diff options
author | trav90 <travawine@palemoon.org> | 2018-10-19 21:52:15 -0500 |
---|---|---|
committer | trav90 <travawine@palemoon.org> | 2018-10-19 21:52:20 -0500 |
commit | bbcc64772580c8a979288791afa02d30bc476d2e (patch) | |
tree | 437ce94c3fdd7497508e5b55de06c6d011678597 /third_party/aom/av1/encoder/rd.c | |
parent | 14805f6ddbfb173c327768fff9f81f40ce5e81b0 (diff) | |
download | UXP-bbcc64772580c8a979288791afa02d30bc476d2e.tar UXP-bbcc64772580c8a979288791afa02d30bc476d2e.tar.gz UXP-bbcc64772580c8a979288791afa02d30bc476d2e.tar.lz UXP-bbcc64772580c8a979288791afa02d30bc476d2e.tar.xz UXP-bbcc64772580c8a979288791afa02d30bc476d2e.zip |
Update aom to v1.0.0
Update aom to commit id d14c5bb4f336ef1842046089849dee4a301fbbf0.
Diffstat (limited to 'third_party/aom/av1/encoder/rd.c')
-rw-r--r-- | third_party/aom/av1/encoder/rd.c | 1215 |
1 files changed, 305 insertions, 910 deletions
diff --git a/third_party/aom/av1/encoder/rd.c b/third_party/aom/av1/encoder/rd.c index 5dd485334..17f23e5ec 100644 --- a/third_party/aom/av1/encoder/rd.c +++ b/third_party/aom/av1/encoder/rd.c @@ -13,7 +13,7 @@ #include <math.h> #include <stdio.h> -#include "./av1_rtcd.h" +#include "config/av1_rtcd.h" #include "aom_dsp/aom_dsp_common.h" #include "aom_mem/aom_mem.h" @@ -36,9 +36,7 @@ #include "av1/encoder/encodemb.h" #include "av1/encoder/encodemv.h" #include "av1/encoder/encoder.h" -#if CONFIG_LV_MAP #include "av1/encoder/encodetxb.h" -#endif #include "av1/encoder/mcomp.h" #include "av1/encoder/ratectrl.h" #include "av1/encoder/rd.h" @@ -54,114 +52,96 @@ // This table is used to correct for block size. // The factors here are << 2 (2 = x0.5, 32 = x8 etc). static const uint8_t rd_thresh_block_size_factor[BLOCK_SIZES_ALL] = { -#if CONFIG_CHROMA_2X2 || CONFIG_CHROMA_SUB8X8 - 2, 2, 2, -#endif - 2, 3, 3, 4, 6, 6, 8, 12, 12, 16, 24, 24, 32, -#if CONFIG_EXT_PARTITION - 48, 48, 64, -#endif // CONFIG_EXT_PARTITION - 4, 4, 8, 8, 16, 16, -#if CONFIG_EXT_PARTITION - 32, 32 -#endif // CONFIG_EXT_PARTITION + 2, 3, 3, 4, 6, 6, 8, 12, 12, 16, 24, 24, 32, 48, 48, 64, 4, 4, 8, 8, 16, 16 }; -#if CONFIG_EXT_TX static const int use_intra_ext_tx_for_txsize[EXT_TX_SETS_INTRA][EXT_TX_SIZES] = { -#if CONFIG_CHROMA_2X2 - { 1, 1, 1, 1, 1 }, // unused - { 0, 1, 1, 0, 0 }, - { 0, 0, 0, 1, 0 }, -#if CONFIG_MRC_TX - { 0, 0, 0, 0, 1 }, -#endif // CONFIG_MRC_TX -#else // CONFIG_CHROMA_2X2 { 1, 1, 1, 1 }, // unused { 1, 1, 0, 0 }, { 0, 0, 1, 0 }, -#if CONFIG_MRC_TX - { 0, 0, 0, 1 }, -#endif // CONFIG_MRC_TX -#endif // CONFIG_CHROMA_2X2 }; static const int use_inter_ext_tx_for_txsize[EXT_TX_SETS_INTER][EXT_TX_SIZES] = { -#if CONFIG_CHROMA_2X2 - { 1, 1, 1, 1, 1 }, // unused - { 0, 1, 1, 0, 0 }, { 0, 0, 0, 1, 0 }, { 0, 0, 0, 0, 1 }, -#if CONFIG_MRC_TX - { 0, 0, 0, 0, 1 }, -#endif // CONFIG_MRC_TX -#else // CONFIG_CHROMA_2X2 { 1, 1, 1, 1 }, // unused - { 1, 1, 0, 0 }, { 0, 0, 1, 0 }, { 0, 0, 0, 1 }, -#if CONFIG_MRC_TX + { 1, 1, 0, 0 }, + { 0, 0, 1, 0 }, { 0, 0, 0, 1 }, -#endif // CONFIG_MRC_TX -#endif // CONFIG_CHROMA_2X2 }; -#endif // CONFIG_EXT_TX + +static const int av1_ext_tx_set_idx_to_type[2][AOMMAX(EXT_TX_SETS_INTRA, + EXT_TX_SETS_INTER)] = { + { + // Intra + EXT_TX_SET_DCTONLY, + EXT_TX_SET_DTT4_IDTX_1DDCT, + EXT_TX_SET_DTT4_IDTX, + }, + { + // Inter + EXT_TX_SET_DCTONLY, + EXT_TX_SET_ALL16, + EXT_TX_SET_DTT9_IDTX_1DDCT, + EXT_TX_SET_DCT_IDTX, + }, +}; void av1_fill_mode_rates(AV1_COMMON *const cm, MACROBLOCK *x, FRAME_CONTEXT *fc) { int i, j; - if (cm->frame_type == KEY_FRAME) { - for (i = 0; i < PARTITION_CONTEXTS_PRIMARY; ++i) - av1_cost_tokens_from_cdf(x->partition_cost[i], fc->partition_cdf[i], + for (i = 0; i < PARTITION_CONTEXTS; ++i) + av1_cost_tokens_from_cdf(x->partition_cost[i], fc->partition_cdf[i], NULL); + + if (cm->skip_mode_flag) { + for (i = 0; i < SKIP_CONTEXTS; ++i) { + av1_cost_tokens_from_cdf(x->skip_mode_cost[i], fc->skip_mode_cdfs[i], NULL); -#if CONFIG_UNPOISON_PARTITION_CTX - for (; i < PARTITION_CONTEXTS_PRIMARY + PARTITION_BLOCK_SIZES; ++i) { - aom_prob p = fc->partition_prob[i][PARTITION_VERT]; - assert(p > 0); - x->partition_cost[i][PARTITION_NONE] = INT_MAX; - x->partition_cost[i][PARTITION_HORZ] = INT_MAX; - x->partition_cost[i][PARTITION_VERT] = av1_cost_bit(p, 0); - x->partition_cost[i][PARTITION_SPLIT] = av1_cost_bit(p, 1); - } - for (; i < PARTITION_CONTEXTS_PRIMARY + 2 * PARTITION_BLOCK_SIZES; ++i) { - aom_prob p = fc->partition_prob[i][PARTITION_HORZ]; - assert(p > 0); - x->partition_cost[i][PARTITION_NONE] = INT_MAX; - x->partition_cost[i][PARTITION_HORZ] = av1_cost_bit(p, 0); - x->partition_cost[i][PARTITION_VERT] = INT_MAX; - x->partition_cost[i][PARTITION_SPLIT] = av1_cost_bit(p, 1); } - x->partition_cost[PARTITION_CONTEXTS][PARTITION_NONE] = INT_MAX; - x->partition_cost[PARTITION_CONTEXTS][PARTITION_HORZ] = INT_MAX; - x->partition_cost[PARTITION_CONTEXTS][PARTITION_VERT] = INT_MAX; - x->partition_cost[PARTITION_CONTEXTS][PARTITION_SPLIT] = 0; -#endif // CONFIG_UNPOISON_PARTITION_CTX } -#if CONFIG_KF_CTX + for (i = 0; i < SKIP_CONTEXTS; ++i) { + av1_cost_tokens_from_cdf(x->skip_cost[i], fc->skip_cdfs[i], NULL); + } + for (i = 0; i < KF_MODE_CONTEXTS; ++i) for (j = 0; j < KF_MODE_CONTEXTS; ++j) av1_cost_tokens_from_cdf(x->y_mode_costs[i][j], fc->kf_y_cdf[i][j], NULL); -#else - for (i = 0; i < INTRA_MODES; ++i) - for (j = 0; j < INTRA_MODES; ++j) - av1_cost_tokens_from_cdf(x->y_mode_costs[i][j], fc->kf_y_cdf[i][j], NULL); -#endif for (i = 0; i < BLOCK_SIZE_GROUPS; ++i) av1_cost_tokens_from_cdf(x->mbmode_cost[i], fc->y_mode_cdf[i], NULL); - for (i = 0; i < INTRA_MODES; ++i) - av1_cost_tokens_from_cdf(x->intra_uv_mode_cost[i], fc->uv_mode_cdf[i], - NULL); + for (i = 0; i < CFL_ALLOWED_TYPES; ++i) + for (j = 0; j < INTRA_MODES; ++j) + av1_cost_tokens_from_cdf(x->intra_uv_mode_cost[i][j], + fc->uv_mode_cdf[i][j], NULL); + + av1_cost_tokens_from_cdf(x->filter_intra_mode_cost, fc->filter_intra_mode_cdf, + NULL); + for (i = 0; i < BLOCK_SIZES_ALL; ++i) { + if (av1_filter_intra_allowed_bsize(cm, i)) + av1_cost_tokens_from_cdf(x->filter_intra_cost[i], + fc->filter_intra_cdfs[i], NULL); + } for (i = 0; i < SWITCHABLE_FILTER_CONTEXTS; ++i) av1_cost_tokens_from_cdf(x->switchable_interp_costs[i], fc->switchable_interp_cdf[i], NULL); - for (i = 0; i < PALETTE_BLOCK_SIZES; ++i) { + for (i = 0; i < PALATTE_BSIZE_CTXS; ++i) { av1_cost_tokens_from_cdf(x->palette_y_size_cost[i], fc->palette_y_size_cdf[i], NULL); av1_cost_tokens_from_cdf(x->palette_uv_size_cost[i], fc->palette_uv_size_cdf[i], NULL); + for (j = 0; j < PALETTE_Y_MODE_CONTEXTS; ++j) { + av1_cost_tokens_from_cdf(x->palette_y_mode_cost[i][j], + fc->palette_y_mode_cdf[i][j], NULL); + } + } + + for (i = 0; i < PALETTE_UV_MODE_CONTEXTS; ++i) { + av1_cost_tokens_from_cdf(x->palette_uv_mode_cost[i], + fc->palette_uv_mode_cdf[i], NULL); } for (i = 0; i < PALETTE_SIZES; ++i) { @@ -172,60 +152,38 @@ void av1_fill_mode_rates(AV1_COMMON *const cm, MACROBLOCK *x, fc->palette_uv_color_index_cdf[i][j], NULL); } } -#if CONFIG_MRC_TX - for (i = 0; i < PALETTE_SIZES; ++i) { - for (j = 0; j < PALETTE_COLOR_INDEX_CONTEXTS; ++j) { - av1_cost_tokens_from_cdf(x->mrc_mask_inter_cost[i][j], - fc->mrc_mask_inter_cdf[i][j], NULL); - av1_cost_tokens_from_cdf(x->mrc_mask_intra_cost[i][j], - fc->mrc_mask_intra_cdf[i][j], NULL); - } - } -#endif // CONFIG_MRC_TX -#if CONFIG_CFL int sign_cost[CFL_JOINT_SIGNS]; av1_cost_tokens_from_cdf(sign_cost, fc->cfl_sign_cdf, NULL); for (int joint_sign = 0; joint_sign < CFL_JOINT_SIGNS; joint_sign++) { - const aom_cdf_prob *cdf_u = fc->cfl_alpha_cdf[CFL_CONTEXT_U(joint_sign)]; - const aom_cdf_prob *cdf_v = fc->cfl_alpha_cdf[CFL_CONTEXT_V(joint_sign)]; int *cost_u = x->cfl_cost[joint_sign][CFL_PRED_U]; int *cost_v = x->cfl_cost[joint_sign][CFL_PRED_V]; - if (CFL_SIGN_U(joint_sign) == CFL_SIGN_ZERO) + if (CFL_SIGN_U(joint_sign) == CFL_SIGN_ZERO) { memset(cost_u, 0, CFL_ALPHABET_SIZE * sizeof(*cost_u)); - else + } else { + const aom_cdf_prob *cdf_u = fc->cfl_alpha_cdf[CFL_CONTEXT_U(joint_sign)]; av1_cost_tokens_from_cdf(cost_u, cdf_u, NULL); - if (CFL_SIGN_V(joint_sign) == CFL_SIGN_ZERO) + } + if (CFL_SIGN_V(joint_sign) == CFL_SIGN_ZERO) { memset(cost_v, 0, CFL_ALPHABET_SIZE * sizeof(*cost_v)); - else + } else { + const aom_cdf_prob *cdf_v = fc->cfl_alpha_cdf[CFL_CONTEXT_V(joint_sign)]; av1_cost_tokens_from_cdf(cost_v, cdf_v, NULL); + } for (int u = 0; u < CFL_ALPHABET_SIZE; u++) cost_u[u] += sign_cost[joint_sign]; } -#endif // CONFIG_CFL - for (i = 0; i < MAX_TX_DEPTH; ++i) + for (i = 0; i < MAX_TX_CATS; ++i) for (j = 0; j < TX_SIZE_CONTEXTS; ++j) av1_cost_tokens_from_cdf(x->tx_size_cost[i][j], fc->tx_size_cdf[i][j], NULL); -#if CONFIG_EXT_TX -#if CONFIG_LGT_FROM_PRED - if (LGT_FROM_PRED_INTRA) { - for (i = 0; i < LGT_SIZES; ++i) { - for (j = 0; j < INTRA_MODES; ++j) { - x->intra_lgt_cost[i][j][0] = av1_cost_bit(fc->intra_lgt_prob[i][j], 0); - x->intra_lgt_cost[i][j][1] = av1_cost_bit(fc->intra_lgt_prob[i][j], 1); - } - } - } - if (LGT_FROM_PRED_INTER) { - for (i = 0; i < LGT_SIZES; ++i) { - x->inter_lgt_cost[i][0] = av1_cost_bit(fc->inter_lgt_prob[i], 0); - x->inter_lgt_cost[i][1] = av1_cost_bit(fc->inter_lgt_prob[i], 1); - } + for (i = 0; i < TXFM_PARTITION_CONTEXTS; ++i) { + av1_cost_tokens_from_cdf(x->txfm_partition_cost[i], + fc->txfm_partition_cdf[i], NULL); } -#endif // CONFIG_LGT_FROM_PRED + for (i = TX_4X4; i < EXT_TX_SIZES; ++i) { int s; for (s = 1; s < EXT_TX_SETS_INTER; ++s) { @@ -245,125 +203,124 @@ void av1_fill_mode_rates(AV1_COMMON *const cm, MACROBLOCK *x, } } } -#else - for (i = TX_4X4; i < EXT_TX_SIZES; ++i) { - for (j = 0; j < TX_TYPES; ++j) - av1_cost_tokens_from_cdf(x->intra_tx_type_costs[i][j], - fc->intra_ext_tx_cdf[i][j], av1_ext_tx_inv); - } - for (i = TX_4X4; i < EXT_TX_SIZES; ++i) { - av1_cost_tokens_from_cdf(x->inter_tx_type_costs[i], fc->inter_ext_tx_cdf[i], - av1_ext_tx_inv); - } -#endif // CONFIG_EXT_TX -#if CONFIG_EXT_INTRA -#if CONFIG_INTRA_INTERP - for (i = 0; i < INTRA_FILTERS + 1; ++i) - av1_cost_tokens_from_cdf(x->intra_filter_cost[i], fc->intra_filter_cdf[i], + for (i = 0; i < DIRECTIONAL_MODES; ++i) { + av1_cost_tokens_from_cdf(x->angle_delta_cost[i], fc->angle_delta_cdf[i], NULL); -#endif // CONFIG_INTRA_INTERP -#endif // CONFIG_EXT_INTRA -#if CONFIG_LOOP_RESTORATION - av1_cost_tokens(x->switchable_restore_cost, fc->switchable_restore_prob, - av1_switchable_restore_tree); -#endif // CONFIG_LOOP_RESTORATION -#if CONFIG_INTRABC + } + av1_cost_tokens_from_cdf(x->switchable_restore_cost, + fc->switchable_restore_cdf, NULL); + av1_cost_tokens_from_cdf(x->wiener_restore_cost, fc->wiener_restore_cdf, + NULL); + av1_cost_tokens_from_cdf(x->sgrproj_restore_cost, fc->sgrproj_restore_cdf, + NULL); av1_cost_tokens_from_cdf(x->intrabc_cost, fc->intrabc_cdf, NULL); -#endif // CONFIG_INTRABC if (!frame_is_intra_only(cm)) { + for (i = 0; i < COMP_INTER_CONTEXTS; ++i) { + av1_cost_tokens_from_cdf(x->comp_inter_cost[i], fc->comp_inter_cdf[i], + NULL); + } + + for (i = 0; i < REF_CONTEXTS; ++i) { + for (j = 0; j < SINGLE_REFS - 1; ++j) { + av1_cost_tokens_from_cdf(x->single_ref_cost[i][j], + fc->single_ref_cdf[i][j], NULL); + } + } + + for (i = 0; i < COMP_REF_TYPE_CONTEXTS; ++i) { + av1_cost_tokens_from_cdf(x->comp_ref_type_cost[i], + fc->comp_ref_type_cdf[i], NULL); + } + + for (i = 0; i < UNI_COMP_REF_CONTEXTS; ++i) { + for (j = 0; j < UNIDIR_COMP_REFS - 1; ++j) { + av1_cost_tokens_from_cdf(x->uni_comp_ref_cost[i][j], + fc->uni_comp_ref_cdf[i][j], NULL); + } + } + + for (i = 0; i < REF_CONTEXTS; ++i) { + for (j = 0; j < FWD_REFS - 1; ++j) { + av1_cost_tokens_from_cdf(x->comp_ref_cost[i][j], fc->comp_ref_cdf[i][j], + NULL); + } + } + + for (i = 0; i < REF_CONTEXTS; ++i) { + for (j = 0; j < BWD_REFS - 1; ++j) { + av1_cost_tokens_from_cdf(x->comp_bwdref_cost[i][j], + fc->comp_bwdref_cdf[i][j], NULL); + } + } + + for (i = 0; i < INTRA_INTER_CONTEXTS; ++i) { + av1_cost_tokens_from_cdf(x->intra_inter_cost[i], fc->intra_inter_cdf[i], + NULL); + } + for (i = 0; i < NEWMV_MODE_CONTEXTS; ++i) { -#if CONFIG_NEW_MULTISYMBOL av1_cost_tokens_from_cdf(x->newmv_mode_cost[i], fc->newmv_cdf[i], NULL); -#else - x->newmv_mode_cost[i][0] = av1_cost_bit(fc->newmv_prob[i], 0); - x->newmv_mode_cost[i][1] = av1_cost_bit(fc->newmv_prob[i], 1); -#endif } - for (i = 0; i < ZEROMV_MODE_CONTEXTS; ++i) { -#if CONFIG_NEW_MULTISYMBOL + for (i = 0; i < GLOBALMV_MODE_CONTEXTS; ++i) { av1_cost_tokens_from_cdf(x->zeromv_mode_cost[i], fc->zeromv_cdf[i], NULL); -#else - x->zeromv_mode_cost[i][0] = av1_cost_bit(fc->zeromv_prob[i], 0); - x->zeromv_mode_cost[i][1] = av1_cost_bit(fc->zeromv_prob[i], 1); -#endif } for (i = 0; i < REFMV_MODE_CONTEXTS; ++i) { -#if CONFIG_NEW_MULTISYMBOL av1_cost_tokens_from_cdf(x->refmv_mode_cost[i], fc->refmv_cdf[i], NULL); -#else - x->refmv_mode_cost[i][0] = av1_cost_bit(fc->refmv_prob[i], 0); - x->refmv_mode_cost[i][1] = av1_cost_bit(fc->refmv_prob[i], 1); -#endif } for (i = 0; i < DRL_MODE_CONTEXTS; ++i) { -#if CONFIG_NEW_MULTISYMBOL av1_cost_tokens_from_cdf(x->drl_mode_cost0[i], fc->drl_cdf[i], NULL); -#else - x->drl_mode_cost0[i][0] = av1_cost_bit(fc->drl_prob[i], 