falkTX
/
FFmpeg
mirror of https://github.com/falkTX/FFmpeg.git
1
0
Fork 0
Code Issues 0 Releases 338 Wiki Activity
You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
44689 Commits
32 Branches
824MB
Tree: c5714b51aa
Branches Tags
${ item.name }
Create branch ${ searchTerm }
from 'c5714b51aa'
${ noResults }
FFmpeg/libavcodec/on2avc.c

1026 lines
42KB
Raw Blame History 

  1. /*

  2.  * On2 Audio for Video Codec decoder

  3.  *

  4.  * Copyright (c) 2013 Konstantin Shishkov

  5.  *

  6.  * This file is part of Libav.

  7.  *

  8.  * Libav is free software; you can redistribute it and/or

  9.  * modify it under the terms of the GNU Lesser General Public

  10.  * License as published by the Free Software Foundation; either

  11.  * version 2.1 of the License, or (at your option) any later version.

  12.  *

  13.  * Libav is distributed in the hope that it will be useful,

  14.  * but WITHOUT ANY WARRANTY; without even the implied warranty of

  15.  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU

  16.  * Lesser General Public License for more details.

  17.  *

  18.  * You should have received a copy of the GNU Lesser General Public

  19.  * License along with Libav; if not, write to the Free Software

  20.  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA

  21.  */

  22. 

  23. #include "libavutil/channel_layout.h"

  24. #include "libavutil/float_dsp.h"

  25. 

  26. #include "avcodec.h"

  27. #include "bitstream.h"

  28. #include "bytestream.h"

  29. #include "fft.h"

  30. #include "internal.h"

  31. #include "vlc.h"

  32. 

  33. #include "on2avcdata.h"

  34. 

  35. #define ON2AVC_SUBFRAME_SIZE   1024

  36. 

  37. enum WindowTypes {

  38.     WINDOW_TYPE_LONG       = 0,

  39.     WINDOW_TYPE_LONG_STOP,

  40.     WINDOW_TYPE_LONG_START,

  41.     WINDOW_TYPE_8SHORT     = 3,

  42.     WINDOW_TYPE_EXT4,

  43.     WINDOW_TYPE_EXT5,

  44.     WINDOW_TYPE_EXT6,

  45.     WINDOW_TYPE_EXT7,

  46. };

  47. 

  48. typedef struct On2AVCContext {

  49.     AVCodecContext *avctx;

  50.     AVFloatDSPContext fdsp;

  51.     FFTContext mdct, mdct_half, mdct_small;

  52.     FFTContext fft128, fft256, fft512, fft1024;

  53.     void (*wtf)(struct On2AVCContext *ctx, float *out, float *in, int size);

  54. 

  55.     int is_av500;

  56. 

  57.     const On2AVCMode *modes;

  58.     int window_type, prev_window_type;

  59.     int num_windows, num_bands;

  60.     int bits_per_section;

  61.     const int *band_start;

  62. 

  63.     int grouping[8];

  64.     int ms_present;

  65.     int ms_info[ON2AVC_MAX_BANDS];

  66. 

  67.     int is_long;

  68. 

  69.     uint8_t band_type[ON2AVC_MAX_BANDS];

  70.     uint8_t band_run_end[ON2AVC_MAX_BANDS];

  71.     int     num_sections;

  72. 

  73.     float band_scales[ON2AVC_MAX_BANDS];

  74. 

  75.     VLC scale_diff;

  76.     VLC cb_vlc[16];

  77. 

  78.     float scale_tab[128];

  79. 

  80.     DECLARE_ALIGNED(32, float, coeffs)[2][ON2AVC_SUBFRAME_SIZE];

  81.     DECLARE_ALIGNED(32, float, delay) [2][ON2AVC_SUBFRAME_SIZE];

  82. 

  83.     DECLARE_ALIGNED(32, float, temp)     [ON2AVC_SUBFRAME_SIZE * 2];

  84.     DECLARE_ALIGNED(32, float, mdct_buf) [ON2AVC_SUBFRAME_SIZE];

  85.     DECLARE_ALIGNED(32, float, long_win) [ON2AVC_SUBFRAME_SIZE];

  86.     DECLARE_ALIGNED(32, float, short_win)[ON2AVC_SUBFRAME_SIZE / 8];

  87. } On2AVCContext;

  88. 

  89. static void on2avc_read_ms_info(On2AVCContext *c, BitstreamContext *bc)

  90. {

  91.     int w, b, band_off = 0;

  92. 

  93.     c->ms_present = bitstream_read_bit(bc);

  94.     if (!c->ms_present)

  95.         return;

  96.     for (w = 0; w < c->num_windows; w++) {

  97.         if (!c->grouping[w]) {

  98.             memcpy(c->ms_info + band_off,

  99.                    c->ms_info + band_off - c->num_bands,

  100.                    c->num_bands * sizeof(*c->ms_info));

  101.             band_off += c->num_bands;

  102.             continue;

  103.         }

  104.         for (b = 0; b < c->num_bands; b++)

  105.             c->ms_info[band_off++] = bitstream_read_bit(bc);

  106.     }

  107. }

  108. 

  109. // do not see Table 17 in ISO/IEC 13818-7

  110. static int on2avc_decode_band_types(On2AVCContext *c, BitstreamContext *bc)

  111. {

  112.     int bits_per_sect = c->is_long ? 5 : 3;

  113.     int esc_val = (1 << bits_per_sect) - 1;

  114.     int num_bands = c->num_bands * c->num_windows;

  115.     int band = 0, i, band_type, run_len, run;

  116. 

  117.     while (band < num_bands) {

  118.         band_type = bitstream_read(bc, 4);

  119.         run_len   = 1;

  120.         do {

  121.             run = bitstream_read(bc, bits_per_sect);

  122.             run_len += run;

  123.         } while (run == esc_val);

  124.         if (band + run_len > num_bands) {

  125.             av_log(c->avctx, AV_LOG_ERROR, "Invalid band type run\n");

  126.             return AVERROR_INVALIDDATA;

  127.         }

  128.         for (i = band; i < band + run_len; i++) {

  129.             c->band_type[i]    = band_type;

  130.             c->band_run_end[i] = band + run_len;

  131.         }

  132.         band += run_len;

  133.     }

  134. 

  135.     return 0;

  136. }

  137. 

  138. // completely not like Table 18 in ISO/IEC 13818-7

  139. // (no intensity stereo, different coding for the first coefficient)

  140. static int on2avc_decode_band_scales(On2AVCContext *c, BitstreamContext *bc)

  141. {

  142.     int w, w2, b, scale, first = 1;

  143.     int band_off = 0;

  144. 

  145.     for (w = 0; w < c->num_windows; w++) {

  146.         if (!c->grouping[w]) {

  147.             memcpy(c->band_scales + band_off,

  148.                    c->band_scales + band_off - c->num_bands,

  149.                    c->num_bands * sizeof(*c->band_scales));

  150.             band_off += c->num_bands;

  151.             continue;

  152.         }

  153.         for (b = 0; b < c->num_bands; b++) {

  154.             if (!c->band_type[band_off]) {

  155.                 int all_zero = 1;

  156.                 for (w2 = w + 1; w2 < c->num_windows; w2++) {

  157.                     if (c->grouping[w2])

  158.                         break;

  159.                     if (c->band_type[w2 * c->num_bands + b]) {

  160.                         all_zero = 0;

  161.                         break;

  162.                     }

  163.                 }

  164.                 if (all_zero) {

  165.                     c->band_scales[band_off++] = 0;

  166.                     continue;

  167.                 }

  168.             }

  169.             if (first) {

  170.                 scale = bitstream_read(bc, 7);

  171.                 first = 0;

  172.             } else {

  173.                 scale += bitstream_read_vlc(bc, c->scale_diff.table, 9, 3) - 60;

  174.             }

  175.             if (scale < 0 || scale > 127) {

  176.                 av_log(c->avctx, AV_LOG_ERROR, "Invalid scale value %d\n",

  177.                        scale);

  178.                 return AVERROR_INVALIDDATA;

  179.             }

  180.             c->band_scales[band_off++] = c->scale_tab[scale];

  181.         }

  182.     }

  183. 

  184.     return 0;

  185. }

  186. 

  187. static inline float on2avc_scale(int v, float scale)

  188. {

  189.     return v * sqrtf(abs(v)) * scale;

  190. }

  191. 

