summaryrefslogtreecommitdiff
path: root/libavcodec/dnxhddec.c
diff options
context:
space:
mode:
Diffstat (limited to 'libavcodec/dnxhddec.c')
-rw-r--r--libavcodec/dnxhddec.c352
1 files changed, 231 insertions, 121 deletions
diff --git a/libavcodec/dnxhddec.c b/libavcodec/dnxhddec.c
index c1bf3afac8..e3dec78f7f 100644
--- a/libavcodec/dnxhddec.c
+++ b/libavcodec/dnxhddec.c
@@ -2,23 +2,25 @@
* VC3/DNxHD decoder.
* Copyright (c) 2007 SmartJog S.A., Baptiste Coudurier <baptiste dot coudurier at smartjog dot com>
* Copyright (c) 2011 MirriAd Ltd
+ * Copyright (c) 2015 Christophe Gisquet
*
* 10 bit support added by MirriAd Ltd, Joseph Artsimovich <joseph@mirriad.com>
+ * Slice multithreading and MB interlaced support added by Christophe Gisquet
*
- * This file is part of Libav.
+ * This file is part of FFmpeg.
*
- * Libav is free software; you can redistribute it and/or
+ * 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.
*
- * Libav is distributed in the hope that it will be useful,
+ * 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 Libav; if not, write to the Free Software
+ * License along with FFmpeg; if not, write to the Free Software
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/
@@ -26,52 +28,73 @@
#include "libavutil/timer.h"
#include "avcodec.h"
#include "blockdsp.h"
+#define UNCHECKED_BITSTREAM_READER 1
#include "get_bits.h"
#include "dnxhddata.h"
#include "idctdsp.h"
#include "internal.h"
+#include "thread.h"
+
+typedef struct RowContext {
+ DECLARE_ALIGNED(16, int16_t, blocks)[12][64];
+ int luma_scale[64];
+ int chroma_scale[64];
+ GetBitContext gb;
+ int last_dc[3];
+ int last_qscale;
+ int errors;
+} RowContext;
typedef struct DNXHDContext {
AVCodecContext *avctx;
- GetBitContext gb;
+ RowContext *rows;
BlockDSPContext bdsp;
- int cid; ///< compression id
+ const uint8_t* buf;
+ int buf_size;
+ int64_t cid; ///< compression id
unsigned int width, height;
+ enum AVPixelFormat pix_fmt;
unsigned int mb_width, mb_height;
uint32_t mb_scan_index[68]; /* max for 1080p */
int cur_field; ///< current interlaced field
VLC ac_vlc, dc_vlc, run_vlc;
- int last_dc[3];
IDCTDSPContext idsp;
- DECLARE_ALIGNED(16, int16_t, blocks)[12][64];
ScanTable scantable;
const CIDEntry *cid_table;
int bit_depth; // 8, 10 or 0 if not initialized at all.
int is_444;
int mbaff;
- void (*decode_dct_block)(struct DNXHDContext *ctx, int16_t *block,
- int n, int qscale);
+ int act;
+ int (*decode_dct_block)(const struct DNXHDContext *ctx,
+ RowContext *row, int n);
} DNXHDContext;
#define DNXHD_VLC_BITS 9
#define DNXHD_DC_VLC_BITS 7
-static void dnxhd_decode_dct_block_8(DNXHDContext *ctx, int16_t *block,
- int n, int qscale);
-static void dnxhd_decode_dct_block_10(DNXHDContext *ctx, int16_t *block,
- int n, int qscale);
-static void dnxhd_decode_dct_block_10_444(DNXHDContext *ctx, int16_t *block,
- int n, int qscale);
+static int dnxhd_decode_dct_block_8(const DNXHDContext *ctx,
+ RowContext *row, int n);
+static int dnxhd_decode_dct_block_10(const DNXHDContext *ctx,