0); - x->drl_mode_cost0[i][1] = av1_cost_bit(fc->drl_prob[i], 1); -#endif } for (i = 0; i < INTER_MODE_CONTEXTS; ++i) av1_cost_tokens_from_cdf(x->inter_compound_mode_cost[i], fc->inter_compound_mode_cdf[i], NULL); -#if CONFIG_WEDGE || CONFIG_COMPOUND_SEGMENT for (i = 0; i < BLOCK_SIZES_ALL; ++i) av1_cost_tokens_from_cdf(x->compound_type_cost[i], fc->compound_type_cdf[i], NULL); -#endif // CONFIG_WEDGE || CONFIG_COMPOUND_SEGMENT -#if CONFIG_COMPOUND_SINGLEREF - for (i = 0; i < INTER_MODE_CONTEXTS; ++i) - av1_cost_tokens_from_cdf(x->inter_singleref_comp_mode_cost[i], - fc->inter_singleref_comp_mode_cdf[i], NULL); -#endif // CONFIG_COMPOUND_SINGLEREF -#if CONFIG_INTERINTRA - for (i = 0; i < BLOCK_SIZE_GROUPS; ++i) + for (i = 0; i < BLOCK_SIZES_ALL; ++i) { + if (get_interinter_wedge_bits(i)) { + av1_cost_tokens_from_cdf(x->wedge_idx_cost[i], fc->wedge_idx_cdf[i], + NULL); + } + } + for (i = 0; i < BLOCK_SIZE_GROUPS; ++i) { + av1_cost_tokens_from_cdf(x->interintra_cost[i], fc->interintra_cdf[i], + NULL); av1_cost_tokens_from_cdf(x->interintra_mode_cost[i], fc->interintra_mode_cdf[i], NULL); -#endif // CONFIG_INTERINTRA -#if CONFIG_MOTION_VAR || CONFIG_WARPED_MOTION + } + for (i = 0; i < BLOCK_SIZES_ALL; ++i) { + av1_cost_tokens_from_cdf(x->wedge_interintra_cost[i], + fc->wedge_interintra_cdf[i], NULL); + } for (i = BLOCK_8X8; i < BLOCK_SIZES_ALL; i++) { av1_cost_tokens_from_cdf(x->motion_mode_cost[i], fc->motion_mode_cdf[i], NULL); } -#if CONFIG_MOTION_VAR && CONFIG_WARPED_MOTION for (i = BLOCK_8X8; i < BLOCK_SIZES_ALL; i++) { -#if CONFIG_NCOBMC_ADAPT_WEIGHT - av1_cost_tokens_from_cdf(x->motion_mode_cost2[i], fc->ncobmc_cdf[i], - NULL); -#endif -#if CONFIG_NEW_MULTISYMBOL || CONFIG_NCOBMC_ADAPT_WEIGHT av1_cost_tokens_from_cdf(x->motion_mode_cost1[i], fc->obmc_cdf[i], NULL); -#else - x->motion_mode_cost1[i][0] = av1_cost_bit(fc->obmc_prob[i], 0); - x->motion_mode_cost1[i][1] = av1_cost_bit(fc->obmc_prob[i], 1); -#endif } -#endif // CONFIG_MOTION_VAR && CONFIG_WARPED_MOTION -#if CONFIG_MOTION_VAR && CONFIG_NCOBMC_ADAPT_WEIGHT - for (i = ADAPT_OVERLAP_BLOCK_8X8; i < ADAPT_OVERLAP_BLOCKS; ++i) { - av1_cost_tokens_from_cdf(x->ncobmc_mode_cost[i], fc->ncobmc_mode_cdf[i], + for (i = 0; i < COMP_INDEX_CONTEXTS; ++i) { + av1_cost_tokens_from_cdf(x->comp_idx_cost[i], fc->compound_index_cdf[i], NULL); } -#endif // CONFIG_MOTION_VAR && CONFIG_NCOBMC_ADAPT_WEIGHT -#endif // CONFIG_MOTION_VAR || CONFIG_WARPED_MOTION + for (i = 0; i < COMP_GROUP_IDX_CONTEXTS; ++i) { + av1_cost_tokens_from_cdf(x->comp_group_idx_cost[i], + fc->comp_group_idx_cdf[i], NULL); + } } } // Values are now correlated to quantizer. static int sad_per_bit16lut_8[QINDEX_RANGE]; static int sad_per_bit4lut_8[QINDEX_RANGE]; - -#if CONFIG_HIGHBITDEPTH static int sad_per_bit16lut_10[QINDEX_RANGE]; static int sad_per_bit4lut_10[QINDEX_RANGE]; static int sad_per_bit16lut_12[QINDEX_RANGE]; static int sad_per_bit4lut_12[QINDEX_RANGE]; -#endif static void init_me_luts_bd(int *bit16lut, int *bit4lut, int range, aom_bit_depth_t bit_depth) { @@ -381,31 +338,26 @@ static void init_me_luts_bd(int *bit16lut, int *bit4lut, int range, void av1_init_me_luts(void) { init_me_luts_bd(sad_per_bit16lut_8, sad_per_bit4lut_8, QINDEX_RANGE, AOM_BITS_8); -#if CONFIG_HIGHBITDEPTH init_me_luts_bd(sad_per_bit16lut_10, sad_per_bit4lut_10, QINDEX_RANGE, AOM_BITS_10); init_me_luts_bd(sad_per_bit16lut_12, sad_per_bit4lut_12, QINDEX_RANGE, AOM_BITS_12); -#endif } static const int rd_boost_factor[16] = { 64, 32, 32, 32, 24, 16, 12, 12, 8, 8, 4, 4, 2, 2, 1, 0 }; static const int rd_frame_type_factor[FRAME_UPDATE_TYPES] = { 128, 144, 128, 128, 144, -#if CONFIG_EXT_REFS // TODO(zoeliu): To adjust further following factor values. 