diff options
Diffstat (limited to 'vp8/encoder')
| -rw-r--r-- | vp8/encoder/bitstream.c | 172 | ||||
| -rw-r--r-- | vp8/encoder/block.h | 22 | ||||
| -rw-r--r-- | vp8/encoder/dct.c | 7 | ||||
| -rw-r--r-- | vp8/encoder/dct.h | 2 | ||||
| -rw-r--r-- | vp8/encoder/encodeframe.c | 63 | ||||
| -rw-r--r-- | vp8/encoder/encodeintra.c | 22 | ||||
| -rw-r--r-- | vp8/encoder/encodemb.c | 44 | ||||
| -rw-r--r-- | vp8/encoder/encodemv.c | 592 | ||||
| -rw-r--r-- | vp8/encoder/encodemv.h | 15 | ||||
| -rw-r--r-- | vp8/encoder/firstpass.c | 10 | ||||
| -rw-r--r-- | vp8/encoder/generic/csystemdependent.c | 79 | ||||
| -rw-r--r-- | vp8/encoder/mbgraph.c | 26 | ||||
| -rw-r--r-- | vp8/encoder/mcomp.c | 137 | ||||
| -rw-r--r-- | vp8/encoder/mcomp.h | 9 | ||||
| -rw-r--r-- | vp8/encoder/modecosts.c | 3 | ||||
| -rw-r--r-- | vp8/encoder/onyx_if.c | 278 | ||||
| -rw-r--r-- | vp8/encoder/onyx_int.h | 41 | ||||
| -rw-r--r-- | vp8/encoder/picklpf.c | 28 | ||||
| -rw-r--r-- | vp8/encoder/quantize.c | 2 | ||||
| -rw-r--r-- | vp8/encoder/quantize.h | 2 | ||||
| -rw-r--r-- | vp8/encoder/ratectrl.c | 40 | ||||
| -rw-r--r-- | vp8/encoder/rdopt.c | 614 | ||||
| -rw-r--r-- | vp8/encoder/sad_c.c | 44 | ||||
| -rw-r--r-- | vp8/encoder/ssim.c | 109 | ||||
| -rw-r--r-- | vp8/encoder/tokenize.c | 97 | ||||
| -rw-r--r-- | vp8/encoder/variance.h | 568 | ||||
| -rw-r--r-- | vp8/encoder/x86/variance_x86.h | 328 | ||||
| -rw-r--r-- | vp8/encoder/x86/x86_csystemdependent.c | 85 |
28 files changed, 634 insertions, 2805 deletions
diff --git a/vp8/encoder/bitstream.c b/vp8/encoder/bitstream.c index 70cdb6aaf..36776ab21 100644 --- a/vp8/encoder/bitstream.c +++ b/vp8/encoder/bitstream.c @@ -45,32 +45,26 @@ unsigned int tree_update_hist [BLOCK_TYPES] [COEF_BANDS] [PREV_COEF_CONTEXTS] [ENTROPY_NODES][2]; -#if CONFIG_HYBRIDTRANSFORM unsigned int hybrid_tree_update_hist [BLOCK_TYPES] [COEF_BANDS] [PREV_COEF_CONTEXTS] [ENTROPY_NODES][2]; -#endif unsigned int tree_update_hist_8x8 [BLOCK_TYPES_8X8] [COEF_BANDS] [PREV_COEF_CONTEXTS] [ENTROPY_NODES] [2]; -#if CONFIG_HYBRIDTRANSFORM8X8 unsigned int hybrid_tree_update_hist_8x8 [BLOCK_TYPES_8X8] [COEF_BANDS] [PREV_COEF_CONTEXTS] [ENTROPY_NODES] [2]; -#endif unsigned int tree_update_hist_16x16 [BLOCK_TYPES_16X16] [COEF_BANDS] [PREV_COEF_CONTEXTS] [ENTROPY_NODES] [2]; -#if CONFIG_HYBRIDTRANSFORM16X16 unsigned int hybrid_tree_update_hist_16x16 [BLOCK_TYPES_16X16] [COEF_BANDS] [PREV_COEF_CONTEXTS] [ENTROPY_NODES] [2]; -#endif extern unsigned int active_section; #endif @@ -265,57 +259,23 @@ void update_skip_probs(VP8_COMP *cpi) { } } -#if CONFIG_SWITCHABLE_INTERP void update_switchable_interp_probs(VP8_COMP *cpi, vp8_writer* const bc) { VP8_COMMON *const pc = &cpi->common; unsigned int branch_ct[32][2]; int i, j; for (j = 0; j <= VP8_SWITCHABLE_FILTERS; ++j) { - //for (j = 0; j <= 0; ++j) { -/* - if (!cpi->dummy_packing) -#if VP8_SWITCHABLE_FILTERS == 3 - printf("HELLO %d %d %d\n", cpi->switchable_interp_count[j][0], - cpi->switchable_interp_count[j][1], cpi->switchable_interp_count[j][2]); -#else - printf("HELLO %d %d\n", cpi->switchable_interp_count[j][0], - cpi->switchable_interp_count[j][1]); -#endif -*/ vp8_tree_probs_from_distribution( VP8_SWITCHABLE_FILTERS, vp8_switchable_interp_encodings, vp8_switchable_interp_tree, - pc->fc.switchable_interp_prob[j], branch_ct, cpi->switchable_interp_count[j], - 256, 1 - ); + pc->fc.switchable_interp_prob[j], branch_ct, + cpi->switchable_interp_count[j], 256, 1); for (i = 0; i < VP8_SWITCHABLE_FILTERS - 1; ++i) { if (pc->fc.switchable_interp_prob[j][i] < 1) pc->fc.switchable_interp_prob[j][i] = 1; vp8_write_literal(bc, pc->fc.switchable_interp_prob[j][i], 8); -/* - if (!cpi->dummy_packing) -#if VP8_SWITCHABLE_FILTERS == 3 - printf("Probs %d %d [%d]\n", - pc->fc.switchable_interp_prob[j][0], - pc->fc.switchable_interp_prob[j][1], pc->frame_type); -#else - printf("Probs %d [%d]\n", pc->fc.switchable_interp_prob[j][0], - pc->frame_type); -#endif -*/ } } - /* - if (!cpi->dummy_packing) -#if VP8_SWITCHABLE_FILTERS == 3 - printf("Probs %d %d [%d]\n", - pc->fc.switchable_interp_prob[0], pc->fc.switchable_interp_prob[1], pc->frame_type); -#else - printf("Probs %d [%d]\n", pc->fc.switchable_interp_prob[0], pc->frame_type); -#endif - */ } -#endif // This function updates the reference frame prediction stats static void update_refpred_stats(VP8_COMP *cpi) { @@ -649,7 +609,6 @@ static void write_sub_mv_ref vp8_sub_mv_ref_encoding_array - LEFT4X4 + m); } -#if CONFIG_NEWMVENTROPY static void write_nmv(vp8_writer *bc, const MV *mv, const int_mv *ref, const nmv_context *nmvc, int usehp) { MV e; @@ -660,31 +619,6 @@ static void write_nmv(vp8_writer *bc, const MV *mv, const int_mv *ref, vp8_encode_nmv_fp(bc, &e, &ref->as_mv, nmvc, usehp); } -#else - -static void write_mv -( - vp8_writer *bc, const MV *mv, const int_mv *ref, const MV_CONTEXT *mvc -) { - MV e; - e.row = mv->row - ref->as_mv.row; - e.col = mv->col - ref->as_mv.col; - - vp8_encode_motion_vector(bc, &e, mvc); -} - -static void write_mv_hp -( - vp8_writer *bc, const MV *mv, const int_mv *ref, const MV_CONTEXT_HP *mvc -) { - MV e; - e.row = mv->row - ref->as_mv.row; - e.col = mv->col - ref->as_mv.col; - - vp8_encode_motion_vector_hp(bc, &e, mvc); -} -#endif /* CONFIG_NEWMVENTROPY */ - // This function writes the current macro block's segnment id to the bitstream // It should only be called if a segment map update is indicated. static void write_mb_segid(vp8_writer *bc, @@ -821,12 +755,7 @@ static void update_ref_probs(VP8_COMP *const cpi) { static void pack_inter_mode_mvs(VP8_COMP *const cpi, vp8_writer *const bc) { int i; VP8_COMMON *const pc = &cpi->common; -#if CONFIG_NEWMVENTROPY const nmv_context *nmvc = &pc->fc.nmvc; -#else - const MV_CONTEXT *mvc = pc->fc.mvc; - const MV_CONTEXT_HP *mvc_hp = pc->fc.mvc_hp; -#endif MACROBLOCK *x = &cpi->mb; MACROBLOCKD *xd = &cpi->mb.e_mbd; MODE_INFO *m; @@ -1042,21 +971,19 @@ static void pack_inter_mode_mvs(VP8_COMP *const cpi, vp8_writer *const bc) { cpi->common.pred_filter_mode); } #endif -#if CONFIG_SWITCHABLE_INTERP if (mode >= NEARESTMV && mode <= SPLITMV) { if (cpi->common.mcomp_filter_type == SWITCHABLE) { vp8_write_token(bc, vp8_switchable_interp_tree, - get_pred_probs(&cpi->common, xd, PRED_SWITCHABLE_INTERP), + get_pred_probs(&cpi->common, xd, + PRED_SWITCHABLE_INTERP), vp8_switchable_interp_encodings + vp8_switchable_interp_map[mi->interp_filter]); - //if (!cpi->dummy_packing) printf("Reading: %d\n", mi->interp_filter); } else { assert (mi->interp_filter == cpi->common.mcomp_filter_type); } } -#endif if (mi->second_ref_frame && (mode == NEWMV || mode == SPLITMV)) { int_mv n1, n2; @@ -1099,17 +1026,9 @@ static void pack_inter_mode_mvs(VP8_COMP *const cpi, vp8_writer *const bc) { } #endif -#if CONFIG_NEWMVENTROPY write_nmv(bc, &mi->mv[0].as_mv, &best_mv, (const nmv_context*) nmvc, xd->allow_high_precision_mv); -#else - if (xd->allow_high_precision_mv) { - write_mv_hp(bc, &mi->mv[0].as_mv, &best_mv, mvc_hp); - } else { - write_mv(bc, &mi->mv[0].as_mv, &best_mv, mvc); - } -#endif if (mi->second_ref_frame) { #if 0 //CONFIG_NEW_MVREF @@ -1126,17 +1045,9 @@ static void pack_inter_mode_mvs(VP8_COMP *const cpi, vp8_writer *const bc) { &best_second_mv); cpi->best_ref_index_counts[best_index]++; #endif -#if CONFIG_NEWMVENTROPY write_nmv(bc, &mi->mv[1].as_mv, &best_second_mv, (const nmv_context*) nmvc, xd->allow_high_precision_mv); -#else - if (xd->allow_high_precision_mv) { - write_mv_hp(bc, &mi->mv[1].as_mv, &best_second_mv, mvc_hp); - } else { - write_mv(bc, &mi->mv[1].as_mv, &best_second_mv, mvc); - } -#endif } break; case SPLITMV: { @@ -1178,40 +1089,16 @@ static void pack_inter_mode_mvs(VP8_COMP *const cpi, vp8_writer *const bc) { #ifdef ENTROPY_STATS active_section = 11; #endif -#if CONFIG_NEWMVENTROPY write_nmv(bc, &blockmv.as_mv, &best_mv, (const nmv_context*) nmvc, xd->allow_high_precision_mv); -#else - if (xd->allow_high_precision_mv) { - write_mv_hp(bc, &blockmv.as_mv, &best_mv, - (const MV_CONTEXT_HP *) mvc_hp); - } else { - write_mv(bc, &blockmv.as_mv, &best_mv, - (const MV_CONTEXT *) mvc); - } -#endif if (mi->second_ref_frame) { -#if CONFIG_NEWMVENTROPY write_nmv(bc, &cpi->mb.partition_info->bmi[j].second_mv.as_mv, &best_second_mv, (const nmv_context*) nmvc, xd->allow_high_precision_mv); -#else - if (xd->allow_high_precision_mv) { - write_mv_hp( - bc, - &cpi->mb.partition_info->bmi[j].second_mv.as_mv, - &best_second_mv, (const MV_CONTEXT_HP *)mvc_hp); - } else { - write_mv( - bc, - &cpi->mb.partition_info->bmi[j].second_mv.as_mv, - &best_second_mv, (const MV_CONTEXT *) mvc); - } -#endif } } } while (++j < cpi->mb.partition_info->count); @@ -1223,9 +1110,9 @@ static void pack_inter_mode_mvs(VP8_COMP *const cpi, vp8_writer *const bc) { } } -#if CONFIG_TX_SELECT if (((rf == INTRA_FRAME && mode <= I8X8_PRED) || - (rf != INTRA_FRAME && mode != SPLITMV)) && + (rf != INTRA_FRAME && !(mode == SPLITMV && + mi->partitioning == PARTITIONING_4X4))) && pc->txfm_mode == TX_MODE_SELECT && !((pc->mb_no_coeff_skip && mi->mb_skip_coeff) || (segfeature_active(xd, segment_id, SEG_LVL_EOB) && @@ -1233,10 +1120,9 @@ static void pack_inter_mode_mvs(VP8_COMP *const cpi, vp8_writer *const bc) { TX_SIZE sz = mi->txfm_size; // FIXME(rbultje) code ternary symbol once all experiments are merged vp8_write(bc, sz != TX_4X4, pc->prob_tx[0]); - if (sz != TX_4X4 && mode != I8X8_PRED) + if (sz != TX_4X4 && mode != I8X8_PRED && mode != SPLITMV) vp8_write(bc, sz != TX_8X8, pc->prob_tx[1]); } -#endif #ifdef ENTROPY_STATS active_section = 1; @@ -1365,7 +1251,6 @@ static void write_mb_modes_kf(const VP8_COMMON *c, } else write_uv_mode(bc, m->mbmi.uv_mode, c->kf_uv_mode_prob[ym]); -#if CONFIG_TX_SELECT if (ym <= I8X8_PRED && c->txfm_mode == TX_MODE_SELECT && !((c->mb_no_coeff_skip && m->mbmi.mb_skip_coeff) || (segfeature_active(xd, segment_id, SEG_LVL_EOB) && @@ -1376,7 +1261,6 @@ static void write_mb_modes_kf(const VP8_COMMON *c, if (sz != TX_4X4 && ym <= TM_PRED) vp8_write(bc, sz != TX_8X8, c->prob_tx[1]); } -#endif } static void write_kfmodes(VP8_COMP* const cpi, vp8_writer* const bc) { @@ -1498,7 +1382,6 @@ void build_coeff_contexts(VP8_COMP *cpi) { } } } -#if CONFIG_HYBRIDTRANSFORM for (i = 0; i < BLOCK_TYPES; ++i) { for (j = 0; j < COEF_BANDS; ++j) { for (k = 0; k < PREV_COEF_CONTEXTS; ++k) { @@ -1519,8 +1402,6 @@ void build_coeff_contexts(VP8_COMP *cpi) { } } } -#endif - if (cpi->common.txfm_mode != ONLY_4X4) { for (i = 0; i < BLOCK_TYPES_8X8; ++i) { @@ -1547,7 +1428,6 @@ void build_coeff_contexts(VP8_COMP *cpi) { } } } -#if CONFIG_HYBRIDTRANSFORM8X8 for (i = 0; i < BLOCK_TYPES_8X8; ++i) { for (j = 0; j < COEF_BANDS; ++j) { for (k = 0; k < PREV_COEF_CONTEXTS; ++k) { @@ -1572,7 +1452,6 @@ void build_coeff_contexts(VP8_COMP *cpi) { } } } -#endif } if (cpi->common.txfm_mode > ALLOW_8X8) { @@ -1595,7 +1474,6 @@ void build_coeff_contexts(VP8_COMP *cpi) { } } } -#if CONFIG_HYBRIDTRANSFORM16X16 for (i = 0; i < BLOCK_TYPES_16X16; ++i) { for (j = 0; j < COEF_BANDS; ++j) { for (k = 0; k < PREV_COEF_CONTEXTS; ++k) { @@ -1614,7 +1492,6 @@ void build_coeff_contexts(VP8_COMP *cpi) { } } } -#endif } #if 0 @@ -1887,7 +1764,6 @@ static void update_coef_probs(VP8_COMP* const cpi, vp8_writer* const bc) { } } -#if CONFIG_HYBRIDTRANSFORM savings = 0; update[0] = update[1] = 0; for (i = 0; i < BLOCK_TYPES; ++i) { @@ -1976,7 +1852,6 @@ static void update_coef_probs(VP8_COMP* const cpi, vp8_writer* const bc) { } } } -#endif /* do not do this if not even allowed */ if (cpi->common.txfm_mode != ONLY_4X4) { @@ -2054,7 +1929,6 @@ static void update_coef_probs(VP8_COMP* const cpi, vp8_writer* const bc) { } } } -#if CONFIG_HYBRIDTRANSFORM8X8 update[0] = update[1] = 0; savings = 0; for (i = 0; i < BLOCK_TYPES_8X8; ++i) { @@ -2128,7 +2002,6 @@ static void update_coef_probs(VP8_COMP* const cpi, vp8_writer* const bc) { } } } -#endif } if (cpi->common.txfm_mode > ALLOW_8X8) { @@ -2206,7 +2079,6 @@ static void update_coef_probs(VP8_COMP* const cpi, vp8_writer* const bc) { } } } -#if CONFIG_HYBRIDTRANSFORM16X16 update[0] = update[1] = 0; savings = 0; for (i = 0; i < BLOCK_TYPES_16X16; ++i) { @@ -2280,7 +2152,6 @@ static void update_coef_probs(VP8_COMP* const cpi, vp8_writer* const bc) { } } } -#endif } } @@ -2561,12 +2432,11 @@ void vp8_pack_bitstream(VP8_COMP *cpi, unsigned char *dest, unsigned long *size) /* sb mode probability */ const int sb_max = (((pc->mb_rows + 1) >> 1) * ((pc->mb_cols + 1) >> 1)); - pc->sb_coded = get_prob(cpi->sb_count, sb_max); + pc->sb_coded = get_prob(sb_max - cpi->sb_count, sb_max); vp8_write_literal(&header_bc, pc->sb_coded, 8); } #endif -#if CONFIG_TX_SELECT { if (pc->txfm_mode == TX_MODE_SELECT) { pc->prob_tx[0] = get_prob(cpi->txfm_count[0] + cpi->txfm_count_8x8p[0], @@ -2583,9 +2453,6 @@ void vp8_pack_bitstream(VP8_COMP *cpi, unsigned char *dest, unsigned long *size) vp8_write_literal(&header_bc, pc->prob_tx[1], 8); } } -#else - vp8_write_bit(&header_bc, !!pc->txfm_mode); -#endif // Encode the loop filter level and type vp8_write_bit(&header_bc, pc->filter_type); @@ -2687,7 +2554,6 @@ void vp8_pack_bitstream(VP8_COMP *cpi, unsigned char *dest, unsigned long *size) // Signal whether to allow high MV precision vp8_write_bit(&header_bc, (xd->allow_high_precision_mv) ? 1 : 0); -#if CONFIG_SWITCHABLE_INTERP if (pc->mcomp_filter_type == SWITCHABLE) { /* Check to see if only one of the filters is actually used */ int count[VP8_SWITCHABLE_FILTERS]; @@ -2712,7 +2578,6 @@ void vp8_pack_bitstream(VP8_COMP *cpi, unsigned char *dest, unsigned long *size) // Signal the type of subpel filter to use vp8_write_bit(&header_bc, (pc->mcomp_filter_type == SWITCHABLE)); if (pc->mcomp_filter_type != SWITCHABLE) -#endif /* CONFIG_SWITCHABLE_INTERP */ vp8_write_literal(&header_bc, (pc->mcomp_filter_type), 2); } @@ -2731,29 +2596,18 @@ void vp8_pack_bitstream(VP8_COMP *cpi, unsigned char *dest, unsigned long *size) vp8_clear_system_state(); // __asm emms; vp8_copy(cpi->common.fc.pre_coef_probs, cpi->common.fc.coef_probs); -#if CONFIG_HYBRIDTRANSFORM vp8_copy(cpi->common.fc.pre_hybrid_coef_probs, cpi->common.fc.hybrid_coef_probs); -#endif vp8_copy(cpi->common.fc.pre_coef_probs_8x8, cpi->common.fc.coef_probs_8x8); -#if CONFIG_HYBRIDTRANSFORM8X8 vp8_copy(cpi->common.fc.pre_hybrid_coef_probs_8x8, cpi->common.fc.hybrid_coef_probs_8x8); -#endif vp8_copy(cpi->common.fc.pre_coef_probs_16x16, cpi->common.fc.coef_probs_16x16); -#if CONFIG_HYBRIDTRANSFORM16X16 vp8_copy(cpi->common.fc.pre_hybrid_coef_probs_16x16, cpi->common.fc.hybrid_coef_probs_16x16); -#endif vp8_copy(cpi->common.fc.pre_ymode_prob, cpi->common.fc.ymode_prob); vp8_copy(cpi->common.fc.pre_uv_mode_prob, cpi->common.fc.uv_mode_prob); vp8_copy(cpi->common.fc.pre_bmode_prob, cpi->common.fc.bmode_prob); vp8_copy(cpi->common.fc.pre_sub_mv_ref_prob, cpi->common.fc.sub_mv_ref_prob); vp8_copy(cpi->common.fc.pre_mbsplit_prob, cpi->common.fc.mbsplit_prob); vp8_copy(cpi->common.fc.pre_i8x8_mode_prob, cpi->common.fc.i8x8_mode_prob); -#if CONFIG_NEWMVENTROPY cpi->common.fc.pre_nmvc = cpi->common.fc.nmvc; -#else - vp8_copy(cpi->common.fc.pre_mvc, cpi->common.fc.mvc); - vp8_copy(cpi->common.fc.pre_mvc_hp, cpi->common.fc.mvc_hp); -#endif vp8_zero(cpi->sub_mv_ref_count); vp8_zero(cpi->mbsplit_count); vp8_zero(cpi->common.fc.mv_ref_ct) @@ -2796,10 +2650,8 @@ void vp8_pack_bitstream(VP8_COMP *cpi, unsigned char *dest, unsigned long *size) vp8_write_literal(&header_bc, pc->prob_pred_filter_off, 8); #endif -#if CONFIG_SWITCHABLE_INTERP if (pc->mcomp_filter_type == SWITCHABLE) update_switchable_interp_probs(cpi, &header_bc); -#endif vp8_write_literal(&header_bc, pc->prob_intra_coded, 8); vp8_write_literal(&header_bc, pc->prob_last_coded, 8); @@ -2825,15 +2677,7 @@ void vp8_pack_bitstream(VP8_COMP *cpi, unsigned char *dest, unsigned long *size) update_mbintra_mode_probs(cpi, &header_bc); -#if CONFIG_NEWMVENTROPY vp8_write_nmvprobs(cpi, xd->allow_high_precision_mv, &header_bc); -#else - if (xd->allow_high_precision_mv) { - vp8_write_mvprobs_hp(cpi, &header_bc); - } else { - vp8_write_mvprobs(cpi, &header_bc); - } -#endif } vp8_stop_encode(&header_bc); diff --git a/vp8/encoder/block.h b/vp8/encoder/block.h index 80f9b75b8..48623be8c 100644 --- a/vp8/encoder/block.h +++ b/vp8/encoder/block.h @@ -79,9 +79,7 @@ typedef struct { int hybrid_pred_diff; int comp_pred_diff; int single_pred_diff; -#if CONFIG_TX_SELECT int64_t txfm_rd_diff[NB_TXFM_MODES]; -#endif } PICK_MODE_CONTEXT; typedef struct { @@ -114,7 +112,6 @@ typedef struct { int *mb_norm_activity_ptr; signed int act_zbin_adj; -#if CONFIG_NEWMVENTROPY int nmvjointcost[MV_JOINTS]; int nmvcosts[2][MV_VALS]; int *nmvcost[2]; @@ -126,28 +123,17 @@ typedef struct { int *nmvsadcost[2]; int nmvsadcosts_hp[2][MV_VALS]; int *nmvsadcost_hp[2]; -#else - int mvcosts[2][MVvals + 1]; - int *mvcost[2]; - int mvsadcosts[2][MVfpvals + 1]; - int *mvsadcost[2]; - int mvcosts_hp[2][MVvals_hp + 1]; - int *mvcost_hp[2]; - int mvsadcosts_hp[2][MVfpvals_hp + 1]; - int *mvsadcost_hp[2]; -#endif /* CONFIG_NEWMVENTROPY */ int mbmode_cost[2][MB_MODE_COUNT]; int intra_uv_mode_cost[2][MB_MODE_COUNT]; int bmode_costs[VP8_BINTRAMODES][VP8_BINTRAMODES][VP8_BINTRAMODES]; int i8x8_mode_costs[MB_MODE_COUNT]; int inter_bmode_costs[B_MODE_COUNT]; -#if CONFIG_SWITCHABLE_INTERP - int switchable_interp_costs[VP8_SWITCHABLE_FILTERS+1] + int switchable_interp_costs[VP8_SWITCHABLE_FILTERS + 1] [VP8_SWITCHABLE_FILTERS]; -#endif - // These define limits to motion vector components to prevent them from extending outside the UMV borders + // These define limits to motion vector components to prevent them + // from extending outside the UMV borders int mv_col_min; int mv_col_max; int mv_row_min; @@ -164,10 +150,8 @@ typedef struct { unsigned int token_costs[TX_SIZE_MAX][BLOCK_TYPES][COEF_BANDS] [PREV_COEF_CONTEXTS][MAX_ENTROPY_TOKENS]; -#if CONFIG_HYBRIDTRANSFORM || CONFIG_HYBRIDTRANSFORM8X8 || CONFIG_HYBRIDTRANSFORM16X16 unsigned int hybrid_token_costs[TX_SIZE_MAX][BLOCK_TYPES][COEF_BANDS] [PREV_COEF_CONTEXTS][MAX_ENTROPY_TOKENS]; -#endif int optimize; diff --git a/vp8/encoder/dct.c b/vp8/encoder/dct.c index cd13fec7c..0983b1c0a 100644 --- a/vp8/encoder/dct.c +++ b/vp8/encoder/dct.c @@ -14,8 +14,6 @@ #include "vp8/common/idct.h" #include "vp8/common/systemdependent.h" -#if CONFIG_HYBRIDTRANSFORM || CONFIG_HYBRIDTRANSFORM8X8 || CONFIG_HYBRIDTRANSFORM16X16 - #include "vp8/common/blockd.h" // TODO: these transforms can be converted into integer forms to reduce @@ -71,9 +69,7 @@ float adst_8[64] = { 0.175227946595735, -0.326790388032145, 0.434217976756762, -0.483002021635509, 0.466553967085785, -0.387095214016348, 0.255357107325376, -0.089131608307532 }; -#endif -#if CONFIG_HYBRIDTRANSFORM16X16 || CONFIG_HYBRIDTRANSFORM || CONFIG_HYBRIDTRANSFORM8X8 float dct_16[256] = { 0.250000, 0.250000, 0.250000, 0.250000, 0.250000, 0.250000, 0.250000, 0.250000, 0.250000, 0.250000, 0.250000, 0.250000, 0.250000, 0.250000, 0.250000, 0.250000, @@ -143,7 +139,6 @@ float adst_16[256] = { 0.065889, -0.129396, 0.188227, -0.240255, 0.283599, -0.316693, 0.338341, -0.347761, 0.344612, -0.329007, 0.301511, -0.263118, 0.215215, -0.159534, 0.098087, -0.033094 }; -#endif static const int xC1S7 = 16069; static const int xC2S6 = 15137; @@ -400,7 +395,6 @@ void vp8_short_fhaar2x2_c(short *input, short *output, int pitch) { // pitch = 8 } -#if CONFIG_HYBRIDTRANSFORM8X8 || CONFIG_HYBRIDTRANSFORM || CONFIG_HYBRIDTRANSFORM16X16 void vp8_fht_c(short *input, short *output, int pitch, TX_TYPE tx_type, int tx_dim) { @@ -518,7 +512,6 @@ void vp8_fht_c(short *input, short *output, int pitch, } vp8_clear_system_state(); // Make it simd safe : __asm emms; } -#endif void vp8_short_fdct4x4_c(short *input, short *output, int pitch) { int i; diff --git a/vp8/encoder/dct.h b/vp8/encoder/dct.h index 180192bbb..4ad1fe85d 100644 --- a/vp8/encoder/dct.h +++ b/vp8/encoder/dct.h @@ -26,10 +26,8 @@ #endif -#if CONFIG_HYBRIDTRANSFORM8X8 || CONFIG_HYBRIDTRANSFORM || CONFIG_HYBRIDTRANSFORM16X16 void vp8_fht_c(short *input, short *output, int pitch, TX_TYPE tx_type, int tx_dim); -#endif #ifndef vp8_fdct_short16x16 #define vp8_fdct_short16x16 vp8_short_fdct16x16_c diff --git a/vp8/encoder/encodeframe.c b/vp8/encoder/encodeframe.c index 619695c33..0910cfd35 100644 --- a/vp8/encoder/encodeframe.c +++ b/vp8/encoder/encodeframe.c @@ -120,8 +120,8 @@ static unsigned int tt_activity_measure(VP8_COMP *cpi, MACROBLOCK *x) { * lambda using a non-linear combination (e.g., the smallest, or second * smallest, etc.). */ - act = VARIANCE_INVOKE(&cpi->rtcd.variance, var16x16)(x->src.y_buffer, - x->src.y_stride, VP8_VAR_OFFS, 0, &sse); + act = vp8_variance16x16(x->src.y_buffer, x->src.y_stride, VP8_VAR_OFFS, 0, + &sse); act = act << 4; /* If the region is flat, lower the activity some more. */ @@ -411,7 +411,6 @@ static void update_state(VP8_COMP *cpi, MACROBLOCK *x, PICK_MODE_CONTEXT *ctx) { mbmi->mv[1].as_int = x->partition_info->bmi[15].second_mv.as_int; } -#if CONFIG_TX_SELECT { int segment_id = mbmi->segment_id; if (!segfeature_active(xd, segment_id, SEG_LVL_EOB) || @@ -421,7 +420,6 @@ static void update_state(VP8_COMP *cpi, MACROBLOCK *x, PICK_MODE_CONTEXT *ctx) { } } } -#endif if (cpi->common.frame_type == KEY_FRAME) { // Restore the coding modes to that held in the coding context @@ -1342,37 +1340,18 @@ static void encode_frame_internal(VP8_COMP *cpi) { cpi->pred_filter_on_count = 0; cpi->pred_filter_off_count = 0; #endif -#if CONFIG_SWITCHABLE_INTERP vp8_zero(cpi->switchable_interp_count); -#endif - -#if 0 - // Experimental code - cpi->frame_distortion = 0; - cpi->last_mb_distortion = 0; -#endif xd->mode_info_context = cm->mi; xd->prev_mode_info_context = cm->prev_mi; -#if CONFIG_NEWMVENTROPY vp8_zero(cpi->NMVcount); -#else - vp8_zero(cpi->MVcount); - vp8_zero(cpi->MVcount_hp); -#endif vp8_zero(cpi->coef_counts); -#if CONFIG_HYBRIDTRANSFORM vp8_zero(cpi->hybrid_coef_counts); -#endif vp8_zero(cpi->coef_counts_8x8); -#if CONFIG_HYBRIDTRANSFORM8X8 vp8_zero(cpi->hybrid_coef_counts_8x8); -#endif vp8_zero(cpi->coef_counts_16x16); -#if CONFIG_HYBRIDTRANSFORM16X16 vp8_zero(cpi->hybrid_coef_counts_16x16); -#endif vp8cx_frame_init_quantizer(cpi); @@ -1393,11 +1372,9 @@ static void encode_frame_internal(VP8_COMP *cpi) { vpx_memset(cpi->rd_comp_pred_diff, 0, sizeof(cpi->rd_comp_pred_diff)); vpx_memset(cpi->single_pred_count, 0, sizeof(cpi->single_pred_count)); vpx_memset(cpi->comp_pred_count, 0, sizeof(cpi->comp_pred_count)); -#if CONFIG_TX_SELECT vpx_memset(cpi->txfm_count, 0, sizeof(cpi->txfm_count)); vpx_memset(cpi->txfm_count_8x8p, 0, sizeof(cpi->txfm_count_8x8p)); vpx_memset(cpi->rd_tx_select_diff, 0, sizeof(cpi->rd_tx_select_diff)); -#endif { struct vpx_usec_timer emr_timer; vpx_usec_timer_start(&emr_timer); @@ -1457,7 +1434,6 @@ static int check_dual_ref_flags(VP8_COMP *cpi) { } } -#if CONFIG_TX_SELECT static void reset_skip_txfm_size(VP8_COMP *cpi, TX_SIZE txfm_max) { VP8_COMMON *cm = &cpi->common; int mb_row, mb_col, mis = cm->mode_info_stride; @@ -1481,7 +1457,6 @@ static void reset_skip_txfm_size(VP8_COMP *cpi, TX_SIZE txfm_max) { } } } -#endif void vp8_encode_frame(VP8_COMP *cpi) { if (cpi->sf.RD) { @@ -1527,7 +1502,6 @@ void vp8_encode_frame(VP8_COMP *cpi) { txfm_type = ONLY_4X4; } else #endif -#if CONFIG_TX_SELECT /* FIXME (rbultje) * this is a hack (no really), basically to work around the complete * nonsense coefficient cost prediction for keyframes. The probabilities @@ -1575,16 +1549,11 @@ void vp8_encode_frame(VP8_COMP *cpi) { cpi->rd_tx_select_threshes[frame_type][TX_MODE_SELECT] ? ALLOW_16X16 : TX_MODE_SELECT; #endif -#else - txfm_type = ALLOW_16X16; -#endif // CONFIG_TX_SELECT cpi->common.txfm_mode = txfm_type; -#if CONFIG_TX_SELECT if (txfm_type != TX_MODE_SELECT) { cpi->common.prob_tx[0] = 128; cpi->common.prob_tx[1] = 128; } -#endif cpi->common.comp_pred_mode = pred_type; encode_frame_internal(cpi); @@ -1594,7 +1563,6 @@ void vp8_encode_frame(VP8_COMP *cpi) { cpi->rd_prediction_type_threshes[frame_type][i] >>= 1; } -#if CONFIG_TX_SELECT for (i = 0; i < NB_TXFM_MODES; ++i) { int64_t pd = cpi->rd_tx_select_diff[i]; int diff; @@ -1604,7 +1572,6 @@ void vp8_encode_frame(VP8_COMP *cpi) { cpi->rd_tx_select_threshes[frame_type][i] += diff; cpi->rd_tx_select_threshes[frame_type][i] /= 2; } -#endif if (cpi->common.comp_pred_mode == HYBRID_PREDICTION) { int single_count_zero = 0; @@ -1622,7 +1589,6 @@ void vp8_encode_frame(VP8_COMP *cpi) { } } -#if CONFIG_TX_SELECT if (cpi->common.txfm_mode == TX_MODE_SELECT) { const int count4x4 = cpi->txfm_count[TX_4X4] + cpi->txfm_count_8x8p[TX_4X4]; const int count8x8 = cpi->txfm_count[TX_8X8]; @@ -1639,7 +1605,6 @@ void vp8_encode_frame(VP8_COMP *cpi) { cpi->common.txfm_mode = ALLOW_16X16; } } -#endif } else { encode_frame_internal(cpi); } @@ -1957,15 +1922,12 @@ void vp8cx_encode_intra_macro_block(VP8_COMP *cpi, } if (output_enabled) { -#if CONFIG_TX_SELECT int segment_id = mbmi->segment_id; -#endif // Tokenize sum_intra_stats(cpi, x); vp8_tokenize_mb(cpi, &x->e_mbd, t, 0); -#if CONFIG_TX_SELECT if (cpi->common.txfm_mode == TX_MODE_SELECT && !((cpi->common.mb_no_coeff_skip && mbmi->mb_skip_coeff) || (segfeature_active(&x->e_mbd, segment_id, SEG_LVL_EOB) && @@ -1975,9 +1937,7 @@ void vp8cx_encode_intra_macro_block(VP8_COMP *cpi, } else if (mbmi->mode == I8X8_PRED) { cpi->txfm_count_8x8p[mbmi->txfm_size]++; } - } else -#endif - if (cpi->common.txfm_mode >= ALLOW_16X16 && mbmi->mode <= TM_PRED) { + } else if (cpi->common.txfm_mode >= ALLOW_16X16 && mbmi->mode <= TM_PRED) { mbmi->txfm_size = TX_16X16; } else if (cpi->common.txfm_mode >= ALLOW_8X8 && mbmi->mode != B_PRED) { @@ -2012,9 +1972,7 @@ void vp8cx_encode_inter_macroblock (VP8_COMP *cpi, MACROBLOCK *x, assert(!xd->mode_info_context->mbmi.encoded_as_sb); #endif -#if CONFIG_SWITCHABLE_INTERP vp8_setup_interp_filters(xd, mbmi->interp_filter, cm); -#endif if (cpi->oxcf.tuning == VP8_TUNE_SSIM) { // Adjust the zbin based on this MB rate. adjust_act_zbin(cpi, x); @@ -2158,7 +2116,6 @@ void vp8cx_encode_inter_macroblock (VP8_COMP *cpi, MACROBLOCK *x, } if (output_enabled) { -#if CONFIG_TX_SELECT int segment_id = mbmi->segment_id; if (cpi->common.txfm_mode == TX_MODE_SELECT && !