• Main Page
  • Related Pages
  • Modules
  • Data Structures
  • Files
  • File List
  • Globals

libavcodec/dnxhdenc.c

Go to the documentation of this file.
00001 /*
00002  * VC3/DNxHD encoder
00003  * Copyright (c) 2007 Baptiste Coudurier <baptiste dot coudurier at smartjog dot com>
00004  *
00005  * VC-3 encoder funded by the British Broadcasting Corporation
00006  *
00007  * This file is part of FFmpeg.
00008  *
00009  * FFmpeg is free software; you can redistribute it and/or
00010  * modify it under the terms of the GNU Lesser General Public
00011  * License as published by the Free Software Foundation; either
00012  * version 2.1 of the License, or (at your option) any later version.
00013  *
00014  * FFmpeg is distributed in the hope that it will be useful,
00015  * but WITHOUT ANY WARRANTY; without even the implied warranty of
00016  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
00017  * Lesser General Public License for more details.
00018  *
00019  * You should have received a copy of the GNU Lesser General Public
00020  * License along with FFmpeg; if not, write to the Free Software
00021  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
00022  */
00023 
00024 //#define DEBUG
00025 #define RC_VARIANCE 1 // use variance or ssd for fast rc
00026 
00027 #include "avcodec.h"
00028 #include "dsputil.h"
00029 #include "mpegvideo.h"
00030 #include "dnxhdenc.h"
00031 
00032 int dct_quantize_c(MpegEncContext *s, DCTELEM *block, int n, int qscale, int *overflow);
00033 
00034 #define LAMBDA_FRAC_BITS 10
00035 
00036 static av_always_inline void dnxhd_get_pixels_8x4(DCTELEM *restrict block, const uint8_t *pixels, int line_size)
00037 {
00038     int i;
00039     for (i = 0; i < 4; i++) {
00040         block[0] = pixels[0]; block[1] = pixels[1];
00041         block[2] = pixels[2]; block[3] = pixels[3];
00042         block[4] = pixels[4]; block[5] = pixels[5];
00043         block[6] = pixels[6]; block[7] = pixels[7];
00044         pixels += line_size;
00045         block += 8;
00046     }
00047     memcpy(block   , block- 8, sizeof(*block)*8);
00048     memcpy(block+ 8, block-16, sizeof(*block)*8);
00049     memcpy(block+16, block-24, sizeof(*block)*8);
00050     memcpy(block+24, block-32, sizeof(*block)*8);
00051 }
00052 
00053 static int dnxhd_init_vlc(DNXHDEncContext *ctx)
00054 {
00055     int i, j, level, run;
00056     int max_level = 1<<(ctx->cid_table->bit_depth+2);
00057 
00058     CHECKED_ALLOCZ(ctx->vlc_codes, max_level*4*sizeof(*ctx->vlc_codes));
00059     CHECKED_ALLOCZ(ctx->vlc_bits,  max_level*4*sizeof(*ctx->vlc_bits));
00060     CHECKED_ALLOCZ(ctx->run_codes, 63*2);
00061     CHECKED_ALLOCZ(ctx->run_bits,    63);
00062 
00063     ctx->vlc_codes += max_level*2;
00064     ctx->vlc_bits  += max_level*2;
00065     for (level = -max_level; level < max_level; level++) {
00066         for (run = 0; run < 2; run++) {
00067             int index = (level<<1)|run;
00068             int sign, offset = 0, alevel = level;
00069 
00070             MASK_ABS(sign, alevel);
00071             if (alevel > 64) {
00072                 offset = (alevel-1)>>6;
00073                 alevel -= offset<<6;
00074             }
00075             for (j = 0; j < 257; j++) {
00076                 if (ctx->cid_table->ac_level[j] == alevel &&
00077                     (!offset || (ctx->cid_table->ac_index_flag[j] && offset)) &&
00078                     (!run    || (ctx->cid_table->ac_run_flag  [j] && run))) {
00079                     assert(!ctx->vlc_codes[index]);
00080                     if (alevel) {
00081                         ctx->vlc_codes[index] = (ctx->cid_table->ac_codes[j]<<1)|(sign&1);
00082                         ctx->vlc_bits [index] = ctx->cid_table->ac_bits[j]+1;
00083                     } else {
00084                         ctx->vlc_codes[index] = ctx->cid_table->ac_codes[j];
00085                         ctx->vlc_bits [index] = ctx->cid_table->ac_bits [j];
00086                     }
00087                     break;
00088                 }
00089             }
