yading@10: /*
yading@10:  * Apple ProRes compatible decoder
yading@10:  *
yading@10:  * Copyright (c) 2010-2011 Maxim Poliakovski
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: /**
yading@10:  * @file
yading@10:  * This is a decoder for Apple ProRes 422 SD/HQ/LT/Proxy and ProRes 4444.
yading@10:  * It is used for storing and editing high definition video data in Apple's Final Cut Pro.
yading@10:  *
yading@10:  * @see http://wiki.multimedia.cx/index.php?title=Apple_ProRes
yading@10:  */
yading@10: 
yading@10: #define LONG_BITSTREAM_READER // some ProRes vlc codes require up to 28 bits to be read at once
yading@10: 
yading@10: #include <stdint.h>
yading@10: 
yading@10: #include "libavutil/intmath.h"
yading@10: #include "avcodec.h"
yading@10: #include "dsputil.h"
yading@10: #include "internal.h"
yading@10: #include "proresdata.h"
yading@10: #include "proresdsp.h"
yading@10: #include "get_bits.h"
yading@10: 
yading@10: typedef struct {
yading@10:     const uint8_t *index;            ///< pointers to the data of this slice
yading@10:     int slice_num;
yading@10:     int x_pos, y_pos;
yading@10:     int slice_width;
yading@10:     int prev_slice_sf;               ///< scalefactor of the previous decoded slice
yading@10:     DECLARE_ALIGNED(16, int16_t, blocks)[8 * 4 * 64];
yading@10:     DECLARE_ALIGNED(16, int16_t, qmat_luma_scaled)[64];
yading@10:     DECLARE_ALIGNED(16, int16_t, qmat_chroma_scaled)[64];
yading@10: } ProresThreadData;
yading@10: 
yading@10: typedef struct {
yading@10:     ProresDSPContext dsp;
yading@10:     AVFrame    *frame;
yading@10:     ScanTable  scantable;
yading@10:     int        scantable_type;           ///< -1 = uninitialized, 0 = progressive, 1/2 = interlaced
yading@10: 
yading@10:     int        frame_type;               ///< 0 = progressive, 1 = top-field first, 2 = bottom-field first
yading@10:     int        pic_format;               ///< 2 = 422, 3 = 444
yading@10:     uint8_t    qmat_luma[64];            ///< dequantization matrix for luma
yading@10:     uint8_t    qmat_chroma[64];          ///< dequantization matrix for chroma
yading@10:     int        qmat_changed;             ///< 1 - global quantization matrices changed
yading@10:     int        total_slices;            ///< total number of slices in a picture
yading@10:     ProresThreadData *slice_data;
yading@10:     int        pic_num;
yading@10:     int        chroma_factor;
yading@10:     int        mb_chroma_factor;
yading@10:     int        num_chroma_blocks;       ///< number of chrominance blocks in a macroblock
yading@10:     int        num_x_slices;
yading@10:     int        num_y_slices;
yading@10:     int        slice_width_factor;
yading@10:     int        slice_height_factor;
yading@10:     int        num_x_mbs;
yading@10:     int        num_y_mbs;
yading@10:     int        alpha_info;
yading@10: } ProresContext;
yading@10: 
yading@10: 
yading@10: static av_cold int decode_init(AVCodecContext *avctx)
yading@10: {
yading@10:     ProresContext *ctx = avctx->priv_data;
yading@10: 
yading@10:     ctx->total_slices     = 0;
yading@10:     ctx->slice_data       = NULL;
yading@10: 
yading@10:     avctx->bits_per_raw_sample = PRORES_BITS_PER_SAMPLE;
yading@10:     ff_proresdsp_init(&ctx->dsp, avctx);
yading@10: 
yading@10:     ctx->scantable_type = -1;   // set scantable type to uninitialized
yading@10:     memset(ctx->qmat_luma, 4, 64);
