[FFmpeg-cvslog] opus: convert encoder and decoder to lavu/tx
Lynne
git at videolan.org
Mon Sep 26 21:28:33 EEST 2022
ffmpeg | branch: master | Lynne <dev at lynne.ee> | Sat Sep 24 00:46:44 2022 +0200| [543a46b4b2ff0f02fa85e45d7b7bcda19d3be9b4] | committer: Lynne
opus: convert encoder and decoder to lavu/tx
This commit changes both the encoder and decoder to use the new lavu/tx code,
which has faster C transforms and more assembly optimizations.
> http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=543a46b4b2ff0f02fa85e45d7b7bcda19d3be9b4
---
libavcodec/opus_celt.c | 20 ++++++++++++--------
libavcodec/opus_celt.h | 5 +++--
libavcodec/opusenc.c | 15 +++++++++------
libavcodec/opusenc_psy.c | 13 ++++++++-----
libavcodec/opusenc_psy.h | 4 +++-
5 files changed, 35 insertions(+), 22 deletions(-)
diff --git a/libavcodec/opus_celt.c b/libavcodec/opus_celt.c
index 9dbeff1927..f1fb88a56d 100644
--- a/libavcodec/opus_celt.c
+++ b/libavcodec/opus_celt.c
@@ -323,7 +323,8 @@ int ff_celt_decode_frame(CeltFrame *f, OpusRangeCoder *rc,
{
int i, j, downmix = 0;
int consumed; // bits of entropy consumed thus far for this frame
- MDCT15Context *imdct;
+ AVTXContext *imdct;
+ av_tx_fn imdct_fn;
if (channels != 1 && channels != 2) {
av_log(f->avctx, AV_LOG_ERROR, "Invalid number of coded channels: %d\n",
@@ -385,7 +386,8 @@ int ff_celt_decode_frame(CeltFrame *f, OpusRangeCoder *rc,
f->blocks = f->transient ? 1 << f->size : 1;
f->blocksize = frame_size / f->blocks;
- imdct = f->imdct[f->transient ? 0 : f->size];
+ imdct = f->tx[f->transient ? 0 : f->size];
+ imdct_fn = f->tx_fn[f->transient ? 0 : f->size];
if (channels == 1) {
for (i = 0; i < CELT_MAX_BANDS; i++)
@@ -440,8 +442,8 @@ int ff_celt_decode_frame(CeltFrame *f, OpusRangeCoder *rc,
for (j = 0; j < f->blocks; j++) {
float *dst = block->buf + 1024 + j * f->blocksize;
- imdct->imdct_half(imdct, dst + CELT_OVERLAP / 2, f->block[i].coeffs + j,
- f->blocks);
+ imdct_fn(imdct, dst + CELT_OVERLAP / 2, f->block[i].coeffs + j,
+ sizeof(float)*f->blocks);
f->dsp->vector_fmul_window(dst, dst, dst + CELT_OVERLAP / 2,
ff_celt_window, CELT_OVERLAP / 2);
}
@@ -526,8 +528,8 @@ void ff_celt_free(CeltFrame **f)
if (!frm)
return;
- for (i = 0; i < FF_ARRAY_ELEMS(frm->imdct); i++)
- ff_mdct15_uninit(&frm->imdct[i]);
+ for (i = 0; i < FF_ARRAY_ELEMS(frm->tx); i++)
+ av_tx_uninit(&frm->tx[i]);
ff_celt_pvq_uninit(&frm->pvq);
@@ -555,9 +557,11 @@ int ff_celt_init(AVCodecContext *avctx, CeltFrame **f, int output_channels,
frm->output_channels = output_channels;
frm->apply_phase_inv = apply_phase_inv;
- for (i = 0; i < FF_ARRAY_ELEMS(frm->imdct); i++)
- if ((ret = ff_mdct15_init(&frm->imdct[i], 1, i + 3, -1.0f/32768)) < 0)
