* @see 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
+#define LONG_BITSTREAM_READER // some ProRes vlc codes require up to 28 bits to be read at once
#include <stdint.h>
#include "libavutil/intmath.h"
#include "avcodec.h"
-#include "dsputil.h"
+#include "idctdsp.h"
+#include "internal.h"
+#include "proresdata.h"
+#include "proresdsp.h"
#include "get_bits.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
-
+typedef struct {
+ const uint8_t *index; ///< pointers to the data of this slice
+ int slice_num;
+ int x_pos, y_pos;
+ int slice_width;
+ int prev_slice_sf; ///< scalefactor of the previous decoded slice
+ DECLARE_ALIGNED(16, int16_t, blocks)[8 * 4 * 64];
+ DECLARE_ALIGNED(16, int16_t, qmat_luma_scaled)[64];
+ DECLARE_ALIGNED(16, int16_t, qmat_chroma_scaled)[64];
+} ProresThreadData;
typedef struct {
- DSPContext dsp;
- AVFrame picture;
+ ProresDSPContext dsp;
+ AVFrame *frame;
ScanTable scantable;
int scantable_type; ///< -1 = uninitialized, 0 = progressive, 1/2 = interlaced
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]);
- DECLARE_ALIGNED(16, DCTELEM, blocks[8 * 4 * 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
+ ProresThreadData *slice_data;
+ int pic_num;
int chroma_factor;
int mb_chroma_factor;
int num_chroma_blocks; ///< number of chrominance blocks in a macroblock
int slice_height_factor;
int num_x_mbs;
int num_y_mbs;
+ int alpha_info;
} 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
-};
-
-
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);
+ ctx->slice_data = NULL;
- avctx->coded_frame = &ctx->picture;
- avcodec_get_frame_defaults(&ctx->picture);
- ctx->picture.type = AV_PICTURE_TYPE_I;
- ctx->picture.key_frame = 1;
+ avctx->bits_per_raw_sample = PRORES_BITS_PER_SAMPLE;
+ ff_proresdsp_init(&ctx->dsp);
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;
}
ctx->chroma_factor = (buf[12] >> 6) & 3;
ctx->mb_chroma_factor = ctx->chroma_factor + 2;
ctx->num_chroma_blocks = (1 << ctx->chroma_factor) >> 1;
+ ctx->alpha_info = buf[17] & 0xf;
+
+ if (ctx->alpha_info > 2) {
+ av_log(avctx, AV_LOG_ERROR, "Invalid alpha mode %d\n", ctx->alpha_info);
+ return AVERROR_INVALIDDATA;
+ }
+
switch (ctx->chroma_factor) {
case 2:
- avctx->pix_fmt = PIX_FMT_YUV422P10;
+ avctx->pix_fmt = ctx->alpha_info ? AV_PIX_FMT_YUVA422P10
+ : AV_PIX_FMT_YUV422P10;
break;
case 3:
- avctx->pix_fmt = PIX_FMT_YUV444P10;
+ avctx->pix_fmt = ctx->alpha_info ? AV_PIX_FMT_YUVA444P10
+ : AV_PIX_FMT_YUV444P10;
break;
default:
av_log(avctx, AV_LOG_ERROR,
if (ctx->scantable_type != ctx->frame_type) {
if (!ctx->frame_type)
ff_init_scantable(ctx->dsp.idct_permutation, &ctx->scantable,
- progressive_scan);
+ ff_prores_progressive_scan);
else
ff_init_scantable(ctx->dsp.idct_permutation, &ctx->scantable,
- interlaced_scan);
+ ff_prores_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->frame->interlaced_frame = 1;
+ ctx->frame->top_field_first = ctx->frame_type & 1;
+ } else {
+ ctx->frame->interlaced_frame = 0;
}
+ avctx->color_primaries = buf[14];
+ avctx->color_trc = buf[15];
+ avctx->colorspace = buf[16];
+
ctx->qmat_changed = 0;
ptr = buf + 20;
flags = buf[19];
ctx->num_x_mbs = (avctx->width + 15) >> 4;
ctx->num_y_mbs = (avctx->height +
- (1 << (4 + ctx->picture.interlaced_frame)) - 1) >>
- (4 + ctx->picture.interlaced_frame);
+ (1 << (4 + ctx->frame->interlaced_frame)) - 1) >>
+ (4 + ctx->frame->interlaced_frame);
remainder = ctx->num_x_mbs & ((1 << slice_width_factor) - 1);
num_x_slices = (ctx->num_x_mbs >> slice_width_factor) + (remainder & 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)
+ av_freep(&ctx->slice_data);
+ ctx->slice_data = av_malloc((num_slices + 1) * sizeof(ctx->slice_data[0]));
+ if (!ctx->slice_data)
return AVERROR(ENOMEM);
ctx->total_slices = num_slices;
}
data_ptr = index_ptr + num_slices * 2;
for (i = 0; i < num_slices; i++) {
- ctx->slice_data_index[i] = data_ptr;
+ ctx->slice_data[i].index = data_ptr;
+ ctx->slice_data[i].prev_slice_sf = 0;
data_ptr += AV_RB16(index_ptr + i * 2);
}
- ctx->slice_data_index[i] = data_ptr;
+ ctx->slice_data[i].index = data_ptr;
+ ctx->slice_data[i].prev_slice_sf = 0;
if (data_ptr > buf + data_size) {
av_log(avctx, AV_LOG_ERROR, "out of slice data\n");
/**
* Read an unsigned rice/exp golomb codeword.
