yading@10: /*
yading@10:  * Apple ProRes encoder
yading@10:  *
yading@10:  * Copyright (c) 2012 Konstantin Shishkov
yading@10:  *
yading@10:  * This encoder appears to be based on Anatoliy Wassermans considering
yading@10:  * similarities in the bugs.
yading@10:  *
yading@10:  * This file is part of Libav.
yading@10:  *
yading@10:  * Libav is free software; you can redistribute it and/or
yading@10:  * modify it under the terms of the GNU Lesser General Public
yading@10:  * License as published by the Free Software Foundation; either
yading@10:  * version 2.1 of the License, or (at your option) any later version.
yading@10:  *
yading@10:  * Libav is distributed in the hope that it will be useful,
yading@10:  * but WITHOUT ANY WARRANTY; without even the implied warranty of
yading@10:  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
yading@10:  * Lesser General Public License for more details.
yading@10:  *
yading@10:  * You should have received a copy of the GNU Lesser General Public
yading@10:  * License along with Libav; if not, write to the Free Software
yading@10:  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
yading@10:  */
yading@10: 
yading@10: #include "libavutil/opt.h"
yading@10: #include "avcodec.h"
yading@10: #include "dsputil.h"
yading@10: #include "put_bits.h"
yading@10: #include "bytestream.h"
yading@10: #include "internal.h"
yading@10: #include "proresdsp.h"
yading@10: #include "proresdata.h"
yading@10: 
yading@10: #define CFACTOR_Y422 2
yading@10: #define CFACTOR_Y444 3
yading@10: 
yading@10: #define MAX_MBS_PER_SLICE 8
yading@10: 
yading@10: #define MAX_PLANES 3 // should be increased to 4 when there's AV_PIX_FMT_YUV444AP10
yading@10: 
yading@10: enum {
yading@10:     PRORES_PROFILE_PROXY = 0,
yading@10:     PRORES_PROFILE_LT,
yading@10:     PRORES_PROFILE_STANDARD,
yading@10:     PRORES_PROFILE_HQ,
yading@10: };
yading@10: 
yading@10: enum {
yading@10:     QUANT_MAT_PROXY = 0,
yading@10:     QUANT_MAT_LT,
yading@10:     QUANT_MAT_STANDARD,
yading@10:     QUANT_MAT_HQ,
yading@10:     QUANT_MAT_DEFAULT,
yading@10: };
yading@10: 
yading@10: static const uint8_t prores_quant_matrices[][64] = {
yading@10:     { // proxy
yading@10:          4,  7,  9, 11, 13, 14, 15, 63,
yading@10:          7,  7, 11, 12, 14, 15, 63, 63,
yading@10:          9, 11, 13, 14, 15, 63, 63, 63,
yading@10:         11, 11, 13, 14, 63, 63, 63, 63,
yading@10:         11, 13, 14, 63, 63, 63, 63, 63,
yading@10:         13, 14, 63, 63, 63, 63, 63, 63,
yading@10:         13, 63, 63, 63, 63, 63, 63, 63,
yading@10:         63, 63, 63, 63, 63, 63, 63, 63,
yading@10:     },
yading@10:     { // LT
yading@10:          4,  5,  6,  7,  9, 11, 13, 15,
yading@10:          5,  5,  7,  8, 11, 13, 15, 17,
yading@10:          6,  7,  9, 11, 13, 15, 15, 17,
yading@10:          7,  7,  9, 11, 13, 15, 17, 19,
yading@10:          7,  9, 11, 13, 14, 16, 19, 23,
yading@10:          9, 11, 13, 14, 16, 19, 23, 29,
yading@10:          9, 11, 13, 15, 17, 21, 28, 35,
yading@10:         11, 13, 16, 17, 21, 28, 35, 41,
yading@10:     },
yading@10:     { // standard
yading@10:          4,  4,  5,  5,  6,  7,  7,  9,
yading@10:          4,  4,  5,  6,  7,  7,  9,  9,
yading@10:          5,  5,  6,  7,  7,  9,  9, 10,
yading@10:          5,  5,  6,  7,  7,  9,  9, 10,
yading@10:          5,  6,  7,  7,  8,  9, 10, 12,
yading@10:          6,  7,  7,  8,  9, 10, 12, 15,
yading@10:          6,  7,  7,  9, 10, 11, 14, 17,
yading@10:          7,  7,  9, 10, 11, 14, 17, 21,
yading@10:     },
yading@10:     { // high quality
yading@10:          4,  4,  4,  4,  4,  4,  4,  4,
yading@10:          4,  4,  4,  4,  4,  4,  4,  4,
yading@10:          4,  4,  4,  4,  4,  4,  4,  4,
yading@10:          4,  4,  4,  4,  4,  4,  4,  5,
yading@10:          4,  4,  4,  4,  4,  4,  5,  5,
yading@10:          4,  4,  4,  4,  4,  5,  5,  6,
yading@10:          4,  4,  4,  4,  5,  5,  6,  7,
yading@10:          4,  4,  4,  4,  5,  6,  7,  7,
yading@10:     },
yading@10:     { // codec default
yading@10:          4,  4,  4,  4,  4,  4,  4,  4,
yading@10:          4,  4,  4,  4,  4,  4,  4,  4,
yading@10:          4,  4,  4,  4,  4,  4,  4,  4,
yading@10:          4,  4,  4,  4,  4,  4,  4,  4,
yading@10:          4,  4,  4,  4,  4,  4,  4,  4,
yading@10:          4,  4,  4,  4,  4,  4,  4,  4,
yading@10:          4,  4,  4,  4,  4,  4,  4,  4,
yading@10:          4,  4,  4,  4,  4,  4,  4,  4,
yading@10:     },
yading@10: };
yading@10: 
yading@10: #define NUM_MB_LIMITS 4
yading@10: static const int prores_mb_limits[NUM_MB_LIMITS] = {
yading@10:     1620, // up to 720x576
yading@10:     2700, // up to 960x720
yading@10:     6075, // up to 1440x1080
yading@10:     9216, // up to 2048x1152
yading@10: };
yading@10: 
yading@10: static const struct prores_profile {
yading@10:     const char *full_name;
yading@10:     uint32_t    tag;
yading@10:     int         min_quant;
yading@10:     int         max_quant;
yading@10:     int         br_tab[NUM_MB_LIMITS];
yading@10:     int         quant;
yading@10: } prores_profile_info[4] = {
yading@10:     {
yading@10:         .full_name = "proxy",
yading@10:         .tag       = MKTAG('a', 'p', 'c', 'o'),
yading@10:         .min_quant = 4,
yading@10:         .max_quant = 8,
yading@10:         .br_tab    = { 300, 242, 220, 194 },
yading@10:         .quant     = QUANT_MAT_PROXY,
yading@10:     },
yading@10:     {
yading@10:         .full_name = "LT",
yading@10:         .tag       = MKTAG('a', 'p', 'c', 's'),
yading@10:         .min_quant = 1,
yading@10:         .max_quant = 9,
yading@10:         .br_tab    = { 720, 560, 490, 440 },
yading@10:         .quant     = QUANT_MAT_LT,
yading@10:     },
yading@10:     {
yading@10:         .full_name = "standard",
yading@10:         .tag       = MKTAG('a', 'p', 'c', 'n'),
