On Wed, Sep 21, 2011 at 03:42:38AM +0200, Maxim wrote:
> Hi crews,
> 
> Yes, it's no mistake! What you see is just another code implementing
> ProRes decoder.
> 
> Now I'm going to explain that confusion.
> Writing a ProRes compatible decoder began as team project in 2010.
> During this period, I created an internal documentation on the coding
> algorithm (entropy coding, slicing and inverse transform); a part of
> this doc has been published at
> 
> http://wiki.multimedia.cx/index.php?title=Apple_ProRes&action=history
> 
> already in 2010 (see my nickname "Maxpol" for a proof), but not to much
> to avoid an implementation by someone else.
> 
> Later I wrote a working decoder. Due to the lack for interlaced frames
> it was incomplete, but it could handle the most of samples available.
> 
> Further I shared my docs and code with another developer, whose name I
> don't want to mention here. He wrote his own code, heavily based on my work.
> 
> The team suffered from poor communication and political intrigues...

It's funny how you call money here...
 
> A few days ago that other developer has submitted his decoder
> anonymously. I feel that the code, containing a significant part of my
> work, was released without proper credit.

Yes, we've suspected that much.
 
> Therefore I decided to post my code as well (without any parts written
> by the other guy though). Hereby I claim copyright on several parts of
> the ProRes decoder (vlc, headers and slice decoding), because due to my
> hard work it has reached its final form.
> 
> My code may be of particular interest because of:
> - compatible license (LGPL)
> - a full description of the codec (especially its headers and encoding
> algorithms) I'm going to make publicitly available soon

Isn't it mostly in the wiki already?

> - maintainership by a real person instead of an anonymous one
> 
> Therefore I let you, crews, to make a final decision about all that...
> 
> Best regards
> Maxim Poliakovski
> 

And here's a quick review (and it's really nice to see this decoder at last).

> /*
>  * Apple ProRes compatible decoder
>  *
>  * Copyright (c) 2010-2011 Maxim Poliakovski
>  *
>  * This file is part of FFmpeg.

Libav please, FFmpeg has its own decoder.

>  * FFmpeg is free software; you can redistribute it and/or
>  * modify it under the terms of the GNU Lesser General Public
>  * License as published by the Free Software Foundation; either
>  * version 2.1 of the License, or (at your option) any later version.
>  *
>  * FFmpeg is distributed in the hope that it will be useful,
>  * but WITHOUT ANY WARRANTY; without even the implied warranty of
>  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
>  * Lesser General Public License for more details.
>  *
>  * You should have received a copy of the GNU Lesser General Public
>  * License along with FFmpeg; if not, write to the Free Software
>  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 
> USA
>  */
> 
> /**
>  * @file libavcodec/proresdec.c
>  * This is a decoder for Apple ProRes 422 SD/HQ/LT/Proxy and ProRes 4444.
>  * It is used for storing and editing high definition video data in Apple's 
> Final Cut Pro.
>  * For a detailed description click here: 
> http://wiki.multimedia.cx/index.php?title=Apple_ProRes
>  */
> 
> #define A32_BITSTREAM_READER // some ProRes vlc codes require up to 28 bits 
> to be read at once
> 
> #include "avcodec.h"
> #include "get_bits.h"
> #include "dsputil.h"
> #include "libavutil/mathematics.h"
> 
> #define       BITS_PER_SAMPLE 10 ///< output precision of that decoder
> #define BIAS (1 << (BITS_PER_SAMPLE - 1)) ///< bias value for converting 
> signed pixels into unsigned ones
> #define CLIP_MIN (1 << (BITS_PER_SAMPLE - 8)) ///< minimum value for clipping 
> resulting pixels
> #define CLIP_MAX (1 << BITS_PER_SAMPLE) - CLIP_MIN - 1 ///< maximum value for 
> clipping resulting pixels