  192. // spectral data is coded completely differently - there are no unsigned codebooks

  193. static int on2avc_decode_quads(On2AVCContext *c, BitstreamContext *bc, float *dst,

  194.                                int dst_size, int type, float band_scale)

  195. {

  196.     int i, j, val, val1;

  197. 

  198.     for (i = 0; i < dst_size; i += 4) {

  199.         val = bitstream_read_vlc(bc, c->cb_vlc[type].table, 9, 3);

  200. 

  201.         for (j = 0; j < 4; j++) {

  202.             val1 = sign_extend((val >> (12 - j * 4)) & 0xF, 4);

  203.             *dst++ = on2avc_scale(val1, band_scale);

  204.         }

  205.     }

  206. 

  207.     return 0;

  208. }

  209. 

  210. static inline int get_egolomb(BitstreamContext *bc)

  211. {

  212.     int v = 4;

  213. 

  214.     while (bitstream_read_bit(bc)) {

  215.         v++;

  216.         if (v > 30) {

  217.             av_log(NULL, AV_LOG_WARNING, "Too large golomb code in get_egolomb.\n");

  218.             v = 30;

  219.             break;

  220.         }

  221.     }

  222. 

  223.     return (1 << v) + bitstream_read(bc, v);

  224. }

  225. 

  226. static int on2avc_decode_pairs(On2AVCContext *c, BitstreamContext *bc, float *dst,

  227.                                int dst_size, int type, float band_scale)

  228. {

  229.     int i, val, val1, val2, sign;

  230. 

  231.     for (i = 0; i < dst_size; i += 2) {

  232.         val = bitstream_read_vlc(bc, c->cb_vlc[type].table, 9, 3);

  233. 

  234.         val1 = sign_extend(val >> 8,   8);

  235.         val2 = sign_extend(val & 0xFF, 8);

  236.         if (type == ON2AVC_ESC_CB) {

  237.             if (val1 <= -16 || val1 >= 16) {

  238.                 sign = 1 - (val1 < 0) * 2;

  239.                 val1 = sign * get_egolomb(bc);

  240.             }

  241.             if (val2 <= -16 || val2 >= 16) {

  242.                 sign = 1 - (val2 < 0) * 2;

  243.                 val2 = sign * get_egolomb(bc);

  244.             }

  245.         }

  246. 

  247.         *dst++ = on2avc_scale(val1, band_scale);

  248.         *dst++ = on2avc_scale(val2, band_scale);

  249.     }

  250. 

  251.     return 0;

  252. }

  253. 

  254. static int on2avc_read_channel_data(On2AVCContext *c, BitstreamContext *bc, int ch)

  255. {

  256.     int ret;

  257.     int w, b, band_idx;

  258.     float *coeff_ptr;

  259. 

  260.     if ((ret = on2avc_decode_band_types(c, bc)) < 0)

  261.         return ret;

  262.     if ((ret = on2avc_decode_band_scales(c, bc)) < 0)

  263.         return ret;

  264. 

  265.     coeff_ptr = c->coeffs[ch];

  266.     band_idx  = 0;

  267.     memset(coeff_ptr, 0, ON2AVC_SUBFRAME_SIZE * sizeof(*coeff_ptr));

  268.     for (w = 0; w < c->num_windows; w++) {

  269.         for (b = 0; b < c->num_bands; b++) {

  270.             int band_size = c->band_start[b + 1] - c->band_start[b];

  271.             int band_type = c->band_type[band_idx + b];

  272. 

  273.             if (!band_type) {

  274.                 coeff_ptr += band_size;

  275.                 continue;

  276.             }

  277.             if (band_type < 9)

  278.                 on2avc_decode_quads(c, bc, coeff_ptr, band_size, band_type,

  279.                                     c->band_scales[band_idx + b]);

  280.             else

  281.                 on2avc_decode_pairs(c, bc, coeff_ptr, band_size, band_type,

  282.                                     c->band_scales[band_idx + b]);

  283.             coeff_ptr += band_size;

  284.         }

  285.         band_idx += c->num_bands;

  286.     }

  287. 

  288.     return 0;

  289. }

  290. 

  291. static int on2avc_apply_ms(On2AVCContext *c)

  292. {

  293.     int w, b, i;

  294.     int band_off = 0;

  295.     float *ch0 = c->coeffs[0];

  296.     float *ch1 = c->coeffs[1];

  297. 

  298.     for (w = 0; w < c->num_windows; w++) {

  299.         for (b = 0; b < c->num_bands; b++) {

  300.             if (c->ms_info[band_off + b]) {

  301.                 for (i = c->band_start[b]; i < c->band_start[b + 1]; i++) {

  302.                     float l = *ch0, r = *ch1;

  303.                     *ch0++ = l + r;

  304.                     *ch1++ = l - r;

  305.                 }

  306.             } else {

  307.                 ch0 += c->band_start[b + 1] - c->band_start[b];

  308.                 ch1 += c->band_start[b + 1] - c->band_start[b];

  309.             }

  310.         }

  311.         band_off += c->num_bands;

  312.     }

  313.     return 0;

  314. }

  315. 

  316. static void zero_head_and_tail(float *src, int len, int order0, int order1)

  317. {

  318.     memset(src,                0, sizeof(*src) * order0);

  319.     memset(src + len - order1, 0, sizeof(*src) * order1);

  320. }

  321. 

  322. static void pretwiddle(float *src, float *dst, int dst_len, int tab_step,

  323.                        int step, int order0, int order1, const double * const *tabs)

  324. {

  325.     float *src2, *out;

  326.     const double *tab;

  327.     int i, j;

  328. 

  329.     out = dst;

  330.     tab = tabs[0];

  331.     for (i = 0; i < tab_step; i++) {

  332.         double sum = 0;

  333.         for (j = 0; j < order0; j++)

  334.             sum += src[j] * tab[j * tab_step + i];

  335.         out[i] += sum;

  336.     }

  337. 

  338.     out = dst + dst_len - tab_step;

  339.     tab = tabs[order0];

  340.     src2 = src + (dst_len - tab_step) / step + 1 + order0;

  341.     for (i = 0; i < tab_step; i++) {

  342.         double sum = 0;

  343.         for (j = 0; j < order1; j++)

  344.             sum += src2[j] * tab[j * tab_step + i];

  345.         out[i] += sum;

  346.     }

  347. }

  348. 

  349. static void twiddle(float *src1, float *src2, int src2_len,

  350.                     const double *tab, int tab_len, int step,

  351.                     int order0, int order1, const double * const *tabs)

  352. {

  353.     int steps;

  354.     int mask;

  355.     int i, j;

  356. 

  357.     steps = (src2_len - tab_len) / step + 1;

  358.     pretwiddle(src1, src2, src2_len, tab_len, step, order0, order1, tabs);

  359.     mask = tab_len - 1;

  360. 

  361.     for (i = 0; i < steps; i++) {

  362.         float in0 = src1[order0 + i];

  363.         int   pos = (src2_len - 1) & mask;

  364. 

  365.         if (pos < tab_len) {

  366.             const double *t = tab;

  367.             for (j = pos; j >= 0; j--)

  368.                 src2[j] += in0 * *t++;

  369.             for (j = 0; j < tab_len - pos - 1; j++)

  370.                 src2[src2_len - j - 1] += in0 * tab[pos + 1 + j];

  371.         } else {

  372.             for (j = 0; j < tab_len; j++)

  373.                 src2[pos - j] += in0 * tab[j];

  374.         }

  375.         mask = pos + step;

  376.     }

  377. }

  378. 

  379. #define CMUL1_R(s, t, is, it) \

  380.     s[is + 0] * t[it + 0] - s[is + 1] * t[it + 1]

  381. #define CMUL1_I(s, t, is, it) \

  382.     s[is + 0] * t[it + 1] + s[is + 1] * t[it + 0]

  383. #define CMUL2_R(s, t, is, it) \

  384.     s[is + 0] * t[it + 0] + s[is + 1] * t[it + 1]

  385. #define CMUL2_I(s, t, is, it) \

  386.     s[is + 0] * t[it + 1] - s[is + 1] * t[it + 0]

  387. 