*/
-static inline int decode_vlc_codeword(GetBitContext *gb, uint8_t codebook)
+static inline int decode_vlc_codeword(GetBitContext *gb, unsigned codebook)
{
unsigned int rice_order, exp_order, switch_bits;
unsigned int buf, 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] = {
- 0x04, // 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
-};
-
-
/**
* Decode DC coefficients for all blocks in a slice.
*/
-static inline void decode_dc_coeffs(GetBitContext *gb, DCTELEM *out,
+static inline void decode_dc_coeffs(GetBitContext *gb, int16_t *out,
int nblocks)
{
- DCTELEM prev_dc;
+ int16_t prev_dc;
int i, sign;
int16_t delta;
unsigned int code;
delta = 3;
for (i = 1; i < nblocks; i++, out += 64) {
- code = decode_vlc_codeword(gb, dc_codebook[FFMIN(FFABS(delta), 3)]);
+ code = decode_vlc_codeword(gb, ff_prores_dc_codebook[FFMIN(FFABS(delta), 3)]);
sign = -(((delta >> 15) & 1) ^ (code & 1));
delta = (((code + 1) >> 1) ^ sign) - sign;
}
-static uint8_t ac_codebook[7] = {
- 0x04, // 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
- 0x05, // rice_order = 0, exp_golomb_order = 1, switch_bits = 1
- 0x29, // rice_order = 1, exp_golomb_order = 2, switch_bits = 1
- 0x06, // rice_order = 0, exp_golomb_order = 1, switch_bits = 2
- 0x0A, // 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)
+static inline int decode_ac_coeffs(GetBitContext *gb, int16_t *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;
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)];
+ run_cb_index = ff_prores_run_to_cb_index[FFMIN(run, 15)];
+ lev_cb_index = ff_prores_lev_to_cb_index[FFMIN(level, 9)];
bits_left = get_bits_left(gb);
if (bits_left <= 0 || (bits_left <= 8 && !show_bits(gb, bits_left)))
- return;
+ return 0;
- run = decode_vlc_codeword(gb, ac_codebook[run_cb_index]);
+ run = decode_vlc_codeword(gb, ff_prores_ac_codebook[run_cb_index]);
+ if (run < 0)
+ return AVERROR_INVALIDDATA;
bits_left = get_bits_left(gb);
if (bits_left <= 0 || (bits_left <= 8 && !show_bits(gb, bits_left)))
- return;
+ return AVERROR_INVALIDDATA;
- level = decode_vlc_codeword(gb, ac_codebook[lev_cb_index]) + 1;
+ level = decode_vlc_codeword(gb, ff_prores_ac_codebook[lev_cb_index]) + 1;
+ if (level < 0)
+ return AVERROR_INVALIDDATA;
pos += run + 1;
if (pos >= max_coeffs)
out[((pos & block_mask) << 6) + scan[pos >> plane_size_factor]] =
(level ^ sign) - sign;
}
-}
+ return 0;
+}
-#define CLIP_AND_BIAS(x) (av_clip((x) + BIAS, CLIP_MIN, CLIP_MAX))
/**
- * Add bias value, clamp and output pixels of a slice
+ * Decode a slice plane (luma or chroma).