it would look better with proper vertical alignment 
 
> typedef struct {
>     DSPContext dsp;
>     AVFrame    picture;
>     ScanTable  scantable;
>     int        scantable_type;  ///< -1 = uninitialized, 0 = progressive, 1/2 
> = interlaced
> 
>     int        frame_type;      ///< 0 = progressive, 1 = top-field first, 2 
> = bottom-field first
>     int        pic_format;      ///< 2 = 422, 3 = 444
>     uint8_t    qmat_luma[64];   ///< dequantization matrix for luma
>     uint8_t    qmat_chroma[64]; ///< dequantization matrix for chroma
>     int        qmat_changed;    ///< 1 - global quantization matrices changed
>     int        prev_slice_sf;   ///< scalefactor of the previous decoded slice
>     DECLARE_ALIGNED(16, int16_t, qmat_luma_scaled[64]);
>     DECLARE_ALIGNED(16, int16_t, qmat_chroma_scaled[64]);
>     int        total_slices;    ///< total number of slices in a picture
>     const uint8_t **slice_data_index; ///< array of pointers to the data of 
> each slice
>     int        chroma_factor;
>     int        mb_chroma_factor;
>     int        num_chroma_blocks; ///< number of chrominance blocks in a 
> macroblock
>     int        num_x_slices;
>     int        num_y_slices;
>     int        slice_width_factor;
>     int        slice_height_factor;
>     int        num_x_mbs;
>     int        num_y_mbs;
> } ProresContext;
> 
> 
> static const uint8_t progressive_scan[64] = {
>      0,  1,  8,  9,  2,  3, 10, 11,
>     16, 17, 24, 25, 18, 19, 26, 27,
>      4,  5, 12, 20, 13,  6,  7, 14,
>     21, 28, 29, 22, 15, 23, 30, 31,
>     32, 33, 40, 48, 41, 34, 35, 42,
>     49, 56, 57, 50, 43, 36, 37, 44,
>     51, 58, 59, 52, 45, 38, 39, 46,
>     53, 60, 61, 54, 47, 55, 62, 63
> };
> 
> static const uint8_t interlaced_scan[64] = {
>      0,  8,  1,  9, 16, 24, 17, 25,
>      2, 10,  3, 11, 18, 26, 19, 27,
>     32, 40, 33, 34, 41, 48, 56, 49,
>     42, 35, 43, 50, 57, 58, 51, 59,
>      4, 12,  5,  6, 13, 20, 28, 21,
>     14,  7, 15, 22, 29, 36, 44, 37,
>     30, 23, 31, 38, 45, 52, 60, 53,
>     46, 39, 47, 54, 61, 62, 55, 63
> };
> 
> 
> /**
>  *  ProRes decoder initializations.
>  */
> static av_cold int decode_init(AVCodecContext * avctx)
> {
>     ProresContext *ctx = avctx->priv_data;
> 
>     ctx->total_slices = 0;
>     ctx->slice_data_index = 0;
> 
>     avctx->pix_fmt = PIX_FMT_YUV422P10; // set default pixel format
> 
>     avctx->bits_per_raw_sample = BITS_PER_SAMPLE;
>     dsputil_init(&ctx->dsp, avctx);
> 
>     avctx->coded_frame = &ctx->picture;
>     avcodec_get_frame_defaults(&ctx->picture);
>     ctx->picture.type = AV_PICTURE_TYPE_I;
>     ctx->picture.key_frame = 1;
> 
>     ctx->scantable_type = -1;   // set scantable type to uninitialized
>     memset(ctx->qmat_luma, 4, 64);
>     memset(ctx->qmat_chroma, 4, 64);
>     ctx->prev_slice_sf = 0;
> 
>     return 0;
> }
> 
> 
> /**
>  *  Parse frame header.
>  */
> static int decode_frame_header(ProresContext * ctx, const uint8_t * buf,
>                                const int data_size, AVCodecContext * avctx)
> {
>     int hdr_size, version, width, height, flags;