  388. #define CMUL0(dst, id, s0, s1, s2, s3, t0, t1, t2, t3, is, it)         \

  389.     dst[id]     = s0[is] * t0[it]     + s1[is] * t1[it]                \

  390.                 + s2[is] * t2[it]     + s3[is] * t3[it];               \

  391.     dst[id + 1] = s0[is] * t0[it + 1] + s1[is] * t1[it + 1]            \

  392.                 + s2[is] * t2[it + 1] + s3[is] * t3[it + 1];

  393. 

  394. #define CMUL1(dst, s0, s1, s2, s3, t0, t1, t2, t3, is, it)             \

  395.     *dst++ = CMUL1_R(s0, t0, is, it)                                   \

  396.            + CMUL1_R(s1, t1, is, it)                                   \

  397.            + CMUL1_R(s2, t2, is, it)                                   \

  398.            + CMUL1_R(s3, t3, is, it);                                  \

  399.     *dst++ = CMUL1_I(s0, t0, is, it)                                   \

  400.            + CMUL1_I(s1, t1, is, it)                                   \

  401.            + CMUL1_I(s2, t2, is, it)                                   \

  402.            + CMUL1_I(s3, t3, is, it);

  403. 

  404. #define CMUL2(dst, s0, s1, s2, s3, t0, t1, t2, t3, is, it)             \

  405.     *dst++ = CMUL2_R(s0, t0, is, it)                                   \

  406.            + CMUL2_R(s1, t1, is, it)                                   \

  407.            + CMUL2_R(s2, t2, is, it)                                   \

  408.            + CMUL2_R(s3, t3, is, it);                                  \

  409.     *dst++ = CMUL2_I(s0, t0, is, it)                                   \

  410.            + CMUL2_I(s1, t1, is, it)                                   \

  411.            + CMUL2_I(s2, t2, is, it)                                   \

  412.            + CMUL2_I(s3, t3, is, it);

  413. 

  414. static void combine_fft(float *s0, float *s1, float *s2, float *s3, float *dst,

  415.                         const float *t0, const float *t1,

  416.                         const float *t2, const float *t3, int len, int step)

  417. {

  418.     const float *h0, *h1, *h2, *h3;

  419.     float *d1, *d2;

  420.     int tmp, half;

  421.     int len2 = len >> 1, len4 = len >> 2;

  422.     int hoff;

  423.     int i, j, k;

  424. 

  425.     tmp = step;

  426.     for (half = len2; tmp > 1; half <<= 1, tmp >>= 1);

  427. 

  428.     h0 = t0 + half;

  429.     h1 = t1 + half;

  430.     h2 = t2 + half;

  431.     h3 = t3 + half;

  432. 

  433.     CMUL0(dst, 0, s0, s1, s2, s3, t0, t1, t2, t3, 0, 0);

  434. 

  435.     hoff = 2 * step * (len4 >> 1);

  436. 

  437.     j = 2;

  438.     k = 2 * step;

  439.     d1 = dst + 2;

  440.     d2 = dst + 2 + (len >> 1);

  441.     for (i = 0; i < (len4 - 1) >> 1; i++) {

  442.         CMUL1(d1, s0, s1, s2, s3, t0, t1, t2, t3, j, k);

  443.         CMUL1(d2, s0, s1, s2, s3, h0, h1, h2, h3, j, k);

  444.         j += 2;

  445.         k += 2 * step;

  446.     }

  447.     CMUL0(dst, len4,        s0, s1, s2, s3, t0, t1, t2, t3, 1, hoff);

  448.     CMUL0(dst, len4 + len2, s0, s1, s2, s3, h0, h1, h2, h3, 1, hoff);

  449. 

  450.     j = len4;

  451.     k = hoff + 2 * step * len4;

  452.     d1 = dst + len4 + 2;

  453.     d2 = dst + len4 + 2 + len2;

  454.     for (i = 0; i < (len4 - 2) >> 1; i++) {

  455.         CMUL2(d1, s0, s1, s2, s3, t0, t1, t2, t3, j, k);

  456.         CMUL2(d2, s0, s1, s2, s3, h0, h1, h2, h3, j, k);

  457.         j -= 2;

  458.         k += 2 * step;

  459.     }

  460.     CMUL0(dst, len2 + 4, s0, s1, s2, s3, t0, t1, t2, t3, 0, k);

  461. }

  462. 

  463. static void wtf_end_512(On2AVCContext *c, float *out, float *src,

  464.                         float *tmp0, float *tmp1)

  465. {

  466.     memcpy(src,        tmp0,      384 * sizeof(*tmp0));

  467.     memcpy(tmp0 + 384, src + 384, 128 * sizeof(*tmp0));

  468. 

  469.     zero_head_and_tail(src,       128, 16, 4);

  470.     zero_head_and_tail(src + 128, 128, 16, 4);

  471.     zero_head_and_tail(src + 256, 128, 13, 7);

  472.     zero_head_and_tail(src + 384, 128, 15, 5);

  473. 

  474.     c->fft128.fft_permute(&c->fft128, (FFTComplex*)src);

  475.     c->fft128.fft_permute(&c->fft128, (FFTComplex*)(src + 128));

  476.     c->fft128.fft_permute(&c->fft128, (FFTComplex*)(src + 256));

  477.     c->fft128.fft_permute(&c->fft128, (FFTComplex*)(src + 384));

  478.     c->fft128.fft_calc(&c->fft128, (FFTComplex*)src);

  479.     c->fft128.fft_calc(&c->fft128, (FFTComplex*)(src + 128));

  480.     c->fft128.fft_calc(&c->fft128, (FFTComplex*)(src + 256));

  481.     c->fft128.fft_calc(&c->fft128, (FFTComplex*)(src + 384));

  482.     combine_fft(src, src + 128, src + 256, src + 384, tmp1,

  483.                 ff_on2avc_ctab_1, ff_on2avc_ctab_2,

  484.                 ff_on2avc_ctab_3, ff_on2avc_ctab_4, 512, 2);

  485.     c->fft512.fft_permute(&c->fft512, (FFTComplex*)tmp1);

  486.     c->fft512.fft_calc(&c->fft512, (FFTComplex*)tmp1);

  487. 

  488.     pretwiddle(&tmp0[  0], tmp1, 512, 84, 4, 16, 4, ff_on2avc_tabs_20_84_1);

  489.     pretwiddle(&tmp0[128], tmp1, 512, 84, 4, 16, 4, ff_on2avc_tabs_20_84_2);

  490.     pretwiddle(&tmp0[256], tmp1, 512, 84, 4, 13, 7, ff_on2avc_tabs_20_84_3);

  491.     pretwiddle(&tmp0[384], tmp1, 512, 84, 4, 15, 5, ff_on2avc_tabs_20_84_4);

  492. 

  493.     memcpy(src, tmp1, 512 * sizeof(float));

  494. }

  495. 

  496. static void wtf_end_1024(On2AVCContext *c, float *out, float *src,

  497.                          float *tmp0, float *tmp1)

  498. {

  499.     memcpy(src,        tmp0,      768 * sizeof(*tmp0));

  500.     memcpy(tmp0 + 768, src + 768, 256 * sizeof(*tmp0));

  501. 

  502.     zero_head_and_tail(src,       256, 16, 4);

  503.     zero_head_and_tail(src + 256, 256, 16, 4);

  504.     zero_head_and_tail(src + 512, 256, 13, 7);

  505.     zero_head_and_tail(src + 768, 256, 15, 5);

  506. 

  507.     c->fft256.fft_permute(&c->fft256, (FFTComplex*)src);

  508.     c->fft256.fft_permute(&c->fft256, (FFTComplex*)(src + 256));

  509.     c->fft256.fft_permute(&c->fft256, (FFTComplex*)(src + 512));

  510.     c->fft256.fft_permute(&c->fft256, (FFTComplex*)(src + 768));

  511.     c->fft256.fft_calc(&c->fft256, (FFTComplex*)src);

  512.     c->fft256.fft_calc(&c->fft256, (FFTComplex*)(src + 256));

  513.     c->fft256.fft_calc(&c->fft256, (FFTComplex*)(src + 512));

  514.     c->fft256.fft_calc(&c->fft256, (FFTComplex*)(src + 768));

  515.     combine_fft(src, src + 256, src + 512, src + 768, tmp1,

  516.                 ff_on2avc_ctab_1, ff_on2avc_ctab_2,

  517.                 ff_on2avc_ctab_3, ff_on2avc_ctab_4, 1024, 1);

  518.     c->fft1024.fft_permute(&c->fft1024, (FFTComplex*)tmp1);

  519.     c->fft1024.fft_calc(&c->fft1024, (FFTComplex*)tmp1);

  520. 