yading@10:     memset(ctx->qmat_chroma, 4, 64);
yading@10: 
yading@10:     return 0;
yading@10: }
yading@10: 
yading@10: 
yading@10: static int decode_frame_header(ProresContext *ctx, const uint8_t *buf,
yading@10:                                const int data_size, AVCodecContext *avctx)
yading@10: {
yading@10:     int hdr_size, version, width, height, flags;
yading@10:     const uint8_t *ptr;
yading@10: 
yading@10:     hdr_size = AV_RB16(buf);
yading@10:     if (hdr_size > data_size) {
yading@10:         av_log(avctx, AV_LOG_ERROR, "frame data too small\n");
yading@10:         return AVERROR_INVALIDDATA;
yading@10:     }
yading@10: 
yading@10:     version = AV_RB16(buf + 2);
yading@10:     if (version >= 2) {
yading@10:         av_log(avctx, AV_LOG_ERROR,
yading@10:                "unsupported header version: %d\n", version);
yading@10:         return AVERROR_INVALIDDATA;
yading@10:     }
yading@10: 
yading@10:     width  = AV_RB16(buf + 8);
yading@10:     height = AV_RB16(buf + 10);
yading@10:     if (width != avctx->width || height != avctx->height) {
yading@10:         av_log(avctx, AV_LOG_ERROR,
yading@10:                "picture dimension changed: old: %d x %d, new: %d x %d\n",
yading@10:                avctx->width, avctx->height, width, height);
yading@10:         return AVERROR_INVALIDDATA;
yading@10:     }
yading@10: 
yading@10:     ctx->frame_type = (buf[12] >> 2) & 3;
yading@10:     if (ctx->frame_type > 2) {
yading@10:         av_log(avctx, AV_LOG_ERROR,
yading@10:                "unsupported frame type: %d\n", ctx->frame_type);
yading@10:         return AVERROR_INVALIDDATA;
yading@10:     }
yading@10: 
yading@10:     ctx->chroma_factor     = (buf[12] >> 6) & 3;
yading@10:     ctx->mb_chroma_factor  = ctx->chroma_factor + 2;
yading@10:     ctx->num_chroma_blocks = (1 << ctx->chroma_factor) >> 1;
yading@10:     switch (ctx->chroma_factor) {
yading@10:     case 2:
yading@10:         avctx->pix_fmt = AV_PIX_FMT_YUV422P10;
yading@10:         break;
yading@10:     case 3:
yading@10:         avctx->pix_fmt = AV_PIX_FMT_YUV444P10;
yading@10:         break;
yading@10:     default:
yading@10:         av_log(avctx, AV_LOG_ERROR,
yading@10:                "unsupported picture format: %d\n", ctx->pic_format);
yading@10:         return AVERROR_INVALIDDATA;
yading@10:     }
yading@10: 
yading@10:     if (ctx->scantable_type != ctx->frame_type) {
yading@10:         if (!ctx->frame_type)
yading@10:             ff_init_scantable(ctx->dsp.idct_permutation, &ctx->scantable,
yading@10:                               ff_prores_progressive_scan);
yading@10:         else
yading@10:             ff_init_scantable(ctx->dsp.idct_permutation, &ctx->scantable,
yading@10:                               ff_prores_interlaced_scan);
yading@10:         ctx->scantable_type = ctx->frame_type;
yading@10:     }
yading@10: 
yading@10:     if (ctx->frame_type) {      /* if interlaced */
yading@10:         ctx->frame->interlaced_frame = 1;
yading@10:         ctx->frame->top_field_first  = ctx->frame_type & 1;
yading@10:     } else {
yading@10:         ctx->frame->interlaced_frame = 0;
yading@10:     }
yading@10: 
yading@10:     avctx->color_primaries = buf[14];
yading@10:     avctx->color_trc       = buf[15];
yading@10:     avctx->colorspace      = buf[16];
yading@10: 
yading@10:     ctx->alpha_info = buf[17] & 0xf;
yading@10:     if (ctx->alpha_info)