>     const uint8_t *ptr;
> 
>     hdr_size = AV_RB16(buf);
>     if (hdr_size > data_size) {
>         av_log(avctx, AV_LOG_ERROR, "frame data too short!\n");
>         return -1;
>     }
> 
>     version = AV_RB16(buf + 2);
>     if (version >= 2) {
>         av_log(avctx, AV_LOG_ERROR, "unsupported header version: %d\n",
>                version);
>         return -1;
>     }
> 
>     width = AV_RB16(buf + 8);
>     height = AV_RB16(buf + 10);
>     if (width != avctx->width || height != avctx->height) {
>         av_log(avctx, AV_LOG_ERROR,
>                "picture dimension changed! Old: %d x %d, new: %d x %d\n",
>                avctx->width, avctx->height, width, height);
>         return -1;
>     }
> 
>     ctx->frame_type = (buf[12] >> 2) & 3;
>     if (ctx->frame_type > 2) {
>         av_log(avctx, AV_LOG_ERROR, "unsupported frame type: %d!\n",
>                ctx->frame_type);
>         return -1;
>     }
> 
>     ctx->chroma_factor = (buf[12] >> 6) & 3;
>     ctx->mb_chroma_factor = ctx->chroma_factor + 2;
>     ctx->num_chroma_blocks = (1 << ctx->chroma_factor) >> 1;
>     switch (ctx->chroma_factor) {
>     case 2:
>         avctx->pix_fmt = PIX_FMT_YUV422P10;
>         break;
>     case 3:
>         avctx->pix_fmt = PIX_FMT_YUV444P10;
>         break;
>     default:
>         av_log(avctx, AV_LOG_ERROR, "unsupported picture format: %d!\n",
>                ctx->pic_format);
>         return -1;
>     }
> 
>     if (ctx->scantable_type != ctx->frame_type) {
>         if (!ctx->frame_type)
>             ff_init_scantable(ctx->dsp.idct_permutation, &ctx->scantable,
>                               progressive_scan);
>         else
>             ff_init_scantable(ctx->dsp.idct_permutation, &ctx->scantable,
>                               interlaced_scan);
>         ctx->scantable_type = ctx->frame_type;
>     }
> 
>     if (ctx->frame_type) {      /* if interlaced */
>         ctx->picture.interlaced_frame = 1;
>         ctx->picture.top_field_first = ctx->frame_type & 1;
>     }
> 
>     ctx->qmat_changed = 0;
>     ptr = buf + 20;
>     flags = buf[19];
>     if (flags & 2) {
>         if (memcmp(ctx->qmat_luma, ptr, 64)) {
>             memcpy(ctx->qmat_luma, ptr, 64);
>             ctx->qmat_changed = 1;
>         }
>         ptr += 64;
>     } else {
>         memset(ctx->qmat_luma, 4, 64);
>         ctx->qmat_changed = 1;
>     }
> 
>     if (flags & 1) {
>         if (memcmp(ctx->qmat_chroma, ptr, 64)) {
>             memcpy(ctx->qmat_chroma, ptr, 64);
>             ctx->qmat_changed = 1;
>         }
>     } else {
>         memset(ctx->qmat_chroma, 4, 64);
>         ctx->qmat_changed = 1;
>     }
> 
>     return hdr_size;
> }
> 
> 
> /**
>  *  Parse picture header.
>  */
> static int decode_picture_header(ProresContext * ctx, const uint8_t * buf,
>                                  const int data_size, AVCodecContext * avctx)
> {
>     int   i, hdr_size, pic_data_size, num_slices;
>     int   slice_width_factor, slice_height_factor;
>     int   remainder, num_x_slices;