  521.     pretwiddle(&tmp0[  0], tmp1, 1024, 84, 4, 16, 4, ff_on2avc_tabs_20_84_1);

  522.     pretwiddle(&tmp0[256], tmp1, 1024, 84, 4, 16, 4, ff_on2avc_tabs_20_84_2);

  523.     pretwiddle(&tmp0[512], tmp1, 1024, 84, 4, 13, 7, ff_on2avc_tabs_20_84_3);

  524.     pretwiddle(&tmp0[768], tmp1, 1024, 84, 4, 15, 5, ff_on2avc_tabs_20_84_4);

  525. 

  526.     memcpy(src, tmp1, 1024 * sizeof(float));

  527. }

  528. 

  529. static void wtf_40(On2AVCContext *c, float *out, float *src, int size)

  530. {

  531.     float *tmp0 = c->temp, *tmp1 = c->temp + 1024;

  532. 

  533.     memset(tmp0, 0, sizeof(*tmp0) * 1024);

  534.     memset(tmp1, 0, sizeof(*tmp1) * 1024);

  535. 

  536.     if (size == 512) {

  537.         twiddle(src,       &tmp0[  0], 16, ff_on2avc_tab_10_1, 10, 2, 1, 3, ff_on2avc_tabs_4_10_1);

  538.         twiddle(src +   8, &tmp0[  0], 16, ff_on2avc_tab_10_2, 10, 2, 3, 1, ff_on2avc_tabs_4_10_2);

  539.         twiddle(src +  16, &tmp0[ 16], 16, ff_on2avc_tab_10_2, 10, 2, 3, 1, ff_on2avc_tabs_4_10_2);

  540.         twiddle(src +  24, &tmp0[ 16], 16, ff_on2avc_tab_10_1, 10, 2, 1, 3, ff_on2avc_tabs_4_10_1);

  541.         twiddle(src +  32, &tmp0[ 32], 16, ff_on2avc_tab_10_1, 10, 2, 1, 3, ff_on2avc_tabs_4_10_1);

  542.         twiddle(src +  40, &tmp0[ 32], 16, ff_on2avc_tab_10_2, 10, 2, 3, 1, ff_on2avc_tabs_4_10_2);

  543.         twiddle(src +  48, &tmp0[ 48], 16, ff_on2avc_tab_10_2, 10, 2, 3, 1, ff_on2avc_tabs_4_10_2);

  544.         twiddle(src +  56, &tmp0[ 48], 16, ff_on2avc_tab_10_1, 10, 2, 1, 3, ff_on2avc_tabs_4_10_1);

  545.         twiddle(&tmp0[ 0], &tmp1[  0], 32, ff_on2avc_tab_20_1, 20, 2, 5, 4, ff_on2avc_tabs_9_20_1);

  546.         twiddle(&tmp0[16], &tmp1[  0], 32, ff_on2avc_tab_20_2, 20, 2, 4, 5, ff_on2avc_tabs_9_20_2);

  547.         twiddle(&tmp0[32], &tmp1[ 32], 32, ff_on2avc_tab_20_2, 20, 2, 4, 5, ff_on2avc_tabs_9_20_2);

  548.         twiddle(&tmp0[48], &tmp1[ 32], 32, ff_on2avc_tab_20_1, 20, 2, 5, 4, ff_on2avc_tabs_9_20_1);

  549.         twiddle(src +  64, &tmp1[ 64], 32, ff_on2avc_tab_20_1, 20, 2, 5, 4, ff_on2avc_tabs_9_20_1);

  550.         twiddle(src +  80, &tmp1[ 64], 32, ff_on2avc_tab_20_2, 20, 2, 4, 5, ff_on2avc_tabs_9_20_2);

  551.         twiddle(src +  96, &tmp1[ 96], 32, ff_on2avc_tab_20_2, 20, 2, 4, 5, ff_on2avc_tabs_9_20_2);

  552.         twiddle(src + 112, &tmp1[ 96], 32, ff_on2avc_tab_20_1, 20, 2, 5, 4, ff_on2avc_tabs_9_20_1);

  553.         twiddle(src + 128, &tmp1[128], 32, ff_on2avc_tab_20_1, 20, 2, 5, 4, ff_on2avc_tabs_9_20_1);

  554.         twiddle(src + 144, &tmp1[128], 32, ff_on2avc_tab_20_2, 20, 2, 4, 5, ff_on2avc_tabs_9_20_2);

  555.         twiddle(src + 160, &tmp1[160], 32, ff_on2avc_tab_20_2, 20, 2, 4, 5, ff_on2avc_tabs_9_20_2);

  556.         twiddle(src + 176, &tmp1[160], 32, ff_on2avc_tab_20_1, 20, 2, 5, 4, ff_on2avc_tabs_9_20_1);

  557. 

  558.         memset(tmp0, 0, 64 * sizeof(*tmp0));

  559. 

  560.         twiddle(&tmp1[  0], &tmp0[  0], 128, ff_on2avc_tab_84_1, 84, 4, 16, 4, ff_on2avc_tabs_20_84_1);

  561.         twiddle(&tmp1[ 32], &tmp0[  0], 128, ff_on2avc_tab_84_2, 84, 4, 16, 4, ff_on2avc_tabs_20_84_2);

  562.         twiddle(&tmp1[ 64], &tmp0[  0], 128, ff_on2avc_tab_84_3, 84, 4, 13, 7, ff_on2avc_tabs_20_84_3);

  563.         twiddle(&tmp1[ 96], &tmp0[  0], 128, ff_on2avc_tab_84_4, 84, 4, 15, 5, ff_on2avc_tabs_20_84_4);

  564.         twiddle(&tmp1[128], &tmp0[128], 128, ff_on2avc_tab_84_4, 84, 4, 15, 5, ff_on2avc_tabs_20_84_4);

  565.         twiddle(&tmp1[160], &tmp0[128], 128, ff_on2avc_tab_84_3, 84, 4, 13, 7, ff_on2avc_tabs_20_84_3);

  566.         twiddle(src + 192,  &tmp0[128], 128, ff_on2avc_tab_84_2, 84, 4, 16, 4, ff_on2avc_tabs_20_84_2);

  567.         twiddle(src + 224,  &tmp0[128], 128, ff_on2avc_tab_84_1, 84, 4, 16, 4, ff_on2avc_tabs_20_84_1);

  568.         twiddle(src + 256,  &tmp0[256], 128, ff_on2avc_tab_84_1, 84, 4, 16, 4, ff_on2avc_tabs_20_84_1);

  569.         twiddle(src + 288,  &tmp0[256], 128, ff_on2avc_tab_84_2, 84, 4, 16, 4, ff_on2avc_tabs_20_84_2);

  570.         twiddle(src + 320,  &tmp0[256], 128, ff_on2avc_tab_84_3, 84, 4, 13, 7, ff_on2avc_tabs_20_84_3);

  571.         twiddle(src + 352,  &tmp0[256], 128, ff_on2avc_tab_84_4, 84, 4, 15, 5, ff_on2avc_tabs_20_84_4);

  572. 