yading@10:         avpriv_report_missing_feature(avctx, "Alpha channel");
yading@10: 
yading@10:     ctx->qmat_changed = 0;
yading@10:     ptr   = buf + 20;
yading@10:     flags = buf[19];
yading@10:     if (flags & 2) {
yading@10:         if (ptr - buf > hdr_size - 64) {
yading@10:             av_log(avctx, AV_LOG_ERROR, "header data too small\n");
yading@10:             return AVERROR_INVALIDDATA;
yading@10:         }
yading@10:         if (memcmp(ctx->qmat_luma, ptr, 64)) {
yading@10:             memcpy(ctx->qmat_luma, ptr, 64);
yading@10:             ctx->qmat_changed = 1;
yading@10:         }
yading@10:         ptr += 64;
yading@10:     } else {
yading@10:         memset(ctx->qmat_luma, 4, 64);
yading@10:         ctx->qmat_changed = 1;
yading@10:     }
yading@10: 
yading@10:     if (flags & 1) {
yading@10:         if (ptr - buf > hdr_size - 64) {
yading@10:             av_log(avctx, AV_LOG_ERROR, "header data too small\n");
yading@10:             return -1;
yading@10:         }
yading@10:         if (memcmp(ctx->qmat_chroma, ptr, 64)) {
yading@10:             memcpy(ctx->qmat_chroma, ptr, 64);
yading@10:             ctx->qmat_changed = 1;
yading@10:         }
yading@10:     } else {
yading@10:         memset(ctx->qmat_chroma, 4, 64);
yading@10:         ctx->qmat_changed = 1;
yading@10:     }
yading@10: 
yading@10:     return hdr_size;
yading@10: }
yading@10: 
yading@10: 
yading@10: static int decode_picture_header(ProresContext *ctx, const uint8_t *buf,
yading@10:                                  const int data_size, AVCodecContext *avctx)
yading@10: {
yading@10:     int   i, hdr_size, pic_data_size, num_slices;
yading@10:     int   slice_width_factor, slice_height_factor;
yading@10:     int   remainder, num_x_slices;
yading@10:     const uint8_t *data_ptr, *index_ptr;
yading@10: 
yading@10:     hdr_size = data_size > 0 ? buf[0] >> 3 : 0;
yading@10:     if (hdr_size < 8 || hdr_size > data_size) {
yading@10:         av_log(avctx, AV_LOG_ERROR, "picture header too small\n");
yading@10:         return AVERROR_INVALIDDATA;
yading@10:     }
yading@10: 
yading@10:     pic_data_size = AV_RB32(buf + 1);
yading@10:     if (pic_data_size > data_size) {
yading@10:         av_log(avctx, AV_LOG_ERROR, "picture data too small\n");
yading@10:         return AVERROR_INVALIDDATA;
yading@10:     }
yading@10: 
yading@10:     slice_width_factor  = buf[7] >> 4;
yading@10:     slice_height_factor = buf[7] & 0xF;
yading@10:     if (slice_width_factor > 3 || slice_height_factor) {
yading@10:         av_log(avctx, AV_LOG_ERROR,
yading@10:                "unsupported slice dimension: %d x %d\n",
yading@10:                1 << slice_width_factor, 1 << slice_height_factor);
yading@10:         return AVERROR_INVALIDDATA;
yading@10:     }
yading@10: 
yading@10:     ctx->slice_width_factor  = slice_width_factor;
yading@10:     ctx->slice_height_factor = slice_height_factor;
yading@10: 
yading@10:     ctx->num_x_mbs = (avctx->width + 15) >> 4;
yading@10:     ctx->num_y_mbs = (avctx->height +
yading@10:                       (1 << (4 + ctx->frame->interlaced_frame)) - 1) >>
yading@10:                      (4 + ctx->frame->interlaced_frame);
yading@10: 
yading@10:     remainder    = ctx->num_x_mbs & ((1 << slice_width_factor) - 1);
yading@10:     num_x_slices = (ctx->num_x_mbs >> slice_width_factor) + (remainder & 1) +
yading@10:                    ((remainder >> 1) & 1) + ((remainder >> 2) & 1);
