From: Suman Sunkara Date: Tue, 16 Nov 2010 20:09:26 +0000 (-0500) Subject: Merge branch 'experimental' of ssh://on2-git.corp.google.com:29418/libvpx into test X-Git-Tag: v1.3.0~1217^2~380^2~474 X-Git-Url: http://review.tizen.org/git/?a=commitdiff_plain;h=4b3f72001de952aaf3874d3ce2dca3cb3fbe3928;p=platform%2Fupstream%2Flibvpx.git Merge branch 'experimental' of ssh://on2-git.corp.google.com:29418/libvpx into test Conflicts: vp8/common/blockd.h vp8/decoder/decodemv.c vp8/decoder/decodframe.c vp8/decoder/demode.c vp8/decoder/onyxd_if.c vp8/decoder/onyxd_int.h vp8/encoder/encodeframe.c Change-Id: Ic379f4dffaded9796dc19d56be304d3f8527c61f --- 4b3f72001de952aaf3874d3ce2dca3cb3fbe3928 diff --cc vp8/common/blockd.h index 7a7b06b,a38f0b7..333f020 --- a/vp8/common/blockd.h +++ b/vp8/common/blockd.h @@@ -167,16 -168,15 +168,16 @@@ typedef struc int as_int; MV as_mv; } mv; - int partitioning; - int partition_count; - int mb_skip_coeff; //does this mb has coefficients at all, 1=no coefficients, 0=need decode tokens - int dc_diff; - unsigned char segment_id; // Which set of segmentation parameters should be used for this MB + unsigned char segment_flag; - int force_no_skip; - B_MODE_INFO partition_bmi[16]; + unsigned char partitioning; + unsigned char mb_skip_coeff; /* does this mb has coefficients at all, 1=no coefficients, 0=need decode tokens */ + unsigned char dc_diff; + unsigned char need_to_clamp_mvs; + unsigned char segment_id; /* Which set of segmentation parameters should be used for this MB */ + + unsigned char force_no_skip; /* encoder only */ } MB_MODE_INFO; @@@ -236,35 -235,29 +236,33 @@@ typedef struc int up_available; int left_available; - // Y,U,V,Y2 - ENTROPY_CONTEXT *above_context[4]; // row of context for each plane - ENTROPY_CONTEXT(*left_context)[4]; // (up to) 4 contexts "" + /* Y,U,V,Y2 */ + ENTROPY_CONTEXT_PLANES *above_context; + ENTROPY_CONTEXT_PLANES *left_context; - // 0 indicates segmentation at MB level is not enabled. Otherwise the individual bits indicate which features are active. + /* 0 indicates segmentation at MB level is not enabled. Otherwise the individual bits indicate which features are active. */ unsigned char segmentation_enabled; - // 0 (do not update) 1 (update) the macroblock segmentation map. + /* 0 (do not update) 1 (update) the macroblock segmentation map. */ unsigned char update_mb_segmentation_map; - // 0 (do not update) 1 (update) the macroblock segmentation feature data. + /* 0 (do not update) 1 (update) the macroblock segmentation feature data. */ unsigned char update_mb_segmentation_data; - // 0 (do not update) 1 (update) the macroblock segmentation feature data. + /* 0 (do not update) 1 (update) the macroblock segmentation feature data. */ unsigned char mb_segement_abs_delta; + unsigned char temporal_update; - // Per frame flags that define which MB level features (such as quantizer or loop filter level) - // are enabled and when enabled the proabilities used to decode the per MB flags in MB_MODE_INFO + /* Per frame flags that define which MB level features (such as quantizer or loop filter level) */ + /* are enabled and when enabled the proabilities used to decode the per MB flags in MB_MODE_INFO */ - vp8_prob mb_segment_tree_probs[MB_FEATURE_TREE_PROBS]; /* Probability Tree used to code Segment number */ - - signed char segment_feature_data[MB_LVL_MAX][MAX_MB_SEGMENTS]; /* Segment parameters */ +#if CONFIG_SEGMENTATION + vp8_prob mb_segment_tree_probs[MB_FEATURE_TREE_PROBS + 3]; // Probability Tree used to code Segment number +#else + vp8_prob mb_segment_tree_probs[MB_FEATURE_TREE_PROBS]; +#endif + signed char segment_feature_data[MB_LVL_MAX][MAX_MB_SEGMENTS]; // Segment parameters - // mode_based Loop filter adjustment + /* mode_based Loop filter adjustment */ unsigned char mode_ref_lf_delta_enabled; unsigned char mode_ref_lf_delta_update; diff --cc vp8/decoder/decodemv.c index f151ef3,203d72d..415392b --- a/vp8/decoder/decodemv.c +++ b/vp8/decoder/decodemv.c @@@ -106,261 -226,147 +226,189 @@@ unsigned int vp8_mv_cont_count[5][4] { 0, 0, 0, 0 }, { 0, 0, 0, 0 } }; + #endif - void vp8_decode_mode_mvs(VP8D_COMP *pbi) - { - const MV Zero = { 0, 0}; + unsigned char vp8_mbsplit_offset[4][16] = { + { 0, 8, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0}, + { 0, 2, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0}, + { 0, 2, 8, 10, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0}, + { 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15} + }; - VP8_COMMON *const pc = & pbi->common; - vp8_reader *const bc = & pbi->bc; - MACROBLOCKD *xd = &pbi->mb; - MODE_INFO *mi = pc->mi, *ms; - const int mis = pc->mode_info_stride; + unsigned char vp8_mbsplit_fill_count[4] = {8, 8, 4, 1}; + unsigned char vp8_mbsplit_fill_offset[4][16] = { + { 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15}, + { 0, 1, 4, 5, 8, 9, 12, 13, 2, 3, 6, 7, 10, 11, 14, 15}, + { 0, 1, 4, 5, 2, 3, 6, 7, 8, 9, 12, 13, 10, 11, 14, 15}, + { 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15} + }; - MV_CONTEXT *const mvc = pc->fc.mvc; - int mb_row = -1; + - + void vp8_mb_mode_mv_init(VP8D_COMP *pbi) + { + vp8_reader *const bc = & pbi->bc; + MV_CONTEXT *const mvc = pbi->common.fc.mvc; +#if CONFIG_SEGMENTATION - int left_id, above_id; - int i; - int sum; - int index = 0; ++ MACROBLOCKD *const xd = & pbi->mb; +#endif - vp8_prob prob_intra; - vp8_prob prob_last; - vp8_prob prob_gf; - vp8_prob prob_skip_false = 0; - - if (pc->mb_no_coeff_skip) - prob_skip_false = (vp8_prob)vp8_read_literal(bc, 8); - - prob_intra = (vp8_prob)vp8_read_literal(bc, 8); - prob_last = (vp8_prob)vp8_read_literal(bc, 8); - prob_gf = (vp8_prob)vp8_read_literal(bc, 8); - ms = pc->mi - 1; + pbi->prob_skip_false = 0; + if (pbi->common.mb_no_coeff_skip) + pbi->prob_skip_false = (vp8_prob)vp8_read_literal(bc, 8); - if (vp8_read_bit(bc)) + if(pbi->common.frame_type != KEY_FRAME) { - int i = 0; + pbi->prob_intra = (vp8_prob)vp8_read_literal(bc, 8); + pbi->prob_last = (vp8_prob)vp8_read_literal(bc, 8); + pbi->prob_gf = (vp8_prob)vp8_read_literal(bc, 8); - do + if (vp8_read_bit(bc)) { - pc->fc.ymode_prob[i] = (vp8_prob) vp8_read_literal(bc, 8); - } - while (++i < 4); - } + int i = 0; - if (vp8_read_bit(bc)) - { - int i = 0; + do + { + pbi->common.fc.ymode_prob[i] = (vp8_prob) vp8_read_literal(bc, 8); + } + while (++i < 4); + } - do + if (vp8_read_bit(bc)) { - pc->fc.uv_mode_prob[i] = (vp8_prob) vp8_read_literal(bc, 8); + int i = 0; + + do + { + pbi->common.fc.uv_mode_prob[i] = (vp8_prob) vp8_read_literal(bc, 8); + } + while (++i < 3); } - while (++i < 3); - } - read_mvcontexts(bc, mvc); + read_mvcontexts(bc, mvc); +#if CONFIG_SEGMENTATION + xd->temporal_update = vp8_read_bit(bc); +#endif - while (++mb_row < pc->mb_rows) - { - int mb_col = -1; + } + } - while (++mb_col < pc->mb_cols) - { - MB_MODE_INFO *const mbmi = & mi->mbmi; - MV *const mv = & mbmi->mv.as_mv; - VP8_COMMON *const pc = &pbi->common; - // MACROBLOCKD *xd = &pbi->mb; - - vp8dx_bool_decoder_fill(bc); - // Distance of Mb to the various image edges. - // These specified to 8th pel as they are always compared to MV values that are in 1/8th pel units - xd->mb_to_left_edge = -((mb_col * 16) << 3); - xd->mb_to_right_edge = ((pc->mb_cols - 1 - mb_col) * 16) << 3; - xd->mb_to_top_edge = -((mb_row * 16)) << 3; - xd->mb_to_bottom_edge = ((pc->mb_rows - 1 - mb_row) * 16) << 3; - - // If required read in new segmentation data for this MB - if (pbi->mb.update_mb_segmentation_map) + void vp8_read_mb_modes_mv(VP8D_COMP *pbi, MODE_INFO *mi, MB_MODE_INFO *mbmi, + int mb_row, int mb_col) + { + const MV Zero = { 0, 0}; + vp8_reader *const bc = & pbi->bc; + MV_CONTEXT *const mvc = pbi->common.fc.mvc; + const int mis = pbi->common.mode_info_stride; - ++#if CONFIG_SEGMENTATION ++ MACROBLOCKD *const xd = & pbi->mb; ++ int sum; ++ int index = mb_row * pbi->common.mb_cols + mb_col; ++#endif + MV *const mv = & mbmi->mv.as_mv; + int mb_to_left_edge; + int mb_to_right_edge; + int mb_to_top_edge; + int mb_to_bottom_edge; + + mb_to_top_edge = pbi->mb.mb_to_top_edge; + mb_to_bottom_edge = pbi->mb.mb_to_bottom_edge; + mb_to_top_edge -= LEFT_TOP_MARGIN; + mb_to_bottom_edge += RIGHT_BOTTOM_MARGIN; - + mbmi->need_to_clamp_mvs = 0; + /* Distance of Mb to the various image edges. + * These specified to 8th pel as they are always compared to MV values that are in 1/8th pel units + */ + pbi->mb.mb_to_left_edge = + mb_to_left_edge = -((mb_col * 16) << 3); + mb_to_left_edge -= LEFT_TOP_MARGIN; + + pbi->mb.mb_to_right_edge = + mb_to_right_edge = ((pbi->common.mb_cols - 1 - mb_col) * 16) << 3; + mb_to_right_edge += RIGHT_BOTTOM_MARGIN; + + /* If required read in new segmentation data for this MB */ + if (pbi->mb.update_mb_segmentation_map) - vp8_read_mb_features(bc, mbmi, &pbi->mb); + { +#if CONFIG_SEGMENTATION + if (xd->temporal_update) + { + sum = 0; + + if (mb_col != 0) + sum += (mi-1)->mbmi.segment_flag; + if (mb_row != 0) - sum += (mi-pc->mb_cols)->mbmi.segment_flag; ++ sum += (mi-pbi->common.mb_cols)->mbmi.segment_flag; + + if (vp8_read(bc, xd->mb_segment_tree_probs[3+sum]) == 0) + { + mbmi->segment_id = pbi->segmentation_map[index]; + mbmi->segment_flag = 0; + } + else + { + vp8_read_mb_features(bc, &mi->mbmi, &pbi->mb); + mbmi->segment_flag = 1; + pbi->segmentation_map[index] = mbmi->segment_id; + } + + } + else + { + vp8_read_mb_features(bc, &mi->mbmi, &pbi->mb); + pbi->segmentation_map[index] = mbmi->segment_id; + } + index++; +#else + vp8_read_mb_features(bc, &mi->mbmi, &pbi->mb); +#endif + } + - // Read the macroblock coeff skip flag if this feature is in use, else default to 0 - if (pc->mb_no_coeff_skip) - mbmi->mb_skip_coeff = vp8_read(bc, prob_skip_false); - else - mbmi->mb_skip_coeff = 0; - - mbmi->uv_mode = DC_PRED; - - if ((mbmi->ref_frame = (MV_REFERENCE_FRAME) vp8_read(bc, prob_intra))) /* inter MB */ - { - int rct[4]; - vp8_prob mv_ref_p [VP8_MVREFS-1]; - MV nearest, nearby, best_mv; - - if (vp8_read(bc, prob_last)) - { - mbmi->ref_frame = (MV_REFERENCE_FRAME)((int)mbmi->ref_frame + (int)(1 + vp8_read(bc, prob_gf))); - } - - vp8_find_near_mvs(xd, mi, &nearest, &nearby, &best_mv, rct, mbmi->ref_frame, pbi->common.ref_frame_sign_bias); - - vp8_mv_ref_probs(mv_ref_p, rct); - - switch (mbmi->mode = read_mv_ref(bc, mv_ref_p)) - { - case SPLITMV: - { - const int s = mbmi->partitioning = vp8_treed_read( - bc, vp8_mbsplit_tree, vp8_mbsplit_probs - ); - const int num_p = vp8_mbsplit_count [s]; - const int *const L = vp8_mbsplits [s]; - int j = 0; - - do /* for each subset j */ - { - B_MODE_INFO *const bmi = mbmi->partition_bmi + j; - MV *const mv = & bmi->mv.as_mv; - - int k = -1; /* first block in subset j */ - int mv_contz; - - while (j != L[++k]) - if (k >= 16) - #if CONFIG_DEBUG - assert(0); - - #else - ; - #endif - - mv_contz = vp8_mv_cont(&(vp8_left_bmi(mi, k)->mv.as_mv), &(vp8_above_bmi(mi, k, mis)->mv.as_mv)); - - switch (bmi->mode = (B_PREDICTION_MODE) sub_mv_ref(bc, vp8_sub_mv_ref_prob2 [mv_contz])) //pc->fc.sub_mv_ref_prob)) - { - case NEW4X4: - read_mv(bc, mv, (const MV_CONTEXT *) mvc); - mv->row += best_mv.row; - mv->col += best_mv.col; - #ifdef VPX_MODE_COUNT - vp8_mv_cont_count[mv_contz][3]++; - #endif - break; - case LEFT4X4: - *mv = vp8_left_bmi(mi, k)->mv.as_mv; - #ifdef VPX_MODE_COUNT - vp8_mv_cont_count[mv_contz][0]++; - #endif - break; - case ABOVE4X4: - *mv = vp8_above_bmi(mi, k, mis)->mv.as_mv; - #ifdef VPX_MODE_COUNT - vp8_mv_cont_count[mv_contz][1]++; - #endif - break; - case ZERO4X4: - *mv = Zero; - #ifdef VPX_MODE_COUNT - vp8_mv_cont_count[mv_contz][2]++; - #endif - break; - default: - break; - } - - /* Fill (uniform) modes, mvs of jth subset. - Must do it here because ensuing subsets can - refer back to us via "left" or "above". */ - do - if (j == L[k]) - mi->bmi[k] = *bmi; - - while (++k < 16); - } - while (++j < num_p); - } - - *mv = mi->bmi[15].mv.as_mv; + /* Read the macroblock coeff skip flag if this feature is in use, else default to 0 */ + if (pbi->common.mb_no_coeff_skip) + mbmi->mb_skip_coeff = vp8_read(bc, pbi->prob_skip_false); + else + mbmi->mb_skip_coeff = 0; - break; /* done with SPLITMV */ - - case NEARMV: - *mv = nearby; - - // Clip "next_nearest" so that it does not extend to far out of image - if (mv->col < (xd->mb_to_left_edge - LEFT_TOP_MARGIN)) - mv->col = xd->mb_to_left_edge - LEFT_TOP_MARGIN; - else if (mv->col > xd->mb_to_right_edge + RIGHT_BOTTOM_MARGIN) - mv->col = xd->mb_to_right_edge + RIGHT_BOTTOM_MARGIN; + if ((mbmi->ref_frame = (MV_REFERENCE_FRAME) vp8_read(bc, pbi->prob_intra))) /* inter MB */ + { + int rct[4]; + vp8_prob mv_ref_p [VP8_MVREFS-1]; + MV nearest, nearby, best_mv; - if (mv->row < (xd->mb_to_top_edge - LEFT_TOP_MARGIN)) - mv->row = xd->mb_to_top_edge - LEFT_TOP_MARGIN; - else if (mv->row > xd->mb_to_bottom_edge + RIGHT_BOTTOM_MARGIN) - mv->row = xd->mb_to_bottom_edge + RIGHT_BOTTOM_MARGIN; + if (vp8_read(bc, pbi->prob_last)) + { + mbmi->ref_frame = (MV_REFERENCE_FRAME)((int)mbmi->ref_frame + (int)(1 + vp8_read(bc, pbi->prob_gf))); + } - goto propagate_mv; + vp8_find_near_mvs(&pbi->mb, mi, &nearest, &nearby, &best_mv, rct, mbmi->ref_frame, pbi->common.ref_frame_sign_bias); - case NEARESTMV: - *mv = nearest; + vp8_mv_ref_probs(mv_ref_p, rct); - // Clip "next_nearest" so that it does not extend to far out of image - if (mv->col < (xd->mb_to_left_edge - LEFT_TOP_MARGIN)) - mv->col = xd->mb_to_left_edge - LEFT_TOP_MARGIN; - else if (mv->col > xd->mb_to_right_edge + RIGHT_BOTTOM_MARGIN) - mv->col = xd->mb_to_right_edge + RIGHT_BOTTOM_MARGIN; + mbmi->uv_mode = DC_PRED; + switch (mbmi->mode = read_mv_ref(bc, mv_ref_p)) + { + case SPLITMV: + { + const int s = mbmi->partitioning = + vp8_treed_read(bc, vp8_mbsplit_tree, vp8_mbsplit_probs); + const int num_p = vp8_mbsplit_count [s]; + int j = 0; - if (mv->row < (xd->mb_to_top_edge - LEFT_TOP_MARGIN)) - mv->row = xd->mb_to_top_edge - LEFT_TOP_MARGIN; - else if (mv->row > xd->mb_to_bottom_edge + RIGHT_BOTTOM_MARGIN) - mv->row = xd->mb_to_bottom_edge + RIGHT_BOTTOM_MARGIN; + do /* for each subset j */ + { + B_MODE_INFO bmi; + MV *const mv = & bmi.mv.as_mv; - goto propagate_mv; + int k; /* first block in subset j */ + int mv_contz; + k = vp8_mbsplit_offset[s][j]; - case ZEROMV: - *mv = Zero; - goto propagate_mv; + mv_contz = vp8_mv_cont(&(vp8_left_bmi(mi, k)->mv.as_mv), &(vp8_above_bmi(mi, k, mis)->mv.as_mv)); - case NEWMV: + switch (bmi.mode = (B_PREDICTION_MODE) sub_mv_ref(bc, vp8_sub_mv_ref_prob2 [mv_contz])) /*pc->fc.sub_mv_ref_prob))*/ + { + case NEW4X4: read_mv(bc, mv, (const MV_CONTEXT *) mvc); mv->row += best_mv.row; mv->col += best_mv.col; diff --cc vp8/decoder/decodframe.c index 01cd7ae,4702fae..06204fe --- a/vp8/decoder/decodframe.c +++ b/vp8/decoder/decodframe.c @@@ -559,7 -565,8 +565,7 @@@ int vp8_decode_frame(VP8D_COMP *pbi MACROBLOCKD *const xd = & pbi->mb; const unsigned char *data = (const unsigned char *)pbi->Source; const unsigned char *const data_end = data + pbi->source_sz; - int first_partition_length_in_bytes; + ptrdiff_t first_partition_length_in_bytes; - int mb_row; int i, j, k, l; const int *const mb_feature_data_bits = vp8_mb_feature_data_bits; @@@ -630,11 -648,12 +647,11 @@@ pc->clamp_type = (CLAMP_TYPE)vp8_read_bit(bc); } - // Is segmentation enabled + /* Is segmentation enabled */ xd->segmentation_enabled = (unsigned char)vp8_read_bit(bc); - if (xd->segmentation_enabled) { - // Signal whether or not the segmentation map is being explicitly updated this frame. + /* Signal whether or not the segmentation map is being explicitly updated this frame. */ xd->update_mb_segmentation_map = (unsigned char)vp8_read_bit(bc); xd->update_mb_segmentation_data = (unsigned char)vp8_read_bit(bc); @@@ -665,16 -684,13 +682,16 @@@ if (xd->update_mb_segmentation_map) { - // Which macro block level features are enabled + /* Which macro block level features are enabled */ vpx_memset(xd->mb_segment_tree_probs, 255, sizeof(xd->mb_segment_tree_probs)); - +#if CONFIG_SEGMENTATION - // Read the probs used to decode the segment id for each macro block. + /* Read the probs used to decode the segment id for each macro block. */ + for (i = 0; i < MB_FEATURE_TREE_PROBS+3; i++) +#else for (i = 0; i < MB_FEATURE_TREE_PROBS; i++) +#endif { - // If not explicitly set value is defaulted to 255 by memset above + /* If not explicitly set value is defaulted to 255 by memset above */ if (vp8_read_bit(bc)) xd->mb_segment_tree_probs[i] = (vp8_prob)vp8_read_literal(bc, 8); } @@@ -793,9 -810,9 +811,8 @@@ fclose(z); } - vp8dx_bool_decoder_fill(bc); - { - // read coef probability tree + /* read coef probability tree */ for (i = 0; i < BLOCK_TYPES; i++) for (j = 0; j < COEF_BANDS; j++) @@@ -813,16 -830,12 +830,17 @@@ } } - vpx_memcpy(&xd->pre, &pc->last_frame, sizeof(YV12_BUFFER_CONFIG)); - vpx_memcpy(&xd->dst, &pc->new_frame, sizeof(YV12_BUFFER_CONFIG)); + vpx_memcpy(&xd->pre, &pc->yv12_fb[pc->lst_fb_idx], sizeof(YV12_BUFFER_CONFIG)); + vpx_memcpy(&xd->dst, &pc->yv12_fb[pc->new_fb_idx], sizeof(YV12_BUFFER_CONFIG)); +#if CONFIG_SEGMENTATION + // Create the encoder segmentation map and set all entries to 0 + CHECK_MEM_ERROR(pbi->segmentation_map, vpx_calloc((pc->mb_rows * pc->mb_cols), 1)); +#endif + - // set up frame new frame for intra coded blocks - vp8_setup_intra_recon(&pc->new_frame); + /* set up frame new frame for intra coded blocks */ + if (!(pbi->b_multithreaded_rd) || pc->multi_token_partition == ONE_PARTITION || !