Chiebot-Mirror
/
FFmpeg
mirror of https://github.com/FFmpeg/FFmpeg.git
8
0
Fork 0
Code Issues Projects Releases Wiki Activity

on2avc: convert to lavu/tx

Browse Source
pull/388/head
Lynne 2 years ago
parent 8cd46c48ac
commit 6ba0aa1770
No known key found for this signature in database
GPG Key ID: A2FEA5F03F034464
2 changed files with 75 additions and 69 deletions
Whitespace
Show all changes Ignore whitespace when comparing lines Ignore changes in amount of whitespace Ignore changes in whitespace at EOL
Split View
Diff Options
Show Stats Download Patch File Download Diff File
  1. 1
      configure
  2. 143
      libavcodec/on2avc.c

1
configure vendored
Unescape View File

@ -2917,7 +2917,6 @@ nellymoser_decoder_select="sinewin"
nellymoser_encoder_select="audio_frame_queue sinewin"
notchlc_decoder_select="lzf"
nuv_decoder_select="idctdsp"
on2avc_decoder_select="mdct"
opus_decoder_deps="swresample"
opus_encoder_select="audio_frame_queue"
png_decoder_select="inflate_wrapper"

143
libavcodec/on2avc.c
Unescape View File

@ -24,12 +24,12 @@
#include "libavutil/ffmath.h"
#include "libavutil/float_dsp.h"
#include "libavutil/mem_internal.h"
#include "libavutil/tx.h"
#include "avcodec.h"
#include "bytestream.h"
#include "codec_internal.h"
#include "decode.h"
#include "fft.h"
#include "get_bits.h"
#include "on2avcdata.h"
@ -50,8 +50,10 @@ enum WindowTypes {
typedef struct On2AVCContext {
AVCodecContext *avctx;
AVFloatDSPContext *fdsp;
FFTContext mdct, mdct_half, mdct_small;
FFTContext fft128, fft256, fft512, fft1024;
AVTXContext *mdct, *mdct_half, *mdct_small;
AVTXContext *fft128, *fft256, *fft512, *fft1024;
av_tx_fn mdct_fn, mdct_half_fn, mdct_small_fn;
av_tx_fn fft128_fn, fft256_fn, fft512_fn, fft1024_fn;
void (*wtf)(struct On2AVCContext *ctx, float *out, float *in, int size);
int is_av500;
@ -465,67 +467,57 @@ static void combine_fft(float *s0, float *s1, float *s2, float *s3, float *dst,
static void wtf_end_512(On2AVCContext *c, float *out, float *src,
float *tmp0, float *tmp1)
{
memcpy(src, tmp0, 384 * sizeof(*tmp0));
memcpy(tmp0 + 384, src + 384, 128 * sizeof(*tmp0));
zero_head_and_tail(src, 128, 16, 4);
zero_head_and_tail(src + 128, 128, 16, 4);
zero_head_and_tail(src + 256, 128, 13, 7);
zero_head_and_tail(src + 384, 128, 15, 5);
c->fft128.fft_permute(&c->fft128, (FFTComplex*)src);
c->fft128.fft_permute(&c->fft128, (FFTComplex*)(src + 128));
c->fft128.fft_permute(&c->fft128, (FFTComplex*)(src + 256));
c->fft128.fft_permute(&c->fft128, (FFTComplex*)(src + 384));
c->fft128.fft_calc(&c->fft128, (FFTComplex*)src);
c->fft128.fft_calc(&c->fft128, (FFTComplex*)(src + 128));
c->fft128.fft_calc(&c->fft128, (FFTComplex*)(src + 256));
c->fft128.fft_calc(&c->fft128, (FFTComplex*)(src + 384));
memcpy(tmp1, tmp0, 384 * sizeof(*tmp0));
memcpy(tmp0 + 384, tmp1 + 384, 128 * sizeof(*tmp0));
zero_head_and_tail(tmp1, 128, 16, 4);
zero_head_and_tail(tmp1 + 128, 128, 16, 4);
zero_head_and_tail(tmp1 + 256, 128, 13, 7);
zero_head_and_tail(tmp1 + 384, 128, 15, 5);
c->fft128_fn(c->fft128, src + 0, tmp1 + 0, sizeof(float));
c->fft128_fn(c->fft128, src + 128, tmp1 + 128, sizeof(float));
c->fft128_fn(c->fft128, src + 256, tmp1 + 256, sizeof(float));
c->fft128_fn(c->fft128, src + 384, tmp1 + 384, sizeof(float));
combine_fft(src, src + 128, src + 256, src + 384, tmp1,