+ for (i = 0; i < FF_ARRAY_ELEMS(frm->tx); i++) {
+ const float scale = -1.0f/32768;
+ if ((ret = av_tx_init(&frm->tx[i], &frm->tx_fn[i], AV_TX_FLOAT_MDCT, 1, 15 << (i + 3), &scale, 0)) < 0)
goto fail;
+ }
if ((ret = ff_celt_pvq_init(&frm->pvq, 0)) < 0)
goto fail;
diff --git a/libavcodec/opus_celt.h b/libavcodec/opus_celt.h
index 661ca251de..291a544298 100644
--- a/libavcodec/opus_celt.h
+++ b/libavcodec/opus_celt.h
@@ -30,10 +30,10 @@
#include "opus_pvq.h"
#include "opusdsp.h"
-#include "mdct15.h"
#include "libavutil/float_dsp.h"
#include "libavutil/libm.h"
#include "libavutil/mem_internal.h"
+#include "libavutil/tx.h"
#define CELT_VECTORS 11
#define CELT_ALLOC_STEPS 6
@@ -93,7 +93,8 @@ typedef struct CeltBlock {
struct CeltFrame {
// constant values that do not change during context lifetime
AVCodecContext *avctx;
- MDCT15Context *imdct[4];
+ AVTXContext *tx[4];
+ av_tx_fn tx_fn[4];
AVFloatDSPContext *dsp;
CeltBlock block[2];
CeltPVQ *pvq;
diff --git a/libavcodec/opusenc.c b/libavcodec/opusenc.c
index a7a9d3a5f5..8cdd27d930 100644
--- a/libavcodec/opusenc.c
+++ b/libavcodec/opusenc.c
@@ -40,7 +40,8 @@ typedef struct OpusEncContext {
AVCodecContext *avctx;
AudioFrameQueue afq;
AVFloatDSPContext *dsp;
- MDCT15Context *mdct[CELT_BLOCK_NB];
+ AVTXContext *tx[CELT_BLOCK_NB];
+ av_tx_fn tx_fn[CELT_BLOCK_NB];
CeltPVQ *pvq;
struct FFBufQueue bufqueue;
@@ -204,7 +205,7 @@ static void celt_frame_mdct(OpusEncContext *s, CeltFrame *f)
s->dsp->vector_fmul_reverse(&win[CELT_OVERLAP], src2,
ff_celt_window - 8, 128);
src1 = src2;
- s->mdct[0]->mdct(s->mdct[0], b->coeffs + t, win, f->blocks);
+ s->tx_fn[0](s->tx[0], b->coeffs + t, win, sizeof(float)*f->blocks);
}
}
} else {
@@ -226,7 +227,7 @@ static void celt_frame_mdct(OpusEncContext *s, CeltFrame *f)
ff_celt_window - 8, 128);
memcpy(win + lap_dst + blk_len, temp, CELT_OVERLAP*sizeof(float));
- s->mdct[f->size]->mdct(s->mdct[f->size], b->coeffs, win, 1);
+ s->tx_fn[f->size](s->tx[f->size], b->coeffs, win, sizeof(float));
}
}
@@ -612,7 +613,7 @@ static av_cold int opus_encode_end(AVCodecContext *avctx)
OpusEncContext *s = avctx->priv_data;
for (int i = 0; i < CELT_BLOCK_NB; i++)
- ff_mdct15_uninit(&s->mdct[i]);
+ av_tx_uninit(&s->tx[i]);
ff_celt_pvq_uninit(&s->pvq);
av_freep(&s->dsp);
@@ -668,9 +669,11 @@ static av_cold int opus_encode_init(AVCodecContext *avctx)
return AVERROR(ENOMEM);
/* I have no idea why a base scaling factor of 68 works, could be the twiddles */
- for (int i = 0; i < CELT_BLOCK_NB; i++)
- if ((ret = ff_mdct15_init(&s->mdct[i], 0, i + 3, 68 << (CELT_BLOCK_NB - 1 - i))))
+ for (int i = 0; i < CELT_BLOCK_NB; i++) {
+ const float scale = 68 << (CELT_BLOCK_NB - 1 - i);
+ if ((ret = av_tx_init(&s->tx[i], &s->tx_fn[i], AV_TX_FLOAT_MDCT, 0, 15 << (i + 3), &scale, 0)))