(pc->filter_level)) + vp8_setup_intra_recon(&pc->yv12_fb[pc->new_fb_idx]); vp8_setup_block_dptrs(xd); diff --cc vp8/decoder/onyxd_if.c index 8d2b267,aa2709f..063b6a4 --- a/vp8/decoder/onyxd_if.c +++ b/vp8/decoder/onyxd_if.c @@@ -137,11 -145,11 +143,16 @@@ void vp8dx_remove_decompressor(VP8D_PT if (!pbi) return; +#if CONFIG_SEGMENTATION + // Delete sementation map + if (pbi->segmentation_map != 0) + vpx_free(pbi->segmentation_map); +#endif + + #if CONFIG_MULTITHREAD + if (pbi->b_multithreaded_rd) + vp8mt_de_alloc_temp_buffers(pbi, pbi->common.mb_rows); + #endif vp8_decoder_remove_threads(pbi); vp8_remove_common(&pbi->common); vpx_free(pbi); @@@ -299,20 -435,13 +438,15 @@@ int vp8dx_receive_compressed_data(VP8D_ cm->last_frame_type = cm->frame_type; cm->last_filter_type = cm->filter_type; cm->last_sharpness_level = cm->sharpness_level; - } - - vpx_usec_timer_mark(&lpftimer); - pbi->time_loop_filtering += vpx_usec_timer_elapsed(&lpftimer); + vp8_yv12_extend_frame_borders_ptr(cm->frame_to_show); } - vp8_yv12_extend_frame_borders_ptr(cm->frame_to_show); - - - // DEBUG code #if 0 + /* DEBUG code */ + /*vp8_recon_write_yuv_frame("recon.yuv", cm->frame_to_show);*/ + vp8_recon_write_yuv_frame("recon.yuv", cm->frame_to_show); + if (cm->current_video_frame <= 5) write_dx_frame_to_file(cm->frame_to_show, cm->current_video_frame); #endif diff --cc vp8/decoder/onyxd_int.h index e8b5d40,7593edf..fc1811d --- a/vp8/decoder/onyxd_int.h +++ b/vp8/decoder/onyxd_int.h @@@ -93,22 -92,29 +92,28 @@@ typedef struct VP8Decompresso int current_mb_col_main; int decoding_thread_count; int allocated_decoding_thread_count; - // variable for threading - DECLARE_ALIGNED(16, MACROBLOCKD, lpfmb); - + /* variable for threading */ #if CONFIG_MULTITHREAD - pthread_t h_thread_lpf; // thread for postprocessing - sem_t h_event_lpf; // Event for post_proc completed - sem_t h_event_start_lpf; - #endif + int mt_baseline_filter_level[MAX_MB_SEGMENTS]; + int sync_range; + int *mt_current_mb_col; /* Each row remembers its already decoded column. */ + + unsigned char **mt_yabove_row; /* mb_rows x width */ + unsigned char **mt_uabove_row; + unsigned char **mt_vabove_row; + unsigned char **mt_yleft_col; /* mb_rows x 16 */ + unsigned char **mt_uleft_col; /* mb_rows x 8 */ + unsigned char **mt_vleft_col; /* mb_rows x 8 */ + MB_ROW_DEC *mb_row_di; - DECODETHREAD_DATA *de_thread_data; - #if CONFIG_MULTITHREAD + DECODETHREAD_DATA *de_thread_data; + pthread_t *h_decoding_thread; - sem_t *h_event_mbrdecoding; - sem_t h_event_main; - // end of threading data + sem_t *h_event_start_decoding; + sem_t h_event_end_decoding; + /* end of threading data */ #endif + vp8_reader *mbc; INT64 last_time_stamp; int ready_for_new_data; diff --cc vp8/encoder/encodeframe.c index faf8e2b,2aac20b..b7d231b --- a/vp8/encoder/encodeframe.c +++ b/vp8/encoder/encodeframe.c @@@ -263,18 -389,15 +397,17 @@@ void encode_mb_row(VP8_COMP *cpi int i; int recon_yoffset, recon_uvoffset; int mb_col; - int recon_y_stride = cm->last_frame.y_stride; - int recon_uv_stride = cm->last_frame.uv_stride; + int ref_fb_idx = cm->lst_fb_idx; + int dst_fb_idx = cm->new_fb_idx; + int recon_y_stride = cm->yv12_fb[ref_fb_idx].y_stride; + int recon_uv_stride = cm->yv12_fb[ref_fb_idx].uv_stride; int seg_map_index = (mb_row * cpi->common.mb_cols); - - +#if CONFIG_SEGMENTATION + int left_id, above_id; + int sum; +#endif // reset above block coeffs - xd->above_context[Y1CONTEXT] = cm->above_context[Y1CONTEXT]; - xd->above_context[UCONTEXT ] = cm->above_context[UCONTEXT ]; - xd->above_context[VCONTEXT ] = cm->above_context[VCONTEXT ]; - xd->above_context[Y2CONTEXT] = cm->above_context[Y2CONTEXT]; + xd->above_context = cm->above_context; xd->up_available = (mb_row != 0); recon_yoffset = (mb_row * recon_y_stride * 16); @@@ -310,15 -443,14 +453,15 @@@ { // Code to set segment id in xd->mbmi.segment_id for current MB (with range checking) if (cpi->segmentation_map[seg_map_index+mb_col] <= 3) - xd->mbmi.segment_id = cpi->segmentation_map[seg_map_index+mb_col]; + xd->mode_info_context->mbmi.segment_id = cpi->segmentation_map[seg_map_index+mb_col]; else - xd->mbmi.segment_id = 0; + xd->mode_info_context->mbmi.segment_id = 0; vp8cx_mb_init_quantizer(cpi, x); + } else - xd->mbmi.segment_id = 0; // Set to Segment 0 by default + xd->mode_info_context->mbmi.segment_id = 0; // Set to Segment 0 by default x->active_ptr = cpi->active_map + seg_map_index + mb_col; @@@ -377,16 -509,8 +520,13 @@@ cpi->tplist[mb_row].stop = *tp; - xd->gf_active_ptr++; // Increment pointer into gf useage flags structure for next mb + x->gf_active_ptr++; // Increment pointer into gf useage flags structure for next mb - if ((xd->mbmi.mode == ZEROMV) && (xd->mbmi.ref_frame == LAST_FRAME)) - xd->mbmi.segment_id = 0; ++ if ((xd->mode_info_context->mbmi.mode == ZEROMV) && (xd->mode_info_context->mbmi.ref_frame == LAST_FRAME)) ++ xd->mode_info_context->mbmi.segment_id = 0; + else - xd->mbmi.segment_id = 1; - - // store macroblock mode info into context array - vpx_memcpy(&xd->mode_info_context->mbmi, &xd->mbmi, sizeof(xd->mbmi)); ++ xd->mode_info_context->mbmi.segment_id = 1; + for (i = 0; i < 16; i++) vpx_memcpy(&xd->mode_info_context->bmi[i], &xd->block[i].bmi, sizeof(xd->block[i].bmi)); @@@ -398,49 -522,14 +538,47 @@@ recon_yoffset += 16; recon_uvoffset += 8; - // Keep track of segment useage - segment_counts[xd->mode_info_context->mbmi.segment_id] ++; +#if CONFIG_SEGMENTATION + //cpi->segmentation_map[mb_row * cm->mb_cols + mb_col] = xd->mbmi.segment_id; + if (cm->frame_type == KEY_FRAME) + { + segment_counts[xd->mode_info_context->mbmi.segment_id] ++; + } + else + { + sum = 0; + if (mb_col != 0) + sum += (xd->mode_info_context-1)->mbmi.segment_flag; + if (mb_row != 0) + sum += (xd->mode_info_context-cm->mb_cols)->mbmi.segment_flag; + - if (xd->mbmi.segment_id == cpi->segmentation_map[(mb_row*cm->mb_cols) + mb_col]) ++ if (xd->mode_info_context->mbmi.segment_id == cpi->segmentation_map[(mb_row*cm->mb_cols) + mb_col]) + xd->mode_info_context->mbmi.segment_flag = 0; + else + xd->mode_info_context->mbmi.segment_flag = 1; + if (xd->mode_info_context->mbmi.segment_flag == 0) + { + segment_counts[SEEK_SAMEID + sum]++; + segment_counts[10]++; + } + else + { + segment_counts[SEEK_DIFFID + sum]++; + segment_counts[11]++; + //calculate individual segment ids + segment_counts[xd->mode_info_context->mbmi.segment_id] ++; + } + } - segment_counts[SEEK_SEGID + xd->mbmi.segment_id] ++; ++ segment_counts[SEEK_SEGID + xd->mode_info_context->mbmi.segment_id] ++; +#else + segment_counts[xd->mode_info_context->mbmi.segment_id] ++; +#endif // skip to next mb xd->mode_info_context++; + x->partition_info++; - xd->above_context[Y1CONTEXT] += 4; - xd->above_context[UCONTEXT ] += 2; - xd->above_context[VCONTEXT ] += 2; - xd->above_context[Y2CONTEXT] ++; + xd->above_context++; cpi->current_mb_col_main = mb_col; } @@@ -453,8 -542,13 +591,9 @@@ // this is to account for the border xd->mode_info_context++; + x->partition_info++; } - - - - void vp8_encode_frame(VP8_COMP *cpi) { int mb_row; @@@ -464,41 -558,34 +603,40 @@@ int i; TOKENEXTRA *tp = cpi->tok; +#if CONFIG_SEGMENTATION + int segment_counts[MAX_MB_SEGMENTS + SEEK_SEGID]; + int prob[3]; + int new_cost, original_cost; +#else int segment_counts[MAX_MB_SEGMENTS]; +#endif int totalrate; - if (cm->frame_type != KEY_FRAME) + // Functions setup for all frame types so we can use MC in AltRef + if (cm->mcomp_filter_type == SIXTAP) { - if (cm->mcomp_filter_type == SIXTAP) - { - xd->subpixel_predict = SUBPIX_INVOKE(&cpi->common.rtcd.subpix, sixtap4x4); - xd->subpixel_predict8x4 = SUBPIX_INVOKE(&cpi->common.rtcd.subpix, sixtap8x4); - xd->subpixel_predict8x8 = SUBPIX_INVOKE(&cpi->common.rtcd.subpix, sixtap8x8); - xd->subpixel_predict16x16 = SUBPIX_INVOKE(&cpi->common.rtcd.subpix, sixtap16x16); - } - else - { - xd->subpixel_predict = SUBPIX_INVOKE(&cpi->common.rtcd.subpix, bilinear4x4); - xd->subpixel_predict8x4 = SUBPIX_INVOKE(&cpi->common.rtcd.subpix, bilinear8x4); - xd->subpixel_predict8x8 = SUBPIX_INVOKE(&cpi->common.rtcd.subpix, bilinear8x8); - xd->subpixel_predict16x16 = SUBPIX_INVOKE(&cpi->common.rtcd.subpix, bilinear16x16); - } + xd->subpixel_predict = SUBPIX_INVOKE( + &cpi->common.rtcd.subpix, sixtap4x4); + xd->subpixel_predict8x4 = SUBPIX_INVOKE( + &cpi->common.rtcd.subpix, sixtap8x4); + xd->subpixel_predict8x8 = SUBPIX_INVOKE( + &cpi->common.rtcd.subpix, sixtap8x8); + xd->subpixel_predict16x16 = SUBPIX_INVOKE( + &cpi->common.rtcd.subpix, sixtap16x16); + } + else + { + xd->subpixel_predict = SUBPIX_INVOKE( + &cpi->common.rtcd.subpix, bilinear4x4); + xd->subpixel_predict8x4 = SUBPIX_INVOKE( + &cpi->common.rtcd.subpix, bilinear8x4); + xd->subpixel_predict8x8 = SUBPIX_INVOKE( + &cpi->common.rtcd.subpix, bilinear8x8); + xd->subpixel_predict16x16 = SUBPIX_INVOKE( + &cpi->common.rtcd.subpix, bilinear16x16); } - //else // Key Frame - //{ - // For key frames make sure the intra ref frame probability value - // is set to "all intra" - //cpi->prob_intra_coded = 255; - //} - - xd->gf_active_ptr = (signed char *)cm->gf_active_flags; // Point to base of GF active flags data structure + x->gf_active_ptr = (signed char *)cpi->gf_active_flags; // Point to base of