  573.         wtf_end_512(c, out, src, tmp0, tmp1);

  574.     } else {

  575.         twiddle(src,       &tmp0[  0], 32, ff_on2avc_tab_20_1, 20, 2,  5,  4, ff_on2avc_tabs_9_20_1);

  576.         twiddle(src +  16, &tmp0[  0], 32, ff_on2avc_tab_20_2, 20, 2,  4,  5, ff_on2avc_tabs_9_20_2);

  577.         twiddle(src +  32, &tmp0[ 32], 32, ff_on2avc_tab_20_2, 20, 2,  4,  5, ff_on2avc_tabs_9_20_2);

  578.         twiddle(src +  48, &tmp0[ 32], 32, ff_on2avc_tab_20_1, 20, 2,  5,  4, ff_on2avc_tabs_9_20_1);

  579.         twiddle(src +  64, &tmp0[ 64], 32, ff_on2avc_tab_20_1, 20, 2,  5,  4, ff_on2avc_tabs_9_20_1);

  580.         twiddle(src +  80, &tmp0[ 64], 32, ff_on2avc_tab_20_2, 20, 2,  4,  5, ff_on2avc_tabs_9_20_2);

  581.         twiddle(src +  96, &tmp0[ 96], 32, ff_on2avc_tab_20_2, 20, 2,  4,  5, ff_on2avc_tabs_9_20_2);

  582.         twiddle(src + 112, &tmp0[ 96], 32, ff_on2avc_tab_20_1, 20, 2,  5,  4, ff_on2avc_tabs_9_20_1);

  583.         twiddle(&tmp0[ 0], &tmp1[  0], 64, ff_on2avc_tab_40_1, 40, 2, 11,  8, ff_on2avc_tabs_19_40_1);

  584.         twiddle(&tmp0[32], &tmp1[  0], 64, ff_on2avc_tab_40_2, 40, 2,  8, 11, ff_on2avc_tabs_19_40_2);

  585.         twiddle(&tmp0[64], &tmp1[ 64], 64, ff_on2avc_tab_40_2, 40, 2,  8, 11, ff_on2avc_tabs_19_40_2);

  586.         twiddle(&tmp0[96], &tmp1[ 64], 64, ff_on2avc_tab_40_1, 40, 2, 11,  8, ff_on2avc_tabs_19_40_1);

  587.         twiddle(src + 128, &tmp1[128], 64, ff_on2avc_tab_40_1, 40, 2, 11,  8, ff_on2avc_tabs_19_40_1);

  588.         twiddle(src + 160, &tmp1[128], 64, ff_on2avc_tab_40_2, 40, 2,  8, 11, ff_on2avc_tabs_19_40_2);

  589.         twiddle(src + 192, &tmp1[192], 64, ff_on2avc_tab_40_2, 40, 2,  8, 11, ff_on2avc_tabs_19_40_2);

  590.         twiddle(src + 224, &tmp1[192], 64, ff_on2avc_tab_40_1, 40, 2, 11,  8, ff_on2avc_tabs_19_40_1);

  591.         twiddle(src + 256, &tmp1[256], 64, ff_on2avc_tab_40_1, 40, 2, 11,  8, ff_on2avc_tabs_19_40_1);

  592.         twiddle(src + 288, &tmp1[256], 64, ff_on2avc_tab_40_2, 40, 2,  8, 11, ff_on2avc_tabs_19_40_2);

  593.         twiddle(src + 320, &tmp1[320], 64, ff_on2avc_tab_40_2, 40, 2,  8, 11, ff_on2avc_tabs_19_40_2);

  594.         twiddle(src + 352, &tmp1[320], 64, ff_on2avc_tab_40_1, 40, 2, 11,  8, ff_on2avc_tabs_19_40_1);

  595. 

  596.         memset(tmp0, 0, 128 * sizeof(*tmp0));

  597. 

  598.         twiddle(&tmp1[  0], &tmp0[  0], 256, ff_on2avc_tab_84_1, 84, 4, 16, 4, ff_on2avc_tabs_20_84_1);

  599.         twiddle(&tmp1[ 64], &tmp0[  0], 256, ff_on2avc_tab_84_2, 84, 4, 16, 4, ff_on2avc_tabs_20_84_2);

  600.         twiddle(&tmp1[128], &tmp0[  0], 256, ff_on2avc_tab_84_3, 84, 4, 13, 7, ff_on2avc_tabs_20_84_3);

  601.         twiddle(&tmp1[192], &tmp0[  0], 256, ff_on2avc_tab_84_4, 84, 4, 15, 5, ff_on2avc_tabs_20_84_4);

  602.         twiddle(&tmp1[256], &tmp0[256], 256, ff_on2avc_tab_84_4, 84, 4, 15, 5, ff_on2avc_tabs_20_84_4);

  603.         twiddle(&tmp1[320], &tmp0[256], 256, ff_on2avc_tab_84_3, 84, 4, 13, 7, ff_on2avc_tabs_20_84_3);

  604.         twiddle(src + 384,  &tmp0[256], 256, ff_on2avc_tab_84_2, 84, 4, 16, 4, ff_on2avc_tabs_20_84_2);

  605.         twiddle(src + 448,  &tmp0[256], 256, ff_on2avc_tab_84_1, 84, 4, 16, 4, ff_on2avc_tabs_20_84_1);

  606.         twiddle(src + 512,  &tmp0[512], 256, ff_on2avc_tab_84_1, 84, 4, 16, 4, ff_on2avc_tabs_20_84_1);

  607.         twiddle(src + 576,  &tmp0[512], 256, ff_on2avc_tab_84_2, 84, 4, 16, 4, ff_on2avc_tabs_20_84_2);

  608.         twiddle(src + 640,  &tmp0[512], 256, ff_on2avc_tab_84_3, 84, 4, 13, 7, ff_on2avc_tabs_20_84_3);

  609.         twiddle(src + 704,  &tmp0[512], 256, ff_on2avc_tab_84_4, 84, 4, 15, 5, ff_on2avc_tabs_20_84_4);

  610. 

  611.         wtf_end_1024(c, out, src, tmp0, tmp1);

  612.     }

  613. }

  614. 

  615. static void wtf_44(On2AVCContext *c, float *out, float *src, int size)

  616. {

  617.     float *tmp0 = c->temp, *tmp1 = c->temp + 1024;

  618. 

  619.     memset(tmp0, 0, sizeof(*tmp0) * 1024);

  620.     memset(tmp1, 0, sizeof(*tmp1) * 1024);

  621. 

  622.     if (size == 512) {

  623.         twiddle(src,       &tmp0[ 0], 16, ff_on2avc_tab_10_1, 10, 2, 1, 3, ff_on2avc_tabs_4_10_1);

  624.         twiddle(src +   8, &tmp0[ 0], 16, ff_on2avc_tab_10_2, 10, 2, 3, 1, ff_on2avc_tabs_4_10_2);

  625.         twiddle(src +  16, &tmp0[16], 16, ff_on2avc_tab_10_2, 10, 2, 3, 1, ff_on2avc_tabs_4_10_2);

  626.         twiddle(src +  24, &tmp0[16], 16, ff_on2avc_tab_10_1, 10, 2, 1, 3, ff_on2avc_tabs_4_10_1);

  627.         twiddle(src +  32, &tmp0[32], 16, ff_on2avc_tab_10_1, 10, 2, 1, 3, ff_on2avc_tabs_4_10_1);

  628.         twiddle(src +  40, &tmp0[32], 16, ff_on2avc_tab_10_2, 10, 2, 3, 1, ff_on2avc_tabs_4_10_2);

  629.         twiddle(src +  48, &tmp0[48], 16, ff_on2avc_tab_10_2, 10, 2, 3, 1, ff_on2avc_tabs_4_10_2);

  630.         twiddle(src +  56, &tmp0[48], 16, ff_on2avc_tab_10_1, 10, 2, 1, 3, ff_on2avc_tabs_4_10_1);

  631.         twiddle(&tmp0[ 0], &tmp1[ 0], 32, ff_on2avc_tab_20_1, 20, 2, 5, 4, ff_on2avc_tabs_9_20_1);

  632.         twiddle(&tmp0[16], &tmp1[ 0], 32, ff_on2avc_tab_20_2, 20, 2, 4, 5, ff_on2avc_tabs_9_20_2);

  633.         twiddle(&tmp0[32], &tmp1[32], 32, ff_on2avc_tab_20_2, 20, 2, 4, 5, ff_on2avc_tabs_9_20_2);

  634.         twiddle(&tmp0[48], &tmp1[32], 32, ff_on2avc_tab_20_1, 20, 2, 5, 4, ff_on2avc_tabs_9_20_1);

  635.         twiddle(src +  64, &tmp1[64], 32, ff_on2avc_tab_20_1, 20, 2, 5, 4, ff_on2avc_tabs_9_20_1);

  636.         twiddle(src +  80, &tmp1[64], 32, ff_on2avc_tab_20_2, 20, 2, 4, 5, ff_on2avc_tabs_9_20_2);

  637.         twiddle(src +  96, &tmp1[96], 32, ff_on2avc_tab_20_2, 20, 2, 4, 5, ff_on2avc_tabs_9_20_2);

  638.         twiddle(src + 112, &tmp1[96], 32, ff_on2avc_tab_20_1, 20, 2, 5, 4, ff_on2avc_tabs_9_20_1);

  639. 