yading@10: 
yading@10:     num_slices = num_x_slices * ctx->num_y_mbs;
yading@10:     if (num_slices != AV_RB16(buf + 5)) {
yading@10:         av_log(avctx, AV_LOG_ERROR, "invalid number of slices\n");
yading@10:         return AVERROR_INVALIDDATA;
yading@10:     }
yading@10: 
yading@10:     if (ctx->total_slices != num_slices) {
yading@10:         av_freep(&ctx->slice_data);
yading@10:         ctx->slice_data = av_malloc((num_slices + 1) * sizeof(ctx->slice_data[0]));
yading@10:         if (!ctx->slice_data)
yading@10:             return AVERROR(ENOMEM);
yading@10:         ctx->total_slices = num_slices;
yading@10:     }
yading@10: 
yading@10:     if (hdr_size + num_slices * 2 > data_size) {
yading@10:         av_log(avctx, AV_LOG_ERROR, "slice table too small\n");
yading@10:         return AVERROR_INVALIDDATA;
yading@10:     }
yading@10: 
yading@10:     /* parse slice table allowing quick access to the slice data */
yading@10:     index_ptr = buf + hdr_size;
yading@10:     data_ptr = index_ptr + num_slices * 2;
yading@10: 
yading@10:     for (i = 0; i < num_slices; i++) {
yading@10:         ctx->slice_data[i].index = data_ptr;
yading@10:         ctx->slice_data[i].prev_slice_sf = 0;
yading@10:         data_ptr += AV_RB16(index_ptr + i * 2);
yading@10:     }
yading@10:     ctx->slice_data[i].index = data_ptr;
yading@10:     ctx->slice_data[i].prev_slice_sf = 0;
yading@10: 
yading@10:     if (data_ptr > buf + data_size) {
yading@10:         av_log(avctx, AV_LOG_ERROR, "out of slice data\n");
yading@10:         return -1;
yading@10:     }
yading@10: 
yading@10:     return pic_data_size;
yading@10: }
yading@10: 
yading@10: 
yading@10: /**
yading@10:  * Read an unsigned rice/exp golomb codeword.
yading@10:  */
yading@10: static inline int decode_vlc_codeword(GetBitContext *gb, unsigned codebook)
yading@10: {
yading@10:     unsigned int rice_order, exp_order, switch_bits;
yading@10:     unsigned int buf, code;
yading@10:     int log, prefix_len, len;
yading@10: 
yading@10:     OPEN_READER(re, gb);
yading@10:     UPDATE_CACHE(re, gb);
yading@10:     buf = GET_CACHE(re, gb);
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:     log = 31 - av_log2(buf); /* count prefix bits (zeroes) */
yading@10: 
yading@10:     if (log < switch_bits) { /* ok, we got a rice code */
yading@10:         if (!rice_order) {
yading@10:             /* shortcut for faster decoding of rice codes without remainder */
yading@10:             code = log;
yading@10:             LAST_SKIP_BITS(re, gb, log + 1);
yading@10:         } else {
yading@10:             prefix_len = log + 1;
yading@10:             code = (log << rice_order) + NEG_USR32(buf << prefix_len, rice_order);
yading@10:             LAST_SKIP_BITS(re, gb, prefix_len + rice_order);
yading@10:         }
yading@10:     } else { /* otherwise we got a exp golomb code */
yading@10:         len  = (log << 1) - switch_bits + exp_order + 1;
yading@10:         code = NEG_USR32(buf, len) - (1 << exp_order) + (switch_bits << rice_order);
yading@10:         LAST_SKIP_BITS(re, gb, len);
yading@10:     }
yading@10: 
yading@10:     CLOSE_READER(re, gb);
yading@10: 
yading@10:     return code;
yading@10: }
yading@10: 
yading@10: #define LSB2SIGN(x) (-((x) & 1))
yading@10: #define TOSIGNED(x) (((x) >> 1) ^ LSB2SIGN(x))
yading@10: 
yading@10: /**
yading@10:  * Decode DC coefficients for all blocks in a slice.
