hashirama/uvg266
1
0
Fork 0
mirror of https://github.com/ultravideo/uvg266.git synced 2024-11-27 19:24:06 +00:00
Code Issues Actions Projects Releases Wiki Activity

Split encoderstate.c in multiple files

Browse source
This commit is contained in:
Laurent Fasnacht 2014-06-03 14:47:49 +02:00
parent 3d07f8cc84
commit 961da05235
4 changed files with 1587 additions and 1538 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

776
src/_encoderstatebitstream.c Normal file
View file

@ -0,0 +1,776 @@
/*****************************************************************************
* This file is part of Kvazaar HEVC encoder.
*
* Copyright (C) 2013-2014 Tampere University of Technology and others (see
* COPYING file).
*
* Kvazaar is free software: you can redistribute it and/or modify
* it under the terms of the GNU General Public License version 2 as published
* by the Free Software Foundation.
*
* Kvazaar 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 General Public License for more details.
*
* You should have received a copy of the GNU General Public License
* along with Kvazaar. If not, see <http://www.gnu.org/licenses/>.
****************************************************************************/
//This file MUST NOT BE COMPILED directly. It's included in encoderstate.c
static void encoder_state_write_bitstream_access_unit_delimiter(encoder_state * const encoder_state)
{
bitstream * const stream = &encoder_state->stream;
uint8_t pic_type = encoder_state->global->slicetype == SLICE_I ? 0
: encoder_state->global->slicetype == SLICE_P ? 1
: 2;
WRITE_U(stream, pic_type, 3, "pic_type");
}
static void encoder_state_write_bitstream_aud(encoder_state * const encoder_state)
{
bitstream * const stream = &encoder_state->stream;
encoder_state_write_bitstream_access_unit_delimiter(encoder_state);
nal_write(stream, AUD_NUT, 0, 1);
bitstream_align(stream);
}
static void encoder_state_write_bitstream_PTL(encoder_state * const encoder_state)
{
bitstream * const stream = &encoder_state->stream;
int i;
// PTL
// Profile Tier
WRITE_U(stream, 0, 2, "general_profile_space");
WRITE_U(stream, 0, 1, "general_tier_flag");
// Main Profile == 1
WRITE_U(stream, 1, 5, "general_profile_idc");
/* Compatibility flags should be set at general_profile_idc
* (so with general_profile_idc = 1, compatibility_flag[1] should be 1)
* According to specification, when compatibility_flag[1] is set,
* compatibility_flag[2] should be set too.
*/
WRITE_U(stream, 3<<29, 32, "general_profile_compatibility_flag[]");
WRITE_U(stream, 1, 1, "general_progressive_source_flag");
WRITE_U(stream, 0, 1, "general_interlaced_source_flag");
WRITE_U(stream, 0, 1, "general_non_packed_constraint_flag");
WRITE_U(stream, 0, 1, "general_frame_only_constraint_flag");
WRITE_U(stream, 0, 32, "XXX_reserved_zero_44bits[0..31]");
WRITE_U(stream, 0, 12, "XXX_reserved_zero_44bits[32..43]");
// end Profile Tier
// Level 6.2 (general_level_idc is 30 * 6.2)
WRITE_U(stream, 186, 8, "general_level_idc");
WRITE_U(stream, 0, 1, "sub_layer_profile_present_flag");
WRITE_U(stream, 0, 1, "sub_layer_level_present_flag");
for (i = 1; i < 8; i++) {
WRITE_U(stream, 0, 2, "reserved_zero_2bits");
}
// end PTL
}
static void encoder_state_write_bitstream_vid_parameter_set(encoder_state * const encoder_state)
{
bitstream * const stream = &encoder_state->stream;
int i;
#ifdef _DEBUG
printf("=========== Video Parameter Set ID: 0 ===========\n");
#endif
WRITE_U(stream, 0, 4, "vps_video_parameter_set_id");
WRITE_U(stream, 3, 2, "vps_reserved_three_2bits" );
WRITE_U(stream, 0, 6, "vps_reserved_zero_6bits" );
WRITE_U(stream, 1, 3, "vps_max_sub_layers_minus1");
WRITE_U(stream, 0, 1, "vps_temporal_id_nesting_flag");
WRITE_U(stream, 0xffff, 16, "vps_reserved_ffff_16bits");
encoder_state_write_bitstream_PTL(encoder_state);
WRITE_U(stream, 0, 1, "vps_sub_layer_ordering_info_present_flag");
//for each layer
for (i = 0; i < 1; i++) {
WRITE_UE(stream, 1, "vps_max_dec_pic_buffering");
WRITE_UE(stream, 0, "vps_num_reorder_pics");
WRITE_UE(stream, 0, "vps_max_latency_increase");
}
WRITE_U(stream, 0, 6, "vps_max_nuh_reserved_zero_layer_id");
WRITE_UE(stream, 0, "vps_max_op_sets_minus1");
WRITE_U(stream, 0, 1, "vps_timing_info_present_flag");
//IF timing info
//END IF
WRITE_U(stream, 0, 1, "vps_extension_flag");
}
static void encoder_state_write_bitstream_scaling_list(encoder_state * const encoder_state)
{
const encoder_control * const encoder = encoder_state->encoder_control;
bitstream * const stream = &encoder_state->stream;
uint32_t size_id;
for (size_id = 0; size_id < SCALING_LIST_SIZE_NUM; size_id++) {
int32_t list_id;
for (list_id = 0; list_id < g_scaling_list_num[size_id]; list_id++) {
uint8_t scaling_list_pred_mode_flag = 1;
int32_t pred_list_idx;
int32_t i;
uint32_t ref_matrix_id = UINT32_MAX;
for (pred_list_idx = list_id; pred_list_idx >= 0; pred_list_idx--) {
const int32_t * const pred_list = (list_id == pred_list_idx) ?
scalinglist_get_default(size_id, pred_list_idx) :
encoder->scaling_list.scaling_list_coeff[size_id][pred_list_idx];
if (!memcmp(encoder->scaling_list.scaling_list_coeff[size_id][list_id], pred_list, sizeof(int32_t) * MIN(8, g_scaling_list_size[size_id])) &&
((size_id < SCALING_LIST_16x16) ||
(encoder->scaling_list.scaling_list_dc[size_id][list_id] == encoder->scaling_list.scaling_list_dc[size_id][pred_list_idx]))) {
ref_matrix_id = pred_list_idx;
scaling_list_pred_mode_flag = 0;
break;
}
}
WRITE_U(stream, scaling_list_pred_mode_flag, 1, "scaling_list_pred_mode_flag" );
if (!scaling_list_pred_mode_flag) {
WRITE_UE(stream, list_id - ref_matrix_id, "scaling_list_pred_matrix_id_delta");
} else {
int32_t delta;
const int32_t coef_num = MIN(MAX_MATRIX_COEF_NUM, g_scaling_list_size[size_id]);
const uint32_t * const scan_cg = (size_id == 0) ? g_sig_last_scan_16x16 : g_sig_last_scan_32x32;
int32_t next_coef = 8;
const int32_t * const coef_list = encoder->scaling_list.scaling_list_coeff[size_id][list_id];
if (size_id >= SCALING_LIST_16x16) {
WRITE_SE(stream, encoder->scaling_list.scaling_list_dc[size_id][list_id] - 8, "scaling_list_dc_coef_minus8");
next_coef = encoder->scaling_list.scaling_list_dc[size_id][list_id];
}
for (i = 0; i < coef_num; i++) {
delta = coef_list[scan_cg[i]] - next_coef;
next_coef = coef_list[scan_cg[i]];
if (delta > 127)
delta -= 256;
if (delta < -128)
delta += 256;
WRITE_SE(stream, delta, "scaling_list_delta_coef");
}
}
}
}
}
static void encoder_state_write_bitstream_VUI(encoder_state * const encoder_state)
{
bitstream * const stream = &encoder_state->stream;
const encoder_control * const encoder = encoder_state->encoder_control;
#ifdef _DEBUG
printf("=========== VUI Set ID: 0 ===========\n");
#endif
if (encoder->vui.sar_width > 0 && encoder->vui.sar_height > 0) {
int i;
static const struct
{
uint8_t width;
uint8_t height;
uint8_t idc;
} sar[] = {
// aspect_ratio_idc = 0 -> unspecified
{ 1, 1, 1 }, { 12, 11, 2 }, { 10, 11, 3 }, { 16, 11, 4 },
{ 40, 33, 5 }, { 24, 11, 6 }, { 20, 11, 7 }, { 32, 11, 8 },
{ 80, 33, 9 }, { 18, 11, 10}, { 15, 11, 11}, { 64, 33, 12},