  640.         memset(tmp0, 0, 64 * sizeof(*tmp0));

  641. 

  642.         twiddle(&tmp1[ 0], &tmp0[  0], 128, ff_on2avc_tab_84_1, 84, 4, 16,  4, ff_on2avc_tabs_20_84_1);

  643.         twiddle(&tmp1[32], &tmp0[  0], 128, ff_on2avc_tab_84_2, 84, 4, 16,  4, ff_on2avc_tabs_20_84_2);

  644.         twiddle(&tmp1[64], &tmp0[  0], 128, ff_on2avc_tab_84_3, 84, 4, 13,  7, ff_on2avc_tabs_20_84_3);

  645.         twiddle(&tmp1[96], &tmp0[  0], 128, ff_on2avc_tab_84_4, 84, 4, 15,  5, ff_on2avc_tabs_20_84_4);

  646.         twiddle(src + 128, &tmp0[128], 128, ff_on2avc_tab_84_4, 84, 4, 15,  5, ff_on2avc_tabs_20_84_4);

  647.         twiddle(src + 160, &tmp0[128], 128, ff_on2avc_tab_84_3, 84, 4, 13,  7, ff_on2avc_tabs_20_84_3);

  648.         twiddle(src + 192, &tmp0[128], 128, ff_on2avc_tab_84_2, 84, 4, 16,  4, ff_on2avc_tabs_20_84_2);

  649.         twiddle(src + 224, &tmp0[128], 128, ff_on2avc_tab_84_1, 84, 4, 16,  4, ff_on2avc_tabs_20_84_1);

  650.         twiddle(src + 256, &tmp0[256], 128, ff_on2avc_tab_40_1, 40, 2, 11,  8, ff_on2avc_tabs_19_40_1);

  651.         twiddle(src + 320, &tmp0[256], 128, ff_on2avc_tab_40_2, 40, 2,  8, 11, ff_on2avc_tabs_19_40_2);

  652. 

  653.         wtf_end_512(c, out, src, tmp0, tmp1);

  654.     } else {

  655.         twiddle(src,       &tmp0[  0], 32, ff_on2avc_tab_20_1, 20, 2,  5,  4, ff_on2avc_tabs_9_20_1);

  656.         twiddle(src +  16, &tmp0[  0], 32, ff_on2avc_tab_20_2, 20, 2,  4,  5, ff_on2avc_tabs_9_20_2);

  657.         twiddle(src +  32, &tmp0[ 32], 32, ff_on2avc_tab_20_2, 20, 2,  4,  5, ff_on2avc_tabs_9_20_2);

  658.         twiddle(src +  48, &tmp0[ 32], 32, ff_on2avc_tab_20_1, 20, 2,  5,  4, ff_on2avc_tabs_9_20_1);

  659.         twiddle(src +  64, &tmp0[ 64], 32, ff_on2avc_tab_20_1, 20, 2,  5,  4, ff_on2avc_tabs_9_20_1);

  660.         twiddle(src +  80, &tmp0[ 64], 32, ff_on2avc_tab_20_2, 20, 2,  4,  5, ff_on2avc_tabs_9_20_2);

  661.         twiddle(src +  96, &tmp0[ 96], 32, ff_on2avc_tab_20_2, 20, 2,  4,  5, ff_on2avc_tabs_9_20_2);

  662.         twiddle(src + 112, &tmp0[ 96], 32, ff_on2avc_tab_20_1, 20, 2,  5,  4, ff_on2avc_tabs_9_20_1);

  663.         twiddle(&tmp0[ 0], &tmp1[  0], 64, ff_on2avc_tab_40_1, 40, 2, 11,  8, ff_on2avc_tabs_19_40_1);

  664.         twiddle(&tmp0[32], &tmp1[  0], 64, ff_on2avc_tab_40_2, 40, 2,  8, 11, ff_on2avc_tabs_19_40_2);

  665.         twiddle(&tmp0[64], &tmp1[ 64], 64, ff_on2avc_tab_40_2, 40, 2,  8, 11, ff_on2avc_tabs_19_40_2);

  666.         twiddle(&tmp0[96], &tmp1[ 64], 64, ff_on2avc_tab_40_1, 40, 2, 11,  8, ff_on2avc_tabs_19_40_1);

  667.         twiddle(src + 128, &tmp1[128], 64, ff_on2avc_tab_40_1, 40, 2, 11,  8, ff_on2avc_tabs_19_40_1);

  668.         twiddle(src + 160, &tmp1[128], 64, ff_on2avc_tab_40_2, 40, 2,  8, 11, ff_on2avc_tabs_19_40_2);

  669.         twiddle(src + 192, &tmp1[192], 64, ff_on2avc_tab_40_2, 40, 2,  8, 11, ff_on2avc_tabs_19_40_2);

  670.         twiddle(src + 224, &tmp1[192], 64, ff_on2avc_tab_40_1, 40, 2, 11,  8, ff_on2avc_tabs_19_40_1);

  671. 

  672.         memset(tmp0, 0, 128 * sizeof(*tmp0));

  673. 

  674.         twiddle(&tmp1[  0], &tmp0[  0], 256, ff_on2avc_tab_84_1, 84, 4, 16,  4, ff_on2avc_tabs_20_84_1);

  675.         twiddle(&tmp1[ 64], &tmp0[  0], 256, ff_on2avc_tab_84_2, 84, 4, 16,  4, ff_on2avc_tabs_20_84_2);

  676.         twiddle(&tmp1[128], &tmp0[  0], 256, ff_on2avc_tab_84_3, 84, 4, 13,  7, ff_on2avc_tabs_20_84_3);

  677.         twiddle(&tmp1[192], &tmp0[  0], 256, ff_on2avc_tab_84_4, 84, 4, 15,  5, ff_on2avc_tabs_20_84_4);

  678.         twiddle(src + 256,  &tmp0[256], 256, ff_on2avc_tab_84_4, 84, 4, 15,  5, ff_on2avc_tabs_20_84_4);

  679.         twiddle(src + 320,  &tmp0[256], 256, ff_on2avc_tab_84_3, 84, 4, 13,  7, ff_on2avc_tabs_20_84_3);

  680.         twiddle(src + 384,  &tmp0[256], 256, ff_on2avc_tab_84_2, 84, 4, 16,  4, ff_on2avc_tabs_20_84_2);

  681.         twiddle(src + 448,  &tmp0[256], 256, ff_on2avc_tab_84_1, 84, 4, 16,  4, ff_on2avc_tabs_20_84_1);

  682.         twiddle(src + 512,  &tmp0[512], 256, ff_on2avc_tab_40_1, 40, 2, 11,  8, ff_on2avc_tabs_19_40_1);

  683.         twiddle(src + 640,  &tmp0[512], 256, ff_on2avc_tab_40_2, 40, 2,  8, 11, ff_on2avc_tabs_19_40_2);

  684. 

  685.         wtf_end_1024(c, out, src, tmp0, tmp1);

  686.     }

  687. }

  688. 

  689. static int on2avc_reconstruct_stereo(On2AVCContext *c, AVFrame *dst, int offset)

  690. {

  691.     int ch, i;

  692. 

  693.     for (ch = 0; ch < 2; ch++) {

  694.         float *out   = (float*)dst->extended_data[ch] + offset;

  695.         float *in    = c->coeffs[ch];

  696.         float *saved = c->delay[ch];

  697.         float *buf   = c->mdct_buf;

  698.         float *wout  = out + 448;

  699. 

  700.         switch (c->window_type) {

  701.         case WINDOW_TYPE_EXT7:

  702.             c->mdct.imdct_half(&c->mdct, buf, in);

  703.             break;

  704.         case WINDOW_TYPE_EXT4:

  705.             c->wtf(c, buf, in, 1024);

  706.             break;

  707.         case WINDOW_TYPE_EXT5:

  708.             c->wtf(c, buf, in, 512);

  709.             c->mdct.imdct_half(&c->mdct_half, buf + 512, in + 512);

  710.             for (i = 0; i < 256; i++) {

  711.                 FFSWAP(float, buf[i + 512], buf[1023 - i]);

  712.             }

  713.             break;

  714.         case WINDOW_TYPE_EXT6:

  715.             c->mdct.imdct_half(&c->mdct_half, buf, in);

  716.             for (i = 0; i < 256; i++) {

  717.                 FFSWAP(float, buf[i], buf[511 - i]);

  718.             }

  719.             c->wtf(c, buf + 512, in + 512, 512);

  720.             break;

  721.         }

  722. 

