From: Lynne via ffmpeg-devel <ffmpeg-devel@ffmpeg.org>
To: ffmpeg-devel@ffmpeg.org
Cc: Lynne <dev@lynne.ee>
Subject: [FFmpeg-devel] [PATCH v2 1/8] aacdec: move from scalefactor ranged arrays to flat arrays
Date: Sun, 19 May 2024 18:54:37 +0200
Message-ID: <20240519165444.829271-2-dev@lynne.ee> (raw)
In-Reply-To: <20240519165444.829271-1-dev@lynne.ee>
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.
---
libavcodec/aac/aacdec.c | 100 ++++++++++++---------------
libavcodec/aac/aacdec.h | 5 +-
libavcodec/aac/aacdec_dsp_template.c | 95 ++++++++++---------------
3 files changed, 83 insertions(+), 117 deletions(-)
diff --git a/libavcodec/aac/aacdec.c b/libavcodec/aac/aacdec.c
index 7457fe6c97..35722f9b9b 100644
--- a/libavcodec/aac/aacdec.c
+++ b/libavcodec/aac/aacdec.c
@@ -1412,13 +1412,13 @@ fail:
*
* @return Returns error status. 0 - OK, !0 - error
*/
-static int decode_band_types(AACDecContext *ac, enum BandType band_type[120],
- int band_type_run_end[120], GetBitContext *gb,
- IndividualChannelStream *ics)
+static int decode_band_types(AACDecContext *ac, SingleChannelElement *sce,
+ GetBitContext *gb)
{
- int g, idx = 0;
+ IndividualChannelStream *ics = &sce->ics;
const int bits = (ics->window_sequence[0] == EIGHT_SHORT_SEQUENCE) ? 3 : 5;
- for (g = 0; g < ics->num_window_groups; g++) {
+
+ for (int g = 0; g < ics->num_window_groups; g++) {
int k = 0;
while (k < ics->max_sfb) {
uint8_t sect_end = k;
@@ -1442,10 +1442,8 @@ static int decode_band_types(AACDecContext *ac, enum BandType band_type[120],
return AVERROR_INVALIDDATA;
}
} while (sect_len_incr == (1 << bits) - 1);
- for (; k < sect_end; k++) {
- band_type [idx] = sect_band_type;
- band_type_run_end[idx++] = sect_end;
- }
+ for (; k < sect_end; k++)
+ sce->band_type[g*ics->max_sfb + k] = sect_band_type;
}
}
return 0;
@@ -1461,69 +1459,59 @@ static int decode_band_types(AACDecContext *ac, enum BandType band_type[120],
*
* @return Returns error status. 0 - OK, !0 - error
*/
-static int decode_scalefactors(AACDecContext *ac, int sfo[120],
- GetBitContext *gb,
- unsigned int global_gain,
- IndividualChannelStream *ics,
- enum BandType band_type[120],
- int band_type_run_end[120])
+static int decode_scalefactors(AACDecContext *ac, SingleChannelElement *sce,
+ GetBitContext *gb, unsigned int global_gain)
{
- int g, i, idx = 0;
+ IndividualChannelStream *ics = &sce->ics;
int offset[3] = { global_gain, global_gain - NOISE_OFFSET, 0 };
int clipped_offset;
int noise_flag = 1;
- for (g = 0; g < ics->num_window_groups; g++) {
- for (i = 0; i < ics->max_sfb;) {
- int run_end = band_type_run_end[idx];
- switch (band_type[idx]) {
+
+ for (int g = 0; g < ics->num_window_groups; g++) {
+ for (int sfb = 0; sfb < ics->max_sfb; sfb++) {
+ switch (sce->band_type[g*ics->max_sfb + sfb]) {
case ZERO_BT:
- for (; i < run_end; i++, idx++)
- sfo[idx] = 0;
+ sce->sfo[g*ics->max_sfb + sfb] = 0;
break;
case INTENSITY_BT: /* fallthrough */
case INTENSITY_BT2:
- for (; i < run_end; i++, idx++) {
- offset[2] += get_vlc2(gb, ff_vlc_scalefactors, 7, 3) - SCALE_DIFF_ZERO;
- clipped_offset = av_clip(offset[2], -155, 100);
- if (offset[2] != clipped_offset) {
- avpriv_request_sample(ac->avctx,
- "If you heard an audible artifact, there may be a bug in the decoder. "
- "Clipped intensity stereo position (%d -> %d)",
- offset[2], clipped_offset);
- }
- sfo[idx] = clipped_offset;
+ offset[2] += get_vlc2(gb, ff_vlc_scalefactors, 7, 3) - SCALE_DIFF_ZERO;