00090             assert(!alevel || j < 257);
00091             if (offset) {
00092                 ctx->vlc_codes[index] = (ctx->vlc_codes[index]<<ctx->cid_table->index_bits)|offset;
00093                 ctx->vlc_bits [index]+= ctx->cid_table->index_bits;
00094             }
00095         }
00096     }
00097     for (i = 0; i < 62; i++) {
00098         int run = ctx->cid_table->run[i];
00099         assert(run < 63);
00100         ctx->run_codes[run] = ctx->cid_table->run_codes[i];
00101         ctx->run_bits [run] = ctx->cid_table->run_bits[i];
00102     }
00103     return 0;
00104  fail:
00105     return -1;
00106 }
00107 
00108 static int dnxhd_init_qmat(DNXHDEncContext *ctx, int lbias, int cbias)
00109 {
00110     // init first elem to 1 to avoid div by 0 in convert_matrix
00111     uint16_t weight_matrix[64] = {1,}; // convert_matrix needs uint16_t*
00112     int qscale, i;
00113 
00114     CHECKED_ALLOCZ(ctx->qmatrix_l,   (ctx->m.avctx->qmax+1) * 64 * sizeof(int));
00115     CHECKED_ALLOCZ(ctx->qmatrix_c,   (ctx->m.avctx->qmax+1) * 64 * sizeof(int));
00116     CHECKED_ALLOCZ(ctx->qmatrix_l16, (ctx->m.avctx->qmax+1) * 64 * 2 * sizeof(uint16_t));
00117     CHECKED_ALLOCZ(ctx->qmatrix_c16, (ctx->m.avctx->qmax+1) * 64 * 2 * sizeof(uint16_t));
00118 
00119     for (i = 1; i < 64; i++) {
00120         int j = ctx->m.dsp.idct_permutation[ff_zigzag_direct[i]];
00121         weight_matrix[j] = ctx->cid_table->luma_weight[i];
00122     }
00123     ff_convert_matrix(&ctx->m.dsp, ctx->qmatrix_l, ctx->qmatrix_l16, weight_matrix,
00124                       ctx->m.intra_quant_bias, 1, ctx->m.avctx->qmax, 1);
00125     for (i = 1; i < 64; i++) {
00126         int j = ctx->m.dsp.idct_permutation[ff_zigzag_direct[i]];
00127         weight_matrix[j] = ctx->cid_table->chroma_weight[i];
00128     }
00129     ff_convert_matrix(&ctx->m.dsp, ctx->qmatrix_c, ctx->qmatrix_c16, weight_matrix,
00130                       ctx->m.intra_quant_bias, 1, ctx->m.avctx->qmax, 1);
00131     for (qscale = 1; qscale <= ctx->m.avctx->qmax; qscale++) {
00132         for (i = 0; i < 64; i++) {
00133             ctx->qmatrix_l  [qscale]   [i] <<= 2; ctx->qmatrix_c  [qscale]   [i] <<= 2;
00134             ctx->qmatrix_l16[qscale][0][i] <<= 2; ctx->qmatrix_l16[qscale][1][i] <<= 2;
00135             ctx->qmatrix_c16[qscale][0][i] <<= 2; ctx->qmatrix_c16[qscale][1][i] <<= 2;
00136         }
00137     }
00138     return 0;
00139  fail:
00140     return -1;
00141 }
00142 
00143 static int dnxhd_init_rc(DNXHDEncContext *ctx)
00144 {
00145     CHECKED_ALLOCZ(ctx->mb_rc, 8160*ctx->m.avctx->qmax*sizeof(RCEntry));
00146     if (ctx->m.avctx->mb_decision != FF_MB_DECISION_RD)
00147         CHECKED_ALLOCZ(ctx->mb_cmp, ctx->m.mb_num*sizeof(RCCMPEntry));
00148 
00149     ctx->frame_bits = (ctx->cid_table->coding_unit_size - 640 - 4) * 8;
00150     ctx->qscale = 1;
00151     ctx->lambda = 2<<LAMBDA_FRAC_BITS; // qscale 2
00152     return 0;
00153  fail:
00154     return -1;
00155 }
00156 
00157 static int dnxhd_encode_init(AVCodecContext *avctx)
00158 {
00159     DNXHDEncContext *ctx = avctx->priv_data;
00160     int i, index;
00161 
00162     ctx->cid = ff_dnxhd_find_cid(avctx);
00163     if (!ctx->cid || avctx->pix_fmt != PIX_FMT_YUV422P) {
00164         av_log(avctx, AV_LOG_ERROR, "video parameters incompatible with DNxHD\n");
00165         return -1;
00166     }
00167     av_log(avctx, AV_LOG_DEBUG, "cid %d\n", ctx->cid);
00168 
00169     index = ff_dnxhd_get_cid_table(ctx->cid);
00170     ctx->cid_table = &ff_dnxhd_cid_table[index];
00171 
00172     ctx->m.avctx = avctx;
00173     ctx->m.mb_intra = 1;
00174     ctx->m.h263_aic = 1;
00175 
00176     ctx->get_pixels_8x4_sym = dnxhd_get_pixels_8x4;
00177 
00178     dsputil_init(&ctx->m.dsp, avctx);
00179     ff_dct_common_init(&ctx->m);
00180 #if HAVE_MMX
00181     ff_dnxhd_init_mmx(ctx);
00182 #endif
00183     if (!ctx->m.dct_quantize)
00184         ctx->m.dct_quantize = dct_quantize_c;
00185 
00186     ctx->m.mb_height = (avctx->height + 15) / 16;
00187     ctx->m.mb_width  = (avctx->width  + 15) / 16;