  723.         memcpy(out, saved, 448 * sizeof(float));

  724.         c->fdsp.vector_fmul_window(wout, saved + 448, buf, c->short_win, 64);

  725.         memcpy(wout + 128,  buf + 64,         448 * sizeof(float));

  726.         memcpy(saved,       buf + 512,        448 * sizeof(float));

  727.         memcpy(saved + 448, buf + 7*128 + 64,  64 * sizeof(float));

  728.     }

  729. 

  730.     return 0;

  731. }

  732. 

  733. // not borrowed from aacdec.c - the codec has original design after all

  734. static int on2avc_reconstruct_channel(On2AVCContext *c, int channel,

  735.                                       AVFrame *dst, int offset)

  736. {

  737.     int i;

  738.     float *out   = (float*)dst->extended_data[channel] + offset;

  739.     float *in    = c->coeffs[channel];

  740.     float *saved = c->delay[channel];

  741.     float *buf   = c->mdct_buf;

  742.     float *temp  = c->temp;

  743. 

  744.     switch (c->window_type) {

  745.     case WINDOW_TYPE_LONG_START:

  746.     case WINDOW_TYPE_LONG_STOP:

  747.     case WINDOW_TYPE_LONG:

  748.         c->mdct.imdct_half(&c->mdct, buf, in);

  749.         break;

  750.     case WINDOW_TYPE_8SHORT:

  751.         for (i = 0; i < ON2AVC_SUBFRAME_SIZE; i += ON2AVC_SUBFRAME_SIZE / 8)

  752.             c->mdct_small.imdct_half(&c->mdct_small, buf + i, in + i);

  753.         break;

  754.     }

  755. 

  756.     if ((c->prev_window_type == WINDOW_TYPE_LONG ||

  757.          c->prev_window_type == WINDOW_TYPE_LONG_STOP) &&

  758.         (c->window_type == WINDOW_TYPE_LONG ||

  759.          c->window_type == WINDOW_TYPE_LONG_START)) {

  760.         c->fdsp.vector_fmul_window(out, saved, buf, c->long_win, 512);

  761.     } else {

  762.         float *wout = out + 448;

  763.         memcpy(out, saved, 448 * sizeof(float));

  764. 

  765.         if (c->window_type == WINDOW_TYPE_8SHORT) {

  766.             c->fdsp.vector_fmul_window(wout + 0*128, saved + 448,      buf + 0*128, c->short_win, 64);

  767.             c->fdsp.vector_fmul_window(wout + 1*128, buf + 0*128 + 64, buf + 1*128, c->short_win, 64);

  768.             c->fdsp.vector_fmul_window(wout + 2*128, buf + 1*128 + 64, buf + 2*128, c->short_win, 64);

  769.             c->fdsp.vector_fmul_window(wout + 3*128, buf + 2*128 + 64, buf + 3*128, c->short_win, 64);

  770.             c->fdsp.vector_fmul_window(temp,         buf + 3*128 + 64, buf + 4*128, c->short_win, 64);

  771.             memcpy(wout + 4*128, temp, 64 * sizeof(float));

  772.         } else {

  773.             c->fdsp.vector_fmul_window(wout, saved + 448, buf, c->short_win, 64);

  774.             memcpy(wout + 128, buf + 64, 448 * sizeof(float));

  775.         }

  776.     }

  777. 

  778.     // buffer update

  779.     switch (c->window_type) {

  780.     case WINDOW_TYPE_8SHORT:

  781.         memcpy(saved,       temp + 64,         64 * sizeof(float));

  782.         c->fdsp.vector_fmul_window(saved + 64,  buf + 4*128 + 64, buf + 5*128, c->short_win, 64);

  783.         c->fdsp.vector_fmul_window(saved + 192, buf + 5*128 + 64, buf + 6*128, c->short_win, 64);

  784.         c->fdsp.vector_fmul_window(saved + 320, buf + 6*128 + 64, buf + 7*128, c->short_win, 64);

  785.         memcpy(saved + 448, buf + 7*128 + 64,  64 * sizeof(float));

  786.         break;

  787.     case WINDOW_TYPE_LONG_START:

  788.         memcpy(saved,       buf + 512,        448 * sizeof(float));

  789.         memcpy(saved + 448, buf + 7*128 + 64,  64 * sizeof(float));

  790.         break;

  791.     case WINDOW_TYPE_LONG_STOP:

  792.     case WINDOW_TYPE_LONG:

  793.         memcpy(saved,       buf + 512,        512 * sizeof(float));

  794.         break;

  795.     }

  796.     return 0;

  797. }

  798. 

  799. static int on2avc_decode_subframe(On2AVCContext *c, const uint8_t *buf,

  800.                                   int buf_size, AVFrame *dst, int offset)

  801. {

  802.     BitstreamContext bc;

  803.     int i, ret;

  804. 

  805.     bitstream_init8(&bc, buf, buf_size);

  806.     if (bitstream_read_bit(&bc)) {

  807.         av_log(c->avctx, AV_LOG_ERROR, "enh bit set\n");

  808.         return AVERROR_INVALIDDATA;

  809.     }

  810.     c->prev_window_type = c->window_type;

  811.     c->window_type      = bitstream_read(&bc, 3);

  812.     if (c->window_type >= WINDOW_TYPE_EXT4 && c->avctx->channels == 1) {

  813.         av_log(c->avctx, AV_LOG_ERROR, "stereo mode window for mono audio\n");

  814.         return AVERROR_INVALIDDATA;

  815.     }

  816. 

  817.     c->band_start  = c->modes[c->window_type].band_start;

  818.     c->num_windows = c->modes[c->window_type].num_windows;

  819.     c->num_bands   = c->modes[c->window_type].num_bands;

  820.     c->is_long     = (c->window_type != WINDOW_TYPE_8SHORT);

  821. 

  822.     c->grouping[0] = 1;

  823.     for (i = 1; i < c->num_windows; i++)

  824.         c->grouping[i] = !bitstream_read_bit(&bc);

  825. 

  826.     on2avc_read_ms_info(c, &bc);

  827.     for (i = 0; i < c->avctx->channels; i++)

  828.         if ((ret = on2avc_read_channel_data(c, &bc, i)) < 0)

  829.             return AVERROR_INVALIDDATA;

  830.     if (c->avctx->channels == 2 && c->ms_present)

  831.         on2avc_apply_ms(c);

  832.     if (c->window_type < WINDOW_TYPE_EXT4) {

  833.         for (i = 0; i < c->avctx->channels; i++)

  834.             on2avc_reconstruct_channel(c, i, dst, offset);

  835.     } else {

  836.         on2avc_reconstruct_stereo(c, dst, offset);

  837.     }

  838. 

  839.     return 0;

  840. }

  841. 

  842. static int on2avc_decode_frame(AVCodecContext * avctx, void *data,

  843.                                int *got_frame_ptr, AVPacket *avpkt)

  844. {

  845.     AVFrame *frame     = data;

  846.     const uint8_t *buf = avpkt->data;

  847.     int buf_size       = avpkt->size;

  848.     On2AVCContext *c   = avctx->priv_data;

  849.     GetByteContext gb;

  850.     int num_frames = 0, frame_size, audio_off;

  851.     int ret;

  852. 

  853.     if (c->is_av500) {

  854.         /* get output buffer */

  855.         frame->nb_samples = ON2AVC_SUBFRAME_SIZE;

  856.         if ((ret = ff_get_buffer(avctx, frame, 0)) < 0) {

  857.             av_log(avctx, AV_LOG_ERROR, "get_buffer() failed\n");

  858.             return ret;

  859.         }

  860. 