+ RowContext *row, int n);
+static int dnxhd_decode_dct_block_10_444(const DNXHDContext *ctx,
+ RowContext *row, int n);
static av_cold int dnxhd_decode_init(AVCodecContext *avctx)
{
DNXHDContext *ctx = avctx->priv_data;
ctx->avctx = avctx;
+ ctx->cid = -1;
+ avctx->colorspace = AVCOL_SPC_BT709;
+
+ ctx->rows = av_mallocz_array(avctx->thread_count, sizeof(RowContext));
+ if (!ctx->rows)
+ return AVERROR(ENOMEM);
+
return 0;
}
-static int dnxhd_init_vlc(DNXHDContext *ctx, int cid)
+static int dnxhd_init_vlc(DNXHDContext *ctx, uint32_t cid)
{
if (cid != ctx->cid) {
int index;
@@ -80,6 +103,10 @@ static int dnxhd_init_vlc(DNXHDContext *ctx, int cid)
av_log(ctx->avctx, AV_LOG_ERROR, "unsupported cid %d\n", cid);
return AVERROR(ENOSYS);
}
+ if (ff_dnxhd_cid_table[index].bit_depth != ctx->bit_depth) {
+ av_log(ctx->avctx, AV_LOG_ERROR, "bit depth mismatches %d %d\n", ff_dnxhd_cid_table[index].bit_depth, ctx->bit_depth);
+ return AVERROR_INVALIDDATA;
+ }
ctx->cid_table = &ff_dnxhd_cid_table[index];
av_log(ctx->avctx, AV_LOG_VERBOSE, "Profile cid %d.\n", cid);
@@ -104,6 +131,20 @@ static int dnxhd_init_vlc(DNXHDContext *ctx, int cid)
return 0;
}
+static av_cold int dnxhd_decode_init_thread_copy(AVCodecContext *avctx)
+{
+ DNXHDContext *ctx = avctx->priv_data;
+
+ // make sure VLC tables will be loaded when cid is parsed
+ ctx->cid = -1;
+
+ ctx->rows = av_mallocz_array(avctx->thread_count, sizeof(RowContext));
+ if (!ctx->rows)
+ return AVERROR(ENOMEM);
+
+ return 0;
+}
+
static int dnxhd_decode_header(DNXHDContext *ctx, AVFrame *frame,
const uint8_t *buf, int buf_size,
int first_field)
@@ -114,8 +155,8 @@ static int dnxhd_decode_header(DNXHDContext *ctx, AVFrame *frame,
int old_bit_depth = ctx->bit_depth;
if (buf_size < 0x280) {
- av_log(ctx->avctx, AV_LOG_ERROR, "buffer too small (%d < 640).\n",
- buf_size);
+ av_log(ctx->avctx, AV_LOG_ERROR,
+ "buffer too small (%d < 640).\n", buf_size);
return AVERROR_INVALIDDATA;
}
@@ -131,6 +172,8 @@ static int dnxhd_decode_header(DNXHDContext *ctx, AVFrame *frame,
frame->top_field_first = first_field ^ ctx->cur_field;
av_log(ctx->avctx, AV_LOG_DEBUG,
"interlaced %d, cur field %d\n", buf[5] & 3, ctx->cur_field);
+ } else {
+ ctx->cur_field = 0;
}
ctx->mbaff = buf[0x6] & 32;
@@ -144,20 +187,22 @@ static int dnxhd_decode_header(DNXHDContext *ctx, AVFrame *frame,
if (buf[0x4] == 0x2) {
ctx->decode_dct_block = dnxhd_decode_dct_block_10_444;
- ctx->avctx->pix_fmt = AV_PIX_FMT_YUV444P10;
+ ctx->pix_fmt = AV_PIX_FMT_YUV444P10;
ctx->is_444 = 1;
} else {
ctx->decode_dct_block = dnxhd_decode_dct_block_10;
- ctx->avctx->pix_fmt = AV_PIX_FMT_YUV422P10;
+ ctx->pix_fmt = AV_PIX_FMT_YUV422P10;
+ ctx->is_444 = 0;
}
} else if (buf[0x21] == 0x38) { /* 8 bit */
ctx->bit_depth = ctx->avctx->bits_per_raw_sample = 8;
- ctx->avctx->pix_fmt = AV_PIX_FMT_YUV422P;
+ ctx->pix_fmt = AV_PIX_FMT_YUV422P;
+ ctx->is_444 = 0;
ctx->decode_dct_block = dnxhd_decode_dct_block_8;
} else {