{160, 99, 13}, { 4, 3, 14}, { 3, 2, 15}, { 2, 1, 16},
// aspect_ratio_idc = [17..254] -> reserved
{ 0, 0, 255 }
};
for (i = 0; sar[i].idc != 255; i++)
if (sar[i].width == encoder->vui.sar_width &&
sar[i].height == encoder->vui.sar_height)
break;
WRITE_U(stream, 1, 1, "aspect_ratio_info_present_flag");
WRITE_U(stream, sar[i].idc, 8, "aspect_ratio_idc");
if (sar[i].idc == 255) {
// EXTENDED_SAR
WRITE_U(stream, encoder->vui.sar_width, 16, "sar_width");
WRITE_U(stream, encoder->vui.sar_height, 16, "sar_height");
}
} else
WRITE_U(stream, 0, 1, "aspect_ratio_info_present_flag");
//IF aspect ratio info
//ENDIF
if (encoder->vui.overscan > 0) {
WRITE_U(stream, 1, 1, "overscan_info_present_flag");
WRITE_U(stream, encoder->vui.overscan - 1, 1, "overscan_appropriate_flag");
} else
WRITE_U(stream, 0, 1, "overscan_info_present_flag");
//IF overscan info
//ENDIF
if (encoder->vui.videoformat != 5 || encoder->vui.fullrange ||
encoder->vui.colorprim != 2 || encoder->vui.transfer != 2 ||
encoder->vui.colormatrix != 2) {
WRITE_U(stream, 1, 1, "video_signal_type_present_flag");
WRITE_U(stream, encoder->vui.videoformat, 3, "video_format");
WRITE_U(stream, encoder->vui.fullrange, 1, "video_full_range_flag");
if (encoder->vui.colorprim != 2 || encoder->vui.transfer != 2 ||
encoder->vui.colormatrix != 2) {
WRITE_U(stream, 1, 1, "colour_description_present_flag");
WRITE_U(stream, encoder->vui.colorprim, 8, "colour_primaries");
WRITE_U(stream, encoder->vui.transfer, 8, "transfer_characteristics");
WRITE_U(stream, encoder->vui.colormatrix, 8, "matrix_coeffs");
} else
WRITE_U(stream, 0, 1, "colour_description_present_flag");
} else
WRITE_U(stream, 0, 1, "video_signal_type_present_flag");
//IF video type
//ENDIF
if (encoder->vui.chroma_loc > 0) {
WRITE_U(stream, 1, 1, "chroma_loc_info_present_flag");
WRITE_UE(stream, encoder->vui.chroma_loc, "chroma_sample_loc_type_top_field");
WRITE_UE(stream, encoder->vui.chroma_loc, "chroma_sample_loc_type_bottom_field");
} else
WRITE_U(stream, 0, 1, "chroma_loc_info_present_flag");
//IF chroma loc info
//ENDIF
WRITE_U(stream, 0, 1, "neutral_chroma_indication_flag");
WRITE_U(stream, 0, 1, "field_seq_flag");
WRITE_U(stream, 0, 1, "frame_field_info_present_flag");
WRITE_U(stream, 0, 1, "default_display_window_flag");
//IF default display window
//ENDIF
WRITE_U(stream, 0, 1, "vui_timing_info_present_flag");
//IF timing info
//ENDIF
WRITE_U(stream, 0, 1, "bitstream_restriction_flag");
//IF bitstream restriction
//ENDIF
}
static void encoder_state_write_bitstream_seq_parameter_set(encoder_state * const encoder_state)
{
bitstream * const stream = &encoder_state->stream;
//FIXME: use encoder_control instead of cur_pic
const picture * const cur_pic = encoder_state->tile->cur_pic;
#ifdef _DEBUG
printf("=========== Sequence Parameter Set ID: 0 ===========\n");
#endif
// TODO: profile IDC and level IDC should be defined later on
WRITE_U(stream, 0, 4, "sps_video_parameter_set_id");
WRITE_U(stream, 1, 3, "sps_max_sub_layers_minus1");
WRITE_U(stream, 0, 1, "sps_temporal_id_nesting_flag");
encoder_state_write_bitstream_PTL(encoder_state);
WRITE_UE(stream, 0, "sps_seq_parameter_set_id");
WRITE_UE(stream, encoder_state->encoder_control->in.video_format,
"chroma_format_idc");
if (encoder_state->encoder_control->in.video_format == 3) {
WRITE_U(stream, 0, 1, "separate_colour_plane_flag");
}
WRITE_UE(stream, cur_pic->width, "pic_width_in_luma_samples");
WRITE_UE(stream, cur_pic->height, "pic_height_in_luma_samples");
if (cur_pic->width != encoder_state->encoder_control->in.real_width || cur_pic->height != encoder_state->encoder_control->in.real_height) {
// The standard does not seem to allow setting conf_win values such that
// the number of luma samples is not a multiple of 2. Options are to either
// hide one line or show an extra line of non-video. Neither seems like a
// very good option, so let's not even try.
assert(!(cur_pic->width % 2));
WRITE_U(stream, 1, 1, "conformance_window_flag");
WRITE_UE(stream, 0, "conf_win_left_offset");
WRITE_UE(stream, (cur_pic->width - encoder_state->encoder_control->in.real_width) >> 1,
"conf_win_right_offset");
WRITE_UE(stream, 0, "conf_win_top_offset");
WRITE_UE(stream, (cur_pic->height - encoder_state->encoder_control->in.real_height) >> 1,
"conf_win_bottom_offset");
} else {
WRITE_U(stream, 0, 1, "conformance_window_flag");
}
//IF window flag
//END IF
WRITE_UE(stream, encoder_state->encoder_control->bitdepth-8, "bit_depth_luma_minus8");
WRITE_UE(stream, encoder_state->encoder_control->bitdepth-8, "bit_depth_chroma_minus8");
WRITE_UE(stream, 0, "log2_max_pic_order_cnt_lsb_minus4");
WRITE_U(stream, 0, 1, "sps_sub_layer_ordering_info_present_flag");
//for each layer
WRITE_UE(stream, 0, "sps_max_dec_pic_buffering");
WRITE_UE(stream, 0, "sps_num_reorder_pics");
WRITE_UE(stream, 0, "sps_max_latency_increase");
//end for
WRITE_UE(stream, MIN_SIZE-3, "log2_min_coding_block_size_minus3");
WRITE_UE(stream, MAX_DEPTH, "log2_diff_max_min_coding_block_size");
WRITE_UE(stream, 0, "log2_min_transform_block_size_minus2"); // 4x4
WRITE_UE(stream, 3, "log2_diff_max_min_transform_block_size"); // 4x4...32x32
WRITE_UE(stream, TR_DEPTH_INTER, "max_transform_hierarchy_depth_inter");
WRITE_UE(stream, TR_DEPTH_INTRA, "max_transform_hierarchy_depth_intra");
// scaling list
WRITE_U(stream, encoder_state->encoder_control->scaling_list.enable, 1, "scaling_list_enable_flag");
if (encoder_state->encoder_control->scaling_list.enable) {
WRITE_U(stream, 1, 1, "sps_scaling_list_data_present_flag");
encoder_state_write_bitstream_scaling_list(encoder_state);
}
WRITE_U(stream, 0, 1, "amp_enabled_flag");
WRITE_U(stream, encoder_state->encoder_control->sao_enable ? 1 : 0, 1,
"sample_adaptive_offset_enabled_flag");
WRITE_U(stream, ENABLE_PCM, 1, "pcm_enabled_flag");
#if ENABLE_PCM == 1
WRITE_U(stream, 7, 4, "pcm_sample_bit_depth_luma_minus1");
WRITE_U(stream, 7, 4, "pcm_sample_bit_depth_chroma_minus1");
WRITE_UE(stream, 0, "log2_min_pcm_coding_block_size_minus3");
WRITE_UE(stream, 2, "log2_diff_max_min_pcm_coding_block_size");
WRITE_U(stream, 1, 1, "pcm_loop_filter_disable_flag");
#endif
WRITE_UE(stream, 0, "num_short_term_ref_pic_sets");
//IF num short term ref pic sets
//ENDIF
WRITE_U(stream, 0, 1, "long_term_ref_pics_present_flag");
//IF long_term_ref_pics_present
//ENDIF
WRITE_U(stream, ENABLE_TEMPORAL_MVP, 1,
"sps_temporal_mvp_enable_flag");
WRITE_U(stream, 0, 1, "sps_strong_intra_smoothing_enable_flag");
WRITE_U(stream, 1, 1, "vui_parameters_present_flag");
encoder_state_write_bitstream_VUI(encoder_state);
WRITE_U(stream, 0, 1, "sps_extension_flag");
}
static void encoder_state_write_bitstream_pic_parameter_set(encoder_state * const encoder_state)
{
const encoder_control * const encoder = encoder_state->encoder_control;
bitstream * const stream = &encoder_state->stream;
#ifdef _DEBUG
printf("=========== Picture Parameter Set ID: 0 ===========\n");
#endif
WRITE_UE(stream, 0, "pic_parameter_set_id");
WRITE_UE(stream, 0, "seq_parameter_set_id");
WRITE_U(stream, 0, 1, "dependent_slice_segments_enabled_flag");