00188 
00189     if (avctx->flags & CODEC_FLAG_INTERLACED_DCT) {
00190         ctx->interlaced = 1;
00191         ctx->m.mb_height /= 2;
00192     }
00193 
00194     ctx->m.mb_num = ctx->m.mb_height * ctx->m.mb_width;
00195 
00196     if (avctx->intra_quant_bias != FF_DEFAULT_QUANT_BIAS)
00197         ctx->m.intra_quant_bias = avctx->intra_quant_bias;
00198     if (dnxhd_init_qmat(ctx, ctx->m.intra_quant_bias, 0) < 0) // XXX tune lbias/cbias
00199         return -1;
00200 
00201     if (dnxhd_init_vlc(ctx) < 0)
00202         return -1;
00203     if (dnxhd_init_rc(ctx) < 0)
00204         return -1;
00205 
00206     CHECKED_ALLOCZ(ctx->slice_size, ctx->m.mb_height*sizeof(uint32_t));
00207     CHECKED_ALLOCZ(ctx->mb_bits,    ctx->m.mb_num   *sizeof(uint16_t));
00208     CHECKED_ALLOCZ(ctx->mb_qscale,  ctx->m.mb_num   *sizeof(uint8_t));
00209 
00210     ctx->frame.key_frame = 1;
00211     ctx->frame.pict_type = FF_I_TYPE;
00212     ctx->m.avctx->coded_frame = &ctx->frame;
00213 
00214     if (avctx->thread_count > MAX_THREADS || (avctx->thread_count > ctx->m.mb_height)) {
00215         av_log(avctx, AV_LOG_ERROR, "too many threads\n");
00216         return -1;
00217     }
00218 
00219     ctx->thread[0] = ctx;
00220     for (i = 1; i < avctx->thread_count; i++) {
00221         ctx->thread[i] =  av_malloc(sizeof(DNXHDEncContext));
00222         memcpy(ctx->thread[i], ctx, sizeof(DNXHDEncContext));
00223     }
00224 
00225     for (i = 0; i < avctx->thread_count; i++) {
00226         ctx->thread[i]->m.start_mb_y = (ctx->m.mb_height*(i  ) + avctx->thread_count/2) / avctx->thread_count;
00227         ctx->thread[i]->m.end_mb_y   = (ctx->m.mb_height*(i+1) + avctx->thread_count/2) / avctx->thread_count;
00228     }
00229 
00230     return 0;
00231  fail: //for CHECKED_ALLOCZ
00232     return -1;
00233 }
00234 
00235 static int dnxhd_write_header(AVCodecContext *avctx, uint8_t *buf)
00236 {
00237     DNXHDEncContext *ctx = avctx->priv_data;
00238     const uint8_t header_prefix[5] = { 0x00,0x00,0x02,0x80,0x01 };
00239 
00240     memcpy(buf, header_prefix, 5);
00241     buf[5] = ctx->interlaced ? ctx->cur_field+2 : 0x01;
00242     buf[6] = 0x80; // crc flag off
00243     buf[7] = 0xa0; // reserved
00244     AV_WB16(buf + 0x18, avctx->height); // ALPF
00245     AV_WB16(buf + 0x1a, avctx->width);  // SPL
00246     AV_WB16(buf + 0x1d, avctx->height); // NAL
00247 
00248     buf[0x21] = 0x38; // FIXME 8 bit per comp
00249     buf[0x22] = 0x88 + (ctx->frame.interlaced_frame<<2);
00250     AV_WB32(buf + 0x28, ctx->cid); // CID
00251     buf[0x2c] = ctx->interlaced ? 0 : 0x80;
00252 
00253     buf[0x5f] = 0x01; // UDL
00254 
00255     buf[0x167] = 0x02; // reserved
00256     AV_WB16(buf + 0x16a, ctx->m.mb_height * 4 + 4); // MSIPS
00257     buf[0x16d] = ctx->m.mb_height; // Ns
00258     buf[0x16f] = 0x10; // reserved
00259 
00260     ctx->msip = buf + 0x170;
00261     return 0;
00262 }
00263 
00264 static av_always_inline void dnxhd_encode_dc(DNXHDEncContext *ctx, int diff)
00265 {
00266     int nbits;
00267     if (diff < 0) {
00268         nbits = av_log2_16bit(-2*diff);
00269         diff--;
00270     } else {
00271         nbits = av_log2_16bit(2*diff);
00272     }
00273     put_bits(&ctx->m.pb, ctx->cid_table->dc_bits[nbits] + nbits,
00274              (ctx->cid_table->dc_codes[nbits]<<nbits) + (diff & ((1 << nbits) - 1)));
00275 }
00276 
00277 static av_always_inline void dnxhd_encode_block(DNXHDEncContext *ctx, DCTELEM *block, int last_index, int n)
00278 {
00279     int last_non_zero = 0;
00280     int slevel, i, j;
00281 
00282     dnxhd_encode_dc(ctx, block[0] - ctx->m.last_dc[n]);
00283     ctx->m.last_dc[n] = block[0];
00284 
00285     for (i = 1; i <= last_index; i++) {
00286         j = ctx->m.intra_scantable.permutated[i];
00287         slevel = block[j];
00288         if (slevel) {
00289             int run_level = i - last_non_zero - 1;
00290             int rlevel = (slevel<<1)|!!run_level;
00291             put_bits(&ctx->m.pb, ctx->vlc_bits[rlevel], ctx->vlc_codes[rlevel]);
00292             if (run_level)