*/
-static void put_pixels(const DCTELEM *in, uint16_t *out, int stride,
- int mbs_per_slice, int blocks_per_mb)
+static int decode_slice_plane(ProresContext *ctx, ProresThreadData *td,
+ 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, int is_chroma)
{
- 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]);
- }
+ GetBitContext gb;
+ int16_t *block_ptr;
+ int mb_num, blocks_per_slice, ret;
+
+ blocks_per_slice = mbs_per_slice * blocks_per_mb;
+
+ memset(td->blocks, 0, 8 * 4 * 64 * sizeof(*td->blocks));
+
+ init_get_bits(&gb, buf, data_size << 3);
+
+ decode_dc_coeffs(&gb, td->blocks, blocks_per_slice);
+
+ ret = decode_ac_coeffs(&gb, td->blocks, blocks_per_slice,
+ plane_size_factor, ctx->scantable.permutated);
+ if (ret < 0)
+ return ret;
+
+ /* inverse quantization, inverse transform and output */
+ block_ptr = td->blocks;
+
+ if (!is_chroma) {
+ for (mb_num = 0; mb_num < mbs_per_slice; mb_num++, out_ptr += blocks_per_mb * 4) {
+ ctx->dsp.idct_put(out_ptr, linesize, block_ptr, qmat);
+ block_ptr += 64;
+ if (blocks_per_mb > 2) {
+ ctx->dsp.idct_put(out_ptr + 8, linesize, block_ptr, qmat);
+ block_ptr += 64;
+ }
+ ctx->dsp.idct_put(out_ptr + linesize * 4, linesize, block_ptr, qmat);
+ block_ptr += 64;
+ if (blocks_per_mb > 2) {
+ ctx->dsp.idct_put(out_ptr + linesize * 4 + 8, linesize, block_ptr, qmat);
+ block_ptr += 64;
+ }
+ }
+ } else {
+ for (mb_num = 0; mb_num < mbs_per_slice; mb_num++, out_ptr += blocks_per_mb * 4) {
+ ctx->dsp.idct_put(out_ptr, linesize, block_ptr, qmat);
+ block_ptr += 64;
+ ctx->dsp.idct_put(out_ptr + linesize * 4, linesize, block_ptr, qmat);
+ block_ptr += 64;
+ if (blocks_per_mb > 2) {
+ ctx->dsp.idct_put(out_ptr + 8, linesize, block_ptr, qmat);
+ block_ptr += 64;
+ ctx->dsp.idct_put(out_ptr + linesize * 4 + 8, linesize, block_ptr, qmat);
+ block_ptr += 64;
}
}
-
- src1 += blocks_per_mb << 6;
- src2 += blocks_per_mb << 6;
- dst1 += blocks_per_mb << 2;
- dst2 += blocks_per_mb << 2;
}
+ return 0;
}
+static void unpack_alpha(GetBitContext *gb, uint16_t *dst, int num_coeffs,
+ const int num_bits)
+{
+ const int mask = (1 << num_bits) - 1;
+ int i, idx, val, alpha_val;
+
+ idx = 0;
+ alpha_val = mask;
+ do {
+ do {
+ if (get_bits1(gb))
+ val = get_bits(gb, num_bits);
+ else {
+ int sign;
+ val = get_bits(gb, num_bits == 16 ? 7 : 4);
+ sign = val & 1;
+ val = (val + 2) >> 1;
+ if (sign)
+ val = -val;
+ }
+ alpha_val = (alpha_val + val) & mask;
+ if (num_bits == 16)
+ dst[idx++] = alpha_val >> 6;
+ else
+ dst[idx++] = (alpha_val << 2) | (alpha_val >> 6);
+ if (idx >= num_coeffs - 1)
+ break;
+ } while (get_bits1(gb));
+ val = get_bits(gb, 4);
+ if (!val)
+ val = get_bits(gb, 11);
+ if (idx + val > num_coeffs)
+ val = num_coeffs - idx;
+ if (num_bits == 16)
+ for (i = 0; i < val; i++)
+ dst[idx++] = alpha_val >> 6;
+ else
+ for (i = 0; i < val; i++)
+ dst[idx++] = (alpha_val << 2) | (alpha_val >> 6);
+ } while (idx < num_coeffs);
+}
+
/**
- * Decode a slice plane (luma or chroma).
+ * Decode alpha slice plane.