WRITE_U(stream, 0, 1, "output_flag_present_flag");
WRITE_U(stream, 0, 3, "num_extra_slice_header_bits");
WRITE_U(stream, ENABLE_SIGN_HIDING, 1, "sign_data_hiding_flag");
WRITE_U(stream, 0, 1, "cabac_init_present_flag");
WRITE_UE(stream, 0, "num_ref_idx_l0_default_active_minus1");
WRITE_UE(stream, 0, "num_ref_idx_l1_default_active_minus1");
WRITE_SE(stream, ((int8_t)encoder_state->global->QP)-26, "pic_init_qp_minus26");
WRITE_U(stream, 0, 1, "constrained_intra_pred_flag");
WRITE_U(stream, encoder_state->encoder_control->trskip_enable, 1, "transform_skip_enabled_flag");
WRITE_U(stream, 0, 1, "cu_qp_delta_enabled_flag");
//if cu_qp_delta_enabled_flag
//WRITE_UE(stream, 0, "diff_cu_qp_delta_depth");
//TODO: add QP offsets
WRITE_SE(stream, 0, "pps_cb_qp_offset");
WRITE_SE(stream, 0, "pps_cr_qp_offset");
WRITE_U(stream, 0, 1, "pps_slice_chroma_qp_offsets_present_flag");
WRITE_U(stream, 0, 1, "weighted_pred_flag");
WRITE_U(stream, 0, 1, "weighted_bipred_idc");
//WRITE_U(stream, 0, 1, "dependent_slices_enabled_flag");
WRITE_U(stream, 0, 1, "transquant_bypass_enable_flag");
WRITE_U(stream, encoder->tiles_enable, 1, "tiles_enabled_flag");
//wavefronts
WRITE_U(stream, encoder->wpp, 1, "entropy_coding_sync_enabled_flag");
if (encoder->tiles_enable) {
WRITE_UE(stream, encoder->tiles_num_tile_columns - 1, "num_tile_columns_minus1");
WRITE_UE(stream, encoder->tiles_num_tile_rows - 1, "num_tile_rows_minus1");
WRITE_U(stream, encoder->tiles_uniform_spacing_flag, 1, "uniform_spacing_flag");
if (!encoder->tiles_uniform_spacing_flag) {
int i;
for (i = 0; i < encoder->tiles_num_tile_columns - 1; ++i) {
WRITE_UE(stream, encoder->tiles_col_width[i] - 1, "column_width_minus1[...]");
}
for (i = 0; i < encoder->tiles_num_tile_rows - 1; ++i) {
WRITE_UE(stream, encoder->tiles_row_height[i] - 1, "row_height_minus1[...]");
}
}
WRITE_U(stream, 0, 1, "loop_filter_across_tiles_enabled_flag");
}
WRITE_U(stream, 0, 1, "loop_filter_across_slice_flag");
WRITE_U(stream, 1, 1, "deblocking_filter_control_present_flag");
//IF deblocking_filter
WRITE_U(stream, 0, 1, "deblocking_filter_override_enabled_flag");
WRITE_U(stream, encoder_state->encoder_control->deblock_enable ? 0 : 1, 1,
"pps_disable_deblocking_filter_flag");
//IF !disabled
if (encoder_state->encoder_control->deblock_enable) {
WRITE_SE(stream, encoder_state->encoder_control->beta_offset_div2, "beta_offset_div2");
WRITE_SE(stream, encoder_state->encoder_control->tc_offset_div2, "tc_offset_div2");
}
//ENDIF
//ENDIF
WRITE_U(stream, 0, 1, "pps_scaling_list_data_present_flag");
//IF scaling_list
//ENDIF
WRITE_U(stream, 0, 1, "lists_modification_present_flag");
WRITE_UE(stream, 0, "log2_parallel_merge_level_minus2");
WRITE_U(stream, 0, 1, "slice_segment_header_extension_present_flag");
WRITE_U(stream, 0, 1, "pps_extension_flag");
}
static void encoder_state_write_bitstream_prefix_sei_version(encoder_state * const encoder_state)
{
#define STR_BUF_LEN 1000
bitstream * const stream = &encoder_state->stream;
int i, length;
char buf[STR_BUF_LEN] = { 0 };
char *s = buf + 16;
const config * const cfg = encoder_state->encoder_control->cfg;
// random uuid_iso_iec_11578 generated with www.famkruithof.net/uuid/uuidgen
static const uint8_t uuid[16] = {
0x32, 0xfe, 0x46, 0x6c, 0x98, 0x41, 0x42, 0x69,
0xae, 0x35, 0x6a, 0x91, 0x54, 0x9e, 0xf3, 0xf1
};
memcpy(buf, uuid, 16);
// user_data_payload_byte
s += sprintf(s, "Kvazaar HEVC Encoder v. " VERSION_STRING " - "
"Copyleft 2012-2014 - http://ultravideo.cs.tut.fi/ - options:");
s += sprintf(s, " %dx%d", cfg->width, cfg->height);
s += sprintf(s, " deblock=%d:%d:%d", cfg->deblock_enable,
cfg->deblock_beta, cfg->deblock_tc);
s += sprintf(s, " sao=%d", cfg->sao_enable);
s += sprintf(s, " intra_period=%d", cfg->intra_period);
s += sprintf(s, " qp=%d", cfg->qp);
s += sprintf(s, " ref=%d", cfg->ref_frames);
length = (int)(s - buf + 1); // length, +1 for \0
// Assert this so that in the future if the message gets longer, we remember
// to increase the buf len. Divide by 2 for margin.
assert(length < STR_BUF_LEN / 2);
// payloadType = 5 -> user_data_unregistered
WRITE_U(stream, 5, 8, "last_payload_type_byte");
// payloadSize
for (i = 0; i <= length - 255; i += 255)
WRITE_U(stream, 255, 8, "ff_byte");
WRITE_U(stream, length - i, 8, "last_payload_size_byte");
for (i = 0; i < length; i++)
WRITE_U(stream, ((uint8_t *)buf)[i], 8, "sei_payload");
#undef STR_BUF_LEN
}
static void encoder_state_entry_points_explore(const encoder_state * const encoder_state, int * const r_count, int * const r_max_length) {
int i;
for (i = 0; encoder_state->children[i].encoder_control; ++i) {
if (encoder_state->children[i].is_leaf) {
const int my_length = bitstream_tell(&encoder_state->children[i].stream)/8;
++(*r_count);
if (my_length > *r_max_length) {
*r_max_length = my_length;
}
} else {
encoder_state_entry_points_explore(&encoder_state->children[i], r_count, r_max_length);
}
}
}
static void encoder_state_write_bitstream_entry_points_write(bitstream * const stream, const encoder_state * const encoder_state, const int num_entry_points, const int write_length, int * const r_count) {
int i;
for (i = 0; encoder_state->children[i].encoder_control; ++i) {
if (encoder_state->children[i].is_leaf) {
const int my_length = bitstream_tell(&encoder_state->children[i].stream)/8;
++(*r_count);
//Don't write the last one
if (*r_count < num_entry_points) {
WRITE_U(stream, my_length - 1, write_length, "entry_point_offset-minus1")
}
} else {
encoder_state_write_bitstream_entry_points_write(stream, &encoder_state->children[i], num_entry_points, write_length, r_count);
}
}
}
static int num_bitcount(unsigned int n) {
int pos = 0;
if (n >= 1<<16) { n >>= 16; pos += 16; }
if (n >= 1<< 8) { n >>= 8; pos += 8; }
if (n >= 1<< 4) { n >>= 4; pos += 4; }
if (n >= 1<< 2) { n >>= 2; pos += 2; }
if (n >= 1<< 1) { pos += 1; }
return ((n == 0) ? (-1) : pos);
}
void encoder_state_write_bitstream_slice_header(encoder_state * const encoder_state)
{
const encoder_control * const encoder = encoder_state->encoder_control;
bitstream * const stream = &encoder_state->stream;
#ifdef _DEBUG
printf("=========== Slice ===========\n");
#endif
WRITE_U(stream, (encoder_state->slice->start_in_rs == 0), 1, "first_slice_segment_in_pic_flag");
if (encoder_state->global->pictype >= NAL_BLA_W_LP
&& encoder_state->global->pictype <= NAL_RSV_IRAP_VCL23) {
WRITE_U(stream, 1, 1, "no_output_of_prior_pics_flag");
}
WRITE_UE(stream, 0, "slice_pic_parameter_set_id");
if (encoder_state->slice->start_in_rs > 0) {
//For now, we don't support dependent slice segments
//WRITE_U(stream, 0, 1, "dependent_slice_segment_flag");
WRITE_UE(stream, encoder_state->slice->start_in_rs, "slice_segment_address");
}
WRITE_UE(stream, encoder_state->global->slicetype, "slice_type");
// if !entropy_slice_flag
//if output_flag_present_flag
//WRITE_U(stream, 1, 1, "pic_output_flag");
//end if
//if( IdrPicFlag ) <- nal_unit_type == 5
if (encoder_state->global->pictype != NAL_IDR_W_RADL
&& encoder_state->global->pictype != NAL_IDR_N_LP) {
int j;
int ref_negative = encoder_state->global->ref->used_size;