yading@10:         .min_quant = 1,
yading@10:         .max_quant = 6,
yading@10:         .br_tab    = { 1050, 808, 710, 632 },
yading@10:         .quant     = QUANT_MAT_STANDARD,
yading@10:     },
yading@10:     {
yading@10:         .full_name = "high quality",
yading@10:         .tag       = MKTAG('a', 'p', 'c', 'h'),
yading@10:         .min_quant = 1,
yading@10:         .max_quant = 6,
yading@10:         .br_tab    = { 1566, 1216, 1070, 950 },
yading@10:         .quant     = QUANT_MAT_HQ,
yading@10:     }
yading@10: // for 4444 profile bitrate numbers are { 2350, 1828, 1600, 1425 }
yading@10: };
yading@10: 
yading@10: #define TRELLIS_WIDTH 16
yading@10: #define SCORE_LIMIT   INT_MAX / 2
yading@10: 
yading@10: struct TrellisNode {
yading@10:     int prev_node;
yading@10:     int quant;
yading@10:     int bits;
yading@10:     int score;
yading@10: };
yading@10: 
yading@10: #define MAX_STORED_Q 16
yading@10: 
yading@10: typedef struct ProresThreadData {
yading@10:     DECLARE_ALIGNED(16, int16_t, blocks)[MAX_PLANES][64 * 4 * MAX_MBS_PER_SLICE];
yading@10:     DECLARE_ALIGNED(16, uint16_t, emu_buf)[16 * 16];
yading@10:     int16_t custom_q[64];
yading@10:     struct TrellisNode *nodes;
yading@10: } ProresThreadData;
yading@10: 
yading@10: typedef struct ProresContext {
yading@10:     AVClass *class;
yading@10:     DECLARE_ALIGNED(16, int16_t, blocks)[MAX_PLANES][64 * 4 * MAX_MBS_PER_SLICE];
yading@10:     DECLARE_ALIGNED(16, uint16_t, emu_buf)[16*16];
yading@10:     int16_t quants[MAX_STORED_Q][64];
yading@10:     int16_t custom_q[64];
yading@10:     const uint8_t *quant_mat;
yading@10: 
yading@10:     ProresDSPContext dsp;
yading@10:     ScanTable  scantable;
yading@10: 
yading@10:     int mb_width, mb_height;
yading@10:     int mbs_per_slice;
yading@10:     int num_chroma_blocks, chroma_factor;
yading@10:     int slices_width;
yading@10:     int slices_per_picture;
yading@10:     int pictures_per_frame; // 1 for progressive, 2 for interlaced
yading@10:     int cur_picture_idx;
yading@10:     int num_planes;
yading@10:     int bits_per_mb;
yading@10:     int force_quant;
yading@10: 
yading@10:     char *vendor;
yading@10:     int quant_sel;
yading@10: 
yading@10:     int frame_size_upper_bound;
yading@10: 
yading@10:     int profile;
yading@10:     const struct prores_profile *profile_info;
yading@10: 
yading@10:     int *slice_q;
yading@10: 
yading@10:     ProresThreadData *tdata;
yading@10: } ProresContext;
yading@10: 
yading@10: static void get_slice_data(ProresContext *ctx, const uint16_t *src,
yading@10:                            int linesize, int x, int y, int w, int h,
yading@10:                            int16_t *blocks, uint16_t *emu_buf,
yading@10:                            int mbs_per_slice, int blocks_per_mb, int is_chroma)
yading@10: {
yading@10:     const uint16_t *esrc;
yading@10:     const int mb_width = 4 * blocks_per_mb;
yading@10:     int elinesize;
yading@10:     int i, j, k;
yading@10: 
yading@10:     for (i = 0; i < mbs_per_slice; i++, src += mb_width) {
yading@10:         if (x >= w) {
yading@10:             memset(blocks, 0, 64 * (mbs_per_slice - i) * blocks_per_mb
yading@10:                               * sizeof(*blocks));
yading@10:             return;
yading@10:         }
yading@10:         if (x + mb_width <= w && y + 16 <= h) {
yading@10:             esrc      = src;
yading@10:             elinesize = linesize;
yading@10:         } else {
yading@10:             int bw, bh, pix;
yading@10: 
yading@10:             esrc      = emu_buf;
yading@10:             elinesize = 16 * sizeof(*emu_buf);
yading@10: 
yading@10:             bw = FFMIN(w - x, mb_width);
yading@10:             bh = FFMIN(h - y, 16);
yading@10: 
yading@10:             for (j = 0; j < bh; j++) {
yading@10:                 memcpy(emu_buf + j * 16,
yading@10:                        (const uint8_t*)src + j * linesize,
yading@10:                        bw * sizeof(*src));
yading@10:                 pix = emu_buf[j * 16 + bw - 1];
yading@10:                 for (k = bw; k < mb_width; k++)
yading@10:                     emu_buf[j * 16 + k] = pix;
yading@10:             }
yading@10:             for (; j < 16; j++)
yading@10:                 memcpy(emu_buf + j * 16,
yading@10:                        emu_buf + (bh - 1) * 16,
yading@10:                        mb_width * sizeof(*emu_buf));
yading@10:         }
yading@10:         if (!is_chroma) {
yading@10:             ctx->dsp.fdct(esrc, elinesize, blocks);
yading@10:             blocks += 64;
yading@10:             if (blocks_per_mb > 2) {
yading@10:                 ctx->dsp.fdct(esrc + 8, elinesize, blocks);
yading@10:                 blocks += 64;
yading@10:             }
yading@10:             ctx->dsp.fdct(esrc + elinesize * 4, elinesize, blocks);
yading@10:             blocks += 64;
yading@10:             if (blocks_per_mb > 2) {
yading@10:                 ctx->dsp.fdct(esrc + elinesize * 4 + 8, elinesize, blocks);
yading@10:                 blocks += 64;
yading@10:             }
yading@10:         } else {
yading@10:             ctx->dsp.fdct(esrc, elinesize, blocks);
yading@10:             blocks += 64;
yading@10:             ctx->dsp.fdct(esrc + elinesize * 4, elinesize, blocks);
yading@10:             blocks += 64;
yading@10:             if (blocks_per_mb > 2) {
yading@10:                 ctx->dsp.fdct(esrc + 8, elinesize, blocks);
yading@10:                 blocks += 64;
yading@10:                 ctx->dsp.fdct(esrc + elinesize * 4 + 8, elinesize, blocks);
yading@10:                 blocks += 64;
yading@10:             }
yading@10:         }
yading@10: 
yading@10:         x += mb_width;
yading@10:     }
yading@10: }
yading@10: 
yading@10: /**
yading@10:  * Write an unsigned rice/exp golomb codeword.