return AVERROR(ENOMEM);
+ }
/* Zero out previous energy (matters for inter first frame) */
for (int ch = 0; ch < s->channels; ch++)
diff --git a/libavcodec/opusenc_psy.c b/libavcodec/opusenc_psy.c
index 1c8f69269c..3bff57d347 100644
--- a/libavcodec/opusenc_psy.c
+++ b/libavcodec/opusenc_psy.c
@@ -22,7 +22,6 @@
#include "opusenc_psy.h"
#include "opus_pvq.h"
#include "opustab.h"
-#include "mdct15.h"
#include "libavutil/qsort.h"
static float pvq_band_cost(CeltPVQ *pvq, CeltFrame *f, OpusRangeCoder *rc, int band,
@@ -99,7 +98,8 @@ static void step_collect_psy_metrics(OpusPsyContext *s, int index)
s->dsp->vector_fmul(s->scratch, s->scratch, s->window[s->bsize_analysis],
(OPUS_BLOCK_SIZE(s->bsize_analysis) << 1));
- s->mdct[s->bsize_analysis]->mdct(s->mdct[s->bsize_analysis], st->coeffs[ch], s->scratch, 1);
+ s->mdct_fn[s->bsize_analysis](s->mdct[s->bsize_analysis], st->coeffs[ch],
+ s->scratch, sizeof(float));
for (i = 0; i < CELT_MAX_BANDS; i++)
st->bands[ch][i] = &st->coeffs[ch][ff_celt_freq_bands[i] << s->bsize_analysis];
@@ -558,13 +558,16 @@ av_cold int ff_opus_psy_init(OpusPsyContext *s, AVCodecContext *avctx,
for (i = 0; i < CELT_BLOCK_NB; i++) {
float tmp;
const int len = OPUS_BLOCK_SIZE(i);
+ const float scale = 68 << (CELT_BLOCK_NB - 1 - i);
s->window[i] = av_malloc(2*len*sizeof(float));
if (!s->window[i]) {
ret = AVERROR(ENOMEM);
goto fail;
}
generate_window_func(s->window[i], 2*len, WFUNC_SINE, &tmp);
- if ((ret = ff_mdct15_init(&s->mdct[i], 0, i + 3, 68 << (CELT_BLOCK_NB - 1 - i))))
+ ret = av_tx_init(&s->mdct[i], &s->mdct_fn[i], AV_TX_FLOAT_MDCT,
+ 0, 15 << (i + 3), &scale, 0);
+ if (ret < 0)
goto fail;
}
@@ -575,7 +578,7 @@ fail:
av_freep(&s->dsp);
for (i = 0; i < CELT_BLOCK_NB; i++) {
- ff_mdct15_uninit(&s->mdct[i]);
+ av_tx_uninit(&s->mdct[i]);
av_freep(&s->window[i]);
}
@@ -598,7 +601,7 @@ av_cold int ff_opus_psy_end(OpusPsyContext *s)
av_freep(&s->dsp);
for (i = 0; i < CELT_BLOCK_NB; i++) {
- ff_mdct15_uninit(&s->mdct[i]);
+ av_tx_uninit(&s->mdct[i]);
av_freep(&s->window[i]);
}
diff --git a/libavcodec/opusenc_psy.h b/libavcodec/opusenc_psy.h
index d4fb096a3d..0a7cdb6f2c 100644
--- a/libavcodec/opusenc_psy.h
+++ b/libavcodec/opusenc_psy.h
@@ -22,6 +22,7 @@
#ifndef AVCODEC_OPUSENC_PSY_H
#define AVCODEC_OPUSENC_PSY_H
+#include "libavutil/tx.h"
#include "libavutil/mem_internal.h"
#include "opusenc.h"
@@ -70,7 +71,8 @@ typedef struct OpusPsyContext {
int max_steps;
float *window[CELT_BLOCK_NB];
- MDCT15Context *mdct[CELT_BLOCK_NB];
+ AVTXContext *mdct[CELT_BLOCK_NB];
+ av_tx_fn mdct_fn[CELT_BLOCK_NB];
int bsize_analysis;
DECLARE_ALIGNED(32, float, scratch)[2048];
More information about the ffmpeg-cvslog
mailing list