GF active flags data structure x->vector_range = 32; diff --cc vp8/encoder/onyx_if.c index 9bc7f52,00ecf97..9dbeeb2 --- a/vp8/encoder/onyx_if.c +++ b/vp8/encoder/onyx_if.c @@@ -2028,12 -2198,8 +2193,9 @@@ VP8_PTR vp8_create_compressor(VP8_CONFI cpi->cyclic_refresh_map = (signed char *) NULL; // Test function for segmentation + //segmentation_test_function((VP8_PTR) cpi); - // Loop filter mode / ref deltas test function - //mode_ref_lf_test_function(cpi); - #ifdef ENTROPY_STATS init_context_counters(); #endif diff --cc vpxenc.c index 0000000,228a8cd..80eba50 mode 000000,100644..100644 --- a/vpxenc.c +++ b/vpxenc.c @@@ -1,0 -1,1698 +1,1699 @@@ + /* + * Copyright (c) 2010 The WebM project authors. All Rights Reserved. + * + * Use of this source code is governed by a BSD-style license + * that can be found in the LICENSE file in the root of the source + * tree. An additional intellectual property rights grant can be found + * in the file PATENTS. All contributing project authors may + * be found in the AUTHORS file in the root of the source tree. + */ + + + /* This is a simple program that encodes YV12 files and generates ivf + * files using the new interface. + */ + #if defined(_WIN32) + #define USE_POSIX_MMAP 0 + #else + #define USE_POSIX_MMAP 1 + #endif + + #include + #include + #include + #include + #include + #include "vpx/vpx_encoder.h" + #if USE_POSIX_MMAP + #include + #include + #include + #include + #include + #endif ++#include "vpx_config.h" + #include "vpx_version.h" + #include "vpx/vp8cx.h" + #include "vpx_ports/mem_ops.h" + #include "vpx_ports/vpx_timer.h" + #include "tools_common.h" + #include "y4minput.h" + #include "libmkv/EbmlWriter.h" + #include "libmkv/EbmlIDs.h" + + /* Need special handling of these functions on Windows */ + #if defined(_MSC_VER) + /* MSVS doesn't define off_t, and uses _f{seek,tell}i64 */ + typedef __int64 off_t; + #define fseeko _fseeki64 + #define ftello _ftelli64 + #elif defined(_WIN32) + /* MinGW defines off_t, and uses f{seek,tell}o64 */ + #define fseeko fseeko64 + #define ftello ftello64 + #endif + + #if defined(_MSC_VER) + #define LITERALU64(n) n + #else + #define LITERALU64(n) n##LLU + #endif + + static const char *exec_name; + + static const struct codec_item + { + char const *name; + const vpx_codec_iface_t *iface; + unsigned int fourcc; + } codecs[] = + { + #if CONFIG_EXPERIMENTAL && CONFIG_VP8_ENCODER + {"vp8x", &vpx_codec_vp8x_cx_algo, 0x78385056}, + #endif + #if CONFIG_VP8_ENCODER + {"vp8", &vpx_codec_vp8_cx_algo, 0x30385056}, + #endif + }; + + static void usage_exit(); + + void die(const char *fmt, ...) + { + va_list ap; + va_start(ap, fmt); + vfprintf(stderr, fmt, ap); + fprintf(stderr, "\n"); + usage_exit(); + } + + static void ctx_exit_on_error(vpx_codec_ctx_t *ctx, const char *s) + { + if (ctx->err) + { + const char *detail = vpx_codec_error_detail(ctx); + + fprintf(stderr, "%s: %s\n", s, vpx_codec_error(ctx)); + + if (detail) + fprintf(stderr, " %s\n", detail); + + exit(EXIT_FAILURE); + } + } + + /* This structure is used to abstract the different ways of handling + * first pass statistics. + */ + typedef struct + { + vpx_fixed_buf_t buf; + int pass; + FILE *file; + char *buf_ptr; + size_t buf_alloc_sz; + } stats_io_t; + + int stats_open_file(stats_io_t *stats, const char *fpf, int pass) + { + int res; + + stats->pass = pass; + + if (pass == 0) + { + stats->file = fopen(fpf, "wb"); + stats->buf.sz = 0; + stats->buf.buf = NULL, + res = (stats->file != NULL); + } + else + { + #if 0 + #elif USE_POSIX_MMAP + struct stat stat_buf; + int fd; + + fd = open(fpf, O_RDONLY); + stats->file = fdopen(fd, "rb"); + fstat(fd, &stat_buf); + stats->buf.sz = stat_buf.st_size; + stats->buf.buf = mmap(NULL, stats->buf.sz, PROT_READ, MAP_PRIVATE, + fd, 0); + res = (stats->buf.buf != NULL); + #else + size_t nbytes; + + stats->file = fopen(fpf, "rb"); + + if (fseek(stats->file, 0, SEEK_END)) + { + fprintf(stderr, "First-pass stats file must be seekable!\n"); + exit(EXIT_FAILURE); + } + + stats->buf.sz = stats->buf_alloc_sz = ftell(stats->file); + rewind(stats->file); + + stats->buf.buf = malloc(stats->buf_alloc_sz); + + if (!stats->buf.buf) + { + fprintf(stderr, "Failed to allocate first-pass stats buffer (%d bytes)\n", + stats->buf_alloc_sz); + exit(EXIT_FAILURE); + } + + nbytes = fread(stats->buf.buf, 1, stats->buf.sz, stats->file); + res = (nbytes == stats->buf.sz); + #endif + } + + return res; + } + + int stats_open_mem(stats_io_t *stats, int pass) + { + int res; + stats->pass = pass; + + if (!pass) + { + stats->buf.sz = 0; + stats->buf_alloc_sz = 64 * 1024; + stats->buf.buf = malloc(stats->buf_alloc_sz); + } + + stats->buf_ptr = stats->buf.buf; + res = (stats->buf.buf != NULL); + return res; + } + + + void stats_close(stats_io_t *stats) + { + if (stats->file) + { + if (stats->pass == 1) + { + #if 0 + #elif USE_POSIX_MMAP + munmap(stats->buf.buf, stats->buf.sz); + #else + free(stats->buf.buf); + #endif + } + + fclose(stats->file); + stats->file = NULL; + } + else + { + if (stats->pass == 1) + free(stats->buf.buf); + } + } + + void stats_write(stats_io_t *stats, const void *pkt, size_t len) + { + if (stats->file) + { + if(fwrite(pkt, 1, len, stats->file)); + } + else + { + if (stats->buf.sz + len > stats->buf_alloc_sz) + { + size_t new_sz = stats->buf_alloc_sz + 64 * 1024; + char *new_ptr = realloc(stats->buf.buf, new_sz); + + if (new_ptr) + { + stats->buf_ptr = new_ptr + (stats->buf_ptr - (char *)stats->buf.buf); + stats->buf.buf = new_ptr; + stats->buf_alloc_sz = new_sz; + } /* else ... */ + } + + memcpy(stats->buf_ptr, pkt, len); + stats->buf.sz += len; + stats->buf_ptr += len; + } + } + + vpx_fixed_buf_t stats_get(stats_io_t *stats) + { + return stats->buf; + } + + enum video_file_type + { + FILE_TYPE_RAW, + FILE_TYPE_IVF, + FILE_TYPE_Y4M + }; + + struct detect_buffer { + char buf[4]; + int valid; + }; + + + #define IVF_FRAME_HDR_SZ (4+8) /* 4 byte size + 8 byte timestamp */ + static int read_frame(FILE *f, vpx_image_t *img, unsigned int file_type, + y4m_input *y4m, struct detect_buffer *detect) + { + int plane = 0; + int shortread = 0; + + if (file_type == FILE_TYPE_Y4M) + { + if (y4m_input_fetch_frame(y4m, f, img) < 1) + return 0; + } + else + { + if (file_type == FILE_TYPE_IVF) + { + char junk[IVF_FRAME_HDR_SZ]; + + /* Skip the frame header. We know how big the frame should be. See + * write_ivf_frame_header() for documentation on the frame header + * layout. + */ + if(fread(junk, 1, IVF_FRAME_HDR_SZ, f)); + } + + for (plane = 0; plane < 3; plane++) + { + unsigned char *ptr; + int w = (plane ? (1 + img->d_w) / 2 : img->d_w); + int h = (plane ? (1 + img->d_h) / 2 : img->d_h); + int r; + + /* Determine the correct plane based on the image format. The for-loop + * always counts in Y,U,V order, but this may not match the order of + * the data on disk. + */ + switch (plane) + { + case 1: + ptr = img->planes[img->fmt==VPX_IMG_FMT_YV12? VPX_PLANE_V : VPX_PLANE_U]; + break; + case 2: + ptr = img->planes[img->fmt==VPX_IMG_FMT_YV12?VPX_PLANE_U : VPX_PLANE_V]; + break; + default: + ptr = img->planes[plane]; + } + + for (r = 0; r < h; r++) + { + if (detect->valid) + { + memcpy(ptr, detect->buf, 4); + shortread |= fread(ptr+4, 1, w-4, f) < w-4; + detect->valid = 0; + } + else + shortread |= fread(ptr, 1, w, f) < w; + + ptr += img->stride[plane]; + } + } + } + + return !shortread; + } + + + unsigned int file_is_y4m(FILE *infile, + y4m_input *y4m, + char detect[4]) + { + if(memcmp(detect, "YUV4", 4) == 0) + { + return 1; + } + return 0; + } + + #define IVF_FILE_HDR_SZ (32) + unsigned int file_is_ivf(FILE *infile, + unsigned int *fourcc, + unsigned int *width, + unsigned int *height, + char detect[4]) + { + char raw_hdr[IVF_FILE_HDR_SZ]; + int is_ivf = 0; + + if(memcmp(detect, "DKIF", 4) != 0) + return 0; + + /* See write_ivf_file_header() for more documentation on the file header + * layout. + */ + if (fread(raw_hdr + 4, 1, IVF_FILE_HDR_SZ - 4, infile) + == IVF_FILE_HDR_SZ - 4) + { + { + is_ivf = 1; + + if (mem_get_le16(raw_hdr + 4) != 0) + fprintf(stderr, "Error: Unrecognized IVF version! This file may not" + " decode properly."); + + *fourcc = mem_get_le32(raw_hdr + 8); + } + } + + if (is_ivf) + { + *width = mem_get_le16(raw_hdr + 12); + *height = mem_get_le16(raw_hdr + 14); + } + + return is_ivf; + } + + + static void write_ivf_file_header(FILE *outfile, + const vpx_codec_enc_cfg_t *cfg, + unsigned int fourcc, + int frame_cnt) + { + char header[32]; + + if (cfg->g_pass != VPX_RC_ONE_PASS && cfg->g_pass != VPX_RC_LAST_PASS) + return; + + header[0] = 'D'; + header[1] = 'K'; + header[2] = 'I'; + header[3] = 'F'; + mem_put_le16(header + 4, 0); /* version */ + mem_put_le16(header + 6, 32); /* headersize */ + mem_put_le32(header + 8, fourcc); /* headersize */ + mem_put_le16(header + 12, cfg->g_w); /* width */ + mem_put_le16(header + 14, cfg->g_h); /* height */ + mem_put_le32(header + 16, cfg->g_timebase.den); /* rate */ + mem_put_le32(header + 20, cfg->g_timebase.num); /* scale */ + mem_put_le32(header + 24, frame_cnt); /* length */ + mem_put_le32(header + 28, 0); /* unused */ + + if(fwrite(header, 1, 32, outfile)); + } + + + static void write_ivf_frame_header(FILE *outfile, + const vpx_codec_cx_pkt_t *pkt) + { + char header[12]; + vpx_codec_pts_t pts; + + if (pkt->kind != VPX_CODEC_CX_FRAME_PKT) + return; + + pts = pkt->data.frame.pts; + mem_put_le32(header, pkt->data.frame.sz); + mem_put_le32(header + 4, pts & 0xFFFFFFFF); + mem_put_le32(header + 8, pts >> 32); + + if(fwrite(header, 1, 12, outfile)); + } + + + typedef off_t EbmlLoc; + + + struct cue_entry + { + unsigned int time; + uint64_t loc; + }; + + + struct EbmlGlobal + { + int debug; + + FILE *stream; + int64_t last_pts_ms; + vpx_rational_t framerate; + + /* These pointers are to the start of an element */ + off_t position_reference; + off_t seek_info_pos; + off_t segment_info_pos; + off_t track_pos; + off_t cue_pos; + off_t cluster_pos; + + /* This pointer is to a specific element to be serialized */ + off_t track_id_pos; + + /* These pointers are to the size field of the element */ + EbmlLoc startSegment; + EbmlLoc startCluster; + + uint32_t cluster_timecode; + int cluster_open; + + struct cue_entry *cue_list; + unsigned int cues; + + }; + + + void Ebml_Write(EbmlGlobal *glob, const void *buffer_in, unsigned long len) + { + if(fwrite(buffer_in, 1, len, glob->stream)); + } + + + void Ebml_Serialize(EbmlGlobal *glob, const void *buffer_in, unsigned long len) + { + const unsigned char *q = (const unsigned char *)buffer_in + len - 1; + + for(; len; len--) + Ebml_Write(glob, q--, 1); + } + + + /* Need a fixed size serializer for the track ID. libmkv provdes a 64 bit + * one, but not a 32 bit one. + */ + static void Ebml_SerializeUnsigned32(EbmlGlobal *glob, unsigned long class_id, uint64_t ui) + { + unsigned char sizeSerialized = 4 | 0x80; + Ebml_WriteID(glob, class_id); + Ebml_Serialize(glob, &sizeSerialized, 1); + Ebml_Serialize(glob, &ui, 4); + } + + + static void + Ebml_StartSubElement(EbmlGlobal *glob, EbmlLoc *ebmlLoc, + unsigned long class_id) + { + //todo this is always taking 8 bytes, this may need later optimization + //this is a key that says lenght unknown + unsigned long long unknownLen = LITERALU64(0x01FFFFFFFFFFFFFF); + + Ebml_WriteID(glob, class_id); + *ebmlLoc = ftello(glob->stream); + Ebml_Serialize(glob, &unknownLen, 8); + } + + static void + Ebml_EndSubElement(EbmlGlobal *glob, EbmlLoc *ebmlLoc) + { + off_t pos; + uint64_t size; + + /* Save the current stream pointer */ + pos = ftello(glob->stream); + + /* Calculate the size of this element */ + size = pos - *ebmlLoc - 8; + size |= LITERALU64(0x0100000000000000); + + /* Seek back to the beginning of the element and write the new size */ + fseeko(glob->stream, *ebmlLoc, SEEK_SET); + Ebml_Serialize(glob, &size, 8); + + /* Reset the stream pointer */ + fseeko(glob->stream, pos, SEEK_SET); + } + + + static void + write_webm_seek_element(EbmlGlobal *ebml, unsigned long id, off_t pos) + { + uint64_t offset = pos - ebml->position_reference; + EbmlLoc start; + Ebml_StartSubElement(ebml, &start, Seek); + Ebml_SerializeBinary(ebml, SeekID, id); + Ebml_SerializeUnsigned64(ebml, SeekPosition, offset); + Ebml_EndSubElement(ebml, &start); + } + + + static void + write_webm_seek_info(EbmlGlobal *ebml) + { + + off_t pos; + + /* Save the current stream pointer */ + pos = ftello(ebml->stream); + + if(ebml->seek_info_pos) + fseeko(ebml->stream, ebml->seek_info_pos, SEEK_SET); + else + ebml->seek_info_pos = pos; + + { + EbmlLoc start; + + Ebml_StartSubElement(ebml, &start, SeekHead); + write_webm_seek_element(ebml, Tracks, ebml->track_pos); + write_webm_seek_element(ebml, Cues, ebml->cue_pos); + write_webm_seek_element(ebml, Info, ebml->segment_info_pos); + Ebml_EndSubElement(ebml, &start); + } + { + //segment info + EbmlLoc startInfo; + uint64_t frame_time; + + frame_time = (uint64_t)1000 * ebml->framerate.den + / ebml->framerate.num; + ebml->segment_info_pos = ftello(ebml->stream); + Ebml_StartSubElement(ebml, &startInfo, Info); + Ebml_SerializeUnsigned(ebml, TimecodeScale, 1000000); + Ebml_SerializeFloat(ebml, Segment_Duration, + ebml->last_pts_ms + frame_time); + Ebml_SerializeString(ebml, 0x4D80, + ebml->debug ? "vpxenc" : "vpxenc" VERSION_STRING); + Ebml_SerializeString(ebml, 0x5741, + ebml->debug ? "vpxenc" : "vpxenc" VERSION_STRING); + Ebml_EndSubElement(ebml, &startInfo); + } + } + + + static void + write_webm_file_header(EbmlGlobal *glob, + const vpx_codec_enc_cfg_t *cfg, + const struct vpx_rational *fps) + { + { + EbmlLoc start; + Ebml_StartSubElement(glob, &start, EBML); + Ebml_SerializeUnsigned(glob, EBMLVersion, 1); + Ebml_SerializeUnsigned(glob, EBMLReadVersion, 1); //EBML Read Version + Ebml_SerializeUnsigned(glob, EBMLMaxIDLength, 4); //EBML Max ID Length + Ebml_SerializeUnsigned(glob, EBMLMaxSizeLength, 8); //EBML Max Size Length + Ebml_SerializeString(glob, DocType, "webm"); //Doc Type + Ebml_SerializeUnsigned(glob, DocTypeVersion, 2); //Doc Type Version + Ebml_SerializeUnsigned(glob, DocTypeReadVersion, 2); //Doc Type Read Version + Ebml_EndSubElement(glob, &start); + } + { + Ebml_StartSubElement(glob, &glob->startSegment, Segment); //segment + glob->position_reference = ftello(glob->stream); + glob->framerate = *fps; + write_webm_seek_info(glob); + + { + EbmlLoc trackStart; + glob->track_pos = ftello(glob->stream); + Ebml_StartSubElement(glob, &trackStart, Tracks); + { + unsigned int trackNumber = 1; + uint64_t trackID = 0; + + EbmlLoc start; + Ebml_StartSubElement(glob, &start, TrackEntry); + Ebml_SerializeUnsigned(glob, TrackNumber, trackNumber); + glob->track_id_pos = ftello(glob->stream); + Ebml_SerializeUnsigned32(glob, TrackUID, trackID); + Ebml_SerializeUnsigned(glob, TrackType, 1); //video is always 1 + Ebml_SerializeString(glob, CodecID, "V_VP8"); + { + unsigned int pixelWidth = cfg->g_w; + unsigned int pixelHeight = cfg->g_h; + float frameRate = (float)fps->num/(float)fps->den; + + EbmlLoc videoStart; + Ebml_StartSubElement(glob, &videoStart, Video); + Ebml_SerializeUnsigned(glob, PixelWidth, pixelWidth); + Ebml_SerializeUnsigned(glob, PixelHeight, pixelHeight); + Ebml_SerializeFloat(glob, FrameRate, frameRate); + Ebml_EndSubElement(glob, &videoStart); //Video + } + Ebml_EndSubElement(glob, &start); //Track Entry + } + Ebml_EndSubElement(glob, &trackStart); + } + // segment element is open + } + } + + + static void + write_webm_block(EbmlGlobal *glob, + const vpx_codec_enc_cfg_t *cfg, + const vpx_codec_cx_pkt_t *pkt) + { + unsigned long block_length; + unsigned char track_number; + unsigned short block_timecode = 0; + unsigned char flags; + int64_t pts_ms; + int start_cluster = 0, is_keyframe; + + /* Calculate the PTS of this frame in milliseconds */ + pts_ms = pkt->data.frame.pts * 1000 + * (uint64_t)cfg->g_timebase.num / (uint64_t)cfg->g_timebase.den; + if(pts_ms <= glob->last_pts_ms) + pts_ms = glob->last_pts_ms + 1; + glob->last_pts_ms = pts_ms; + + /* Calculate the relative time of this block */ + if(pts_ms - glob->cluster_timecode > SHRT_MAX) + start_cluster = 1; + else + block_timecode = pts_ms - glob->cluster_timecode; + + is_keyframe = (pkt->data.frame.flags & VPX_FRAME_IS_KEY); + if(start_cluster || is_keyframe) + { + if(glob->cluster_open) + Ebml_EndSubElement(glob, &glob->startCluster); + + /* Open the new cluster */ + block_timecode = 0; + glob->cluster_open = 1; + glob->cluster_timecode = pts_ms; + glob->cluster_pos = ftello(glob->stream); + Ebml_StartSubElement(glob, &glob->startCluster, Cluster); //cluster + Ebml_SerializeUnsigned(glob, Timecode, glob->cluster_timecode); + + /* Save a cue point if this is a keyframe. */ + if(is_keyframe) + { + struct cue_entry *cue; + + glob->cue_list = realloc(glob->cue_list, + (glob->cues+1) * sizeof(struct cue_entry)); + cue = &glob->cue_list[glob->cues]; + cue->time = glob->cluster_timecode; + cue->loc = glob->cluster_pos; + glob->cues++; + } + } + + /* Write the Simple Block */ + Ebml_WriteID(glob, SimpleBlock); + + block_length = pkt->data.frame.sz + 4; + block_length |= 0x10000000; + Ebml_Serialize(glob, &block_length, 4); + + track_number = 1; + track_number |= 0x80; + Ebml_Write(glob, &track_number, 1); + + Ebml_Serialize(glob, &block_timecode, 2); + + flags = 0; + if(is_keyframe) + flags |= 0x80; + if(pkt->data.frame.flags & VPX_FRAME_IS_INVISIBLE) + flags |= 0x08; + Ebml_Write(glob, &flags, 1); + + Ebml_Write(glob, pkt->data.frame.buf, pkt->data.frame.sz); + } + + + static void + write_webm_file_footer(EbmlGlobal *glob, long hash) + { + + if(glob->cluster_open) + Ebml_EndSubElement(glob, &glob->startCluster); + + { + EbmlLoc start; + int i; + + glob->cue_pos = ftello(glob->stream); + Ebml_StartSubElement(glob, &start, Cues); + for(i=0; icues; i++) + { + struct cue_entry *cue = &glob->cue_list[i]; + EbmlLoc start; + + Ebml_StartSubElement(glob, &start, CuePoint); + { + EbmlLoc start; + + Ebml_SerializeUnsigned(glob, CueTime, cue->time); + + Ebml_StartSubElement(glob, &start, CueTrackPositions); + Ebml_SerializeUnsigned(glob, CueTrack, 1); + Ebml_SerializeUnsigned64(glob, CueClusterPosition, + cue->loc - glob->position_reference); + //Ebml_SerializeUnsigned(glob, CueBlockNumber, cue->blockNumber); + Ebml_EndSubElement(glob, &start); + } + Ebml_EndSubElement(glob, &start); + } + Ebml_EndSubElement(glob, &start); + } + + Ebml_EndSubElement(glob, &glob->startSegment); + + /* Patch up the seek info block */ + write_webm_seek_info(glob); + + /* Patch up the track id */ + fseeko(glob->stream, glob->track_id_pos, SEEK_SET); + Ebml_SerializeUnsigned32(glob, TrackUID, glob->debug ? 