int ref_positive = 0;
WRITE_U(stream, encoder_state->global->poc&0xf, 4, "pic_order_cnt_lsb");
WRITE_U(stream, 0, 1, "short_term_ref_pic_set_sps_flag");
WRITE_UE(stream, ref_negative, "num_negative_pics");
WRITE_UE(stream, ref_positive, "num_positive_pics");
for (j = 0; j < ref_negative; j++) {
int32_t delta_poc_minus1 = 0;
WRITE_UE(stream, delta_poc_minus1, "delta_poc_s0_minus1");
WRITE_U(stream,1,1, "used_by_curr_pic_s0_flag");
}
//WRITE_UE(stream, 0, "short_term_ref_pic_set_idx");
}
//end if
//end if
if (encoder->sao_enable) {
WRITE_U(stream, 1, 1, "slice_sao_luma_flag");
WRITE_U(stream, 1, 1, "slice_sao_chroma_flag");
}
if (encoder_state->global->slicetype != SLICE_I) {
WRITE_U(stream, 1, 1, "num_ref_idx_active_override_flag");
WRITE_UE(stream, encoder_state->global->ref->used_size-1, "num_ref_idx_l0_active_minus1");
WRITE_UE(stream, 5-MRG_MAX_NUM_CANDS, "five_minus_max_num_merge_cand");
}
if (encoder_state->global->slicetype == SLICE_B) {
WRITE_U(stream, 0, 1, "mvd_l1_zero_flag");
}
// Skip flags that are not present
// if !entropy_slice_flag
WRITE_SE(stream, 0, "slice_qp_delta");
//WRITE_U(stream, 1, 1, "alignment");
if (encoder->tiles_enable || encoder->wpp) {
int num_entry_points = 0;
int max_length_seen = 0;
encoder_state_entry_points_explore(encoder_state, &num_entry_points, &max_length_seen);
WRITE_UE(stream, num_entry_points - 1, "num_entry_point_offsets");
if (num_entry_points > 0) {
int entry_points_written = 0;
int offset_len = num_bitcount(max_length_seen) + 1;
WRITE_UE(stream, offset_len - 1, "offset_len_minus1");
encoder_state_write_bitstream_entry_points_write(stream, encoder_state, num_entry_points, offset_len, &entry_points_written);
}
}
}
static void encoder_state_write_bitstream_main(encoder_state * const main_state) {
const encoder_control * const encoder = main_state->encoder_control;
bitstream * const stream = &main_state->stream;
int i;
if (main_state->global->is_radl_frame) {
// Access Unit Delimiter (AUD)
if (encoder->aud_enable)
encoder_state_write_bitstream_aud(main_state);
// Video Parameter Set (VPS)
nal_write(stream, NAL_VPS_NUT, 0, 1);
encoder_state_write_bitstream_vid_parameter_set(main_state);
bitstream_align(stream);
// Sequence Parameter Set (SPS)
nal_write(stream, NAL_SPS_NUT, 0, 1);
encoder_state_write_bitstream_seq_parameter_set(main_state);
bitstream_align(stream);
// Picture Parameter Set (PPS)
nal_write(stream, NAL_PPS_NUT, 0, 1);
encoder_state_write_bitstream_pic_parameter_set(main_state);
bitstream_align(stream);
if (main_state->global->frame == 0) {
// Prefix SEI
nal_write(stream, PREFIX_SEI_NUT, 0, 0);
encoder_state_write_bitstream_prefix_sei_version(main_state);
bitstream_align(stream);
}
} else {
// Access Unit Delimiter (AUD)
if (encoder->aud_enable)
encoder_state_write_bitstream_aud(main_state);
}
{
// Not quite sure if this is correct, but it seems to have worked so far
// so I tried to not change it's behavior.
int long_start_code = main_state->global->is_radl_frame || encoder->aud_enable ? 0 : 1;
nal_write(stream,
main_state->global->is_radl_frame ? NAL_IDR_W_RADL : NAL_TRAIL_R, 0, long_start_code);
}
{
PERFORMANCE_MEASURE_START();
for (i = 0; main_state->children[i].encoder_control; ++i) {
//Append bitstream to main stream
bitstream_append(&main_state->stream, &main_state->children[i].stream);
//FIXME: Move this...
bitstream_clear(&main_state->children[i].stream);
}
PERFORMANCE_MEASURE_END(main_state->encoder_control->threadqueue, "type=write_bitstream_append,frame=%d,type=%c", main_state->global->frame, main_state->type);
}
{
PERFORMANCE_MEASURE_START();
// Calculate checksum
add_checksum(main_state);
PERFORMANCE_MEASURE_END(main_state->encoder_control->threadqueue, "type=write_bitstream_checksum,frame=%d,type=%c", main_state->global->frame, main_state->type);
}
//FIXME: Why is this needed?
main_state->tile->cur_pic->poc = main_state->global->poc;
}
static void encoder_state_worker_write_bitstream_leaf(void * opaque) {
encoder_state_write_bitstream_leaf((encoder_state *) opaque);
}
static void encoder_state_write_bitstream_leaf(encoder_state * const encoder_state) {
const encoder_control * const encoder = encoder_state->encoder_control;
//Write terminator of the leaf
assert(encoder_state->is_leaf);
//Last LCU
{
const lcu_order_element * const lcu = &encoder_state->lcu_order[encoder_state->lcu_order_count - 1];
const int lcu_addr_in_ts = lcu->id + encoder_state->tile->lcu_offset_in_ts;
const int end_of_slice_segment_flag = lcu_at_slice_end(encoder, lcu_addr_in_ts);
cabac_encode_bin_trm(&encoder_state->cabac, end_of_slice_segment_flag); // end_of_slice_segment_flag
if (!end_of_slice_segment_flag) {
assert(lcu_at_tile_end(encoder, lcu_addr_in_ts) || lcu->position.x == (encoder_state->tile->cur_pic->width_in_lcu - 1));
cabac_encode_bin_trm(&encoder_state->cabac, 1); // end_of_sub_stream_one_bit == 1
cabac_flush(&encoder_state->cabac);
} else {
cabac_flush(&encoder_state->cabac);
bitstream_align(&encoder_state->stream);
}
}
}
static void encoder_state_write_bitstream_tile(encoder_state * const main_state) {
//If it's not a leaf, a tile is "nothing". We only have to write sub elements
int i;
for (i = 0; main_state->children[i].encoder_control; ++i) {
//Append bitstream to main stream
bitstream_append(&main_state->stream, &main_state->children[i].stream);
}
}
static void encoder_state_write_bitstream_slice(encoder_state * const main_state) {
int i;
encoder_state_write_bitstream_slice_header(main_state);
bitstream_align(&main_state->stream);
for (i = 0; main_state->children[i].encoder_control; ++i) {
//Append bitstream to main stream
bitstream_append(&main_state->stream, &main_state->children[i].stream);
}
}
static void encoder_state_write_bitstream(encoder_state * const main_state) {
int i;
if (!main_state->is_leaf) {
for (i=0; main_state->children[i].encoder_control; ++i) {
encoder_state *sub_state = &(main_state->children[i]);
encoder_state_write_bitstream(sub_state);
}
switch (main_state->type) {
case ENCODER_STATE_TYPE_MAIN:
encoder_state_write_bitstream_main(main_state);
break;
case ENCODER_STATE_TYPE_TILE:
encoder_state_write_bitstream_tile(main_state);
break;
case ENCODER_STATE_TYPE_SLICE:
encoder_state_write_bitstream_slice(main_state);
break;
default:
fprintf(stderr, "Unsupported node type %c!\n", main_state->type);
assert(0);
}
}
}

672
src/_encoderstatectorsdtors.c Normal file
View file

@ -0,0 +1,672 @@
/*****************************************************************************
* This file is part of Kvazaar HEVC encoder.
*
* Copyright (C) 2013-2014 Tampere University of Technology and others (see
* COPYING file).
*
* Kvazaar is free software: you can redistribute it and/or modify
* it under the terms of the GNU General Public License version 2 as published
* by the Free Software Foundation.
*
* Kvazaar 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 General Public License for more details.
*
* You should have received a copy of the GNU General Public License
* along with Kvazaar. If not, see <http://www.gnu.org/licenses/>.