((cpi->common.mb_no_coeff_skip && mbmi->mb_skip_coeff) || @@ -2167,16 +2124,18 @@ void vp8cx_encode_inter_macroblock (VP8_COMP *cpi, MACROBLOCK *x, if (mbmi->mode != B_PRED && mbmi->mode != I8X8_PRED && mbmi->mode != SPLITMV) { cpi->txfm_count[mbmi->txfm_size]++; - } else if (mbmi->mode == I8X8_PRED) { + } else if (mbmi->mode == I8X8_PRED || + (mbmi->mode == SPLITMV && + mbmi->partitioning != PARTITIONING_4X4)) { cpi->txfm_count_8x8p[mbmi->txfm_size]++; } - } else -#endif - if (mbmi->mode != B_PRED && mbmi->mode != I8X8_PRED && + } else if (mbmi->mode != B_PRED && mbmi->mode != I8X8_PRED && mbmi->mode != SPLITMV && cpi->common.txfm_mode >= ALLOW_16X16) { mbmi->txfm_size = TX_16X16; - } else if (mbmi->mode != B_PRED && mbmi->mode != SPLITMV && - cpi->common.txfm_mode >= ALLOW_8X8) { + } else if (mbmi->mode != B_PRED && + !(mbmi->mode == SPLITMV && + mbmi->partitioning == PARTITIONING_4X4) && + cpi->common.txfm_mode >= ALLOW_8X8) { mbmi->txfm_size = TX_8X8; } else { mbmi->txfm_size = TX_4X4; diff --git a/vp8/encoder/encodeintra.c b/vp8/encoder/encodeintra.c index 9076780d9..f44df22ea 100644 --- a/vp8/encoder/encodeintra.c +++ b/vp8/encoder/encodeintra.c @@ -48,7 +48,7 @@ int vp8_encode_intra(VP8_COMP *cpi, MACROBLOCK *x, int use_16x16_pred) { } } - intra_pred_var = VARIANCE_INVOKE(&cpi->rtcd.variance, getmbss)(x->src_diff); + intra_pred_var = vp8_get_mb_ss(x->src_diff); return intra_pred_var; } @@ -57,9 +57,7 @@ void vp8_encode_intra4x4block(const VP8_ENCODER_RTCD *rtcd, MACROBLOCK *x, int ib) { BLOCKD *b = &x->e_mbd.block[ib]; BLOCK *be = &x->block[ib]; -#if CONFIG_HYBRIDTRANSFORM TX_TYPE tx_type; -#endif #if CONFIG_COMP_INTRA_PRED if (b->bmi.as_mode.second == (B_PREDICTION_MODE)(B_DC_PRED - 1)) { @@ -74,15 +72,12 @@ void vp8_encode_intra4x4block(const VP8_ENCODER_RTCD *rtcd, ENCODEMB_INVOKE(&rtcd->encodemb, subb)(be, b, 16); -#if CONFIG_HYBRIDTRANSFORM tx_type = get_tx_type(&x->e_mbd, b); if (tx_type != DCT_DCT) { vp8_fht_c(be->src_diff, be->coeff, 32, tx_type, 4); vp8_ht_quantize_b_4x4(be, b, tx_type); vp8_ihtllm_c(b->dqcoeff, b->diff, 32, tx_type, 4); - } else -#endif - { + } else { x->vp8_short_fdct4x4(be->src_diff, be->coeff, 32) ; x->quantize_b_4x4(be, b) ; vp8_inverse_transform_b_4x4(IF_RTCD(&rtcd->common->idct), b, 32) ; @@ -103,9 +98,7 @@ void vp8_encode_intra16x16mby(const VP8_ENCODER_RTCD *rtcd, MACROBLOCK *x) { MACROBLOCKD *xd = &x->e_mbd; BLOCK *b = &x->block[0]; TX_SIZE tx_size = xd->mode_info_context->mbmi.txfm_size; -#if CONFIG_HYBRIDTRANSFORM16X16 TX_TYPE tx_type; -#endif #if CONFIG_COMP_INTRA_PRED if (xd->mode_info_context->mbmi.second_mode == (MB_PREDICTION_MODE)(DC_PRED - 1)) @@ -120,7 +113,6 @@ void vp8_encode_intra16x16mby(const VP8_ENCODER_RTCD *rtcd, MACROBLOCK *x) { xd->predictor, b->src_stride); if (tx_size == TX_16X16) { -#if CONFIG_HYBRIDTRANSFORM16X16 BLOCKD *bd = &xd->block[0]; tx_type = get_tx_type(xd, bd); if (tx_type != DCT_DCT) { @@ -129,9 +121,7 @@ void vp8_encode_intra16x16mby(const VP8_ENCODER_RTCD *rtcd, MACROBLOCK *x) { if (x->optimize) vp8_optimize_mby_16x16(x, rtcd); vp8_ihtllm_c(bd->dqcoeff, bd->diff, 32, tx_type, 16); - } else -#endif - { + } else { vp8_transform_mby_16x16(x); vp8_quantize_mby_16x16(x); if (x->optimize) @@ -196,9 +186,7 @@ void vp8_encode_intra8x8(const VP8_ENCODER_RTCD *rtcd, BLOCK *be = &x->block[ib]; const int iblock[4] = {0, 1, 4, 5}; int i; -#if CONFIG_HYBRIDTRANSFORM8X8 TX_TYPE tx_type; -#endif #if CONFIG_COMP_INTRA_PRED if (b->bmi.as_mode.second == (MB_PREDICTION_MODE)(DC_PRED - 1)) { @@ -217,7 +205,6 @@ void vp8_encode_intra8x8(const VP8_ENCODER_RTCD *rtcd, // generate residual blocks vp8_subtract_4b_c(be, b, 16); -#if CONFIG_HYBRIDTRANSFORM8X8 tx_type = get_tx_type(xd, xd->block + idx); if (tx_type != DCT_DCT) { vp8_fht_c(be->src_diff, (x->block + idx)->coeff, 32, @@ -226,13 +213,10 @@ void vp8_encode_intra8x8(const VP8_ENCODER_RTCD *rtcd, vp8_ihtllm_c(xd->block[idx].dqcoeff, xd->block[ib].diff, 32, tx_type, 8); } else { -#endif x->vp8_short_fdct8x8(be->src_diff, (x->block + idx)->coeff, 32); x->quantize_b_8x8(x->block + idx, xd->block + idx); vp8_idct_idct8(xd->block[idx].dqcoeff, xd->block[ib].diff, 32); -#if CONFIG_HYBRIDTRANSFORM8X8 } -#endif } else { for (i = 0; i < 4; i++) { b = &xd->block[ib + iblock[i]]; diff --git a/vp8/encoder/encodemb.c b/vp8/encoder/encodemb.c index dc54d05a2..d3bd0f1dd 100644 --- a/vp8/encoder/encodemb.c +++ b/vp8/encoder/encodemb.c @@ -304,7 +304,6 @@ void optimize_b(MACROBLOCK *mb, int i, PLANE_TYPE type, scan = vp8_default_zig_zag1d; bands = vp8_coef_bands; default_eob = 16; -#if CONFIG_HYBRIDTRANSFORM // TODO: this isn't called (for intra4x4 modes), but will be left in // since it could be used later { @@ -327,7 +326,6 @@ void optimize_b(MACROBLOCK *mb, int i, PLANE_TYPE type, scan = vp8_default_zig_zag1d; } } -#endif break; case TX_8X8: scan = vp8_default_zig_zag1d_8x8; @@ -638,6 +636,7 @@ void vp8_optimize_mby_8x8(MACROBLOCK *x, const VP8_ENCODER_RTCD *rtcd) { ENTROPY_CONTEXT_PLANES t_above, t_left; ENTROPY_CONTEXT *ta; ENTROPY_CONTEXT *tl; + int has_2nd_order = x->e_mbd.mode_info_context->mbmi.mode != SPLITMV; if (!x->e_mbd.above_context || !x->e_mbd.left_context) return; @@ -647,18 +646,21 @@ void vp8_optimize_mby_8x8(MACROBLOCK *x, const VP8_ENCODER_RTCD *rtcd) { ta = (ENTROPY_CONTEXT *)&t_above; tl = (ENTROPY_CONTEXT *)&t_left; - type = PLANE_TYPE_Y_NO_DC; + type = has_2nd_order ? PLANE_TYPE_Y_NO_DC : PLANE_TYPE_Y_WITH_DC; for (b = 0; b < 16; b += 4) { optimize_b(x, b, type, ta + vp8_block2above_8x8[b], tl + vp8_block2left_8x8[b], rtcd, TX_8X8); - *(ta + vp8_block2above_8x8[b] + 1) = *(ta + vp8_block2above_8x8[b]); - *(tl + vp8_block2left_8x8[b] + 1) = *(tl + vp8_block2left_8x8[b]); + ta[vp8_block2above_8x8[b] + 1] = ta[vp8_block2above_8x8[b]]; + tl[vp8_block2left_8x8[b] + 1] = tl[vp8_block2left_8x8[b]]; } // 8x8 always have 2nd roder haar block - check_reset_8x8_2nd_coeffs(&x->e_mbd, - ta + vp8_block2above_8x8[24], tl + vp8_block2left_8x8[24]); + if (has_2nd_order) { + check_reset_8x8_2nd_coeffs(&x->e_mbd, + ta + vp8_block2above_8x8[24], + tl + vp8_block2left_8x8[24]); + } } void vp8_optimize_mbuv_8x8(MACROBLOCK *x, const VP8_ENCODER_RTCD *rtcd) { @@ -680,8 +682,8 @@ void vp8_optimize_mbuv_8x8(MACROBLOCK *x, const VP8_ENCODER_RTCD *rtcd) { optimize_b(x, b, PLANE_TYPE_UV, ta + vp8_block2above_8x8[b], tl + vp8_block2left_8x8[b], rtcd, TX_8X8); - *(ta + vp8_block2above_8x8[b] + 1) = *(ta + vp8_block2above_8x8[b]); - *(tl + vp8_block2left_8x8[b] + 1) = *(tl + vp8_block2left_8x8[b]); + ta[vp8_block2above_8x8[b] + 1] = ta[vp8_block2above_8x8[b]]; + tl[vp8_block2left_8x8[b] + 1] = tl[vp8_block2left_8x8[b]]; } } @@ -898,11 +900,25 @@ void vp8_encode_inter16x16(const VP8_ENCODER_RTCD *rtcd, MACROBLOCK *x) { optimize_mb_16x16(x, rtcd); vp8_inverse_transform_mb_16x16(IF_RTCD(&rtcd->common->idct), xd); } else if (tx_size == TX_8X8) { - vp8_transform_mb_8x8(x); - vp8_quantize_mb_8x8(x); - if (x->optimize) - optimize_mb_8x8(x, rtcd); - vp8_inverse_transform_mb_8x8(IF_RTCD(&rtcd->common->idct), xd); + if (xd->mode_info_context->mbmi.mode == SPLITMV) { + assert(xd->mode_info_context->mbmi.partitioning != PARTITIONING_4X4); + vp8_transform_mby_8x8(x); + vp8_transform_mbuv_4x4(x); + vp8_quantize_mby_8x8(x); + vp8_quantize_mbuv_4x4(x); + if (x->optimize) { + vp8_optimize_mby_8x8(x, rtcd); + vp8_optimize_mbuv_4x4(x, rtcd); + } + vp8_inverse_transform_mby_8x8(IF_RTCD(&rtcd->common->idct), xd); + vp8_inverse_transform_mbuv_4x4(IF_RTCD(&rtcd->common->idct), xd); + } else { + vp8_transform_mb_8x8(x); + vp8_quantize_mb_8x8(x); + if (x->optimize) + optimize_mb_8x8(x, rtcd); + vp8_inverse_transform_mb_8x8(IF_RTCD(&rtcd->common->idct), xd); + } } else { transform_mb_4x4(x); vp8_quantize_mb_4x4(x); diff --git a/vp8/encoder/encodemv.c b/vp8/encoder/encodemv.c index d520d995a..75dad2f9b 100644 --- a/vp8/encoder/encodemv.c +++ b/vp8/encoder/encodemv.c @@ -22,8 +22,6 @@ extern unsigned int active_section; //extern int final_packing; -#if CONFIG_NEWMVENTROPY - #ifdef NMV_STATS nmv_context_counts tnmvcounts; #endif @@ -549,593 +547,3 @@ void vp8_build_nmv_cost_table(int *mvjoint, if (mvc_flag_h) build_nmv_component_cost_table(mvcost[1], &mvctx->comps[1], usehp); } - -#else /* CONFIG_NEWMVENTROPY */ - -static void encode_mvcomponent( - vp8_writer *const bc, - const int v, - const struct mv_context *mvc -) { - const vp8_prob *p = mvc->prob; - const int x = v < 0 ? -v : v; - - if (x < mvnum_short) { // Small - vp8_write(bc, 0, p[mvpis_short]); - vp8_treed_write(bc, vp8_small_mvtree, p + MVPshort, x, mvnum_short_bits); - if (!x) - return; // no sign bit - } else { // Large - int i = 0; - - vp8_write(bc, 1, p[mvpis_short]); - - do - vp8_write(bc, (x >> i) & 1, p[MVPbits + i]); - - while (++i < mvnum_short_bits); - - i = mvlong_width - 1; /* Skip bit 3, which is sometimes implicit */ - - do - vp8_write(bc, (x >> i) & 1, p[MVPbits + i]); - - while (--i > mvnum_short_bits); - - if (x & ~((2 << mvnum_short_bits) - 1)) - vp8_write(bc, (x >> mvnum_short_bits) & 1, p[MVPbits + mvnum_short_bits]); - } - - vp8_write(bc, v < 0, p[MVPsign]); -} - -void vp8_encode_motion_vector(vp8_writer* const bc, - const MV* const mv, - const MV_CONTEXT* const mvc) { - encode_mvcomponent(bc, mv->row >> 1, &mvc[0]); - encode_mvcomponent(bc, mv->col >> 1, &mvc[1]); -} - - -static unsigned int cost_mvcomponent(const int v, - const struct mv_context* const mvc) { - const vp8_prob *p = mvc->prob; - const int x = v; // v<0? -v:v; - unsigned int cost; - - if (x < mvnum_short) { - cost = vp8_cost_zero(p [mvpis_short]) - + vp8_treed_cost(vp8_small_mvtree, p + MVPshort, x, mvnum_short_bits); - - if (!x) - return cost; - } else { - int i = 0; - cost = vp8_cost_one(p [mvpis_short]); - - do - cost += vp8_cost_bit(p [MVPbits + i], (x >> i) & 1); - - while (++i < mvnum_short_bits); - - i = mvlong_width - 1; /* Skip bit 3, which is sometimes implicit */ - - do - cost += vp8_cost_bit(p [MVPbits + i], (x >> i) & 1); - - while (--i > mvnum_short_bits); - - if (x & ~((2 << mvnum_short_bits) - 1)) - cost += vp8_cost_bit(p [MVPbits + mvnum_short_bits], (x >> mvnum_short_bits) & 1); - } - - return cost; // + vp8_cost_bit( p [MVPsign], v < 0); -} - -void vp8_build_component_cost_table(int *mvcost[2], const MV_CONTEXT *mvc, - const int mvc_flag[2]) { - int i = 1; // -mv_max; - unsigned int cost0 = 0; - unsigned int cost1 = 0; - - vp8_clear_system_state(); - - i = 1; - - if (mvc_flag[0]) { - mvcost [0] [0] = cost_mvcomponent(0, &mvc[0]); - - do { - // mvcost [0] [i] = cost_mvcomponent( i, &mvc[0]); - cost0 = cost_mvcomponent(i, &mvc[0]); - - mvcost [0] [i] = cost0 + vp8_cost_zero(mvc[0].prob[MVPsign]); - mvcost [0] [-i] = cost0 + vp8_cost_one(mvc[0].prob[MVPsign]); - } while (++i <= mv_max); - } - - i = 1; - - if (mvc_flag[1]) { - mvcost [1] [0] = cost_mvcomponent(0, &mvc[1]); - - do { - // mvcost [1] [i] = cost_mvcomponent( i, mvc[1]); - cost1 = cost_mvcomponent(i, &mvc[1]); - - mvcost [1] [i] = cost1 + vp8_cost_zero(mvc[1].prob[MVPsign]); - mvcost [1] [-i] = cost1 + vp8_cost_one(mvc[1].prob[MVPsign]); - } while (++i <= mv_max); - } -} - - -// Motion vector probability table update depends on benefit. -// Small correction allows for the fact that an update to an MV probability -// may have benefit in subsequent frames as well as the current one. - -#define MV_PROB_UPDATE_CORRECTION -1 - - -__inline static void calc_prob(vp8_prob *p, const unsigned int ct[2]) { - const unsigned int tot = ct[0] + ct[1]; - - if (tot) { - const vp8_prob x = ((ct[0] * 255) / tot) & -2; - *p = x ? x : 1; - } -} - -static void update( - vp8_writer *const bc, - const unsigned int ct[2], - vp8_prob *const cur_p, - const vp8_prob new_p, - const vp8_prob update_p, - int *updated -) { - const int cur_b = vp8_cost_branch(ct, *cur_p); - const int new_b = vp8_cost_branch(ct, new_p); - const int cost = 7 + MV_PROB_UPDATE_CORRECTION + ((vp8_cost_one(update_p) - vp8_cost_zero(update_p) + 128) >> 8); - - if (cur_b - new_b > cost) { - *cur_p = new_p; - vp8_write(bc, 1, update_p); - vp8_write_literal(bc, new_p >> 1, 7); - *updated = 1; - - } else - vp8_write(bc, 0, update_p); -} - -static void write_component_probs( - vp8_writer *const bc, - struct mv_context *cur_mvc, - const struct mv_context *default_mvc_, - const struct mv_context *update_mvc, - const unsigned int events [MVvals], - unsigned int rc, - int *updated -) { - vp8_prob *Pcur = cur_mvc->prob; - const vp8_prob *default_mvc = default_mvc_->prob; - const vp8_prob *Pupdate = update_mvc->prob; - unsigned int is_short_ct[2], sign_ct[2]; - - unsigned int bit_ct [mvlong_width] [2]; - - unsigned int short_ct [mvnum_short]; - unsigned int short_bct [mvnum_short - 1] [2]; - - vp8_prob Pnew [MVPcount]; - - (void) rc; - vp8_copy_array(Pnew, default_mvc, MVPcount); - - vp8_zero(is_short_ct) - vp8_zero(sign_ct) - vp8_zero(bit_ct) - vp8_zero(short_ct) - vp8_zero(short_bct) - - - // j=0 - { - const int c = events [mv_max]; - - is_short_ct [0] += c; // Short vector - short_ct [0] += c; // Magnitude distribution - } - - // j: 1 ~ mv_max (1023) - { - int j = 1; - - do { - const int c1 = events [mv_max + j]; // positive - const int c2 = events [mv_max - j]; // negative - const int c = c1 + c2; - int a = j; - - sign_ct [0] += c1; - sign_ct [1] += c2; - - if (a < mvnum_short) { - is_short_ct [0] += c; // Short vector - short_ct [a] += c; // Magnitude distribution - } else { - int k = mvlong_width - 1; - is_short_ct [1] += c; // Long vector - - /* bit 3 not always encoded. */ - do - bit_ct [k] [(a >> k) & 1] += c; - - while (--k >= 0); - } - } while (++j <= mv_max); - } - - calc_prob(Pnew + mvpis_short, is_short_ct); - - calc_prob(Pnew + MVPsign, sign_ct); - - { - vp8_prob p [mvnum_short - 1]; /* actually only need branch ct */ - int j = 0; - - vp8_tree_probs_from_distribution( - mvnum_short, vp8_small_mvencodings, vp8_small_mvtree, - p, short_bct, short_ct, - 256, 1 - ); - - do - calc_prob(Pnew + MVPshort + j, short_bct[j]); - - while (++j < mvnum_short - 1); - } - - { - int j = 0; - - do - calc_prob(Pnew + MVPbits + j, bit_ct[j]); - - while (++j < mvlong_width); - } - - update(bc, is_short_ct, Pcur + mvpis_short, Pnew[mvpis_short], - *Pupdate++, updated); - - update(bc, sign_ct, Pcur + MVPsign, Pnew[MVPsign], - *Pupdate++, updated); - - { - const vp8_prob *const new_p = Pnew + MVPshort; - vp8_prob *const cur_p = Pcur + MVPshort; - - int j = 0; - - do - - update(bc, short_bct[j], cur_p + j, new_p[j], *Pupdate++, updated); - - while (++j < mvnum_short - 1); - } - - { - const vp8_prob *const new_p = Pnew + MVPbits; - vp8_prob *const cur_p = Pcur + MVPbits; - - int j = 0; - - do - - update(bc, bit_ct[j], cur_p + j, new_p[j], *Pupdate++, updated); - - while (++j < mvlong_width); - } -} - -void vp8_write_mvprobs(VP8_COMP* const cpi, vp8_writer* const bc) { - MV_CONTEXT *mvc = cpi->common.fc.mvc; - int flags[2] = {0, 0}; -#ifdef ENTROPY_STATS - active_section = 4; -#endif - write_component_probs( - bc, &mvc[0], &vp8_default_mv_context[0], &vp8_mv_update_probs[0], - cpi->MVcount[0], 0, &flags[0]); - - write_component_probs( - bc, &mvc[1], &vp8_default_mv_context[1], &vp8_mv_update_probs[1], - cpi->MVcount[1], 1, &flags[1]); - - if (flags[0] || flags[1]) - vp8_build_component_cost_table(cpi->mb.mvcost, (const MV_CONTEXT *) cpi->common.fc.mvc, flags); - -#ifdef ENTROPY_STATS - active_section = 5; -#endif -} - - -static void encode_mvcomponent_hp( - vp8_writer *const bc, - const int v, - const struct mv_context_hp *mvc -) { - const vp8_prob *p = mvc->prob; - const int x = v < 0 ? -v : v; - - if (x < mvnum_short_hp) { // Small - vp8_write(bc, 0, p[mvpis_short_hp]); - vp8_treed_write(bc, vp8_small_mvtree_hp, p + MVPshort_hp, x, - mvnum_short_bits_hp); - if (!x) - return; // no sign bit - } else { // Large - int i = 0; - - vp8_write(bc, 1, p[mvpis_short_hp]); - - do - vp8_write(bc, (x >> i) & 1, p[MVPbits_hp + i]); - - while (++i < mvnum_short_bits_hp); - - i = mvlong_width_hp - 1; /* Skip bit 3, which is sometimes implicit */ - - do - vp8_write(bc, (x >> i) & 1, p[MVPbits_hp + i]); - - while (--i > mvnum_short_bits_hp); - - if (x & ~((2 << mvnum_short_bits_hp) - 1)) - vp8_write(bc, (x >> mvnum_short_bits_hp) & 1, - p[MVPbits_hp + mvnum_short_bits_hp]); - } - - vp8_write(bc, v < 0, p[MVPsign_hp]); -} - -void vp8_encode_motion_vector_hp(vp8_writer *bc, const MV *mv, - const MV_CONTEXT_HP *mvc) { - - encode_mvcomponent_hp(bc, mv->row, &mvc[0]); - encode_mvcomponent_hp(bc, mv->col, &mvc[1]); -} - - -static unsigned int cost_mvcomponent_hp(const int v, - const struct mv_context_hp *mvc) { - const vp8_prob *p = mvc->prob; - const int x = v; // v<0? -v:v; - unsigned int cost; - - if (x < mvnum_short_hp) { - cost = vp8_cost_zero(p [mvpis_short_hp]) - + vp8_treed_cost(vp8_small_mvtree_hp, p + MVPshort_hp, x, - mvnum_short_bits_hp); - - if (!x) - return cost; - } else { - int i = 0; - cost = vp8_cost_one(p [mvpis_short_hp]); - - do - cost += vp8_cost_bit(p [MVPbits_hp + i], (x >> i) & 1); - - while (++i < mvnum_short_bits_hp); - - i = mvlong_width_hp - 1; /* Skip bit 3, which is sometimes implicit */ - - do - cost += vp8_cost_bit(p [MVPbits_hp + i], (x >> i) & 1); - - while (--i > mvnum_short_bits_hp); - - if (x & ~((2 << mvnum_short_bits_hp) - 1)) - cost += vp8_cost_bit(p [MVPbits_hp + mvnum_short_bits_hp], - (x >> mvnum_short_bits_hp) & 1); - } - - return cost; // + vp8_cost_bit( p [MVPsign], v < 0); -} - -void vp8_build_component_cost_table_hp(int *mvcost[2], - const MV_CONTEXT_HP *mvc, - const int mvc_flag[2]) { - int i = 1; // -mv_max; - unsigned int cost0 = 0; - unsigned int cost1 = 0; - - vp8_clear_system_state(); - - i = 1; - - if (mvc_flag[0]) { - mvcost [0] [0] = cost_mvcomponent_hp(0, &mvc[0]); - - do { - // mvcost [0] [i] = cost_mvcomponent( i, &mvc[0]); - cost0 = cost_mvcomponent_hp(i, &mvc[0]); - - mvcost [0] [i] = cost0 + vp8_cost_zero(mvc[0].prob[MVPsign_hp]); - mvcost [0] [-i] = cost0 + vp8_cost_one(mvc[0].prob[MVPsign_hp]); - } while (++i <= mv_max_hp); - } - - i = 1; - - if (mvc_flag[1]) { - mvcost [1] [0] = cost_mvcomponent_hp(0, &mvc[1]); - - do { - // mvcost [1] [i] = cost_mvcomponent( i, mvc[1]); - cost1 = cost_mvcomponent_hp(i, &mvc[1]); - - mvcost [1] [i] = cost1 + vp8_cost_zero(mvc[1].prob[MVPsign_hp]); - mvcost [1] [-i] = cost1 + vp8_cost_one(mvc[1].prob[MVPsign_hp]); - } while (++i <= mv_max_hp); - } -} - - -static void write_component_probs_hp( - vp8_writer *const bc, - struct mv_context_hp *cur_mvc, - const struct mv_context_hp *default_mvc_, - const struct mv_context_hp *update_mvc, - const unsigned int events [MVvals_hp], - unsigned int rc, - int *updated -) { - vp8_prob *Pcur = cur_mvc->prob; - const vp8_prob *default_mvc = default_mvc_->prob; - const vp8_prob *Pupdate = update_mvc->prob; - unsigned int is_short_ct[2], sign_ct[2]; - - unsigned int bit_ct [mvlong_width_hp] [2]; - - unsigned int short_ct [mvnum_short_hp]; - unsigned int short_bct [mvnum_short_hp - 1] [2]; - - vp8_prob Pnew [MVPcount_hp]; - - (void) rc; - vp8_copy_array(Pnew, default_mvc, MVPcount_hp); - - vp8_zero(is_short_ct) - vp8_zero(sign_ct) - vp8_zero(bit_ct) - vp8_zero(short_ct) - vp8_zero(short_bct) - - - // j=0 - { - const int c = events [mv_max_hp]; - - is_short_ct [0] += c; // Short vector - short_ct [0] += c; // Magnitude distribution - } - - // j: 1 ~ mv_max (1023) - { - int j = 1; - - do { - const int c1 = events [mv_max_hp + j]; // positive - const int c2 = events [mv_max_hp - j]; // negative - const int c = c1 + c2; - int a = j; - - sign_ct [0] += c1; - sign_ct [1] += c2; - - if (a < mvnum_short_hp) { - is_short_ct [0] += c; // Short vector - short_ct [a] += c; // Magnitude distribution - } else { - int k = mvlong_width_hp - 1; - is_short_ct [1] += c; // Long vector - - /* bit 3 not always encoded. */ - do - bit_ct [k] [(a >> k) & 1] += c; - - while (--k >= 0); - } - } while (++j <= mv_max_hp); - } - - calc_prob(Pnew + mvpis_short_hp, is_short_ct); - - calc_prob(Pnew + MVPsign_hp, sign_ct); - - { - vp8_prob p [mvnum_short_hp - 1]; /* actually only need branch ct */ - int j = 0; - - vp8_tree_probs_from_distribution( - mvnum_short_hp, vp8_small_mvencodings_hp, vp8_small_mvtree_hp, - p, short_bct, short_ct, - 256, 1 - ); - - do - calc_prob(Pnew + MVPshort_hp + j, short_bct[j]); - - while (++j < mvnum_short_hp - 1); - } - - { - int j = 0; - - do - calc_prob(Pnew + MVPbits_hp + j, bit_ct[j]); - - while (++j < mvlong_width_hp); - } - - update(bc, is_short_ct, Pcur + mvpis_short_hp, Pnew[mvpis_short_hp], - *Pupdate++, updated); - - update(bc, sign_ct, Pcur + MVPsign_hp, Pnew[MVPsign_hp], *Pupdate++, - updated); - - { - const vp8_prob *const new_p = Pnew + MVPshort_hp; - vp8_prob *const cur_p = Pcur + MVPshort_hp; - - int j = 0; - - do - - update(bc, short_bct[j], cur_p + j, new_p[j], *Pupdate++, updated); - - while (++j < mvnum_short_hp - 1); - } - - { - const vp8_prob *const new_p = Pnew + MVPbits_hp; - vp8_prob *const cur_p = Pcur + MVPbits_hp; - - int j = 0; - - do - - update(bc, bit_ct[j], cur_p + j, new_p[j], *Pupdate++, updated); - - while (++j < mvlong_width_hp); - } -} - -void vp8_write_mvprobs_hp(VP8_COMP* const cpi, vp8_writer* const bc) { - MV_CONTEXT_HP *mvc = cpi->common.fc.mvc_hp; - int flags[2] = {0, 0}; -#ifdef ENTROPY_STATS - active_section = 4; -#endif - write_component_probs_hp( - bc, &mvc[0], &vp8_default_mv_context_hp[0], &vp8_mv_update_probs_hp[0], - cpi->MVcount_hp[0], 0, &flags[0] - ); - write_component_probs_hp( - bc, &mvc[1], &vp8_default_mv_context_hp[1], &vp8_mv_update_probs_hp[1], - cpi->MVcount_hp[1], 1, &flags[1] - ); - - if (flags[0] || flags[1]) - vp8_build_component_cost_table_hp(cpi->mb.mvcost_hp, - (const MV_CONTEXT_HP *) - cpi->common.fc.mvc_hp, flags); -#ifdef ENTROPY_STATS - active_section = 5; -#endif -} - -#endif /* CONFIG_NEWMVENTROPY */ diff --git a/vp8/encoder/encodemv.h b/vp8/encoder/encodemv.h index c06831cb2..254536580 100644 --- a/vp8/encoder/encodemv.h +++ b/vp8/encoder/encodemv.h @@ -14,7 +14,6 @@ #include "onyx_int.h" -#if CONFIG_NEWMVENTROPY void vp8_write_nmvprobs(VP8_COMP* const, int usehp, vp8_writer* const); void vp8_encode_nmv(vp8_writer* const w, const MV* const mv, const MV* const ref, const nmv_context* const mvctx); @@ -27,19 +26,5 @@ void vp8_build_nmv_cost_table(int *mvjoint, int usehp, int mvc_flag_v, int mvc_flag_h); -#else /* CONFIG_NEWMVENTROPY */ -void vp8_write_mvprobs(VP8_COMP* const, vp8_writer* const); -void vp8_encode_motion_vector(vp8_writer* const, const MV* const, - const MV_CONTEXT* const); -void vp8_build_component_cost_table(int *mvcost[2], - const MV_CONTEXT*, - const int mvc_flag[2]); -void vp8_write_mvprobs_hp(VP8_COMP* const, vp8_writer* const); -void vp8_encode_motion_vector_hp(vp8_writer* const, const MV* const, - const MV_CONTEXT_HP* const); -void vp8_build_component_cost_table_hp(int *mvcost[2], - const MV_CONTEXT_HP*, - const int mvc_flag[2]); -#endif /* CONFIG_NEWMVENTROPY */ #endif diff --git a/vp8/encoder/firstpass.c b/vp8/encoder/firstpass.c index 95f6d97d3..09d5a762e 100644 --- a/vp8/encoder/firstpass.c +++ b/vp8/encoder/firstpass.c @@ -364,7 +364,8 @@ static void zz_motion_search(VP8_COMP *cpi, MACROBLOCK *x, YV12_BUFFER_CONFIG *r ref_ptr = (unsigned char *)(*(d->base_pre) + d->pre); - VARIANCE_INVOKE(IF_RTCD(&cpi->rtcd.variance), mse16x16)(src_ptr, src_stride, ref_ptr, ref_stride, (unsigned int *)(best_motion_err)); + vp8_mse16x16(src_ptr, src_stride, ref_ptr, ref_stride, + (unsigned int *)(best_motion_err)); } static void first_pass_motion_search(VP8_COMP *cpi, MACROBLOCK *x, @@ -387,7 +388,7 @@ static void first_pass_motion_search(VP8_COMP *cpi, MACROBLOCK *x, int new_mv_mode_penalty = 256; // override the default variance function to use MSE - v_fn_ptr.vf = VARIANCE_INVOKE(IF_RTCD(&cpi->rtcd.variance), mse16x16); + v_fn_ptr.vf = vp8_mse16x16; // Set up pointers for this macro block recon buffer xd->pre.y_buffer = recon_buffer->y_buffer + recon_yoffset; @@ -492,12 +493,7 @@ void vp8_first_pass(VP8_COMP *cpi) { { int flag[2] = {1, 1}; vp8_init_mv_probs(cm); -#if CONFIG_NEWMVENTROPY vp8_initialize_rd_consts(cpi, cm->base_qindex + cm->y1dc_delta_q); -#else - vp8_build_component_cost_table(cpi->mb.mvcost, (const MV_CONTEXT *) cm->fc.mvc, flag); - vp8_build_component_cost_table_hp(cpi->mb.mvcost_hp, (const MV_CONTEXT_HP *) cm->fc.mvc_hp, flag); -#endif } // for each macroblock row in image diff --git a/vp8/encoder/generic/csystemdependent.c b/vp8/encoder/generic/csystemdependent.c index 356e32c3f..44e83fdc7 100644 --- a/vp8/encoder/generic/csystemdependent.c +++ b/vp8/encoder/generic/csystemdependent.c @@ -23,80 +23,6 @@ extern void vp8_yv12_copy_partial_frame(YV12_BUFFER_CONFIG *src_ybc, YV12_BUFFER void vp8_cmachine_specific_config(VP8_COMP *cpi) { #if CONFIG_RUNTIME_CPU_DETECT cpi->rtcd.common = &cpi->common.rtcd; -#if CONFIG_SUPERBLOCKS - cpi->rtcd.variance.sad32x32 = vp8_sad32x32_c; -#endif - cpi->rtcd.variance.sad16x16 = vp8_sad16x16_c; - cpi->rtcd.variance.sad16x8 = vp8_sad16x8_c; - cpi->rtcd.variance.sad8x16 = vp8_sad8x16_c; - cpi->rtcd.variance.sad8x8 = vp8_sad8x8_c; - cpi->rtcd.variance.sad4x4 = vp8_sad4x4_c; - -#if CONFIG_SUPERBLOCKS - cpi->rtcd.variance.sad32x32x3 = vp8_sad32x32x3_c; -#endif - cpi->rtcd.variance.sad16x16x3 = vp8_sad16x16x3_c; - cpi->rtcd.variance.sad16x8x3 = vp8_sad16x8x3_c; - cpi->rtcd.variance.sad8x16x3 = vp8_sad8x16x3_c; - cpi->rtcd.variance.sad8x8x3 = vp8_sad8x8x3_c; - cpi->rtcd.variance.sad4x4x3 = vp8_sad4x4x3_c; - -#if CONFIG_SUPERBLOCKS - cpi->rtcd.variance.sad32x32x8 = vp8_sad32x32x8_c; -#endif - cpi->rtcd.variance.sad16x16x8 = vp8_sad16x16x8_c; - cpi->rtcd.variance.sad16x8x8 = vp8_sad16x8x8_c; - cpi->rtcd.variance.sad8x16x8 = vp8_sad8x16x8_c; - cpi->rtcd.variance.sad8x8x8 = vp8_sad8x8x8_c; - cpi->rtcd.variance.sad4x4x8 = vp8_sad4x4x8_c; - -#if CONFIG_SUPERBLOCKS - cpi->rtcd.variance.sad32x32x4d = vp8_sad32x32x4d_c; -#endif - cpi->rtcd.variance.sad16x16x4d = vp8_sad16x16x4d_c; - cpi->rtcd.variance.sad16x8x4d = vp8_sad16x8x4d_c; - cpi->rtcd.variance.sad8x16x4d = vp8_sad8x16x4d_c; - cpi->rtcd.variance.sad8x8x4d = vp8_sad8x8x4d_c; - cpi->rtcd.variance.sad4x4x4d = vp8_sad4x4x4d_c; -#if ARCH_X86 || ARCH_X86_64 - cpi->rtcd.variance.copy32xn = vp8_copy32xn_c; -#endif - cpi->rtcd.variance.var4x4 = vp8_variance4x4_c; - cpi->rtcd.variance.var8x8 = vp8_variance8x8_c; - cpi->rtcd.variance.var8x16 = vp8_variance8x16_c; - cpi->rtcd.variance.var16x8 = vp8_variance16x8_c; - cpi->rtcd.variance.var16x16 = vp8_variance16x16_c; -#if CONFIG_SUPERBLOCKS - cpi->rtcd.variance.var32x32 = vp8_variance32x32_c; -#endif - - cpi->rtcd.variance.subpixvar4x4 = vp8_sub_pixel_variance4x4_c; - cpi->rtcd.variance.subpixvar8x8 = vp8_sub_pixel_variance8x8_c; - cpi->rtcd.variance.subpixvar8x16 = vp8_sub_pixel_variance8x16_c; - cpi->rtcd.variance.subpixvar16x8 = vp8_sub_pixel_variance16x8_c; - cpi->rtcd.variance.subpixvar16x16 = vp8_sub_pixel_variance16x16_c; -#if CONFIG_SUPERBLOCKS - cpi->rtcd.variance.subpixvar32x32 = vp8_sub_pixel_variance32x32_c; -#endif - cpi->rtcd.variance.halfpixvar16x16_h = vp8_variance_halfpixvar16x16_h_c; -#if CONFIG_SUPERBLOCKS - cpi->rtcd.variance.halfpixvar32x32_h = vp8_variance_halfpixvar32x32_h_c; -#endif - cpi->rtcd.variance.halfpixvar16x16_v = vp8_variance_halfpixvar16x16_v_c; -#if CONFIG_SUPERBLOCKS - cpi->rtcd.variance.halfpixvar32x32_v = vp8_variance_halfpixvar32x32_v_c; -#endif - cpi->rtcd.variance.halfpixvar16x16_hv = vp8_variance_halfpixvar16x16_hv_c; -#if CONFIG_SUPERBLOCKS - cpi->rtcd.variance.halfpixvar32x32_hv = vp8_variance_halfpixvar32x32_hv_c; -#endif - cpi->rtcd.variance.subpixmse16x16 = vp8_sub_pixel_mse16x16_c; -#if CONFIG_SUPERBLOCKS - cpi->rtcd.variance.subpixmse32x32 = vp8_sub_pixel_mse32x32_c; -#endif - - cpi->rtcd.variance.mse16x16 = vp8_mse16x16_c; - cpi->rtcd.variance.getmbss = vp8_get_mb_ss_c; cpi->rtcd.fdct.short8x8 = vp8_short_fdct8x8_c; cpi->rtcd.fdct.short16x16 = vp8_short_fdct16x16_c; @@ -118,16 +44,11 @@ void vp8_cmachine_specific_config(VP8_COMP *cpi) { cpi->rtcd.search.refining_search = vp8_refining_search_sad; cpi->rtcd.search.diamond_search = vp8_diamond_search_sad; cpi->rtcd.temporal.apply = vp8_temporal_filter_apply_c; - cpi->rtcd.variance.satd16x16 = vp8_satd16x16_c; cpi->rtcd.fdct.short4x4 = vp8_short_fdct4x4_c; cpi->rtcd.fdct.short8x4 = vp8_short_fdct8x4_c; cpi->rtcd.fdct.fast4x4 = vp8_short_fdct4x4_c; cpi->rtcd.fdct.fast8x4 = vp8_short_fdct8x4_c; cpi->rtcd.fdct.walsh_short4x4 = vp8_short_walsh4x4_c; -#if CONFIG_INTERNAL_STATS - cpi->rtcd.variance.ssimpf_8x8 = vp8_ssim_parms_8x8_c; - cpi->rtcd.variance.ssimpf_16x16 = vp8_ssim_parms_16x16_c; -#endif #endif vp8_yv12_copy_partial_frame_ptr = vp8_yv12_copy_partial_frame; diff --git a/vp8/encoder/mbgraph.c b/vp8/encoder/mbgraph.c index 180ee5870..2eecfcdad 100644 --- a/vp8/encoder/mbgraph.c +++ b/vp8/encoder/mbgraph.c @@ -83,10 +83,8 @@ static unsigned int do_16x16_motion_iteration vp8_set_mbmode_and_mvs(x, NEWMV, dst_mv); vp8_build_1st_inter16x16_predictors_mby(xd, xd->predictor, 16, 0); - // VARIANCE_INVOKE(&cpi->rtcd.variance, satd16x16) - best_err = VARIANCE_INVOKE(&cpi->rtcd.variance, sad16x16) - (xd->dst.y_buffer, xd->dst.y_stride, - xd->predictor, 16, INT_MAX); + best_err = vp8_sad16x16(xd->dst.y_buffer, xd->dst.y_stride, + xd->predictor, 16, INT_MAX); /* restore UMV window */ x->mv_col_min = tmp_col_min; @@ -130,11 +128,8 @@ static int do_16x16_motion_search // FIXME should really use something like near/nearest MV and/or MV prediction xd->pre.y_buffer = ref->y_buffer + mb_y_offset; xd->pre.y_stride = ref->y_stride; - // VARIANCE_INVOKE(&cpi->rtcd.variance, satd16x16) - err = VARIANCE_INVOKE(&cpi->rtcd.variance, sad16x16) - (ref->y_buffer + mb_y_offset, - ref->y_stride, xd->dst.y_buffer, - xd->dst.y_stride, INT_MAX); + err = vp8_sad16x16(ref->y_buffer + mb_y_offset, ref->y_stride, + xd->dst.y_buffer, xd->dst.y_stride, INT_MAX); dst_mv->as_int = 0; // Test last reference frame using the previous best mv as the @@ -193,10 +188,8 @@ static int do_16x16_zerozero_search xd->pre.y_buffer = ref->y_buffer + mb_y_offset; xd->pre.y_stride = ref->y_stride; // VARIANCE_INVOKE(&cpi->rtcd.variance, satd16x16) - err = VARIANCE_INVOKE(&cpi->rtcd.variance, sad16x16) - (ref->y_buffer + mb_y_offset, - ref->y_stride, xd->dst.y_buffer, - xd->dst.y_stride, INT_MAX); + err = vp8_sad16x16(ref->y_buffer + mb_y_offset, ref->y_stride, + xd->dst.y_buffer, xd->dst.y_stride, INT_MAX); dst_mv->as_int = 0; @@ -221,11 +214,8 @@ static int find_best_16x16_intra xd->mode_info_context->mbmi.mode = mode; vp8_build_intra_predictors_mby(xd); - // VARIANCE_INVOKE(&cpi->rtcd.variance, satd16x16) - err = VARIANCE_INVOKE(&cpi->rtcd.variance, sad16x16) - (xd->predictor, 16, - buf->y_buffer + mb_y_offset, - buf->y_stride, best_err); + err = vp8_sad16x16(xd->predictor, 16, buf->y_buffer + mb_y_offset, + buf->y_stride, best_err); // find best if (err < best_err) { best_err = err; diff --git a/vp8/encoder/mcomp.c b/vp8/encoder/mcomp.c index a6cf2f18b..210887491 100644 --- a/vp8/encoder/mcomp.c +++ b/vp8/encoder/mcomp.c @@ -47,15 +47,9 @@ int vp8_mv_bit_cost(int_mv *mv, int_mv *ref, DEC_MVCOSTS, MV v; v.row = (mv->as_mv.row - ref->as_mv.row); v.col = (mv->as_mv.col - ref->as_mv.col); -#if CONFIG_NEWMVENTROPY return ((mvjcost[vp8_get_mv_joint(v)] + mvcost[0][v.row] + mvcost[1][v.col]) * Weight) >> 7; -#else - return ((mvcost[0][v.row >> (ishp == 0)] + - mvcost[1][v.col >> (ishp == 0)]) - * Weight) >> 7; -#endif } static int mv_err_cost(int_mv *mv, int_mv *ref, DEC_MVCOSTS, @@ -64,14 +58,9 @@ static int mv_err_cost(int_mv *mv, int_mv *ref, DEC_MVCOSTS, MV v; v.row = (mv->as_mv.row - ref->as_mv.row); v.col = (mv->as_mv.col - ref->as_mv.col); -#if CONFIG_NEWMVENTROPY return ((mvjcost[vp8_get_mv_joint(v)] + mvcost[0][v.row] + mvcost[1][v.col]) * error_per_bit + 128) >> 8; -#else - return ((mvcost[0][v.row >> (ishp == 0)] + - mvcost[1][v.col >> (ishp == 0)]) * error_per_bit + 128) >> 8; -#endif } return 0; } @@ -83,14 +72,9 @@ static int mvsad_err_cost(int_mv *mv, int_mv *ref, DEC_MVSADCOSTS, MV v; v.row = (mv->as_mv.row - ref->as_mv.row); v.col = (mv->as_mv.col - ref->as_mv.col); -#if CONFIG_NEWMVENTROPY return ((mvjsadcost[vp8_get_mv_joint(v)] + mvsadcost[0][v.row] + mvsadcost[1][v.col]) * error_per_bit + 128) >> 8; -#else - return ((mvsadcost[0][v.row] + mvsadcost[1][v.col]) - * error_per_bit + 128) >> 8; -#endif } return 0; } @@ -220,35 +204,42 @@ void vp8_init3smotion_compensation(MACROBLOCK *x, int stride) { * could reduce the area. */ -#if CONFIG_NEWMVENTROPY /* estimated cost of a motion vector (r,c) */ -#define MVC(r,c) \ - (mvcost ? \ - ((mvjcost[((r)!=rr)*2 + ((c)!=rc)] + \ - mvcost[0][((r)-rr)] + mvcost[1][((c)-rc)]) * error_per_bit + 128 )>>8 : 0) -#else -#define MVC(r,c) \ - (mvcost ? \ - ((mvcost[0][((r)-rr)>>(xd->allow_high_precision_mv==0)] + \ - mvcost[1][((c)-rc)>>(xd->allow_high_precision_mv==0)]) * \ - error_per_bit + 128 )>>8 : 0) -#endif /* CONFIG_NEWMVENTROPY */ +#define MVC(r, c) \ + (mvcost ? \ + ((mvjcost[((r) != rr) * 2 + ((c) != rc)] + \ + mvcost[0][((r) - rr)] + mvcost[1][((c) - rc)]) * \ + error_per_bit + 128) >> 8 : 0) -#define SP(x) (((x)&7)<<1) // convert motion vector component to offset for svf calc +#define SP(x) (((x) & 7) << 1) // convert motion vector component to offset + // for svf calc -#define IFMVCV(r,c,s,e) \ - if ( c >= minc && c <= maxc && r >= minr && r <= maxr) s else e; +#define IFMVCV(r, c, s, e) \ + if (c >= minc && c <= maxc && r >= minr && r <= maxr) \ + s \ + else \ + e; /* pointer to predictor base of a motionvector */ -#define PRE(r,c) (y + (((r)>>3) * y_stride + ((c)>>3) -(offset))) +#define PRE(r, c) (y + (((r) >> 3) * y_stride + ((c) >> 3) -(offset))) /* returns subpixel variance error function */ -#define DIST(r,c) vfp->svf( PRE(r,c), y_stride, SP(c),SP(r), z,b->src_stride,&sse) - -/* checks if (r,c) has better score than previous best */ -#define CHECK_BETTER(v,r,c) \ - IFMVCV(r,c,{thismse = (DIST(r,c)); if((v = MVC(r,c)+thismse) < besterr) \ - { besterr = v; br=r; bc=c; *distortion = thismse; *sse1 = sse; }}, v=INT_MAX;) +#define DIST(r, c) \ + vfp->svf(PRE(r, c), y_stride, SP(c), SP(r), z, b->src_stride, &sse) + +/* checks if (r, c) has better score than previous best */ +#define CHECK_BETTER(v, r, c) \ + IFMVCV(r, c, { \ + thismse = (DIST(r, c)); \ + if ((v = MVC(r, c) + thismse) < besterr) { \ + besterr = v; \ + br = r; \ + bc = c; \ + *distortion = thismse; \ + *sse1 = sse; \ + } \ + }, \ + v = INT_MAX;) #define MIN(x,y) (((x)<(y))?(x):(y)) #define MAX(x,y) (((x)>(y))?(x):(y)) @@ -307,17 +298,10 @@ int vp8_find_best_sub_pixel_step_iteratively(MACROBLOCK *x, BLOCK *b, BLOCKD *d, br = bestmv->as_mv.row << 3; bc = bestmv->as_mv.col << 3; hstep = 4; -#if CONFIG_NEWMVENTROPY minc = MAX(x->mv_col_min << 3, (ref_mv->as_mv.col) - ((1 << MV_MAX_BITS) - 1)); maxc = MIN(x->mv_col_max << 3, (ref_mv->as_mv.col) + ((1 << MV_MAX_BITS) - 1)); minr = MAX(x->mv_row_min << 3, (ref_mv->as_mv.row) - ((1 << MV_MAX_BITS) - 1)); maxr = MIN(x->mv_row_max << 3, (ref_mv->as_mv.row) + ((1 << MV_MAX_BITS) - 1)); -#else - minc = MAX(x->mv_col_min << 3, (ref_mv->as_mv.col) - ((1 << mvlong_width_hp) - 1)); - maxc = MIN(x->mv_col_max << 3, (ref_mv->as_mv.col) + ((1 << mvlong_width_hp) - 1)); - minr = MAX(x->mv_row_min << 3, (ref_mv->as_mv.row) - ((1 << mvlong_width_hp) - 1)); - maxr = MIN(x->mv_row_max << 3, (ref_mv->as_mv.row) + ((1 << mvlong_width_hp) - 1)); -#endif tr = br; tc = bc; @@ -403,13 +387,11 @@ int vp8_find_best_sub_pixel_step_iteratively(MACROBLOCK *x, BLOCK *b, BLOCKD *d, tc = bc; } -#if CONFIG_NEWMVENTROPY if (xd->allow_high_precision_mv) { usehp = vp8_use_nmv_hp(&ref_mv->as_mv); } else { usehp = 0; } -#endif if (usehp) { hstep >>= 1; @@ -771,13 +753,11 @@ int vp8_find_best_sub_pixel_step(MACROBLOCK *x, BLOCK *b, BLOCKD *d, *sse1 = sse; } -#if CONFIG_NEWMVENTROPY if (x->e_mbd.allow_high_precision_mv) { usehp = vp8_use_nmv_hp(&ref_mv->as_mv); } else { usehp = 0; } -#endif if (!usehp) return bestmse; @@ -1304,16 +1284,8 @@ int vp8_diamond_search_sad(MACROBLOCK *x, BLOCK *b, BLOCKD *d, MACROBLOCKD *xd = &x->e_mbd; int_mv fcenter_mv; -#if CONFIG_NEWMVENTROPY int *mvjsadcost = x->nmvjointsadcost; int *mvsadcost[2] = {x->nmvsadcost[0], x->nmvsadcost[1]}; -#else - int *mvsadcost[2] = {x->mvsadcost[0], x->mvsadcost[1]}; - if (xd->allow_high_precision_mv) { - mvsadcost[0] = x->mvsadcost_hp[0]; - mvsadcost[1] = x->mvsadcost_hp[1]; - } -#endif fcenter_mv.as_mv.row = center_mv->as_mv.row >> 3; fcenter_mv.as_mv.col = center_mv->as_mv.col >> 3; @@ -1423,16 +1395,8 @@ int vp8_diamond_search_sadx4(MACROBLOCK *x, BLOCK *b, BLOCKD *d, MACROBLOCKD *xd = &x->e_mbd; int_mv fcenter_mv; -#if CONFIG_NEWMVENTROPY int *mvjsadcost = x->nmvjointsadcost; int *mvsadcost[2] = {x->nmvsadcost[0], x->nmvsadcost[1]}; -#else - int *mvsadcost[2] = {x->mvsadcost[0], x->mvsadcost[1]}; - if (xd->allow_high_precision_mv) { - mvsadcost[0] = x->mvsadcost_hp[0]; - mvsadcost[1] = x->mvsadcost_hp[1]; - } -#endif fcenter_mv.as_mv.row = center_mv->as_mv.row >> 3; fcenter_mv.as_mv.col = center_mv->as_mv.col >> 3; @@ -1479,7 +1443,8 @@ int vp8_diamond_search_sadx4(MACROBLOCK *x, BLOCK *b, BLOCKD *d, for (t = 0; t < 4; t++) block_offset[t] = ss[i + t].offset + best_address; - fn_ptr->sdx4df(what, what_stride, block_offset, in_what_stride, sad_array); + fn_ptr->sdx4df(what, what_stride, block_offset, in_what_stride, + sad_array); for (t = 0; t < 4; t++, i++) { if (sad_array[t] < bestsad) { @@ -1631,16 +1596,8 @@ int vp8_full_search_sad(MACROBLOCK *x, BLOCK *b, BLOCKD *d, int_mv *ref_mv, int col_max = ref_col + distance; int_mv fcenter_mv; -#if CONFIG_NEWMVENTROPY int *mvjsadcost = x->nmvjointsadcost; int *mvsadcost[2] = {x->nmvsadcost[0], x->nmvsadcost[1]}; -#else - int *mvsadcost[2] = {x->mvsadcost[0], x->mvsadcost[1]}; - if (xd->allow_high_precision_mv) { - mvsadcost[0] = x->mvsadcost_hp[0]; - mvsadcost[1] = x->mvsadcost_hp[1]; - } -#endif fcenter_mv.as_mv.row = center_mv->as_mv.row >> 3; fcenter_mv.as_mv.col = center_mv->as_mv.col >> 3; @@ -1735,16 +1692,8 @@ int vp8_full_search_sadx3(MACROBLOCK *x, BLOCK *b, BLOCKD *d, int_mv *ref_mv, unsigned int sad_array[3]; int_mv fcenter_mv; -#if CONFIG_NEWMVENTROPY int *mvjsadcost = x->nmvjointsadcost; int *mvsadcost[2] = {x->nmvsadcost[0], x->nmvsadcost[1]}; -#else - int *mvsadcost[2] = {x->mvsadcost[0], x->mvsadcost[1]}; - if (xd->allow_high_precision_mv) { - mvsadcost[0] = x->mvsadcost_hp[0]; - mvsadcost[1] = x->mvsadcost_hp[1]; - } -#endif fcenter_mv.as_mv.row = center_mv->as_mv.row >> 3; fcenter_mv.as_mv.col = center_mv->as_mv.col >> 3; @@ -1872,16 +1821,8 @@ int vp8_full_search_sadx8(MACROBLOCK *x, BLOCK *b, BLOCKD *d, int_mv *ref_mv, unsigned int sad_array[3]; int_mv fcenter_mv; -#if CONFIG_NEWMVENTROPY int *mvjsadcost = x->nmvjointsadcost; int *mvsadcost[2] = {x->nmvsadcost[0], x->nmvsadcost[1]}; -#else - int *mvsadcost[2] = {x->mvsadcost[0], x->mvsadcost[1]}; - if (xd->allow_high_precision_mv) { - mvsadcost[0] = x->mvsadcost_hp[0]; - mvsadcost[1] = x->mvsadcost_hp[1]; - } -#endif fcenter_mv.as_mv.row = center_mv->as_mv.row >> 3; fcenter_mv.as_mv.col = center_mv->as_mv.col >> 3; @@ -2022,16 +1963,8 @@ int vp8_refining_search_sad(MACROBLOCK *x, BLOCK *b, BLOCKD *d, int_mv *ref_mv, MACROBLOCKD *xd = &x->e_mbd; int_mv fcenter_mv; -#if CONFIG_NEWMVENTROPY int *mvjsadcost = x->nmvjointsadcost; int *mvsadcost[2] = {x->nmvsadcost[0], x->nmvsadcost[1]}; -#else - int *mvsadcost[2] = {x->mvsadcost[0], x->mvsadcost[1]}; - if (xd->allow_high_precision_mv) { - mvsadcost[0] = x->mvsadcost_hp[0]; - mvsadcost[1] = x->mvsadcost_hp[1]; - } -#endif fcenter_mv.as_mv.row = center_mv->as_mv.row >> 3; fcenter_mv.as_mv.col = center_mv->as_mv.col >> 3; @@ -2106,16 +2039,8 @@ int vp8_refining_search_sadx4(MACROBLOCK *x, BLOCK *b, BLOCKD *d, MACROBLOCKD *xd = &x->e_mbd; int_mv fcenter_mv; -#if CONFIG_NEWMVENTROPY int *mvjsadcost = x->nmvjointsadcost; int *mvsadcost[2] = {x->nmvsadcost[0], x->nmvsadcost[1]}; -#else - int *mvsadcost[2] = {x->mvsadcost[0], x->mvsadcost[1]}; - if (xd->allow_high_precision_mv) { - mvsadcost[0] = x->mvsadcost_hp[0]; - mvsadcost[1] = x->mvsadcost_hp[1]; - } -#endif fcenter_mv.as_mv.row = center_mv->as_mv.row >> 3; fcenter_mv.as_mv.col = center_mv->as_mv.col >> 3; diff --git a/vp8/encoder/mcomp.h b/vp8/encoder/mcomp.h index afca58084..f09106927 100644 --- a/vp8/encoder/mcomp.h +++ b/vp8/encoder/mcomp.h @@ -15,21 +15,12 @@ #include "block.h" #include "variance.h" -#if CONFIG_NEWMVENTROPY #define MVCOSTS mvjcost, mvcost #define MVSADCOSTS mvjsadcost, mvsadcost #define DEC_MVCOSTS int *mvjcost, int *mvcost[2] #define DEC_MVSADCOSTS int *mvjsadcost, int *mvsadcost[2] #define NULLMVCOST NULL, NULL #define XMVCOST x->nmvjointcost, (x->e_mbd.allow_high_precision_mv?x->nmvcost_hp:x->nmvcost) -#else -#define MVCOSTS mvcost -#define MVSADCOSTS mvsadcost -#define DEC_MVCOSTS int *mvcost[2] -#define DEC_MVSADCOSTS int *mvsadcost[2] -#define NULLMVCOST NULL -#define XMVCOST (x->e_mbd.allow_high_precision_mv?x->mvcost_hp:x->mvcost) -#endif /* CONFIG_NEWMVENTROPY */ #ifdef ENTROPY_STATS extern void init_mv_ref_counts(); diff --git a/vp8/encoder/modecosts.c b/vp8/encoder/modecosts.c index b1abd1e2a..23b9973c3 100644 --- a/vp8/encoder/modecosts.c +++ b/vp8/encoder/modecosts.c @@ -46,14 +46,11 @@ void vp8_init_mode_costs(VP8_COMP *c) { vp8_cost_tokens(c->mb.i8x8_mode_costs, x->fc.i8x8_mode_prob, vp8_i8x8_mode_tree); -#if CONFIG_SWITCHABLE_INTERP { int i; for (i = 0; i <= VP8_SWITCHABLE_FILTERS; ++i) - //for (i = 0; i <= 0; ++i) vp8_cost_tokens((int *)c->mb.switchable_interp_costs[i], x->fc.switchable_interp_prob[i], vp8_switchable_interp_tree); } -#endif } diff --git a/vp8/encoder/onyx_if.c b/vp8/encoder/onyx_if.c index 14e9e784a..f11ff5936 100644 --- a/vp8/encoder/onyx_if.c +++ b/vp8/encoder/onyx_if.c @@ -77,7 +77,7 @@ extern void vp8_yv12_copy_frame_func_neon(YV12_BUFFER_CONFIG *src_ybc, YV12_BUFF extern void vp8_yv12_copy_src_frame_func_neon(YV12_BUFFER_CONFIG *src_ybc, YV12_BUFFER_CONFIG *dst_ybc); #endif -int vp8_calc_ss_err(YV12_BUFFER_CONFIG *source, YV12_BUFFER_CONFIG *dest, const vp8_variance_rtcd_vtable_t *rtcd); +int vp8_calc_ss_err(YV12_BUFFER_CONFIG *source, YV12_BUFFER_CONFIG *dest); extern void vp8_temporal_filter_prepare_c(VP8_COMP *cpi, int distance); @@ -85,6 +85,7 @@ static void set_default_lf_deltas(VP8_COMP *cpi); extern const int vp8_gf_interval_table[101]; +#define DEFAULT_INTERP_FILTER EIGHTTAP /* SWITCHABLE for better performance */ #define SEARCH_BEST_FILTER 0 /* to search exhaustively for best filter */ #define RESET_FOREACH_FILTER 0 /* whether to reset the encoder state @@ -101,25 +102,14 @@ extern const int vp8_gf_interval_table[101]; #if CONFIG_INTERNAL_STATS #include "math.h" -extern double vp8_calc_ssim -( - YV12_BUFFER_CONFIG *source, - YV12_BUFFER_CONFIG *dest, - int lumamask, - double *weight, - const vp8_variance_rtcd_vtable_t *rtcd -); +extern double vp8_calc_ssim(YV12_BUFFER_CONFIG *source, + YV12_BUFFER_CONFIG *dest, int lumamask, + double *weight); -extern double vp8_calc_ssimg -( - YV12_BUFFER_CONFIG *source, - YV12_BUFFER_CONFIG *dest, - double *ssim_y, - double *ssim_u, - double *ssim_v, - const vp8_variance_rtcd_vtable_t *rtcd -); +extern double vp8_calc_ssimg(YV12_BUFFER_CONFIG *source, + YV12_BUFFER_CONFIG *dest, double *ssim_y, + double *ssim_u, double *ssim_v); #endif @@ -149,12 +139,10 @@ extern int skip_false_count; extern int intra_mode_stats[VP8_BINTRAMODES][VP8_BINTRAMODES][VP8_BINTRAMODES]; #endif -#if CONFIG_NEWMVENTROPY #ifdef NMV_STATS extern void init_nmvstats(); extern void print_nmvstats(); #endif -#endif #ifdef SPEEDSTATS unsigned int frames_at_speed[16] = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0}; @@ -1630,7 +1618,7 @@ void vp8_change_config(VP8_PTR ptr, VP8_CONFIG *oxcf) { cpi->cq_target_quality = cpi->oxcf.cq_level; if (!cm->use_bilinear_mc_filter) - cm->mcomp_filter_type = EIGHTTAP; + cm->mcomp_filter_type = DEFAULT_INTERP_FILTER; else cm->mcomp_filter_type = BILINEAR; @@ -1700,8 +1688,6 @@ void vp8_change_config(VP8_PTR ptr, VP8_CONFIG *oxcf) { #define M_LOG2_E 0.693147180559945309417 #define log2f(x) (log (x) / (float) M_LOG2_E) -#if CONFIG_NEWMVENTROPY - static void cal_nmvjointsadcost(int *mvjointsadcost) { mvjointsadcost[0] = 600; mvjointsadcost[1] = 300; @@ -1739,40 +1725,6 @@ static void cal_nmvsadcosts_hp(int *mvsadcost[2]) { } while (++i <= MV_MAX); } -#else - -static void cal_mvsadcosts(int *mvsadcost[2]) { - int i = 1; - - mvsadcost [0] [0] = 300; - mvsadcost [1] [0] = 300; - - do { - double z = 256 * (2 * (log2f(8 * i) + .6)); - mvsadcost [0][i] = (int) z; - mvsadcost [1][i] = (int) z; - mvsadcost [0][-i] = (int) z; - mvsadcost [1][-i] = (int) z; - } while (++i <= mvfp_max); -} - -static void cal_mvsadcosts_hp(int *mvsadcost[2]) { - int i = 1; - - mvsadcost [0] [0] = 300; - mvsadcost [1] [0] = 300; - - do { - double z = 256 * (2 * (log2f(8 * i) + .6)); - mvsadcost [0][i] = (int) z; - mvsadcost [1][i] = (int) z; - mvsadcost [0][-i] = (int) z; - mvsadcost [1][-i] = (int) z; - } while (++i <= mvfp_max_hp); -} - -#endif /* CONFIG_NEWMVENTROPY */ - VP8_PTR vp8_create_compressor(VP8_CONFIG *oxcf) { int i; volatile union { @@ -1824,10 +1776,8 @@ VP8_PTR vp8_create_compressor(VP8_CONFIG *oxcf) { #endif for (i = 0; i < COMP_PRED_CONTEXTS; i++) cm->prob_comppred[i] = 128; -#if CONFIG_TX_SELECT for (i = 0; i < TX_SIZE_MAX - 1; i++) cm->prob_tx[i] = 128; -#endif // Prime the recent reference frame useage counters. // Hereafter they will be maintained as a sort of moving average @@ -1888,11 +1838,9 @@ VP8_PTR vp8_create_compressor(VP8_CONFIG *oxcf) { vp8_zero(inter_uv_modes); vp8_zero(inter_b_modes); #endif -#if CONFIG_NEWMVENTROPY #ifdef NMV_STATS init_nmvstats(); #endif -#endif /*Initialize the feed-forward activity masking.*/ cpi->activity_avg = 90 << 12; @@ -1958,7 +1906,6 @@ VP8_PTR vp8_create_compressor(VP8_CONFIG *oxcf) { cpi->gf_rate_correction_factor = 1.0; cpi->twopass.est_max_qcorrection_factor = 1.0; -#if CONFIG_NEWMVENTROPY cal_nmvjointsadcost(cpi->mb.nmvjointsadcost); cpi->mb.nmvcost[0] = &cpi->mb.nmvcosts[0][MV_MAX]; cpi->mb.nmvcost[1] = &cpi->mb.nmvcosts[1][MV_MAX]; @@ -1971,19 +1918,6 @@ VP8_PTR vp8_create_compressor(VP8_CONFIG *oxcf) { cpi->mb.nmvsadcost_hp[0] = &cpi->mb.nmvsadcosts_hp[0][MV_MAX]; cpi->mb.nmvsadcost_hp[1] = &cpi->mb.nmvsadcosts_hp[1][MV_MAX]; cal_nmvsadcosts_hp(cpi->mb.nmvsadcost_hp); -#else - cpi->mb.mvcost[0] = &cpi->mb.mvcosts[0][mv_max + 1]; - cpi->mb.mvcost[1] = &cpi->mb.mvcosts[1][mv_max + 1]; - cpi->mb.mvsadcost[0] = &cpi->mb.mvsadcosts[0][mvfp_max + 1]; - cpi->mb.mvsadcost[1] = &cpi->mb.mvsadcosts[1][mvfp_max + 1]; - cal_mvsadcosts(cpi->mb.mvsadcost); - - cpi->mb.mvcost_hp[0] = &cpi->mb.mvcosts_hp[0][mv_max_hp + 1]; - cpi->mb.mvcost_hp[1] = &cpi->mb.mvcosts_hp[1][mv_max_hp + 1]; - cpi->mb.mvsadcost_hp[0] = &cpi->mb.mvsadcosts_hp[0][mvfp_max_hp + 1]; - cpi->mb.mvsadcost_hp[1] = &cpi->mb.mvsadcosts_hp[1][mvfp_max_hp + 1]; - cal_mvsadcosts_hp(cpi->mb.mvsadcost_hp); -#endif /* CONFIG_NEWMVENTROPY */ for (i = 0; i < KEY_FRAME_CONTEXT; i++) { cpi->prior_key_frame_distance[i] = (int)cpi->output_frame_rate; @@ -2027,74 +1961,48 @@ VP8_PTR vp8_create_compressor(VP8_CONFIG *oxcf) { init_mv_ref_counts(); #endif +#define BFP(BT, SDF, VF, SVF, SVFHH, SVFHV, SVFHHV, SDX3F, SDX8F, SDX4DF) \ + cpi->fn_ptr[BT].sdf = SDF; \ + cpi->fn_ptr[BT].vf = VF; \ + cpi->fn_ptr[BT].svf = SVF; \ + cpi->fn_ptr[BT].svf_halfpix_h = SVFHH; \ + cpi->fn_ptr[BT].svf_halfpix_v = SVFHV; \ + cpi->fn_ptr[BT].svf_halfpix_hv = SVFHHV; \ + cpi->fn_ptr[BT].sdx3f = SDX3F; \ + cpi->fn_ptr[BT].sdx8f = SDX8F; \ + cpi->fn_ptr[BT].sdx4df = SDX4DF; + + #if CONFIG_SUPERBLOCKS - cpi->fn_ptr[BLOCK_32X32].sdf = VARIANCE_INVOKE(&cpi->rtcd.variance, sad32x32); - cpi->fn_ptr[BLOCK_32X32].vf = VARIANCE_INVOKE(&cpi->rtcd.variance, var32x32); - cpi->fn_ptr[BLOCK_32X32].svf = VARIANCE_INVOKE(&cpi->rtcd.variance, subpixvar32x32); - cpi->fn_ptr[BLOCK_32X32].svf_halfpix_h = VARIANCE_INVOKE(&cpi->rtcd.variance, halfpixvar32x32_h); - cpi->fn_ptr[BLOCK_32X32].svf_halfpix_v = VARIANCE_INVOKE(&cpi->rtcd.variance, halfpixvar32x32_v); - cpi->fn_ptr[BLOCK_32X32].svf_halfpix_hv = VARIANCE_INVOKE(&cpi->rtcd.variance, halfpixvar32x32_hv); - cpi->fn_ptr[BLOCK_32X32].sdx3f = VARIANCE_INVOKE(&cpi->rtcd.variance, sad32x32x3); - cpi->fn_ptr[BLOCK_32X32].sdx8f = VARIANCE_INVOKE(&cpi->rtcd.variance, sad32x32x8); - cpi->fn_ptr[BLOCK_32X32].sdx4df = VARIANCE_INVOKE(&cpi->rtcd.variance, sad32x32x4d); + BFP(BLOCK_32X32, vp8_sad32x32, vp8_variance32x32, vp8_sub_pixel_variance32x32, + vp8_variance_halfpixvar32x32_h, vp8_variance_halfpixvar32x32_v, + vp8_variance_halfpixvar32x32_hv, vp8_sad32x32x3, vp8_sad32x32x8, + vp8_sad32x32x4d) #endif - cpi->fn_ptr[BLOCK_16X16].sdf = VARIANCE_INVOKE(&cpi->rtcd.variance, sad16x16); - cpi->fn_ptr[BLOCK_16X16].vf = VARIANCE_INVOKE(&cpi->rtcd.variance, var16x16); - cpi->fn_ptr[BLOCK_16X16].svf = VARIANCE_INVOKE(&cpi->rtcd.variance, subpixvar16x16); - cpi->fn_ptr[BLOCK_16X16].svf_halfpix_h = VARIANCE_INVOKE(&cpi->rtcd.variance, halfpixvar16x16_h); - cpi->fn_ptr[BLOCK_16X16].svf_halfpix_v = VARIANCE_INVOKE(&cpi->rtcd.variance, halfpixvar16x16_v); - cpi->fn_ptr[BLOCK_16X16].svf_halfpix_hv = VARIANCE_INVOKE(&cpi->rtcd.variance, halfpixvar16x16_hv); - cpi->fn_ptr[BLOCK_16X16].sdx3f = VARIANCE_INVOKE(&cpi->rtcd.variance, sad16x16x3); - cpi->fn_ptr[BLOCK_16X16].sdx8f = VARIANCE_INVOKE(&cpi->rtcd.variance, sad16x16x8); - cpi->fn_ptr[BLOCK_16X16].sdx4df = VARIANCE_INVOKE(&cpi->rtcd.variance, sad16x16x4d); - - cpi->fn_ptr[BLOCK_16X8].sdf = VARIANCE_INVOKE(&cpi->rtcd.variance, sad16x8); - cpi->fn_ptr[BLOCK_16X8].vf = VARIANCE_INVOKE(&cpi->rtcd.variance, var16x8); - cpi->fn_ptr[BLOCK_16X8].svf = VARIANCE_INVOKE(&cpi->rtcd.variance, subpixvar16x8); - cpi->fn_ptr[BLOCK_16X8].svf_halfpix_h = NULL; - cpi->fn_ptr[BLOCK_16X8].svf_halfpix_v = NULL; - cpi->fn_ptr[BLOCK_16X8].svf_halfpix_hv = NULL; - cpi->fn_ptr[BLOCK_16X8].sdx3f = VARIANCE_INVOKE(&cpi->rtcd.variance, sad16x8x3); - cpi->fn_ptr[BLOCK_16X8].sdx8f = VARIANCE_INVOKE(&cpi->rtcd.variance, sad16x8x8); - cpi->fn_ptr[BLOCK_16X8].sdx4df = VARIANCE_INVOKE(&cpi->rtcd.variance, sad16x8x4d); - - cpi->fn_ptr[BLOCK_8X16].sdf = VARIANCE_INVOKE(&cpi->rtcd.variance, sad8x16); - cpi->fn_ptr[BLOCK_8X16].vf = VARIANCE_INVOKE(&cpi->rtcd.variance, var8x16); - cpi->fn_ptr[BLOCK_8X16].svf = VARIANCE_INVOKE(&cpi->rtcd.variance, subpixvar8x16); - cpi->fn_ptr[BLOCK_8X16].svf_halfpix_h = NULL; - cpi->fn_ptr[BLOCK_8X16].svf_halfpix_v = NULL; - cpi->fn_ptr[BLOCK_8X16].svf_halfpix_hv = NULL; - cpi->fn_ptr[BLOCK_8X16].sdx3f = VARIANCE_INVOKE(&cpi->rtcd.variance, sad8x16x3); - cpi->fn_ptr[BLOCK_8X16].sdx8f = VARIANCE_INVOKE(&cpi->rtcd.variance, sad8x16x8); - cpi->fn_ptr[BLOCK_8X16].sdx4df = VARIANCE_INVOKE(&cpi->rtcd.variance, sad8x16x4d); - - cpi->fn_ptr[BLOCK_8X8].sdf = VARIANCE_INVOKE(&cpi->rtcd.variance, sad8x8); - cpi->fn_ptr[BLOCK_8X8].vf = VARIANCE_INVOKE(&cpi->rtcd.variance, var8x8); - cpi->fn_ptr[BLOCK_8X8].svf = VARIANCE_INVOKE(&cpi->rtcd.variance, subpixvar8x8); - cpi->fn_ptr[BLOCK_8X8].svf_halfpix_h = NULL; - cpi->fn_ptr[BLOCK_8X8].svf_halfpix_v = NULL; - cpi->fn_ptr[BLOCK_8X8].svf_halfpix_hv = NULL; - cpi->fn_ptr[BLOCK_8X8].sdx3f = VARIANCE_INVOKE(&cpi->rtcd.variance, sad8x8x3); - cpi->fn_ptr[BLOCK_8X8].sdx8f = VARIANCE_INVOKE(&cpi->rtcd.variance, sad8x8x8); - cpi->fn_ptr[BLOCK_8X8].sdx4df = VARIANCE_INVOKE(&cpi->rtcd.variance, sad8x8x4d); - - cpi->fn_ptr[BLOCK_4X4].sdf = VARIANCE_INVOKE(&cpi->rtcd.variance, sad4x4); - cpi->fn_ptr[BLOCK_4X4].vf = VARIANCE_INVOKE(&cpi->rtcd.variance, var4x4); - cpi->fn_ptr[BLOCK_4X4].svf = VARIANCE_INVOKE(&cpi->rtcd.variance, subpixvar4x4); - cpi->fn_ptr[BLOCK_4X4].svf_halfpix_h = NULL; - cpi->fn_ptr[BLOCK_4X4].svf_halfpix_v = NULL; - cpi->fn_ptr[BLOCK_4X4].svf_halfpix_hv = NULL; - cpi->fn_ptr[BLOCK_4X4].sdx3f = VARIANCE_INVOKE(&cpi->rtcd.variance, sad4x4x3); - cpi->fn_ptr[BLOCK_4X4].sdx8f = VARIANCE_INVOKE(&cpi->rtcd.variance, sad4x4x8); - cpi->fn_ptr[BLOCK_4X4].sdx4df = VARIANCE_INVOKE(&cpi->rtcd.variance, sad4x4x4d); + BFP(BLOCK_16X16, vp8_sad16x16, vp8_variance16x16, vp8_sub_pixel_variance16x16, + vp8_variance_halfpixvar16x16_h, vp8_variance_halfpixvar16x16_v, + vp8_variance_halfpixvar16x16_hv, vp8_sad16x16x3, vp8_sad16x16x8, + vp8_sad16x16x4d) + + BFP(BLOCK_16X8, vp8_sad16x8, vp8_variance16x8, vp8_sub_pixel_variance16x8, + NULL, NULL, NULL, vp8_sad16x8x3, vp8_sad16x8x8, vp8_sad16x8x4d) + + BFP(BLOCK_8X16, vp8_sad8x16, vp8_variance8x16, vp8_sub_pixel_variance8x16, + NULL, NULL, NULL, vp8_sad8x16x3, vp8_sad8x16x8, vp8_sad8x16x4d) + + BFP(BLOCK_8X8, vp8_sad8x8, vp8_variance8x8, vp8_sub_pixel_variance8x8, + NULL, NULL, NULL, vp8_sad8x8x3, vp8_sad8x8x8, vp8_sad8x8x4d) + + BFP(BLOCK_4X4, vp8_sad4x4, vp8_variance4x4, vp8_sub_pixel_variance4x4, + NULL, NULL, NULL, vp8_sad4x4x3, vp8_sad4x4x8, vp8_sad4x4x4d) #if ARCH_X86 || ARCH_X86_64 - cpi->fn_ptr[BLOCK_16X16].copymem = VARIANCE_INVOKE(&cpi->rtcd.variance, copy32xn); - cpi->fn_ptr[BLOCK_16X8].copymem = VARIANCE_INVOKE(&cpi->rtcd.variance, copy32xn); - cpi->fn_ptr[BLOCK_8X16].copymem = VARIANCE_INVOKE(&cpi->rtcd.variance, copy32xn); - cpi->fn_ptr[BLOCK_8X8].copymem = VARIANCE_INVOKE(&cpi->rtcd.variance, copy32xn); - cpi->fn_ptr[BLOCK_4X4].copymem = VARIANCE_INVOKE(&cpi->rtcd.variance, copy32xn); + cpi->fn_ptr[BLOCK_16X16].copymem = vp8_copy32xn; + cpi->fn_ptr[BLOCK_16X8].copymem = vp8_copy32xn; + cpi->fn_ptr[BLOCK_8X16].copymem = vp8_copy32xn; + cpi->fn_ptr[BLOCK_8X8].copymem = vp8_copy32xn; + cpi->fn_ptr[BLOCK_4X4].copymem = vp8_copy32xn; #endif cpi->full_search_sad = SEARCH_INVOKE(&cpi->rtcd.search, full_search); @@ -2136,12 +2044,10 @@ void vp8_remove_compressor(VP8_PTR *ptr) { print_mode_context(); } #endif -#if CONFIG_NEWMVENTROPY #ifdef NMV_STATS if (cpi->pass != 1) print_nmvstats(); #endif -#endif #if CONFIG_INTERNAL_STATS @@ -2370,8 +2276,7 @@ void vp8_remove_compressor(VP8_PTR *ptr) { static uint64_t calc_plane_error(unsigned char *orig, int orig_stride, unsigned char *recon, int recon_stride, - unsigned int cols, unsigned int rows, - vp8_variance_rtcd_vtable_t *rtcd) { + unsigned int cols, unsigned int rows) { unsigned int row, col; uint64_t total_sse = 0; int diff; @@ -2380,9 +2285,7 @@ static uint64_t calc_plane_error(unsigned char *orig, int orig_stride, for (col = 0; col + 16 <= cols; col += 16) { unsigned int sse; - VARIANCE_INVOKE(rtcd, mse16x16)(orig + col, orig_stride, - recon + col, recon_stride, - &sse); + vp8_mse16x16(orig + col, orig_stride, recon + col, recon_stride, &sse); total_sse += sse; } @@ -2434,8 +2337,7 @@ static void generate_psnr_packet(VP8_COMP *cpi) { pkt.kind = VPX_CODEC_PSNR_PKT; sse = calc_plane_error(orig->y_buffer, orig->y_stride, recon->y_buffer, recon->y_stride, - width, height, - IF_RTCD(&cpi->rtcd.variance)); + width, height); pkt.data.psnr.sse[0] = sse; pkt.data.psnr.sse[1] = sse; pkt.data.psnr.samples[0] = width * height; @@ -2446,8 +2348,7 @@ static void generate_psnr_packet(VP8_COMP *cpi) { sse = calc_plane_error(orig->u_buffer, orig->uv_stride, recon->u_buffer, recon->uv_stride, - width, height, - IF_RTCD(&cpi->rtcd.variance)); + width, height); pkt.data.psnr.sse[0] += sse; pkt.data.psnr.sse[2] = sse; pkt.data.psnr.samples[0] += width * height; @@ -2455,8 +2356,7 @@ static void generate_psnr_packet(VP8_COMP *cpi) { sse = calc_plane_error(orig->v_buffer, orig->uv_stride, recon->v_buffer, recon->uv_stride, - width, height, - IF_RTCD(&cpi->rtcd.variance)); + width, height); pkt.data.psnr.sse[0] += sse; pkt.data.psnr.sse[3] = sse; pkt.data.psnr.samples[0] += width * height; @@ -3034,13 +2934,10 @@ static void encode_frame_to_data_rate /* list of filters to search over */ int mcomp_filters_to_search[] = { -#if CONFIG_SWITCHABLE_INTERP EIGHTTAP, EIGHTTAP_SHARP, SIXTAP, SWITCHABLE -#else - EIGHTTAP, EIGHTTAP_SHARP, SIXTAP, -#endif }; - int mcomp_filters = sizeof(mcomp_filters_to_search) / sizeof(*mcomp_filters_to_search); + int mcomp_filters = sizeof(mcomp_filters_to_search) / + sizeof(*mcomp_filters_to_search); int mcomp_filter_index = 0; INT64 mcomp_filter_cost[4]; @@ -3265,12 +3162,7 @@ static void encode_frame_to_data_rate cm->mcomp_filter_type = mcomp_filters_to_search[0]; mcomp_filter_index = 0; } else { -#if CONFIG_SWITCHABLE_INTERP - cm->mcomp_filter_type = SWITCHABLE; -#else - cm->mcomp_filter_type = - (Q < SHARP_FILTER_QTHRESH ? EIGHTTAP_SHARP : EIGHTTAP); -#endif + cm->mcomp_filter_type = DEFAULT_INTERP_FILTER; } /* TODO: Decide this more intelligently */ xd->allow_high_precision_mv = (Q < HIGH_PRECISION_MV_QTHRESH); @@ -3428,8 +3320,7 @@ static void encode_frame_to_data_rate if ((cm->frame_type == KEY_FRAME) && cpi->this_key_frame_forced) { int last_q = Q; int kf_err = vp8_calc_ss_err(cpi->Source, - &cm->yv12_fb[cm->new_fb_idx], - IF_RTCD(&cpi->rtcd.variance)); + &cm->yv12_fb[cm->new_fb_idx]); int high_err_target = cpi->ambient_err; int low_err_target = (cpi->ambient_err >> 1); @@ -3584,7 +3475,6 @@ static void encode_frame_to_data_rate if (cpi->is_src_frame_alt_ref) Loop = FALSE; -#if CONFIG_SWITCHABLE_INTERP if (cm->frame_type != KEY_FRAME && !sf->search_best_filter && cm->mcomp_filter_type == SWITCHABLE) { @@ -3610,19 +3500,16 @@ static void encode_frame_to_data_rate if (count[i]) { cm->mcomp_filter_type = vp8_switchable_interp[i]; Loop = TRUE; /* Make sure to loop since the filter changed */ - //loop_count = -1; break; } } } } -#endif if (Loop == FALSE && cm->frame_type != KEY_FRAME && sf->search_best_filter) { if (mcomp_filter_index < mcomp_filters) { INT64 err = vp8_calc_ss_err(cpi->Source, - &cm->yv12_fb[cm->new_fb_idx], - IF_RTCD(&cpi->rtcd.variance)); + &cm->yv12_fb[cm->new_fb_idx]); INT64 rate = cpi->projected_frame_size << 8; mcomp_filter_cost[mcomp_filter_index] = (RDCOST(cpi->RDMULT, cpi->RDDIV, rate, err)); @@ -3684,8 +3571,7 @@ static void encode_frame_to_data_rate // the force key frame if (cpi->next_key_frame_forced && (cpi->twopass.frames_to_key == 0)) { cpi->ambient_err = vp8_calc_ss_err(cpi->Source, - &cm->yv12_fb[cm->new_fb_idx], - IF_RTCD(&cpi->rtcd.variance)); + &cm->yv12_fb[cm->new_fb_idx]); } // This frame's MVs are saved and will be used in next frame's MV @@ -3758,18 +3644,12 @@ static void encode_frame_to_data_rate update_reference_frames(cm); vp8_copy(cpi->common.fc.coef_counts, cpi->coef_counts); -#if CONFIG_HYBRIDTRANSFORM vp8_copy(cpi->common.fc.hybrid_coef_counts, cpi->hybrid_coef_counts); -#endif vp8_copy(cpi->common.fc.coef_counts_8x8, cpi->coef_counts_8x8); -#if CONFIG_HYBRIDTRANSFORM8X8 vp8_copy(cpi->common.fc.hybrid_coef_counts_8x8, cpi->hybrid_coef_counts_8x8); -#endif vp8_copy(cpi->common.fc.coef_counts_16x16, cpi->coef_counts_16x16); -#if CONFIG_HYBRIDTRANSFORM16X16 vp8_copy(cpi->common.fc.hybrid_coef_counts_16x16, cpi->hybrid_coef_counts_16x16); -#endif vp8_adapt_coef_probs(&cpi->common); if (cpi->common.frame_type != KEY_FRAME) { vp8_copy(cpi->common.fc.ymode_counts, cpi->ymode_count); @@ -3780,14 +3660,8 @@ static void encode_frame_to_data_rate vp8_copy(cpi->common.fc.mbsplit_counts, cpi->mbsplit_count); vp8_adapt_mode_probs(&cpi->common); -#if CONFIG_NEWMVENTROPY cpi->common.fc.NMVcount = cpi->NMVcount; vp8_adapt_nmv_probs(&cpi->common, cpi->mb.e_mbd.allow_high_precision_mv); -#else - vp8_copy(cpi->common.fc.MVcount, cpi->MVcount); - vp8_copy(cpi->common.fc.MVcount_hp, cpi->MVcount_hp); - vp8_adapt_mv_probs(&cpi->common); -#endif /* CONFIG_NEWMVENTROPY */ vp8_update_mode_context(&cpi->common); } @@ -3903,8 +3777,7 @@ static void encode_frame_to_data_rate vp8_clear_system_state(); // __asm emms; recon_err = vp8_calc_ss_err(cpi->Source, - &cm->yv12_fb[cm->new_fb_idx], - IF_RTCD(&cpi->rtcd.variance)); + &cm->yv12_fb[cm->new_fb_idx]); if (cpi->twopass.total_left_stats->coded_error != 0.0) fprintf(f, "%10d %10d %10d %10d %10d %10d %10d %10d" @@ -4390,16 +4263,16 @@ int vp8_get_compressed_data(VP8_PTR ptr, unsigned int *frame_flags, unsigned lon int64_t sq_error; ye = calc_plane_error(orig->y_buffer, orig->y_stride, - recon->y_buffer, recon->y_stride, orig->y_width, orig->y_height, - IF_RTCD(&cpi->rtcd.variance)); + recon->y_buffer, recon->y_stride, orig->y_width, + orig->y_height); ue = calc_plane_error(orig->u_buffer, orig->uv_stride, - recon->u_buffer, recon->uv_stride, orig->uv_width, orig->uv_height, - IF_RTCD(&cpi->rtcd.variance)); + recon->u_buffer, recon->uv_stride, orig->uv_width, + orig->uv_height); ve = calc_plane_error(orig->v_buffer, orig->uv_stride, - recon->v_buffer, recon->uv_stride, orig->uv_width, orig->uv_height, - IF_RTCD(&cpi->rtcd.variance)); + recon->v_buffer, recon->uv_stride, orig->uv_width, + orig->uv_height); sq_error = ye + ue + ve; @@ -4419,16 +4292,16 @@ int vp8_get_compressed_data(VP8_PTR ptr, unsigned int *frame_flags, unsigned lon vp8_clear_system_state(); ye = calc_plane_error(orig->y_buffer, orig->y_stride, - pp->y_buffer, pp->y_stride, orig->y_width, orig->y_height, - IF_RTCD(&cpi->rtcd.variance)); + pp->y_buffer, pp->y_stride, orig->y_width, + orig->y_height); ue = calc_plane_error(orig->u_buffer, orig->uv_stride, - pp->u_buffer, pp->uv_stride, orig->uv_width, orig->uv_height, - IF_RTCD(&cpi->rtcd.variance)); + pp->u_buffer, pp->uv_stride, orig->uv_width, + orig->uv_height); ve = calc_plane_error(orig->v_buffer, orig->uv_stride, - pp->v_buffer, pp->uv_stride, orig->uv_width, orig->uv_height, - IF_RTCD(&cpi->rtcd.variance)); + pp->v_buffer, pp->uv_stride, orig->uv_width, + orig->uv_height); sq_error = ye + ue + ve; @@ -4441,8 +4314,7 @@ int vp8_get_compressed_data(VP8_PTR ptr, unsigned int *frame_flags, unsigned lon cpi->totalp += frame_psnr2; frame_ssim2 = vp8_calc_ssim(cpi->Source, - &cm->post_proc_buffer, 1, &weight, - IF_RTCD(&cpi->rtcd.variance)); + &cm->post_proc_buffer, 1, &weight); cpi->summed_quality += frame_ssim2 * weight; cpi->summed_weights += weight; @@ -4461,7 +4333,7 @@ int vp8_get_compressed_data(VP8_PTR ptr, unsigned int *frame_flags, unsigned lon if (cpi->b_calculate_ssimg) { double y, u, v, frame_all; frame_all = vp8_calc_ssimg(cpi->Source, cm->frame_to_show, - &y, &u, &v, IF_RTCD(&cpi->rtcd.variance)); + &y, &u, &v); cpi->total_ssimg_y += y; cpi->total_ssimg_u += u; cpi->total_ssimg_v += v; @@ -4604,19 +4476,19 @@ int vp8_set_internal_size(VP8_PTR comp, VPX_SCALING horiz_mode, VPX_SCALING vert -int vp8_calc_ss_err(YV12_BUFFER_CONFIG *source, YV12_BUFFER_CONFIG *dest, const vp8_variance_rtcd_vtable_t *rtcd) { +int vp8_calc_ss_err(YV12_BUFFER_CONFIG *source, YV12_BUFFER_CONFIG *dest) { int i, j; int Total = 0; unsigned char *src = source->y_buffer; unsigned char *dst = dest->y_buffer; - (void)rtcd; // Loop through the Y plane raw and reconstruction data summing (square differences) for (i = 0; i < source->y_height; i += 16) { for (j = 0; j < source->y_width; j += 16) { unsigned int sse; - Total += VARIANCE_INVOKE(rtcd, mse16x16)(src + j, source->y_stride, dst + j, dest->y_stride, &sse); + Total += vp8_mse16x16(src + j, source->y_stride, dst + j, dest->y_stride, + &sse); } src += 16 * source->y_stride; diff --git a/vp8/encoder/onyx_int.h b/vp8/encoder/onyx_int.h index 01151280c..ab6802509 100644 --- a/vp8/encoder/onyx_int.h +++ b/vp8/encoder/onyx_int.h @@ -60,17 +60,10 @@ #define VP8_TEMPORAL_ALT_REF 1 typedef struct { -#if CONFIG_NEWMVENTROPY nmv_context nmvc; int nmvjointcost[MV_JOINTS]; int nmvcosts[2][MV_VALS]; int nmvcosts_hp[2][MV_VALS]; -#else - MV_CONTEXT mvc[2]; - int mvcosts[2][MVvals + 1]; - MV_CONTEXT_HP mvc_hp[2]; - int mvcosts_hp[2][MVvals_hp + 1]; -#endif #ifdef MODE_STATS // Stats @@ -97,24 +90,18 @@ typedef struct { vp8_prob coef_probs[BLOCK_TYPES] [COEF_BANDS][PREV_COEF_CONTEXTS][ENTROPY_NODES]; -#if CONFIG_HYBRIDTRANSFORM vp8_prob hybrid_coef_probs[BLOCK_TYPES] [COEF_BANDS][PREV_COEF_CONTEXTS][ENTROPY_NODES]; -#endif vp8_prob coef_probs_8x8[BLOCK_TYPES_8X8] [COEF_BANDS][PREV_COEF_CONTEXTS][ENTROPY_NODES]; -#if CONFIG_HYBRIDTRANSFORM8X8 vp8_prob hybrid_coef_probs_8x8[BLOCK_TYPES_8X8] [COEF_BANDS][PREV_COEF_CONTEXTS][ENTROPY_NODES]; -#endif vp8_prob coef_probs_16x16[BLOCK_TYPES_16X16] [COEF_BANDS][PREV_COEF_CONTEXTS][ENTROPY_NODES]; -#if CONFIG_HYBRIDTRANSFORM16X16 vp8_prob hybrid_coef_probs_16x16[BLOCK_TYPES_16X16] [COEF_BANDS][PREV_COEF_CONTEXTS][ENTROPY_NODES]; -#endif vp8_prob ymode_prob [VP8_YMODES - 1]; /* interframe intra mode probs */ vp8_prob uv_mode_prob [VP8_YMODES][VP8_UV_MODES - 1]; @@ -123,10 +110,8 @@ typedef struct { vp8_prob sub_mv_ref_prob [SUBMVREF_COUNT][VP8_SUBMVREFS - 1]; vp8_prob mbsplit_prob [VP8_NUMMBSPLITS - 1]; -#if CONFIG_SWITCHABLE_INTERP vp8_prob switchable_interp_prob[VP8_SWITCHABLE_FILTERS + 1] [VP8_SWITCHABLE_FILTERS - 1]; -#endif int mv_ref_ct[6][4][2]; int mode_context[6][4]; @@ -365,7 +350,6 @@ typedef struct { typedef struct VP8_ENCODER_RTCD { VP8_COMMON_RTCD *common; - vp8_variance_rtcd_vtable_t variance; vp8_fdct_rtcd_vtable_t fdct; vp8_encodemb_rtcd_vtable_t encodemb; vp8_search_rtcd_vtable_t search; @@ -373,10 +357,10 @@ typedef struct VP8_ENCODER_RTCD { } VP8_ENCODER_RTCD; enum { - BLOCK_16X8, - BLOCK_8X16, - BLOCK_8X8, - BLOCK_4X4, + BLOCK_16X8 = PARTITIONING_16X8, + BLOCK_8X16 = PARTITIONING_8X16, + BLOCK_8X8 = PARTITIONING_8X8, + BLOCK_4X4 = PARTITIONING_4X4, BLOCK_16X16, BLOCK_MAX_SEGMENTS, BLOCK_32X32 = BLOCK_MAX_SEGMENTS, @@ -465,13 +449,11 @@ typedef struct VP8_COMP { int rd_prediction_type_threshes[4][NB_PREDICTION_TYPES]; int comp_pred_count[COMP_PRED_CONTEXTS]; int single_pred_count[COMP_PRED_CONTEXTS]; -#if CONFIG_TX_SELECT // FIXME contextualize int txfm_count[TX_SIZE_MAX]; int txfm_count_8x8p[TX_SIZE_MAX - 1]; int64_t rd_tx_select_diff[NB_TXFM_MODES]; int rd_tx_select_threshes[4][NB_TXFM_MODES]; -#endif int RDMULT; int RDDIV; @@ -563,39 +545,28 @@ typedef struct VP8_COMP { // int uv_mode_count[VP8_UV_MODES]; /* intra MB type cts this frame */ int y_uv_mode_count[VP8_YMODES][VP8_UV_MODES]; -#if CONFIG_NEWMVENTROPY nmv_context_counts NMVcount; -#else - unsigned int MVcount [2] [MVvals]; /* (row,col) MV cts this frame */ - unsigned int MVcount_hp [2] [MVvals_hp]; /* (row,col) MV cts this frame */ -#endif unsigned int coef_counts [BLOCK_TYPES] [COEF_BANDS] [PREV_COEF_CONTEXTS] [MAX_ENTROPY_TOKENS]; /* for this frame */ vp8_prob frame_coef_probs [BLOCK_TYPES] [COEF_BANDS] [PREV_COEF_CONTEXTS] [ENTROPY_NODES]; unsigned int frame_branch_ct [BLOCK_TYPES] [COEF_BANDS] [PREV_COEF_CONTEXTS] [ENTROPY_NODES][2]; -#if CONFIG_HYBRIDTRANSFORM unsigned int hybrid_coef_counts [BLOCK_TYPES] [COEF_BANDS] [PREV_COEF_CONTEXTS] [MAX_ENTROPY_TOKENS]; /* for this frame */ vp8_prob frame_hybrid_coef_probs [BLOCK_TYPES] [COEF_BANDS] [PREV_COEF_CONTEXTS] [ENTROPY_NODES]; unsigned int frame_hybrid_branch_ct [BLOCK_TYPES] [COEF_BANDS] [PREV_COEF_CONTEXTS] [ENTROPY_NODES][2]; -#endif unsigned int coef_counts_8x8 [BLOCK_TYPES_8X8] [COEF_BANDS] [PREV_COEF_CONTEXTS] [MAX_ENTROPY_TOKENS]; /* for this frame */ vp8_prob frame_coef_probs_8x8 [BLOCK_TYPES_8X8] [COEF_BANDS] [PREV_COEF_CONTEXTS] [ENTROPY_NODES]; unsigned int frame_branch_ct_8x8 [BLOCK_TYPES_8X8] [COEF_BANDS] [PREV_COEF_CONTEXTS] [ENTROPY_NODES][2]; -#if CONFIG_HYBRIDTRANSFORM8X8 unsigned int hybrid_coef_counts_8x8 [BLOCK_TYPES_8X8] [COEF_BANDS] [PREV_COEF_CONTEXTS] [MAX_ENTROPY_TOKENS]; /* for this frame */ vp8_prob frame_hybrid_coef_probs_8x8 [BLOCK_TYPES_8X8] [COEF_BANDS] [PREV_COEF_CONTEXTS] [ENTROPY_NODES]; unsigned int frame_hybrid_branch_ct_8x8 [BLOCK_TYPES_8X8] [COEF_BANDS] [PREV_COEF_CONTEXTS] [ENTROPY_NODES][2]; -#endif unsigned int coef_counts_16x16 [BLOCK_TYPES_16X16] [COEF_BANDS] [PREV_COEF_CONTEXTS] [MAX_ENTROPY_TOKENS]; /* for this frame */ vp8_prob frame_coef_probs_16x16 [BLOCK_TYPES_16X16] [COEF_BANDS] [PREV_COEF_CONTEXTS] [ENTROPY_NODES]; unsigned int frame_branch_ct_16x16 [BLOCK_TYPES_16X16] [COEF_BANDS] [PREV_COEF_CONTEXTS] [ENTROPY_NODES][2]; -#if CONFIG_HYBRIDTRANSFORM16X16 unsigned int hybrid_coef_counts_16x16 [BLOCK_TYPES_16X16] [COEF_BANDS] [PREV_COEF_CONTEXTS] [MAX_ENTROPY_TOKENS]; /* for this frame */ vp8_prob frame_hybrid_coef_probs_16x16 [BLOCK_TYPES_16X16] [COEF_BANDS] [PREV_COEF_CONTEXTS] [ENTROPY_NODES]; unsigned int frame_hybrid_branch_ct_16x16 [BLOCK_TYPES_16X16] [COEF_BANDS] [PREV_COEF_CONTEXTS] [ENTROPY_NODES][2]; -#endif int gfu_boost; int last_boost; @@ -780,10 +751,8 @@ typedef struct VP8_COMP { int pred_filter_on_count; int pred_filter_off_count; #endif -#if CONFIG_SWITCHABLE_INTERP - unsigned int switchable_interp_count[VP8_SWITCHABLE_FILTERS+1] + unsigned int switchable_interp_count[VP8_SWITCHABLE_FILTERS + 1] [VP8_SWITCHABLE_FILTERS]; -#endif #if CONFIG_NEW_MVREF unsigned int best_ref_index_counts[MAX_MV_REFS]; diff --git a/vp8/encoder/picklpf.c b/vp8/encoder/picklpf.c index 954997889..57bd41468 100644 --- a/vp8/encoder/picklpf.c +++ b/vp8/encoder/picklpf.c @@ -21,7 +21,8 @@ #include "vpx_ports/arm.h" #endif -extern int vp8_calc_ss_err(YV12_BUFFER_CONFIG *source, YV12_BUFFER_CONFIG *dest, const vp8_variance_rtcd_vtable_t *rtcd); +extern int vp8_calc_ss_err(YV12_BUFFER_CONFIG *source, + YV12_BUFFER_CONFIG *dest); #if HAVE_ARMV7 extern void vp8_yv12_copy_frame_yonly_no_extend_frame_borders_neon(YV12_BUFFER_CONFIG *src_ybc, YV12_BUFFER_CONFIG *dst_ybc); #endif @@ -71,7 +72,8 @@ vp8_yv12_copy_partial_frame(YV12_BUFFER_CONFIG *src_ybc, YV12_BUFFER_CONFIG *dst vpx_memcpy(dst_y, src_y, ystride * (linestocopy + 16)); } -static int vp8_calc_partial_ssl_err(YV12_BUFFER_CONFIG *source, YV12_BUFFER_CONFIG *dest, int Fraction, const vp8_variance_rtcd_vtable_t *rtcd) { +static int vp8_calc_partial_ssl_err(YV12_BUFFER_CONFIG *source, + YV12_BUFFER_CONFIG *dest, int Fraction) { int i, j; int Total = 0; int srcoffset, dstoffset; @@ -79,7 +81,6 @@ static int vp8_calc_partial_ssl_err(YV12_BUFFER_CONFIG *source, YV12_BUFFER_CONF unsigned char *dst = dest->y_buffer; int linestocopy = (source->y_height >> (Fraction + 4)); - (void)rtcd; if (linestocopy < 1) linestocopy = 1; @@ -97,7 +98,8 @@ static int vp8_calc_partial_ssl_err(YV12_BUFFER_CONFIG *source, YV12_BUFFER_CONF for (i = 0; i < linestocopy; i += 16) { for (j = 0; j < source->y_width; j += 16) { unsigned int sse; - Total += VARIANCE_INVOKE(rtcd, mse16x16)(src + j, source->y_stride, dst + j, dest->y_stride, &sse); + Total += vp8_mse16x16(src + j, source->y_stride, dst + j, dest->y_stride, + &sse); } src += 16 * source->y_stride; @@ -179,7 +181,7 @@ void vp8cx_pick_filter_level_fast(YV12_BUFFER_CONFIG *sd, VP8_COMP *cpi) { // Get the err using the previous frame's filter value. vp8_loop_filter_partial_frame(cm, &cpi->mb.e_mbd, filt_val); - best_err = vp8_calc_partial_ssl_err(sd, cm->frame_to_show, 3, IF_RTCD(&cpi->rtcd.variance)); + best_err = vp8_calc_partial_ssl_err(sd, cm->frame_to_show, 3); // Re-instate the unfiltered frame vp8_yv12_copy_partial_frame_ptr(&cpi->last_frame_uf, cm->frame_to_show, 3); @@ -192,7 +194,7 @@ void vp8cx_pick_filter_level_fast(YV12_BUFFER_CONFIG *sd, VP8_COMP *cpi) { vp8_loop_filter_partial_frame(cm, &cpi->mb.e_mbd, filt_val); // Get the err for filtered frame - filt_err = vp8_calc_partial_ssl_err(sd, cm->frame_to_show, 3, IF_RTCD(&cpi->rtcd.variance)); + filt_err = vp8_calc_partial_ssl_err(sd, cm->frame_to_show, 3); // Re-instate the unfiltered frame vp8_yv12_copy_partial_frame_ptr(&cpi->last_frame_uf, cm->frame_to_show, 3); @@ -221,7 +223,7 @@ void vp8cx_pick_filter_level_fast(YV12_BUFFER_CONFIG *sd, VP8_COMP *cpi) { vp8_loop_filter_partial_frame(cm, &cpi->mb.e_mbd, filt_val); // Get the err for filtered frame - filt_err = vp8_calc_partial_ssl_err(sd, cm->frame_to_show, 3, IF_RTCD(&cpi->rtcd.variance)); + filt_err = vp8_calc_partial_ssl_err(sd, cm->frame_to_show, 3); // Re-instate the unfiltered frame vp8_yv12_copy_partial_frame_ptr(&cpi->last_frame_uf, cm->frame_to_show, 3); @@ -308,7 +310,7 @@ void vp8cx_pick_filter_level_sg(YV12_BUFFER_CONFIG *sd, VP8_COMP *cpi, int segme vp8cx_set_alt_lf_level(cpi, filt_mid); vp8_loop_filter_frame_segment(cm, &cpi->mb.e_mbd, filt_mid, segment); - best_err = vp8_calc_ss_err(sd, cm->frame_to_show, IF_RTCD(&cpi->rtcd.variance)); + best_err = vp8_calc_ss_err(sd, cm->frame_to_show); filt_best = filt_mid; // Re-instate the unfiltered frame @@ -348,7 +350,7 @@ void vp8cx_pick_filter_level_sg(YV12_BUFFER_CONFIG *sd, VP8_COMP *cpi, int segme vp8cx_set_alt_lf_level(cpi, filt_low); vp8_loop_filter_frame_segment(cm, &cpi->mb.e_mbd, filt_low, segment); - filt_err = vp8_calc_ss_err(sd, cm->frame_to_show, IF_RTCD(&cpi->rtcd.variance)); + filt_err = vp8_calc_ss_err(sd, cm->frame_to_show); // Re-instate the unfiltered frame #if HAVE_ARMV7 @@ -383,7 +385,7 @@ void vp8cx_pick_filter_level_sg(YV12_BUFFER_CONFIG *sd, VP8_COMP *cpi, int segme vp8cx_set_alt_lf_level(cpi, filt_high); vp8_loop_filter_frame_segment(cm, &cpi->mb.e_mbd, filt_high, segment); - filt_err = vp8_calc_ss_err(sd, cm->frame_to_show, IF_RTCD(&cpi->rtcd.variance)); + filt_err = vp8_calc_ss_err(sd, cm->frame_to_show); // Re-instate the unfiltered frame #if HAVE_ARMV7 @@ -517,7 +519,7 @@ void vp8cx_pick_filter_level(YV12_BUFFER_CONFIG *sd, VP8_COMP *cpi) { vp8cx_set_alt_lf_level(cpi, filt_mid); vp8_loop_filter_frame_yonly(cm, &cpi->mb.e_mbd, filt_mid); - best_err = vp8_calc_ss_err(sd, cm->frame_to_show, IF_RTCD(&cpi->rtcd.variance)); + best_err = vp8_calc_ss_err(sd, cm->frame_to_show); filt_best = filt_mid; // Re-instate the unfiltered frame @@ -557,7 +559,7 @@ void vp8cx_pick_filter_level(YV12_BUFFER_CONFIG *sd, VP8_COMP *cpi) { vp8cx_set_alt_lf_level(cpi, filt_low); vp8_loop_filter_frame_yonly(cm, &cpi->mb.e_mbd, filt_low); - filt_err = vp8_calc_ss_err(sd, cm->frame_to_show, IF_RTCD(&cpi->rtcd.variance)); + filt_err = vp8_calc_ss_err(sd, cm->frame_to_show); // Re-instate the unfiltered frame #if HAVE_ARMV7 @@ -592,7 +594,7 @@ void vp8cx_pick_filter_level(YV12_BUFFER_CONFIG *sd, VP8_COMP *cpi) { vp8cx_set_alt_lf_level(cpi, filt_high); vp8_loop_filter_frame_yonly(cm, &cpi->mb.e_mbd, filt_high); - filt_err = vp8_calc_ss_err(sd, cm->frame_to_show, IF_RTCD(&cpi->rtcd.variance)); + filt_err = vp8_calc_ss_err(sd, cm->frame_to_show); // Re-instate the unfiltered frame #if HAVE_ARMV7 diff --git a/vp8/encoder/quantize.c b/vp8/encoder/quantize.c index b6a1f27f8..16b4e6e1d 100644 --- a/vp8/encoder/quantize.c +++ b/vp8/encoder/quantize.c @@ -21,7 +21,6 @@ extern int enc_debug; #endif -#if CONFIG_HYBRIDTRANSFORM void vp8_ht_quantize_b_4x4(BLOCK *b, BLOCKD *d, TX_TYPE tx_type) { int i, rc, eob; int zbin; @@ -85,7 +84,6 @@ void vp8_ht_quantize_b_4x4(BLOCK *b, BLOCKD *d, TX_TYPE tx_type) { d->eob = eob + 1; } -#endif void vp8_regular_quantize_b_4x4(BLOCK *b, BLOCKD *d) { int i, rc, eob; diff --git a/vp8/encoder/quantize.h b/vp8/encoder/quantize.h index 1375ed0b0..e39433fc3 100644 --- a/vp8/encoder/quantize.h +++ b/vp8/encoder/quantize.h @@ -30,11 +30,9 @@ #include "arm/quantize_arm.h" #endif -#if CONFIG_HYBRIDTRANSFORM #define prototype_quantize_block_type(sym) \ void (sym)(BLOCK *b, BLOCKD *d, TX_TYPE type) extern prototype_quantize_block_type(vp8_ht_quantize_b_4x4); -#endif #ifndef vp8_quantize_quantb_4x4 #define vp8_quantize_quantb_4x4 vp8_regular_quantize_b_4x4 diff --git a/vp8/encoder/ratectrl.c b/vp8/encoder/ratectrl.c index 570bedfe9..cc3c82e74 100644 --- a/vp8/encoder/ratectrl.c +++ b/vp8/encoder/ratectrl.c @@ -132,17 +132,10 @@ void vp8_save_coding_context(VP8_COMP *cpi) { // intended for use in a re-code loop in vp8_compress_frame where the // quantizer value is adjusted between loop iterations. -#if CONFIG_NEWMVENTROPY cc->nmvc = cm->fc.nmvc; vp8_copy(cc->nmvjointcost, cpi->mb.nmvjointcost); vp8_copy(cc->nmvcosts, cpi->mb.nmvcosts); vp8_copy(cc->nmvcosts_hp, cpi->mb.nmvcosts_hp); -#else - vp8_copy(cc->mvc, cm->fc.mvc); - vp8_copy(cc->mvcosts, cpi->mb.mvcosts); - vp8_copy(cc->mvc_hp, cm->fc.mvc_hp); - vp8_copy(cc->mvcosts_hp, cpi->mb.mvcosts_hp); -#endif vp8_copy(cc->mv_ref_ct, cm->fc.mv_ref_ct); vp8_copy(cc->mode_context, cm->fc.mode_context); @@ -178,20 +171,12 @@ void vp8_save_coding_context(VP8_COMP *cpi) { vp8_copy(cc->last_mode_lf_deltas, xd->last_mode_lf_deltas); vp8_copy(cc->coef_probs, cm->fc.coef_probs); -#if CONFIG_HYBRIDTRANSFORM vp8_copy(cc->hybrid_coef_probs, cm->fc.hybrid_coef_probs); -#endif vp8_copy(cc->coef_probs_8x8, cm->fc.coef_probs_8x8); -#if CONFIG_HYBRIDTRANSFORM8X8 vp8_copy(cc->hybrid_coef_probs_8x8, cm->fc.hybrid_coef_probs_8x8); -#endif vp8_copy(cc->coef_probs_16x16, cm->fc.coef_probs_16x16); -#if CONFIG_HYBRIDTRANSFORM16X16 vp8_copy(cc->hybrid_coef_probs_16x16, cm->fc.hybrid_coef_probs_16x16); -#endif -#if CONFIG_SWITCHABLE_INTERP vp8_copy(cc->switchable_interp_prob, cm->fc.switchable_interp_prob); -#endif } void vp8_restore_coding_context(VP8_COMP *cpi) { @@ -202,17 +187,10 @@ void vp8_restore_coding_context(VP8_COMP *cpi) { // Restore key state variables to the snapshot state stored in the // previous call to vp8_save_coding_context. -#if CONFIG_NEWMVENTROPY cm->fc.nmvc = cc->nmvc; vp8_copy(cpi->mb.nmvjointcost, cc->nmvjointcost); vp8_copy(cpi->mb.nmvcosts, cc->nmvcosts); vp8_copy(cpi->mb.nmvcosts_hp, cc->nmvcosts_hp); -#else - vp8_copy(cm->fc.mvc, cc->mvc); - vp8_copy(cpi->mb.mvcosts, cc->mvcosts); - vp8_copy(cm->fc.mvc_hp, cc->mvc_hp); - vp8_copy(cpi->mb.mvcosts_hp, cc->mvcosts_hp); -#endif