00293                 put_bits(&ctx->m.pb, ctx->run_bits[run_level], ctx->run_codes[run_level]);
00294             last_non_zero = i;
00295         }
00296     }
00297     put_bits(&ctx->m.pb, ctx->vlc_bits[0], ctx->vlc_codes[0]); // EOB
00298 }
00299 
00300 static av_always_inline void dnxhd_unquantize_c(DNXHDEncContext *ctx, DCTELEM *block, int n, int qscale, int last_index)
00301 {
00302     const uint8_t *weight_matrix;
00303     int level;
00304     int i;
00305 
00306     weight_matrix = (n&2) ? ctx->cid_table->chroma_weight : ctx->cid_table->luma_weight;
00307 
00308     for (i = 1; i <= last_index; i++) {
00309         int j = ctx->m.intra_scantable.permutated[i];
00310         level = block[j];
00311         if (level) {
00312             if (level < 0) {
00313                 level = (1-2*level) * qscale * weight_matrix[i];
00314                 if (weight_matrix[i] != 32)
00315                     level += 32;
00316                 level >>= 6;
00317                 level = -level;
00318             } else {
00319                 level = (2*level+1) * qscale * weight_matrix[i];
00320                 if (weight_matrix[i] != 32)
00321                     level += 32;
00322                 level >>= 6;
00323             }
00324             block[j] = level;
00325         }
00326     }
00327 }
00328 
00329 static av_always_inline int dnxhd_ssd_block(DCTELEM *qblock, DCTELEM *block)
00330 {
00331     int score = 0;
00332     int i;
00333     for (i = 0; i < 64; i++)
00334         score += (block[i]-qblock[i])*(block[i]-qblock[i]);
00335     return score;
00336 }
00337 
00338 static av_always_inline int dnxhd_calc_ac_bits(DNXHDEncContext *ctx, DCTELEM *block, int last_index)
00339 {
00340     int last_non_zero = 0;
00341     int bits = 0;
00342     int i, j, level;
00343     for (i = 1; i <= last_index; i++) {
00344         j = ctx->m.intra_scantable.permutated[i];
00345         level = block[j];
00346         if (level) {
00347             int run_level = i - last_non_zero - 1;
00348             bits += ctx->vlc_bits[(level<<1)|!!run_level]+ctx->run_bits[run_level];
00349             last_non_zero = i;
00350         }
00351     }
00352     return bits;
00353 }
00354 
00355 static av_always_inline void dnxhd_get_blocks(DNXHDEncContext *ctx, int mb_x, int mb_y)
00356 {
00357     const uint8_t *ptr_y = ctx->thread[0]->src[0] + ((mb_y << 4) * ctx->m.linesize)   + (mb_x << 4);
00358     const uint8_t *ptr_u = ctx->thread[0]->src[1] + ((mb_y << 4) * ctx->m.uvlinesize) + (mb_x << 3);
00359     const uint8_t *ptr_v = ctx->thread[0]->src[2] + ((mb_y << 4) * ctx->m.uvlinesize) + (mb_x << 3);
00360     DSPContext *dsp = &ctx->m.dsp;
00361 
00362     dsp->get_pixels(ctx->blocks[0], ptr_y    , ctx->m.linesize);
00363     dsp->get_pixels(ctx->blocks[1], ptr_y + 8, ctx->m.linesize);
00364     dsp->get_pixels(ctx->blocks[2], ptr_u    , ctx->m.uvlinesize);
00365     dsp->get_pixels(ctx->blocks[3], ptr_v    , ctx->m.uvlinesize);
00366 
00367     if (mb_y+1 == ctx->m.mb_height && ctx->m.avctx->height == 1080) {
00368         if (ctx->interlaced) {
00369             ctx->get_pixels_8x4_sym(ctx->blocks[4], ptr_y + ctx->dct_y_offset    , ctx->m.linesize);
00370             ctx->get_pixels_8x4_sym(ctx->blocks[5], ptr_y + ctx->dct_y_offset + 8, ctx->m.linesize);
00371             ctx->get_pixels_8x4_sym(ctx->blocks[6], ptr_u + ctx->dct_uv_offset   , ctx->m.uvlinesize);
00372             ctx->get_pixels_8x4_sym(ctx->blocks[7], ptr_v + ctx->dct_uv_offset   , ctx->m.uvlinesize);
00373         } else {
00374             dsp->clear_block(ctx->blocks[4]); dsp->clear_block(ctx->blocks[5]);
00375             dsp->clear_block(ctx->blocks[6]); dsp->clear_block(ctx->blocks[7]);
00376         }
00377     } else {
00378         dsp->get_pixels(ctx->blocks[4], ptr_y + ctx->dct_y_offset    , ctx->m.linesize);
00379         dsp->get_pixels(ctx->blocks[5], ptr_y + ctx->dct_y_offset + 8, ctx->m.linesize);
00380         dsp->get_pixels(ctx->blocks[6], ptr_u + ctx->dct_uv_offset   , ctx->m.uvlinesize);