yading@10:  */
yading@10: static inline void encode_vlc_codeword(PutBitContext *pb, unsigned codebook, int val)
yading@10: {
yading@10:     unsigned int rice_order, exp_order, switch_bits, switch_val;
yading@10:     int exponent;
yading@10: 
yading@10:     /* number of prefix bits to switch between Rice and expGolomb */
yading@10:     switch_bits = (codebook & 3) + 1;
yading@10:     rice_order  =  codebook >> 5;       /* rice code order */
yading@10:     exp_order   = (codebook >> 2) & 7;  /* exp golomb code order */
yading@10: 
yading@10:     switch_val  = switch_bits << rice_order;
yading@10: 
yading@10:     if (val >= switch_val) {
yading@10:         val -= switch_val - (1 << exp_order);
yading@10:         exponent = av_log2(val);
yading@10: 
yading@10:         put_bits(pb, exponent - exp_order + switch_bits, 0);
yading@10:         put_bits(pb, exponent + 1, val);
yading@10:     } else {
yading@10:         exponent = val >> rice_order;
yading@10: 
yading@10:         if (exponent)
yading@10:             put_bits(pb, exponent, 0);
yading@10:         put_bits(pb, 1, 1);
yading@10:         if (rice_order)
yading@10:             put_sbits(pb, rice_order, val);
yading@10:     }
yading@10: }
yading@10: 
yading@10: #define GET_SIGN(x)  ((x) >> 31)
yading@10: #define MAKE_CODE(x) (((x) << 1) ^ GET_SIGN(x))
yading@10: 
yading@10: static void encode_dcs(PutBitContext *pb, int16_t *blocks,
yading@10:                        int blocks_per_slice, int scale)
yading@10: {
yading@10:     int i;
yading@10:     int codebook = 3, code, dc, prev_dc, delta, sign, new_sign;
yading@10: 
yading@10:     prev_dc = (blocks[0] - 0x4000) / scale;
yading@10:     encode_vlc_codeword(pb, FIRST_DC_CB, MAKE_CODE(prev_dc));
yading@10:     sign     = 0;
yading@10:     codebook = 3;
yading@10:     blocks  += 64;
yading@10: 
yading@10:     for (i = 1; i < blocks_per_slice; i++, blocks += 64) {
yading@10:         dc       = (blocks[0] - 0x4000) / scale;
yading@10:         delta    = dc - prev_dc;
yading@10:         new_sign = GET_SIGN(delta);
yading@10:         delta    = (delta ^ sign) - sign;
yading@10:         code     = MAKE_CODE(delta);
yading@10:         encode_vlc_codeword(pb, ff_prores_dc_codebook[codebook], code);
yading@10:         codebook = (code + (code & 1)) >> 1;
yading@10:         codebook = FFMIN(codebook, 3);
yading@10:         sign     = new_sign;
yading@10:         prev_dc  = dc;
yading@10:     }
yading@10: }
yading@10: 
yading@10: static void encode_acs(PutBitContext *pb, int16_t *blocks,
yading@10:                        int blocks_per_slice,
yading@10:                        int plane_size_factor,
yading@10:                        const uint8_t *scan, const int16_t *qmat)
yading@10: {
yading@10:     int idx, i;
yading@10:     int run, level, run_cb, lev_cb;
yading@10:     int max_coeffs, abs_level;
yading@10: 
yading@10:     max_coeffs = blocks_per_slice << 6;
yading@10:     run_cb     = ff_prores_run_to_cb_index[4];
yading@10:     lev_cb     = ff_prores_lev_to_cb_index[2];
yading@10:     run        = 0;
yading@10: 
yading@10:     for (i = 1; i < 64; i++) {
yading@10:         for (idx = scan[i]; idx < max_coeffs; idx += 64) {
yading@10:             level = blocks[idx] / qmat[scan[i]];
yading@10:             if (level) {
yading@10:                 abs_level = FFABS(level);
yading@10:                 encode_vlc_codeword(pb, ff_prores_ac_codebook[run_cb], run);
yading@10:                 encode_vlc_codeword(pb, ff_prores_ac_codebook[lev_cb],
yading@10:                                     abs_level - 1);
yading@10:                 put_sbits(pb, 1, GET_SIGN(level));
yading@10: 
yading@10:                 run_cb = ff_prores_run_to_cb_index[FFMIN(run, 15)];
yading@10:                 lev_cb = ff_prores_lev_to_cb_index[FFMIN(abs_level, 9)];
yading@10:                 run    = 0;
yading@10:             } else {
yading@10:                 run++;
yading@10:             }
yading@10:         }
yading@10:     }
yading@10: }
yading@10: 
yading@10: static int encode_slice_plane(ProresContext *ctx, PutBitContext *pb,
yading@10:                               const uint16_t *src, int linesize,
yading@10:                               int mbs_per_slice, int16_t *blocks,
yading@10:                               int blocks_per_mb, int plane_size_factor,
yading@10:                               const int16_t *qmat)
yading@10: {
yading@10:     int blocks_per_slice, saved_pos;
yading@10: 
yading@10:     saved_pos = put_bits_count(pb);
yading@10:     blocks_per_slice = mbs_per_slice * blocks_per_mb;
yading@10: 
yading@10:     encode_dcs(pb, blocks, blocks_per_slice, qmat[0]);
yading@10:     encode_acs(pb, blocks, blocks_per_slice, plane_size_factor,
yading@10:                ctx->scantable.permutated, qmat);
yading@10:     flush_put_bits(pb);
yading@10: 
yading@10:     return (put_bits_count(pb) - saved_pos) >> 3;
yading@10: }
yading@10: 
yading@10: static int encode_slice(AVCodecContext *avctx, const AVFrame *pic,
yading@10:                         PutBitContext *pb,
yading@10:                         int sizes[4], int x, int y, int quant,
yading@10:                         int mbs_per_slice)
yading@10: {
yading@10:     ProresContext *ctx = avctx->priv_data;
yading@10:     int i, xp, yp;
yading@10:     int total_size = 0;
yading@10:     const uint16_t *src;
yading@10:     int slice_width_factor = av_log2(mbs_per_slice);
yading@10:     int num_cblocks, pwidth, linesize, line_add;
yading@10:     int plane_factor, is_chroma;
yading@10:     uint16_t *qmat;
yading@10: 
yading@10:     if (ctx->pictures_per_frame == 1)
yading@10:         line_add = 0;
yading@10:     else
yading@10:         line_add = ctx->cur_picture_idx ^ !pic->top_field_first;
yading@10: 
yading@10:     if (ctx->force_quant) {
yading@10:         qmat = ctx->quants[0];
yading@10:     } else if (quant < MAX_STORED_Q) {
yading@10:         qmat = ctx->quants[quant];
yading@10:     } else {
yading@10:         qmat = ctx->custom_q;
yading@10:         for (i = 0; i < 64; i++)
yading@10:             qmat[i] = ctx->quant_mat[i] * quant;
yading@10:     }
yading@10: 
yading@10:     for (i = 0; i < ctx->num_planes; i++) {
yading@10:         is_chroma    = (i == 1 || i == 2);