yading@10:  */
yading@10: static inline void decode_dc_coeffs(GetBitContext *gb, int16_t *out,
yading@10:                                     int nblocks)
yading@10: {
yading@10:     int16_t prev_dc;
yading@10:     int     i, sign;
yading@10:     int16_t delta;
yading@10:     unsigned int code;
yading@10: 
yading@10:     code   = decode_vlc_codeword(gb, FIRST_DC_CB);
yading@10:     out[0] = prev_dc = TOSIGNED(code);
yading@10: 
yading@10:     out   += 64; /* move to the DC coeff of the next block */
yading@10:     delta  = 3;
yading@10: 
yading@10:     for (i = 1; i < nblocks; i++, out += 64) {
yading@10:         code = decode_vlc_codeword(gb, ff_prores_dc_codebook[FFMIN(FFABS(delta), 3)]);
yading@10: 
yading@10:         sign     = -(((delta >> 15) & 1) ^ (code & 1));
yading@10:         delta    = (((code + 1) >> 1) ^ sign) - sign;
yading@10:         prev_dc += delta;
yading@10:         out[0]   = prev_dc;
yading@10:     }
yading@10: }
yading@10: 
yading@10: 
yading@10: /**
yading@10:  * Decode AC coefficients for all blocks in a slice.
yading@10:  */
yading@10: static inline void decode_ac_coeffs(GetBitContext *gb, int16_t *out,
yading@10:                                     int blocks_per_slice,
yading@10:                                     int plane_size_factor,
yading@10:                                     const uint8_t *scan)
yading@10: {
yading@10:     int pos, block_mask, run, level, sign, run_cb_index, lev_cb_index;
yading@10:     int max_coeffs, bits_left;
yading@10: 
yading@10:     /* set initial prediction values */
yading@10:     run   = 4;
yading@10:     level = 2;
yading@10: 
yading@10:     max_coeffs = blocks_per_slice << 6;
yading@10:     block_mask = blocks_per_slice - 1;
yading@10: 
yading@10:     for (pos = blocks_per_slice - 1; pos < max_coeffs;) {
yading@10:         run_cb_index = ff_prores_run_to_cb_index[FFMIN(run, 15)];
yading@10:         lev_cb_index = ff_prores_lev_to_cb_index[FFMIN(level, 9)];
yading@10: 
yading@10:         bits_left = get_bits_left(gb);
yading@10:         if (bits_left <= 0 || (bits_left <= 8 && !show_bits(gb, bits_left)))
yading@10:             return;
yading@10: 
yading@10:         run = decode_vlc_codeword(gb, ff_prores_ac_codebook[run_cb_index]);
yading@10: 
yading@10:         bits_left = get_bits_left(gb);
yading@10:         if (bits_left <= 0 || (bits_left <= 8 && !show_bits(gb, bits_left)))
yading@10:             return;
yading@10: 
yading@10:         level = decode_vlc_codeword(gb, ff_prores_ac_codebook[lev_cb_index]) + 1;
yading@10: 
yading@10:         pos += run + 1;
yading@10:         if (pos >= max_coeffs)
yading@10:             break;
yading@10: 
yading@10:         sign = get_sbits(gb, 1);
yading@10:         out[((pos & block_mask) << 6) + scan[pos >> plane_size_factor]] =
yading@10:             (level ^ sign) - sign;
yading@10:     }
yading@10: }
yading@10: 
yading@10: 
yading@10: /**
yading@10:  * Decode a slice plane (luma or chroma).
yading@10:  */
yading@10: static void decode_slice_plane(ProresContext *ctx, ProresThreadData *td,
yading@10:                                const uint8_t *buf,
yading@10:                                int data_size, uint16_t *out_ptr,
yading@10:                                int linesize, int mbs_per_slice,
yading@10:                                int blocks_per_mb, int plane_size_factor,
yading@10:                                const int16_t *qmat, int is_chroma)
yading@10: {
yading@10:     GetBitContext gb;
yading@10:     int16_t *block_ptr;
yading@10:     int mb_num, blocks_per_slice;
yading@10: 
yading@10:     blocks_per_slice = mbs_per_slice * blocks_per_mb;
yading@10: 