>     const uint8_t *data_ptr, *index_ptr;
> 
>     hdr_size = data_size > 0 ? buf[0] >> 3 : 0;
>     if (hdr_size < 8 || hdr_size > data_size) {
>         av_log(avctx, AV_LOG_ERROR, "picture header too short!\n");
>         return -1;
>     }
> 
>     pic_data_size = AV_RB32(buf + 1);
>     if (pic_data_size > data_size) {
>         av_log(avctx, AV_LOG_ERROR, "picture data too short!\n");
>         return -1;
>     }
> 
>     slice_width_factor  = buf[7] >> 4;
>     slice_height_factor = buf[7] & 0xF;
>     if (slice_width_factor > 3 || slice_height_factor) {
>         av_log(avctx, AV_LOG_ERROR,
>                "unsupported slice dimension: %d x %d!\n",
>                1 << slice_width_factor, 1 << slice_height_factor);
>         return -1;
>     }
> 
>     ctx->slice_width_factor  = slice_width_factor;
>     ctx->slice_height_factor = slice_height_factor;
> 
>     ctx->num_x_mbs = (avctx->width + 15) >> 4;
>     ctx->num_y_mbs =
>         (avctx->height + 15) >> (4 + ctx->picture.interlaced_frame);
> 
>     remainder    = ctx->num_x_mbs & ((1 << slice_width_factor) - 1);
>     num_x_slices = (ctx->num_x_mbs >> slice_width_factor) + (remainder & 1) +
>                    ((remainder >> 1) & 1) + ((remainder >> 2) & 1);
> 
>     num_slices = num_x_slices * ctx->num_y_mbs;
>     if (num_slices != AV_RB16(buf + 5)) {
>         av_log(avctx, AV_LOG_ERROR, "invalid number of slices!\n");
>         return -1;
>     }
> 
>     if (ctx->total_slices != num_slices) {
>         av_freep(&ctx->slice_data_index);
>         ctx->slice_data_index =
>             av_malloc((num_slices + 1) * sizeof(uint8_t *));
>         if (!ctx->slice_data_index)
>             return AVERROR(ENOMEM);
>         ctx->total_slices = num_slices;
>     }
> 
>     if (hdr_size + num_slices * 2 > data_size) {
>         av_log(avctx, AV_LOG_ERROR, "slice table too short!\n");
>         return -1;
>     }
> 
>     /* parse slice table allowing quick access to the slice data */
>     index_ptr = buf + hdr_size;
>     data_ptr = index_ptr + num_slices * 2;
> 
>     for (i = 0; i < num_slices; i++) {
>         ctx->slice_data_index[i] = data_ptr;
>         data_ptr += AV_RB16(index_ptr + i * 2);
>     }
>     ctx->slice_data_index[i] = data_ptr;
> 
>     if (data_ptr > buf + data_size) {
>         av_log(avctx, AV_LOG_ERROR, "out of slice data!\n");
>         return -1;
>     }
> 
>     return pic_data_size;
> }
> 
> 
> /**
>  *  Read an unsigned rice/exp golomb codeword.
>  */
> static inline int decode_vlc_codeword(GetBitContext * gb, uint8_t codebook)
> {
>     unsigned int rice_order, exp_order, switch_bits;
>     unsigned int buf, code;
>     int log, prefix_len, len;
> 
>     OPEN_READER(re, gb);
>     UPDATE_CACHE(re, gb);
>     buf = GET_CACHE(re, gb);
> 
>     /* number of prefix bits to switch
>        between Rice and expGolomb */
>     switch_bits = (codebook & 3) + 1;
>     rice_order = codebook >> 5;       /* rice code order */
>     exp_order = (codebook >> 2) & 7;  /* exp golomb code order */
> 
>     log = __builtin_clz(buf); /* count prefix bits (zeroes) */