yading@10:         plane_factor = slice_width_factor + 2;
yading@10:         if (is_chroma)
yading@10:             plane_factor += ctx->chroma_factor - 3;
yading@10:         if (!is_chroma || ctx->chroma_factor == CFACTOR_Y444) {
yading@10:             xp          = x << 4;
yading@10:             yp          = y << 4;
yading@10:             num_cblocks = 4;
yading@10:             pwidth      = avctx->width;
yading@10:         } else {
yading@10:             xp          = x << 3;
yading@10:             yp          = y << 4;
yading@10:             num_cblocks = 2;
yading@10:             pwidth      = avctx->width >> 1;
yading@10:         }
yading@10: 
yading@10:         linesize = pic->linesize[i] * ctx->pictures_per_frame;
yading@10:         src = (const uint16_t*)(pic->data[i] + yp * linesize +
yading@10:                                 line_add * pic->linesize[i]) + xp;
yading@10: 
yading@10:         get_slice_data(ctx, src, linesize, xp, yp,
yading@10:                        pwidth, avctx->height / ctx->pictures_per_frame,
yading@10:                        ctx->blocks[0], ctx->emu_buf,
yading@10:                        mbs_per_slice, num_cblocks, is_chroma);
yading@10:         sizes[i] = encode_slice_plane(ctx, pb, src, linesize,
yading@10:                                       mbs_per_slice, ctx->blocks[0],
yading@10:                                       num_cblocks, plane_factor,
yading@10:                                       qmat);
yading@10:         total_size += sizes[i];
yading@10:     }
yading@10:     return total_size;
yading@10: }
yading@10: 
yading@10: static inline int estimate_vlc(unsigned codebook, int val)
yading@10: {
yading@10:     unsigned int rice_order, exp_order, switch_bits, switch_val;
yading@10:     int exponent;
yading@10: 
yading@10:     /* number of prefix bits to switch between Rice and expGolomb */
yading@10:     switch_bits = (codebook & 3) + 1;
yading@10:     rice_order  =  codebook >> 5;       /* rice code order */
yading@10:     exp_order   = (codebook >> 2) & 7;  /* exp golomb code order */
yading@10: 
yading@10:     switch_val  = switch_bits << rice_order;
yading@10: 
yading@10:     if (val >= switch_val) {
yading@10:         val -= switch_val - (1 << exp_order);
yading@10:         exponent = av_log2(val);
yading@10: 
yading@10:         return exponent * 2 - exp_order + switch_bits + 1;
yading@10:     } else {
yading@10:         return (val >> rice_order) + rice_order + 1;
yading@10:     }
yading@10: }
yading@10: 
yading@10: static int estimate_dcs(int *error, int16_t *blocks, int blocks_per_slice,
yading@10:                         int scale)
yading@10: {
yading@10:     int i;
yading@10:     int codebook = 3, code, dc, prev_dc, delta, sign, new_sign;
yading@10:     int bits;
yading@10: 
yading@10:     prev_dc  = (blocks[0] - 0x4000) / scale;
yading@10:     bits     = estimate_vlc(FIRST_DC_CB, MAKE_CODE(prev_dc));
yading@10:     sign     = 0;
yading@10:     codebook = 3;
yading@10:     blocks  += 64;
yading@10:     *error  += FFABS(blocks[0] - 0x4000) % scale;
yading@10: 
yading@10:     for (i = 1; i < blocks_per_slice; i++, blocks += 64) {
yading@10:         dc       = (blocks[0] - 0x4000) / scale;
yading@10:         *error  += FFABS(blocks[0] - 0x4000) % scale;
yading@10:         delta    = dc - prev_dc;
yading@10:         new_sign = GET_SIGN(delta);
yading@10:         delta    = (delta ^ sign) - sign;
yading@10:         code     = MAKE_CODE(delta);
yading@10:         bits    += estimate_vlc(ff_prores_dc_codebook[codebook], code);
yading@10:         codebook = (code + (code & 1)) >> 1;
yading@10:         codebook = FFMIN(codebook, 3);
yading@10:         sign     = new_sign;
yading@10:         prev_dc  = dc;
yading@10:     }
yading@10: 
yading@10:     return bits;
yading@10: }
yading@10: 
yading@10: static int estimate_acs(int *error, int16_t *blocks, int blocks_per_slice,
yading@10:                         int plane_size_factor,
yading@10:                         const uint8_t *scan, const int16_t *qmat)
yading@10: {
yading@10:     int idx, i;
yading@10:     int run, level, run_cb, lev_cb;
yading@10:     int max_coeffs, abs_level;
yading@10:     int bits = 0;
yading@10: 
yading@10:     max_coeffs = blocks_per_slice << 6;
yading@10:     run_cb     = ff_prores_run_to_cb_index[4];
yading@10:     lev_cb     = ff_prores_lev_to_cb_index[2];
yading@10:     run        = 0;
yading@10: 
yading@10:     for (i = 1; i < 64; i++) {
yading@10:         for (idx = scan[i]; idx < max_coeffs; idx += 64) {
yading@10:             level   = blocks[idx] / qmat[scan[i]];
yading@10:             *error += FFABS(blocks[idx]) % qmat[scan[i]];
yading@10:             if (level) {
yading@10:                 abs_level = FFABS(level);
yading@10:                 bits += estimate_vlc(ff_prores_ac_codebook[run_cb], run);
yading@10:                 bits += estimate_vlc(ff_prores_ac_codebook[lev_cb],
yading@10:                                      abs_level - 1) + 1;
yading@10: 
yading@10:                 run_cb = ff_prores_run_to_cb_index[FFMIN(run, 15)];
yading@10:                 lev_cb = ff_prores_lev_to_cb_index[FFMIN(abs_level, 9)];
yading@10:                 run    = 0;
yading@10:             } else {
yading@10:                 run++;
yading@10:             }
yading@10:         }
yading@10:     }
yading@10: 
yading@10:     return bits;
yading@10: }
yading@10: 
yading@10: static int estimate_slice_plane(ProresContext *ctx, int *error, int plane,
yading@10:                                 const uint16_t *src, int linesize,
yading@10:                                 int mbs_per_slice,
yading@10:                                 int blocks_per_mb, int plane_size_factor,
yading@10:                                 const int16_t *qmat, ProresThreadData *td)
yading@10: {
yading@10:     int blocks_per_slice;
yading@10:     int bits;
yading@10: 
yading@10:     blocks_per_slice = mbs_per_slice * blocks_per_mb;
yading@10: 
yading@10:     bits  = estimate_dcs(error, td->blocks[plane], blocks_per_slice, qmat[0]);
yading@10:     bits += estimate_acs(error, td->blocks[plane], blocks_per_slice,
yading@10:                          plane_size_factor, ctx->scantable.permutated, qmat);