00381         dsp->get_pixels(ctx->blocks[7], ptr_v + ctx->dct_uv_offset   , ctx->m.uvlinesize);
00382     }
00383 }
00384 
00385 static av_always_inline int dnxhd_switch_matrix(DNXHDEncContext *ctx, int i)
00386 {
00387     if (i&2) {
00388         ctx->m.q_intra_matrix16 = ctx->qmatrix_c16;
00389         ctx->m.q_intra_matrix   = ctx->qmatrix_c;
00390         return 1 + (i&1);
00391     } else {
00392         ctx->m.q_intra_matrix16 = ctx->qmatrix_l16;
00393         ctx->m.q_intra_matrix   = ctx->qmatrix_l;
00394         return 0;
00395     }
00396 }
00397 
00398 static int dnxhd_calc_bits_thread(AVCodecContext *avctx, void *arg)
00399 {
00400     DNXHDEncContext *ctx = *(void**)arg;
00401     int mb_y, mb_x;
00402     int qscale = ctx->thread[0]->qscale;
00403 
00404     for (mb_y = ctx->m.start_mb_y; mb_y < ctx->m.end_mb_y; mb_y++) {
00405         ctx->m.last_dc[0] =
00406         ctx->m.last_dc[1] =
00407         ctx->m.last_dc[2] = 1024;
00408 
00409         for (mb_x = 0; mb_x < ctx->m.mb_width; mb_x++) {
00410             unsigned mb = mb_y * ctx->m.mb_width + mb_x;
00411             int ssd     = 0;
00412             int ac_bits = 0;
00413             int dc_bits = 0;
00414             int i;
00415 
00416             dnxhd_get_blocks(ctx, mb_x, mb_y);
00417 
00418             for (i = 0; i < 8; i++) {
00419                 DECLARE_ALIGNED_16(DCTELEM, block[64]);
00420                 DCTELEM *src_block = ctx->blocks[i];
00421                 int overflow, nbits, diff, last_index;
00422                 int n = dnxhd_switch_matrix(ctx, i);
00423 
00424                 memcpy(block, src_block, sizeof(block));
00425                 last_index = ctx->m.dct_quantize((MpegEncContext*)ctx, block, i, qscale, &overflow);
00426                 ac_bits += dnxhd_calc_ac_bits(ctx, block, last_index);
00427 
00428                 diff = block[0] - ctx->m.last_dc[n];
00429                 if (diff < 0) nbits = av_log2_16bit(-2*diff);
00430                 else          nbits = av_log2_16bit( 2*diff);
00431                 dc_bits += ctx->cid_table->dc_bits[nbits] + nbits;
00432 
00433                 ctx->m.last_dc[n] = block[0];
00434 
00435                 if (avctx->mb_decision == FF_MB_DECISION_RD || !RC_VARIANCE) {
00436                     dnxhd_unquantize_c(ctx, block, i, qscale, last_index);
00437                     ctx->m.dsp.idct(block);
00438                     ssd += dnxhd_ssd_block(block, src_block);
00439                 }
00440             }
00441             ctx->mb_rc[qscale][mb].ssd = ssd;
00442             ctx->mb_rc[qscale][mb].bits = ac_bits+dc_bits+12+8*ctx->vlc_bits[0];
00443         }
00444     }
00445     return 0;
00446 }
00447 
00448 static int dnxhd_encode_thread(AVCodecContext *avctx, void *arg)
00449 {
00450     DNXHDEncContext *ctx = *(void**)arg;
00451     int mb_y, mb_x;
00452 
00453     for (mb_y = ctx->m.start_mb_y; mb_y < ctx->m.end_mb_y; mb_y++) {
00454         ctx->m.last_dc[0] =
00455         ctx->m.last_dc[1] =
00456         ctx->m.last_dc[2] = 1024;
00457         for (mb_x = 0; mb_x < ctx->m.mb_width; mb_x++) {
00458             unsigned mb = mb_y * ctx->m.mb_width + mb_x;
00459             int qscale = ctx->mb_qscale[mb];
00460             int i;
00461 
00462             put_bits(&ctx->m.pb, 12, qscale<<1);
00463 
00464             dnxhd_get_blocks(ctx, mb_x, mb_y);
00465 
00466             for (i = 0; i < 8; i++) {
00467                 DCTELEM *block = ctx->blocks[i];
00468                 int last_index, overflow;
00469                 int n = dnxhd_switch_matrix(ctx, i);
00470                 last_index = ctx->m.dct_quantize((MpegEncContext*)ctx, block, i, qscale, &overflow);
00471                 //START_TIMER;
00472                 dnxhd_encode_block(ctx, block, last_index, n);
00473                 //STOP_TIMER("encode_block");
00474             }
00475         }
00476         if (put_bits_count(&ctx->m.pb)&31)
00477             put_bits(&ctx->m.pb, 32-(put_bits_count(&ctx->m.pb)&31), 0);
00478     }
00479     flush_put_bits(&ctx->m.pb);
00480     return 0;
00481 }