yading@10:     memset(td->blocks, 0, 8 * 4 * 64 * sizeof(*td->blocks));
yading@10: 
yading@10:     init_get_bits(&gb, buf, data_size << 3);
yading@10: 
yading@10:     decode_dc_coeffs(&gb, td->blocks, blocks_per_slice);
yading@10: 
yading@10:     decode_ac_coeffs(&gb, td->blocks, blocks_per_slice,
yading@10:                      plane_size_factor, ctx->scantable.permutated);
yading@10: 
yading@10:     /* inverse quantization, inverse transform and output */
yading@10:     block_ptr = td->blocks;
yading@10: 
yading@10:     if (!is_chroma) {
yading@10:         for (mb_num = 0; mb_num < mbs_per_slice; mb_num++, out_ptr += blocks_per_mb * 4) {
yading@10:             ctx->dsp.idct_put(out_ptr,                    linesize, block_ptr, qmat);
yading@10:             block_ptr += 64;
yading@10:             if (blocks_per_mb > 2) {
yading@10:                 ctx->dsp.idct_put(out_ptr + 8,            linesize, block_ptr, qmat);
yading@10:                 block_ptr += 64;
yading@10:             }
yading@10:             ctx->dsp.idct_put(out_ptr + linesize * 4,     linesize, block_ptr, qmat);
yading@10:             block_ptr += 64;
yading@10:             if (blocks_per_mb > 2) {
yading@10:                 ctx->dsp.idct_put(out_ptr + linesize * 4 + 8, linesize, block_ptr, qmat);
yading@10:                 block_ptr += 64;
yading@10:             }
yading@10:         }
yading@10:     } else {
yading@10:         for (mb_num = 0; mb_num < mbs_per_slice; mb_num++, out_ptr += blocks_per_mb * 4) {
yading@10:             ctx->dsp.idct_put(out_ptr,                    linesize, block_ptr, qmat);
yading@10:             block_ptr += 64;
yading@10:             ctx->dsp.idct_put(out_ptr + linesize * 4,     linesize, block_ptr, qmat);
yading@10:             block_ptr += 64;
yading@10:             if (blocks_per_mb > 2) {
yading@10:                 ctx->dsp.idct_put(out_ptr + 8,            linesize, block_ptr, qmat);
yading@10:                 block_ptr += 64;
yading@10:                 ctx->dsp.idct_put(out_ptr + linesize * 4 + 8, linesize, block_ptr, qmat);
yading@10:                 block_ptr += 64;
yading@10:             }
yading@10:         }
yading@10:     }
yading@10: }
yading@10: 
yading@10: 
yading@10: static int decode_slice(AVCodecContext *avctx, void *tdata)
yading@10: {
yading@10:     ProresThreadData *td = tdata;
yading@10:     ProresContext *ctx = avctx->priv_data;
yading@10:     int mb_x_pos  = td->x_pos;
yading@10:     int mb_y_pos  = td->y_pos;
yading@10:     int pic_num   = ctx->pic_num;
yading@10:     int slice_num = td->slice_num;
yading@10:     int mbs_per_slice = td->slice_width;
yading@10:     const uint8_t *buf;
yading@10:     uint8_t *y_data, *u_data, *v_data;
yading@10:     AVFrame *pic = ctx->frame;
yading@10:     int i, sf, slice_width_factor;
yading@10:     int slice_data_size, hdr_size, y_data_size, u_data_size, v_data_size;
yading@10:     int y_linesize, u_linesize, v_linesize;
yading@10: 
yading@10:     buf             = ctx->slice_data[slice_num].index;
yading@10:     slice_data_size = ctx->slice_data[slice_num + 1].index - buf;
yading@10: 
yading@10:     slice_width_factor = av_log2(mbs_per_slice);
yading@10: 
yading@10:     y_data     = pic->data[0];
yading@10:     u_data     = pic->data[1];
yading@10:     v_data     = pic->data[2];
yading@10:     y_linesize = pic->linesize[0];
yading@10:     u_linesize = pic->linesize[1];
yading@10:     v_linesize = pic->linesize[2];
yading@10: 
yading@10:     if (pic->interlaced_frame) {
yading@10:         if (!(pic_num ^ pic->top_field_first)) {
yading@10:             y_data += y_linesize;
yading@10:             u_data += u_linesize;
yading@10:             v_data += v_linesize;