yading@10: 
yading@10:     return FFALIGN(bits, 8);
yading@10: }
yading@10: 
yading@10: static int find_slice_quant(AVCodecContext *avctx, const AVFrame *pic,
yading@10:                             int trellis_node, int x, int y, int mbs_per_slice,
yading@10:                             ProresThreadData *td)
yading@10: {
yading@10:     ProresContext *ctx = avctx->priv_data;
yading@10:     int i, q, pq, xp, yp;
yading@10:     const uint16_t *src;
yading@10:     int slice_width_factor = av_log2(mbs_per_slice);
yading@10:     int num_cblocks[MAX_PLANES], pwidth;
yading@10:     int plane_factor[MAX_PLANES], is_chroma[MAX_PLANES];
yading@10:     const int min_quant = ctx->profile_info->min_quant;
yading@10:     const int max_quant = ctx->profile_info->max_quant;
yading@10:     int error, bits, bits_limit;
yading@10:     int mbs, prev, cur, new_score;
yading@10:     int slice_bits[TRELLIS_WIDTH], slice_score[TRELLIS_WIDTH];
yading@10:     int overquant;
yading@10:     uint16_t *qmat;
yading@10:     int linesize[4], line_add;
yading@10: 
yading@10:     if (ctx->pictures_per_frame == 1)
yading@10:         line_add = 0;
yading@10:     else
yading@10:         line_add = ctx->cur_picture_idx ^ !pic->top_field_first;
yading@10:     mbs = x + mbs_per_slice;
yading@10: 
yading@10:     for (i = 0; i < ctx->num_planes; i++) {
yading@10:         is_chroma[i]    = (i == 1 || i == 2);
yading@10:         plane_factor[i] = slice_width_factor + 2;
yading@10:         if (is_chroma[i])
yading@10:             plane_factor[i] += ctx->chroma_factor - 3;
yading@10:         if (!is_chroma[i] || ctx->chroma_factor == CFACTOR_Y444) {
yading@10:             xp             = x << 4;
yading@10:             yp             = y << 4;
yading@10:             num_cblocks[i] = 4;
yading@10:             pwidth         = avctx->width;
yading@10:         } else {
yading@10:             xp             = x << 3;
yading@10:             yp             = y << 4;
yading@10:             num_cblocks[i] = 2;
yading@10:             pwidth         = avctx->width >> 1;
yading@10:         }
yading@10: 
yading@10:         linesize[i] = pic->linesize[i] * ctx->pictures_per_frame;
yading@10:         src = (const uint16_t*)(pic->data[i] + yp * linesize[i] +
yading@10:                                 line_add * pic->linesize[i]) + xp;
yading@10: 
yading@10:         get_slice_data(ctx, src, linesize[i], xp, yp,
yading@10:                        pwidth, avctx->height / ctx->pictures_per_frame,
yading@10:                        td->blocks[i], td->emu_buf,
yading@10:                        mbs_per_slice, num_cblocks[i], is_chroma[i]);
yading@10:     }
yading@10: 
yading@10:     for (q = min_quant; q < max_quant + 2; q++) {
yading@10:         td->nodes[trellis_node + q].prev_node = -1;
yading@10:         td->nodes[trellis_node + q].quant     = q;
yading@10:     }
yading@10: 
yading@10:     // todo: maybe perform coarser quantising to fit into frame size when needed
yading@10:     for (q = min_quant; q <= max_quant; q++) {
yading@10:         bits  = 0;
yading@10:         error = 0;
yading@10:         for (i = 0; i < ctx->num_planes; i++) {
yading@10:             bits += estimate_slice_plane(ctx, &error, i,
yading@10:                                          src, linesize[i],
yading@10:                                          mbs_per_slice,
yading@10:                                          num_cblocks[i], plane_factor[i],
yading@10:                                          ctx->quants[q], td);
yading@10:         }
yading@10:         if (bits > 65000 * 8) {
yading@10:             error = SCORE_LIMIT;
yading@10:             break;
yading@10:         }
yading@10:         slice_bits[q]  = bits;
yading@10:         slice_score[q] = error;
yading@10:     }
yading@10:     if (slice_bits[max_quant] <= ctx->bits_per_mb * mbs_per_slice) {
yading@10:         slice_bits[max_quant + 1]  = slice_bits[max_quant];
yading@10:         slice_score[max_quant + 1] = slice_score[max_quant] + 1;
yading@10:         overquant = max_quant;
yading@10:     } else {
yading@10:         for (q = max_quant + 1; q < 128; q++) {
yading@10:             bits  = 0;
yading@10:             error = 0;
yading@10:             if (q < MAX_STORED_Q) {
yading@10:                 qmat = ctx->quants[q];
yading@10:             } else {
yading@10:                 qmat = td->custom_q;
yading@10:                 for (i = 0; i < 64; i++)
yading@10:                     qmat[i] = ctx->quant_mat[i] * q;
yading@10:             }
yading@10:             for (i = 0; i < ctx->num_planes; i++) {
yading@10:                 bits += estimate_slice_plane(ctx, &error, i,
yading@10:                                              src, linesize[i],
yading@10:                                              mbs_per_slice,
yading@10:                                              num_cblocks[i], plane_factor[i],
yading@10:                                              qmat, td);
yading@10:             }
yading@10:             if (bits <= ctx->bits_per_mb * mbs_per_slice)
yading@10:                 break;
yading@10:         }
yading@10: 
yading@10:         slice_bits[max_quant + 1]  = bits;
yading@10:         slice_score[max_quant + 1] = error;
yading@10:         overquant = q;
yading@10:     }
yading@10:     td->nodes[trellis_node + max_quant + 1].quant = overquant;
yading@10: 
yading@10:     bits_limit = mbs * ctx->bits_per_mb;
yading@10:     for (pq = min_quant; pq < max_quant + 2; pq++) {
yading@10:         prev = trellis_node - TRELLIS_WIDTH + pq;
yading@10: 
yading@10:         for (q = min_quant; q < max_quant + 2; q++) {
yading@10:             cur = trellis_node + q;
yading@10: 
yading@10:             bits  = td->nodes[prev].bits + slice_bits[q];
yading@10:             error = slice_score[q];
yading@10:             if (bits > bits_limit)
yading@10:                 error = SCORE_LIMIT;
yading@10: 
yading@10:             if (td->nodes[prev].score < SCORE_LIMIT && error < SCORE_LIMIT)
yading@10:                 new_score = td->nodes[prev].score + error;
yading@10:             else
yading@10:                 new_score = SCORE_LIMIT;
yading@10:             if (td->nodes[cur].prev_node == -1 ||