128, 128, 128, // TODO(weitinglin): We should investigate if the values should be the same // as the value used by OVERLAY frame 144, // INTNL_OVERLAY_UPDATE 128 // INTNL_ARF_UPDATE -#endif // CONFIG_EXT_REFS }; int av1_compute_rd_mult(const AV1_COMP *cpi, int qindex) { - const int64_t q = av1_dc_quant(qindex, 0, cpi->common.bit_depth); -#if CONFIG_HIGHBITDEPTH + const int64_t q = av1_dc_quant_Q3(qindex, 0, cpi->common.bit_depth); int64_t rdmult = 0; switch (cpi->common.bit_depth) { case AOM_BITS_8: rdmult = 88 * q * q / 24; break; @@ -415,9 +367,6 @@ int av1_compute_rd_mult(const AV1_COMP *cpi, int qindex) { assert(0 && "bit_depth should be AOM_BITS_8, AOM_BITS_10 or AOM_BITS_12"); return -1; } -#else - int64_t rdmult = 88 * q * q / 24; -#endif // CONFIG_HIGHBITDEPTH if (cpi->oxcf.pass == 2 && (cpi->common.frame_type != KEY_FRAME)) { const GF_GROUP *const gf_group = &cpi->twopass.gf_group; const FRAME_UPDATE_TYPE frame_type = gf_group->update_type[gf_group->index]; @@ -432,25 +381,19 @@ int av1_compute_rd_mult(const AV1_COMP *cpi, int qindex) { static int compute_rd_thresh_factor(int qindex, aom_bit_depth_t bit_depth) { double q; -#if CONFIG_HIGHBITDEPTH switch (bit_depth) { - case AOM_BITS_8: q = av1_dc_quant(qindex, 0, AOM_BITS_8) / 4.0; break; - case AOM_BITS_10: q = av1_dc_quant(qindex, 0, AOM_BITS_10) / 16.0; break; - case AOM_BITS_12: q = av1_dc_quant(qindex, 0, AOM_BITS_12) / 64.0; break; + case AOM_BITS_8: q = av1_dc_quant_Q3(qindex, 0, AOM_BITS_8) / 4.0; break; + case AOM_BITS_10: q = av1_dc_quant_Q3(qindex, 0, AOM_BITS_10) / 16.0; break; + case AOM_BITS_12: q = av1_dc_quant_Q3(qindex, 0, AOM_BITS_12) / 64.0; break; default: assert(0 && "bit_depth should be AOM_BITS_8, AOM_BITS_10 or AOM_BITS_12"); return -1; } -#else - (void)bit_depth; - q = av1_dc_quant(qindex, 0, AOM_BITS_8) / 4.0; -#endif // CONFIG_HIGHBITDEPTH // TODO(debargha): Adjust the function below. return AOMMAX((int)(pow(q, RD_THRESH_POW) * 5.12), 8); } void av1_initialize_me_consts(const AV1_COMP *cpi, MACROBLOCK *x, int qindex) { -#if CONFIG_HIGHBITDEPTH switch (cpi->common.bit_depth) { case AOM_BITS_8: x->sadperbit16 = sad_per_bit16lut_8[qindex]; @@ -467,11 +410,6 @@ void av1_initialize_me_consts(const AV1_COMP *cpi, MACROBLOCK *x, int qindex) { default: assert(0 && "bit_depth should be AOM_BITS_8, AOM_BITS_10 or AOM_BITS_12"); } -#else - (void)cpi; - x->sadperbit16 = sad_per_bit16lut_8[qindex]; - x->sadperbit4 = sad_per_bit4lut_8[qindex]; -#endif // CONFIG_HIGHBITDEPTH } static void set_block_thresholds(const AV1_COMMON *cm, RD_OPT *rd) { @@ -490,195 +428,89 @@ static void set_block_thresholds(const AV1_COMMON *cm, RD_OPT *rd) { const int t = q * rd_thresh_block_size_factor[bsize]; const int thresh_max = INT_MAX / t; -#if CONFIG_CB4X4 for (i = 0; i < MAX_MODES; ++i) rd->threshes[segment_id][bsize][i] = rd->thresh_mult[i] < thresh_max ? rd->thresh_mult[i] * t / 4 : INT_MAX; -#else - if (bsize >= BLOCK_8X8) { - for (i = 0; i < MAX_MODES; ++i) - rd->threshes[segment_id][bsize][i] = rd->thresh_mult[i] < thresh_max - ? rd->thresh_mult[i] * t / 4 - : INT_MAX; - } else { - for (i = 0; i < MAX_REFS; ++i) - rd->threshes[segment_id][bsize][i] = - rd->thresh_mult_sub8x8[i] < thresh_max - ? rd->thresh_mult_sub8x8[i] * t / 4 - : INT_MAX; - } -#endif } } } -void av1_set_mvcost(MACROBLOCK *x, MV_REFERENCE_FRAME ref_frame, int ref, - int ref_mv_idx) { - MB_MODE_INFO_EXT *mbmi_ext = x->mbmi_ext; - int8_t rf_type = av1_ref_frame_type(x->e_mbd.mi[0]->mbmi.ref_frame); - int nmv_ctx = av1_nmv_ctx(mbmi_ext->ref_mv_count[rf_type], - mbmi_ext->ref_mv_stack[rf_type], ref, ref_mv_idx); - (void)ref_frame; - x->mvcost = x->mv_cost_stack[nmv_ctx]; - x->nmvjointcost = x->nmv_vec_cost[nmv_ctx]; +void av1_set_mvcost(MACROBLOCK *x, int ref, int ref_mv_idx) { + (void)ref; + (void)ref_mv_idx; + x->mvcost = x->mv_cost_stack; + x->nmvjointcost = x->nmv_vec_cost; } -#if CONFIG_LV_MAP -#if !LV_MAP_PROB -static void get_rate_cost(aom_prob p, int cost[2]) { - cost[0] = av1_cost_bit(p, 0); - cost[1] = av1_cost_bit(p, 1); -} -#endif // !LV_MAP_PROB - -void av1_fill_coeff_costs(MACROBLOCK *x, FRAME_CONTEXT *fc) { +void av1_fill_coeff_costs(MACROBLOCK *x, FRAME_CONTEXT *fc, + const int num_planes) { + const int nplanes = AOMMIN(num_planes, PLANE_TYPES); + for (int eob_multi_size = 0; eob_multi_size < 7; ++eob_multi_size) { + for (int plane = 0; plane < nplanes; ++plane) { + LV_MAP_EOB_COST *pcost = &x->eob_costs[eob_multi_size][plane]; + + for (int ctx = 0; ctx < 2; ++ctx) { + aom_cdf_prob *pcdf; + switch (eob_multi_size) { + case 0: pcdf = fc->eob_flag_cdf16[plane][ctx]; break; + case 1: pcdf = fc->eob_flag_cdf32[plane][ctx]; break; + case 2: pcdf = fc->eob_flag_cdf64[plane][ctx]; break; + case 3: pcdf = fc->eob_flag_cdf128[plane][ctx]; break; + case 4: pcdf = fc->eob_flag_cdf256[plane][ctx]; break; + case 5: pcdf = fc->eob_flag_cdf512[plane][ctx]; break; + case 6: + default: pcdf = fc->eob_flag_cdf1024[plane][ctx]; break; + } + av1_cost_tokens_from_cdf(pcost->eob_cost[ctx], pcdf, NULL); + } + } + } for (int tx_size = 0; tx_size < TX_SIZES; ++tx_size) { - for (int plane = 0; plane < PLANE_TYPES; ++plane) { + for (int plane = 0; plane < nplanes; ++plane) { LV_MAP_COEFF_COST *pcost = &x->coeff_costs[tx_size][plane]; -#if LV_MAP_PROB for (int ctx = 0; ctx < TXB_SKIP_CONTEXTS; ++ctx) av1_cost_tokens_from_cdf(pcost->txb_skip_cost[ctx], fc->txb_skip_cdf[tx_size][ctx], NULL); + for (int ctx = 0; ctx < SIG_COEF_CONTEXTS_EOB; ++ctx) + av1_cost_tokens_from_cdf(pcost->base_eob_cost[ctx], + fc->coeff_base_eob_cdf[tx_size][plane][ctx], + NULL); for (int ctx = 0; ctx < SIG_COEF_CONTEXTS; ++ctx) - av1_cost_tokens_from_cdf(pcost->nz_map_cost[ctx], - fc->nz_map_cdf[tx_size][plane][ctx], NULL); + av1_cost_tokens_from_cdf(pcost->base_cost[ctx], + fc->coeff_base_cdf[tx_size][plane][ctx], NULL); for (int ctx = 0; ctx < EOB_COEF_CONTEXTS; ++ctx) - av1_cost_tokens_from_cdf(pcost->eob_cost[ctx], - fc->eob_flag_cdf[tx_size][plane][ctx], NULL); + av1_cost_tokens_from_cdf(pcost->eob_extra_cost[ctx], + fc->eob_extra_cdf[tx_size][plane][ctx], NULL); for (int ctx = 0; ctx < DC_SIGN_CONTEXTS; ++ctx) av1_cost_tokens_from_cdf(pcost->dc_sign_cost[ctx], fc->dc_sign_cdf[plane][ctx], NULL); - for (int layer = 0; layer < NUM_BASE_LEVELS; ++layer) - for (int ctx = 0; ctx < COEFF_BASE_CONTEXTS; ++ctx) - av1_cost_tokens_from_cdf( - pcost->base_cost[layer][ctx], - fc->coeff_base_cdf[tx_size][plane][layer][ctx], NULL); - -#if BR_NODE - for (int br = 0; br < BASE_RANGE_SETS; ++br) - for (int ctx = 0; ctx < LEVEL_CONTEXTS; ++ctx) - av1_cost_tokens_from_cdf(pcost->br_cost[br][ctx], - fc->coeff_br_cdf[tx_size][plane][br][ctx], - NULL); - for (int ctx = 0; ctx < LEVEL_CONTEXTS; ++ctx) { - int lps_rate[2]; - av1_cost_tokens_from_cdf(lps_rate, - fc->coeff_lps_cdf[tx_size][plane][ctx], NULL); - - for (int base_range = 0; base_range < COEFF_BASE_RANGE + 1; - ++base_range) { - int br_set_idx = base_range < COEFF_BASE_RANGE - ? coeff_to_br_index[base_range] - : BASE_RANGE_SETS; - - pcost->lps_cost[ctx][base_range] = 0; - - for (int idx = 0; idx < BASE_RANGE_SETS; ++idx) { - if (idx == br_set_idx) { - pcost->lps_cost[ctx][base_range] += pcost->br_cost[idx][ctx][1]; - - int br_base = br_index_to_coeff[br_set_idx]; - int br_offset = base_range - br_base; - int extra_bits = (1 << br_extra_bits[idx]) - 1; - for (int tok = 0; tok < extra_bits; ++tok) { - if (tok == br_offset) { - pcost->lps_cost[ctx][base_range] += lps_rate[1]; - break; - } else { - pcost->lps_cost[ctx][base_range] += lps_rate[0]; - } - } - break; - } else { - pcost->lps_cost[ctx][base_range] += pcost->br_cost[idx][ctx][0]; - } - } - // load the base range cost - } - } -#else // BR_NODE - for (int ctx = 0; ctx < LEVEL_CONTEXTS; ++ctx) - av1_cost_tokens_from_cdf(pcost->lps_cost[ctx], - fc->coeff_lps_cdf[tx_size][plane][ctx], NULL); -#endif // BR_NODE -#if CONFIG_CTX1D - for (int tx_class = 0; tx_class < TX_CLASSES; ++tx_class) - av1_cost_tokens_from_cdf(pcost->eob_mode_cost[tx_class], - fc->eob_mode_cdf[tx_size][plane][tx_class], + int br_rate[BR_CDF_SIZE]; + int prev_cost = 0; + int i, j; + av1_cost_tokens_from_cdf(br_rate, fc->coeff_br_cdf[tx_size][plane][ctx], NULL); - - for (int tx_class = 0; tx_class < TX_CLASSES; ++tx_class) - for (int ctx = 0; ctx < EMPTY_LINE_CONTEXTS; ++ctx) - av1_cost_tokens_from_cdf( - pcost->empty_line_cost[tx_class][ctx], - fc->empty_line_cdf[tx_size][plane][tx_class][ctx], NULL); - - for (int tx_class = 0; tx_class < TX_CLASSES; ++tx_class) - for (int ctx = 0; ctx < HV_EOB_CONTEXTS; ++ctx) - av1_cost_tokens_from_cdf( - pcost->hv_eob_cost[tx_class][ctx], - fc->hv_eob_cdf[tx_size][plane][tx_class][ctx], NULL); -#endif // CONFIG_CTX1D -#else // LV_MAP_PROB - for (int ctx = 0; ctx < TXB_SKIP_CONTEXTS; ++ctx) - get_rate_cost(fc->txb_skip[tx_size][ctx], pcost->txb_skip_cost[ctx]); - - for (int ctx = 0; ctx < SIG_COEF_CONTEXTS; ++ctx) - get_rate_cost(fc->nz_map[tx_size][plane][ctx], pcost->nz_map_cost[ctx]); - - for (int ctx = 0; ctx < EOB_COEF_CONTEXTS; ++ctx) - get_rate_cost(fc->eob_flag[tx_size][plane][ctx], pcost->eob_cost[ctx]); - - for (int ctx = 0; ctx < DC_SIGN_CONTEXTS; ++ctx) - get_rate_cost(fc->dc_sign[plane][ctx], pcost->dc_sign_cost[ctx]); - - for (int layer = 0; layer < NUM_BASE_LEVELS; ++layer) - for (int ctx = 0; ctx < COEFF_BASE_CONTEXTS; ++ctx) - get_rate_cost(fc->coeff_base[tx_size][plane][layer][ctx], - pcost->base_cost[layer][ctx]); - - for (int ctx = 0; ctx < LEVEL_CONTEXTS; ++ctx) - get_rate_cost(fc->coeff_lps[tx_size][plane][ctx], pcost->lps_cost[ctx]); - -#if CONFIG_CTX1D - for (int tx_class = 0; tx_class < TX_CLASSES; ++tx_class) - get_rate_cost(fc->eob_mode[tx_size][plane][tx_class], - pcost->eob_mode_cost[tx_class]); - - for (int tx_class = 0; tx_class < TX_CLASSES; ++tx_class) - for (int ctx = 0; ctx < EMPTY_LINE_CONTEXTS; ++ctx) - get_rate_cost(fc->empty_line[tx_size][plane][tx_class][ctx], - pcost->empty_line_cost[tx_class][ctx]); - - for (int tx_class = 0; tx_class < TX_CLASSES; ++tx_class) - for (int ctx = 0; ctx < HV_EOB_CONTEXTS; ++ctx) - get_rate_cost(fc->hv_eob[tx_size][plane][tx_class][ctx], - pcost->hv_eob_cost[tx_class][ctx]); -#endif // CONFIG_CTX1D -#endif // LV_MAP_PROB - } - } -} -#endif // CONFIG_LV_MAP - -void av1_fill_token_costs_from_cdf(av1_coeff_cost *cost, - coeff_cdf_model (*cdf)[PLANE_TYPES]) { - for (int tx = 0; tx < TX_SIZES; ++tx) { - for (int pt = 0; pt < PLANE_TYPES; ++pt) { - for (int rt = 0; rt < REF_TYPES; ++rt) { - for (int band = 0; band < COEF_BANDS; ++band) { - for (int ctx = 0; ctx < BAND_COEFF_CONTEXTS(band); ++ctx) { - av1_cost_tokens_from_cdf(cost[tx][pt][rt][band][ctx], - cdf[tx][pt][rt][band][ctx], NULL); + // printf("br_rate: "); + // for(j = 0; j < BR_CDF_SIZE; j++) + // printf("%4d ", br_rate[j]); + // printf("\n"); + for (i = 0; i < COEFF_BASE_RANGE; i += BR_CDF_SIZE - 1) { + for (j = 0; j < BR_CDF_SIZE - 1; j++) { + pcost->lps_cost[ctx][i + j] = prev_cost + br_rate[j]; } + prev_cost += br_rate[j]; } + pcost->lps_cost[ctx][i] = prev_cost; + // printf("lps_cost: %d %d %2d : ", tx_size, plane, ctx); + // for (i = 0; i <= COEFF_BASE_RANGE; i++) + // printf("%5d ", pcost->lps_cost[ctx][i]); + // printf("\n"); } } } @@ -688,7 +520,6 @@ void av1_initialize_rd_consts(AV1_COMP *cpi) { AV1_COMMON *const cm = &cpi->common; MACROBLOCK *const x = &cpi->td.mb; RD_OPT *const rd = &cpi->rd; - int nmv_ctx; aom_clear_system_state(); @@ -698,56 +529,35 @@ void av1_initialize_rd_consts(AV1_COMP *cpi) { set_block_thresholds(cm, rd); - for (nmv_ctx = 0; nmv_ctx < NMV_CONTEXTS; ++nmv_ctx) { -#if CONFIG_AMVR - if (cm->cur_frame_mv_precision_level) { - av1_build_nmv_cost_table(x->nmv_vec_cost[nmv_ctx], x->nmvcost[nmv_ctx], - &cm->fc->nmvc[nmv_ctx], MV_SUBPEL_NONE); - } else { - av1_build_nmv_cost_table( - x->nmv_vec_cost[nmv_ctx], - cm->allow_high_precision_mv ? x->nmvcost_hp[nmv_ctx] - : x->nmvcost[nmv_ctx], - &cm->fc->nmvc[nmv_ctx], cm->allow_high_precision_mv); - } - -#else + if (cm->cur_frame_force_integer_mv) { + av1_build_nmv_cost_table(x->nmv_vec_cost, x->nmvcost, &cm->fc->nmvc, + MV_SUBPEL_NONE); + } else { av1_build_nmv_cost_table( - x->nmv_vec_cost[nmv_ctx], - cm->allow_high_precision_mv ? x->nmvcost_hp[nmv_ctx] - : x->nmvcost[nmv_ctx], - &cm->fc->nmvc[nmv_ctx], cm->allow_high_precision_mv); -#endif + x->nmv_vec_cost, + cm->allow_high_precision_mv ? x->nmvcost_hp : x->nmvcost, &cm->fc->nmvc, + cm->allow_high_precision_mv); } - x->mvcost = x->mv_cost_stack[0]; - x->nmvjointcost = x->nmv_vec_cost[0]; -#if CONFIG_INTRABC + x->mvcost = x->mv_cost_stack; + x->nmvjointcost = x->nmv_vec_cost; + if (frame_is_intra_only(cm) && cm->allow_screen_content_tools && cpi->oxcf.pass != 1) { - av1_build_nmv_cost_table( - x->nmv_vec_cost[0], - cm->allow_high_precision_mv ? x->nmvcost_hp[0] : x->nmvcost[0], - &cm->fc->ndvc, MV_SUBPEL_NONE); + int *dvcost[2] = { &cpi->dv_cost[0][MV_MAX], &cpi->dv_cost[1][MV_MAX] }; + av1_build_nmv_cost_table(cpi->dv_joint_cost, dvcost, &cm->fc->ndvc, + MV_SUBPEL_NONE); } -#endif -#if CONFIG_GLOBAL_MOTION if (cpi->oxcf.pass != 1) { for (int i = 0; i < TRANS_TYPES; ++i) -#if GLOBAL_TRANS_TYPES > 4 - cpi->gmtype_cost[i] = (1 + (i > 0 ? GLOBAL_TYPE_BITS : 0)) - << AV1_PROB_COST_SHIFT; -#else // IDENTITY: 1 bit // TRANSLATION: 3 bits // ROTZOOM: 2 bits // AFFINE: 3 bits cpi->gmtype_cost[i] = (1 + (i > 0 ? (i == ROTZOOM ? 