00482 
00483 static void dnxhd_setup_threads_slices(DNXHDEncContext *ctx, uint8_t *buf)
00484 {
00485     int mb_y, mb_x;
00486     int i, offset = 0;
00487     for (i = 0; i < ctx->m.avctx->thread_count; i++) {
00488         int thread_size = 0;
00489         for (mb_y = ctx->thread[i]->m.start_mb_y; mb_y < ctx->thread[i]->m.end_mb_y; mb_y++) {
00490             ctx->slice_size[mb_y] = 0;
00491             for (mb_x = 0; mb_x < ctx->m.mb_width; mb_x++) {
00492                 unsigned mb = mb_y * ctx->m.mb_width + mb_x;
00493                 ctx->slice_size[mb_y] += ctx->mb_bits[mb];
00494             }
00495             ctx->slice_size[mb_y] = (ctx->slice_size[mb_y]+31)&~31;
00496             ctx->slice_size[mb_y] >>= 3;
00497             thread_size += ctx->slice_size[mb_y];
00498         }
00499         init_put_bits(&ctx->thread[i]->m.pb, buf + 640 + offset, thread_size);
00500         offset += thread_size;
00501     }
00502 }
00503 
00504 static int dnxhd_mb_var_thread(AVCodecContext *avctx, void *arg)
00505 {
00506     DNXHDEncContext *ctx = *(void**)arg;
00507     int mb_y, mb_x;
00508     for (mb_y = ctx->m.start_mb_y; mb_y < ctx->m.end_mb_y; mb_y++) {
00509         for (mb_x = 0; mb_x < ctx->m.mb_width; mb_x++) {
00510             unsigned mb  = mb_y * ctx->m.mb_width + mb_x;
00511             uint8_t *pix = ctx->thread[0]->src[0] + ((mb_y<<4) * ctx->m.linesize) + (mb_x<<4);
00512             int sum      = ctx->m.dsp.pix_sum(pix, ctx->m.linesize);
00513             int varc     = (ctx->m.dsp.pix_norm1(pix, ctx->m.linesize) - (((unsigned)(sum*sum))>>8)+128)>>8;
00514             ctx->mb_cmp[mb].value = varc;
00515             ctx->mb_cmp[mb].mb = mb;
00516         }
00517     }
00518     return 0;
00519 }
00520 
00521 static int dnxhd_encode_rdo(AVCodecContext *avctx, DNXHDEncContext *ctx)
00522 {
00523     int lambda, up_step, down_step;
00524     int last_lower = INT_MAX, last_higher = 0;
00525     int x, y, q;
00526 
00527     for (q = 1; q < avctx->qmax; q++) {
00528         ctx->qscale = q;
00529         avctx->execute(avctx, dnxhd_calc_bits_thread, (void**)&ctx->thread[0], NULL, avctx->thread_count, sizeof(void*));
00530     }
00531     up_step = down_step = 2<<LAMBDA_FRAC_BITS;
00532     lambda = ctx->lambda;
00533 
00534     for (;;) {
00535         int bits = 0;
00536         int end = 0;
00537         if (lambda == last_higher) {
00538             lambda++;
00539             end = 1; // need to set final qscales/bits
00540         }
00541         for (y = 0; y < ctx->m.mb_height; y++) {
00542             for (x = 0; x < ctx->m.mb_width; x++) {
00543                 unsigned min = UINT_MAX;
00544                 int qscale = 1;
00545                 int mb = y*ctx->m.mb_width+x;
00546                 for (q = 1; q < avctx->qmax; q++) {
00547                     unsigned score = ctx->mb_rc[q][mb].bits*lambda+(ctx->mb_rc[q][mb].ssd<<LAMBDA_FRAC_BITS);
00548                     if (score < min) {
00549                         min = score;
00550                         qscale = q;
00551                     }
00552                 }
00553                 bits += ctx->mb_rc[qscale][mb].bits;
00554                 ctx->mb_qscale[mb] = qscale;
00555                 ctx->mb_bits[mb] = ctx->mb_rc[qscale][mb].bits;
00556             }
00557             bits = (bits+31)&~31; // padding
00558             if (bits > ctx->frame_bits)
00559                 break;
00560         }
00561         //dprintf(ctx->m.avctx, "lambda %d, up %u, down %u, bits %d, frame %d\n",
00562         //        lambda, last_higher, last_lower, bits, ctx->frame_bits);
00563         if (end) {
00564             if (bits > ctx->frame_bits)
00565                 return -1;
00566             break;
00567         }
00568         if (bits < ctx->frame_bits) {
00569             last_lower = FFMIN(lambda, last_lower);
00570             if (last_higher != 0)
00571                 lambda = (lambda+last_higher)>>1;
00572             else
00573                 lambda -= down_step;
00574             down_step *= 5; // XXX tune ?