please use standard av_log2() instead

>     if (log < switch_bits) { /* ok, we got a rice code */
>         if (!rice_order) {
>             /* shortcut for faster decoding of rice codes without remainder */
>             code = log;
>             LAST_SKIP_BITS(re, gb, log + 1);
>         } else {
>             prefix_len = log + 1;
>             code = (log << rice_order) + NEG_USR32((buf << prefix_len), 
> rice_order);
>             LAST_SKIP_BITS(re, gb, prefix_len + rice_order);
>         }
>     } else { /* otherwise we got a exp golomb code */
>         len  = (log << 1) - switch_bits + exp_order + 1;
>         code = NEG_USR32(buf, len) - (1 << exp_order) +
>                (switch_bits << rice_order);
>         LAST_SKIP_BITS(re, gb, len);
>     }
> 
>     CLOSE_READER(re, gb);
> 
>     return code;
> }
> 
> #define       LSB2SIGN(x) (-((x) & 1))
> #define TOSIGNED(x) (((x) >> 1) ^ LSB2SIGN(x))
> 
> #define FIRST_DC_CB 0xB8 // rice_order = 5, exp_golomb_order = 6, switch_bits 
> = 0
> 
> static uint8_t dc_codebook[4] = {
>      0x4, // rice_order = 0, exp_golomb_order = 1, switch_bits = 0
>     0x28, // rice_order = 1, exp_golomb_order = 2, switch_bits = 0
>     0x4D, // rice_order = 2, exp_golomb_order = 3, switch_bits = 1
>     0x70  // rice_order = 3, exp_golomb_order = 4, switch_bits = 0
> };