+ clipped_offset = av_clip(offset[2], -155, 100);
+ if (offset[2] != clipped_offset) {
+ avpriv_request_sample(ac->avctx,
+ "If you heard an audible artifact, there may be a bug in the decoder. "
+ "Clipped intensity stereo position (%d -> %d)",
+ offset[2], clipped_offset);
}
+ sce->sfo[g*ics->max_sfb + sfb] = clipped_offset - 100;
break;
case NOISE_BT:
- for (; i < run_end; i++, idx++) {
- if (noise_flag-- > 0)
- offset[1] += get_bits(gb, NOISE_PRE_BITS) - NOISE_PRE;
- else
- offset[1] += get_vlc2(gb, ff_vlc_scalefactors, 7, 3) - SCALE_DIFF_ZERO;
- clipped_offset = av_clip(offset[1], -100, 155);
- if (offset[1] != clipped_offset) {
- avpriv_request_sample(ac->avctx,
- "If you heard an audible artifact, there may be a bug in the decoder. "
- "Clipped noise gain (%d -> %d)",
- offset[1], clipped_offset);
- }
- sfo[idx] = clipped_offset;
+ if (noise_flag-- > 0)
+ offset[1] += get_bits(gb, NOISE_PRE_BITS) - NOISE_PRE;
+ else
+ offset[1] += get_vlc2(gb, ff_vlc_scalefactors, 7, 3) - SCALE_DIFF_ZERO;
+ clipped_offset = av_clip(offset[1], -100, 155);
+ if (offset[1] != clipped_offset) {
+ avpriv_request_sample(ac->avctx,
+ "If you heard an audible artifact, there may be a bug in the decoder. "
+ "Clipped noise gain (%d -> %d)",
+ offset[1], clipped_offset);
}
+ sce->sfo[g*ics->max_sfb + sfb] = clipped_offset;
break;
default:
- for (; i < run_end; i++, idx++) {
- offset[0] += get_vlc2(gb, ff_vlc_scalefactors, 7, 3) - SCALE_DIFF_ZERO;
- if (offset[0] > 255U) {
- av_log(ac->avctx, AV_LOG_ERROR,
- "Scalefactor (%d) out of range.\n", offset[0]);
- return AVERROR_INVALIDDATA;
- }
- sfo[idx] = offset[0];
+ offset[0] += get_vlc2(gb, ff_vlc_scalefactors, 7, 3) - SCALE_DIFF_ZERO;
+ if (offset[0] > 255U) {
+ av_log(ac->avctx, AV_LOG_ERROR,
+ "Scalefactor (%d) out of range.\n", offset[0]);
+ return AVERROR_INVALIDDATA;
}
+ sce->sfo[g*ics->max_sfb + sfb] = offset[0] - 100;
break;
}
}
}
+
return 0;
}
@@ -1680,11 +1668,9 @@ int ff_aac_decode_ics(AACDecContext *ac, SingleChannelElement *sce,
goto fail;
}
- if ((ret = decode_band_types(ac, sce->band_type,
- sce->band_type_run_end, gb, ics)) < 0)
+ if ((ret = decode_band_types(ac, sce, gb)) < 0)
goto fail;
- if ((ret = decode_scalefactors(ac, sce->sfo, gb, global_gain, ics,
- sce->band_type, sce->band_type_run_end)) < 0)
+ if ((ret = decode_scalefactors(ac, sce, gb, global_gain)) < 0)
goto fail;
ac->dsp.dequant_scalefactors(sce);
diff --git a/libavcodec/aac/aacdec.h b/libavcodec/aac/aacdec.h
index eed53c6c96..bbb7ea358f 100644
--- a/libavcodec/aac/aacdec.h
+++ b/libavcodec/aac/aacdec.h
@@ -146,9 +146,8 @@ typedef struct SingleChannelElement {
IndividualChannelStream ics;
TemporalNoiseShaping tns;
enum BandType band_type[128]; ///< band types
- int band_type_run_end[120]; ///< band type run end points
- int sfo[120]; ///< scalefactor offsets
- INTFLOAT_UNION(sf, [120]); ///< scalefactors
+ int sfo[128]; ///< scalefactor offsets
+ INTFLOAT_UNION(sf, [128]); ///< scalefactors (8 windows * 16 sfb max)
INTFLOAT_ALIGNED_UNION(32, coeffs, 1024); ///< coefficients for IMDCT, maybe processed
INTFLOAT_ALIGNED_UNION(32, saved, 1536); ///< overlap
INTFLOAT_ALIGNED_UNION(32, ret_buf, 2048); ///< PCM output buffer
diff --git a/libavcodec/aac/aacdec_dsp_template.c b/libavcodec/aac/aacdec_dsp_template.c
index 621baef8ca..e69970472c 100644
--- a/libavcodec/aac/aacdec_dsp_template.c
+++ b/libavcodec/aac/aacdec_dsp_template.c
@@ -41,47 +41,37 @@
static void AAC_RENAME(dequant_scalefactors)(SingleChannelElement *sce)