ff_on2avc_ctab_1, ff_on2avc_ctab_2,
ff_on2avc_ctab_3, ff_on2avc_ctab_4, 512, 2);
c->fft512.fft_permute(&c->fft512, (FFTComplex*)tmp1);
c->fft512.fft_calc(&c->fft512, (FFTComplex*)tmp1);
pretwiddle(&tmp0[ 0], tmp1, 512, 84, 4, 16, 4, ff_on2avc_tabs_20_84_1);
pretwiddle(&tmp0[128], tmp1, 512, 84, 4, 16, 4, ff_on2avc_tabs_20_84_2);
pretwiddle(&tmp0[256], tmp1, 512, 84, 4, 13, 7, ff_on2avc_tabs_20_84_3);
pretwiddle(&tmp0[384], tmp1, 512, 84, 4, 15, 5, ff_on2avc_tabs_20_84_4);
c->fft512_fn(c->fft512, src, tmp1, sizeof(float));
memcpy(src, tmp1, 512 * sizeof(float));
pretwiddle(&tmp0[ 0], src, 512, 84, 4, 16, 4, ff_on2avc_tabs_20_84_1);
pretwiddle(&tmp0[128], src, 512, 84, 4, 16, 4, ff_on2avc_tabs_20_84_2);
pretwiddle(&tmp0[256], src, 512, 84, 4, 13, 7, ff_on2avc_tabs_20_84_3);
pretwiddle(&tmp0[384], src, 512, 84, 4, 15, 5, ff_on2avc_tabs_20_84_4);
}
static void wtf_end_1024(On2AVCContext *c, float *out, float *src,
float *tmp0, float *tmp1)
{
memcpy(src, tmp0, 768 * sizeof(*tmp0));
memcpy(tmp0 + 768, src + 768, 256 * sizeof(*tmp0));
zero_head_and_tail(src, 256, 16, 4);
zero_head_and_tail(src + 256, 256, 16, 4);
zero_head_and_tail(src + 512, 256, 13, 7);
zero_head_and_tail(src + 768, 256, 15, 5);
c->fft256.fft_permute(&c->fft256, (FFTComplex*)src);
c->fft256.fft_permute(&c->fft256, (FFTComplex*)(src + 256));
c->fft256.fft_permute(&c->fft256, (FFTComplex*)(src + 512));
c->fft256.fft_permute(&c->fft256, (FFTComplex*)(src + 768));
c->fft256.fft_calc(&c->fft256, (FFTComplex*)src);
c->fft256.fft_calc(&c->fft256, (FFTComplex*)(src + 256));
c->fft256.fft_calc(&c->fft256, (FFTComplex*)(src + 512));
c->fft256.fft_calc(&c->fft256, (FFTComplex*)(src + 768));
memcpy(tmp1, tmp0, 768 * sizeof(*tmp0));
memcpy(tmp0 + 768, tmp1 + 768, 256 * sizeof(*tmp0));
zero_head_and_tail(tmp1, 256, 16, 4);
zero_head_and_tail(tmp1 + 256, 256, 16, 4);
zero_head_and_tail(tmp1 + 512, 256, 13, 7);
zero_head_and_tail(tmp1 + 768, 256, 15, 5);
c->fft256_fn(c->fft256, src + 0, tmp1 + 0, sizeof(float));
c->fft256_fn(c->fft256, src + 256, tmp1 + 256, sizeof(float));
c->fft256_fn(c->fft256, src + 512, tmp1 + 512, sizeof(float));
c->fft256_fn(c->fft256, src + 768, tmp1 + 768, sizeof(float));
combine_fft(src, src + 256, src + 512, src + 768, tmp1,
ff_on2avc_ctab_1, ff_on2avc_ctab_2,
ff_on2avc_ctab_3, ff_on2avc_ctab_4, 1024, 1);
c->fft1024.fft_permute(&c->fft1024, (FFTComplex*)tmp1);
c->fft1024.fft_calc(&c->fft1024, (FFTComplex*)tmp1);
pretwiddle(&tmp0[ 0], tmp1, 1024, 84, 4, 16, 4, ff_on2avc_tabs_20_84_1);
pretwiddle(&tmp0[256], tmp1, 1024, 84, 4, 16, 4, ff_on2avc_tabs_20_84_2);
pretwiddle(&tmp0[512], tmp1, 1024, 84, 4, 13, 7, ff_on2avc_tabs_20_84_3);
pretwiddle(&tmp0[768], tmp1, 1024, 84, 4, 15, 5, ff_on2avc_tabs_20_84_4);
c->fft1024_fn(c->fft1024, src, tmp1, sizeof(float));
memcpy(src, tmp1, 1024 * sizeof(float));
pretwiddle(&tmp0[ 0], src, 1024, 84, 4, 16, 4, ff_on2avc_tabs_20_84_1);
pretwiddle(&tmp0[256], src, 1024, 84, 4, 16, 4, ff_on2avc_tabs_20_84_2);
pretwiddle(&tmp0[512], src, 1024, 84, 4, 13, 7, ff_on2avc_tabs_20_84_3);
pretwiddle(&tmp0[768], src, 1024, 84, 4, 15, 5, ff_on2avc_tabs_20_84_4);
}
static void wtf_40(On2AVCContext *c, float *out, float *src, int size)
@ -701,20 +693,20 @@ static int on2avc_reconstruct_channel_ext(On2AVCContext *c, AVFrame *dst, int of