nit: 0x04 would look slightly better
 
> /**
>  *  Decode DC coefficients for all blocks in a slice.
>  */
> static inline void decode_dc_coeffs(GetBitContext * gb, DCTELEM *out,
>                                     int nblocks)
> {
>     DCTELEM prev_dc;
>     int     i, sign;
>     int16_t delta;
>     unsigned int code;
> 
>     code = decode_vlc_codeword(gb, FIRST_DC_CB);
>     out[0] = prev_dc = TOSIGNED(code);
> 
>     out += 64; /* move to the DC coeff of the next block */
> 
>     delta = 3;
> 
>     for (i = 1; i < nblocks; i++, out += 64) {
>         code = decode_vlc_codeword(gb, dc_codebook[FFMIN(FFABS(delta), 3)]);
>         if (code < 0)
>             return;
> 
>         sign = -(((delta >> 15) & 1) ^ (code & 1));
>         delta = (((code + 1) >> 1) ^ sign) - sign;
>         prev_dc += delta;
>         out[0] = prev_dc;
>     }
> }
> 
> 
> static uint8_t ac_codebook[7] = {
>      0x4, // rice_order = 0, exp_golomb_order = 1, switch_bits = 0
>     0x28, // rice_order = 1, exp_golomb_order = 2, switch_bits = 0
>     0x4C, // rice_order = 2, exp_golomb_order = 3, switch_bits = 0
>      0x5, // rice_order = 0, exp_golomb_order = 1, switch_bits = 1
>     0x29, // rice_order = 1, exp_golomb_order = 2, switch_bits = 1
>      0x6, // rice_order = 0, exp_golomb_order = 1, switch_bits = 2
>      0xA, // rice_order = 0, exp_golomb_order = 2, switch_bits = 2
> };
> 
> /**
>     Lookup tables for adaptive switching between codebooks
>     according with previous run/level value.
>  */
> static uint8_t run_to_cb_index[16] =
>     { 5, 5, 3, 3, 0, 4, 4, 4, 4, 1, 1, 1, 1, 1, 1, 2 };
> 
> static uint8_t lev_to_cb_index[10] = { 0, 6, 3, 5, 0, 1, 1, 1, 1, 2 };
> 
> 
> /**
>  *  Decode AC coefficients for all blocks in a slice.
>  */
> static inline void decode_ac_coeffs(GetBitContext * gb, DCTELEM * out,
>                                     int blocks_per_slice,
>                                     int plane_size_factor,
>                                     const uint8_t * scan)
> {
>     int pos, block_mask, run, level, sign, run_cb_index, lev_cb_index;
>     int max_coeffs, bits_left;
> 
>     /* set initial prediction values */
>     run = 4;
>     level = 2;
>     max_coeffs = blocks_per_slice << 6;
>     block_mask = blocks_per_slice - 1;
> 
>     for (pos = blocks_per_slice - 1; pos < max_coeffs;) {
>         run_cb_index = run_to_cb_index[FFMIN(run, 15)];
>         lev_cb_index = lev_to_cb_index[FFMIN(level, 9)];
> 
>         bits_left = get_bits_left(gb);
>         if (bits_left <= 8 && !show_bits(gb, bits_left))
>             return;
> 
>         run = decode_vlc_codeword(gb, ac_codebook[run_cb_index]);
> 
>         bits_left = get_bits_left(gb);
>         if (bits_left <= 8 && !show_bits(gb, bits_left))
>             return;
> 
>         level = decode_vlc_codeword(gb, ac_codebook[lev_cb_index]) + 1;
> 
>         pos += run + 1;
>         if (pos >= max_coeffs)
>             break;
> 
>         sign = get_sbits(gb, 1);
>         out[((pos & block_mask) << 6) + scan[pos >> plane_size_factor]] =
>             (level ^ sign) - sign;
>     }
> }
> 
> 
> #define CLIP_AND_BIAS(x) (av_clip((x) + BIAS, CLIP_MIN, CLIP_MAX))
> 
> /**
>  *  Add bias value, clamp and output pixels of a slice
>  */
> static void put_pixels(const DCTELEM * in, uint16_t * out, int stride,
>                        int mbs_per_slice, int blocks_per_mb)
> {
>     int mb, x, y, src_offset, dst_offset;
>     const DCTELEM *src1, *src2;
>     uint16_t *dst1, *dst2;
> 
>     src1 = in;
>     src2 = in + (blocks_per_mb << 5);
>     dst1 = out;
>     dst2 = out + (stride << 3);
> 
>     for (mb = 0; mb < mbs_per_slice; mb++) {
>         for (y = 0, dst_offset = 0; y < 8; y++, dst_offset += stride) {
>             for (x = 0; x < 8; x++) {
>                 src_offset = (y << 3) + x;
> 
>                 dst1[dst_offset + x] = CLIP_AND_BIAS(src1[src_offset]);
>                 dst2[dst_offset + x] = CLIP_AND_BIAS(src2[src_offset]);
> 
>                 if (blocks_per_mb > 2) {
>                     dst1[dst_offset + x + 8] =
>                         CLIP_AND_BIAS(src1[src_offset + 64]);
>                     dst2[dst_offset + x + 8] =
>                         CLIP_AND_BIAS(src2[src_offset + 64]);
>                 }
>             }
>         }
> 
>         src1 += blocks_per_mb << 6;
>         src2 += blocks_per_mb << 6;
>         dst1 += blocks_per_mb << 2;
>         dst2 += blocks_per_mb << 2;
>     }
> }