{
IndividualChannelStream *ics = &sce->ics;
- const enum BandType *band_type = sce->band_type;
- const int *band_type_run_end = sce->band_type_run_end;
const int *sfo = sce->sfo;
INTFLOAT *sf = sce->AAC_RENAME(sf);
- int g, i, idx = 0;
- for (g = 0; g < ics->num_window_groups; g++) {
- for (i = 0; i < ics->max_sfb;) {
- int run_end = band_type_run_end[idx];
- switch (band_type[idx]) {
+ 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]) {
case ZERO_BT:
- for (; i < run_end; i++, idx++)
- sf[idx] = FIXR(0.);
+ sf[idx] = FIXR(0.);
break;
case INTENSITY_BT: /* fallthrough */
case INTENSITY_BT2:
- for (; i < run_end; i++, idx++) {
#if USE_FIXED
- sf[idx] = 100 - sfo[idx];
+ sf[idx] = 100 - (sfo[idx] + 100);
#else
- sf[idx] = ff_aac_pow2sf_tab[-sfo[idx] + POW_SF2_ZERO];
+ sf[idx] = ff_aac_pow2sf_tab[-sfo[idx] - 100 + POW_SF2_ZERO];
#endif /* USE_FIXED */
- }
break;
case NOISE_BT:
- for (; i < run_end; i++, idx++) {
#if USE_FIXED
- sf[idx] = -(100 + sfo[idx]);
+ sf[idx] = -(100 + sfo[idx]);
#else
- sf[idx] = -ff_aac_pow2sf_tab[sfo[idx] + POW_SF2_ZERO];
+ sf[idx] = -ff_aac_pow2sf_tab[sfo[idx] + POW_SF2_ZERO];
#endif /* USE_FIXED */
- }
break;
default:
- for (; i < run_end; i++, idx++) {
#if USE_FIXED
- sf[idx] = -sfo[idx];
+ sf[idx] = -sfo[idx] - 100;
#else
- sf[idx] = -ff_aac_pow2sf_tab[sfo[idx] - 100 + POW_SF2_ZERO];
+ sf[idx] = -ff_aac_pow2sf_tab[sfo[idx] + POW_SF2_ZERO];
#endif /* USE_FIXED */
- }
break;
}
}
@@ -96,25 +86,23 @@ static void AAC_RENAME(apply_mid_side_stereo)(AACDecContext *ac, ChannelElement
const IndividualChannelStream *ics = &cpe->ch[0].ics;
INTFLOAT *ch0 = cpe->ch[0].AAC_RENAME(coeffs);
INTFLOAT *ch1 = cpe->ch[1].AAC_RENAME(coeffs);
- int g, i, group, idx = 0;
const uint16_t *offsets = ics->swb_offset;
- for (g = 0; g < ics->num_window_groups; g++) {
- for (i = 0; i < ics->max_sfb; i++, idx++) {
+ 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;
if (cpe->ms_mask[idx] &&
cpe->ch[0].band_type[idx] < NOISE_BT &&
cpe->ch[1].band_type[idx] < NOISE_BT) {
+ for (int group = 0; group < ics->group_len[g]; group++)
#if USE_FIXED
- for (group = 0; group < ics->group_len[g]; group++) {
- ac->fdsp->butterflies_fixed(ch0 + group * 128 + offsets[i],
- ch1 + group * 128 + offsets[i],
- offsets[i+1] - offsets[i]);
+ ac->fdsp->butterflies_fixed(ch0 + group * 128 + offsets[sfb],
+ ch1 + group * 128 + offsets[sfb],
+ offsets[sfb+1] - offsets[sfb]);
#else
- for (group = 0; group < ics->group_len[g]; group++) {
- ac->fdsp->butterflies_float(ch0 + group * 128 + offsets[i],
- ch1 + group * 128 + offsets[i],
- offsets[i+1] - offsets[i]);
+ ac->fdsp->butterflies_float(ch0 + group * 128 + offsets[sfb],
+ ch1 + group * 128 + offsets[sfb],
+ offsets[sfb+1] - offsets[sfb]);
#endif /* USE_FIXED */
- }
}
}
ch0 += ics->group_len[g] * 128;
@@ -136,37 +124,30 @@ static void AAC_RENAME(apply_intensity_stereo)(AACDecContext *ac,
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 g, group, i, idx = 0;
int c;
INTFLOAT scale;
- for (g = 0; g < ics->num_window_groups; g++) {
- for (i = 0; i < ics->max_sfb;) {
+ 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;
if (sce1->band_type[idx] == INTENSITY_BT ||
sce1->band_type[idx] == INTENSITY_BT2) {
- const int bt_run_end = sce1->band_type_run_end[idx];
- for (; i < bt_run_end; i++, idx++) {
- 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 (group = 0; group < ics->group_len[g]; group++)
+ 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++)