00575             up_step = 1<<LAMBDA_FRAC_BITS;
00576             lambda = FFMAX(1, lambda);
00577             if (lambda == last_lower)
00578                 break;
00579         } else {
00580             last_higher = FFMAX(lambda, last_higher);
00581             if (last_lower != INT_MAX)
00582                 lambda = (lambda+last_lower)>>1;
00583             else
00584                 lambda += up_step;
00585             up_step *= 5;
00586             down_step = 1<<LAMBDA_FRAC_BITS;
00587         }
00588     }
00589     //dprintf(ctx->m.avctx, "out lambda %d\n", lambda);
00590     ctx->lambda = lambda;
00591     return 0;
00592 }
00593 
00594 static int dnxhd_find_qscale(DNXHDEncContext *ctx)
00595 {
00596     int bits = 0;
00597     int up_step = 1;
00598     int down_step = 1;
00599     int last_higher = 0;
00600     int last_lower = INT_MAX;
00601     int qscale;
00602     int x, y;
00603 
00604     qscale = ctx->qscale;
00605     for (;;) {
00606         bits = 0;
00607         ctx->qscale = qscale;
00608         // XXX avoid recalculating bits
00609         ctx->m.avctx->execute(ctx->m.avctx, dnxhd_calc_bits_thread, (void**)&ctx->thread[0], NULL, ctx->m.avctx->thread_count, sizeof(void*));
00610         for (y = 0; y < ctx->m.mb_height; y++) {
00611             for (x = 0; x < ctx->m.mb_width; x++)
00612                 bits += ctx->mb_rc[qscale][y*ctx->m.mb_width+x].bits;
00613             bits = (bits+31)&~31; // padding
00614             if (bits > ctx->frame_bits)
00615                 break;
00616         }
00617         //dprintf(ctx->m.avctx, "%d, qscale %d, bits %d, frame %d, higher %d, lower %d\n",
00618         //        ctx->m.avctx->frame_number, qscale, bits, ctx->frame_bits, last_higher, last_lower);
00619         if (bits < ctx->frame_bits) {
00620             if (qscale == 1)
00621                 return 1;
00622             if (last_higher == qscale - 1) {
00623                 qscale = last_higher;
00624                 break;
00625             }
00626             last_lower = FFMIN(qscale, last_lower);
00627             if (last_higher != 0)
00628                 qscale = (qscale+last_higher)>>1;
00629             else
00630                 qscale -= down_step++;
00631             if (qscale < 1)
00632                 qscale = 1;
00633             up_step = 1;
00634         } else {
00635             if (last_lower == qscale + 1)
00636                 break;
00637             last_higher = FFMAX(qscale, last_higher);
00638             if (last_lower != INT_MAX)
00639                 qscale = (qscale+last_lower)>>1;
00640             else
00641                 qscale += up_step++;
00642             down_step = 1;
00643             if (qscale >= ctx->m.avctx->qmax)
00644                 return -1;
00645         }
00646     }
00647     //dprintf(ctx->m.avctx, "out qscale %d\n", qscale);
00648     ctx->qscale = qscale;
00649     return 0;
00650 }
00651 
00652 static int dnxhd_rc_cmp(const void *a, const void *b)
00653 {
00654     return ((const RCCMPEntry *)b)->value - ((const RCCMPEntry *)a)->value;
00655 }
00656 
00657 static int dnxhd_encode_fast(AVCodecContext *avctx, DNXHDEncContext *ctx)
00658 {
00659     int max_bits = 0;
00660     int ret, x, y;
00661     if ((ret = dnxhd_find_qscale(ctx)) < 0)
00662         return -1;
00663     for (y = 0; y < ctx->m.mb_height; y++) {
00664         for (x = 0; x < ctx->m.mb_width; x++) {
00665             int mb = y*ctx->m.mb_width+x;
00666             int delta_bits;
00667             ctx->mb_qscale[mb] = ctx->qscale;
00668             ctx->mb_bits[mb] = ctx->mb_rc[ctx->qscale][mb].bits;
00669             max_bits += ctx->mb_rc[ctx->qscale][mb].bits;
00670             if (!RC_VARIANCE) {
00671                 delta_bits = ctx->mb_rc[ctx->qscale][mb].bits-ctx->mb_rc[ctx->qscale+1][mb].bits;
00672                 ctx->mb_cmp[mb].mb = mb;
00673                 ctx->mb_cmp[mb].value = delta_bits ?