****************************************************************************/
//This file MUST NOT BE COMPILED directly. It's included in encoderstate.c
static int encoder_state_config_global_init(encoder_state * const encoder_state) {
encoder_state->global->ref = picture_list_init(MAX_REF_PIC_COUNT);
if(!encoder_state->global->ref) {
fprintf(stderr, "Failed to allocate the picture list!\n");
return 0;
}
encoder_state->global->ref_list = REF_PIC_LIST_0;
encoder_state->global->frame = 0;
encoder_state->global->poc = 0;
return 1;
}
static void encoder_state_config_global_finalize(encoder_state * const encoder_state) {
picture_list_destroy(encoder_state->global->ref);
}
static int encoder_state_config_tile_init(encoder_state * const encoder_state,
const int lcu_offset_x, const int lcu_offset_y,
const int width, const int height, const int width_in_lcu, const int height_in_lcu) {
const encoder_control * const encoder = encoder_state->encoder_control;
encoder_state->tile->cur_pic = picture_alloc(width, height, width_in_lcu, height_in_lcu);
if (!encoder_state->tile->cur_pic) {
printf("Error allocating picture!\r\n");
return 0;
}
// Init coeff data table
//FIXME: move them
encoder_state->tile->cur_pic->coeff_y = MALLOC(coefficient, width * height);
encoder_state->tile->cur_pic->coeff_u = MALLOC(coefficient, (width * height) >> 2);
encoder_state->tile->cur_pic->coeff_v = MALLOC(coefficient, (width * height) >> 2);
encoder_state->tile->lcu_offset_x = lcu_offset_x;
encoder_state->tile->lcu_offset_y = lcu_offset_y;
encoder_state->tile->lcu_offset_in_ts = encoder->tiles_ctb_addr_rs_to_ts[lcu_offset_x + lcu_offset_y * encoder->in.width_in_lcu];
//Allocate buffers
//order by row of (LCU_WIDTH * cur_pic->width_in_lcu) pixels
encoder_state->tile->hor_buf_search = yuv_t_alloc(LCU_WIDTH * encoder_state->tile->cur_pic->width_in_lcu * encoder_state->tile->cur_pic->height_in_lcu);
//order by column of (LCU_WIDTH * encoder_state->height_in_lcu) pixels (there is no more extra pixel, since we can use a negative index)
encoder_state->tile->ver_buf_search = yuv_t_alloc(LCU_WIDTH * encoder_state->tile->cur_pic->height_in_lcu * encoder_state->tile->cur_pic->width_in_lcu);
if (encoder->sao_enable) {
encoder_state->tile->hor_buf_before_sao = yuv_t_alloc(LCU_WIDTH * encoder_state->tile->cur_pic->width_in_lcu * encoder_state->tile->cur_pic->height_in_lcu);
} else {
encoder_state->tile->hor_buf_before_sao = NULL;
}
if (encoder->wpp) {
encoder_state->tile->wf_jobs = MALLOC(threadqueue_job*, encoder_state->tile->cur_pic->width_in_lcu * encoder_state->tile->cur_pic->height_in_lcu);
if (!encoder_state->tile->wf_jobs) {
printf("Error allocating wf_jobs array!\n");
return 0;
}
} else {
encoder_state->tile->wf_jobs = NULL;
}
encoder_state->tile->id = encoder->tiles_tile_id[encoder_state->tile->lcu_offset_in_ts];
return 1;
}
static void encoder_state_config_tile_finalize(encoder_state * const encoder_state) {
if (encoder_state->tile->hor_buf_before_sao) yuv_t_free(encoder_state->tile->hor_buf_before_sao);
yuv_t_free(encoder_state->tile->hor_buf_search);
yuv_t_free(encoder_state->tile->ver_buf_search);
picture_free(encoder_state->tile->cur_pic);
encoder_state->tile->cur_pic = NULL;
FREE_POINTER(encoder_state->tile->wf_jobs);
}
static int encoder_state_config_slice_init(encoder_state * const encoder_state,
const int start_address_in_ts, const int end_address_in_ts) {
int i = 0, slice_found=0;
for (i = 0; i < encoder_state->encoder_control->slice_count; ++i) {
if (encoder_state->encoder_control->slice_addresses_in_ts[i] == start_address_in_ts) {
encoder_state->slice->id = i;
slice_found = 1;
break;
}
}
assert(slice_found);
encoder_state->slice->start_in_ts = start_address_in_ts;
encoder_state->slice->end_in_ts = end_address_in_ts;
encoder_state->slice->start_in_rs = encoder_state->encoder_control->tiles_ctb_addr_ts_to_rs[start_address_in_ts];
encoder_state->slice->end_in_rs = encoder_state->encoder_control->tiles_ctb_addr_ts_to_rs[end_address_in_ts];
return 1;
}
static void encoder_state_config_slice_finalize(encoder_state * const encoder_state) {
//Nothing to do (yet?)
}
static int encoder_state_config_wfrow_init(encoder_state * const encoder_state,
const int lcu_offset_y) {
encoder_state->wfrow->lcu_offset_y = lcu_offset_y;
return 1;
}
static void encoder_state_config_wfrow_finalize(encoder_state * const encoder_state) {
//Nothing to do (yet?)
}
#ifdef _DEBUG
static void encoder_state_dump_graphviz(const encoder_state * const encoder_state) {
int i;
if (!encoder_state->parent) {
const encoder_control * const encoder = encoder_state->encoder_control;
int y,x;
//Empty lines (easier to copy-paste)
printf("\n\n\n\n\n");
//Some styling...
printf("digraph EncoderStates {\n");
printf(" fontname = \"Bitstream Vera Sans\"\n");
printf(" fontsize = 8\n\n");
printf(" node [\n");
printf(" fontname = \"Bitstream Vera Sans\"\n");
printf(" fontsize = 8\n");
printf(" shape = \"record\"\n");
printf(" ]\n\n");
printf(" edge [\n");
printf(" arrowtail = \"empty\"\n");
printf(" ]\n\n");
printf(" \"Map\" [\n");
printf(" shape=plaintext\n");
printf(" label = <<table cellborder=\"1\" cellspacing=\"0\" border=\"0\">");
printf("<tr><td colspan=\"%d\" height=\"20\" valign=\"bottom\"><b>RS Map</b></td></tr>", encoder->in.width_in_lcu);
for (y = 0; y < encoder->in.height_in_lcu; ++y) {
printf("<tr>");
for (x = 0; x < encoder->in.width_in_lcu; ++x) {
const int lcu_id_rs = y * encoder->in.width_in_lcu + x;
printf("<td>%d</td>", lcu_id_rs);
}
printf("</tr>");
}
printf("<tr><td colspan=\"%d\" height=\"20\" valign=\"bottom\"><b>TS Map</b></td></tr>", encoder->in.width_in_lcu);
for (y = 0; y < encoder->in.height_in_lcu; ++y) {
printf("<tr>");
for (x = 0; x < encoder->in.width_in_lcu; ++x) {
const int lcu_id_rs = y * encoder->in.width_in_lcu + x;
const int lcu_id_ts = encoder->tiles_ctb_addr_rs_to_ts[lcu_id_rs];
printf("<td>%d</td>", lcu_id_ts);
}
printf("</tr>");
}
printf("<tr><td colspan=\"%d\" height=\"20\" valign=\"bottom\"><b>Tile map</b></td></tr>", encoder->in.width_in_lcu);
for (y = 0; y < encoder->in.height_in_lcu; ++y) {
printf("<tr>");
for (x = 0; x < encoder->in.width_in_lcu; ++x) {
const int lcu_id_rs = y * encoder->in.width_in_lcu + x;
const int lcu_id_ts = encoder->tiles_ctb_addr_rs_to_ts[lcu_id_rs];
printf("<td>%d</td>", encoder->tiles_tile_id[lcu_id_ts]);
}
printf("</tr>");
}
printf("<tr><td colspan=\"%d\" height=\"20\" valign=\"bottom\"><b>Slice map</b></td></tr>", encoder->in.width_in_lcu);
for (y = 0; y < encoder->in.height_in_lcu; ++y) {
printf("<tr>");
for (x = 0; x < encoder->in.width_in_lcu; ++x) {
const int lcu_id_rs = y * encoder->in.width_in_lcu + x;
const int lcu_id_ts = encoder->tiles_ctb_addr_rs_to_ts[lcu_id_rs];
int slice_id = 0;
//Not efficient, but who cares
for (i=0; i < encoder->slice_count; ++i) {
if (encoder->slice_addresses_in_ts[i] <= lcu_id_ts) {
slice_id = i;
}
}
printf("<td>%d</td>", slice_id);
}
printf("</tr>");
}
printf("</table>>\n ]\n");
}
printf(" \"%p\" [\n", encoder_state);
printf(" label = \"{encoder_state|");
printf("+ type=%c\\l", encoder_state->type);
if (!encoder_state->parent || encoder_state->global != encoder_state->parent->global) {
printf("|+ global\\l");
}
if (!encoder_state->parent || encoder_state->tile != encoder_state->parent->tile) {
printf("|+ tile\\l");
printf(" - id = %d\\l", encoder_state->tile->id);
printf(" - lcu_offset_x = %d\\l", encoder_state->tile->lcu_offset_x);
printf(" - lcu_offset_y = %d\\l", encoder_state->tile->lcu_offset_y);
printf(" - lcu_offset_in_ts = %d\\l", encoder_state->tile->lcu_offset_in_ts);
}
if (!encoder_state->parent || encoder_state->slice != encoder_state->parent->slice) {
printf("|+ slice\\l");
printf(" - id = %d\\l", encoder_state->slice->id);