yading@10:                 td->nodes[cur].score >= new_score) {
yading@10: 
yading@10:                 td->nodes[cur].bits      = bits;
yading@10:                 td->nodes[cur].score     = new_score;
yading@10:                 td->nodes[cur].prev_node = prev;
yading@10:             }
yading@10:         }
yading@10:     }
yading@10: 
yading@10:     error = td->nodes[trellis_node + min_quant].score;
yading@10:     pq    = trellis_node + min_quant;
yading@10:     for (q = min_quant + 1; q < max_quant + 2; q++) {
yading@10:         if (td->nodes[trellis_node + q].score <= error) {
yading@10:             error = td->nodes[trellis_node + q].score;
yading@10:             pq    = trellis_node + q;
yading@10:         }
yading@10:     }
yading@10: 
yading@10:     return pq;
yading@10: }
yading@10: 
yading@10: static int find_quant_thread(AVCodecContext *avctx, void *arg,
yading@10:                              int jobnr, int threadnr)
yading@10: {
yading@10:     ProresContext *ctx = avctx->priv_data;
yading@10:     ProresThreadData *td = ctx->tdata + threadnr;
yading@10:     int mbs_per_slice = ctx->mbs_per_slice;
yading@10:     int x, y = jobnr, mb, q = 0;
yading@10: 
yading@10:     for (x = mb = 0; x < ctx->mb_width; x += mbs_per_slice, mb++) {
yading@10:         while (ctx->mb_width - x < mbs_per_slice)
yading@10:             mbs_per_slice >>= 1;
yading@10:         q = find_slice_quant(avctx, avctx->coded_frame,
yading@10:                              (mb + 1) * TRELLIS_WIDTH, x, y,
yading@10:                              mbs_per_slice, td);
yading@10:     }
yading@10: 
yading@10:     for (x = ctx->slices_width - 1; x >= 0; x--) {
yading@10:         ctx->slice_q[x + y * ctx->slices_width] = td->nodes[q].quant;
yading@10:         q = td->nodes[q].prev_node;
yading@10:     }
yading@10: 
yading@10:     return 0;
yading@10: }
yading@10: 
yading@10: static int encode_frame(AVCodecContext *avctx, AVPacket *pkt,
yading@10:                         const AVFrame *pic, int *got_packet)
yading@10: {
yading@10:     ProresContext *ctx = avctx->priv_data;
yading@10:     uint8_t *orig_buf, *buf, *slice_hdr, *slice_sizes, *tmp;
yading@10:     uint8_t *picture_size_pos;
yading@10:     PutBitContext pb;
yading@10:     int x, y, i, mb, q = 0;
yading@10:     int sizes[4] = { 0 };
yading@10:     int slice_hdr_size = 2 + 2 * (ctx->num_planes - 1);
yading@10:     int frame_size, picture_size, slice_size;
yading@10:     int pkt_size, ret;
yading@10:     uint8_t frame_flags;
yading@10: 
yading@10:     *avctx->coded_frame           = *pic;
yading@10:     avctx->coded_frame->pict_type = AV_PICTURE_TYPE_I;
yading@10:     avctx->coded_frame->key_frame = 1;
yading@10: 
yading@10:     pkt_size = ctx->frame_size_upper_bound + FF_MIN_BUFFER_SIZE;
yading@10: 
yading@10:     if ((ret = ff_alloc_packet2(avctx, pkt, pkt_size)) < 0)
yading@10:         return ret;
yading@10: 
yading@10:     orig_buf = pkt->data;
yading@10: 
yading@10:     // frame atom
yading@10:     orig_buf += 4;                              // frame size
yading@10:     bytestream_put_be32  (&orig_buf, FRAME_ID); // frame container ID
yading@10:     buf = orig_buf;
yading@10: 
yading@10:     // frame header
yading@10:     tmp = buf;
yading@10:     buf += 2;                                   // frame header size will be stored here
yading@10:     bytestream_put_be16  (&buf, 0);             // version 1
yading@10:     bytestream_put_buffer(&buf, ctx->vendor, 4);
yading@10:     bytestream_put_be16  (&buf, avctx->width);
yading@10:     bytestream_put_be16  (&buf, avctx->height);
yading@10: 
yading@10:     frame_flags = ctx->chroma_factor << 6;
yading@10:     if (avctx->flags & CODEC_FLAG_INTERLACED_DCT)
yading@10:         frame_flags |= pic->top_field_first ? 0x04 : 0x08;
yading@10:     bytestream_put_byte  (&buf, frame_flags);
yading@10: 
yading@10:     bytestream_put_byte  (&buf, 0);             // reserved
yading@10:     bytestream_put_byte  (&buf, avctx->color_primaries);
yading@10:     bytestream_put_byte  (&buf, avctx->color_trc);
yading@10:     bytestream_put_byte  (&buf, avctx->colorspace);
yading@10:     bytestream_put_byte  (&buf, 0x40);          // source format and alpha information
yading@10:     bytestream_put_byte  (&buf, 0);             // reserved
yading@10:     if (ctx->quant_sel != QUANT_MAT_DEFAULT) {
yading@10:         bytestream_put_byte  (&buf, 0x03);      // matrix flags - both matrices are present
yading@10:         // luma quantisation matrix
yading@10:         for (i = 0; i < 64; i++)
yading@10:             bytestream_put_byte(&buf, ctx->quant_mat[i]);
yading@10:         // chroma quantisation matrix
yading@10:         for (i = 0; i < 64; i++)
yading@10:             bytestream_put_byte(&buf, ctx->quant_mat[i]);
yading@10:     } else {
yading@10:         bytestream_put_byte  (&buf, 0x00);      // matrix flags - default matrices are used
yading@10:     }
yading@10:     bytestream_put_be16  (&tmp, buf - orig_buf); // write back frame header size
yading@10: 
yading@10:     for (ctx->cur_picture_idx = 0;
yading@10:          ctx->cur_picture_idx < ctx->pictures_per_frame;
yading@10:          ctx->cur_picture_idx++) {
yading@10:         // picture header
yading@10:         picture_size_pos = buf + 1;
yading@10:         bytestream_put_byte  (&buf, 0x40);          // picture header size (in bits)
yading@10:         buf += 4;                                   // picture data size will be stored here
yading@10:         bytestream_put_be16  (&buf, ctx->slices_per_picture);
yading@10:         bytestream_put_byte  (&buf, av_log2(ctx->mbs_per_slice) << 4); // slice width and height in MBs
yading@10: 
yading@10:         // seek table - will be filled during slice encoding
yading@10:         slice_sizes = buf;
yading@10:         buf += ctx->slices_per_picture * 2;
yading@10: 
yading@10:         // slices
yading@10:         if (!ctx->force_quant) {
yading@10:             ret = avctx->execute2(avctx, find_quant_thread, NULL, NULL,
yading@10:                                   ctx->mb_height);
yading@10:             if (ret)
yading@10:                 return ret;
yading@10:         }
yading@10: 