0xDEADBEEF : hash); + + fseeko(glob->stream, 0, SEEK_END); + } + + + /* Murmur hash derived from public domain reference implementation at + * http://sites.google.com/site/murmurhash/ + */ + static unsigned int murmur ( const void * key, int len, unsigned int seed ) + { + const unsigned int m = 0x5bd1e995; + const int r = 24; + + unsigned int h = seed ^ len; + + const unsigned char * data = (const unsigned char *)key; + + while(len >= 4) + { + unsigned int k; + + k = data[0]; + k |= data[1] << 8; + k |= data[2] << 16; + k |= data[3] << 24; + + k *= m; + k ^= k >> r; + k *= m; + + h *= m; + h ^= k; + + data += 4; + len -= 4; + } + + switch(len) + { + case 3: h ^= data[2] << 16; + case 2: h ^= data[1] << 8; + case 1: h ^= data[0]; + h *= m; + }; + + h ^= h >> 13; + h *= m; + h ^= h >> 15; + + return h; + } + + #include "math.h" + + static double vp8_mse2psnr(double Samples, double Peak, double Mse) + { + double psnr; + + if ((double)Mse > 0.0) + psnr = 10.0 * log10(Peak * Peak * Samples / Mse); + else + psnr = 60; // Limit to prevent / 0 + + if (psnr > 60) + psnr = 60; + + return psnr; + } + + + #include "args.h" + + static const arg_def_t debugmode = ARG_DEF("D", "debug", 0, + "Debug mode (makes output deterministic)"); + static const arg_def_t outputfile = ARG_DEF("o", "output", 1, + "Output filename"); + static const arg_def_t use_yv12 = ARG_DEF(NULL, "yv12", 0, + "Input file is YV12 "); + static const arg_def_t use_i420 = ARG_DEF(NULL, "i420", 0, + "Input file is I420 (default)"); + static const arg_def_t codecarg = ARG_DEF(NULL, "codec", 1, + "Codec to use"); + static const arg_def_t passes = ARG_DEF("p", "passes", 1, + "Number of passes (1/2)"); + static const arg_def_t pass_arg = ARG_DEF(NULL, "pass", 1, + "Pass to execute (1/2)"); + static const arg_def_t fpf_name = ARG_DEF(NULL, "fpf", 1, + "First pass statistics file name"); + static const arg_def_t limit = ARG_DEF(NULL, "limit", 1, + "Stop encoding after n input frames"); + static const arg_def_t deadline = ARG_DEF("d", "deadline", 1, + "Deadline per frame (usec)"); + static const arg_def_t best_dl = ARG_DEF(NULL, "best", 0, + "Use Best Quality Deadline"); + static const arg_def_t good_dl = ARG_DEF(NULL, "good", 0, + "Use Good Quality Deadline"); + static const arg_def_t rt_dl = ARG_DEF(NULL, "rt", 0, + "Use Realtime Quality Deadline"); + static const arg_def_t verbosearg = ARG_DEF("v", "verbose", 0, + "Show encoder parameters"); + static const arg_def_t psnrarg = ARG_DEF(NULL, "psnr", 0, + "Show PSNR in status line"); + static const arg_def_t framerate = ARG_DEF(NULL, "fps", 1, + "Stream frame rate (rate/scale)"); + static const arg_def_t use_ivf = ARG_DEF(NULL, "ivf", 0, + "Output IVF (default is WebM)"); + static const arg_def_t *main_args[] = + { + &debugmode, + &outputfile, &codecarg, &passes, &pass_arg, &fpf_name, &limit, &deadline, + &best_dl, &good_dl, &rt_dl, + &verbosearg, &psnrarg, &use_ivf, &framerate, + NULL + }; + + static const arg_def_t usage = ARG_DEF("u", "usage", 1, + "Usage profile number to use"); + static const arg_def_t threads = ARG_DEF("t", "threads", 1, + "Max number of threads to use"); + static const arg_def_t profile = ARG_DEF(NULL, "profile", 1, + "Bitstream profile number to use"); + static const arg_def_t width = ARG_DEF("w", "width", 1, + "Frame width"); + static const arg_def_t height = ARG_DEF("h", "height", 1, + "Frame height"); + static const arg_def_t timebase = ARG_DEF(NULL, "timebase", 1, + "Stream timebase (frame duration)"); + static const arg_def_t error_resilient = ARG_DEF(NULL, "error-resilient", 1, + "Enable error resiliency features"); + static const arg_def_t lag_in_frames = ARG_DEF(NULL, "lag-in-frames", 1, + "Max number of frames to lag"); + + static const arg_def_t *global_args[] = + { + &use_yv12, &use_i420, &usage, &threads, &profile, + &width, &height, &timebase, &framerate, &error_resilient, + &lag_in_frames, NULL + }; + + static const arg_def_t dropframe_thresh = ARG_DEF(NULL, "drop-frame", 1, + "Temporal resampling threshold (buf %)"); + static const arg_def_t resize_allowed = ARG_DEF(NULL, "resize-allowed", 1, + "Spatial resampling enabled (bool)"); + static const arg_def_t resize_up_thresh = ARG_DEF(NULL, "resize-up", 1, + "Upscale threshold (buf %)"); + static const arg_def_t resize_down_thresh = ARG_DEF(NULL, "resize-down", 1, + "Downscale threshold (buf %)"); + static const arg_def_t end_usage = ARG_DEF(NULL, "end-usage", 1, + "VBR=0 | CBR=1"); + static const arg_def_t target_bitrate = ARG_DEF(NULL, "target-bitrate", 1, + "Bitrate (kbps)"); + static const arg_def_t min_quantizer = ARG_DEF(NULL, "min-q", 1, + "Minimum (best) quantizer"); + static const arg_def_t max_quantizer = ARG_DEF(NULL, "max-q", 1, + "Maximum (worst) quantizer"); + static const arg_def_t undershoot_pct = ARG_DEF(NULL, "undershoot-pct", 1, + "Datarate undershoot (min) target (%)"); + static const arg_def_t overshoot_pct = ARG_DEF(NULL, "overshoot-pct", 1, + "Datarate overshoot (max) target (%)"); + static const arg_def_t buf_sz = ARG_DEF(NULL, "buf-sz", 1, + "Client buffer size (ms)"); + static const arg_def_t buf_initial_sz = ARG_DEF(NULL, "buf-initial-sz", 1, + "Client initial buffer size (ms)"); + static const arg_def_t buf_optimal_sz = ARG_DEF(NULL, "buf-optimal-sz", 1, + "Client optimal buffer size (ms)"); + static const arg_def_t *rc_args[] = + { + &dropframe_thresh, &resize_allowed, &resize_up_thresh, &resize_down_thresh, + &end_usage, &target_bitrate, &min_quantizer, &max_quantizer, + &undershoot_pct, &overshoot_pct, &buf_sz, &buf_initial_sz, &buf_optimal_sz, + NULL + }; + + + static const arg_def_t bias_pct = ARG_DEF(NULL, "bias-pct", 1, + "CBR/VBR bias (0=CBR, 100=VBR)"); + static const arg_def_t minsection_pct = ARG_DEF(NULL, "minsection-pct", 1, + "GOP min bitrate (% of target)"); + static const arg_def_t maxsection_pct = ARG_DEF(NULL, "maxsection-pct", 1, + "GOP max bitrate (% of target)"); + static const arg_def_t *rc_twopass_args[] = + { + &bias_pct, &minsection_pct, &maxsection_pct, NULL + }; + + + static const arg_def_t kf_min_dist = ARG_DEF(NULL, "kf-min-dist", 1, + "Minimum keyframe interval (frames)"); + static const arg_def_t kf_max_dist = ARG_DEF(NULL, "kf-max-dist", 1, + "Maximum keyframe interval (frames)"); + static const arg_def_t kf_disabled = ARG_DEF(NULL, "disable-kf", 0, + "Disable keyframe placement"); + static const arg_def_t *kf_args[] = + { + &kf_min_dist, &kf_max_dist, &kf_disabled, NULL + }; + + + #if CONFIG_VP8_ENCODER + static const arg_def_t noise_sens = ARG_DEF(NULL, "noise-sensitivity", 1, + "Noise sensitivity (frames to blur)"); + static const arg_def_t sharpness = ARG_DEF(NULL, "sharpness", 1, + "Filter sharpness (0-7)"); + static const arg_def_t static_thresh = ARG_DEF(NULL, "static-thresh", 1, + "Motion detection threshold"); + #endif + + #if CONFIG_VP8_ENCODER + static const arg_def_t cpu_used = ARG_DEF(NULL, "cpu-used", 1, + "CPU Used (-16..16)"); + #endif + + + #if CONFIG_VP8_ENCODER + static const arg_def_t token_parts = ARG_DEF(NULL, "token-parts", 1, + "Number of token partitions to use, log2"); + static const arg_def_t auto_altref = ARG_DEF(NULL, "auto-alt-ref", 1, + "Enable automatic alt reference frames"); + static const arg_def_t arnr_maxframes = ARG_DEF(NULL, "arnr-maxframes", 1, + "AltRef Max Frames"); + static const arg_def_t arnr_strength = ARG_DEF(NULL, "arnr-strength", 1, + "AltRef Strength"); + static const arg_def_t arnr_type = ARG_DEF(NULL, "arnr-type", 1, + "AltRef Type"); + + static const arg_def_t *vp8_args[] = + { + &cpu_used, &auto_altref, &noise_sens, &sharpness, &static_thresh, + &token_parts, &arnr_maxframes, &arnr_strength, &arnr_type, NULL + }; + static const int vp8_arg_ctrl_map[] = + { + VP8E_SET_CPUUSED, VP8E_SET_ENABLEAUTOALTREF, + VP8E_SET_NOISE_SENSITIVITY, VP8E_SET_SHARPNESS, VP8E_SET_STATIC_THRESHOLD, + VP8E_SET_TOKEN_PARTITIONS, + VP8E_SET_ARNR_MAXFRAMES, VP8E_SET_ARNR_STRENGTH , VP8E_SET_ARNR_TYPE, 0 + }; + #endif + + static const arg_def_t *no_args[] = { NULL }; + + static void