vp8_copy(cm->fc.mv_ref_ct, cc->mv_ref_ct); vp8_copy(cm->fc.mode_context, cc->mode_context); @@ -249,20 +227,12 @@ void vp8_restore_coding_context(VP8_COMP *cpi) { vp8_copy(xd->last_mode_lf_deltas, cc->last_mode_lf_deltas); vp8_copy(cm->fc.coef_probs, cc->coef_probs); -#if CONFIG_HYBRIDTRANSFORM vp8_copy(cm->fc.hybrid_coef_probs, cc->hybrid_coef_probs); -#endif vp8_copy(cm->fc.coef_probs_8x8, cc->coef_probs_8x8); -#if CONFIG_HYBRIDTRANSFORM8X8 vp8_copy(cm->fc.hybrid_coef_probs_8x8, cc->hybrid_coef_probs_8x8); -#endif vp8_copy(cm->fc.coef_probs_16x16, cc->coef_probs_16x16); -#if CONFIG_HYBRIDTRANSFORM16X16 vp8_copy(cm->fc.hybrid_coef_probs_16x16, cc->hybrid_coef_probs_16x16); -#endif -#if CONFIG_SWITCHABLE_INTERP vp8_copy(cm->fc.switchable_interp_prob, cc->switchable_interp_prob); -#endif } @@ -275,16 +245,6 @@ void vp8_setup_key_frame(VP8_COMP *cpi) { vp8_default_bmode_probs(cm->fc.bmode_prob); vp8_init_mv_probs(& cpi->common); -#if CONFIG_NEWMVENTROPY == 0 - /* this is not really required */ - { - int flag[2] = {1, 1}; - vp8_build_component_cost_table( - cpi->mb.mvcost, (const MV_CONTEXT *) cpi->common.fc.mvc, flag); - vp8_build_component_cost_table_hp( - cpi->mb.mvcost_hp, (const MV_CONTEXT_HP *) cpi->common.fc.mvc_hp, flag); - } -#endif // cpi->common.filter_level = 0; // Reset every key frame. cpi->common.filter_level = cpi->common.base_qindex * 3 / 8; diff --git a/vp8/encoder/rdopt.c b/vp8/encoder/rdopt.c index 4b9e90725..e919de36f 100644 --- a/vp8/encoder/rdopt.c +++ b/vp8/encoder/rdopt.c @@ -60,10 +60,8 @@ extern void vp8_update_zbin_extra(VP8_COMP *cpi, MACROBLOCK *x); #define INVALID_MV 0x80008000 -#if CONFIG_SWITCHABLE_INTERP /* Factor to weigh the rate for switchable interp filters */ #define SWITCHABLE_INTERP_RATE_FACTOR 1 -#endif static const int auto_speed_thresh[17] = { 1000, @@ -355,37 +353,31 @@ void vp8_initialize_rd_consts(VP8_COMP *cpi, int QIndex) { cpi->mb.token_costs[TX_4X4], (const vp8_prob( *)[8][PREV_COEF_CONTEXTS][11]) cpi->common.fc.coef_probs, BLOCK_TYPES); -#if CONFIG_HYBRIDTRANSFORM fill_token_costs( cpi->mb.hybrid_token_costs[TX_4X4], (const vp8_prob( *)[8][PREV_COEF_CONTEXTS][11]) cpi->common.fc.hybrid_coef_probs, BLOCK_TYPES); -#endif fill_token_costs( cpi->mb.token_costs[TX_8X8], (const vp8_prob( *)[8][PREV_COEF_CONTEXTS][11]) cpi->common.fc.coef_probs_8x8, BLOCK_TYPES_8X8); -#if CONFIG_HYBRIDTRANSFORM8X8 fill_token_costs( cpi->mb.hybrid_token_costs[TX_8X8], (const vp8_prob( *)[8][PREV_COEF_CONTEXTS][11]) cpi->common.fc.hybrid_coef_probs_8x8, BLOCK_TYPES_8X8); -#endif fill_token_costs( cpi->mb.token_costs[TX_16X16], (const vp8_prob(*)[8][PREV_COEF_CONTEXTS][11]) cpi->common.fc.coef_probs_16x16, BLOCK_TYPES_16X16); -#if CONFIG_HYBRIDTRANSFORM16X16 fill_token_costs( cpi->mb.hybrid_token_costs[TX_16X16], (const vp8_prob(*)[8][PREV_COEF_CONTEXTS][11]) cpi->common.fc.hybrid_coef_probs_16x16, BLOCK_TYPES_16X16); -#endif /*rough estimate for costing*/ cpi->common.kf_ymode_probs_index = cpi->common.base_qindex >> 4; @@ -393,14 +385,12 @@ void vp8_initialize_rd_consts(VP8_COMP *cpi, int QIndex) { if (cpi->common.frame_type != KEY_FRAME) { -#if CONFIG_NEWMVENTROPY vp8_build_nmv_cost_table( cpi->mb.nmvjointcost, cpi->mb.e_mbd.allow_high_precision_mv ? cpi->mb.nmvcost_hp : cpi->mb.nmvcost, &cpi->common.fc.nmvc, cpi->mb.e_mbd.allow_high_precision_mv, 1, 1); -#endif } } @@ -409,19 +399,6 @@ void vp8_auto_select_speed(VP8_COMP *cpi) { milliseconds_for_compress = milliseconds_for_compress * (16 - cpi->oxcf.cpu_used) / 16; -#if 0 - - if (0) { - FILE *f; - - f = fopen("speed.stt", "a"); - fprintf(f, " %8ld %10ld %10ld %10ld\n", - cpi->common.current_video_frame, cpi->Speed, milliseconds_for_compress, cpi->avg_pick_mode_time); - fclose(f); - } - -#endif - /* // this is done during parameter valid check if( cpi->oxcf.cpu_used > 16) @@ -520,7 +497,7 @@ int vp8_mbuverror_c(MACROBLOCK *mb) { return error; } -int VP8_UVSSE(MACROBLOCK *x, const vp8_variance_rtcd_vtable_t *rtcd) { +int vp8_uvsse(MACROBLOCK *x) { unsigned char *uptr, *vptr; unsigned char *upred_ptr = (*(x->block[16].base_src) + x->block[16].src); unsigned char *vpred_ptr = (*(x->block[20].base_src) + x->block[20].src); @@ -551,16 +528,14 @@ int VP8_UVSSE(MACROBLOCK *x, const vp8_variance_rtcd_vtable_t *rtcd) { vptr = x->e_mbd.pre.v_buffer + offset; if ((mv_row | mv_col) & 7) { - VARIANCE_INVOKE(rtcd, subpixvar8x8)(uptr, pre_stride, - (mv_col & 7) << 1, (mv_row & 7) << 1, upred_ptr, uv_stride, &sse2); - VARIANCE_INVOKE(rtcd, subpixvar8x8)(vptr, pre_stride, - (mv_col & 7) << 1, (mv_row & 7) << 1, vpred_ptr, uv_stride, &sse1); + vp8_sub_pixel_variance8x8(uptr, pre_stride, (mv_col & 7) << 1, + (mv_row & 7) << 1, upred_ptr, uv_stride, &sse2); + vp8_sub_pixel_variance8x8(vptr, pre_stride, (mv_col & 7) << 1, + (mv_row & 7) << 1, vpred_ptr, uv_stride, &sse1); sse2 += sse1; } else { - VARIANCE_INVOKE(rtcd, var8x8)(uptr, pre_stride, - upred_ptr, uv_stride, &sse2); - VARIANCE_INVOKE(rtcd, var8x8)(vptr, pre_stride, - vpred_ptr, uv_stride, &sse1); + vp8_variance8x8(uptr, pre_stride, upred_ptr, uv_stride, &sse2); + vp8_variance8x8(vptr, pre_stride, vpred_ptr, uv_stride, &sse1); sse2 += sse1; } return sse2; @@ -607,9 +582,7 @@ static int cost_coeffs(MACROBLOCK *mb, BLOCKD *b, PLANE_TYPE type, short *qcoeff_ptr = b->qcoeff; MACROBLOCKD *xd = &mb->e_mbd; MB_MODE_INFO *mbmi = &mb->e_mbd.mode_info_context->mbmi; -#if CONFIG_HYBRIDTRANSFORM || CONFIG_HYBRIDTRANSFORM8X8 || CONFIG_HYBRIDTRANSFORM16X16 TX_TYPE tx_type = DCT_DCT; -#endif int segment_id = mbmi->segment_id; switch (tx_size) { @@ -617,7 +590,6 @@ static int cost_coeffs(MACROBLOCK *mb, BLOCKD *b, PLANE_TYPE type, scan = vp8_default_zig_zag1d; band = vp8_coef_bands; default_eob = 16; -#if CONFIG_HYBRIDTRANSFORM if (type == PLANE_TYPE_Y_WITH_DC) { tx_type = get_tx_type_4x4(xd, b); if (tx_type != DCT_DCT) { @@ -636,14 +608,12 @@ static int cost_coeffs(MACROBLOCK *mb, BLOCKD *b, PLANE_TYPE type, } } } -#endif break; case TX_8X8: scan = vp8_default_zig_zag1d_8x8; band = vp8_coef_bands_8x8; default_eob = 64; -#if CONFIG_HYBRIDTRANSFORM8X8 if (type == PLANE_TYPE_Y_WITH_DC) { BLOCKD *bb; int ib = (b - xd->block); @@ -653,17 +623,14 @@ static int cost_coeffs(MACROBLOCK *mb, BLOCKD *b, PLANE_TYPE type, tx_type = get_tx_type_8x8(xd, bb); } } -#endif break; case TX_16X16: scan = vp8_default_zig_zag1d_16x16; band = vp8_coef_bands_16x16; default_eob = 256; -#if CONFIG_HYBRIDTRANSFORM16X16 if (type == PLANE_TYPE_Y_WITH_DC) { tx_type = get_tx_type_16x16(xd, b); } -#endif break; default: break; @@ -675,7 +642,6 @@ static int cost_coeffs(MACROBLOCK *mb, BLOCKD *b, PLANE_TYPE type, VP8_COMBINEENTROPYCONTEXTS(pt, *a, *l); -#if CONFIG_HYBRIDTRANSFORM || CONFIG_HYBRIDTRANSFORM8X8 || CONFIG_HYBRIDTRANSFORM16X16 if (tx_type != DCT_DCT) { for (; c < eob; c++) { int v = qcoeff_ptr[scan[c]]; @@ -687,9 +653,7 @@ static int cost_coeffs(MACROBLOCK *mb, BLOCKD *b, PLANE_TYPE type, if (c < seg_eob) cost += mb->hybrid_token_costs[tx_size][type][band[c]] [pt][DCT_EOB_TOKEN]; - } else -#endif - { + } else { for (; c < eob; c++) { int v = qcoeff_ptr[scan[c]]; int t = vp8_dct_value_tokens_ptr[v].Token; @@ -870,9 +834,7 @@ static void macro_block_yrd_16x16(MACROBLOCK *mb, int *Rate, int *Distortion, MACROBLOCKD *xd = &mb->e_mbd; BLOCKD *b = &mb->e_mbd.block[0]; BLOCK *be = &mb->block[0]; -#if CONFIG_HYBRIDTRANSFORM16X16 TX_TYPE tx_type; -#endif ENCODEMB_INVOKE(&rtcd->encodemb, submby)( mb->src_diff, @@ -880,24 +842,18 @@ static void macro_block_yrd_16x16(MACROBLOCK *mb, int *Rate, int *Distortion, mb->e_mbd.predictor, mb->block[0].src_stride); -#if CONFIG_HYBRIDTRANSFORM16X16 tx_type = get_tx_type_16x16(xd, b); if (tx_type != DCT_DCT) { vp8_fht_c(be->src_diff, be->coeff, 32, tx_type, 16); } else vp8_transform_mby_16x16(mb); -#else - vp8_transform_mby_16x16(mb); -#endif vp8_quantize_mby_16x16(mb); -#if CONFIG_HYBRIDTRANSFORM16X16 // TODO(jingning) is it possible to quickly determine whether to force // trailing coefficients to be zero, instead of running trellis // optimization in the rate-distortion optimization loop? if (mb->e_mbd.mode_info_context->mbmi.mode < I8X8_PRED) vp8_optimize_mby_16x16(mb, rtcd); -#endif d = ENCODEMB_INVOKE(&rtcd->encodemb, mberr)(mb, 0); @@ -913,8 +869,6 @@ static void macro_block_yrd(VP8_COMP *cpi, MACROBLOCK *x, int *rate, VP8_COMMON *cm = &cpi->common; MB_MODE_INFO *mbmi = &x->e_mbd.mode_info_context->mbmi; -#if CONFIG_TX_SELECT - MACROBLOCKD *xd = &x->e_mbd; int can_skip = cm->mb_no_coeff_skip; vp8_prob skip_prob = can_skip ? get_pred_prob(cm, xd, PRED_MBSKIP) : 128; @@ -1022,25 +976,6 @@ static void macro_block_yrd(VP8_COMP *cpi, MACROBLOCK *x, int *rate, else txfm_cache[TX_MODE_SELECT] = rd4x4s < rd8x8s ? rd4x4s : rd8x8s; -#else /* CONFIG_TX_SELECT */ - - switch (cpi->common.txfm_mode) { - case ALLOW_16X16: - macro_block_yrd_16x16(x, rate, distortion, IF_RTCD(&cpi->rtcd), skippable); - mbmi->txfm_size = TX_16X16; - break; - case ALLOW_8X8: - macro_block_yrd_8x8(x, rate, distortion, IF_RTCD(&cpi->rtcd), skippable); - mbmi->txfm_size = TX_8X8; - break; - default: - case ONLY_4X4: - macro_block_yrd_4x4(x, rate, distortion, IF_RTCD(&cpi->rtcd), skippable); - mbmi->txfm_size = TX_4X4; - break; - } - -#endif /* CONFIG_TX_SELECT */ } static void copy_predictor(unsigned char *dst, const unsigned char *predictor) { @@ -1155,10 +1090,8 @@ static int64_t rd_pick_intra4x4block(VP8_COMP *cpi, MACROBLOCK *x, BLOCK *be, ENTROPY_CONTEXT ta = *a, tempa = *a; ENTROPY_CONTEXT tl = *l, templ = *l; -#if CONFIG_HYBRIDTRANSFORM TX_TYPE tx_type = DCT_DCT; TX_TYPE best_tx_type = DCT_DCT; -#endif /* * The predictor buffer is a 2d buffer with a stride of 16. Create * a temp buffer that meets the stride requirements, but we are only @@ -1191,7 +1124,6 @@ static int64_t rd_pick_intra4x4block(VP8_COMP *cpi, MACROBLOCK *x, BLOCK *be, ENCODEMB_INVOKE(IF_RTCD(&cpi->rtcd.encodemb), subb)(be, b, 16); b->bmi.as_mode.first = mode; -#if CONFIG_HYBRIDTRANSFORM tx_type = get_tx_type_4x4(xd, b); if (tx_type != DCT_DCT) { vp8_fht_c(be->src_diff, be->coeff, 32, tx_type, 4); @@ -1200,10 +1132,6 @@ static int64_t rd_pick_intra4x4block(VP8_COMP *cpi, MACROBLOCK *x, BLOCK *be, x->vp8_short_fdct4x4(be->src_diff, be->coeff, 32); x->quantize_b_4x4(be, b); } -#else - x->vp8_short_fdct4x4(be->src_diff, be->coeff, 32); - x->quantize_b_4x4(be, b); -#endif tempa = ta; templ = tl; @@ -1221,9 +1149,7 @@ static int64_t rd_pick_intra4x4block(VP8_COMP *cpi, MACROBLOCK *x, BLOCK *be, *bestdistortion = distortion; best_rd = this_rd; *best_mode = mode; -#if CONFIG_HYBRIDTRANSFORM best_tx_type = tx_type; -#endif #if CONFIG_COMP_INTRA_PRED *best_second_mode = mode2; @@ -1242,17 +1168,12 @@ static int64_t rd_pick_intra4x4block(VP8_COMP *cpi, MACROBLOCK *x, BLOCK *be, b->bmi.as_mode.second = (B_PREDICTION_MODE)(*best_second_mode); #endif -#if CONFIG_HYBRIDTRANSFORM // inverse transform if (best_tx_type != DCT_DCT) vp8_ihtllm_c(best_dqcoeff, b->diff, 32, best_tx_type, 4); else IDCT_INVOKE(IF_RTCD(&cpi->rtcd.common->idct), idct16)( best_dqcoeff, b->diff, 32); -#else - IDCT_INVOKE(IF_RTCD(&cpi->rtcd.common->idct), idct16)( - best_dqcoeff, b->diff, 32); -#endif vp8_recon_b(best_predictor, b->diff, *(b->base_dst) + b->dst, b->dst_stride); @@ -1405,11 +1326,9 @@ static int64_t rd_pick_intra16x16mby_mode(VP8_COMP *cpi, int64_t this_rd; MACROBLOCKD *xd = &x->e_mbd; -#if CONFIG_TX_SELECT int i; for (i = 0; i < NB_TXFM_MODES; i++) txfm_cache[i] = INT64_MAX; -#endif // Y Search for 16x16 intra prediction mode for (mode = DC_PRED; mode <= TM_PRED; mode++) { @@ -1452,7 +1371,6 @@ static int64_t rd_pick_intra16x16mby_mode(VP8_COMP *cpi, *skippable = skip; } -#if CONFIG_TX_SELECT for (i = 0; i < NB_TXFM_MODES; i++) { int64_t adj_rd = this_rd + local_txfm_cache[i] - local_txfm_cache[cpi->common.txfm_mode]; @@ -1460,7 +1378,6 @@ static int64_t rd_pick_intra16x16mby_mode(VP8_COMP *cpi, txfm_cache[i] = adj_rd; } } -#endif #if CONFIG_COMP_INTRA_PRED } @@ -1535,22 +1452,18 @@ static int64_t rd_pick_intra8x8block(VP8_COMP *cpi, MACROBLOCK *x, int ib, vp8_subtract_4b_c(be, b, 16); if (xd->mode_info_context->mbmi.txfm_size == TX_8X8) { -#if CONFIG_HYBRIDTRANSFORM8X8 TX_TYPE tx_type = get_tx_type_8x8(xd, b); if (tx_type != DCT_DCT) vp8_fht_c(be->src_diff, (x->block + idx)->coeff, 32, tx_type, 8); else x->vp8_short_fdct8x8(be->src_diff, (x->block + idx)->coeff, 32); -#else - x->vp8_short_fdct8x8(be->src_diff, (x->block + idx)->coeff, 32); -#endif x->quantize_b_8x8(x->block + idx, xd->block + idx); // compute quantization mse of 8x8 block distortion = vp8_block_error_c((x->block + idx)->coeff, (xd->block + idx)->dqcoeff, 64); - ta0 = *(a + vp8_block2above_8x8[idx]); - tl0 = *(l + vp8_block2left_8x8 [idx]); + ta0 = a[vp8_block2above_8x8[idx]]; + tl0 = l[vp8_block2left_8x8[idx]]; rate_t = cost_coeffs(x, xd->block + idx, PLANE_TYPE_Y_WITH_DC, &ta0, &tl0, TX_8X8); @@ -1576,10 +1489,10 @@ static int64_t rd_pick_intra8x8block(VP8_COMP *cpi, MACROBLOCK *x, int ib, distortion += vp8_block_error_c((x->block + ib + 5)->coeff, (xd->block + ib + 5)->dqcoeff, 16); - ta0 = *(a + vp8_block2above[ib]); - ta1 = *(a + vp8_block2above[ib + 1]); - tl0 = *(l + vp8_block2above[ib]); - tl1 = *(l + vp8_block2above[ib + 4]); + ta0 = a[vp8_block2above[ib]]; + ta1 = a[vp8_block2above[ib + 1]]; + tl0 = l[vp8_block2left[ib]]; + tl1 = l[vp8_block2left[ib + 4]]; rate_t = cost_coeffs(x, xd->block + ib, PLANE_TYPE_Y_WITH_DC, &ta0, &tl0, TX_4X4); rate_t += cost_coeffs(x, xd->block + ib + 1, PLANE_TYPE_Y_WITH_DC, @@ -1621,15 +1534,15 @@ static int64_t rd_pick_intra8x8block(VP8_COMP *cpi, MACROBLOCK *x, int ib, vp8_encode_intra8x8(IF_RTCD(&cpi->rtcd), x, ib); if (xd->mode_info_context->mbmi.txfm_size == TX_8X8) { - *(a + vp8_block2above_8x8[idx]) = besta0; - *(a + vp8_block2above_8x8[idx] + 1) = besta1; - *(l + vp8_block2left_8x8 [idx]) = bestl0; - *(l + vp8_block2left_8x8 [idx] + 1) = bestl1; + a[vp8_block2above_8x8[idx]] = besta0; + a[vp8_block2above_8x8[idx] + 1] = besta1; + l[vp8_block2left_8x8[idx]] = bestl0; + l[vp8_block2left_8x8[idx] + 1] = bestl1; } else { - *(a + vp8_block2above[ib]) = besta0; - *(a + vp8_block2above[ib + 1]) = besta1; - *(l + vp8_block2above[ib]) = bestl0; - *(l + vp8_block2above[ib + 4]) = bestl1; + a[vp8_block2above[ib]] = besta0; + a[vp8_block2above[ib + 1]] = besta1; + l[vp8_block2left[ib]] = bestl0; + l[vp8_block2left[ib + 4]] = bestl1; } return best_rd; @@ -2223,12 +2136,22 @@ static int64_t encode_inter_mb_segment_8x8(MACROBLOCK *x, int which_label, int *labelyrate, int *distortion, + int64_t *otherrd, ENTROPY_CONTEXT *ta, ENTROPY_CONTEXT *tl, const VP8_ENCODER_RTCD *rtcd) { int i, j; MACROBLOCKD *xd = &x->e_mbd; const int iblock[4] = { 0, 1, 4, 5 }; + int othercost = 0, otherdist = 0; + ENTROPY_CONTEXT_PLANES tac, tlc; + ENTROPY_CONTEXT *tacp = (ENTROPY_CONTEXT *) &tac, + *tlcp = (ENTROPY_CONTEXT *) &tlc; + + if (otherrd) { + memcpy(&tac, ta, sizeof(ENTROPY_CONTEXT_PLANES)); + memcpy(&tlc, tl, sizeof(ENTROPY_CONTEXT_PLANES)); + } *distortion = 0; *labelyrate = 0; @@ -2236,8 +2159,9 @@ static int64_t encode_inter_mb_segment_8x8(MACROBLOCK *x, int ib = vp8_i8x8_block[i]; if (labels[ib] == which_label) { - BLOCKD *bd = &xd->block[ib]; - BLOCK *be = &x->block[ib]; + int idx = (ib & 8) + ((ib & 2) << 1); + BLOCKD *bd = &xd->block[ib], *bd2 = &xd->block[idx]; + BLOCK *be = &x->block[ib], *be2 = &x->block[idx]; int thisdistortion; vp8_build_inter_predictors4b(xd, bd, 16); @@ -2245,24 +2169,66 @@ static int64_t encode_inter_mb_segment_8x8(MACROBLOCK *x, vp8_build_2nd_inter_predictors4b(xd, bd, 16); vp8_subtract_4b_c(be, bd, 16); - for (j = 0; j < 4; j += 2) { - bd = &xd->block[ib + iblock[j]]; - be = &x->block[ib + iblock[j]]; - x->vp8_short_fdct8x4(be->src_diff, be->coeff, 32); - x->quantize_b_4x4_pair(be, be + 1, bd, bd + 1); - thisdistortion = vp8_block_error_c(be->coeff, bd->dqcoeff, 32); + if (xd->mode_info_context->mbmi.txfm_size == TX_4X4) { + if (otherrd) { + x->vp8_short_fdct8x8(be->src_diff, be2->coeff, 32); + x->quantize_b_8x8(be2, bd2); + thisdistortion = vp8_block_error_c(be2->coeff, bd2->dqcoeff, 64); + otherdist += thisdistortion; + othercost += cost_coeffs(x, bd2, PLANE_TYPE_Y_WITH_DC, + tacp + vp8_block2above_8x8[idx], + tlcp + vp8_block2left_8x8[idx], TX_8X8); + } + for (j = 0; j < 4; j += 2) { + bd = &xd->block[ib + iblock[j]]; + be = &x->block[ib + iblock[j]]; + x->vp8_short_fdct8x4(be->src_diff, be->coeff, 32); + x->quantize_b_4x4_pair(be, be + 1, bd, bd + 1); + thisdistortion = vp8_block_error_c(be->coeff, bd->dqcoeff, 32); + *distortion += thisdistortion; + *labelyrate += cost_coeffs(x, bd, PLANE_TYPE_Y_WITH_DC, + ta + vp8_block2above[ib + iblock[j]], + tl + vp8_block2left[ib + iblock[j]], + TX_4X4); + *labelyrate += cost_coeffs(x, bd + 1, PLANE_TYPE_Y_WITH_DC, + ta + vp8_block2above[ib + iblock[j] + 1], + tl + vp8_block2left[ib + iblock[j]], + TX_4X4); + } + } else /* 8x8 */ { + if (otherrd) { + for (j = 0; j < 4; j += 2) { + BLOCKD *bd3 = &xd->block[ib + iblock[j]]; + BLOCK *be3 = &x->block[ib + iblock[j]]; + x->vp8_short_fdct8x4(be3->src_diff, be3->coeff, 32); + x->quantize_b_4x4_pair(be3, be3 + 1, bd3, bd3 + 1); + thisdistortion = vp8_block_error_c(be3->coeff, bd3->dqcoeff, 32); + otherdist += thisdistortion; + othercost += cost_coeffs(x, bd3, PLANE_TYPE_Y_WITH_DC, + tacp + vp8_block2above[ib + iblock[j]], + tlcp + vp8_block2left[ib + iblock[j]], + TX_4X4); + othercost += cost_coeffs(x, bd3 + 1, PLANE_TYPE_Y_WITH_DC, + tacp + vp8_block2above[ib + iblock[j] + 1], + tlcp + vp8_block2left[ib + iblock[j]], + TX_4X4); + } + } + x->vp8_short_fdct8x8(be->src_diff, be2->coeff, 32); + x->quantize_b_8x8(be2, bd2); + thisdistortion = vp8_block_error_c(be2->coeff, bd2->dqcoeff, 64); *distortion += thisdistortion; - *labelyrate += cost_coeffs(x, bd, PLANE_TYPE_Y_WITH_DC, - ta + vp8_block2above[ib + iblock[j]], - tl + vp8_block2left[ib + iblock[j]], TX_4X4); - *labelyrate += cost_coeffs(x, bd + 1, PLANE_TYPE_Y_WITH_DC, - ta + vp8_block2above[ib + iblock[j] + 1], - tl + vp8_block2left[ib + iblock[j]], - TX_4X4); + *labelyrate += cost_coeffs(x, bd2, PLANE_TYPE_Y_WITH_DC, + ta + vp8_block2above_8x8[idx], + tl + vp8_block2left_8x8[idx], TX_8X8); } } } *distortion >>= 2; + if (otherrd) { + othercost >>= 2; + *otherrd = RDCOST(x->rdmult, x->rddiv, othercost, otherdist); + } return RDCOST(x->rdmult, x->rddiv, *labelyrate, *distortion); } @@ -2274,7 +2240,8 @@ typedef struct { int_mv mvp; int64_t segment_rd; - int segment_num; + SPLITMV_PARTITIONING_TYPE segment_num; + TX_SIZE txfm_size; int r; int d; int segment_yrate; @@ -2300,9 +2267,14 @@ int mv_check_bounds(MACROBLOCK *x, int_mv *mv) { return r; } -static void rd_check_segment(VP8_COMP *cpi, MACROBLOCK *x, - BEST_SEG_INFO *bsi, unsigned int segmentation, - int_mv seg_mvs[16 /* n_blocks */][MAX_REF_FRAMES - 1]) { +static void rd_check_segment_txsize(VP8_COMP *cpi, MACROBLOCK *x, + BEST_SEG_INFO *bsi, + SPLITMV_PARTITIONING_TYPE segmentation, + TX_SIZE tx_size, int64_t *otherrds, + int64_t *rds, int *completed, + /* 16 = n_blocks */ + int_mv seg_mvs[16 /* n_blocks */] + [MAX_REF_FRAMES - 1]) { int i, j; int const *labels; int br = 0, bd = 0; @@ -2310,12 +2282,12 @@ static void rd_check_segment(VP8_COMP *cpi, MACROBLOCK *x, MB_MODE_INFO * mbmi = &x->e_mbd.mode_info_context->mbmi; int label_count; - int64_t this_segment_rd = 0; + int64_t this_segment_rd = 0, other_segment_rd; int label_mv_thresh; int rate = 0; int sbr = 0, sbd = 0; int segmentyrate = 0; - uint8_t best_eobs[16]; + uint8_t best_eobs[16] = { 0 }; vp8_variance_fn_ptr_t *v_fn_ptr; @@ -2343,20 +2315,23 @@ static void rd_check_segment(VP8_COMP *cpi, MACROBLOCK *x, label_mv_thresh = 1 * bsi->mvthresh / label_count; // Segmentation method overheads - rate = vp8_cost_token(vp8_mbsplit_tree, vp8_mbsplit_probs, vp8_mbsplit_encodings + segmentation); + rate = vp8_cost_token(vp8_mbsplit_tree, vp8_mbsplit_probs, + vp8_mbsplit_encodings + segmentation); rate += vp8_cost_mv_ref(cpi, SPLITMV, bsi->mdcounts); this_segment_rd += RDCOST(x->rdmult, x->rddiv, rate, 0); br += rate; + other_segment_rd = this_segment_rd; - for (i = 0; i < label_count; i++) { + mbmi->txfm_size = tx_size; + for (i = 0; i < label_count && this_segment_rd < bsi->segment_rd; i++) { int_mv mode_mv[B_MODE_COUNT], second_mode_mv[B_MODE_COUNT]; - int64_t best_label_rd = INT64_MAX; + int64_t best_label_rd = INT64_MAX, best_other_rd = INT64_MAX; B_PREDICTION_MODE mode_selected = ZERO4X4; int bestlabelyrate = 0; // search for the best motion vector on this segment for (this_mode = LEFT4X4; this_mode <= NEW4X4; this_mode ++) { - int64_t this_rd; + int64_t this_rd, other_rd; int distortion; int labelyrate; ENTROPY_CONTEXT_PLANES t_above_s, t_left_s; @@ -2378,21 +2353,23 @@ static void rd_check_segment(VP8_COMP *cpi, MACROBLOCK *x, BLOCK *c; BLOCKD *e; - // Is the best so far sufficiently good that we cant justify doing and new motion search. + /* Is the best so far sufficiently good that we cant justify doing + * and new motion search. */ if (best_label_rd < label_mv_thresh) break; if (cpi->compressor_speed) { - if (segmentation == BLOCK_8X16 || segmentation == BLOCK_16X8) { + if (segmentation == PARTITIONING_8X16 || + segmentation == PARTITIONING_16X8) { bsi->mvp.as_int = bsi->sv_mvp[i].as_int; - if (i == 1 && segmentation == BLOCK_16X8) + if (i == 1 && segmentation == PARTITIONING_16X8) bsi->mvp.as_int = bsi->sv_mvp[2].as_int; step_param = bsi->sv_istep[i]; } // use previous block's result as next block's MV predictor. - if (segmentation == BLOCK_4X4 && i > 0) { + if (segmentation == PARTITIONING_4X4 && i > 0) { bsi->mvp.as_int = x->e_mbd.block[i - 1].bmi.as_mv.first.as_int; if (i == 4 || i == 8 || i == 12) bsi->mvp.as_int = x->e_mbd.block[i - 4].bmi.as_mv.first.as_int; @@ -2424,7 +2401,8 @@ static void rd_check_segment(VP8_COMP *cpi, MACROBLOCK *x, // Should we do a full search (best quality only) if ((cpi->compressor_speed == 0) && (bestsme >> sseshift) > 4000) { /* Check if mvp_full is within the range. */ - vp8_clamp_mv(&mvp_full, x->mv_col_min, x->mv_col_max, x->mv_row_min, x->mv_row_max); + vp8_clamp_mv(&mvp_full, x->mv_col_min, x->mv_col_max, + x->mv_row_min, x->mv_row_max); thissme = cpi->full_search_sad(x, c, e, &mvp_full, sadpb, 16, v_fn_ptr, @@ -2434,7 +2412,8 @@ static void rd_check_segment(VP8_COMP *cpi, MACROBLOCK *x, bestsme = thissme; mode_mv[NEW4X4].as_int = e->bmi.as_mv.first.as_int; } else { - // The full search result is actually worse so re-instate the previous best vector + /* The full search result is actually worse so re-instate the + * previous best vector */ e->bmi.as_mv.first.as_int = mode_mv[NEW4X4].as_int; } } @@ -2444,15 +2423,16 @@ static void rd_check_segment(VP8_COMP *cpi, MACROBLOCK *x, int distortion; unsigned int sse; cpi->find_fractional_mv_step(x, c, e, &mode_mv[NEW4X4], - bsi->ref_mv, x->errorperbit, v_fn_ptr, XMVCOST, - &distortion, &sse); + bsi->ref_mv, x->errorperbit, v_fn_ptr, + XMVCOST, &distortion, &sse); // safe motion search result for use in compound prediction seg_mvs[i][mbmi->ref_frame - 1].as_int = mode_mv[NEW4X4].as_int; } } /* NEW4X4 */ else if (mbmi->second_ref_frame && this_mode == NEW4X4) { - // motion search not completed? Then skip newmv for this block with comppred + /* motion search not completed? Then skip newmv for this block with + * comppred */ if (seg_mvs[i][mbmi->second_ref_frame - 1].as_int == INVALID_MV || seg_mvs[i][mbmi->ref_frame - 1].as_int == INVALID_MV) { continue; @@ -2474,14 +2454,15 @@ static void rd_check_segment(VP8_COMP *cpi, MACROBLOCK *x, mv_check_bounds(x, &second_mode_mv[this_mode])) continue; - if (segmentation == BLOCK_4X4) { + if (segmentation == PARTITIONING_4X4) { this_rd = encode_inter_mb_segment(x, labels, i, &labelyrate, &distortion, ta_s, tl_s, IF_RTCD(&cpi->rtcd)); + other_rd = this_rd; } else { this_rd = encode_inter_mb_segment_8x8(x, labels, i, &labelyrate, - &distortion, ta_s, tl_s, - IF_RTCD(&cpi->rtcd)); + &distortion, &other_rd, + ta_s, tl_s, IF_RTCD(&cpi->rtcd)); } this_rd += RDCOST(x->rdmult, x->rddiv, rate, 0); rate += labelyrate; @@ -2492,9 +2473,20 @@ static void rd_check_segment(VP8_COMP *cpi, MACROBLOCK *x, bestlabelyrate = labelyrate; mode_selected = this_mode; best_label_rd = this_rd; - for (j = 0; j < 16; j++) - if (labels[j] == i) - best_eobs[j] = x->e_mbd.block[j].eob; + if (x->e_mbd.mode_info_context->mbmi.txfm_size == TX_4X4) { + for (j = 0; j < 16; j++) + if (labels[j] == i) + best_eobs[j] = x->e_mbd.block[j].eob; + } else { + for (j = 0; j < 4; j++) { + int ib = vp8_i8x8_block[j], idx = j * 4; + + if (labels[ib] == i) + best_eobs[idx] = x->e_mbd.block[idx].eob; + } + } + if (other_rd < best_other_rd) + best_other_rd = other_rd; vpx_memcpy(ta_b, ta_s, sizeof(ENTROPY_CONTEXT_PLANES)); vpx_memcpy(tl_b, tl_s, sizeof(ENTROPY_CONTEXT_PLANES)); @@ -2506,18 +2498,18 @@ static void rd_check_segment(VP8_COMP *cpi, MACROBLOCK *x, vpx_memcpy(tl, tl_b, sizeof(ENTROPY_CONTEXT_PLANES)); labels2mode(x, labels, i, mode_selected, &mode_mv[mode_selected], - &second_mode_mv[mode_selected], seg_mvs[i], bsi->ref_mv, bsi->second_ref_mv, XMVCOST); + &second_mode_mv[mode_selected], seg_mvs[i], + bsi->ref_mv, bsi->second_ref_mv, XMVCOST); br += sbr; bd += sbd; segmentyrate += bestlabelyrate; this_segment_rd += best_label_rd; - - if (this_segment_rd >= bsi->segment_rd) { - break; - } - - + other_segment_rd += best_other_rd; + if (rds) + rds[i] = this_segment_rd; + if (otherrds) + rds[i] = other_segment_rd; } /* for each label */ if (this_segment_rd < bsi->segment_rd) { @@ -2526,6 +2518,7 @@ static void rd_check_segment(VP8_COMP *cpi, MACROBLOCK *x, bsi->segment_yrate = segmentyrate; bsi->segment_rd = this_segment_rd; bsi->segment_num = segmentation; + bsi->txfm_size = mbmi->txfm_size; // store everything needed to come back to this!! for (i = 0; i < 16; i++) { @@ -2538,6 +2531,105 @@ static void rd_check_segment(VP8_COMP *cpi, MACROBLOCK *x, bsi->eobs[i] = best_eobs[i]; } } + + if (completed) { + *completed = i; + } +} + +static void rd_check_segment(VP8_COMP *cpi, MACROBLOCK *x, + BEST_SEG_INFO *bsi, + unsigned int segmentation, + /* 16 = n_blocks */ + int_mv seg_mvs[16][MAX_REF_FRAMES - 1], + int64_t txfm_cache[NB_TXFM_MODES]) { + int i, n, c = vp8_mbsplit_count[segmentation]; + + if (segmentation == PARTITIONING_4X4) { + int64_t rd[16]; + + rd_check_segment_txsize(cpi, x, bsi, segmentation, TX_4X4, NULL, + rd, &n, seg_mvs); + if (n == c) { + for (i = 0; i < NB_TXFM_MODES; i++) { + if (rd[c - 1] < txfm_cache[i]) + txfm_cache[i] = rd[c - 1]; + } + } + } else { + int64_t diff, base_rd; + int cost4x4 = vp8_cost_bit(cpi->common.prob_tx[0], 0); + int cost8x8 = vp8_cost_bit(cpi->common.prob_tx[0], 1); + + if (cpi->common.txfm_mode == TX_MODE_SELECT) { + int64_t rd4x4[4], rd8x8[4]; + int n4x4, n8x8, nmin; + BEST_SEG_INFO bsi4x4, bsi8x8; + + /* factor in cost of cost4x4/8x8 in decision */ + vpx_memcpy(&bsi4x4, bsi, sizeof(*bsi)); + vpx_memcpy(&bsi8x8, bsi, sizeof(*bsi)); + rd_check_segment_txsize(cpi, x, &bsi4x4, segmentation, + TX_4X4, NULL, rd4x4, &n4x4, seg_mvs); + rd_check_segment_txsize(cpi, x, &bsi8x8, segmentation, + TX_8X8, NULL, rd8x8, &n8x8, seg_mvs); + if (bsi4x4.segment_num == segmentation) { + bsi4x4.segment_rd += RDCOST(x->rdmult, x->rddiv, cost4x4, 0); + if (bsi4x4.segment_rd < bsi->segment_rd) + vpx_memcpy(bsi, &bsi4x4, sizeof(*bsi)); + } + if (bsi8x8.segment_num == segmentation) { + bsi8x8.segment_rd += RDCOST(x->rdmult, x->rddiv, cost8x8, 0); + if (bsi8x8.segment_rd < bsi->segment_rd) + vpx_memcpy(bsi, &bsi8x8, sizeof(*bsi)); + } + n = n4x4 > n8x8 ? n4x4 : n8x8; + if (n == c) { + nmin = n4x4 < n8x8 ? n4x4 : n8x8; + diff = rd8x8[nmin - 1] - rd4x4[nmin - 1]; + if (n == n4x4) { + base_rd = rd4x4[c - 1]; + } else { + base_rd = rd8x8[c - 1] - diff; + } + } + } else { + int64_t rd[4], otherrd[4]; + + if (cpi->common.txfm_mode == ONLY_4X4) { + rd_check_segment_txsize(cpi, x, bsi, segmentation, TX_4X4, otherrd, + rd, &n, seg_mvs); + if (n == c) { + base_rd = rd[c - 1]; + diff = otherrd[c - 1] - rd[c - 1]; + } + } else /* use 8x8 transform */ { + rd_check_segment_txsize(cpi, x, bsi, segmentation, TX_8X8, otherrd, + rd, &n, seg_mvs); + if (n == c) { + diff = rd[c - 1] - otherrd[c - 1]; + base_rd = otherrd[c - 1]; + } + } + } + + if (n == c) { + if (base_rd < txfm_cache[ONLY_4X4]) { + txfm_cache[ONLY_4X4] = base_rd; + } + if (base_rd + diff < txfm_cache[1]) { + txfm_cache[ALLOW_8X8] = txfm_cache[ALLOW_16X16] = base_rd + diff; + } + if (diff < 0) { + base_rd += diff + RDCOST(x->rdmult, x->rddiv, cost8x8, 0); + } else { + base_rd += RDCOST(x->rdmult, x->rddiv, cost4x4, 0); + } + if (base_rd < txfm_cache[TX_MODE_SELECT]) { + txfm_cache[TX_MODE_SELECT] = base_rd; + } + } + } } static __inline @@ -2553,17 +2645,26 @@ void vp8_cal_step_param(int sr, int *sp) { *sp = MAX_MVSEARCH_STEPS - 1 - step; } -static int vp8_rd_pick_best_mbsegmentation(VP8_COMP *cpi, MACROBLOCK *x, - int_mv *best_ref_mv, int_mv *second_best_ref_mv, int64_t best_rd, - int *mdcounts, int *returntotrate, - int *returnyrate, int *returndistortion, - int *skippable, int mvthresh, - int_mv seg_mvs[BLOCK_MAX_SEGMENTS - 1][16 /* n_blocks */][MAX_REF_FRAMES - 1]) { +static int rd_pick_best_mbsegmentation(VP8_COMP *cpi, MACROBLOCK *x, + int_mv *best_ref_mv, + int_mv *second_best_ref_mv, + int64_t best_rd, + int *mdcounts, + int *returntotrate, + int *returnyrate, + int *returndistortion, + int *skippable, int mvthresh, + int_mv seg_mvs[NB_PARTITIONINGS] + [16 /* n_blocks */] + [MAX_REF_FRAMES - 1], + int64_t txfm_cache[NB_TXFM_MODES]) { int i; BEST_SEG_INFO bsi; MB_MODE_INFO * mbmi = &x->e_mbd.mode_info_context->mbmi; vpx_memset(&bsi, 0, sizeof(bsi)); + for (i = 0; i < NB_TXFM_MODES; i++) + txfm_cache[i] = INT64_MAX; bsi.segment_rd = best_rd; bsi.ref_mv = best_ref_mv; @@ -2571,6 +2672,7 @@ static int vp8_rd_pick_best_mbsegmentation(VP8_COMP *cpi, MACROBLOCK *x, bsi.mvp.as_int = best_ref_mv->as_int; bsi.mvthresh = mvthresh; bsi.mdcounts = mdcounts; + bsi.txfm_size = TX_4X4; for (i = 0; i < 16; i++) bsi.modes[i] = ZERO4X4; @@ -2578,15 +2680,19 @@ static int vp8_rd_pick_best_mbsegmentation(VP8_COMP *cpi, MACROBLOCK *x, if (cpi->compressor_speed == 0) { /* for now, we will keep the original segmentation order when in best quality mode */ - rd_check_segment(cpi, x, &bsi, BLOCK_16X8, seg_mvs[BLOCK_16X8]); - rd_check_segment(cpi, x, &bsi, BLOCK_8X16, seg_mvs[BLOCK_8X16]); - rd_check_segment(cpi, x, &bsi, BLOCK_8X8, seg_mvs[BLOCK_8X8]); - rd_check_segment(cpi, x, &bsi, BLOCK_4X4, seg_mvs[BLOCK_4X4]); + rd_check_segment(cpi, x, &bsi, PARTITIONING_16X8, + seg_mvs[PARTITIONING_16X8], txfm_cache); + rd_check_segment(cpi, x, &bsi, PARTITIONING_8X16, + seg_mvs[PARTITIONING_8X16], txfm_cache); + rd_check_segment(cpi, x, &bsi, PARTITIONING_8X8, + seg_mvs[PARTITIONING_8X8], txfm_cache); + rd_check_segment(cpi, x, &bsi, PARTITIONING_4X4, + seg_mvs[PARTITIONING_4X4], txfm_cache); } else { int sr; - rd_check_segment(cpi, x, &bsi, BLOCK_8X8, seg_mvs[BLOCK_8X8]); - + rd_check_segment(cpi, x, &bsi, PARTITIONING_8X8, + seg_mvs[PARTITIONING_8X8], txfm_cache); if (bsi.segment_rd < best_rd) { int tmp_col_min = x->mv_col_min; @@ -2602,34 +2708,40 @@ static int vp8_rd_pick_best_mbsegmentation(VP8_COMP *cpi, MACROBLOCK *x, bsi.sv_mvp[2].as_int = bsi.mvs[8].as_int; bsi.sv_mvp[3].as_int = bsi.mvs[10].as_int; - /* Use 8x8 result as 16x8/8x16's predictor MV. Adjust search range according to the closeness of 2 MV. */ + /* Use 8x8 result as 16x8/8x16's predictor MV. Adjust search range + * according to the closeness of 2 MV. */ /* block 8X16 */ - { - sr = MAXF((abs(bsi.sv_mvp[0].as_mv.row - bsi.sv_mvp[2].as_mv.row)) >> 3, (abs(bsi.sv_mvp[0].as_mv.col - bsi.sv_mvp[2].as_mv.col)) >> 3); - vp8_cal_step_param(sr, &bsi.sv_istep[0]); + sr = MAXF((abs(bsi.sv_mvp[0].as_mv.row - bsi.sv_mvp[2].as_mv.row)) >> 3, + (abs(bsi.sv_mvp[0].as_mv.col - bsi.sv_mvp[2].as_mv.col)) >> 3); + vp8_cal_step_param(sr, &bsi.sv_istep[0]); - sr = MAXF((abs(bsi.sv_mvp[1].as_mv.row - bsi.sv_mvp[3].as_mv.row)) >> 3, (abs(bsi.sv_mvp[1].as_mv.col - bsi.sv_mvp[3].as_mv.col)) >> 3); - vp8_cal_step_param(sr, &bsi.sv_istep[1]); + sr = MAXF((abs(bsi.sv_mvp[1].as_mv.row - bsi.sv_mvp[3].as_mv.row)) >> 3, + (abs(bsi.sv_mvp[1].as_mv.col - bsi.sv_mvp[3].as_mv.col)) >> 3); + vp8_cal_step_param(sr, &bsi.sv_istep[1]); - rd_check_segment(cpi, x, &bsi, BLOCK_8X16, seg_mvs[BLOCK_8X16]); - } + rd_check_segment(cpi, x, &bsi, PARTITIONING_8X16, + seg_mvs[PARTITIONING_8X16], txfm_cache); /* block 16X8 */ - { - sr = MAXF((abs(bsi.sv_mvp[0].as_mv.row - bsi.sv_mvp[1].as_mv.row)) >> 3, (abs(bsi.sv_mvp[0].as_mv.col - bsi.sv_mvp[1].as_mv.col)) >> 3); - vp8_cal_step_param(sr, &bsi.sv_istep[0]); + sr = MAXF((abs(bsi.sv_mvp[0].as_mv.row - bsi.sv_mvp[1].as_mv.row)) >> 3, + (abs(bsi.sv_mvp[0].as_mv.col - bsi.sv_mvp[1].as_mv.col)) >> 3); + vp8_cal_step_param(sr, &bsi.sv_istep[0]); - sr = MAXF((abs(bsi.sv_mvp[2].as_mv.row - bsi.sv_mvp[3].as_mv.row)) >> 3, (abs(bsi.sv_mvp[2].as_mv.col - bsi.sv_mvp[3].as_mv.col)) >> 3); - vp8_cal_step_param(sr, &bsi.sv_istep[1]); + sr = MAXF((abs(bsi.sv_mvp[2].as_mv.row - bsi.sv_mvp[3].as_mv.row)) >> 3, + (abs(bsi.sv_mvp[2].as_mv.col - bsi.sv_mvp[3].as_mv.col)) >> 3); + vp8_cal_step_param(sr, &bsi.sv_istep[1]); - rd_check_segment(cpi, x, &bsi, BLOCK_16X8, seg_mvs[BLOCK_16X8]); - } + rd_check_segment(cpi, x, &bsi, PARTITIONING_16X8, + seg_mvs[PARTITIONING_16X8], txfm_cache); /* If 8x8 is better than 16x8/8x16, then do 4x4 search */ /* Not skip 4x4 if speed=0 (good quality) */ - if (cpi->sf.no_skip_block4x4_search || bsi.segment_num == BLOCK_8X8) { /* || (sv_segment_rd8x8-bsi.segment_rd) < sv_segment_rd8x8>>5) */ + if (cpi->sf.no_skip_block4x4_search || + bsi.segment_num == PARTITIONING_8X8) { + /* || (sv_segment_rd8x8-bsi.segment_rd) < sv_segment_rd8x8>>5) */ bsi.mvp.as_int = bsi.sv_mvp[0].as_int; - rd_check_segment(cpi, x, &bsi, BLOCK_4X4, seg_mvs[BLOCK_4X4]); + rd_check_segment(cpi, x, &bsi, PARTITIONING_4X4, + seg_mvs[PARTITIONING_4X4], txfm_cache); } /* restore UMV window */ @@ -2653,9 +2765,12 @@ static int vp8_rd_pick_best_mbsegmentation(VP8_COMP *cpi, MACROBLOCK *x, *returntotrate = bsi.r; *returndistortion = bsi.d; *returnyrate = bsi.segment_yrate; - *skippable = mby_is_skippable_4x4(&x->e_mbd, 0); + *skippable = bsi.txfm_size == TX_4X4 ? + mby_is_skippable_4x4(&x->e_mbd, 0) : + mby_is_skippable_8x8(&x->e_mbd, 0); /* save partitions */ + mbmi->txfm_size = bsi.txfm_size; mbmi->partitioning = bsi.segment_num; x->partition_info->count = vp8_mbsplit_count[bsi.segment_num]; @@ -2901,9 +3016,7 @@ void vp8_cal_sad(VP8_COMP *cpi, MACROBLOCKD *xd, MACROBLOCK *x, int recon_yoffse void rd_update_mvcount(VP8_COMP *cpi, MACROBLOCK *x, int_mv *best_ref_mv, int_mv *second_best_ref_mv) { MB_MODE_INFO * mbmi = &x->e_mbd.mode_info_context->mbmi; -#if CONFIG_NEWMVENTROPY MV mv; -#endif if (mbmi->mode == SPLITMV) { int i; @@ -2911,7 +3024,6 @@ void rd_update_mvcount(VP8_COMP *cpi, MACROBLOCK *x, for (i = 0; i < x->partition_info->count; i++) { if (x->partition_info->bmi[i].mode == NEW4X4) { if (x->e_mbd.allow_high_precision_mv) { -#if CONFIG_NEWMVENTROPY mv.row = (x->partition_info->bmi[i].mv.as_mv.row - best_ref_mv->as_mv.row); mv.col = (x->partition_info->bmi[i].mv.as_mv.col @@ -2925,20 +3037,7 @@ void rd_update_mvcount(VP8_COMP *cpi, MACROBLOCK *x, vp8_increment_nmv(&mv, &second_best_ref_mv->as_mv, &cpi->NMVcount, 1); } -#else - cpi->MVcount_hp[0][mv_max_hp + (x->partition_info->bmi[i].mv.as_mv.row - - best_ref_mv->as_mv.row)]++; - cpi->MVcount_hp[1][mv_max_hp + (x->partition_info->bmi[i].mv.as_mv.col - - best_ref_mv->as_mv.col)]++; - if (mbmi->second_ref_frame) { - cpi->MVcount_hp[0][mv_max_hp + (x->partition_info->bmi[i].second_mv.as_mv.row - - second_best_ref_mv->as_mv.row)]++; - cpi->MVcount_hp[1][mv_max_hp + (x->partition_info->bmi[i].second_mv.as_mv.col - - second_best_ref_mv->as_mv.col)]++; - } -#endif } else { -#if CONFIG_NEWMVENTROPY mv.row = (x->partition_info->bmi[i].mv.as_mv.row - best_ref_mv->as_mv.row); mv.col = (x->partition_info->bmi[i].mv.as_mv.col @@ -2952,24 +3051,11 @@ void rd_update_mvcount(VP8_COMP *cpi, MACROBLOCK *x, vp8_increment_nmv(&mv, &second_best_ref_mv->as_mv, &cpi->NMVcount, 0); } -#else - cpi->MVcount[0][mv_max + ((x->partition_info->bmi[i].mv.as_mv.row - - best_ref_mv->as_mv.row) >> 1)]++; - cpi->MVcount[1][mv_max + ((x->partition_info->bmi[i].mv.as_mv.col - - best_ref_mv->as_mv.col) >> 1)]++; - if (mbmi->second_ref_frame) { - cpi->MVcount[0][mv_max + ((x->partition_info->bmi[i].second_mv.as_mv.row - - second_best_ref_mv->as_mv.row) >> 1)]++; - cpi->MVcount[1][mv_max + ((x->partition_info->bmi[i].second_mv.as_mv.col - - second_best_ref_mv->as_mv.col) >> 1)]++; - } -#endif } } } } else if (mbmi->mode == NEWMV) { if (x->e_mbd.allow_high_precision_mv) { -#if CONFIG_NEWMVENTROPY mv.row = (mbmi->mv[0].as_mv.row - best_ref_mv->as_mv.row); mv.col = (mbmi->mv[0].as_mv.col - best_ref_mv->as_mv.col); vp8_increment_nmv(&mv, &best_ref_mv->as_mv, &cpi->NMVcount, 1); @@ -2978,20 +3064,7 @@ void rd_update_mvcount(VP8_COMP *cpi, MACROBLOCK *x, mv.col = (mbmi->mv[1].as_mv.col - second_best_ref_mv->as_mv.col); vp8_increment_nmv(&mv, &second_best_ref_mv->as_mv, &cpi->NMVcount, 1); } -#else - cpi->MVcount_hp[0][mv_max_hp + (mbmi->mv[0].as_mv.row - - best_ref_mv->as_mv.row)]++; - cpi->MVcount_hp[1][mv_max_hp + (mbmi->mv[0].as_mv.col - - best_ref_mv->as_mv.col)]++; - if (mbmi->second_ref_frame) { - cpi->MVcount_hp[0][mv_max_hp + (mbmi->mv[1].as_mv.row - - second_best_ref_mv->as_mv.row)]++; - cpi->MVcount_hp[1][mv_max_hp + (mbmi->mv[1].as_mv.col - - second_best_ref_mv->as_mv.col)]++; - } -#endif } else { -#if CONFIG_NEWMVENTROPY mv.row = (mbmi->mv[0].as_mv.row - best_ref_mv->as_mv.row); mv.col = (mbmi->mv[0].as_mv.col - best_ref_mv->as_mv.col); vp8_increment_nmv(&mv, &best_ref_mv->as_mv, &cpi->NMVcount, 0); @@ -3000,18 +3073,6 @@ void rd_update_mvcount(VP8_COMP *cpi, MACROBLOCK *x, mv.col = (mbmi->mv[1].as_mv.col - second_best_ref_mv->as_mv.col); vp8_increment_nmv(&mv, &second_best_ref_mv->as_mv, &cpi->NMVcount, 0); } -#else - cpi->MVcount[0][mv_max + ((mbmi->mv[0].as_mv.row - - best_ref_mv->as_mv.row) >> 1)]++; - cpi->MVcount[1][mv_max + ((mbmi->mv[0].as_mv.col - - best_ref_mv->as_mv.col) >> 1)]++; - if (mbmi->second_ref_frame) { - cpi->MVcount[0][mv_max + ((mbmi->mv[1].as_mv.row - - second_best_ref_mv->as_mv.row) >> 1)]++; - cpi->MVcount[1][mv_max + ((mbmi->mv[1].as_mv.col - - second_best_ref_mv->as_mv.col) >> 1)]++; - } -#endif } } } @@ -3185,9 +3246,7 @@ static void store_coding_context(MACROBLOCK *x, PICK_MODE_CONTEXT *ctx, int hybrid_pred_diff, int64_t txfm_size_diff[NB_TXFM_MODES]) { MACROBLOCKD *xd = &x->e_mbd; -#if CONFIG_TX_SELECT MB_MODE_INFO *mbmi = &xd->mode_info_context->mbmi; -#endif // Take a snapshot of the coding context so it can be // restored if we decide to encode this way @@ -3207,9 +3266,7 @@ static void store_coding_context(MACROBLOCK *x, PICK_MODE_CONTEXT *ctx, ctx->comp_pred_diff = comp_pred_diff; ctx->hybrid_pred_diff = hybrid_pred_diff; -#if CONFIG_TX_SELECT memcpy(ctx->txfm_rd_diff, txfm_size_diff, sizeof(ctx->txfm_rd_diff)); -#endif } static void inter_mode_cost(VP8_COMP *cpi, MACROBLOCK *x, int this_mode, @@ -3326,9 +3383,7 @@ void vp8_rd_pick_inter_mode(VP8_COMP *cpi, MACROBLOCK *x, int recon_yoffset, int int_mv ref_mv[MAX_REF_FRAMES] = {{0}}; #endif -#if CONFIG_SWITCHABLE_INTERP int switchable_filter_index = 0; -#endif MB_PREDICTION_MODE uv_intra_mode; MB_PREDICTION_MODE uv_intra_mode_8x8 = 0; @@ -3344,7 +3399,7 @@ void vp8_rd_pick_inter_mode(VP8_COMP *cpi, MACROBLOCK *x, int recon_yoffset, int unsigned char *y_buffer[4], *u_buffer[4], *v_buffer[4]; unsigned int ref_costs[MAX_REF_FRAMES]; - int_mv seg_mvs[BLOCK_MAX_SEGMENTS - 1][16 /* n_blocks */][MAX_REF_FRAMES - 1]; + int_mv seg_mvs[NB_PARTITIONINGS][16 /* n_blocks */][MAX_REF_FRAMES - 1]; vpx_memset(mode8x8, 0, sizeof(mode8x8)); vpx_memset(&frame_mv, 0, sizeof(frame_mv)); @@ -3359,7 +3414,7 @@ void vp8_rd_pick_inter_mode(VP8_COMP *cpi, MACROBLOCK *x, int recon_yoffset, int for (i = 0; i < NB_TXFM_MODES; i++) best_txfm_rd[i] = INT64_MAX; - for (i = 0; i < BLOCK_MAX_SEGMENTS - 1; i++) { + for (i = 0; i < NB_PARTITIONINGS; i++) { int j, k; for (j = 0; j < 16; j++) @@ -3425,12 +3480,8 @@ void vp8_rd_pick_inter_mode(VP8_COMP *cpi, MACROBLOCK *x, int recon_yoffset, int // that depend on the current prediction etc. vp8_estimate_ref_frame_costs(cpi, segment_id, ref_costs); -#if CONFIG_SWITCHABLE_INTERP for (mode_index = 0; mode_index < MAX_MODES; mode_index += (!switchable_filter_index)) { -#else - for (mode_index = 0; mode_index < MAX_MODES; ++mode_index) { -#endif int64_t this_rd = INT64_MAX; int is_comp_pred; int disable_skip = 0, skippable = 0; @@ -3458,19 +3509,16 @@ void vp8_rd_pick_inter_mode(VP8_COMP *cpi, MACROBLOCK *x, int recon_yoffset, int #if CONFIG_PRED_FILTER mbmi->pred_filter_enabled = 0; #endif -#if CONFIG_SWITCHABLE_INTERP if (cpi->common.mcomp_filter_type == SWITCHABLE && this_mode >= NEARESTMV && this_mode <= SPLITMV) { mbmi->interp_filter = vp8_switchable_interp[switchable_filter_index++]; if (switchable_filter_index == VP8_SWITCHABLE_FILTERS) switchable_filter_index = 0; - //printf("Searching %d (%d)\n", this_mode, switchable_filter_index); } else { mbmi->interp_filter = cpi->common.mcomp_filter_type; } vp8_setup_interp_filters(xd, mbmi->interp_filter, &cpi->common); -#endif // Test best rd so far against threshold for trying this mode. if (best_rd <= cpi->rd_threshes[mode_index]) @@ -3612,11 +3660,9 @@ void vp8_rd_pick_inter_mode(VP8_COMP *cpi, MACROBLOCK *x, int recon_yoffset, int } break; case I8X8_PRED: { -#if CONFIG_TX_SELECT int cost0 = vp8_cost_bit(cm->prob_tx[0], 0); int cost1 = vp8_cost_bit(cm->prob_tx[0], 1); int64_t tmp_rd_4x4s, tmp_rd_8x8s; -#endif int64_t tmp_rd_4x4, tmp_rd_8x8, tmp_rd; int r4x4, tok4x4, d4x4, r8x8, tok8x8, d8x8; mbmi->txfm_size = TX_4X4; @@ -3638,7 +3684,6 @@ void vp8_rd_pick_inter_mode(VP8_COMP *cpi, MACROBLOCK *x, int recon_yoffset, int txfm_cache[ONLY_4X4] = tmp_rd_4x4; txfm_cache[ALLOW_8X8] = tmp_rd_8x8; txfm_cache[ALLOW_16X16] = tmp_rd_8x8; -#if CONFIG_TX_SELECT tmp_rd_4x4s = tmp_rd_4x4 + RDCOST(x->rdmult, x->rddiv, cost0, 0); tmp_rd_8x8s = tmp_rd_8x8 + RDCOST(x->rdmult, x->rddiv, cost1, 0); txfm_cache[TX_MODE_SELECT] = tmp_rd_4x4s < tmp_rd_8x8s ? tmp_rd_4x4s : tmp_rd_8x8s; @@ -3667,9 +3712,7 @@ void vp8_rd_pick_inter_mode(VP8_COMP *cpi, MACROBLOCK *x, int recon_yoffset, int mode8x8[1][3] = x->e_mbd.mode_info_context->bmi[10].as_mode.second; #endif } - } else -#endif - if (cm->txfm_mode == ONLY_4X4) { + } else if (cm->txfm_mode == ONLY_4X4) { rate = r4x4; rate_y = tok4x4; distortion = d4x4; @@ -3725,21 +3768,19 @@ void vp8_rd_pick_inter_mode(VP8_COMP *cpi, MACROBLOCK *x, int recon_yoffset, int (mbmi->ref_frame == GOLDEN_FRAME) ? cpi->rd_threshes[THR_NEWG] : this_rd_thresh; - mbmi->txfm_size = TX_4X4; // FIXME use 8x8 in case of 8x8/8x16/16x8 - tmp_rd = vp8_rd_pick_best_mbsegmentation(cpi, x, &best_ref_mv, - second_ref, best_yrd, mdcounts, - &rate, &rate_y, &distortion, - &skippable, - this_rd_thresh, seg_mvs); + tmp_rd = rd_pick_best_mbsegmentation(cpi, x, &best_ref_mv, + second_ref, best_yrd, mdcounts, + &rate, &rate_y, &distortion, + &skippable, + this_rd_thresh, seg_mvs, + txfm_cache); rate2 += rate; distortion2 += distortion; -#if CONFIG_SWITCHABLE_INTERP if (cpi->common.mcomp_filter_type == SWITCHABLE) rate2 += SWITCHABLE_INTERP_RATE_FACTOR * x->switchable_interp_costs [get_pred_context(&cpi->common, xd, PRED_SWITCHABLE_INTERP)] [vp8_switchable_interp_map[mbmi->interp_filter]]; -#endif // If even the 'Y' rd value of split is higher than best so far // then dont bother looking at UV if (tmp_rd < best_yrd) { @@ -3877,13 +3918,11 @@ void vp8_rd_pick_inter_mode(VP8_COMP *cpi, MACROBLOCK *x, int recon_yoffset, int rate2 += vp8_cost_bit(cpi->common.prob_pred_filter_off, xd->mode_info_context->mbmi.pred_filter_enabled); #endif -#if CONFIG_SWITCHABLE_INTERP if (cpi->common.mcomp_filter_type == SWITCHABLE) rate2 += SWITCHABLE_INTERP_RATE_FACTOR * x->switchable_interp_costs [get_pred_context(&cpi->common, xd, PRED_SWITCHABLE_INTERP)] [vp8_switchable_interp_map[ x->e_mbd.mode_info_context->mbmi.interp_filter]]; -#endif /* We don't include the cost of the second reference here, because there are only * three options: Last/Golden, ARF/Last or Golden/ARF, or in other words if you @@ -3908,8 +3947,7 @@ void vp8_rd_pick_inter_mode(VP8_COMP *cpi, MACROBLOCK *x, int recon_yoffset, int if (threshold < x->encode_breakout) threshold = x->encode_breakout; - var = VARIANCE_INVOKE(&cpi->rtcd.variance, var16x16) - (*(b->base_src), b->src_stride, + var = vp8_variance16x16(*(b->base_src), b->src_stride, x->e_mbd.predictor, 16, &sse); if (sse < threshold) { @@ -3919,7 +3957,7 @@ void vp8_rd_pick_inter_mode(VP8_COMP *cpi, MACROBLOCK *x, int recon_yoffset, int if ((sse - var < q2dc *q2dc >> 4) || (sse / 2 > var && sse - var < 64)) { // Check u and v to make sure skip is ok - int sse2 = VP8_UVSSE(x, IF_RTCD(&cpi->rtcd.variance)); + int sse2 = vp8_uvsse(x); if (sse2 * 2 < threshold) { x->skip = 1; distortion2 = sse + sse2; @@ -4127,7 +4165,7 @@ void vp8_rd_pick_inter_mode(VP8_COMP *cpi, MACROBLOCK *x, int recon_yoffset, int if (!mode_excluded && this_rd != INT64_MAX) { for (i = 0; i < NB_TXFM_MODES; i++) { int64_t adj_rd; - if (this_mode != B_PRED && this_mode != SPLITMV) { + if (this_mode != B_PRED) { adj_rd = this_rd + txfm_cache[i] - txfm_cache[cm->txfm_mode]; } else { adj_rd = this_rd; @@ -4151,7 +4189,6 @@ void vp8_rd_pick_inter_mode(VP8_COMP *cpi, MACROBLOCK *x, int recon_yoffset, int else ++cpi->pred_filter_off_count; #endif -#if CONFIG_SWITCHABLE_INTERP if (cpi->common.mcomp_filter_type == SWITCHABLE && best_mbmode.mode >= NEARESTMV && best_mbmode.mode <= SPLITMV) { @@ -4159,7 +4196,6 @@ void vp8_rd_pick_inter_mode(VP8_COMP *cpi, MACROBLOCK *x, int recon_yoffset, int [get_pred_context(&cpi->common, xd, PRED_SWITCHABLE_INTERP)] [vp8_switchable_interp_map[best_mbmode.interp_filter]]; } -#endif // Reduce the activation RD thresholds for the best choice mode if ((cpi->rd_baseline_thresh[best_mode_index] > 0) && @@ -4185,11 +4221,9 @@ void vp8_rd_pick_inter_mode(VP8_COMP *cpi, MACROBLOCK *x, int recon_yoffset, int (cpi->oxcf.arnr_max_frames == 0) && (best_mbmode.mode != ZEROMV || best_mbmode.ref_frame != ALTREF_FRAME)) { mbmi->mode = ZEROMV; -#if CONFIG_TX_SELECT if (cm->txfm_mode != TX_MODE_SELECT) mbmi->txfm_size = cm->txfm_mode; else -#endif mbmi->txfm_size = TX_16X16; mbmi->ref_frame = ALTREF_FRAME; mbmi->mv[0].as_int = 0; @@ -4239,7 +4273,6 @@ void vp8_rd_pick_inter_mode(VP8_COMP *cpi, MACROBLOCK *x, int recon_yoffset, int best_pred_diff[i] = best_rd - best_pred_rd[i]; } -#if CONFIG_TX_SELECT if (!x->skip) { for (i = 0; i < NB_TXFM_MODES; i++) { if (best_txfm_rd[i] == INT64_MAX) @@ -4250,7 +4283,6 @@ void vp8_rd_pick_inter_mode(VP8_COMP *cpi, MACROBLOCK *x, int recon_yoffset, int } else { vpx_memset(best_txfm_diff, 0, sizeof(best_txfm_diff)); } -#endif end: store_coding_context(x, &x->mb_context[xd->mb_index], best_mode_index, &best_partition, @@ -4381,10 +4413,8 @@ void vp8_rd_pick_intra_mode(VP8_COMP *cpi, MACROBLOCK *x, vp8_cost_bit(get_pred_prob(cm, xd, PRED_MBSKIP), 1); dist = dist16x16 + (distuv8x8 >> 2); mbmi->txfm_size = txfm_size_16x16; -#if CONFIG_TX_SELECT memset(x->mb_context[xd->mb_index].txfm_rd_diff, 0, sizeof(x->mb_context[xd->mb_index].txfm_rd_diff)); -#endif } else if (error8x8 > error16x16) { if (error4x4 < error16x16) { rate = rateuv; @@ -4401,20 +4431,16 @@ void vp8_rd_pick_intra_mode(VP8_COMP *cpi, MACROBLOCK *x, mbmi->mode = B_PRED; mbmi->txfm_size = TX_4X4; dist = dist4x4 + (distuv >> 2); -#if CONFIG_TX_SELECT memset(x->mb_context[xd->mb_index].txfm_rd_diff, 0, sizeof(x->mb_context[xd->mb_index].txfm_rd_diff)); -#endif } else { mbmi->txfm_size = txfm_size_16x16; mbmi->mode = mode16x16; rate = rate16x16 + rateuv8x8; dist = dist16x16 + (distuv8x8 >> 2); -#if CONFIG_TX_SELECT for (i = 0; i < NB_TXFM_MODES; i++) { x->mb_context[xd->mb_index].txfm_rd_diff[i] = error16x16 - txfm_cache[i]; } -#endif } if (cpi->common.mb_no_coeff_skip) rate += vp8_cost_bit(get_pred_prob(cm, xd, PRED_MBSKIP), 0); @@ -4434,10 +4460,8 @@ void vp8_rd_pick_intra_mode(VP8_COMP *cpi, MACROBLOCK *x, mbmi->mode = B_PRED; mbmi->txfm_size = TX_4X4; dist = dist4x4 + (distuv >> 2); -#if CONFIG_TX_SELECT memset(x->mb_context[xd->mb_index].txfm_rd_diff, 0, sizeof(x->mb_context[xd->mb_index].txfm_rd_diff)); -#endif } else { // FIXME(rbultje) support transform-size selection mbmi->mode = I8X8_PRED; @@ -4445,10 +4469,8 @@ void vp8_rd_pick_intra_mode(VP8_COMP *cpi, MACROBLOCK *x, set_i8x8_block_modes(x, mode8x8); rate = rate8x8 + rateuv; dist = dist8x8 + (distuv >> 2); -#if CONFIG_TX_SELECT memset(x->mb_context[xd->mb_index].txfm_rd_diff, 0, sizeof(x->mb_context[xd->mb_index].txfm_rd_diff)); -#endif } if (cpi->common.mb_no_coeff_skip) rate += vp8_cost_bit(get_pred_prob(cm, xd, PRED_MBSKIP), 0); @@ -4805,8 +4827,8 @@ int64_t vp8_rd_pick_inter_mode_sb(VP8_COMP *cpi, MACROBLOCK *x, if (threshold < x->encode_breakout) threshold = x->encode_breakout; - var = VARIANCE_INVOKE(&cpi->rtcd.variance, var32x32)(*(b->base_src), - b->src_stride, xd->dst.y_buffer, xd->dst.y_stride, &sse); + var = vp8_variance32x32(*(b->base_src), b->src_stride, + xd->dst.y_buffer, xd->dst.y_stride, &sse); if (sse < threshold) { unsigned int q2dc = xd->block[24].dequant[0]; @@ -4816,11 +4838,9 @@ int64_t vp8_rd_pick_inter_mode_sb(VP8_COMP *cpi, MACROBLOCK *x, (sse / 2 > var && sse - var < 64)) { // Check u and v to make sure skip is ok unsigned int sse2, sse3; - var += VARIANCE_INVOKE(&cpi->rtcd.variance, var16x16) - (x->src.u_buffer, x->src.uv_stride, + var += vp8_variance16x16(x->src.u_buffer, x->src.uv_stride, xd->dst.u_buffer, xd->dst.uv_stride, &sse2); - var += VARIANCE_INVOKE(&cpi->rtcd.variance, var16x16) - (x->src.v_buffer, x->src.uv_stride, + var += vp8_variance16x16(x->src.v_buffer, x->src.uv_stride, xd->dst.v_buffer, xd->dst.uv_stride, &sse3); sse2 += sse3; if (sse2 * 2 < threshold) { diff --git a/vp8/encoder/sad_c.c b/vp8/encoder/sad_c.c index 2e86a16c0..f15e687c6 100644 --- a/vp8/encoder/sad_c.c +++ b/vp8/encoder/sad_c.c @@ -10,33 +10,10 @@ #include <stdlib.h> +#include "vp8/common/sadmxn.h" #include "vpx_ports/config.h" #include "vpx/vpx_integer.h" -static __inline -unsigned int sad_mx_n_c( - const unsigned char *src_ptr, - int src_stride, - const unsigned char *ref_ptr, - int ref_stride, - int m, - int n) { - - int r, c; - unsigned int sad = 0; - - for (r = 0; r < n; r++) { - for (c = 0; c < m; c++) { - sad += abs(src_ptr[c] - ref_ptr[c]); - } - - src_ptr += src_stride; - ref_ptr += ref_stride; - } - - return sad; -} - unsigned int vp8_sad32x32_c(const unsigned char *src_ptr, int src_stride, const unsigned char *ref_ptr, @@ -97,25 +74,6 @@ unsigned int vp8_sad4x4_c( return sad_mx_n_c(src_ptr, src_stride, ref_ptr, ref_stride, 4, 4); } -#if CONFIG_NEWBESTREFMV -unsigned int vp8_sad3x16_c( - const unsigned char *src_ptr, - int src_stride, - const unsigned char *ref_ptr, - int ref_stride, - int max_sad){ - return sad_mx_n_c(src_ptr, src_stride, ref_ptr, ref_stride, 3, 16); -} -unsigned int vp8_sad16x3_c( - const unsigned char *src_ptr, - int src_stride, - const unsigned char *ref_ptr, - int ref_stride, - int max_sad){ - return sad_mx_n_c(src_ptr, src_stride, ref_ptr, ref_stride, 16, 3); -} -#endif - void vp8_sad32x32x3_c(const unsigned char *src_ptr, int src_stride, const unsigned char *ref_ptr, diff --git