- av_log(ctx->avctx, AV_LOG_ERROR, "invalid bit depth value (%d).\n",
- buf[0x21]);
+ av_log(ctx->avctx, AV_LOG_ERROR,
+ "invalid bit depth value (%d).\n", buf[0x21]);
return AVERROR_INVALIDDATA;
}
if (ctx->bit_depth != old_bit_depth) {
@@ -174,6 +219,11 @@ static int dnxhd_decode_header(DNXHDContext *ctx, AVFrame *frame,
av_log(ctx->avctx, AV_LOG_WARNING,
"Adaptive MB interlace flag in an unsupported profile.\n");
+ ctx->act = buf[0x2C] & 7;
+ if (ctx->act && ctx->cid_table->cid != 1256)
+ av_log(ctx->avctx, AV_LOG_WARNING,
+ "Adaptive color transform in an unsupported profile.\n");
+
// make sure profile size constraints are respected
// DNx100 allows 1920->1440 and 1280->960 subsampling
if (ctx->width != ctx->cid_table->width) {
@@ -208,7 +258,7 @@ static int dnxhd_decode_header(DNXHDContext *ctx, AVFrame *frame,
for (i = 0; i < ctx->mb_height; i++) {
ctx->mb_scan_index[i] = AV_RB32(buf + 0x170 + (i << 2));
ff_dlog(ctx->avctx, "mb scan index %d\n", ctx->mb_scan_index[i]);
- if (buf_size < ctx->mb_scan_index[i] + 0x280) {
+ if (buf_size < ctx->mb_scan_index[i] + 0x280LL) {
av_log(ctx->avctx, AV_LOG_ERROR,
"invalid mb scan index (%d < %d).\n",
buf_size, ctx->mb_scan_index[i] + 0x280);
@@ -219,132 +269,162 @@ static int dnxhd_decode_header(DNXHDContext *ctx, AVFrame *frame,
return 0;
}
-static av_always_inline void dnxhd_decode_dct_block(DNXHDContext *ctx,
- int16_t *block, int n,
- int qscale,
- int index_bits,
- int level_bias,
- int level_shift)
+static av_always_inline int dnxhd_decode_dct_block(const DNXHDContext *ctx,
+ RowContext *row,
+ int n,
+ int index_bits,
+ int level_bias,
+ int level_shift)
{
- int i, j, index1, index2, len;
+ int i, j, index1, index2, len, flags;
int level, component, sign;
+ const int *scale;
const uint8_t *weight_matrix;
- OPEN_READER(bs, &ctx->gb);
+ const uint8_t *ac_level = ctx->cid_table->ac_level;
+ const uint8_t *ac_flags = ctx->cid_table->ac_flags;
+ int16_t *block = row->blocks[n];
+ const int eob_index = ctx->cid_table->eob_index;
+ int ret = 0;
+ OPEN_READER(bs, &row->gb);
+
+ ctx->bdsp.clear_block(block);
if (!ctx->is_444) {
if (n & 2) {
component = 1 + (n & 1);
+ scale = row->chroma_scale;
weight_matrix = ctx->cid_table->chroma_weight;
} else {
component = 0;
+ scale = row->luma_scale;
weight_matrix = ctx->cid_table->luma_weight;
}
} else {
component = (n >> 1) % 3;
if (component) {
+ scale = row->chroma_scale;
weight_matrix = ctx->cid_table->chroma_weight;
} else {
+ scale = row->luma_scale;
weight_matrix = ctx->cid_table->luma_weight;
}
}
- UPDATE_CACHE(bs, &ctx->gb);
- GET_VLC(len, bs, &ctx->gb, ctx->dc_vlc.table, DNXHD_DC_VLC_BITS, 1);
+ UPDATE_CACHE(bs, &row->gb);
+ GET_VLC(len, bs, &row->gb, ctx->dc_vlc.table, DNXHD_DC_VLC_BITS, 1);
if (len) {
- level = GET_CACHE(bs, &ctx->gb);
- LAST_SKIP_BITS(bs, &ctx->gb, len);
+ level = GET_CACHE(bs, &row->gb);
+ LAST_SKIP_BITS(bs, &row->gb, len);
sign = ~level >> 31;
level = (NEG_USR32(sign ^ level, len) ^ sign) - sign;
- ctx->last_dc[component] += level;
