < sce->ics.swb_offset[sfb + 1];
k++) {
predict(&sce->AAC_RENAME(predictor_state)[k],
&sce->AAC_RENAME(coeffs)[k],
sce->ics.predictor_present &&
sce->ics.prediction_used[sfb]);
}
}
if (sce->ics.predictor_reset_group)
reset_predictor_group(sce->AAC_RENAME(predictor_state),
sce->ics.predictor_reset_group);
} else
reset_all_predictors(sce->AAC_RENAME(predictor_state));
}
avcodec/aac/aacdec_(fixed|float): Set AACDecDSP, AACDecProc directly This is more in line with how we initialize DSP functions and avoids tables of function pointers as well as relocations for these. Reviewed-by: Lynne <dev@lynne.ee> Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2024-05-06 13:59:08 +02:00
static av_cold void AAC_RENAME(aac_dsp_init)(AACDecDSP *aac_dsp)
{
#define SET(member) aac_dsp->member = AAC_RENAME(member)
SET(dequant_scalefactors);
SET(apply_mid_side_stereo);
SET(apply_intensity_stereo);
SET(apply_tns);
SET(apply_ltp);
SET(update_ltp);
SET(apply_prediction);
SET(imdct_and_windowing);
aacdec_dsp: implement 768-point transform and windowing Required for USAC
2024-03-25 07:13:38 +02:00
SET(imdct_and_windowing_768);
avcodec/aac/aacdec_(fixed|float): Set AACDecDSP, AACDecProc directly This is more in line with how we initialize DSP functions and avoids tables of function pointers as well as relocations for these. Reviewed-by: Lynne <dev@lynne.ee> Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2024-05-06 13:59:08 +02:00
SET(imdct_and_windowing_960);
SET(imdct_and_windowing_ld);
SET(imdct_and_windowing_eld);
SET(apply_dependent_coupling);
SET(apply_independent_coupling);
SET(clip_output);
#undef SET
}
Reference in New Issue Copy Permalink