yading@10:         for (y = 0; y < ctx->mb_height; y++) {
yading@10:             int mbs_per_slice = ctx->mbs_per_slice;
yading@10:             for (x = mb = 0; x < ctx->mb_width; x += mbs_per_slice, mb++) {
yading@10:                 q = ctx->force_quant ? ctx->force_quant
yading@10:                                      : ctx->slice_q[mb + y * ctx->slices_width];
yading@10: 
yading@10:                 while (ctx->mb_width - x < mbs_per_slice)
yading@10:                     mbs_per_slice >>= 1;
yading@10: 
yading@10:                 bytestream_put_byte(&buf, slice_hdr_size << 3);
yading@10:                 slice_hdr = buf;
yading@10:                 buf += slice_hdr_size - 1;
yading@10:                 init_put_bits(&pb, buf, (pkt_size - (buf - orig_buf)) * 8);
yading@10:                 encode_slice(avctx, pic, &pb, sizes, x, y, q, mbs_per_slice);
yading@10: 
yading@10:                 bytestream_put_byte(&slice_hdr, q);
yading@10:                 slice_size = slice_hdr_size + sizes[ctx->num_planes - 1];
yading@10:                 for (i = 0; i < ctx->num_planes - 1; i++) {
yading@10:                     bytestream_put_be16(&slice_hdr, sizes[i]);
yading@10:                     slice_size += sizes[i];
yading@10:                 }
yading@10:                 bytestream_put_be16(&slice_sizes, slice_size);
yading@10:                 buf += slice_size - slice_hdr_size;
yading@10:             }
yading@10:         }
yading@10: 
yading@10:         picture_size = buf - (picture_size_pos - 1);
yading@10:         bytestream_put_be32(&picture_size_pos, picture_size);
yading@10:     }
yading@10: 
yading@10:     orig_buf -= 8;
yading@10:     frame_size = buf - orig_buf;
yading@10:     bytestream_put_be32(&orig_buf, frame_size);
yading@10: 
yading@10:     pkt->size   = frame_size;
yading@10:     pkt->flags |= AV_PKT_FLAG_KEY;
yading@10:     *got_packet = 1;
yading@10: 
yading@10:     return 0;
yading@10: }
yading@10: 
yading@10: static av_cold int encode_close(AVCodecContext *avctx)
yading@10: {
yading@10:     ProresContext *ctx = avctx->priv_data;
yading@10:     int i;
yading@10: 
yading@10:     av_freep(&avctx->coded_frame);
yading@10: 
yading@10:     if (ctx->tdata) {
yading@10:         for (i = 0; i < avctx->thread_count; i++)
yading@10:             av_free(ctx->tdata[i].nodes);
yading@10:     }
yading@10:     av_freep(&ctx->tdata);
yading@10:     av_freep(&ctx->slice_q);
yading@10: 
yading@10:     return 0;
yading@10: }
yading@10: 
yading@10: static av_cold int encode_init(AVCodecContext *avctx)
yading@10: {
yading@10:     ProresContext *ctx = avctx->priv_data;
yading@10:     int mps;
yading@10:     int i, j;
yading@10:     int min_quant, max_quant;
yading@10:     int interlaced = !!(avctx->flags & CODEC_FLAG_INTERLACED_DCT);
yading@10: 
yading@10:     avctx->bits_per_raw_sample = 10;
yading@10:     avctx->coded_frame = avcodec_alloc_frame();
yading@10:     if (!avctx->coded_frame)
yading@10:         return AVERROR(ENOMEM);
yading@10: 
yading@10:     ff_proresdsp_init(&ctx->dsp, avctx);
yading@10:     ff_init_scantable(ctx->dsp.dct_permutation, &ctx->scantable,
yading@10:                       interlaced ? ff_prores_interlaced_scan
yading@10:                                  : ff_prores_progressive_scan);
yading@10: 
yading@10:     mps = ctx->mbs_per_slice;
yading@10:     if (mps & (mps - 1)) {
yading@10:         av_log(avctx, AV_LOG_ERROR,
yading@10:                "there should be an integer power of two MBs per slice\n");
yading@10:         return AVERROR(EINVAL);
yading@10:     }
yading@10: 
yading@10:     ctx->chroma_factor = avctx->pix_fmt == AV_PIX_FMT_YUV422P10
yading@10:                          ? CFACTOR_Y422
yading@10:                          : CFACTOR_Y444;
yading@10:     ctx->profile_info  = prores_profile_info + ctx->profile;
yading@10:     ctx->num_planes    = 3;
yading@10: 
yading@10:     ctx->mb_width      = FFALIGN(avctx->width,  16) >> 4;
yading@10: 
yading@10:     if (interlaced)
yading@10:         ctx->mb_height = FFALIGN(avctx->height, 32) >> 5;
yading@10:     else
yading@10:         ctx->mb_height = FFALIGN(avctx->height, 16) >> 4;
yading@10: 
yading@10:     ctx->slices_width  = ctx->mb_width / mps;
yading@10:     ctx->slices_width += av_popcount(ctx->mb_width - ctx->slices_width * mps);
yading@10:     ctx->slices_per_picture = ctx->mb_height * ctx->slices_width;
yading@10:     ctx->pictures_per_frame = 1 + interlaced;
yading@10: 
yading@10:     if (ctx->quant_sel == -1)
yading@10:         ctx->quant_mat = prores_quant_matrices[ctx->profile_info->quant];
yading@10:     else
yading@10:         ctx->quant_mat = prores_quant_matrices[ctx->quant_sel];
yading@10: 
yading@10:     if (strlen(ctx->vendor) != 4) {
yading@10:         av_log(avctx, AV_LOG_ERROR, "vendor ID should be 4 bytes\n");
yading@10:         return AVERROR_INVALIDDATA;
yading@10:     }
yading@10: 
yading@10:     ctx->force_quant = avctx->global_quality / FF_QP2LAMBDA;
yading@10:     if (!ctx->force_quant) {
yading@10:         if (!ctx->bits_per_mb) {
yading@10:             for (i = 0; i < NUM_MB_LIMITS - 1; i++)
yading@10:                 if (prores_mb_limits[i] >= ctx->mb_width * ctx->mb_height *
yading@10:                                            ctx->pictures_per_frame)
yading@10:                     break;
yading@10:             ctx->bits_per_mb   = ctx->profile_info->br_tab[i];
yading@10:         } else if (ctx->bits_per_mb < 128) {
yading@10:             av_log(avctx, AV_LOG_ERROR, "too few bits per MB, please set at least 128\n");
yading@10:             return AVERROR_INVALIDDATA;
yading@10:         }
yading@10: 
yading@10:         min_quant = ctx->profile_info->min_quant;
yading@10:         max_quant = ctx->profile_info->max_quant;
yading@10:         for (i = min_quant; i < MAX_STORED_Q; i++) {
yading@10:             for (j = 0; j < 64; j++)
yading@10:                 ctx->quants[i][j] = ctx->quant_mat[j] * i;
yading@10:         }
yading@10: 
yading@10:         ctx->slice_q = av_malloc(ctx->slices_per_picture * sizeof(*ctx->slice_q));
yading@10:         if (!ctx->slice_q) {
yading@10:             encode_close(avctx);
yading@10:             return AVERROR(ENOMEM);
yading@10:         }
yading@10: 