+ row->last_dc[component] += level;
}
- block[0] = ctx->last_dc[component];
+ block[0] = row->last_dc[component];
- for (i = 1; ; i++) {
- UPDATE_CACHE(bs, &ctx->gb);
- GET_VLC(index1, bs, &ctx->gb, ctx->ac_vlc.table,
- DNXHD_VLC_BITS, 2);
- level = ctx->cid_table->ac_level[index1];
- if (!level) /* EOB */
- break;
+ i = 0;
+
+ UPDATE_CACHE(bs, &row->gb);
+ GET_VLC(index1, bs, &row->gb, ctx->ac_vlc.table,
+ DNXHD_VLC_BITS, 2);
+
+ while (index1 != eob_index) {
+ level = ac_level[index1];
+ flags = ac_flags[index1];
- sign = SHOW_SBITS(bs, &ctx->gb, 1);
- SKIP_BITS(bs, &ctx->gb, 1);
+ sign = SHOW_SBITS(bs, &row->gb, 1);
+ SKIP_BITS(bs, &row->gb, 1);
- if (ctx->cid_table->ac_index_flag[index1]) {
- level += SHOW_UBITS(bs, &ctx->gb, index_bits) << 6;
- SKIP_BITS(bs, &ctx->gb, index_bits);
+ if (flags & 1) {
+ level += SHOW_UBITS(bs, &row->gb, index_bits) << 7;
+ SKIP_BITS(bs, &row->gb, index_bits);
}
- if (ctx->cid_table->ac_run_flag[index1]) {
- UPDATE_CACHE(bs, &ctx->gb);
- GET_VLC(index2, bs, &ctx->gb, ctx->run_vlc.table,
+ if (flags & 2) {
+ UPDATE_CACHE(bs, &row->gb);
+ GET_VLC(index2, bs, &row->gb, ctx->run_vlc.table,
DNXHD_VLC_BITS, 2);
i += ctx->cid_table->run[index2];
}
- if (i > 63) {
+ if (++i > 63) {
av_log(ctx->avctx, AV_LOG_ERROR, "ac tex damaged %d, %d\n", n, i);
+ ret = -1;
break;
}
j = ctx->scantable.permutated[i];
- level = (2 * level + 1) * qscale * weight_matrix[i];
+ level *= scale[i];
if (level_bias < 32 || weight_matrix[i] != level_bias)
level += level_bias;
level >>= level_shift;
block[j] = (level ^ sign) - sign;
+
+ UPDATE_CACHE(bs, &row->gb);
+ GET_VLC(index1, bs, &row->gb, ctx->ac_vlc.table,
+ DNXHD_VLC_BITS, 2);
}
- CLOSE_READER(bs, &ctx->gb);
+ CLOSE_READER(bs, &row->gb);
+ return ret;
}
-static void dnxhd_decode_dct_block_8(DNXHDContext *ctx, int16_t *block,
- int n, int qscale)
+static int dnxhd_decode_dct_block_8(const DNXHDContext *ctx,
+ RowContext *row, int n)
{
- dnxhd_decode_dct_block(ctx, block, n, qscale, 4, 32, 6);
+ return dnxhd_decode_dct_block(ctx, row, n, 4, 32, 6);
}
-static void dnxhd_decode_dct_block_10(DNXHDContext *ctx, int16_t *block,
- int n, int qscale)
+static int dnxhd_decode_dct_block_10(const DNXHDContext *ctx,
+ RowContext *row, int n)
{
- dnxhd_decode_dct_block(ctx, block, n, qscale, 6, 8, 4);
+ return dnxhd_decode_dct_block(ctx, row, n, 6, 8, 4);
}
-static void dnxhd_decode_dct_block_10_444(DNXHDContext *ctx, int16_t *block,
- int n, int qscale)
+static int dnxhd_decode_dct_block_10_444(const DNXHDContext *ctx,
+ RowContext *row, int n)
{
- dnxhd_decode_dct_block(ctx, block, n, qscale, 6, 32, 6);
+ return dnxhd_decode_dct_block(ctx, row, n, 6, 32, 6);
}
-static int dnxhd_decode_macroblock(DNXHDContext *ctx, AVFrame *frame,
- int x, int y)
+static int dnxhd_decode_macroblock(const DNXHDContext *ctx, RowContext *row,
+ AVFrame *frame, int x, int y)
{
int shift1 = ctx->bit_depth == 10;
int dct_linesize_luma = frame->linesize[0];
int dct_linesize_chroma = frame->linesize[1];
uint8_t *dest_y, *dest_u, *dest_v;