printf(" - start_in_ts = %d\\l", encoder_state->slice->start_in_ts);
printf(" - end_in_ts = %d\\l", encoder_state->slice->end_in_ts);
printf(" - start_in_rs = %d\\l", encoder_state->slice->start_in_rs);
printf(" - end_in_rs = %d\\l", encoder_state->slice->end_in_rs);
}
if (!encoder_state->parent || encoder_state->wfrow != encoder_state->parent->wfrow) {
printf("|+ wfrow\\l");
printf(" - lcu_offset_y = %d\\l", encoder_state->wfrow->lcu_offset_y);
}
printf("}\"\n");
printf(" ]\n");
if (encoder_state->parent) {
printf(" \"%p\" -> \"%p\"\n", encoder_state->parent, encoder_state);
}
for (i = 0; encoder_state->children[i].encoder_control; ++i) {
encoder_state_dump_graphviz(&encoder_state->children[i]);
}
if (!encoder_state->parent) {
printf("}\n");
//Empty lines (easier to copy-paste)
printf("\n\n\n\n\n");
}
}
#endif //_DEBUG
int encoder_state_init(encoder_state * const child_state, encoder_state * const parent_state) {
//We require that, if parent_state is NULL:
//child_state->encoder_control is set
//
//If parent_state is not NULL, the following variable should either be set to NULL,
//in order to inherit from parent, or should point to a valid structure:
//child_state->global
//child_state->tile
//child_state->slice
//child_state->wfrow
child_state->parent = parent_state;
child_state->children = MALLOC(encoder_state, 1);
child_state->children[0].encoder_control = NULL;
if (!parent_state) {
const encoder_control * const encoder = child_state->encoder_control;
child_state->type = ENCODER_STATE_TYPE_MAIN;
assert(child_state->encoder_control);
child_state->global = MALLOC(encoder_state_config_global, 1);
if (!child_state->global || !encoder_state_config_global_init(child_state)) {
fprintf(stderr, "Could not initialize encoder_state->global!\n");
return 0;
}
child_state->tile = MALLOC(encoder_state_config_tile, 1);
if (!child_state->tile || !encoder_state_config_tile_init(child_state, 0, 0, encoder->in.width, encoder->in.height, encoder->in.width_in_lcu, encoder->in.height_in_lcu)) {
fprintf(stderr, "Could not initialize encoder_state->tile!\n");
return 0;
}
child_state->slice = MALLOC(encoder_state_config_slice, 1);
if (!child_state->slice || !encoder_state_config_slice_init(child_state, 0, encoder->in.width_in_lcu * encoder->in.height_in_lcu - 1)) {
fprintf(stderr, "Could not initialize encoder_state->slice!\n");
return 0;
}
child_state->wfrow = MALLOC(encoder_state_config_wfrow, 1);
if (!child_state->wfrow || !encoder_state_config_wfrow_init(child_state, 0)) {
fprintf(stderr, "Could not initialize encoder_state->wfrow!\n");
return 0;
}
} else {
child_state->encoder_control = parent_state->encoder_control;
if (!child_state->global) child_state->global = parent_state->global;
if (!child_state->tile) child_state->tile = parent_state->tile;
if (!child_state->slice) child_state->slice = parent_state->slice;
if (!child_state->wfrow) child_state->wfrow = parent_state->wfrow;
}
//Allocate bitstream
if (child_state->type == ENCODER_STATE_TYPE_MAIN) {
//Main encoder outputs to file
if (!bitstream_init(&child_state->stream, BITSTREAM_TYPE_FILE)) {
fprintf(stderr, "Could not initialize stream!\n");
return 0;
}
child_state->stream.file.output = child_state->encoder_control->out.file;
} else {
//Other encoders use a memory bitstream
if (!bitstream_init(&child_state->stream, BITSTREAM_TYPE_MEMORY)) {
fprintf(stderr, "Could not initialize stream!\n");
return 0;
}
}
// Set CABAC output bitstream
child_state->cabac.stream = &child_state->stream;
//Create sub-encoders
{
const encoder_control * const encoder = child_state->encoder_control;
int child_count = 0;
//We first check the type of this element.
//If it's a MAIN, it can allow both slices or tiles as child
//If it's a TILE, it can allow slices as child, if its parent is not a slice, or wavefront rows if there is no other children
//If it's a SLICE, it can allow tiles as child, if its parent is not a tile, or wavefront rows if there is no other children
//If it's a WAVEFRONT_ROW, it doesn't allow any children
int children_allow_wavefront_row = 0;
int children_allow_slice = 0;
int children_allow_tile = 0;
int range_start;
int start_in_ts, end_in_ts;
switch(child_state->type) {
case ENCODER_STATE_TYPE_MAIN:
children_allow_slice = 1;
children_allow_tile = 1;
start_in_ts = 0;
end_in_ts = child_state->tile->cur_pic->width_in_lcu * child_state->tile->cur_pic->height_in_lcu;
break;
case ENCODER_STATE_TYPE_SLICE:
assert(child_state->parent);
if (child_state->parent->type != ENCODER_STATE_TYPE_TILE) children_allow_tile = 1;
children_allow_wavefront_row = encoder->wpp;
start_in_ts = child_state->slice->start_in_ts;
end_in_ts = child_state->slice->end_in_ts;
break;
case ENCODER_STATE_TYPE_TILE:
assert(child_state->parent);
if (child_state->parent->type != ENCODER_STATE_TYPE_SLICE) children_allow_slice = 1;
children_allow_wavefront_row = encoder->wpp;
start_in_ts = child_state->tile->lcu_offset_in_ts;
end_in_ts = child_state->tile->lcu_offset_in_ts + child_state->tile->cur_pic->width_in_lcu * child_state->tile->cur_pic->height_in_lcu;
break;
case ENCODER_STATE_TYPE_WAVEFRONT_ROW:
//GCC tries to be too clever...
start_in_ts = -1;
end_in_ts = -1;
break;
default:
fprintf(stderr, "Invalid encoder_state->type %d!\n", child_state->type);
assert(0);
return 0;
}
range_start = start_in_ts;
//printf("%c-%p: start_in_ts=%d, end_in_ts=%d\n",child_state->type, child_state, start_in_ts, end_in_ts);
while (range_start < end_in_ts && (children_allow_slice || children_allow_tile)) {
encoder_state *new_child = NULL;
int range_end_slice = range_start; //Will be incremented to get the range of the "thing"
int range_end_tile = range_start; //Will be incremented to get the range of the "thing"
int tile_allowed = lcu_at_tile_start(encoder, range_start) && children_allow_tile;
int slice_allowed = lcu_at_slice_start(encoder, range_start) && children_allow_slice;
//Find the smallest structure following the cursor
if (slice_allowed) {
while(!lcu_at_slice_end(encoder, range_end_slice)) {
++range_end_slice;
}
}
if (tile_allowed) {
while(!lcu_at_tile_end(encoder, range_end_tile)) {
++range_end_tile;
}
}
//printf("range_start=%d, range_end_slice=%d, range_end_tile=%d, tile_allowed=%d, slice_allowed=%d end_in_ts=%d\n",range_start,range_end_slice,range_end_tile,tile_allowed,slice_allowed,end_in_ts);
if ((!tile_allowed || (range_end_slice >= range_end_tile)) && !new_child && slice_allowed) {
//Create a slice
new_child = &child_state->children[child_count];
new_child->encoder_control = encoder;
new_child->type = ENCODER_STATE_TYPE_SLICE;
new_child->global = child_state->global;
new_child->tile = child_state->tile;
new_child->wfrow = child_state->wfrow;
new_child->slice = MALLOC(encoder_state_config_slice, 1);
if (!new_child->slice || !encoder_state_config_slice_init(new_child, range_start, range_end_slice)) {
fprintf(stderr, "Could not initialize encoder_state->slice!\n");
return 0;
}
}
if ((!slice_allowed || (range_end_slice < range_end_tile)) && !new_child && tile_allowed) {
//Create a tile
int tile_id = encoder->tiles_tile_id[range_start];
int tile_x = tile_id % encoder->tiles_num_tile_columns;
int tile_y = tile_id / encoder->tiles_num_tile_columns;
int lcu_offset_x = encoder->tiles_col_bd[tile_x];
int lcu_offset_y = encoder->tiles_row_bd[tile_y];
int width_in_lcu = encoder->tiles_col_bd[tile_x+1]-encoder->tiles_col_bd[tile_x];
int height_in_lcu = encoder->tiles_row_bd[tile_y+1]-encoder->tiles_row_bd[tile_y];
int width = MIN(width_in_lcu * LCU_WIDTH, encoder->in.width - lcu_offset_x * LCU_WIDTH);
int height = MIN(height_in_lcu * LCU_WIDTH, encoder->in.height - lcu_offset_y * LCU_WIDTH);
new_child = &child_state->children[child_count];
new_child->encoder_control = encoder;