switch (c->window_type) {
case WINDOW_TYPE_EXT7:
c->mdct.imdct_half(&c->mdct, buf, in);
c->mdct_fn(c->mdct, buf, in, sizeof(float));
break;
case WINDOW_TYPE_EXT4:
c->wtf(c, buf, in, 1024);
break;
case WINDOW_TYPE_EXT5:
c->wtf(c, buf, in, 512);
c->mdct.imdct_half(&c->mdct_half, buf + 512, in + 512);
c->mdct_half_fn(c->mdct, buf + 512, in + 512, sizeof(float));
for (i = 0; i < 256; i++) {
FFSWAP(float, buf[i + 512], buf[1023 - i]);
}
break;
case WINDOW_TYPE_EXT6:
c->mdct.imdct_half(&c->mdct_half, buf, in);
c->mdct_half_fn(c->mdct_half, buf, in, sizeof(float));
for (i = 0; i < 256; i++) {
FFSWAP(float, buf[i], buf[511 - i]);
}
@ -747,11 +739,11 @@ static int on2avc_reconstruct_channel(On2AVCContext *c, int channel,
case WINDOW_TYPE_LONG_START:
case WINDOW_TYPE_LONG_STOP:
case WINDOW_TYPE_LONG:
c->mdct.imdct_half(&c->mdct, buf, in);
c->mdct_fn(c->mdct, buf, in, sizeof(float));
break;
case WINDOW_TYPE_8SHORT:
for (i = 0; i < ON2AVC_SUBFRAME_SIZE; i += ON2AVC_SUBFRAME_SIZE / 8)
c->mdct_small.imdct_half(&c->mdct_small, buf + i, in + i);
c->mdct_small_fn(c->mdct_small, buf + i, in + i, sizeof(float));
break;
}
@ -911,6 +903,7 @@ static av_cold int on2avc_decode_init(AVCodecContext *avctx)
const uint8_t *lens = ff_on2avc_cb_lens;
const uint16_t *syms = ff_on2avc_cb_syms;
int channels = avctx->ch_layout.nb_channels;
float scale;
int i, ret;
if (channels > 2U) {
@ -951,13 +944,27 @@ static av_cold int on2avc_decode_init(AVCodecContext *avctx)
c->wtf = (avctx->sample_rate <= 40000) ? wtf_40
: wtf_44;
ff_mdct_init(&c->mdct, 11, 1, 1.0 / (32768.0 * 1024.0));
ff_mdct_init(&c->mdct_half, 10, 1, 1.0 / (32768.0 * 512.0));
ff_mdct_init(&c->mdct_small, 8, 1, 1.0 / (32768.0 * 128.0));
ff_fft_init(&c->fft128, 6, 0);
ff_fft_init(&c->fft256, 7, 0);
ff_fft_init(&c->fft512, 8, 1);
ff_fft_init(&c->fft1024, 9, 1);
scale = 1.0 / (1024*32768);
if ((ret = av_tx_init(&c->mdct, &c->mdct_fn, AV_TX_FLOAT_MDCT, 1, 1024, &scale, 0)) < 0)
return ret;
scale = 1.0 / (512*32768);
if ((ret = av_tx_init(&c->mdct_half, &c->mdct_half_fn, AV_TX_FLOAT_MDCT, 1, 512, &scale, 0)) < 0)
return ret;
scale = 1.0 / (128*32768);
if ((ret = av_tx_init(&c->mdct_small, &c->mdct_small_fn, AV_TX_FLOAT_MDCT, 1, 128, &scale, 0)) < 0)
return ret;
if ((ret = av_tx_init(&c->fft1024, &c->fft1024_fn, AV_TX_FLOAT_FFT, 1, 1024, NULL, 0)) < 0)
return ret;
if ((ret = av_tx_init(&c->fft512, &c->fft512_fn, AV_TX_FLOAT_FFT, 1, 512, NULL, 0)) < 0)
return ret;
if ((ret = av_tx_init(&c->fft256, &c->fft256_fn, AV_TX_FLOAT_FFT, 0, 256, NULL, 0)) < 0)
return ret;
if ((ret = av_tx_init(&c->fft128, &c->fft128_fn, AV_TX_FLOAT_FFT, 0, 128, NULL, 0)) < 0)
return ret;
c->fdsp = avpriv_float_dsp_alloc(avctx->flags & AV_CODEC_FLAG_BITEXACT);
if (!c->fdsp)
return AVERROR(ENOMEM);
@ -988,13 +995,13 @@ static av_cold int on2avc_decode_close(AVCodecContext *avctx)
{
On2AVCContext *c = avctx->priv_data;
ff_mdct_end(&c->mdct);
ff_mdct_end(&c->mdct_half);
ff_mdct_end(&c->mdct_small);
ff_fft_end(&c->fft128);
ff_fft_end(&c->fft256);
ff_fft_end(&c->fft512);
ff_fft_end(&c->fft1024);
av_tx_uninit(&c->mdct);
av_tx_uninit(&c->mdct_half);
av_tx_uninit(&c->mdct_small);
av_tx_uninit(&c->fft128);
av_tx_uninit(&c->fft256);
av_tx_uninit(&c->fft512);
av_tx_uninit(&c->fft1024);
av_freep(&c->fdsp);

Write Preview
Loading…
Cancel
Save