int dct_y_offset, dct_x_offset;
- int qscale, i;
+ int qscale, i, act;
int interlaced_mb = 0;
if (ctx->mbaff) {
- interlaced_mb = get_bits1(&ctx->gb);
- qscale = get_bits(&ctx->gb, 10);
+ interlaced_mb = get_bits1(&row->gb);
+ qscale = get_bits(&row->gb, 10);
} else {
- qscale = get_bits(&ctx->gb, 11);
+ qscale = get_bits(&row->gb, 11);
}
- skip_bits1(&ctx->gb);
-
- for (i = 0; i < 8; i++) {
- ctx->bdsp.clear_block(ctx->blocks[i]);
- ctx->decode_dct_block(ctx, ctx->blocks[i], i, qscale);
+ act = get_bits1(&row->gb);
+ if (act) {
+ static int warned = 0;
+ if (!warned) {
+ warned = 1;
+ av_log(ctx->avctx, AV_LOG_ERROR,
+ "Unsupported adaptive color transform, patch welcome.\n");
+ }
}
- if (ctx->is_444) {
- for (; i < 12; i++) {
- ctx->bdsp.clear_block(ctx->blocks[i]);
- ctx->decode_dct_block(ctx, ctx->blocks[i], i, qscale);
+
+ if (qscale != row->last_qscale) {
+ for (i = 0; i < 64; i++) {
+ row->luma_scale[i] = qscale * ctx->cid_table->luma_weight[i];
+ row->chroma_scale[i] = qscale * ctx->cid_table->chroma_weight[i];
}
+ row->last_qscale = qscale;
+ }
+
+ for (i = 0; i < 8 + 4 * ctx->is_444; i++) {
+ if (ctx->decode_dct_block(ctx, row, i) < 0)
+ return AVERROR_INVALIDDATA;
}
if (frame->interlaced_frame) {
@@ -356,7 +436,7 @@ static int dnxhd_decode_macroblock(DNXHDContext *ctx, AVFrame *frame,
dest_u = frame->data[1] + ((y * dct_linesize_chroma) << 4) + (x << (3 + shift1 + ctx->is_444));
dest_v = frame->data[2] + ((y * dct_linesize_chroma) << 4) + (x << (3 + shift1 + ctx->is_444));
- if (ctx->cur_field) {
+ if (frame->interlaced_frame && ctx->cur_field) {
dest_y += frame->linesize[0];
dest_u += frame->linesize[1];
dest_v += frame->linesize[2];
@@ -369,55 +449,62 @@ static int dnxhd_decode_macroblock(DNXHDContext *ctx, AVFrame *frame,
dct_y_offset = interlaced_mb ? frame->linesize[0] : (dct_linesize_luma << 3);
dct_x_offset = 8 << shift1;
if (!ctx->is_444) {
- ctx->idsp.idct_put(dest_y, dct_linesize_luma, ctx->blocks[0]);
- ctx->idsp.idct_put(dest_y + dct_x_offset, dct_linesize_luma, ctx->blocks[1]);
- ctx->idsp.idct_put(dest_y + dct_y_offset, dct_linesize_luma, ctx->blocks[4]);
- ctx->idsp.idct_put(dest_y + dct_y_offset + dct_x_offset, dct_linesize_luma, ctx->blocks[5]);
+ ctx->idsp.idct_put(dest_y, dct_linesize_luma, row->blocks[0]);
+ ctx->idsp.idct_put(dest_y + dct_x_offset, dct_linesize_luma, row->blocks[1]);
+ ctx->idsp.idct_put(dest_y + dct_y_offset, dct_linesize_luma, row->blocks[4]);
+ ctx->idsp.idct_put(dest_y + dct_y_offset + dct_x_offset, dct_linesize_luma, row->blocks[5]);
if (!(ctx->avctx->flags & AV_CODEC_FLAG_GRAY)) {
dct_y_offset = interlaced_mb ? frame->linesize[1] : (dct_linesize_chroma << 3);
- ctx->idsp.idct_put(dest_u, dct_linesize_chroma, ctx->blocks[2]);
- ctx->idsp.idct_put(dest_v, dct_linesize_chroma, ctx->blocks[3]);
- ctx->idsp.idct_put(dest_u + dct_y_offset, dct_linesize_chroma, ctx->blocks[6]);
- ctx->idsp.idct_put(dest_v + dct_y_offset, dct_linesize_chroma, ctx->blocks[7]);