why cannot you use put_pixels_clamped() or something similar? This looks like
a perfect candidate for dsputil.
 
> /**
>  *  Decode a slice plane (luma or chroma).
>  */
> static void decode_slice_plane(ProresContext * ctx, const uint8_t * buf,
>                                int data_size, uint16_t * out_ptr,
>                                int linesize, int mbs_per_slice,
>                                int blocks_per_mb, int plane_size_factor,
>                                const int16_t * qmat)
> {
>     GetBitContext gb;
>     DECLARE_ALIGNED(16, DCTELEM, blocks[8 * 4 * 64]);
>     DCTELEM *block_ptr;
>     int i, blk_num, blocks_per_slice;
> 
>     blocks_per_slice = mbs_per_slice * blocks_per_mb;
> 
>     memset(blocks, 0, sizeof(blocks));
> 
>     init_get_bits(&gb, buf, data_size << 3);
> 
>     decode_dc_coeffs(&gb, blocks, blocks_per_slice);
> 
>     decode_ac_coeffs(&gb, blocks, blocks_per_slice,
>                      plane_size_factor, ctx->scantable.permutated);
> 
>     /* inverse quantization, inverse transform and output */
>     block_ptr = blocks;
> 
>     for (blk_num = 0; blk_num < blocks_per_slice;
>          blk_num++, block_ptr += 64) {
>         /* TODO: the correct solution shoud be (block_ptr[i] * qmat[i]) >> 1
>            and the input of the inverse transform should be scaled by 2
>            in order to avoid rounding errors.
>            Due to the fact the existing FFmpeg transforms are incompatible 
> with that
>            input I temporally introduced the coarse solution below...
>          */
>         for (i = 0; i < 64; i++)
>             block_ptr[i] = (block_ptr[i] * qmat[i]) >> 2;
> 
>         ctx->dsp.idct(block_ptr);
>     }
> 
>     put_pixels(blocks, out_ptr, linesize >> 1, mbs_per_slice,
>                blocks_per_mb);
> }
> 
> 
> /**
>  *  Decode a slice.
>  */
> static int decode_slice(ProresContext * ctx, int pic_num, int slice_num,
>                         int mb_x_pos, int mb_y_pos, int mbs_per_slice,
>                         AVCodecContext * avctx)
> {
>     const uint8_t *buf;
>     uint8_t *y_data, *u_data, *v_data;
>     AVFrame *pic = avctx->coded_frame;
>     int i, sf, blocks_per_slice, slice_width_factor;
>     int slice_data_size, hdr_size, y_data_size, u_data_size, v_data_size;
>     int y_linesize, u_linesize, v_linesize;
> 
>     buf = ctx->slice_data_index[slice_num];
>     slice_data_size = ctx->slice_data_index[slice_num + 1] - buf;
> 
>     slice_width_factor = av_log2(mbs_per_slice);
> 
>     y_data = pic->data[0];
>     u_data = pic->data[1];
>     v_data = pic->data[2];
>     y_linesize = pic->linesize[0];
>     u_linesize = pic->linesize[1];
>     v_linesize = pic->linesize[2];
> 
>     if (pic->interlaced_frame) {
>         if (!(pic_num ^ pic->top_field_first)) {
>             y_data += y_linesize;
>             u_data += u_linesize;
>             v_data += v_linesize;
>         }
>         y_linesize <<= 1;
>         u_linesize <<= 1;
>         v_linesize <<= 1;
>     }
> 
>     if (slice_data_size < 6) {
>         av_log(avctx, AV_LOG_ERROR, "slice data too short!\n");
>         return -1;
>     }
> 
>     /* parse slice header */
>     hdr_size = buf[0] >> 3;
>     y_data_size = AV_RB16(buf + 2);
>     u_data_size = AV_RB16(buf + 4);
>     v_data_size = slice_data_size - y_data_size - u_data_size - hdr_size;
> 
>     if (v_data_size < 0 || hdr_size < 6) {
>         av_log(avctx, AV_LOG_ERROR, "invalid data sizes!\n");
>         return -1;
>     }
> 
>     sf = av_clip(buf[1], 1, 224);
>     sf = sf > 128 ? (sf - 96) << 2 : sf;
> 
>     /* scale quantization matrixes according with slice's scale factor */
>     /* TODO: this can be SIMD-optimized alot */
>     if (ctx->qmat_changed || sf != ctx->prev_slice_sf) {
>         ctx->prev_slice_sf = sf;
>         for (i = 0; i < 64; i++) {
>             ctx->qmat_luma_scaled[i] = ctx->qmat_luma[i] * sf;
>             ctx->qmat_chroma_scaled[i] = ctx->qmat_chroma[i] * sf;
>         }
>     }
> 
>     /* decode luma plane */
>     decode_slice_plane(ctx, buf + hdr_size, y_data_size,
>                        (uint16_t *) (y_data + (mb_y_pos << 4) * y_linesize +
>                                      (mb_x_pos << 5)), y_linesize,
>                        mbs_per_slice, 4, slice_width_factor + 2,
>                        ctx->qmat_luma_scaled);
> 
>     /* decode U chroma plane */