yading@10:         }
yading@10:         y_linesize <<= 1;
yading@10:         u_linesize <<= 1;
yading@10:         v_linesize <<= 1;
yading@10:     }
yading@10: 
yading@10:     if (slice_data_size < 6) {
yading@10:         av_log(avctx, AV_LOG_ERROR, "slice data too small\n");
yading@10:         return AVERROR_INVALIDDATA;
yading@10:     }
yading@10: 
yading@10:     /* parse slice header */
yading@10:     hdr_size    = buf[0] >> 3;
yading@10:     y_data_size = AV_RB16(buf + 2);
yading@10:     u_data_size = AV_RB16(buf + 4);
yading@10:     v_data_size = hdr_size > 7 ? AV_RB16(buf + 6) :
yading@10:         slice_data_size - y_data_size - u_data_size - hdr_size;
yading@10: 
yading@10:     if (hdr_size + y_data_size + u_data_size + v_data_size > slice_data_size ||
yading@10:         v_data_size < 0 || hdr_size < 6) {
yading@10:         av_log(avctx, AV_LOG_ERROR, "invalid data size\n");
yading@10:         return AVERROR_INVALIDDATA;
yading@10:     }
yading@10: 
yading@10:     sf = av_clip(buf[1], 1, 224);
yading@10:     sf = sf > 128 ? (sf - 96) << 2 : sf;
yading@10: 
yading@10:     /* scale quantization matrixes according with slice's scale factor */
yading@10:     /* TODO: this can be SIMD-optimized a lot */
yading@10:     if (ctx->qmat_changed || sf != td->prev_slice_sf) {
yading@10:         td->prev_slice_sf = sf;
yading@10:         for (i = 0; i < 64; i++) {
yading@10:             td->qmat_luma_scaled[ctx->dsp.idct_permutation[i]]   = ctx->qmat_luma[i]   * sf;
yading@10:             td->qmat_chroma_scaled[ctx->dsp.idct_permutation[i]] = ctx->qmat_chroma[i] * sf;
yading@10:         }
yading@10:     }
yading@10: 
yading@10:     /* decode luma plane */
yading@10:     decode_slice_plane(ctx, td, buf + hdr_size, y_data_size,
yading@10:                        (uint16_t*) (y_data + (mb_y_pos << 4) * y_linesize +
yading@10:                                     (mb_x_pos << 5)), y_linesize,
yading@10:                        mbs_per_slice, 4, slice_width_factor + 2,
yading@10:                        td->qmat_luma_scaled, 0);
yading@10: 
yading@10:     /* decode U chroma plane */
yading@10:     decode_slice_plane(ctx, td, buf + hdr_size + y_data_size, u_data_size,
yading@10:                        (uint16_t*) (u_data + (mb_y_pos << 4) * u_linesize +
yading@10:                                     (mb_x_pos << ctx->mb_chroma_factor)),
yading@10:                        u_linesize, mbs_per_slice, ctx->num_chroma_blocks,
yading@10:                        slice_width_factor + ctx->chroma_factor - 1,
yading@10:                        td->qmat_chroma_scaled, 1);
yading@10: 
yading@10:     /* decode V chroma plane */
yading@10:     decode_slice_plane(ctx, td, buf + hdr_size + y_data_size + u_data_size,
yading@10:                        v_data_size,
yading@10:                        (uint16_t*) (v_data + (mb_y_pos << 4) * v_linesize +
yading@10:                                     (mb_x_pos << ctx->mb_chroma_factor)),
yading@10:                        v_linesize, mbs_per_slice, ctx->num_chroma_blocks,
yading@10:                        slice_width_factor + ctx->chroma_factor - 1,
yading@10:                        td->qmat_chroma_scaled, 1);
yading@10: 
yading@10:     return 0;
yading@10: }
yading@10: 
yading@10: 
yading@10: static int decode_picture(ProresContext *ctx, int pic_num,
yading@10:                           AVCodecContext *avctx)
yading@10: {
yading@10:     int slice_num, slice_width, x_pos, y_pos;
yading@10: 
yading@10:     slice_num = 0;
yading@10: 
yading@10:     ctx->pic_num = pic_num;
yading@10:     for (y_pos = 0; y_pos < ctx->num_y_mbs; y_pos++) {
yading@10:         slice_width = 1 << ctx->slice_width_factor;