+ ctx->idsp.idct_put(dest_u, dct_linesize_chroma, row->blocks[2]);
+ ctx->idsp.idct_put(dest_v, dct_linesize_chroma, row->blocks[3]);
+ ctx->idsp.idct_put(dest_u + dct_y_offset, dct_linesize_chroma, row->blocks[6]);
+ ctx->idsp.idct_put(dest_v + dct_y_offset, dct_linesize_chroma, row->blocks[7]);
}
} else {
- ctx->idsp.idct_put(dest_y, dct_linesize_luma, ctx->blocks[0]);
- ctx->idsp.idct_put(dest_y + dct_x_offset, dct_linesize_luma, ctx->blocks[1]);
- ctx->idsp.idct_put(dest_y + dct_y_offset, dct_linesize_luma, ctx->blocks[6]);
- ctx->idsp.idct_put(dest_y + dct_y_offset + dct_x_offset, dct_linesize_luma, ctx->blocks[7]);
+ ctx->idsp.idct_put(dest_y, dct_linesize_luma, row->blocks[0]);
+ ctx->idsp.idct_put(dest_y + dct_x_offset, dct_linesize_luma, row->blocks[1]);
+ ctx->idsp.idct_put(dest_y + dct_y_offset, dct_linesize_luma, row->blocks[6]);
+ ctx->idsp.idct_put(dest_y + dct_y_offset + dct_x_offset, dct_linesize_luma, row->blocks[7]);
if (!(ctx->avctx->flags & AV_CODEC_FLAG_GRAY)) {
dct_y_offset = interlaced_mb ? frame->linesize[1] : (dct_linesize_chroma << 3);
- ctx->idsp.idct_put(dest_u, dct_linesize_chroma, ctx->blocks[2]);
- ctx->idsp.idct_put(dest_u + dct_x_offset, dct_linesize_chroma, ctx->blocks[3]);
- ctx->idsp.idct_put(dest_u + dct_y_offset, dct_linesize_chroma, ctx->blocks[8]);
- ctx->idsp.idct_put(dest_u + dct_y_offset + dct_x_offset, dct_linesize_chroma, ctx->blocks[9]);
- ctx->idsp.idct_put(dest_v, dct_linesize_chroma, ctx->blocks[4]);
- ctx->idsp.idct_put(dest_v + dct_x_offset, dct_linesize_chroma, ctx->blocks[5]);
- ctx->idsp.idct_put(dest_v + dct_y_offset, dct_linesize_chroma, ctx->blocks[10]);
- ctx->idsp.idct_put(dest_v + dct_y_offset + dct_x_offset, dct_linesize_chroma, ctx->blocks[11]);
+ ctx->idsp.idct_put(dest_u, dct_linesize_chroma, row->blocks[2]);
+ ctx->idsp.idct_put(dest_u + dct_x_offset, dct_linesize_chroma, row->blocks[3]);
+ ctx->idsp.idct_put(dest_u + dct_y_offset, dct_linesize_chroma, row->blocks[8]);
+ ctx->idsp.idct_put(dest_u + dct_y_offset + dct_x_offset, dct_linesize_chroma, row->blocks[9]);
+ ctx->idsp.idct_put(dest_v, dct_linesize_chroma, row->blocks[4]);
+ ctx->idsp.idct_put(dest_v + dct_x_offset, dct_linesize_chroma, row->blocks[5]);
+ ctx->idsp.idct_put(dest_v + dct_y_offset, dct_linesize_chroma, row->blocks[10]);
+ ctx->idsp.idct_put(dest_v + dct_y_offset + dct_x_offset, dct_linesize_chroma, row->blocks[11]);
}
}
return 0;
}
-static int dnxhd_decode_macroblocks(DNXHDContext *ctx, AVFrame *frame,
- const uint8_t *buf, int buf_size)
+static int dnxhd_decode_row(AVCodecContext *avctx, void *data,
+ int rownb, int threadnb)
{
- int x, y;
- for (y = 0; y < ctx->mb_height; y++) {
- ctx->last_dc[0] =
- ctx->last_dc[1] =
- ctx->last_dc[2] = 1 << (ctx->bit_depth + 2); // for levels +2^(bitdepth-1)
- init_get_bits(&ctx->gb, buf + ctx->mb_scan_index[y], (buf_size - ctx->mb_scan_index[y]) << 3);
- for (x = 0; x < ctx->mb_width; x++) {
- //START_TIMER;
- dnxhd_decode_macroblock(ctx, frame, x, y);
- //STOP_TIMER("decode macroblock");