a/vp8/encoder/ssim.c b/vp8/encoder/ssim.c index d3d9711dc..865496ae2 100644 --- a/vp8/encoder/ssim.c +++ b/vp8/encoder/ssim.c @@ -11,18 +11,10 @@ #include "onyx_int.h" -void vp8_ssim_parms_16x16_c -( - unsigned char *s, - int sp, - unsigned char *r, - int rp, - unsigned long *sum_s, - unsigned long *sum_r, - unsigned long *sum_sq_s, - unsigned long *sum_sq_r, - unsigned long *sum_sxr -) { +void vp8_ssim_parms_16x16_c(unsigned char *s, int sp, unsigned char *r, + int rp, unsigned long *sum_s, unsigned long *sum_r, + unsigned long *sum_sq_s, unsigned long *sum_sq_r, + unsigned long *sum_sxr) { int i, j; for (i = 0; i < 16; i++, s += sp, r += rp) { for (j = 0; j < 16; j++) { @@ -34,18 +26,10 @@ void vp8_ssim_parms_16x16_c } } } -void vp8_ssim_parms_8x8_c -( - unsigned char *s, - int sp, - unsigned char *r, - int rp, - unsigned long *sum_s, - unsigned long *sum_r, - unsigned long *sum_sq_s, - unsigned long *sum_sq_r, - unsigned long *sum_sxr -) { +void vp8_ssim_parms_8x8_c(unsigned char *s, int sp, unsigned char *r, int rp, + unsigned long *sum_s, unsigned long *sum_r, + unsigned long *sum_sq_s, unsigned long *sum_sq_r, + unsigned long *sum_sxr) { int i, j; for (i = 0; i < 8; i++, s += sp, r += rp) { for (j = 0; j < 8; j++) { @@ -61,15 +45,9 @@ void vp8_ssim_parms_8x8_c const static int64_t cc1 = 26634; // (64^2*(.01*255)^2 const static int64_t cc2 = 239708; // (64^2*(.03*255)^2 -static double similarity -( - unsigned long sum_s, - unsigned long sum_r, - unsigned long sum_sq_s, - unsigned long sum_sq_r, - unsigned long sum_sxr, - int count -) { +static double similarity(unsigned long sum_s, unsigned long sum_r, + unsigned long sum_sq_s, unsigned long sum_sq_r, + unsigned long sum_sxr, int count) { int64_t ssim_n, ssim_d; int64_t c1, c2; @@ -87,23 +65,22 @@ static double similarity return ssim_n * 1.0 / ssim_d; } -static double ssim_16x16(unsigned char *s, int sp, unsigned char *r, int rp, - const vp8_variance_rtcd_vtable_t *rtcd) { +static double ssim_16x16(unsigned char *s, int sp, unsigned char *r, int rp) { unsigned long sum_s = 0, sum_r = 0, sum_sq_s = 0, sum_sq_r = 0, sum_sxr = 0; - SSIMPF_INVOKE(rtcd, 16x16)(s, sp, r, rp, &sum_s, &sum_r, &sum_sq_s, &sum_sq_r, &sum_sxr); + vp8_ssim_parms_16x16(s, sp, r, rp, &sum_s, &sum_r, &sum_sq_s, &sum_sq_r, + &sum_sxr); return similarity(sum_s, sum_r, sum_sq_s, sum_sq_r, sum_sxr, 256); } -static double ssim_8x8(unsigned char *s, int sp, unsigned char *r, int rp, - const vp8_variance_rtcd_vtable_t *rtcd) { +static double ssim_8x8(unsigned char *s, int sp, unsigned char *r, int rp) { unsigned long sum_s = 0, sum_r = 0, sum_sq_s = 0, sum_sq_r = 0, sum_sxr = 0; - SSIMPF_INVOKE(rtcd, 8x8)(s, sp, r, rp, &sum_s, &sum_r, &sum_sq_s, &sum_sq_r, &sum_sxr); + vp8_ssim_parms_8x8(s, sp, r, rp, &sum_s, &sum_r, &sum_sq_s, &sum_sq_r, + &sum_sxr); return similarity(sum_s, sum_r, sum_sq_s, sum_sq_r, sum_sxr, 64); } // TODO: (jbb) tried to scale this function such that we may be able to use it // for distortion metric in mode selection code ( provided we do a reconstruction) -long dssim(unsigned char *s, int sp, unsigned char *r, int rp, - const vp8_variance_rtcd_vtable_t *rtcd) { +long dssim(unsigned char *s, int sp, unsigned char *r, int rp) { unsigned long sum_s = 0, sum_r = 0, sum_sq_s = 0, sum_sq_r = 0, sum_sxr = 0; int64_t ssim3; int64_t ssim_n1, ssim_n2; @@ -115,7 +92,8 @@ long dssim(unsigned char *s, int sp, unsigned char *r, int rp, c1 = cc1 * 16; c2 = cc2 * 16; - SSIMPF_INVOKE(rtcd, 16x16)(s, sp, r, rp, &sum_s, &sum_r, &sum_sq_s, &sum_sq_r, &sum_sxr); + vp8_ssim_parms_16x16(s, sp, r, rp, &sum_s, &sum_r, &sum_sq_s, &sum_sq_r, + &sum_sxr); ssim_n1 = (2 * sum_s * sum_r + c1); ssim_n2 = ((int64_t) 2 * 256 * sum_sxr - (int64_t) 2 * sum_s * sum_r + c2); @@ -137,16 +115,8 @@ long dssim(unsigned char *s, int sp, unsigned char *r, int rp, // We are using a 8x8 moving window with starting location of each 8x8 window // on the 4x4 pixel grid. Such arrangement allows the windows to overlap // block boundaries to penalize blocking artifacts. -double vp8_ssim2 -( - unsigned char *img1, - unsigned char *img2, - int stride_img1, - int stride_img2, - int width, - int height, - const vp8_variance_rtcd_vtable_t *rtcd -) { +double vp8_ssim2(unsigned char *img1, unsigned char *img2, int stride_img1, + int stride_img2, int width, int height) { int i, j; int samples = 0; double ssim_total = 0; @@ -154,7 +124,7 @@ double vp8_ssim2 // sample point start with each 4x4 location for (i = 0; i < height - 8; i += 4, img1 += stride_img1 * 4, img2 += stride_img2 * 4) { for (j = 0; j < width - 8; j += 4) { - double v = ssim_8x8(img1 + j, stride_img1, img2 + j, stride_img2, rtcd); + double v = ssim_8x8(img1 + j, stride_img1, img2 + j, stride_img2); ssim_total += v; samples++; } @@ -162,28 +132,22 @@ double vp8_ssim2 ssim_total /= samples; return ssim_total; } -double vp8_calc_ssim -( - YV12_BUFFER_CONFIG *source, - YV12_BUFFER_CONFIG *dest, - int lumamask, - double *weight, - const vp8_variance_rtcd_vtable_t *rtcd -) { +double vp8_calc_ssim(YV12_BUFFER_CONFIG *source, YV12_BUFFER_CONFIG *dest, + int lumamask, double *weight) { double a, b, c; double ssimv; a = vp8_ssim2(source->y_buffer, dest->y_buffer, source->y_stride, dest->y_stride, source->y_width, - source->y_height, rtcd); + source->y_height); b = vp8_ssim2(source->u_buffer, dest->u_buffer, source->uv_stride, dest->uv_stride, source->uv_width, - source->uv_height, rtcd); + source->uv_height); c = vp8_ssim2(source->v_buffer, dest->v_buffer, source->uv_stride, dest->uv_stride, source->uv_width, - source->uv_height, rtcd); + source->uv_height); ssimv = a * .8 + .1 * (b + c); @@ -192,29 +156,22 @@ double vp8_calc_ssim return ssimv; } -double vp8_calc_ssimg -( - YV12_BUFFER_CONFIG *source, - YV12_BUFFER_CONFIG *dest, - double *ssim_y, - double *ssim_u, - double *ssim_v, - const vp8_variance_rtcd_vtable_t *rtcd -) { +double vp8_calc_ssimg(YV12_BUFFER_CONFIG *source, YV12_BUFFER_CONFIG *dest, + double *ssim_y, double *ssim_u, double *ssim_v) { double ssim_all = 0; double a, b, c; a = vp8_ssim2(source->y_buffer, dest->y_buffer, source->y_stride, dest->y_stride, source->y_width, - source->y_height, rtcd); + source->y_height); b = vp8_ssim2(source->u_buffer, dest->u_buffer, source->uv_stride, dest->uv_stride, source->uv_width, - source->uv_height, rtcd); + source->uv_height); c = vp8_ssim2(source->v_buffer, dest->v_buffer, source->uv_stride, dest->uv_stride, source->uv_width, - source->uv_height, rtcd); + source->uv_height); *ssim_y = a; *ssim_u = b; *ssim_v = c; diff --git a/vp8/encoder/tokenize.c b/vp8/encoder/tokenize.c index d46637a3e..2ddae1cbd 100644 --- a/vp8/encoder/tokenize.c +++ b/vp8/encoder/tokenize.c @@ -26,38 +26,26 @@ #ifdef ENTROPY_STATS INT64 context_counters[BLOCK_TYPES] [COEF_BANDS] [PREV_COEF_CONTEXTS] [MAX_ENTROPY_TOKENS]; -#if CONFIG_HYBRIDTRANSFORM INT64 hybrid_context_counters[BLOCK_TYPES] [COEF_BANDS] [PREV_COEF_CONTEXTS] [MAX_ENTROPY_TOKENS]; -#endif INT64 context_counters_8x8[BLOCK_TYPES_8X8] [COEF_BANDS] [PREV_COEF_CONTEXTS] [MAX_ENTROPY_TOKENS]; -#if CONFIG_HYBRIDTRANSFORM8X8 INT64 hybrid_context_counters_8x8[BLOCK_TYPES_8X8] [COEF_BANDS] [PREV_COEF_CONTEXTS] [MAX_ENTROPY_TOKENS]; -#endif INT64 context_counters_16x16[BLOCK_TYPES_16X16] [COEF_BANDS] [PREV_COEF_CONTEXTS] [MAX_ENTROPY_TOKENS]; -#if CONFIG_HYBRIDTRANSFORM16X16 INT64 hybrid_context_counters_16x16[BLOCK_TYPES_16X16] [COEF_BANDS] [PREV_COEF_CONTEXTS] [MAX_ENTROPY_TOKENS]; -#endif extern unsigned int tree_update_hist[BLOCK_TYPES][COEF_BANDS] [PREV_COEF_CONTEXTS][ENTROPY_NODES][2]; -#if CONFIG_HYBRIDTRANSFORM extern unsigned int hybrid_tree_update_hist[BLOCK_TYPES][COEF_BANDS] [PREV_COEF_CONTEXTS][ENTROPY_NODES][2]; -#endif extern unsigned int tree_update_hist_8x8[BLOCK_TYPES_8X8][COEF_BANDS] [PREV_COEF_CONTEXTS][ENTROPY_NODES] [2]; -#if CONFIG_HYBRIDTRANSFORM8X8 extern unsigned int hybrid_tree_update_hist_8x8[BLOCK_TYPES_8X8][COEF_BANDS] [PREV_COEF_CONTEXTS][ENTROPY_NODES] [2]; -#endif extern unsigned int tree_update_hist_16x16[BLOCK_TYPES_16X16][COEF_BANDS] [PREV_COEF_CONTEXTS][ENTROPY_NODES] [2]; -#if CONFIG_HYBRIDTRANSFORM16X16 extern unsigned int hybrid_tree_update_hist_16x16[BLOCK_TYPES_16X16][COEF_BANDS] [PREV_COEF_CONTEXTS][ENTROPY_NODES] [2]; -#endif #endif /* ENTROPY_STATS */ void vp8_stuff_mb(VP8_COMP *cpi, MACROBLOCKD *xd, TOKENEXTRA **t, int dry_run); @@ -134,9 +122,7 @@ static void tokenize1st_order_b_16x16(MACROBLOCKD *xd, const int eob = b->eob; /* one beyond last nonzero coeff */ TOKENEXTRA *t = *tp; /* store tokens starting here */ const short *qcoeff_ptr = b->qcoeff; -#if CONFIG_HYBRIDTRANSFORM16X16 TX_TYPE tx_type = get_tx_type(xd, b); -#endif int seg_eob = 256; int segment_id = xd->mode_info_context->mbmi.segment_id; @@ -162,22 +148,18 @@ static void tokenize1st_order_b_16x16(MACROBLOCKD *xd, } t->Token = x; -#if CONFIG_HYBRIDTRANSFORM16X16 if (tx_type != DCT_DCT) t->context_tree = cpi->common.fc.hybrid_coef_probs_16x16[type][band][pt]; else -#endif t->context_tree = cpi->common.fc.coef_probs_16x16[type][band][pt]; t->skip_eob_node = pt == 0 && ((band > 0 && type != PLANE_TYPE_Y_NO_DC) || (band > 1 && type == PLANE_TYPE_Y_NO_DC)); assert(vp8_coef_encodings[t->Token].Len - t->skip_eob_node > 0); if (!dry_run) { -#if CONFIG_HYBRIDTRANSFORM16X16 if (tx_type != DCT_DCT) ++cpi->hybrid_coef_counts_16x16[type][band][pt][x]; else -#endif ++cpi->coef_counts_16x16[type][band][pt][x]; } pt = vp8_prev_token_class[x]; @@ -310,9 +292,7 @@ static void tokenize1st_order_b_8x8(MACROBLOCKD *xd, int c = (type == PLANE_TYPE_Y_NO_DC) ? 1 : 0; /* start at DC unless type 0 */ TOKENEXTRA *t = *tp; /* store tokens starting here */ const short *qcoeff_ptr = b->qcoeff; -#if CONFIG_HYBRIDTRANSFORM8X8 TX_TYPE tx_type = get_tx_type(xd, b); -#endif const int eob = b->eob; int seg_eob = 64; int segment_id = xd->mode_info_context->mbmi.segment_id; @@ -338,11 +318,9 @@ static void tokenize1st_order_b_8x8(MACROBLOCKD *xd, x = DCT_EOB_TOKEN; t->Token = x; -#if CONFIG_HYBRIDTRANSFORM8X8 if (tx_type != DCT_DCT) t->context_tree = cpi->common.fc.hybrid_coef_probs_8x8[type][band][pt]; else -#endif t->context_tree = cpi->common.fc.coef_probs_8x8[type][band][pt]; t->skip_eob_node = pt == 0 && ((band > 0 && type != PLANE_TYPE_Y_NO_DC) || @@ -350,11 +328,9 @@ static void tokenize1st_order_b_8x8(MACROBLOCKD *xd, assert(vp8_coef_encodings[t->Token].Len - t->skip_eob_node > 0); if (!dry_run) { -#if CONFIG_HYBRIDTRANSFORM8X8 if (tx_type != DCT_DCT) ++cpi->hybrid_coef_counts_8x8[type][band][pt][x]; else -#endif ++cpi->coef_counts_8x8[type][band][pt][x]; } pt = vp8_prev_token_class[x]; @@ -451,7 +427,6 @@ static void tokenize1st_order_b_4x4(MACROBLOCKD *xd, const int16_t *qcoeff_ptr = b->qcoeff; int c = (type == PLANE_TYPE_Y_NO_DC) ? 1 : 0; -#if CONFIG_HYBRIDTRANSFORM TX_TYPE tx_type = get_tx_type(xd, &xd->block[block]); switch (tx_type) { case ADST_DCT: @@ -464,7 +439,6 @@ static void tokenize1st_order_b_4x4(MACROBLOCKD *xd, pt_scan = vp8_default_zig_zag1d; break; } -#endif a = (ENTROPY_CONTEXT *)xd->above_context + vp8_block2above[block]; l = (ENTROPY_CONTEXT *)xd->left_context + vp8_block2left[block]; VP8_COMBINEENTROPYCONTEXTS(pt, *a, *l); @@ -485,22 +459,18 @@ static void tokenize1st_order_b_4x4(MACROBLOCKD *xd, token = DCT_EOB_TOKEN; t->Token = token; -#if CONFIG_HYBRIDTRANSFORM if (tx_type != DCT_DCT) t->context_tree = cpi->common.fc.hybrid_coef_probs[type][band][pt]; else -#endif t->context_tree = cpi->common.fc.coef_probs[type][band][pt]; t->skip_eob_node = pt == 0 && ((band > 0 && type != PLANE_TYPE_Y_NO_DC) || (band > 1 && type == PLANE_TYPE_Y_NO_DC)); assert(vp8_coef_encodings[t->Token].Len - t->skip_eob_node > 0); if (!dry_run) { -#if CONFIG_HYBRIDTRANSFORM if (tx_type != DCT_DCT) ++cpi->hybrid_coef_counts[type][band][pt][token]; else -#endif ++cpi->coef_counts[type][band][pt][token]; } pt = vp8_prev_token_class[token]; @@ -619,7 +589,8 @@ void vp8_tokenize_mb(VP8_COMP *cpi, xd->mode_info_context->mbmi.mb_skip_coeff = mb_is_skippable_16x16(xd); break; case TX_8X8: - if (xd->mode_info_context->mbmi.mode == I8X8_PRED) + if (xd->mode_info_context->mbmi.mode == I8X8_PRED || + xd->mode_info_context->mbmi.mode == SPLITMV) xd->mode_info_context->mbmi.mb_skip_coeff = mb_is_skippable_8x8_4x4uv(xd, 0); else xd->mode_info_context->mbmi.mb_skip_coeff = mb_is_skippable_8x8(xd, has_y2_block); @@ -668,17 +639,15 @@ void vp8_tokenize_mb(VP8_COMP *cpi, tokenize1st_order_b_16x16(xd, xd->block, t, PLANE_TYPE_Y_WITH_DC, A, L, cpi, dry_run); + A[1] = A[2] = A[3] = A[0]; + L[1] = L[2] = L[3] = L[0]; - for (b = 1; b < 16; b++) { - *(A + vp8_block2above[b]) = *(A); - *(L + vp8_block2left[b] ) = *(L); - } for (b = 16; b < 24; b += 4) { tokenize1st_order_b_8x8(xd, xd->block + b, t, PLANE_TYPE_UV, A + vp8_block2above_8x8[b], L + vp8_block2left_8x8[b], cpi, dry_run); - *(A + vp8_block2above_8x8[b]+1) = *(A + vp8_block2above_8x8[b]); - *(L + vp8_block2left_8x8[b]+1 ) = *(L + vp8_block2left_8x8[b]); + A[vp8_block2above_8x8[b] + 1] = A[vp8_block2above_8x8[b]]; + L[vp8_block2left_8x8[b] + 1] = L[vp8_block2left_8x8[b]]; } vpx_memset(&A[8], 0, sizeof(A[8])); vpx_memset(&L[8], 0, sizeof(L[8])); @@ -692,18 +661,19 @@ void vp8_tokenize_mb(VP8_COMP *cpi, A + vp8_block2above_8x8[b], L + vp8_block2left_8x8[b], cpi, dry_run); - *(A + vp8_block2above_8x8[b] + 1) = *(A + vp8_block2above_8x8[b]); - *(L + vp8_block2left_8x8[b] + 1) = *(L + vp8_block2left_8x8[b]); + A[vp8_block2above_8x8[b] + 1] = A[vp8_block2above_8x8[b]]; + L[vp8_block2left_8x8[b] + 1] = L[vp8_block2left_8x8[b]]; } - if (xd->mode_info_context->mbmi.mode == I8X8_PRED) { + if (xd->mode_info_context->mbmi.mode == I8X8_PRED || + xd->mode_info_context->mbmi.mode == SPLITMV) { tokenize1st_order_chroma_4x4(xd, t, cpi, dry_run); } else { for (b = 16; b < 24; b += 4) { tokenize1st_order_b_8x8(xd, xd->block + b, t, PLANE_TYPE_UV, A + vp8_block2above_8x8[b], L + vp8_block2left_8x8[b], cpi, dry_run); - *(A + vp8_block2above_8x8[b] + 1) = *(A + vp8_block2above_8x8[b]); - *(L + vp8_block2left_8x8[b] + 1) = *(L + vp8_block2left_8x8[b]); + A[vp8_block2above_8x8[b] + 1] = A[vp8_block2above_8x8[b]]; + L[vp8_block2left_8x8[b] + 1] = L[vp8_block2left_8x8[b]]; } } } else { @@ -995,30 +965,24 @@ static __inline void stuff1st_order_b_8x8(MACROBLOCKD *xd, int dry_run) { int pt; /* near block/prev token context index */ TOKENEXTRA *t = *tp; /* store tokens starting here */ -#if CONFIG_HYBRIDTRANSFORM8X8 TX_TYPE tx_type = get_tx_type(xd, b); -#endif const int band = vp8_coef_bands_8x8[(type == PLANE_TYPE_Y_NO_DC) ? 1 : 0]; VP8_COMBINEENTROPYCONTEXTS(pt, *a, *l); (void) b; t->Token = DCT_EOB_TOKEN; -#if CONFIG_HYBRIDTRANSFORM8X8 if (tx_type != DCT_DCT) t->context_tree = cpi->common.fc.hybrid_coef_probs_8x8[type][band][pt]; else -#endif t->context_tree = cpi->common.fc.coef_probs_8x8[type][band][pt]; // t->section = 8; t->skip_eob_node = 0; ++t; *tp = t; if (!dry_run) { -#if CONFIG_HYBRIDTRANSFORM8X8 if (tx_type == DCT_DCT) ++cpi->hybrid_coef_counts_8x8[type][band][pt][DCT_EOB_TOKEN]; else -#endif ++cpi->coef_counts_8x8[type][band][pt][DCT_EOB_TOKEN]; } pt = 0; /* 0 <-> all coeff data is zero */ @@ -1074,8 +1038,8 @@ static void vp8_stuff_mb_8x8(VP8_COMP *cpi, MACROBLOCKD *xd, A + vp8_block2above_8x8[b], L + vp8_block2left_8x8[b], cpi, dry_run); - *(A + vp8_block2above_8x8[b] + 1) = *(A + vp8_block2above_8x8[b]); - *(L + vp8_block2left_8x8[b] + 1) = *(L + vp8_block2left_8x8[b]); + A[vp8_block2above_8x8[b] + 1] = A[vp8_block2above_8x8[b]]; + L[vp8_block2left_8x8[b] + 1] = L[vp8_block2left_8x8[b]]; } for (b = 16; b < 24; b += 4) { @@ -1083,8 +1047,8 @@ static void vp8_stuff_mb_8x8(VP8_COMP *cpi, MACROBLOCKD *xd, A + vp8_block2above[b], L + vp8_block2left[b], cpi, dry_run); - *(A + vp8_block2above_8x8[b] + 1) = *(A + vp8_block2above_8x8[b]); - *(L + vp8_block2left_8x8[b] + 1) = *(L + vp8_block2left_8x8[b]); + A[vp8_block2above_8x8[b] + 1] = A[vp8_block2above_8x8[b]]; + L[vp8_block2left_8x8[b] + 1] = L[vp8_block2left_8x8[b]]; } if (dry_run) *t = t_backup; @@ -1100,29 +1064,23 @@ static __inline void stuff1st_order_b_16x16(MACROBLOCKD *xd, int dry_run) { int pt; /* near block/prev token context index */ TOKENEXTRA *t = *tp; /* store tokens starting here */ -#if CONFIG_HYBRIDTRANSFORM16X16 TX_TYPE tx_type = get_tx_type(xd, b); -#endif const int band = vp8_coef_bands_16x16[(type == PLANE_TYPE_Y_NO_DC) ? 1 : 0]; VP8_COMBINEENTROPYCONTEXTS(pt, *a, *l); (void) b; t->Token = DCT_EOB_TOKEN; -#if CONFIG_HYBRIDTRANSFORM16X16 if (tx_type != DCT_DCT) t->context_tree = cpi->common.fc.hybrid_coef_probs_16x16[type][band][pt]; else -#endif t->context_tree = cpi->common.fc.coef_probs_16x16[type][band][pt]; t->skip_eob_node = 0; ++t; *tp = t; if (!dry_run) { -#if CONFIG_HYBRIDTRANSFORM16X16 if (tx_type != DCT_DCT) ++cpi->hybrid_coef_counts_16x16[type][band][pt][DCT_EOB_TOKEN]; else -#endif ++cpi->coef_counts_16x16[type][band][pt][DCT_EOB_TOKEN]; } pt = 0; /* 0 <-> all coeff data is zero */ @@ -1138,17 +1096,15 @@ static void vp8_stuff_mb_16x16(VP8_COMP *cpi, MACROBLOCKD *xd, stuff1st_order_b_16x16(xd, xd->block, t, PLANE_TYPE_Y_WITH_DC, A, L, cpi, dry_run); - for (i = 1; i < 16; i++) { - *(A + vp8_block2above[i]) = *(A); - *(L + vp8_block2left[i]) = *(L); - } + A[1] = A[2] = A[3] = A[0]; + L[1] = L[2] = L[3] = L[0]; for (b = 16; b < 24; b += 4) { stuff1st_order_buv_8x8(xd, xd->block + b, t, A + vp8_block2above[b], L + vp8_block2left[b], cpi, dry_run); - *(A + vp8_block2above_8x8[b]+1) = *(A + vp8_block2above_8x8[b]); - *(L + vp8_block2left_8x8[b]+1 ) = *(L + vp8_block2left_8x8[b]); + A[vp8_block2above_8x8[b] + 1] = A[vp8_block2above_8x8[b]]; + L[vp8_block2left_8x8[b] + 1] = L[vp8_block2left_8x8[b]]; } vpx_memset(&A[8], 0, sizeof(A[8])); vpx_memset(&L[8], 0, sizeof(L[8])); @@ -1189,28 +1145,22 @@ static __inline void stuff1st_order_b_4x4(MACROBLOCKD *xd, int dry_run) { int pt; /* near block/prev token context index */ TOKENEXTRA *t = *tp; /* store tokens starting here */ -#if CONFIG_HYBRIDTRANSFORM TX_TYPE tx_type = get_tx_type(xd, b); -#endif const int band = vp8_coef_bands[(type == PLANE_TYPE_Y_NO_DC) ? 1 : 0]; VP8_COMBINEENTROPYCONTEXTS(pt, *a, *l); t->Token = DCT_EOB_TOKEN; -#if CONFIG_HYBRIDTRANSFORM if (tx_type != DCT_DCT) t->context_tree = cpi->common.fc.hybrid_coef_probs[type][band][pt]; else -#endif t->context_tree = cpi->common.fc.coef_probs[type][band][pt]; t->skip_eob_node = 0; ++t; *tp = t; if (!dry_run) { -#if CONFIG_HYBRIDTRANSFORM if (tx_type != DCT_DCT) ++cpi->hybrid_coef_counts[type][band][pt][DCT_EOB_TOKEN]; else -#endif ++cpi->coef_counts[type][band][pt][DCT_EOB_TOKEN]; } pt = 0; /* 0 <-> all coeff data is zero */ @@ -1288,8 +1238,8 @@ static void vp8_stuff_mb_8x8_4x4uv(VP8_COMP *cpi, MACROBLOCKD *xd, A + vp8_block2above_8x8[b], L + vp8_block2left_8x8[b], cpi, dry_run); - *(A + vp8_block2above_8x8[b] + 1) = *(A + vp8_block2above_8x8[b]); - *(L + vp8_block2left_8x8[b] + 1) = *(L + vp8_block2left_8x8[b]); + A[vp8_block2above_8x8[b] + 1] = A[vp8_block2above_8x8[b]]; + L[vp8_block2left_8x8[b] + 1] = L[vp8_block2left_8x8[b]]; } for (b = 16; b < 24; b++) @@ -1308,7 +1258,8 @@ void vp8_stuff_mb(VP8_COMP *cpi, MACROBLOCKD *xd, TOKENEXTRA **t, int dry_run) { if (tx_size == TX_16X16) { vp8_stuff_mb_16x16(cpi, xd, t, dry_run); } else if (tx_size == TX_8X8) { - if (xd->mode_info_context->mbmi.mode == I8X8_PRED) { + if (xd->mode_info_context->mbmi.mode == I8X8_PRED || + xd->mode_info_context->mbmi.mode == SPLITMV) { vp8_stuff_mb_8x8_4x4uv(cpi, xd, t, dry_run); } else { vp8_stuff_mb_8x8(cpi, xd, t, dry_run); diff --git a/vp8/encoder/variance.h b/vp8/encoder/variance.h index a2fadfc4c..cdeb390c3 100644 --- a/vp8/encoder/variance.h +++ b/vp8/encoder/variance.h @@ -12,507 +12,73 @@ #ifndef VARIANCE_H #define VARIANCE_H -#include "vpx_config.h" - -#define prototype_sad(sym)\ - unsigned int (sym)\ - (\ - const unsigned char *src_ptr, \ - int source_stride, \ - const unsigned char *ref_ptr, \ - int ref_stride, \ - int max_sad\ - ) - -#define prototype_sad_multi_same_address(sym)\ - void (sym)\ - (\ - const unsigned char *src_ptr, \ - int source_stride, \ - const unsigned char *ref_ptr, \ - int ref_stride, \ - unsigned int *sad_array\ - ) - -#define prototype_sad_multi_same_address_1(sym)\ - void (sym)\ - (\ - const unsigned char *src_ptr, \ - int source_stride, \ - const unsigned char *ref_ptr, \ - int ref_stride, \ - unsigned short *sad_array\ - ) - -#define prototype_sad_multi_dif_address(sym)\ - void (sym)\ - (\ - const unsigned char *src_ptr, \ - int source_stride, \ - unsigned char *ref_ptr[4], \ - int ref_stride, \ - unsigned int *sad_array\ - ) - -#define prototype_variance(sym) \ - unsigned int (sym) \ - (\ - const unsigned char *src_ptr, \ - int source_stride, \ - const unsigned char *ref_ptr, \ - int ref_stride, \ - unsigned int *sse\ - ) - -#define prototype_variance2(sym) \ - unsigned int (sym) \ - (\ - const unsigned char *src_ptr, \ - int source_stride, \ - const unsigned char *ref_ptr, \ - int ref_stride, \ - unsigned int *sse,\ - int *sum\ - ) - -#define prototype_subpixvariance(sym) \ - unsigned int (sym) \ - ( \ - const unsigned char *src_ptr, \ - int source_stride, \ - int xoffset, \ - int yoffset, \ - const unsigned char *ref_ptr, \ - int Refstride, \ - unsigned int *sse \ - ); - -#define prototype_ssimpf(sym) \ - void (sym) \ - ( \ - unsigned char *s, \ - int sp, \ - unsigned char *r, \ - int rp, \ - unsigned long *sum_s, \ - unsigned long *sum_r, \ - unsigned long *sum_sq_s, \ - unsigned long *sum_sq_r, \ - unsigned long *sum_sxr \ - ); - -#define prototype_getmbss(sym) unsigned int (sym)(const short *) - -#define prototype_get16x16prederror(sym)\ - unsigned int (sym)\ - (\ - const unsigned char *src_ptr, \ - int source_stride, \ - const unsigned char *ref_ptr, \ - int ref_stride \ - ) - -#if ARCH_X86 || ARCH_X86_64 -#include "x86/variance_x86.h" -#endif - -#if ARCH_ARM -#include "arm/variance_arm.h" -#endif - -#ifndef vp8_variance_sad4x4 -#define vp8_variance_sad4x4 vp8_sad4x4_c -#endif -extern prototype_sad(vp8_variance_sad4x4); - -#ifndef vp8_variance_sad8x8 -#define vp8_variance_sad8x8 vp8_sad8x8_c -#endif -extern prototype_sad(vp8_variance_sad8x8); - -#ifndef vp8_variance_sad8x16 -#define vp8_variance_sad8x16 vp8_sad8x16_c -#endif -extern prototype_sad(vp8_variance_sad8x16); - -#ifndef vp8_variance_sad16x8 -#define vp8_variance_sad16x8 vp8_sad16x8_c -#endif -extern prototype_sad(vp8_variance_sad16x8); - -#ifndef vp8_variance_sad16x16 -#define vp8_variance_sad16x16 vp8_sad16x16_c -#endif -extern prototype_sad(vp8_variance_sad16x16); - -#ifndef vp8_variance_sad32x32 -#define vp8_variance_sad32x32 vp8_sad32x32_c -#endif -extern prototype_sad(vp8_variance_sad32x32); - -// -=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=- - -#ifndef vp8_variance_sad32x32x3 -#define vp8_variance_sad32x32x3 vp8_sad32x32x3_c -#endif -extern prototype_sad_multi_same_address(vp8_variance_sad32x32x3); - -#ifndef vp8_variance_sad16x16x3 -#define vp8_variance_sad16x16x3 vp8_sad16x16x3_c -#endif -extern prototype_sad_multi_same_address(vp8_variance_sad16x16x3); - -#ifndef vp8_variance_sad16x8x3 -#define vp8_variance_sad16x8x3 vp8_sad16x8x3_c -#endif -extern prototype_sad_multi_same_address(vp8_variance_sad16x8x3); - -#ifndef vp8_variance_sad8x8x3 -#define vp8_variance_sad8x8x3 vp8_sad8x8x3_c -#endif -extern prototype_sad_multi_same_address(vp8_variance_sad8x8x3); - -#ifndef vp8_variance_sad8x16x3 -#define vp8_variance_sad8x16x3 vp8_sad8x16x3_c -#endif -extern prototype_sad_multi_same_address(vp8_variance_sad8x16x3); - -#ifndef vp8_variance_sad4x4x3 -#define vp8_variance_sad4x4x3 vp8_sad4x4x3_c -#endif -extern prototype_sad_multi_same_address(vp8_variance_sad4x4x3); - -#ifndef vp8_variance_sad32x32x8 -#define vp8_variance_sad32x32x8 vp8_sad32x32x8_c -#endif -extern prototype_sad_multi_same_address_1(vp8_variance_sad32x32x8); - -#ifndef vp8_variance_sad16x16x8 -#define vp8_variance_sad16x16x8 vp8_sad16x16x8_c -#endif -extern prototype_sad_multi_same_address_1(vp8_variance_sad16x16x8); - -#ifndef vp8_variance_sad16x8x8 -#define vp8_variance_sad16x8x8 vp8_sad16x8x8_c -#endif -extern prototype_sad_multi_same_address_1(vp8_variance_sad16x8x8); - -#ifndef vp8_variance_sad8x8x8 -#define vp8_variance_sad8x8x8 vp8_sad8x8x8_c -#endif -extern prototype_sad_multi_same_address_1(vp8_variance_sad8x8x8); - -#ifndef vp8_variance_sad8x16x8 -#define vp8_variance_sad8x16x8 vp8_sad8x16x8_c -#endif -extern prototype_sad_multi_same_address_1(vp8_variance_sad8x16x8); - -#ifndef vp8_variance_sad4x4x8 -#define vp8_variance_sad4x4x8 vp8_sad4x4x8_c -#endif -extern prototype_sad_multi_same_address_1(vp8_variance_sad4x4x8); - -// -=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=- - -#ifndef vp8_variance_sad32x32x4d -#define vp8_variance_sad32x32x4d vp8_sad32x32x4d_c -#endif -extern prototype_sad_multi_dif_address(vp8_variance_sad32x32x4d); - -#ifndef vp8_variance_sad16x16x4d -#define vp8_variance_sad16x16x4d vp8_sad16x16x4d_c -#endif -extern prototype_sad_multi_dif_address(vp8_variance_sad16x16x4d); - -#ifndef vp8_variance_sad16x8x4d -#define vp8_variance_sad16x8x4d vp8_sad16x8x4d_c -#endif -extern prototype_sad_multi_dif_address(vp8_variance_sad16x8x4d); - -#ifndef vp8_variance_sad8x8x4d -#define vp8_variance_sad8x8x4d vp8_sad8x8x4d_c -#endif -extern prototype_sad_multi_dif_address(vp8_variance_sad8x8x4d); - -#ifndef vp8_variance_sad8x16x4d -#define vp8_variance_sad8x16x4d vp8_sad8x16x4d_c -#endif -extern prototype_sad_multi_dif_address(vp8_variance_sad8x16x4d); - -#ifndef vp8_variance_sad4x4x4d -#define vp8_variance_sad4x4x4d vp8_sad4x4x4d_c -#endif -extern prototype_sad_multi_dif_address(vp8_variance_sad4x4x4d); - -#if ARCH_X86 || ARCH_X86_64 -#ifndef vp8_variance_copy32xn -#define vp8_variance_copy32xn vp8_copy32xn_c -#endif -extern