  861.         if ((ret = on2avc_decode_subframe(c, buf, buf_size, frame, 0)) < 0)

  862.             return ret;

  863.     } else {

  864.         bytestream2_init(&gb, buf, buf_size);

  865.         while (bytestream2_get_bytes_left(&gb) > 2) {

  866.             frame_size = bytestream2_get_le16(&gb);

  867.             if (!frame_size || frame_size > bytestream2_get_bytes_left(&gb)) {

  868.                 av_log(avctx, AV_LOG_ERROR, "Invalid subframe size %d\n",

  869.                        frame_size);

  870.                 return AVERROR_INVALIDDATA;

  871.             }

  872.             num_frames++;

  873.             bytestream2_skip(&gb, frame_size);

  874.         }

  875.         if (!num_frames) {

  876.             av_log(avctx, AV_LOG_ERROR, "No subframes present\n");

  877.             return AVERROR_INVALIDDATA;

  878.         }

  879. 

  880.         /* get output buffer */

  881.         frame->nb_samples = ON2AVC_SUBFRAME_SIZE * num_frames;

  882.         if ((ret = ff_get_buffer(avctx, frame, 0)) < 0) {

  883.             av_log(avctx, AV_LOG_ERROR, "get_buffer() failed\n");

  884.             return ret;

  885.         }

  886. 

  887.         audio_off = 0;

  888.         bytestream2_init(&gb, buf, buf_size);

  889.         while (bytestream2_get_bytes_left(&gb) > 2) {

  890.             frame_size = bytestream2_get_le16(&gb);

  891.             if ((ret = on2avc_decode_subframe(c, gb.buffer, frame_size,

  892.                                               frame, audio_off)) < 0)

  893.                 return ret;

  894.             audio_off += ON2AVC_SUBFRAME_SIZE;

  895.             bytestream2_skip(&gb, frame_size);

  896.         }

  897.     }

  898. 

  899.     *got_frame_ptr = 1;

  900. 

  901.     return buf_size;

  902. }

  903. 

  904. static av_cold void on2avc_free_vlcs(On2AVCContext *c)

  905. {

  906.     int i;

  907. 

  908.     ff_free_vlc(&c->scale_diff);

  909.     for (i = 1; i < 16; i++)

  910.         ff_free_vlc(&c->cb_vlc[i]);

  911. }

  912. 

  913. static av_cold int on2avc_decode_init(AVCodecContext *avctx)

  914. {

  915.     On2AVCContext *c = avctx->priv_data;

  916.     int i;

  917. 

  918.     c->avctx = avctx;

  919.     avctx->sample_fmt     = AV_SAMPLE_FMT_FLTP;

  920.     avctx->channel_layout = (avctx->channels == 2) ? AV_CH_LAYOUT_STEREO

  921.                                                    : AV_CH_LAYOUT_MONO;

  922. 

  923.     c->is_av500 = (avctx->codec_tag == 0x500);

  924.     if (c->is_av500 && avctx->channels == 2) {

  925.         av_log(avctx, AV_LOG_ERROR, "0x500 version should be mono\n");

  926.         return AVERROR_INVALIDDATA;

  927.     }

  928.     if (avctx->channels > 2) {

  929.         av_log(avctx, AV_LOG_ERROR, "Only 1 or 2 channels are supported.\n");

  930.         return AVERROR(EINVAL);

  931.     }

  932.     if (avctx->channels == 2)

  933.         av_log(avctx, AV_LOG_WARNING,

  934.                "Stereo mode support is not good, patch is welcome\n");

  935. 

  936.     for (i = 0; i < 20; i++)

  937.         c->scale_tab[i] = ceil(pow(10.0, i * 0.1) * 16) / 32;

  938.     for (; i < 128; i++)

  939.         c->scale_tab[i] = ceil(pow(10.0, i * 0.1) * 0.5);

  940. 

  941.     if (avctx->sample_rate < 32000 || avctx->channels == 1)

  942.         memcpy(c->long_win, ff_on2avc_window_long_24000,

  943.                1024 * sizeof(*c->long_win));

  944.     else

  945.         memcpy(c->long_win, ff_on2avc_window_long_32000,

  946.                1024 * sizeof(*c->long_win));

  947.     memcpy(c->short_win, ff_on2avc_window_short, 128 * sizeof(*c->short_win));

  948. 

  949.     c->modes = (avctx->sample_rate <= 40000) ? ff_on2avc_modes_40

  950.                                              : ff_on2avc_modes_44;

  951.     c->wtf   = (avctx->sample_rate <= 40000) ? wtf_40

  952.                                              : wtf_44;

  953. 

  954.     ff_mdct_init(&c->mdct,       11, 1, 1.0 / (32768.0 * 1024.0));

  955.     ff_mdct_init(&c->mdct_half,  10, 1, 1.0 / (32768.0 * 512.0));

  956.     ff_mdct_init(&c->mdct_small,  8, 1, 1.0 / (32768.0 * 128.0));

  957.     ff_fft_init(&c->fft128,  6, 0);

  958.     ff_fft_init(&c->fft256,  7, 0);

  959.     ff_fft_init(&c->fft512,  8, 1);

  960.     ff_fft_init(&c->fft1024, 9, 1);

  961.     avpriv_float_dsp_init(&c->fdsp, avctx->flags & AV_CODEC_FLAG_BITEXACT);

  962. 

  963.     if (init_vlc(&c->scale_diff, 9, ON2AVC_SCALE_DIFFS,

  964.                  ff_on2avc_scale_diff_bits,  1, 1,

  965.                  ff_on2avc_scale_diff_codes, 4, 4, 0)) {

  966.         av_log(avctx, AV_LOG_ERROR, "Cannot init VLC\n");

  967.         return AVERROR(ENOMEM);

  968.     }

  969.     for (i = 1; i < 9; i++) {

  970.         int idx = i - 1;

  971.         if (ff_init_vlc_sparse(&c->cb_vlc[i], 9, ff_on2avc_quad_cb_elems[idx],

  972.                                ff_on2avc_quad_cb_bits[idx],  1, 1,

  973.                                ff_on2avc_quad_cb_codes[idx], 4, 4,

  974.                                ff_on2avc_quad_cb_syms[idx],  2, 2, 0)) {

  975.             av_log(avctx, AV_LOG_ERROR, "Cannot init VLC\n");

  976.             on2avc_free_vlcs(c);

  977.             return AVERROR(ENOMEM);

  978.         }

  979.     }

  980.     for (i = 9; i < 16; i++) {

  981.         int idx = i - 9;

  982.         if (ff_init_vlc_sparse(&c->cb_vlc[i], 9, ff_on2avc_pair_cb_elems[idx],

  983.                                ff_on2avc_pair_cb_bits[idx],  1, 1,

  984.                                ff_on2avc_pair_cb_codes[idx], 2, 2,

  985.                                ff_on2avc_pair_cb_syms[idx],  2, 2, 0)) {

  986.             av_log(avctx, AV_LOG_ERROR, "Cannot init VLC\n");

  987.             on2avc_free_vlcs(c);

  988.             return AVERROR(ENOMEM);

  989.         }

  990.     }

  991. 

  992.     return 0;

  993. }

  994. 

  995. static av_cold int on2avc_decode_close(AVCodecContext *avctx)

  996. {

  997.     On2AVCContext *c = avctx->priv_data;

  998. 

  999.     ff_mdct_end(&c->mdct);

  1000.     ff_mdct_end(&c->mdct_half);

  1001.     ff_mdct_end(&c->mdct_small);

  1002.     ff_fft_end(&c->fft128);

  1003.     ff_fft_end(&c->fft256);

  1004.     ff_fft_end(&c->fft512);

  1005.     ff_fft_end(&c->fft1024);

  1006. 

  1007.     on2avc_free_vlcs(c);

  1008. 

  1009.     return 0;

  1010. }

  1011. 

  1012. 

  1013. AVCodec ff_on2avc_decoder = {

  1014.     .name           = "on2avc",

  1015.     .long_name      = NULL_IF_CONFIG_SMALL("On2 Audio for Video Codec"),

  1016.     .type           = AVMEDIA_TYPE_AUDIO,

  1017.     .id             = AV_CODEC_ID_ON2AVC,

  1018.     .priv_data_size = sizeof(On2AVCContext),

  1019.     .init           = on2avc_decode_init,

  1020.     .decode         = on2avc_decode_frame,

  1021.     .close          = on2avc_decode_close,

  1022.     .capabilities   = AV_CODEC_CAP_DR1,

  1023.     .sample_fmts    = (const enum AVSampleFormat[]) { AV_SAMPLE_FMT_FLTP,

  1024.                                                       AV_SAMPLE_FMT_NONE },

  1025. };