00674                     ((ctx->mb_rc[ctx->qscale][mb].ssd-ctx->mb_rc[ctx->qscale+1][mb].ssd)*100)/delta_bits
00675                     : INT_MIN; //avoid increasing qscale
00676             }
00677         }
00678         max_bits += 31; //worst padding
00679     }
00680     if (!ret) {
00681         if (RC_VARIANCE)
00682             avctx->execute(avctx, dnxhd_mb_var_thread, (void**)&ctx->thread[0], NULL, avctx->thread_count, sizeof(void*));
00683         qsort(ctx->mb_cmp, ctx->m.mb_num, sizeof(RCEntry), dnxhd_rc_cmp);
00684         for (x = 0; x < ctx->m.mb_num && max_bits > ctx->frame_bits; x++) {
00685             int mb = ctx->mb_cmp[x].mb;
00686             max_bits -= ctx->mb_rc[ctx->qscale][mb].bits - ctx->mb_rc[ctx->qscale+1][mb].bits;
00687             ctx->mb_qscale[mb] = ctx->qscale+1;
00688             ctx->mb_bits[mb] = ctx->mb_rc[ctx->qscale+1][mb].bits;
00689         }
00690     }
00691     return 0;
00692 }
00693 
00694 static void dnxhd_load_picture(DNXHDEncContext *ctx, const AVFrame *frame)
00695 {
00696     int i;
00697 
00698     for (i = 0; i < 3; i++) {
00699         ctx->frame.data[i]     = frame->data[i];
00700         ctx->frame.linesize[i] = frame->linesize[i];
00701     }
00702 
00703     for (i = 0; i < ctx->m.avctx->thread_count; i++) {
00704         ctx->thread[i]->m.linesize    = ctx->frame.linesize[0]<<ctx->interlaced;
00705         ctx->thread[i]->m.uvlinesize  = ctx->frame.linesize[1]<<ctx->interlaced;
00706         ctx->thread[i]->dct_y_offset  = ctx->m.linesize  *8;
00707         ctx->thread[i]->dct_uv_offset = ctx->m.uvlinesize*8;
00708     }
00709 
00710     ctx->frame.interlaced_frame = frame->interlaced_frame;
00711     ctx->cur_field = frame->interlaced_frame && !frame->top_field_first;
00712 }
00713 
00714 static int dnxhd_encode_picture(AVCodecContext *avctx, unsigned char *buf, int buf_size, const void *data)
00715 {
00716     DNXHDEncContext *ctx = avctx->priv_data;
00717     int first_field = 1;
00718     int offset, i, ret;
00719 
00720     if (buf_size < ctx->cid_table->frame_size) {
00721         av_log(avctx, AV_LOG_ERROR, "output buffer is too small to compress picture\n");
00722         return -1;
00723     }
00724 
00725     dnxhd_load_picture(ctx, data);
00726 
00727  encode_coding_unit:
00728     for (i = 0; i < 3; i++) {
00729         ctx->src[i] = ctx->frame.data[i];
00730         if (ctx->interlaced && ctx->cur_field)
00731             ctx->src[i] += ctx->frame.linesize[i];
00732     }
00733 
00734     dnxhd_write_header(avctx, buf);
00735 
00736     if (avctx->mb_decision == FF_MB_DECISION_RD)
00737         ret = dnxhd_encode_rdo(avctx, ctx);
00738     else
00739         ret = dnxhd_encode_fast(avctx, ctx);
00740     if (ret < 0) {
00741         av_log(avctx, AV_LOG_ERROR, "picture could not fit ratecontrol constraints\n");
00742         return -1;
00743     }
00744 
00745     dnxhd_setup_threads_slices(ctx, buf);
00746 
00747     offset = 0;
00748     for (i = 0; i < ctx->m.mb_height; i++) {
00749         AV_WB32(ctx->msip + i * 4, offset);
00750         offset += ctx->slice_size[i];
00751         assert(!(ctx->slice_size[i] & 3));
00752     }
00753 
00754     avctx->execute(avctx, dnxhd_encode_thread, (void**)&ctx->thread[0], NULL, avctx->thread_count, sizeof(void*));
00755 
00756     AV_WB32(buf + ctx->cid_table->coding_unit_size - 4, 0x600DC0DE); // EOF
00757 
00758     if (ctx->interlaced && first_field) {
00759         first_field     = 0;
00760         ctx->cur_field ^= 1;
00761         buf      += ctx->cid_table->coding_unit_size;
00762         buf_size -= ctx->cid_table->coding_unit_size;
00763         goto encode_coding_unit;
00764     }
00765 
00766     ctx->frame.quality = ctx->qscale*FF_QP2LAMBDA;
00767 
00768     return ctx->cid_table->frame_size;
00769 }
00770 
00771 static int dnxhd_encode_end(AVCodecContext *avctx)
00772 {
00773     DNXHDEncContext *ctx = avctx->priv_data;
00774     int max_level = 1<<(ctx->cid_table->bit_depth+2);
00775     int i;
00776 
00777     av_free(ctx->vlc_codes-max_level*2);
00778     av_free(ctx->vlc_bits -max_level*2);
00779     av_freep(&ctx->run_codes);
00780     av_freep(&ctx->run_bits);
00781 
00782     av_freep(&ctx->mb_bits);
00783     av_freep(&ctx->mb_qscale);
00784     av_freep(&ctx->mb_rc);
00785     av_freep(&ctx->mb_cmp);
00786     av_freep(&ctx->slice_size);
00787 
00788     av_freep(&ctx->qmatrix_c);
00789     av_freep(&ctx->qmatrix_l);
00790     av_freep(&ctx->qmatrix_c16);
00791     av_freep(&ctx->qmatrix_l16);
00792 
00793     for (i = 1; i < avctx->thread_count; i++)
00794         av_freep(&ctx->thread[i]);
00795 
00796     return 0;
00797 }
00798 
00799 AVCodec dnxhd_encoder = {
00800     "dnxhd",
00801     CODEC_TYPE_VIDEO,
00802     CODEC_ID_DNXHD,
00803     sizeof(DNXHDEncContext),
00804     dnxhd_encode_init,
00805     dnxhd_encode_picture,
00806     dnxhd_encode_end,
00807     .pix_fmts = (enum PixelFormat[]){PIX_FMT_YUV422P, PIX_FMT_NONE},
00808     .long_name = NULL_IF_CONFIG_SMALL("VC3/DNxHD"),
00809 };

Generated on Sat Feb 16 2013 09:23:11 for ffmpeg by  doxygen 1.7.1