>     decode_slice_plane(ctx, buf + hdr_size + y_data_size, u_data_size,
>                        (uint16_t *) (u_data + (mb_y_pos << 4) * u_linesize +
>                                      (mb_x_pos << ctx->mb_chroma_factor)),
>                        u_linesize, mbs_per_slice, ctx->num_chroma_blocks,
>                        slice_width_factor + ctx->chroma_factor - 1,
>                        ctx->qmat_chroma_scaled);
> 
>     /* decode V chroma plane */
>     decode_slice_plane(ctx, buf + hdr_size + y_data_size + u_data_size,
>                        v_data_size,
>                        (uint16_t *) (v_data + (mb_y_pos << 4) * v_linesize +
>                                      (mb_x_pos << ctx->mb_chroma_factor)),
>                        v_linesize, mbs_per_slice, ctx->num_chroma_blocks,
>                        slice_width_factor + ctx->chroma_factor - 1,
>                        ctx->qmat_chroma_scaled);
> 
>     return 0;
> }
> 
> 
> /**
>  *  decode a picture
>  */
> static int decode_picture(ProresContext * ctx, int pic_num,
>                           AVCodecContext * avctx)
> {
>     int slice_num, slice_width, x_pos, y_pos;
> 
>     slice_num = 0;
> 
>     for (y_pos = 0; y_pos < ctx->num_y_mbs; y_pos++) {
>         slice_width = 1 << ctx->slice_width_factor;
> 
>         for (x_pos = 0; x_pos < ctx->num_x_mbs && slice_width;
>              x_pos += slice_width) {
>             while (ctx->num_x_mbs - x_pos < slice_width)
>                 slice_width >>= 1;
> 
>             if (decode_slice
>                 (ctx, pic_num, slice_num, x_pos, y_pos, slice_width,
>                  avctx) < 0)
>                 return -1;
> 
>             slice_num++;
>         }
>     }
> 
>     return 0;
> }
> 
> 
> #define FRAME_ID MKBETAG('i', 'c', 'p', 'f')
> #define MOVE_DATA_PTR(nbytes) buf += (nbytes); buf_size -= (nbytes)
> 
> /**
>  *  main decoder function
>  */
> static int decode_frame(AVCodecContext * avctx, void *data, int *data_size,
>                         AVPacket * avpkt)
> {
>     ProresContext *ctx = avctx->priv_data;
>     AVFrame *picture = avctx->coded_frame;
>     const uint8_t *buf = avpkt->data;
>     int buf_size = avpkt->size;
>     int frame_hdr_size, pic_num, pic_data_size;
> 
>     /* check frame atom container */
>     if (buf_size < 28 || buf_size < AV_RB32(buf) ||
>         AV_RB32(buf + 4) != FRAME_ID) {
>         av_log(avctx, AV_LOG_ERROR, "invalid frame\n");
>         return -1;
>     }
> 
>     MOVE_DATA_PTR(8);
> 
>     frame_hdr_size = decode_frame_header(ctx, buf, buf_size, avctx);
>     if (frame_hdr_size < 0)
>         return -1;
> 
>     MOVE_DATA_PTR(frame_hdr_size);
> 
>     if (picture->data[0])
>         avctx->release_buffer(avctx, picture);
> 
>     picture->reference = 0;
>     if (avctx->get_buffer(avctx, picture) < 0)
>         return -1;
> 
>     for (pic_num = 0; ctx->picture.interlaced_frame - pic_num + 1;
>          pic_num++) {
>         pic_data_size = decode_picture_header(ctx, buf, buf_size, avctx);
>         if (pic_data_size < 0)
>             return -1;
> 
>         if (decode_picture(ctx, pic_num, avctx))
>             return -1;
> 
>         MOVE_DATA_PTR(pic_data_size);
>     }
> 
>     *data_size = sizeof(AVPicture);
>     *(AVFrame *) data = *avctx->coded_frame;
> 
>     return avpkt->size;
> }
> 
> 
> /**
>  *  Decoder close function.
>  */
> static av_cold int decode_close(AVCodecContext * avctx)
> {
>     ProresContext *ctx = avctx->priv_data;
> 
>     if (ctx->picture.data[0])
>         avctx->release_buffer(avctx, &ctx->picture);
> 
>     av_freep(&ctx->slice_data_index);
> 
>     return 0;
> }
> 
> 
> AVCodec ff_prores_decoder = {
>     .name           = "ProRes",
>     .type           = AVMEDIA_TYPE_VIDEO,
>     .id             = CODEC_ID_PRORES,
>     .priv_data_size = sizeof(ProresContext),
>     .init           = decode_init,
>     .close          = decode_close,
>     .decode         = decode_frame,
>     .capabilities   = CODEC_CAP_DR1,
>     .long_name = NULL_IF_CONFIG_SMALL("Apple ProRes (iCodec Pro)")
> };

> _______________________________________________
> libav-devel mailing list
> [email protected]
> https://lists.libav.org/mailman/listinfo/libav-devel

_______________________________________________
libav-devel mailing list
[email protected]
https://lists.libav.org/mailman/listinfo/libav-devel

Reply via email to