yading@10:         ctx->tdata = av_mallocz(avctx->thread_count * sizeof(*ctx->tdata));
yading@10:         if (!ctx->tdata) {
yading@10:             encode_close(avctx);
yading@10:             return AVERROR(ENOMEM);
yading@10:         }
yading@10: 
yading@10:         for (j = 0; j < avctx->thread_count; j++) {
yading@10:             ctx->tdata[j].nodes = av_malloc((ctx->slices_width + 1)
yading@10:                                             * TRELLIS_WIDTH
yading@10:                                             * sizeof(*ctx->tdata->nodes));
yading@10:             if (!ctx->tdata[j].nodes) {
yading@10:                 encode_close(avctx);
yading@10:                 return AVERROR(ENOMEM);
yading@10:             }
yading@10:             for (i = min_quant; i < max_quant + 2; i++) {
yading@10:                 ctx->tdata[j].nodes[i].prev_node = -1;
yading@10:                 ctx->tdata[j].nodes[i].bits      = 0;
yading@10:                 ctx->tdata[j].nodes[i].score     = 0;
yading@10:             }
yading@10:         }
yading@10:     } else {
yading@10:         int ls = 0;
yading@10: 
yading@10:         if (ctx->force_quant > 64) {
yading@10:             av_log(avctx, AV_LOG_ERROR, "too large quantiser, maximum is 64\n");
yading@10:             return AVERROR_INVALIDDATA;
yading@10:         }
yading@10: 
yading@10:         for (j = 0; j < 64; j++) {
yading@10:             ctx->quants[0][j] = ctx->quant_mat[j] * ctx->force_quant;
yading@10:             ls += av_log2((1 << 11)  / ctx->quants[0][j]) * 2 + 1;
yading@10:         }
yading@10: 
yading@10:         ctx->bits_per_mb = ls * 8;
yading@10:         if (ctx->chroma_factor == CFACTOR_Y444)
yading@10:             ctx->bits_per_mb += ls * 4;
yading@10:         if (ctx->num_planes == 4)
yading@10:             ctx->bits_per_mb += ls * 4;
yading@10:     }
yading@10: 
yading@10:     ctx->frame_size_upper_bound = ctx->pictures_per_frame *
yading@10:                                   ctx->slices_per_picture *
yading@10:                                   (2 + 2 * ctx->num_planes +
yading@10:                                    (mps * ctx->bits_per_mb) / 8)
yading@10:                                   + 200;
yading@10: 
yading@10:     avctx->codec_tag   = ctx->profile_info->tag;
yading@10: 
yading@10:     av_log(avctx, AV_LOG_DEBUG,
yading@10:            "profile %d, %d slices, interlacing: %s, %d bits per MB\n",
yading@10:            ctx->profile, ctx->slices_per_picture * ctx->pictures_per_frame,
yading@10:            interlaced ? "yes" : "no", ctx->bits_per_mb);
yading@10:     av_log(avctx, AV_LOG_DEBUG, "frame size upper bound: %d\n",
yading@10:            ctx->frame_size_upper_bound);
yading@10: 
yading@10:     return 0;
yading@10: }
yading@10: 
yading@10: #define OFFSET(x) offsetof(ProresContext, x)
yading@10: #define VE     AV_OPT_FLAG_VIDEO_PARAM | AV_OPT_FLAG_ENCODING_PARAM
yading@10: 
yading@10: static const AVOption options[] = {
yading@10:     { "mbs_per_slice", "macroblocks per slice", OFFSET(mbs_per_slice),
yading@10:         AV_OPT_TYPE_INT, { .i64 = 8 }, 1, MAX_MBS_PER_SLICE, VE },
yading@10:     { "profile",       NULL, OFFSET(profile), AV_OPT_TYPE_INT,
yading@10:         { .i64 = PRORES_PROFILE_STANDARD },
yading@10:         PRORES_PROFILE_PROXY, PRORES_PROFILE_HQ, VE, "profile" },
yading@10:     { "proxy",         NULL, 0, AV_OPT_TYPE_CONST, { .i64 = PRORES_PROFILE_PROXY },
yading@10:         0, 0, VE, "profile" },
yading@10:     { "lt",            NULL, 0, AV_OPT_TYPE_CONST, { .i64 = PRORES_PROFILE_LT },
yading@10:         0, 0, VE, "profile" },
yading@10:     { "standard",      NULL, 0, AV_OPT_TYPE_CONST, { .i64 = PRORES_PROFILE_STANDARD },
yading@10:         0, 0, VE, "profile" },
yading@10:     { "hq",            NULL, 0, AV_OPT_TYPE_CONST, { .i64 = PRORES_PROFILE_HQ },
yading@10:         0, 0, VE, "profile" },
yading@10:     { "vendor", "vendor ID", OFFSET(vendor),
yading@10:         AV_OPT_TYPE_STRING, { .str = "Lavc" }, CHAR_MIN, CHAR_MAX, VE },
yading@10:     { "bits_per_mb", "desired bits per macroblock", OFFSET(bits_per_mb),
yading@10:         AV_OPT_TYPE_INT, { .i64 = 0 }, 0, 8192, VE },
yading@10:     { "quant_mat", "quantiser matrix", OFFSET(quant_sel), AV_OPT_TYPE_INT,
yading@10:         { .i64 = -1 }, -1, QUANT_MAT_DEFAULT, VE, "quant_mat" },
yading@10:     { "auto",          NULL, 0, AV_OPT_TYPE_CONST, { .i64 = -1 },
yading@10:         0, 0, VE, "quant_mat" },
yading@10:     { "proxy",         NULL, 0, AV_OPT_TYPE_CONST, { .i64 = QUANT_MAT_PROXY },
yading@10:         0, 0, VE, "quant_mat" },
yading@10:     { "lt",            NULL, 0, AV_OPT_TYPE_CONST, { .i64 = QUANT_MAT_LT },
yading@10:         0, 0, VE, "quant_mat" },
yading@10:     { "standard",      NULL, 0, AV_OPT_TYPE_CONST, { .i64 = QUANT_MAT_STANDARD },
yading@10:         0, 0, VE, "quant_mat" },
yading@10:     { "hq",            NULL, 0, AV_OPT_TYPE_CONST, { .i64 = QUANT_MAT_HQ },
yading@10:         0, 0, VE, "quant_mat" },
yading@10:     { "default",       NULL, 0, AV_OPT_TYPE_CONST, { .i64 = QUANT_MAT_DEFAULT },
yading@10:         0, 0, VE, "quant_mat" },
yading@10:     { NULL }
yading@10: };
yading@10: 
yading@10: static const AVClass proresenc_class = {
yading@10:     .class_name = "ProRes encoder",
yading@10:     .item_name  = av_default_item_name,
yading@10:     .option     = options,
yading@10:     .version    = LIBAVUTIL_VERSION_INT,
yading@10: };
yading@10: 
yading@10: AVCodec ff_prores_ks_encoder = {
yading@10:     .name           = "prores_ks",
yading@10:     .type           = AVMEDIA_TYPE_VIDEO,
yading@10:     .id             = AV_CODEC_ID_PRORES,
yading@10:     .priv_data_size = sizeof(ProresContext),
yading@10:     .init           = encode_init,
yading@10:     .close          = encode_close,
yading@10:     .encode2        = encode_frame,
yading@10:     .capabilities   = CODEC_CAP_SLICE_THREADS,
yading@10:     .long_name      = NULL_IF_CONFIG_SMALL("Apple ProRes (iCodec Pro)"),
yading@10:     .pix_fmts       = (const enum AVPixelFormat[]) {
yading@10:                           AV_PIX_FMT_YUV422P10, AV_PIX_FMT_YUV444P10, AV_PIX_FMT_NONE
yading@10:                       },
yading@10:     .priv_class     = &proresenc_class,
yading@10: };