usage_exit() + { + int i; + + fprintf(stderr, "Usage: %s -o dst_filename src_filename \n", + exec_name); + + fprintf(stderr, "\nOptions:\n"); + arg_show_usage(stdout, main_args); + fprintf(stderr, "\nEncoder Global Options:\n"); + arg_show_usage(stdout, global_args); + fprintf(stderr, "\nRate Control Options:\n"); + arg_show_usage(stdout, rc_args); + fprintf(stderr, "\nTwopass Rate Control Options:\n"); + arg_show_usage(stdout, rc_twopass_args); + fprintf(stderr, "\nKeyframe Placement Options:\n"); + arg_show_usage(stdout, kf_args); + #if CONFIG_VP8_ENCODER + fprintf(stderr, "\nVP8 Specific Options:\n"); + arg_show_usage(stdout, vp8_args); + #endif + fprintf(stderr, "\n" + "Included encoders:\n" + "\n"); + + for (i = 0; i < sizeof(codecs) / sizeof(codecs[0]); i++) + fprintf(stderr, " %-6s - %s\n", + codecs[i].name, + vpx_codec_iface_name(codecs[i].iface)); + + exit(EXIT_FAILURE); + } + + #define ARG_CTRL_CNT_MAX 10 + + + int main(int argc, const char **argv_) + { + vpx_codec_ctx_t encoder; + const char *in_fn = NULL, *out_fn = NULL, *stats_fn = NULL; + int i; + FILE *infile, *outfile; + vpx_codec_enc_cfg_t cfg; + vpx_codec_err_t res; + int pass, one_pass_only = 0; + stats_io_t stats; + vpx_image_t raw; + const struct codec_item *codec = codecs; + int frame_avail, got_data; + + struct arg arg; + char **argv, **argi, **argj; + int arg_usage = 0, arg_passes = 1, arg_deadline = 0; + int arg_ctrls[ARG_CTRL_CNT_MAX][2], arg_ctrl_cnt = 0; + int arg_limit = 0; + static const arg_def_t **ctrl_args = no_args; + static const int *ctrl_args_map = NULL; + int verbose = 0, show_psnr = 0; + int arg_use_i420 = 1; + unsigned long cx_time = 0; + unsigned int file_type, fourcc; + y4m_input y4m; + struct vpx_rational arg_framerate = {30, 1}; + int arg_have_framerate = 0; + int write_webm = 1; + EbmlGlobal ebml = {0}; + uint32_t hash = 0; + uint64_t psnr_sse_total = 0; + uint64_t psnr_samples_total = 0; + double psnr_totals[4] = {0, 0, 0, 0}; + int psnr_count = 0; + + exec_name = argv_[0]; + ebml.last_pts_ms = -1; + + if (argc < 3) + usage_exit(); + + + /* First parse the codec and usage values, because we want to apply other + * parameters on top of the default configuration provided by the codec. + */ + argv = argv_dup(argc - 1, argv_ + 1); + + for (argi = argj = argv; (*argj = *argi); argi += arg.argv_step) + { + arg.argv_step = 1; + + if (arg_match(&arg, &codecarg, argi)) + { + int j, k = -1; + + for (j = 0; j < sizeof(codecs) / sizeof(codecs[0]); j++) + if (!strcmp(codecs[j].name, arg.val)) + k = j; + + if (k >= 0) + codec = codecs + k; + else + die("Error: Unrecognized argument (%s) to --codec\n", + arg.val); + + } + else if (arg_match(&arg, &passes, argi)) + { + arg_passes = arg_parse_uint(&arg); + + if (arg_passes < 1 || arg_passes > 2) + die("Error: Invalid number of passes (%d)\n", arg_passes); + } + else if (arg_match(&arg, &pass_arg, argi)) + { + one_pass_only = arg_parse_uint(&arg); + + if (one_pass_only < 1 || one_pass_only > 2) + die("Error: Invalid pass selected (%d)\n", one_pass_only); + } + else if (arg_match(&arg, &fpf_name, argi)) + stats_fn = arg.val; + else if (arg_match(&arg, &usage, argi)) + arg_usage = arg_parse_uint(&arg); + else if (arg_match(&arg, &deadline, argi)) + arg_deadline = arg_parse_uint(&arg); + else if (arg_match(&arg, &best_dl, argi)) + arg_deadline = VPX_DL_BEST_QUALITY; + else if (arg_match(&arg, &good_dl, argi)) + arg_deadline = VPX_DL_GOOD_QUALITY; + else if (arg_match(&arg, &rt_dl, argi)) + arg_deadline = VPX_DL_REALTIME; + else if (arg_match(&arg, &use_yv12, argi)) + { + arg_use_i420 = 0; + } + else if (arg_match(&arg, &use_i420, argi)) + { + arg_use_i420 = 1; + } + else if (arg_match(&arg, &verbosearg, argi)) + verbose = 1; + else if (arg_match(&arg, &limit, argi)) + arg_limit = arg_parse_uint(&arg); + else if (arg_match(&arg, &psnrarg, argi)) + show_psnr = 1; + else if (arg_match(&arg, &framerate, argi)) + { + arg_framerate = arg_parse_rational(&arg); + arg_have_framerate = 1; + } + else if (arg_match(&arg, &use_ivf, argi)) + write_webm = 0; + else if (arg_match(&arg, &outputfile, argi)) + out_fn = arg.val; + else if (arg_match(&arg, &debugmode, argi)) + ebml.debug = 1; + else + argj++; + } + + /* Ensure that --passes and --pass are consistent. If --pass is set and --passes=2, + * ensure --fpf was set. + */ + if (one_pass_only) + { + /* DWIM: Assume the user meant passes=2 if pass=2 is specified */ + if (one_pass_only > arg_passes) + { + fprintf(stderr, "Warning: Assuming --pass=%d implies --passes=%d\n", + one_pass_only, one_pass_only); + arg_passes = one_pass_only; + } + + if (arg_passes == 2 && !stats_fn) + die("Must specify --fpf when --pass=%d and --passes=2\n", one_pass_only); + } + + /* Populate encoder configuration */ + res = vpx_codec_enc_config_default(codec->iface, &cfg, arg_usage); + + if (res) + { + fprintf(stderr, "Failed to get config: %s\n", + vpx_codec_err_to_string(res)); + return EXIT_FAILURE; + } + + /* Change the default timebase to a high enough value so that the encoder + * will always create strictly increasing timestamps. + */ + cfg.g_timebase.den = 1000; + + /* Never use the library's default resolution, require it be parsed + * from the file or set on the command line. + */ + cfg.g_w = 0; + cfg.g_h = 0; + + /* Now parse the remainder of the parameters. */ + for (argi = argj = argv; (*argj = *argi); argi += arg.argv_step) + { + arg.argv_step = 1; + + if (0); + else if (arg_match(&arg, &threads, argi)) + cfg.g_threads = arg_parse_uint(&arg); + else if (arg_match(&arg, &profile, argi)) + cfg.g_profile = arg_parse_uint(&arg); + else if (arg_match(&arg, &width, argi)) + cfg.g_w = arg_parse_uint(&arg); + else if (arg_match(&arg, &height, argi)) + cfg.g_h = arg_parse_uint(&arg); + else if (arg_match(&arg, &timebase, argi)) + cfg.g_timebase = arg_parse_rational(&arg); + else if (arg_match(&arg, &error_resilient, argi)) + cfg.g_error_resilient = arg_parse_uint(&arg); + else if (arg_match(&arg, &lag_in_frames, argi)) + cfg.g_lag_in_frames = arg_parse_uint(&arg); + else if (arg_match(&arg, &dropframe_thresh, argi)) + cfg.rc_dropframe_thresh = arg_parse_uint(&arg); + else if (arg_match(&arg, &resize_allowed, argi)) + cfg.rc_resize_allowed = arg_parse_uint(&arg); + else if (arg_match(&arg, &resize_up_thresh, argi)) + cfg.rc_resize_up_thresh = arg_parse_uint(&arg); + else if (arg_match(&arg, &resize_down_thresh, argi)) + cfg.rc_resize_down_thresh = arg_parse_uint(&arg); + else if (arg_match(&arg, &resize_down_thresh, argi)) + cfg.rc_resize_down_thresh = arg_parse_uint(&arg); + else if (arg_match(&arg, &end_usage, argi)) + cfg.rc_end_usage = arg_parse_uint(&arg); + else if (arg_match(&arg, &target_bitrate, argi)) + cfg.rc_target_bitrate = arg_parse_uint(&arg); + else if (arg_match(&arg, &min_quantizer, argi)) + cfg.rc_min_quantizer = arg_parse_uint(&arg); + else if (arg_match(&arg, &max_quantizer, argi)) + cfg.rc_max_quantizer = arg_parse_uint(&arg); + else if (arg_match(&arg, &undershoot_pct, argi)) + cfg.rc_undershoot_pct = arg_parse_uint(&arg); + else if (arg_match(&arg, &overshoot_pct, argi)) + cfg.rc_overshoot_pct = arg_parse_uint(&arg); + else if (arg_match(&arg, &buf_sz, argi)) + cfg.rc_buf_sz = arg_parse_uint(&arg); + else if (arg_match(&arg, &buf_initial_sz, argi)) + cfg.rc_buf_initial_sz = arg_parse_uint(&arg); + else if (arg_match(&arg, &buf_optimal_sz, argi)) + cfg.rc_buf_optimal_sz = arg_parse_uint(&arg); + else if (arg_match(&arg, &bias_pct, argi)) + { + cfg.rc_2pass_vbr_bias_pct = arg_parse_uint(&arg); + + if (arg_passes < 2) + fprintf(stderr, + "Warning: option %s ignored in one-pass mode.\n", + arg.name); + } + else if (arg_match(&arg, &minsection_pct, argi)) + { + cfg.rc_2pass_vbr_minsection_pct = arg_parse_uint(&arg); + + if (arg_passes < 2) + fprintf(stderr, + "Warning: option %s ignored in one-pass mode.\n", + arg.name); + } + else if (arg_match(&arg, &maxsection_pct, argi)) + { + cfg.rc_2pass_vbr_maxsection_pct = arg_parse_uint(&arg); + + if (arg_passes < 2) + fprintf(stderr, + "Warning: option %s ignored in one-pass mode.