new_child->type = ENCODER_STATE_TYPE_TILE;
new_child->global = child_state->global;
new_child->tile = MALLOC(encoder_state_config_tile, 1);
new_child->slice = child_state->slice;
new_child->wfrow = child_state->wfrow;
if (!new_child->tile || !encoder_state_config_tile_init(new_child, lcu_offset_x, lcu_offset_y, width, height, width_in_lcu, height_in_lcu)) {
fprintf(stderr, "Could not initialize encoder_state->tile!\n");
return 0;
}
}
if (new_child) {
child_state->children = realloc(child_state->children, sizeof(encoder_state) * (2+child_count));
child_state->children[1+child_count].encoder_control = NULL;
if (!child_state->children) {
fprintf(stderr, "Failed to allocate memory for children...\n");
return 0;
}
//Fix children parent (since we changed the address), except for the last one which is not ready yet
{
int i, j;
for (i = 0; child_state->children[i].encoder_control && i < child_count; ++i) {
for (j = 0; child_state->children[i].children[j].encoder_control; ++j) {
child_state->children[i].children[j].parent = &child_state->children[i];
}
for (j = 0; j < child_state->children[i].lcu_order_count; ++j) {
child_state->children[i].lcu_order[j].encoder_state = &child_state->children[i];
}
child_state->children[i].cabac.stream = &child_state->children[i].stream;
}
}
if (!encoder_state_init(&child_state->children[child_count], child_state)) {
fprintf(stderr, "Unable to init child...\n");
return 0;
}
child_count += 1;
}
range_start = MAX(range_end_slice, range_end_tile) + 1;
}
//We create wavefronts only if we have no children
if (children_allow_wavefront_row && child_count == 0) {
int first_row = encoder->tiles_ctb_addr_ts_to_rs[start_in_ts] / encoder->in.width_in_lcu;
int last_row = encoder->tiles_ctb_addr_ts_to_rs[start_in_ts] / encoder->in.width_in_lcu;
int num_rows;
int i;
assert(!(children_allow_slice || children_allow_tile));
assert(child_count == 0);
for (i=start_in_ts; i<end_in_ts; ++i) {
const int row = encoder->tiles_ctb_addr_ts_to_rs[i] / encoder->in.width_in_lcu;
if (row < first_row) first_row = row;
if (row > last_row) last_row = row;
}
num_rows = last_row - first_row + 1;
//When entropy_coding_sync_enabled_flag is equal to 1 and the first coding tree block in a slice is not the first coding
//tree block of a row of coding tree blocks in a tile, it is a requirement of bitstream conformance that the last coding tree
//block in the slice shall belong to the same row of coding tree blocks as the first coding tree block in the slice.
if (encoder->tiles_ctb_addr_ts_to_rs[start_in_ts] % encoder->in.width_in_lcu != child_state->tile->lcu_offset_x) {
if (num_rows > 1) {
fprintf(stderr, "Invalid: first CTB in slice %d is not at the tile %d edge, and the slice spans on more than one row.\n", child_state->slice->id, child_state->tile->id);
return 0;
}
}
//FIXME Do the same kind of check if we implement slice segments
child_count = num_rows;
child_state->children = realloc(child_state->children, sizeof(encoder_state) * (num_rows + 1));
child_state->children[num_rows].encoder_control = NULL;
for (i=0; i < num_rows; ++i) {
encoder_state *new_child = &child_state->children[i];
new_child->encoder_control = encoder;
new_child->type = ENCODER_STATE_TYPE_WAVEFRONT_ROW;
new_child->global = child_state->global;
new_child->tile = child_state->tile;
new_child->slice = child_state->slice;
new_child->wfrow = MALLOC(encoder_state_config_wfrow, 1);
if (!new_child->wfrow || !encoder_state_config_wfrow_init(new_child, i)) {
fprintf(stderr, "Could not initialize encoder_state->wfrow!\n");
return 0;
}
if (!encoder_state_init(new_child, child_state)) {
fprintf(stderr, "Unable to init child...\n");
return 0;
}
}
}
child_state->is_leaf = (child_count == 0);
//This node is a leaf, compute LCU-order
if (child_state->is_leaf) {
//All LCU computations are relative to the tile
//Remark: this could be optimized, but since it's run only once, it's better to do it in a understandable way.
//By default, the full tile
int i;
int lcu_id;
int lcu_start = 0;
//End is the element AFTER the end (iterate < lcu_end)
int lcu_end = child_state->tile->cur_pic->width_in_lcu * child_state->tile->cur_pic->height_in_lcu;
//Restrict to the current slice if needed
lcu_start = MAX(lcu_start, child_state->slice->start_in_ts - child_state->tile->lcu_offset_in_ts);
lcu_end = MIN(lcu_end, child_state->slice->end_in_ts - child_state->tile->lcu_offset_in_ts + 1);
//Restrict to the current wavefront row if needed
if (child_state->type == ENCODER_STATE_TYPE_WAVEFRONT_ROW) {
lcu_start = MAX(lcu_start, (child_state->wfrow->lcu_offset_y) * child_state->tile->cur_pic->width_in_lcu);
lcu_end = MIN(lcu_end, (child_state->wfrow->lcu_offset_y + 1) * child_state->tile->cur_pic->width_in_lcu);
}
child_state->lcu_order_count = lcu_end - lcu_start;
child_state->lcu_order = MALLOC(lcu_order_element, child_state->lcu_order_count);
assert(child_state->lcu_order);
for (i = 0; i < child_state->lcu_order_count; ++i) {
lcu_id = lcu_start + i;
child_state->lcu_order[i].encoder_state = child_state;
child_state->lcu_order[i].id = lcu_id;
child_state->lcu_order[i].index = i;
child_state->lcu_order[i].position.x = lcu_id % child_state->tile->cur_pic->width_in_lcu;
child_state->lcu_order[i].position.y = lcu_id / child_state->tile->cur_pic->width_in_lcu;
child_state->lcu_order[i].position_px.x = child_state->lcu_order[i].position.x * LCU_WIDTH;
child_state->lcu_order[i].position_px.y = child_state->lcu_order[i].position.y * LCU_WIDTH;
child_state->lcu_order[i].size.x = MIN(LCU_WIDTH, encoder->in.width - (child_state->tile->lcu_offset_x * LCU_WIDTH + child_state->lcu_order[i].position_px.x));
child_state->lcu_order[i].size.y = MIN(LCU_WIDTH, encoder->in.height - (child_state->tile->lcu_offset_y * LCU_WIDTH + child_state->lcu_order[i].position_px.y));
child_state->lcu_order[i].first_row = lcu_in_first_row(child_state, child_state->tile->lcu_offset_in_ts + lcu_id);
child_state->lcu_order[i].last_row = lcu_in_last_row(child_state, child_state->tile->lcu_offset_in_ts + lcu_id);
child_state->lcu_order[i].first_column = lcu_in_first_column(child_state, child_state->tile->lcu_offset_in_ts + lcu_id);
child_state->lcu_order[i].last_column = lcu_in_last_column(child_state, child_state->tile->lcu_offset_in_ts + lcu_id);
child_state->lcu_order[i].above = NULL;
child_state->lcu_order[i].below = NULL;
child_state->lcu_order[i].left = NULL;
child_state->lcu_order[i].right = NULL;
if (!child_state->lcu_order[i].first_row) {
//Find LCU above
if (child_state->type == ENCODER_STATE_TYPE_WAVEFRONT_ROW) {
int j;
//For all previous wavefront rows
for (j=0; &child_state->parent->children[j] != child_state && child_state->parent->children[j].encoder_control; ++j) {
if (child_state->parent->children[j].wfrow->lcu_offset_y == child_state->wfrow->lcu_offset_y - 1) {
int k;
for (k=0; k < child_state->parent->children[j].lcu_order_count; ++k) {
if (child_state->parent->children[j].lcu_order[k].position.x == child_state->lcu_order[i].position.x) {
assert(child_state->parent->children[j].lcu_order[k].position.y == child_state->lcu_order[i].position.y - 1);
child_state->lcu_order[i].above = &child_state->parent->children[j].lcu_order[k];
}
}
}
}
} else {
child_state->lcu_order[i].above = &child_state->lcu_order[i-child_state->tile->cur_pic->width_in_lcu];
}
assert(child_state->lcu_order[i].above);
child_state->lcu_order[i].above->below = &child_state->lcu_order[i];
}
if (!child_state->lcu_order[i].first_column) {
child_state->lcu_order[i].left = &child_state->lcu_order[i-1];
assert(child_state->lcu_order[i].left->position.x == child_state->lcu_order[i].position.x - 1);
child_state->lcu_order[i].left->right = &child_state->lcu_order[i];
}
}
} else {