#if USE_FIXED
- subband_scale(coef1 + group * 128 + offsets[i],
- coef0 + group * 128 + offsets[i],
- scale,
- 23,
- offsets[i + 1] - offsets[i] ,ac->avctx);
+ subband_scale(coef1 + group * 128 + offsets[sfb],
+ coef0 + group * 128 + offsets[sfb],
+ scale,
+ 23,
+ offsets[sfb + 1] - offsets[sfb], ac->avctx);
#else
- ac->fdsp->vector_fmul_scalar(coef1 + group * 128 + offsets[i],
- coef0 + group * 128 + offsets[i],
- scale,
- offsets[i + 1] - offsets[i]);
+ ac->fdsp->vector_fmul_scalar(coef1 + group * 128 + offsets[sfb],
+ coef0 + group * 128 + offsets[sfb],
+ scale,
+ offsets[sfb + 1] - offsets[sfb]);
#endif /* USE_FIXED */
- }
- } else {
- int bt_run_end = sce1->band_type_run_end[idx];
- idx += bt_run_end - i;
- i = bt_run_end;
}
}
coef0 += ics->group_len[g] * 128;
--
2.43.0.381.gb435a96ce8
_______________________________________________
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
To unsubscribe, visit link above, or email
ffmpeg-devel-request@ffmpeg.org with subject "unsubscribe".
next prev parent reply other threads:[~2024-05-19 16:55 UTC|newest]
Thread overview: 22+ messages / expand[flat|nested] mbox.gz Atom feed top
2024-05-19 16:54 [FFmpeg-devel] [PATCH v2 0/8] aacdec: add a native xHE-AAC decoder Lynne via ffmpeg-devel
2024-05-19 16:54 ` Lynne via ffmpeg-devel [this message]
2024-05-19 16:54 ` [FFmpeg-devel] [PATCH v2 2/8] aacdec: expose channel layout related functions Lynne via ffmpeg-devel
2024-05-19 16:54 ` [FFmpeg-devel] [PATCH v2 3/8] aacdec: expose decode_tns Lynne via ffmpeg-devel
2024-05-19 16:54 ` [FFmpeg-devel] [PATCH v2 4/8] aacdec_dsp: implement 768-point transform and windowing Lynne via ffmpeg-devel
2024-05-19 16:54 ` [FFmpeg-devel] [PATCH v2 5/8] aactab: add deemphasis tables for USAC Lynne via ffmpeg-devel
2024-05-19 16:54 ` [FFmpeg-devel] [PATCH v2 6/8] aactab: add tables for the new USAC arithmetic coder Lynne via ffmpeg-devel
2024-05-19 16:54 ` [FFmpeg-devel] [PATCH v2 7/8] aactab: add new scalefactor offset tables for 96/768pt windows Lynne via ffmpeg-devel
2024-05-19 16:54 ` [FFmpeg-devel] [PATCH v2 8/8] aacdec: add a decoder for AAC USAC (xHE-AAC) Lynne via ffmpeg-devel
2024-05-19 19:39 ` Marton Balint
2024-05-19 19:50 ` Lynne via ffmpeg-devel
2024-05-21 7:16 ` Marton Balint
2024-05-21 17:58 ` Lynne via ffmpeg-devel
2024-05-21 19:40 ` Marton Balint
2024-05-21 19:52 ` Lynne via ffmpeg-devel
2024-05-21 20:12 ` Marton Balint
2024-05-21 21:33 ` Hendrik Leppkes
2024-05-21 22:09 ` Lynne via ffmpeg-devel
2024-05-22 20:15 ` Marton Balint
2024-05-22 20:25 ` Lynne via ffmpeg-devel
2024-05-19 23:19 ` Michael Niedermayer
2024-05-20 1:11 ` Lynne via ffmpeg-devel
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20240519165444.829271-2-dev@lynne.ee \
--to=ffmpeg-devel@ffmpeg.org \
--cc=dev@lynne.ee \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Git Inbox Mirror of the ffmpeg-devel mailing list - see https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
This inbox may be cloned and mirrored by anyone:
git clone --mirror https://master.gitmailbox.com/ffmpegdev/0 ffmpegdev/git/0.git
# If you have public-inbox 1.1+ installed, you may
# initialize and index your mirror using the following commands:
public-inbox-init -V2 ffmpegdev ffmpegdev/ https://master.gitmailbox.com/ffmpegdev \
ffmpegdev@gitmailbox.com
public-inbox-index ffmpegdev
Example config snippet for mirrors.
AGPL code for this site: git clone https://public-inbox.org/public-inbox.git