yading@10: 
yading@10:         for (x_pos = 0; x_pos < ctx->num_x_mbs && slice_width;
yading@10:              x_pos += slice_width) {
yading@10:             while (ctx->num_x_mbs - x_pos < slice_width)
yading@10:                 slice_width >>= 1;
yading@10: 
yading@10:             ctx->slice_data[slice_num].slice_num   = slice_num;
yading@10:             ctx->slice_data[slice_num].x_pos       = x_pos;
yading@10:             ctx->slice_data[slice_num].y_pos       = y_pos;
yading@10:             ctx->slice_data[slice_num].slice_width = slice_width;
yading@10: 
yading@10:             slice_num++;
yading@10:         }
yading@10:     }
yading@10: 
yading@10:     return avctx->execute(avctx, decode_slice,
yading@10:                           ctx->slice_data, NULL, slice_num,
yading@10:                           sizeof(ctx->slice_data[0]));
yading@10: }
yading@10: 
yading@10: 
yading@10: #define MOVE_DATA_PTR(nbytes) buf += (nbytes); buf_size -= (nbytes)
yading@10: 
yading@10: static int decode_frame(AVCodecContext *avctx, void *data, int *got_frame,
yading@10:                         AVPacket *avpkt)
yading@10: {
yading@10:     ProresContext *ctx = avctx->priv_data;
yading@10:     const uint8_t *buf = avpkt->data;
yading@10:     int buf_size       = avpkt->size;
yading@10:     int frame_hdr_size, pic_num, pic_data_size;
yading@10: 
yading@10:     ctx->frame            = data;
yading@10:     ctx->frame->pict_type = AV_PICTURE_TYPE_I;
yading@10:     ctx->frame->key_frame = 1;
yading@10: 
yading@10:     /* check frame atom container */
yading@10:     if (buf_size < 28 || buf_size < AV_RB32(buf) ||
yading@10:         AV_RB32(buf + 4) != FRAME_ID) {
yading@10:         av_log(avctx, AV_LOG_ERROR, "invalid frame\n");
yading@10:         return AVERROR_INVALIDDATA;
yading@10:     }
yading@10: 
yading@10:     MOVE_DATA_PTR(8);
yading@10: 
yading@10:     frame_hdr_size = decode_frame_header(ctx, buf, buf_size, avctx);
yading@10:     if (frame_hdr_size < 0)
yading@10:         return AVERROR_INVALIDDATA;
yading@10: 
yading@10:     MOVE_DATA_PTR(frame_hdr_size);
yading@10: 
yading@10:     if (ff_get_buffer(avctx, ctx->frame, 0) < 0)
yading@10:         return -1;
yading@10: 
yading@10:     for (pic_num = 0; ctx->frame->interlaced_frame - pic_num + 1; pic_num++) {
yading@10:         pic_data_size = decode_picture_header(ctx, buf, buf_size, avctx);
yading@10:         if (pic_data_size < 0)
yading@10:             return AVERROR_INVALIDDATA;
yading@10: 
yading@10:         if (decode_picture(ctx, pic_num, avctx))
yading@10:             return -1;
yading@10: 
yading@10:         MOVE_DATA_PTR(pic_data_size);
yading@10:     }
yading@10: 
yading@10:     ctx->frame = NULL;
yading@10:     *got_frame = 1;
yading@10: 
yading@10:     return avpkt->size;
yading@10: }
yading@10: 
yading@10: 
yading@10: static av_cold int decode_close(AVCodecContext *avctx)
yading@10: {
yading@10:     ProresContext *ctx = avctx->priv_data;
yading@10: 
yading@10:     av_freep(&ctx->slice_data);
yading@10: 
yading@10:     return 0;
yading@10: }
yading@10: 
yading@10: 
yading@10: AVCodec ff_prores_lgpl_decoder = {
yading@10:     .name           = "prores_lgpl",
yading@10:     .type           = AVMEDIA_TYPE_VIDEO,
yading@10:     .id             = AV_CODEC_ID_PRORES,
yading@10:     .priv_data_size = sizeof(ProresContext),
yading@10:     .init           = decode_init,
yading@10:     .close          = decode_close,
yading@10:     .decode         = decode_frame,
yading@10:     .capabilities   = CODEC_CAP_DR1 | CODEC_CAP_SLICE_THREADS,
yading@10:     .long_name      = NULL_IF_CONFIG_SMALL("Apple ProRes (iCodec Pro)")
yading@10: };