*/
-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)
+static void decode_alpha_plane(ProresContext *ctx, ProresThreadData *td,
+ const uint8_t *buf, int data_size,
+ uint16_t *out_ptr, int linesize,
+ int mbs_per_slice)
{
GetBitContext gb;
- DCTELEM *block_ptr;
- int i, blk_num, blocks_per_slice;
+ int i;
+ uint16_t *block_ptr;
- blocks_per_slice = mbs_per_slice * blocks_per_mb;
-
- memset(ctx->blocks, 0, 8 * 4 * 64 * sizeof(*ctx->blocks));
+ memset(td->blocks, 0, 8 * 4 * 64 * sizeof(*td->blocks));
init_get_bits(&gb, buf, data_size << 3);
- decode_dc_coeffs(&gb, ctx->blocks, blocks_per_slice);
+ if (ctx->alpha_info == 2)
+ unpack_alpha(&gb, td->blocks, mbs_per_slice * 4 * 64, 16);
+ else
+ unpack_alpha(&gb, td->blocks, mbs_per_slice * 4 * 64, 8);
- decode_ac_coeffs(&gb, ctx->blocks, blocks_per_slice,
- plane_size_factor, ctx->scantable.permutated);
+ block_ptr = td->blocks;
- /* inverse quantization, inverse transform and output */
- block_ptr = ctx->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 Libav 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);
+ for (i = 0; i < 16; i++) {
+ memcpy(out_ptr, block_ptr, 16 * mbs_per_slice * sizeof(*out_ptr));
+ out_ptr += linesize >> 1;
+ block_ptr += 16 * mbs_per_slice;
}
-
- put_pixels(ctx->blocks, out_ptr, linesize >> 1, mbs_per_slice,
- blocks_per_mb);
}
-
-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)
+static int decode_slice(AVCodecContext *avctx, void *tdata)
{
+ ProresThreadData *td = tdata;
+ ProresContext *ctx = avctx->priv_data;
+ int mb_x_pos = td->x_pos;
+ int mb_y_pos = td->y_pos;
+ int pic_num = ctx->pic_num;
+ int slice_num = td->slice_num;
+ int mbs_per_slice = td->slice_width;
const uint8_t *buf;
- uint8_t *y_data, *u_data, *v_data;
- AVFrame *pic = avctx->coded_frame;
+ uint8_t *y_data, *u_data, *v_data, *a_data;
+ AVFrame *pic = ctx->frame;
int i, sf, 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;
+ int slice_data_size, hdr_size;
+ int y_data_size, u_data_size, v_data_size, a_data_size;
+ int y_linesize, u_linesize, v_linesize, a_linesize;
+ int coff[4];
+ int ret;
- buf = ctx->slice_data_index[slice_num];
- slice_data_size = ctx->slice_data_index[slice_num + 1] - buf;
+ buf = ctx->slice_data[slice_num].index;
+ slice_data_size = ctx->slice_data[slice_num + 1].index - buf;
slice_width_factor = av_log2(mbs_per_slice);
y_data = pic->data[0];
u_data = pic->data[1];
v_data = pic->data[2];
+ a_data = pic->data[3];
y_linesize = pic->linesize[0];
u_linesize = pic->linesize[1];
v_linesize = pic->linesize[2];
+ a_linesize = pic->linesize[3];
if (pic->interlaced_frame) {
if (!(pic_num ^ pic->top_field_first)) {
y_data += y_linesize;
u_data += u_linesize;
v_data += v_linesize;
+ if (a_data)
+ a_data += a_linesize;
}
y_linesize <<= 1;
u_linesize <<= 1;
v_linesize <<= 1;
+ a_linesize <<= 1;
}
+ y_data += (mb_y_pos << 4) * y_linesize + (mb_x_pos << 5);
+ u_data += (mb_y_pos << 4) * u_linesize + (mb_x_pos << ctx->mb_chroma_factor);
+ v_data += (mb_y_pos << 4) * v_linesize + (mb_x_pos << ctx->mb_chroma_factor);
+ if (a_data)
+ a_data += (mb_y_pos << 4) * a_linesize + (mb_x_pos << 5);
if (slice_data_size < 6) {
av_log(avctx, AV_LOG_ERROR, "slice data too small\n");
/* parse slice header */
hdr_size = buf[0] >> 3;
+ coff[0] = hdr_size;
y_data_size = AV_RB16(buf + 2);
+ coff[1] = coff[0] + y_data_size;
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) {
+ coff[2] = coff[1] + u_data_size;
+ v_data_size = hdr_size > 7 ? AV_RB16(buf + 6) : slice_data_size - coff[2];
+ coff[3] = coff[2] + v_data_size;
+ a_data_size = slice_data_size - coff[3];
+
+ /* if V or alpha component size is negative that means that previous