\n", + arg.name); + } + else if (arg_match(&arg, &kf_min_dist, argi)) + cfg.kf_min_dist = arg_parse_uint(&arg); + else if (arg_match(&arg, &kf_max_dist, argi)) + cfg.kf_max_dist = arg_parse_uint(&arg); + else if (arg_match(&arg, &kf_disabled, argi)) + cfg.kf_mode = VPX_KF_DISABLED; + else + argj++; + } + + /* Handle codec specific options */ + #if CONFIG_VP8_ENCODER + + if (codec->iface == &vpx_codec_vp8_cx_algo + #if CONFIG_EXPERIMENTAL + || codec->iface == &vpx_codec_vp8x_cx_algo + #endif + ) + { + ctrl_args = vp8_args; + ctrl_args_map = vp8_arg_ctrl_map; + } + + #endif + + for (argi = argj = argv; (*argj = *argi); argi += arg.argv_step) + { + int match = 0; + + arg.argv_step = 1; + + for (i = 0; ctrl_args[i]; i++) + { + if (arg_match(&arg, ctrl_args[i], argi)) + { + match = 1; + + if (arg_ctrl_cnt < ARG_CTRL_CNT_MAX) + { + arg_ctrls[arg_ctrl_cnt][0] = ctrl_args_map[i]; + arg_ctrls[arg_ctrl_cnt][1] = arg_parse_int(&arg); + arg_ctrl_cnt++; + } + } + } + + if (!match) + argj++; + } + + /* Check for unrecognized options */ + for (argi = argv; *argi; argi++) + if (argi[0][0] == '-' && argi[0][1]) + die("Error: Unrecognized option %s\n", *argi); + + /* Handle non-option arguments */ + in_fn = argv[0]; + + if (!in_fn) + usage_exit(); + + if(!out_fn) + die("Error: Output file is required (specify with -o)\n"); + + memset(&stats, 0, sizeof(stats)); + + for (pass = one_pass_only ? one_pass_only - 1 : 0; pass < arg_passes; pass++) + { + int frames_in = 0, frames_out = 0; + unsigned long nbytes = 0; + size_t detect_bytes; + struct detect_buffer detect; + + /* Parse certain options from the input file, if possible */ + infile = strcmp(in_fn, "-") ? fopen(in_fn, "rb") + : set_binary_mode(stdin); + + if (!infile) + { + fprintf(stderr, "Failed to open input file\n"); + return EXIT_FAILURE; + } + + /* For RAW input sources, these bytes will applied on the first frame + * in read_frame(). + * We can always read 4 bytes because the minimum supported frame size + * is 2x2. + */ + detect_bytes = fread(detect.buf, 1, 4, infile); + detect.valid = 0; + + if (detect_bytes == 4 && file_is_y4m(infile, &y4m, detect.buf)) + { + if (y4m_input_open(&y4m, infile, detect.buf, 4) >= 0) + { + file_type = FILE_TYPE_Y4M; + cfg.g_w = y4m.pic_w; + cfg.g_h = y4m.pic_h; + + /* Use the frame rate from the file only if none was specified + * on the command-line. + */ + if (!arg_have_framerate) + { + arg_framerate.num = y4m.fps_n; + arg_framerate.den = y4m.fps_d; + } + + arg_use_i420 = 0; + } + else + { + fprintf(stderr, "Unsupported Y4M stream.\n"); + return EXIT_FAILURE; + } + } + else if (detect_bytes == 4 && + file_is_ivf(infile, &fourcc, &cfg.g_w, &cfg.g_h, detect.buf)) + { + file_type = FILE_TYPE_IVF; + switch (fourcc) + { + case 0x32315659: + arg_use_i420 = 0; + break; + case 0x30323449: + arg_use_i420 = 1; + break; + default: + fprintf(stderr, "Unsupported fourcc (%08x) in IVF\n", fourcc); + return EXIT_FAILURE; + } + } + else + { + file_type = FILE_TYPE_RAW; + detect.valid = 1; + } + + if(!cfg.g_w || !cfg.g_h) + { + fprintf(stderr, "Specify stream dimensions with --width (-w) " + " and --height (-h).\n"); + return EXIT_FAILURE; + } + + #define SHOW(field) fprintf(stderr, " %-28s = %d\n", #field, cfg.field) + + if (verbose && pass == 0) + { + fprintf(stderr, "Codec: %s\n", vpx_codec_iface_name(codec->iface)); + fprintf(stderr, "Source file: %s Format: %s\n", in_fn, + arg_use_i420 ? "I420" : "YV12"); + fprintf(stderr, "Destination file: %s\n", out_fn); + fprintf(stderr, "Encoder parameters:\n"); + + SHOW(g_usage); + SHOW(g_threads); + SHOW(g_profile); + SHOW(g_w); + SHOW(g_h); + SHOW(g_timebase.num); + SHOW(g_timebase.den); + SHOW(g_error_resilient); + SHOW(g_pass); + SHOW(g_lag_in_frames); + SHOW(rc_dropframe_thresh); + SHOW(rc_resize_allowed); + SHOW(rc_resize_up_thresh); + SHOW(rc_resize_down_thresh); + SHOW(rc_end_usage); + SHOW(rc_target_bitrate); + SHOW(rc_min_quantizer); + SHOW(rc_max_quantizer); + SHOW(rc_undershoot_pct); + SHOW(rc_overshoot_pct); + SHOW(rc_buf_sz); + SHOW(rc_buf_initial_sz); + SHOW(rc_buf_optimal_sz); + SHOW(rc_2pass_vbr_bias_pct); + SHOW(rc_2pass_vbr_minsection_pct); + SHOW(rc_2pass_vbr_maxsection_pct); + SHOW(kf_mode); + SHOW(kf_min_dist); + SHOW(kf_max_dist); + } + + if(pass == (one_pass_only ? one_pass_only - 1 : 0)) { + if (file_type == FILE_TYPE_Y4M) + /*The Y4M reader does its own allocation. + Just initialize this here to avoid problems if we never read any + frames.*/ + memset(&raw, 0, sizeof(raw)); + else + vpx_img_alloc(&raw, arg_use_i420 ? VPX_IMG_FMT_I420 : VPX_IMG_FMT_YV12, + cfg.g_w, cfg.g_h, 1); + } + + outfile = strcmp(out_fn, "-") ? fopen(out_fn, "wb") + : set_binary_mode(stdout); + + if (!outfile) + { + fprintf(stderr, "Failed to open output file\n"); + return EXIT_FAILURE; + } + + if(write_webm && fseek(outfile, 0, SEEK_CUR)) + { + fprintf(stderr, "WebM output to pipes not supported.\n"); + return EXIT_FAILURE; + } + + if (stats_fn) + { + if (!stats_open_file(&stats, stats_fn, pass)) + { + fprintf(stderr, "Failed to open statistics store\n"); + return EXIT_FAILURE; + } + } + else + { + if (!stats_open_mem(&stats, pass)) + { + fprintf(stderr, "Failed to open statistics store\n"); + return EXIT_FAILURE; + } + } + + cfg.g_pass = arg_passes == 2 + ? pass ? VPX_RC_LAST_PASS : VPX_RC_FIRST_PASS + : VPX_RC_ONE_PASS; + #if VPX_ENCODER_ABI_VERSION > (1 + VPX_CODEC_ABI_VERSION) + + if (pass) + { + cfg.rc_twopass_stats_in = stats_get(&stats); + } + + #endif + + if(write_webm) + { + ebml.stream = outfile; + write_webm_file_header(&ebml, &cfg, &arg_framerate); + } + else + write_ivf_file_header(outfile, &cfg, codec->fourcc, 0); + + + /* Construct Encoder Context */ + vpx_codec_enc_init(&encoder, codec->iface, &cfg, + show_psnr ? VPX_CODEC_USE_PSNR : 0); + ctx_exit_on_error(&encoder, "Failed to initialize encoder"); + + /* Note that we bypass the vpx_codec_control wrapper macro because + * we're being clever to store the control IDs in an array. Real + * applications will want to make use of the enumerations directly + */ + for (i = 0; i < arg_ctrl_cnt; i++) + { + if (vpx_codec_control_(&encoder, arg_ctrls[i][0], arg_ctrls[i][1])) + fprintf(stderr, "Error: Tried to set control %d = %d\n", + arg_ctrls[i][0], arg_ctrls[i][1]); + + ctx_exit_on_error(&encoder, "Failed to control codec"); + } + + frame_avail = 1; + got_data = 0; + + while (frame_avail || got_data) + { + vpx_codec_iter_t iter = NULL; + const vpx_codec_cx_pkt_t *pkt; + struct vpx_usec_timer timer; + int64_t frame_start, next_frame_start; + + if (!arg_limit || frames_in < arg_limit) + { + frame_avail = read_frame(infile, &raw, file_type, &y4m, + &detect); + + if (frame_avail) + frames_in++; + + fprintf(stderr, + "\rPass %d/%d frame %4d/%-4d %7ldB \033[K", pass + 1, + arg_passes, frames_in, frames_out, nbytes); + } + else + frame_avail = 0; + + vpx_usec_timer_start(&timer); + + frame_start = (cfg.g_timebase.den * (int64_t)(frames_in - 1) + * arg_framerate.den) / cfg.g_timebase.num / arg_framerate.num; + next_frame_start = (cfg.g_timebase.den * (int64_t)(frames_in) + * arg_framerate.den) + / cfg.g_timebase.num / arg_framerate.num; + vpx_codec_encode(&encoder, frame_avail ? &raw : NULL, frame_start, + next_frame_start - frame_start, + 0, arg_deadline); + vpx_usec_timer_mark(&timer); + cx_time += vpx_usec_timer_elapsed(&timer); + ctx_exit_on_error(&encoder, "Failed to encode frame"); + got_data = 0; + + while ((pkt = vpx_codec_get_cx_data(&encoder, &iter))) + { + got_data = 1; + + switch (pkt->kind) + { + case VPX_CODEC_CX_FRAME_PKT: + frames_out++; + fprintf(stderr, " %6luF", + (unsigned long)pkt->data.frame.sz); + + if(write_webm) + { + /* Update the hash */ + if(!ebml.debug) + hash = murmur(pkt->data.frame.buf, + pkt->data.frame.sz, hash); + + write_webm_block(&ebml, &cfg, pkt); + } + else + { + write_ivf_frame_header(outfile, pkt); + if(fwrite(pkt->data.frame.buf, 1, + pkt->data.frame.sz, outfile)); + } + nbytes += pkt->data.raw.sz; + break; + case VPX_CODEC_STATS_PKT: + frames_out++; + fprintf(stderr, " %6luS", + (unsigned long)pkt->data.twopass_stats.sz); + stats_write(&stats, + pkt->data.twopass_stats.buf, + pkt->data.twopass_stats.sz); + nbytes += pkt->data.raw.sz; + break; + case VPX_CODEC_PSNR_PKT: + + if (show_psnr) + { + int i; + + psnr_sse_total += pkt->data.psnr.sse[0]; + psnr_samples_total += pkt->data.psnr.samples[0]; + for (i = 0; i < 4; i++) + { + fprintf(stderr, "%.3lf ", pkt->data.psnr.psnr[i]); + psnr_totals[i] += pkt->data.psnr.psnr[i]; + } + psnr_count++; + } + + break; + default: + break; + } + } + + fflush(stdout); + } + + fprintf(stderr, + "\rPass %d/%d frame %4d/%-4d %7ldB %7ldb/f %7"PRId64"b/s" + " %7lu %s (%.2f fps)\033[K", pass + 1, + arg_passes, frames_in, frames_out, nbytes, nbytes * 8 / frames_in, + nbytes * 8 *(int64_t)arg_framerate.num / arg_framerate.den / frames_in, + cx_time > 9999999 ? cx_time / 1000 : cx_time, + cx_time > 9999999 ? "ms" : "us", + (float)frames_in * 1000000.0 / (float)cx_time); + + if ( (show_psnr) && (psnr_count>0) ) + { + int i; + double ovpsnr = vp8_mse2psnr(psnr_samples_total, 255.0, + psnr_sse_total); + + fprintf(stderr, "\nPSNR (Overall/Avg/Y/U/V)"); + + fprintf(stderr, " %.3lf", ovpsnr); + for (i = 0; i < 4; i++) + { + fprintf(stderr, " %.3lf", psnr_totals[i]/psnr_count); + } + } + + vpx_codec_destroy(&encoder); + + fclose(infile); + + if(write_webm) + { + write_webm_file_footer(&ebml, hash); + } + else + { + if (!fseek(outfile, 0, SEEK_SET)) + write_ivf_file_header(outfile, &cfg, codec->fourcc, frames_out); + } + + fclose(outfile); + stats_close(&stats); + fprintf(stderr, "\n"); + + if (one_pass_only) + break; + } + + vpx_img_free(&raw); + free(argv); + return EXIT_SUCCESS; + }