child_state->lcu_order_count = 0;
child_state->lcu_order = NULL;
}
}
//Validate the structure
if (child_state->type == ENCODER_STATE_TYPE_TILE) {
if (child_state->tile->lcu_offset_in_ts < child_state->slice->start_in_ts) {
fprintf(stderr, "Tile %d starts before slice %d, in which it should be included!\n", child_state->tile->id, child_state->slice->id);
return 0;
}
if (child_state->tile->lcu_offset_in_ts + child_state->tile->cur_pic->width_in_lcu * child_state->tile->cur_pic->height_in_lcu - 1 > child_state->slice->end_in_ts) {
fprintf(stderr, "Tile %d ends after slice %d, in which it should be included!\n", child_state->tile->id, child_state->slice->id);
return 0;
}
}
if (child_state->type == ENCODER_STATE_TYPE_SLICE) {
if (child_state->slice->start_in_ts < child_state->tile->lcu_offset_in_ts) {
fprintf(stderr, "Slice %d starts before tile %d, in which it should be included!\n", child_state->slice->id, child_state->tile->id);
return 0;
}
if (child_state->slice->end_in_ts > child_state->tile->lcu_offset_in_ts + child_state->tile->cur_pic->width_in_lcu * child_state->tile->cur_pic->height_in_lcu - 1) {
fprintf(stderr, "Slice %d ends after tile %d, in which it should be included!\n", child_state->slice->id, child_state->tile->id);
return 0;
}
}
#ifdef _DEBUG
if (!parent_state) encoder_state_dump_graphviz(child_state);
#endif //_DEBUG
return 1;
}
void encoder_state_finalize(encoder_state * const encoder_state) {
if (encoder_state->children) {
int i=0;
for (i = 0; encoder_state->children[i].encoder_control; ++i) {
encoder_state_finalize(&encoder_state->children[i]);
}
FREE_POINTER(encoder_state->children);
}
FREE_POINTER(encoder_state->lcu_order);
encoder_state->lcu_order_count = 0;
if (!encoder_state->parent || (encoder_state->parent->wfrow != encoder_state->wfrow)) {
encoder_state_config_wfrow_finalize(encoder_state);
FREE_POINTER(encoder_state->wfrow);
}
if (!encoder_state->parent || (encoder_state->parent->slice != encoder_state->slice)) {
encoder_state_config_slice_finalize(encoder_state);
FREE_POINTER(encoder_state->slice);
}
if (!encoder_state->parent || (encoder_state->parent->tile != encoder_state->tile)) {
encoder_state_config_tile_finalize(encoder_state);
FREE_POINTER(encoder_state->tile);
}
if (!encoder_state->parent || (encoder_state->parent->global != encoder_state->global)) {
encoder_state_config_global_finalize(encoder_state);
FREE_POINTER(encoder_state->global);
}
bitstream_finalize(&encoder_state->stream);
}

133
src/_encoderstategeometry.c Normal file
View file

@ -0,0 +1,133 @@
/*****************************************************************************
* This file is part of Kvazaar HEVC encoder.
*
* Copyright (C) 2013-2014 Tampere University of Technology and others (see
* COPYING file).
*
* Kvazaar is free software: you can redistribute it and/or modify
* it under the terms of the GNU General Public License version 2 as published
* by the Free Software Foundation.
*
* Kvazaar 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 General Public License for more details.
*
* You should have received a copy of the GNU General Public License
* along with Kvazaar. If not, see <http://www.gnu.org/licenses/>.
****************************************************************************/
//This file MUST NOT BE COMPILED directly. It's included in encoderstate.c
static int lcu_at_slice_start(const encoder_control * const encoder, int lcu_addr_in_ts) {
int i;
assert(lcu_addr_in_ts >= 0 && lcu_addr_in_ts < encoder->in.height_in_lcu * encoder->in.width_in_lcu);
if (lcu_addr_in_ts == 0) return 1;
for (i = 0; i < encoder->slice_count; ++i) {
if (encoder->slice_addresses_in_ts[i] == lcu_addr_in_ts) return 1;
}
return 0;
}
static int lcu_at_slice_end(const encoder_control * const encoder, int lcu_addr_in_ts) {
int i;
assert(lcu_addr_in_ts >= 0 && lcu_addr_in_ts < encoder->in.height_in_lcu * encoder->in.width_in_lcu);
if (lcu_addr_in_ts == encoder->in.height_in_lcu * encoder->in.width_in_lcu - 1) return 1;
for (i = 0; i < encoder->slice_count; ++i) {
if (encoder->slice_addresses_in_ts[i] == lcu_addr_in_ts + 1) return 1;
}
return 0;
}
static int lcu_at_tile_start(const encoder_control * const encoder, int lcu_addr_in_ts) {
assert(lcu_addr_in_ts >= 0 && lcu_addr_in_ts < encoder->in.height_in_lcu * encoder->in.width_in_lcu);
if (lcu_addr_in_ts == 0) return 1;
if (encoder->tiles_tile_id[lcu_addr_in_ts - 1] != encoder->tiles_tile_id[lcu_addr_in_ts]) {
return 1;
}
return 0;
}
static int lcu_at_tile_end(const encoder_control * const encoder, int lcu_addr_in_ts) {
assert(lcu_addr_in_ts >= 0 && lcu_addr_in_ts < encoder->in.height_in_lcu * encoder->in.width_in_lcu);
if (lcu_addr_in_ts == encoder->in.height_in_lcu * encoder->in.width_in_lcu - 1) return 1;
if (encoder->tiles_tile_id[lcu_addr_in_ts + 1] != encoder->tiles_tile_id[lcu_addr_in_ts]) {
return 1;
}
return 0;
}
//Return 1 if the LCU is at the first row of a structure (tile or slice)
static int lcu_in_first_row(const encoder_state * const encoder_state, int lcu_addr_in_ts) {
const int lcu_addr_in_rs = encoder_state->encoder_control->tiles_ctb_addr_ts_to_rs[lcu_addr_in_ts];
if (lcu_addr_in_rs / encoder_state->encoder_control->in.width_in_lcu == encoder_state->tile->lcu_offset_y) {
return 1;
}
if (lcu_addr_in_rs / encoder_state->encoder_control->in.width_in_lcu == encoder_state->slice->start_in_rs / encoder_state->encoder_control->in.width_in_lcu) {
return 1;
}
//One row above is before the start of the slice => it's also a boundary
if (lcu_addr_in_rs - encoder_state->encoder_control->in.width_in_lcu < encoder_state->slice->start_in_rs) {
return 1;
}
return 0;
}
//Return 1 if the LCU is at the first row of a structure (tile or slice)
static int lcu_in_last_row(const encoder_state * const encoder_state, int lcu_addr_in_ts) {
const int lcu_addr_in_rs = encoder_state->encoder_control->tiles_ctb_addr_ts_to_rs[lcu_addr_in_ts];
if (lcu_addr_in_rs / encoder_state->encoder_control->in.width_in_lcu == encoder_state->tile->lcu_offset_y + encoder_state->tile->cur_pic->height_in_lcu - 1) {
return 1;
}
if (lcu_addr_in_rs / encoder_state->encoder_control->in.width_in_lcu == encoder_state->slice->end_in_rs / encoder_state->encoder_control->in.width_in_lcu) {
return 1;
}
//One row below is before the end of the slice => it's also a boundary
if (lcu_addr_in_rs + encoder_state->encoder_control->in.width_in_lcu > encoder_state->slice->end_in_rs) {
return 1;
}
return 0;
}
//Return 1 if the LCU is at the first column of a structure (tile or slice)
static int lcu_in_first_column(const encoder_state * const encoder_state, int lcu_addr_in_ts) {
const int lcu_addr_in_rs = encoder_state->encoder_control->tiles_ctb_addr_ts_to_rs[lcu_addr_in_ts];
//First column of tile?
if (lcu_addr_in_rs % encoder_state->encoder_control->in.width_in_lcu == encoder_state->tile->lcu_offset_x) {
return 1;
}
//Slice start may not be aligned with the tile, so we need to allow this
if (lcu_addr_in_rs == encoder_state->slice->start_in_rs) {
return 1;
}
return 0;
}
//Return 1 if the LCU is at the last column of a structure (tile or slice)
static int lcu_in_last_column(const encoder_state * const encoder_state, int lcu_addr_in_ts) {
const int lcu_addr_in_rs = encoder_state->encoder_control->tiles_ctb_addr_ts_to_rs[lcu_addr_in_ts];
//First column of tile?
if (lcu_addr_in_rs % encoder_state->encoder_control->in.width_in_lcu == encoder_state->tile->lcu_offset_x + encoder_state->tile->cur_pic->width_in_lcu - 1) {
return 1;
}
//Slice start may not be aligned with the tile, so we need to allow this
if (lcu_addr_in_rs == encoder_state->slice->end_in_rs) {
return 1;
}
return 0;
}

1544
src/encoderstate.c
View file

File diff suppressed because it is too large Load diff