+ component sizes are too large */
+ if (v_data_size < 0 || a_data_size < 0 || hdr_size < 6) {
av_log(avctx, AV_LOG_ERROR, "invalid data size\n");
return AVERROR_INVALIDDATA;
}
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;
+ /* TODO: this can be SIMD-optimized a lot */
+ if (ctx->qmat_changed || sf != td->prev_slice_sf) {
+ td->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;
+ td->qmat_luma_scaled[ctx->dsp.idct_permutation[i]] = ctx->qmat_luma[i] * sf;
+ td->qmat_chroma_scaled[ctx->dsp.idct_permutation[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);
+ ret = decode_slice_plane(ctx, td, buf + coff[0], y_data_size,
+ (uint16_t*) y_data, y_linesize,
+ mbs_per_slice, 4, slice_width_factor + 2,
+ td->qmat_luma_scaled, 0);
+
+ if (ret < 0)
+ return ret;
/* 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);
+ ret = decode_slice_plane(ctx, td, buf + coff[1], u_data_size,
+ (uint16_t*) u_data, u_linesize,
+ mbs_per_slice, ctx->num_chroma_blocks,
+ slice_width_factor + ctx->chroma_factor - 1,
+ td->qmat_chroma_scaled, 1);
+ if (ret < 0)
+ return ret;
/* 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);
+ ret = decode_slice_plane(ctx, td, buf + coff[2], v_data_size,
+ (uint16_t*) v_data, v_linesize,
+ mbs_per_slice, ctx->num_chroma_blocks,
+ slice_width_factor + ctx->chroma_factor - 1,
+ td->qmat_chroma_scaled, 1);
+ if (ret < 0)
+ return ret;
+
+ /* decode alpha plane if available */
+ if (a_data && a_data_size)
+ decode_alpha_plane(ctx, td, buf + coff[3], a_data_size,
+ (uint16_t*) a_data, a_linesize,
+ mbs_per_slice);
return 0;
}
slice_num = 0;
+ ctx->pic_num = pic_num;
for (y_pos = 0; y_pos < ctx->num_y_mbs; y_pos++) {
slice_width = 1 << ctx->slice_width_factor;
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;
+ ctx->slice_data[slice_num].slice_num = slice_num;
+ ctx->slice_data[slice_num].x_pos = x_pos;
+ ctx->slice_data[slice_num].y_pos = y_pos;
+ ctx->slice_data[slice_num].slice_width = slice_width;
slice_num++;
}
}
- return 0;
+ return avctx->execute(avctx, decode_slice,
+ ctx->slice_data, NULL, slice_num,
+ sizeof(ctx->slice_data[0]));
}
-#define FRAME_ID MKBETAG('i', 'c', 'p', 'f')
#define MOVE_DATA_PTR(nbytes) buf += (nbytes); buf_size -= (nbytes)
-static int decode_frame(AVCodecContext *avctx, void *data, int *data_size,
+static int decode_frame(AVCodecContext *avctx, void *data, int *got_frame,
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;
+ ctx->frame = data;
+ ctx->frame->pict_type = AV_PICTURE_TYPE_I;
+ ctx->frame->key_frame = 1;
+
/* check frame atom container */
if (buf_size < 28 || buf_size < AV_RB32(buf) ||
AV_RB32(buf + 4) != FRAME_ID) {
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)
+ if (ff_get_buffer(avctx, ctx->frame, 0) < 0)
return -1;
- for (pic_num = 0; ctx->picture.interlaced_frame - pic_num + 1; pic_num++) {
+ for (pic_num = 0; ctx->frame->interlaced_frame - pic_num + 1; pic_num++) {
pic_data_size = decode_picture_header(ctx, buf, buf_size, avctx);
if (pic_data_size < 0)
return AVERROR_INVALIDDATA;
MOVE_DATA_PTR(pic_data_size);
}
- *data_size = sizeof(AVPicture);
- *(AVFrame*) data = *avctx->coded_frame;
+ ctx->frame = NULL;
+ *got_frame = 1;
return avpkt->size;
}
{
ProresContext *ctx = avctx->priv_data;
- if (ctx->picture.data[0])
- avctx->release_buffer(avctx, &ctx->picture);
-
- av_freep(&ctx->slice_data_index);
+ av_freep(&ctx->slice_data);
return 0;
}
AVCodec ff_prores_decoder = {
.name = "prores",
+ .long_name = NULL_IF_CONFIG_SMALL("Apple ProRes (iCodec Pro)"),
.type = AVMEDIA_TYPE_VIDEO,
- .id = CODEC_ID_PRORES,
+ .id = AV_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)")
+ .capabilities = CODEC_CAP_DR1 | CODEC_CAP_SLICE_THREADS,
};