prototype_sad(vp8_variance_copy32xn); -#endif - -// -=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=- - -#ifndef vp8_variance_var4x4 -#define vp8_variance_var4x4 vp8_variance4x4_c -#endif -extern prototype_variance(vp8_variance_var4x4); - -#ifndef vp8_variance_var8x8 -#define vp8_variance_var8x8 vp8_variance8x8_c -#endif -extern prototype_variance(vp8_variance_var8x8); - -#ifndef vp8_variance_var8x16 -#define vp8_variance_var8x16 vp8_variance8x16_c -#endif -extern prototype_variance(vp8_variance_var8x16); - -#ifndef vp8_variance_var16x8 -#define vp8_variance_var16x8 vp8_variance16x8_c -#endif -extern prototype_variance(vp8_variance_var16x8); - -#ifndef vp8_variance_var16x16 -#define vp8_variance_var16x16 vp8_variance16x16_c -#endif -extern prototype_variance(vp8_variance_var16x16); - -#ifndef vp8_variance_var32x32 -#define vp8_variance_var32x32 vp8_variance32x32_c -#endif -extern prototype_variance(vp8_variance_var32x32); - -// -=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=- - -#ifndef vp8_variance_subpixvar4x4 -#define vp8_variance_subpixvar4x4 vp8_sub_pixel_variance4x4_c -#endif -extern prototype_subpixvariance(vp8_variance_subpixvar4x4); - -#ifndef vp8_variance_subpixvar8x8 -#define vp8_variance_subpixvar8x8 vp8_sub_pixel_variance8x8_c -#endif -extern prototype_subpixvariance(vp8_variance_subpixvar8x8); - -#ifndef vp8_variance_subpixvar8x16 -#define vp8_variance_subpixvar8x16 vp8_sub_pixel_variance8x16_c -#endif -extern prototype_subpixvariance(vp8_variance_subpixvar8x16); - -#ifndef vp8_variance_subpixvar16x8 -#define vp8_variance_subpixvar16x8 vp8_sub_pixel_variance16x8_c -#endif -extern prototype_subpixvariance(vp8_variance_subpixvar16x8); - -#ifndef vp8_variance_subpixvar16x16 -#define vp8_variance_subpixvar16x16 vp8_sub_pixel_variance16x16_c -#endif -extern prototype_subpixvariance(vp8_variance_subpixvar16x16); - -#ifndef vp8_variance_subpixvar32x32 -#define vp8_variance_subpixvar32x32 vp8_sub_pixel_variance32x32_c -#endif -extern prototype_subpixvariance(vp8_variance_subpixvar32x32); - -#ifndef vp8_variance_halfpixvar16x16_h -#define vp8_variance_halfpixvar16x16_h vp8_variance_halfpixvar16x16_h_c -#endif -extern prototype_variance(vp8_variance_halfpixvar16x16_h); - -#ifndef vp8_variance_halfpixvar32x32_h -#define vp8_variance_halfpixvar32x32_h vp8_variance_halfpixvar32x32_h_c -#endif -extern prototype_variance(vp8_variance_halfpixvar32x32_h); - -#ifndef vp8_variance_halfpixvar16x16_v -#define vp8_variance_halfpixvar16x16_v vp8_variance_halfpixvar16x16_v_c -#endif -extern prototype_variance(vp8_variance_halfpixvar16x16_v); - -#ifndef vp8_variance_halfpixvar32x32_v -#define vp8_variance_halfpixvar32x32_v vp8_variance_halfpixvar32x32_v_c -#endif -extern prototype_variance(vp8_variance_halfpixvar32x32_v); - -#ifndef vp8_variance_halfpixvar16x16_hv -#define vp8_variance_halfpixvar16x16_hv vp8_variance_halfpixvar16x16_hv_c -#endif -extern prototype_variance(vp8_variance_halfpixvar16x16_hv); - -#ifndef vp8_variance_halfpixvar32x32_hv -#define vp8_variance_halfpixvar32x32_hv vp8_variance_halfpixvar32x32_hv_c -#endif -extern prototype_variance(vp8_variance_halfpixvar32x32_hv); - -#ifndef vp8_variance_subpixmse16x16 -#define vp8_variance_subpixmse16x16 vp8_sub_pixel_mse16x16_c -#endif -extern prototype_subpixvariance(vp8_variance_subpixmse16x16); - -#ifndef vp8_variance_subpixmse32x32 -#define vp8_variance_subpixmse32x32 vp8_sub_pixel_mse32x32_c -#endif -extern prototype_subpixvariance(vp8_variance_subpixmse32x32); - -// -=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=- - -#ifndef vp8_variance_getmbss -#define vp8_variance_getmbss vp8_get_mb_ss_c -#endif -extern prototype_getmbss(vp8_variance_getmbss); - -#ifndef vp8_variance_mse16x16 -#define vp8_variance_mse16x16 vp8_mse16x16_c -#endif -extern prototype_variance(vp8_variance_mse16x16); - -#ifndef vp8_ssimpf_8x8 -#define vp8_ssimpf_8x8 vp8_ssim_parms_8x8_c -#endif -extern prototype_ssimpf(vp8_ssimpf_8x8) - -#ifndef vp8_ssimpf_16x16 -#define vp8_ssimpf_16x16 vp8_ssim_parms_16x16_c -#endif -extern prototype_ssimpf(vp8_ssimpf_16x16) - -#ifndef vp8_variance_satd16x16 -#define vp8_variance_satd16x16 vp8_satd16x16_c -#endif -extern prototype_variance(vp8_variance_satd16x16); - -typedef prototype_sad(*vp8_sad_fn_t); -typedef prototype_sad_multi_same_address(*vp8_sad_multi_fn_t); -typedef prototype_sad_multi_same_address_1(*vp8_sad_multi1_fn_t); -typedef prototype_sad_multi_dif_address(*vp8_sad_multi_d_fn_t); -typedef prototype_variance(*vp8_variance_fn_t); -typedef prototype_variance2(*vp8_variance2_fn_t); -typedef prototype_subpixvariance(*vp8_subpixvariance_fn_t); -typedef prototype_getmbss(*vp8_getmbss_fn_t); -typedef prototype_ssimpf(*vp8_ssimpf_fn_t); -typedef prototype_get16x16prederror(*vp8_get16x16prederror_fn_t); - -typedef struct { - vp8_sad_fn_t sad4x4; - vp8_sad_fn_t sad8x8; - vp8_sad_fn_t sad8x16; - vp8_sad_fn_t sad16x8; - vp8_sad_fn_t sad16x16; -#if CONFIG_SUPERBLOCKS - vp8_sad_fn_t sad32x32; -#endif - - vp8_variance_fn_t var4x4; - vp8_variance_fn_t var8x8; - vp8_variance_fn_t var8x16; - vp8_variance_fn_t var16x8; - vp8_variance_fn_t var16x16; -#if CONFIG_SUPERBLOCKS - vp8_variance_fn_t var32x32; -#endif - - vp8_subpixvariance_fn_t subpixvar4x4; - vp8_subpixvariance_fn_t subpixvar8x8; - vp8_subpixvariance_fn_t subpixvar8x16; - vp8_subpixvariance_fn_t subpixvar16x8; - vp8_subpixvariance_fn_t subpixvar16x16; -#if CONFIG_SUPERBLOCKS - vp8_subpixvariance_fn_t subpixvar32x32; -#endif - vp8_variance_fn_t halfpixvar16x16_h; - vp8_variance_fn_t halfpixvar32x32_h; - vp8_variance_fn_t halfpixvar16x16_v; -#if CONFIG_SUPERBLOCKS - vp8_variance_fn_t halfpixvar32x32_v; -#endif - vp8_variance_fn_t halfpixvar16x16_hv; -#if CONFIG_SUPERBLOCKS - vp8_variance_fn_t halfpixvar32x32_hv; -#endif - vp8_subpixvariance_fn_t subpixmse16x16; -#if CONFIG_SUPERBLOCKS - vp8_subpixvariance_fn_t subpixmse32x32; -#endif - - vp8_getmbss_fn_t getmbss; - vp8_variance_fn_t mse16x16; - -#if CONFIG_SUPERBLOCKS - vp8_sad_multi_fn_t sad32x32x3; -#endif - vp8_sad_multi_fn_t sad16x16x3; - vp8_sad_multi_fn_t sad16x8x3; - vp8_sad_multi_fn_t sad8x16x3; - vp8_sad_multi_fn_t sad8x8x3; - vp8_sad_multi_fn_t sad4x4x3; - -#if CONFIG_SUPERBLOCKS - vp8_sad_multi1_fn_t sad32x32x8; -#endif - vp8_sad_multi1_fn_t sad16x16x8; - vp8_sad_multi1_fn_t sad16x8x8; - vp8_sad_multi1_fn_t sad8x16x8; - vp8_sad_multi1_fn_t sad8x8x8; - vp8_sad_multi1_fn_t sad4x4x8; - -#if CONFIG_SUPERBLOCKS - vp8_sad_multi_d_fn_t sad32x32x4d; -#endif - vp8_sad_multi_d_fn_t sad16x16x4d; - vp8_sad_multi_d_fn_t sad16x8x4d; - vp8_sad_multi_d_fn_t sad8x16x4d; - vp8_sad_multi_d_fn_t sad8x8x4d; - vp8_sad_multi_d_fn_t sad4x4x4d; - -#if ARCH_X86 || ARCH_X86_64 - vp8_sad_fn_t copy32xn; -#endif - -#if CONFIG_INTERNAL_STATS - vp8_ssimpf_fn_t ssimpf_8x8; - vp8_ssimpf_fn_t ssimpf_16x16; -#endif - - vp8_variance_fn_t satd16x16; -} vp8_variance_rtcd_vtable_t; - -typedef struct { - vp8_sad_fn_t sdf; - vp8_variance_fn_t vf; - vp8_subpixvariance_fn_t svf; - vp8_variance_fn_t svf_halfpix_h; - vp8_variance_fn_t svf_halfpix_v; - vp8_variance_fn_t svf_halfpix_hv; - vp8_sad_multi_fn_t sdx3f; - vp8_sad_multi1_fn_t sdx8f; - vp8_sad_multi_d_fn_t sdx4df; -#if ARCH_X86 || ARCH_X86_64 - vp8_sad_fn_t copymem; -#endif +typedef unsigned int(*vp8_sad_fn_t)(const unsigned char *src_ptr, + int source_stride, + const unsigned char *ref_ptr, + int ref_stride, + unsigned int max_sad); + +typedef void (*vp8_copy32xn_fn_t)(const unsigned char *src_ptr, + int source_stride, + const unsigned char *ref_ptr, + int ref_stride, + int n); + +typedef void (*vp8_sad_multi_fn_t)(const unsigned char *src_ptr, + int source_stride, + const unsigned char *ref_ptr, + int ref_stride, + unsigned int *sad_array); + +typedef void (*vp8_sad_multi1_fn_t)(const unsigned char *src_ptr, + int source_stride, + const unsigned char *ref_ptr, + int ref_stride, + unsigned short *sad_array); + +typedef void (*vp8_sad_multi_d_fn_t)(const unsigned char *src_ptr, + int source_stride, + const unsigned char * const ref_ptr[], + int ref_stride, unsigned int *sad_array); + +typedef unsigned int (*vp8_variance_fn_t)(const unsigned char *src_ptr, + int source_stride, + const unsigned char *ref_ptr, + int ref_stride, + unsigned int *sse); + +typedef unsigned int (*vp8_subpixvariance_fn_t)(const unsigned char *src_ptr, + int source_stride, + int xoffset, + int yoffset, + const unsigned char *ref_ptr, + int Refstride, + unsigned int *sse); + +typedef void (*vp8_ssimpf_fn_t)(unsigned char *s, int sp, unsigned char *r, + int rp, unsigned long *sum_s, + unsigned long *sum_r, unsigned long *sum_sq_s, + unsigned long *sum_sq_r, + unsigned long *sum_sxr); + +typedef unsigned int (*vp8_getmbss_fn_t)(const short *); + +typedef unsigned int (*vp8_get16x16prederror_fn_t)(const unsigned char *src_ptr, + int source_stride, + const unsigned char *ref_ptr, + int ref_stride); + +typedef struct variance_vtable { + vp8_sad_fn_t sdf; + vp8_variance_fn_t vf; + vp8_subpixvariance_fn_t svf; + vp8_variance_fn_t svf_halfpix_h; + vp8_variance_fn_t svf_halfpix_v; + vp8_variance_fn_t svf_halfpix_hv; + vp8_sad_multi_fn_t sdx3f; + vp8_sad_multi1_fn_t sdx8f; + vp8_sad_multi_d_fn_t sdx4df; + vp8_copy32xn_fn_t copymem; } vp8_variance_fn_ptr_t; -#if CONFIG_RUNTIME_CPU_DETECT -#define VARIANCE_INVOKE(ctx,fn) (ctx)->fn -#define SSIMPF_INVOKE(ctx,fn) (ctx)->ssimpf_##fn -#else -#define VARIANCE_INVOKE(ctx,fn) vp8_variance_##fn -#define SSIMPF_INVOKE(ctx,fn) vp8_ssimpf_##fn -#endif - -#if CONFIG_NEWBESTREFMV -unsigned int vp8_sad2x16_c( - const unsigned char *src_ptr, - int src_stride, - const unsigned char *ref_ptr, - int ref_stride, - int max_sad); -unsigned int vp8_sad16x2_c( - const unsigned char *src_ptr, - int src_stride, - const unsigned char *ref_ptr, - int ref_stride, - int max_sad); -#endif - #endif diff --git a/vp8/encoder/x86/variance_x86.h b/vp8/encoder/x86/variance_x86.h deleted file mode 100644 index 0971f11b0..000000000 --- a/vp8/encoder/x86/variance_x86.h +++ /dev/null @@ -1,328 +0,0 @@ -/* - * 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. - */ - - -#ifndef VARIANCE_X86_H -#define VARIANCE_X86_H - - -/* Note: - * - * This platform is commonly built for runtime CPU detection. If you modify - * any of the function mappings present in this file, be sure to also update - * them in the function pointer initialization code - */ -#if HAVE_MMX -extern prototype_sad(vp8_sad4x4_mmx); -extern prototype_sad(vp8_sad8x8_mmx); -extern prototype_sad(vp8_sad8x16_mmx); -extern prototype_sad(vp8_sad16x8_mmx); -extern prototype_sad(vp8_sad16x16_mmx); -extern prototype_variance(vp8_variance4x4_mmx); -extern prototype_variance(vp8_variance8x8_mmx); -extern prototype_variance(vp8_variance8x16_mmx); -extern prototype_variance(vp8_variance16x8_mmx); -extern prototype_variance(vp8_variance16x16_mmx); -extern prototype_subpixvariance(vp8_sub_pixel_variance4x4_mmx); -extern prototype_subpixvariance(vp8_sub_pixel_variance8x8_mmx); -extern prototype_subpixvariance(vp8_sub_pixel_variance8x16_mmx); -extern prototype_subpixvariance(vp8_sub_pixel_variance16x8_mmx); -extern prototype_subpixvariance(vp8_sub_pixel_variance16x16_mmx); -extern prototype_variance(vp8_variance_halfpixvar16x16_h_mmx); -extern prototype_variance(vp8_variance_halfpixvar16x16_v_mmx); -extern prototype_variance(vp8_variance_halfpixvar16x16_hv_mmx); -extern prototype_subpixvariance(vp8_sub_pixel_mse16x16_mmx); -extern prototype_getmbss(vp8_get_mb_ss_mmx); -extern prototype_variance(vp8_mse16x16_mmx); -extern prototype_variance2(vp8_get8x8var_mmx); - -#if !CONFIG_RUNTIME_CPU_DETECT -#undef vp8_variance_sad4x4 -#define vp8_variance_sad4x4 vp8_sad4x4_mmx - -#undef vp8_variance_sad8x8 -#define vp8_variance_sad8x8 vp8_sad8x8_mmx - -#undef vp8_variance_sad8x16 -#define vp8_variance_sad8x16 vp8_sad8x16_mmx - -#undef vp8_variance_sad16x8 -#define vp8_variance_sad16x8 vp8_sad16x8_mmx - -#undef vp8_variance_sad16x16 -#define vp8_variance_sad16x16 vp8_sad16x16_mmx - -#undef vp8_variance_var4x4 -#define vp8_variance_var4x4 vp8_variance4x4_mmx - -#undef vp8_variance_var8x8 -#define vp8_variance_var8x8 vp8_variance8x8_mmx - -#undef vp8_variance_var8x16 -#define vp8_variance_var8x16 vp8_variance8x16_mmx - -#undef vp8_variance_var16x8 -#define vp8_variance_var16x8 vp8_variance16x8_mmx - -#undef vp8_variance_var16x16 -#define vp8_variance_var16x16 vp8_variance16x16_mmx - -#undef vp8_variance_subpixvar4x4 -#define vp8_variance_subpixvar4x4 vp8_sub_pixel_variance4x4_mmx - -#undef vp8_variance_subpixvar8x8 -#define vp8_variance_subpixvar8x8 vp8_sub_pixel_variance8x8_mmx - -#undef vp8_variance_subpixvar8x16 -#define vp8_variance_subpixvar8x16 vp8_sub_pixel_variance8x16_mmx - -#undef vp8_variance_subpixvar16x8 -#define vp8_variance_subpixvar16x8 vp8_sub_pixel_variance16x8_mmx - -#undef vp8_variance_subpixvar16x16 -#define vp8_variance_subpixvar16x16 vp8_sub_pixel_variance16x16_mmx - -#undef vp8_variance_halfpixvar16x16_h -#define vp8_variance_halfpixvar16x16_h vp8_variance_halfpixvar16x16_h_mmx - -#undef vp8_variance_halfpixvar16x16_v -#define vp8_variance_halfpixvar16x16_v vp8_variance_halfpixvar16x16_v_mmx - -#undef vp8_variance_halfpixvar16x16_hv -#define vp8_variance_halfpixvar16x16_hv vp8_variance_halfpixvar16x16_hv_mmx - -#undef vp8_variance_subpixmse16x16 -#define vp8_variance_subpixmse16x16 vp8_sub_pixel_mse16x16_mmx - -#undef vp8_variance_getmbss -#define vp8_variance_getmbss vp8_get_mb_ss_mmx - -#undef vp8_variance_mse16x16 -#define vp8_variance_mse16x16 vp8_mse16x16_mmx - -#endif -#endif - - -#if HAVE_SSE2 -extern prototype_sad(vp8_sad4x4_wmt); -extern prototype_sad(vp8_sad8x8_wmt); -extern prototype_sad(vp8_sad8x16_wmt); -extern prototype_sad(vp8_sad16x8_wmt); -extern prototype_sad(vp8_sad16x16_wmt); -extern prototype_sad(vp8_copy32xn_sse2); -extern prototype_variance(vp8_variance4x4_wmt); -extern prototype_variance(vp8_variance8x8_wmt); -extern prototype_variance(vp8_variance8x16_wmt); -extern prototype_variance(vp8_variance16x8_wmt); -extern prototype_variance(vp8_variance16x16_wmt); -extern prototype_subpixvariance(vp8_sub_pixel_variance4x4_wmt); -extern prototype_subpixvariance(vp8_sub_pixel_variance8x8_wmt); -extern prototype_subpixvariance(vp8_sub_pixel_variance8x16_wmt); -extern prototype_subpixvariance(vp8_sub_pixel_variance16x8_wmt); -extern prototype_subpixvariance(vp8_sub_pixel_variance16x16_wmt); -extern prototype_variance(vp8_variance_halfpixvar16x16_h_wmt); -extern prototype_variance(vp8_variance_halfpixvar16x16_v_wmt); -extern prototype_variance(vp8_variance_halfpixvar16x16_hv_wmt); -extern prototype_subpixvariance(vp8_sub_pixel_mse16x16_wmt); -extern prototype_getmbss(vp8_get_mb_ss_sse2); -extern prototype_variance(vp8_mse16x16_wmt); -extern prototype_variance2(vp8_get8x8var_sse2); -extern prototype_variance2(vp8_get16x16var_sse2); -extern prototype_ssimpf(vp8_ssim_parms_8x8_sse2) -extern prototype_ssimpf(vp8_ssim_parms_16x16_sse2) - -#if !CONFIG_RUNTIME_CPU_DETECT -#undef vp8_variance_sad4x4 -#define vp8_variance_sad4x4 vp8_sad4x4_wmt - -#undef vp8_variance_sad8x8 -#define vp8_variance_sad8x8 vp8_sad8x8_wmt - -#undef vp8_variance_sad8x16 -#define vp8_variance_sad8x16 vp8_sad8x16_wmt - -#undef vp8_variance_sad16x8 -#define vp8_variance_sad16x8 vp8_sad16x8_wmt - -#undef vp8_variance_sad16x16 -#define vp8_variance_sad16x16 vp8_sad16x16_wmt - -#undef vp8_variance_copy32xn -#define vp8_variance_copy32xn vp8_copy32xn_sse2 - -#undef vp8_variance_var4x4 -#define vp8_variance_var4x4 vp8_variance4x4_wmt - -#undef vp8_variance_var8x8 -#define vp8_variance_var8x8 vp8_variance8x8_wmt - -#undef vp8_variance_var8x16 -#define vp8_variance_var8x16 vp8_variance8x16_wmt - -#undef vp8_variance_var16x8 -#define vp8_variance_var16x8 vp8_variance16x8_wmt - -#undef vp8_variance_var16x16 -#define vp8_variance_var16x16 vp8_variance16x16_wmt - -#undef vp8_variance_subpixvar4x4 -#define vp8_variance_subpixvar4x4 vp8_sub_pixel_variance4x4_wmt - -#undef vp8_variance_subpixvar8x8 -#define vp8_variance_subpixvar8x8 vp8_sub_pixel_variance8x8_wmt - -#undef vp8_variance_subpixvar8x16 -#define vp8_variance_subpixvar8x16 vp8_sub_pixel_variance8x16_wmt - -#undef vp8_variance_subpixvar16x8 -#define vp8_variance_subpixvar16x8 vp8_sub_pixel_variance16x8_wmt - -#undef vp8_variance_subpixvar16x16 -#define vp8_variance_subpixvar16x16 vp8_sub_pixel_variance16x16_wmt - -#undef vp8_variance_halfpixvar16x16_h -#define vp8_variance_halfpixvar16x16_h vp8_variance_halfpixvar16x16_h_wmt - -#undef vp8_variance_halfpixvar16x16_v -#define vp8_variance_halfpixvar16x16_v vp8_variance_halfpixvar16x16_v_wmt - -#undef vp8_variance_halfpixvar16x16_hv -#define vp8_variance_halfpixvar16x16_hv vp8_variance_halfpixvar16x16_hv_wmt - -#undef vp8_variance_subpixmse16x16 -#define vp8_variance_subpixmse16x16 vp8_sub_pixel_mse16x16_wmt - -#undef vp8_variance_getmbss -#define vp8_variance_getmbss vp8_get_mb_ss_sse2 - -#undef vp8_variance_mse16x16 -#define vp8_variance_mse16x16 vp8_mse16x16_wmt - -#if ARCH_X86_64 -#undef vp8_ssimpf_8x8 -#define vp8_ssimpf_8x8 vp8_ssim_parms_8x8_sse2 - -#undef vp8_ssimpf_16x16 -#define vp8_ssimpf_16x16 vp8_ssim_parms_16x16_sse2 -#endif - -#endif -#endif - - -#if HAVE_SSE3 -extern prototype_sad(vp8_sad16x16_sse3); -extern prototype_sad(vp8_sad16x8_sse3); -extern prototype_sad_multi_same_address(vp8_sad16x16x3_sse3); -extern prototype_sad_multi_same_address(vp8_sad16x8x3_sse3); -extern prototype_sad_multi_same_address(vp8_sad8x16x3_sse3); -extern prototype_sad_multi_same_address(vp8_sad8x8x3_sse3); -extern prototype_sad_multi_same_address(vp8_sad4x4x3_sse3); - -extern prototype_sad_multi_dif_address(vp8_sad16x16x4d_sse3); -extern prototype_sad_multi_dif_address(vp8_sad16x8x4d_sse3); -extern prototype_sad_multi_dif_address(vp8_sad8x16x4d_sse3); -extern prototype_sad_multi_dif_address(vp8_sad8x8x4d_sse3); -extern prototype_sad_multi_dif_address(vp8_sad4x4x4d_sse3); -extern prototype_sad(vp8_copy32xn_sse3); - -#if !CONFIG_RUNTIME_CPU_DETECT - -#undef vp8_variance_sad16x16 -#define vp8_variance_sad16x16 vp8_sad16x16_sse3 - -#undef vp8_variance_sad16x16x3 -#define vp8_variance_sad16x16x3 vp8_sad16x16x3_sse3 - -#undef vp8_variance_sad16x8x3 -#define vp8_variance_sad16x8x3 vp8_sad16x8x3_sse3 - -#undef vp8_variance_sad8x16x3 -#define vp8_variance_sad8x16x3 vp8_sad8x16x3_sse3 - -#undef vp8_variance_sad8x8x3 -#define vp8_variance_sad8x8x3 vp8_sad8x8x3_sse3 - -#undef vp8_variance_sad4x4x3 -#define vp8_variance_sad4x4x3 vp8_sad4x4x3_sse3 - -#undef vp8_variance_sad16x16x4d -#define vp8_variance_sad16x16x4d vp8_sad16x16x4d_sse3 - -#undef vp8_variance_sad16x8x4d -#define vp8_variance_sad16x8x4d vp8_sad16x8x4d_sse3 - -#undef vp8_variance_sad8x16x4d -#define vp8_variance_sad8x16x4d vp8_sad8x16x4d_sse3 - -#undef vp8_variance_sad8x8x4d -#define vp8_variance_sad8x8x4d vp8_sad8x8x4d_sse3 - -#undef vp8_variance_sad4x4x4d -#define vp8_variance_sad4x4x4d vp8_sad4x4x4d_sse3 - -#undef vp8_variance_copy32xn -#define vp8_variance_copy32xn vp8_copy32xn_sse3 - -#endif -#endif - - -#if HAVE_SSSE3 -extern prototype_sad_multi_same_address(vp8_sad16x16x3_ssse3); -extern prototype_sad_multi_same_address(vp8_sad16x8x3_ssse3); -extern prototype_subpixvariance(vp8_sub_pixel_variance16x8_ssse3); -extern prototype_subpixvariance(vp8_sub_pixel_variance16x16_ssse3); - -#if !CONFIG_RUNTIME_CPU_DETECT -#undef vp8_variance_sad16x16x3 -#define vp8_variance_sad16x16x3 vp8_sad16x16x3_ssse3 - -#undef vp8_variance_sad16x8x3 -#define vp8_variance_sad16x8x3 vp8_sad16x8x3_ssse3 - -#undef vp8_variance_subpixvar16x8 -#define vp8_variance_subpixvar16x8 vp8_sub_pixel_variance16x8_ssse3 - -#undef vp8_variance_subpixvar16x16 -#define vp8_variance_subpixvar16x16 vp8_sub_pixel_variance16x16_ssse3 - -#endif -#endif - - -#if HAVE_SSE4_1 -extern prototype_sad_multi_same_address_1(vp8_sad16x16x8_sse4); -extern prototype_sad_multi_same_address_1(vp8_sad16x8x8_sse4); -extern prototype_sad_multi_same_address_1(vp8_sad8x16x8_sse4); -extern prototype_sad_multi_same_address_1(vp8_sad8x8x8_sse4); -extern prototype_sad_multi_same_address_1(vp8_sad4x4x8_sse4); - -#if !CONFIG_RUNTIME_CPU_DETECT -#undef vp8_variance_sad16x16x8 -#define vp8_variance_sad16x16x8 vp8_sad16x16x8_sse4 - -#undef vp8_variance_sad16x8x8 -#define vp8_variance_sad16x8x8 vp8_sad16x8x8_sse4 - -#undef vp8_variance_sad8x16x8 -#define vp8_variance_sad8x16x8 vp8_sad8x16x8_sse4 - -#undef vp8_variance_sad8x8x8 -#define vp8_variance_sad8x8x8 vp8_sad8x8x8_sse4 - -#undef vp8_variance_sad4x4x8 -#define vp8_variance_sad4x4x8 vp8_sad4x4x8_sse4 - -#endif -#endif - -#endif diff --git a/vp8/encoder/x86/x86_csystemdependent.c b/vp8/encoder/x86/x86_csystemdependent.c index 71c51c14f..a169b493e 100644 --- a/vp8/encoder/x86/x86_csystemdependent.c +++ b/vp8/encoder/x86/x86_csystemdependent.c @@ -90,31 +90,6 @@ void vp8_arch_x86_encoder_init(VP8_COMP *cpi) { /* Override default functions with fastest ones for this CPU. */ #if HAVE_MMX if (flags & HAS_MMX) { - cpi->rtcd.variance.sad16x16 = vp8_sad16x16_mmx; - cpi->rtcd.variance.sad16x8 = vp8_sad16x8_mmx; - cpi->rtcd.variance.sad8x16 = vp8_sad8x16_mmx; - cpi->rtcd.variance.sad8x8 = vp8_sad8x8_mmx; - cpi->rtcd.variance.sad4x4 = vp8_sad4x4_mmx; - - cpi->rtcd.variance.var4x4 = vp8_variance4x4_mmx; - cpi->rtcd.variance.var8x8 = vp8_variance8x8_mmx; - cpi->rtcd.variance.var8x16 = vp8_variance8x16_mmx; - cpi->rtcd.variance.var16x8 = vp8_variance16x8_mmx; - cpi->rtcd.variance.var16x16 = vp8_variance16x16_mmx; - - cpi->rtcd.variance.subpixvar4x4 = vp8_sub_pixel_variance4x4_mmx; - cpi->rtcd.variance.subpixvar8x8 = vp8_sub_pixel_variance8x8_mmx; - cpi->rtcd.variance.subpixvar8x16 = vp8_sub_pixel_variance8x16_mmx; - cpi->rtcd.variance.subpixvar16x8 = vp8_sub_pixel_variance16x8_mmx; - cpi->rtcd.variance.subpixvar16x16 = vp8_sub_pixel_variance16x16_mmx; - cpi->rtcd.variance.halfpixvar16x16_h = vp8_variance_halfpixvar16x16_h_mmx; - cpi->rtcd.variance.halfpixvar16x16_v = vp8_variance_halfpixvar16x16_v_mmx; - cpi->rtcd.variance.halfpixvar16x16_hv = vp8_variance_halfpixvar16x16_hv_mmx; - cpi->rtcd.variance.subpixmse16x16 = vp8_sub_pixel_mse16x16_mmx; - - cpi->rtcd.variance.mse16x16 = vp8_mse16x16_mmx; - cpi->rtcd.variance.getmbss = vp8_get_mb_ss_mmx; - cpi->rtcd.encodemb.berr = vp8_block_error_mmx; cpi->rtcd.encodemb.mberr = vp8_mbblock_error_mmx; cpi->rtcd.encodemb.mbuverr = vp8_mbuverror_mmx; @@ -126,32 +101,6 @@ void vp8_arch_x86_encoder_init(VP8_COMP *cpi) { #if HAVE_SSE2 if (flags & HAS_SSE2) { - cpi->rtcd.variance.sad16x16 = vp8_sad16x16_wmt; - cpi->rtcd.variance.sad16x8 = vp8_sad16x8_wmt; - cpi->rtcd.variance.sad8x16 = vp8_sad8x16_wmt; - cpi->rtcd.variance.sad8x8 = vp8_sad8x8_wmt; - cpi->rtcd.variance.sad4x4 = vp8_sad4x4_wmt; - cpi->rtcd.variance.copy32xn = vp8_copy32xn_sse2; - - cpi->rtcd.variance.var4x4 = vp8_variance4x4_wmt; - cpi->rtcd.variance.var8x8 = vp8_variance8x8_wmt; - cpi->rtcd.variance.var8x16 = vp8_variance8x16_wmt; - cpi->rtcd.variance.var16x8 = vp8_variance16x8_wmt; - cpi->rtcd.variance.var16x16 = vp8_variance16x16_wmt; - - cpi->rtcd.variance.subpixvar4x4 = vp8_sub_pixel_variance4x4_wmt; - cpi->rtcd.variance.subpixvar8x8 = vp8_sub_pixel_variance8x8_wmt; - cpi->rtcd.variance.subpixvar8x16 = vp8_sub_pixel_variance8x16_wmt; - cpi->rtcd.variance.subpixvar16x8 = vp8_sub_pixel_variance16x8_wmt; - cpi->rtcd.variance.subpixvar16x16 = vp8_sub_pixel_variance16x16_wmt; - cpi->rtcd.variance.halfpixvar16x16_h = vp8_variance_halfpixvar16x16_h_wmt; - cpi->rtcd.variance.halfpixvar16x16_v = vp8_variance_halfpixvar16x16_v_wmt; - cpi->rtcd.variance.halfpixvar16x16_hv = vp8_variance_halfpixvar16x16_hv_wmt; - cpi->rtcd.variance.subpixmse16x16 = vp8_sub_pixel_mse16x16_wmt; - - cpi->rtcd.variance.mse16x16 = vp8_mse16x16_wmt; - cpi->rtcd.variance.getmbss = vp8_get_mb_ss_sse2; - cpi->rtcd.encodemb.berr = vp8_block_error_xmm; cpi->rtcd.encodemb.mberr = vp8_mbblock_error_xmm; cpi->rtcd.encodemb.mbuverr = vp8_mbuverror_xmm; @@ -160,54 +109,20 @@ void vp8_arch_x86_encoder_init(VP8_COMP *cpi) { cpi->rtcd.encodemb.submbuv = vp8_subtract_mbuv_sse2; cpi->rtcd.temporal.apply = vp8_temporal_filter_apply_sse2; -#if CONFIG_INTERNAL_STATS -#if ARCH_X86_64 - cpi->rtcd.variance.ssimpf_8x8 = vp8_ssim_parms_8x8_sse2; - cpi->rtcd.variance.ssimpf_16x16 = vp8_ssim_parms_16x16_sse2; -#endif -#endif } #endif #if HAVE_SSE3 if (flags & HAS_SSE3) { - cpi->rtcd.variance.sad16x16 = vp8_sad16x16_sse3; - cpi->rtcd.variance.sad16x16x3 = vp8_sad16x16x3_sse3; - cpi->rtcd.variance.sad16x8x3 = vp8_sad16x8x3_sse3; - cpi->rtcd.variance.sad8x16x3 = vp8_sad8x16x3_sse3; - cpi->rtcd.variance.sad8x8x3 = vp8_sad8x8x3_sse3; - cpi->rtcd.variance.sad4x4x3 = vp8_sad4x4x3_sse3; cpi->rtcd.search.full_search = vp8_full_search_sadx3; - cpi->rtcd.variance.sad16x16x4d = vp8_sad16x16x4d_sse3; - cpi->rtcd.variance.sad16x8x4d = vp8_sad16x8x4d_sse3; - cpi->rtcd.variance.sad8x16x4d = vp8_sad8x16x4d_sse3; - cpi->rtcd.variance.sad8x8x4d = vp8_sad8x8x4d_sse3; - cpi->rtcd.variance.sad4x4x4d = vp8_sad4x4x4d_sse3; - cpi->rtcd.variance.copy32xn = vp8_copy32xn_sse3; cpi->rtcd.search.diamond_search = vp8_diamond_search_sadx4; cpi->rtcd.search.refining_search = vp8_refining_search_sadx4; } #endif -#if HAVE_SSSE3 - if (flags & HAS_SSSE3) { - cpi->rtcd.variance.sad16x16x3 = vp8_sad16x16x3_ssse3; - cpi->rtcd.variance.sad16x8x3 = vp8_sad16x8x3_ssse3; - - cpi->rtcd.variance.subpixvar16x8 = vp8_sub_pixel_variance16x8_ssse3; - cpi->rtcd.variance.subpixvar16x16 = vp8_sub_pixel_variance16x16_ssse3; - } -#endif - - #if HAVE_SSE4_1 if (flags & HAS_SSE4_1) { - cpi->rtcd.variance.sad16x16x8 = vp8_sad16x16x8_sse4; - cpi->rtcd.variance.sad16x8x8 = vp8_sad16x8x8_sse4; - cpi->rtcd.variance.sad8x16x8 = vp8_sad8x16x8_sse4; - cpi->rtcd.variance.sad8x8x8 = vp8_sad8x8x8_sse4; - cpi->rtcd.variance.sad4x4x8 = vp8_sad4x4x8_sse4; cpi->rtcd.search.full_search = vp8_full_search_sadx8; } #endif |