+ const DNXHDContext *ctx = avctx->priv_data;
+ uint32_t offset = ctx->mb_scan_index[rownb];
+ RowContext *row = ctx->rows + threadnb;
+ int x;
+
+ row->last_dc[0] =
+ row->last_dc[1] =
+ row->last_dc[2] = 1 << (ctx->bit_depth + 2); // for levels +2^(bitdepth-1)
+ init_get_bits(&row->gb, ctx->buf + offset, (ctx->buf_size - offset) << 3);
+ for (x = 0; x < ctx->mb_width; x++) {
+ //START_TIMER;
+ int ret = dnxhd_decode_macroblock(ctx, row, data, x, rownb);
+ if (ret < 0) {
+ row->errors++;
+ return ret;
}
+ //STOP_TIMER("decode macroblock");
}
+
return 0;
}
@@ -427,9 +514,10 @@ static int dnxhd_decode_frame(AVCodecContext *avctx, void *data,
const uint8_t *buf = avpkt->data;
int buf_size = avpkt->size;
DNXHDContext *ctx = avctx->priv_data;
+ ThreadFrame frame = { .f = data };
AVFrame *picture = data;
int first_field = 1;
- int ret;
+ int ret, i;
ff_dlog(avctx, "frame size %d\n", buf_size);
@@ -443,21 +531,27 @@ decode_coding_unit:
avctx->width, avctx->height, ctx->width, ctx->height);
first_field = 1;
}
+ if (avctx->pix_fmt != AV_PIX_FMT_NONE && avctx->pix_fmt != ctx->pix_fmt) {
+ av_log(avctx, AV_LOG_WARNING, "pix_fmt changed: %s -> %s\n",
+ av_get_pix_fmt_name(avctx->pix_fmt), av_get_pix_fmt_name(ctx->pix_fmt));
+ first_field = 1;
+ }
+ avctx->pix_fmt = ctx->pix_fmt;
ret = ff_set_dimensions(avctx, ctx->width, ctx->height);
if (ret < 0)
return ret;
if (first_field) {
- if ((ret = ff_get_buffer(avctx, picture, 0)) < 0) {
- av_log(avctx, AV_LOG_ERROR, "get_buffer() failed\n");
+ if ((ret = ff_thread_get_buffer(avctx, &frame, 0)) < 0)
return ret;
- }
picture->pict_type = AV_PICTURE_TYPE_I;
picture->key_frame = 1;
}
- dnxhd_decode_macroblocks(ctx, picture, buf + 0x280, buf_size - 0x280);
+ ctx->buf_size = buf_size - 0x280;
+ ctx->buf = buf + 0x280;
+ avctx->execute2(avctx, dnxhd_decode_row, picture, NULL, ctx->mb_height);
if (first_field && picture->interlaced_frame) {
buf += ctx->cid_table->coding_unit_size;
@@ -466,8 +560,19 @@ decode_coding_unit:
goto decode_coding_unit;
}
+ ret = 0;
+ for (i = 0; i < avctx->thread_count; i++) {
+ ret += ctx->rows[i].errors;
+ ctx->rows[i].errors = 0;
+ }
+
+ if (ret) {
+ av_log(ctx->avctx, AV_LOG_ERROR, "%d lines with errors\n", ret);
+ return AVERROR_INVALIDDATA;
+ }
+
*got_frame = 1;
- return buf_size;
+ return avpkt->size;
}
static av_cold int dnxhd_decode_close(AVCodecContext *avctx)
@@ -477,6 +582,9 @@ static av_cold int dnxhd_decode_close(AVCodecContext *avctx)
ff_free_vlc(&ctx->ac_vlc);
ff_free_vlc(&ctx->dc_vlc);
ff_free_vlc(&ctx->run_vlc);
+
+ av_freep(&ctx->rows);
+
return 0;
}
@@ -489,5 +597,7 @@ AVCodec ff_dnxhd_decoder = {
.init = dnxhd_decode_init,
.close = dnxhd_decode_close,
.decode = dnxhd_decode_frame,
- .capabilities = AV_CODEC_CAP_DR1,
+ .capabilities = AV_CODEC_CAP_DR1 | AV_CODEC_CAP_FRAME_THREADS |
+ AV_CODEC_CAP_SLICE_THREADS,
+ .init_thread_copy = ONLY_IF_THREADS_ENABLED(dnxhd_decode_init_thread_copy),
};
generated by cgit v1.2.3 (git 2.39.1) at 2024-07-10 17:10:18 +0000