1 : 2) : 0)) << AV1_PROB_COST_SHIFT; -#endif // GLOBAL_TRANS_TYPES > 4 } -#endif // CONFIG_GLOBAL_MOTION } static void model_rd_norm(int xsq_q10, int *r_q10, int *d_q10) { @@ -840,288 +650,32 @@ void av1_model_rd_from_var_lapndz(int64_t var, unsigned int n_log2, } } -static void get_entropy_contexts_plane( - BLOCK_SIZE plane_bsize, TX_SIZE tx_size, const struct macroblockd_plane *pd, - ENTROPY_CONTEXT t_above[2 * MAX_MIB_SIZE], - ENTROPY_CONTEXT t_left[2 * MAX_MIB_SIZE]) { +static void get_entropy_contexts_plane(BLOCK_SIZE plane_bsize, + const struct macroblockd_plane *pd, + ENTROPY_CONTEXT t_above[MAX_MIB_SIZE], + ENTROPY_CONTEXT t_left[MAX_MIB_SIZE]) { const int num_4x4_w = block_size_wide[plane_bsize] >> tx_size_wide_log2[0]; const int num_4x4_h = block_size_high[plane_bsize] >> tx_size_high_log2[0]; const ENTROPY_CONTEXT *const above = pd->above_context; const ENTROPY_CONTEXT *const left = pd->left_context; -#if CONFIG_LV_MAP memcpy(t_above, above, sizeof(ENTROPY_CONTEXT) * num_4x4_w); memcpy(t_left, left, sizeof(ENTROPY_CONTEXT) * num_4x4_h); - return; -#endif // CONFIG_LV_MAP - - int i; - -#if CONFIG_CHROMA_2X2 - switch (tx_size) { - case TX_2X2: - memcpy(t_above, above, sizeof(ENTROPY_CONTEXT) * num_4x4_w); - memcpy(t_left, left, sizeof(ENTROPY_CONTEXT) * num_4x4_h); - break; - case TX_4X4: - for (i = 0; i < num_4x4_w; i += 2) - t_above[i] = !!*(const uint16_t *)&above[i]; - for (i = 0; i < num_4x4_h; i += 2) - t_left[i] = !!*(const uint16_t *)&left[i]; - break; - case TX_8X8: - for (i = 0; i < num_4x4_w; i += 4) - t_above[i] = !!*(const uint32_t *)&above[i]; - for (i = 0; i < num_4x4_h; i += 4) - t_left[i] = !!*(const uint32_t *)&left[i]; - break; - case TX_16X16: - for (i = 0; i < num_4x4_w; i += 8) - t_above[i] = !!*(const uint64_t *)&above[i]; - for (i = 0; i < num_4x4_h; i += 8) - t_left[i] = !!*(const uint64_t *)&left[i]; - break; - case TX_32X32: - for (i = 0; i < num_4x4_w; i += 16) - t_above[i] = - !!(*(const uint64_t *)&above[i] | *(const uint64_t *)&above[i + 8]); - for (i = 0; i < num_4x4_h; i += 16) - t_left[i] = - !!(*(const uint64_t *)&left[i] | *(const uint64_t *)&left[i + 8]); - break; -#if CONFIG_TX64X64 - case TX_32X64: - for (i = 0; i < num_4x4_w; i += 16) - t_above[i] = - !!(*(const uint64_t *)&above[i] | *(const uint64_t *)&above[i + 8]); - for (i = 0; i < num_4x4_h; i += 32) - t_left[i] = - !!(*(const uint64_t *)&left[i] | *(const uint64_t *)&left[i + 8] | - *(const uint64_t *)&left[i + 16] | - *(const uint64_t *)&left[i + 24]); - break; - case TX_64X32: - for (i = 0; i < num_4x4_w; i += 32) - t_above[i] = - !!(*(const uint64_t *)&above[i] | *(const uint64_t *)&above[i + 8] | - *(const uint64_t *)&above[i + 16] | - *(const uint64_t *)&above[i + 24]); - for (i = 0; i < num_4x4_h; i += 16) - t_left[i] = - !!(*(const uint64_t *)&left[i] | *(const uint64_t *)&left[i + 8]); - break; - case TX_64X64: - for (i = 0; i < num_4x4_w; i += 32) - t_above[i] = - !!(*(const uint64_t *)&above[i] | *(const uint64_t *)&above[i + 8] | - *(const uint64_t *)&above[i + 16] | - *(const uint64_t *)&above[i + 24]); - for (i = 0; i < num_4x4_h; i += 32) - t_left[i] = - !!(*(const uint64_t *)&left[i] | *(const uint64_t *)&left[i + 8] | - *(const uint64_t *)&left[i + 16] | - *(const uint64_t *)&left[i + 24]); - break; -#endif // CONFIG_TX64X64 - case TX_4X8: - for (i = 0; i < num_4x4_w; i += 2) - t_above[i] = !!*(const uint16_t *)&above[i]; - for (i = 0; i < num_4x4_h; i += 4) - t_left[i] = !!*(const uint32_t *)&left[i]; - break; - case TX_8X4: - for (i = 0; i < num_4x4_w; i += 4) - t_above[i] = !!*(const uint32_t *)&above[i]; - for (i = 0; i < num_4x4_h; i += 2) - t_left[i] = !!*(const uint16_t *)&left[i]; - break; - case TX_8X16: - for (i = 0; i < num_4x4_w; i += 4) - t_above[i] = !!*(const uint32_t *)&above[i]; - for (i = 0; i < num_4x4_h; i += 8) - t_left[i] = !!*(const uint64_t *)&left[i]; - break; - case TX_16X8: - for (i = 0; i < num_4x4_w; i += 8) - t_above[i] = !!*(const uint64_t *)&above[i]; - for (i = 0; i < num_4x4_h; i += 4) - t_left[i] = !!*(const uint32_t *)&left[i]; - break; - case TX_16X32: - for (i = 0; i < num_4x4_w; i += 8) - t_above[i] = !!*(const uint64_t *)&above[i]; - for (i = 0; i < num_4x4_h; i += 16) - t_left[i] = - !!(*(const uint64_t *)&left[i] | *(const uint64_t *)&left[i + 8]); - break; - case TX_32X16: - for (i = 0; i < num_4x4_w; i += 16) - t_above[i] = - !!(*(const uint64_t *)&above[i] | *(const uint64_t *)&above[i + 8]); - for (i = 0; i < num_4x4_h; i += 8) - t_left[i] = !!*(const uint64_t *)&left[i]; - break; -#if CONFIG_RECT_TX_EXT && (CONFIG_EXT_TX || CONFIG_VAR_TX) - case TX_4X16: - for (i = 0; i < num_4x4_w; i += 2) - t_above[i] = !!*(const uint16_t *)&above[i]; - for (i = 0; i < num_4x4_h; i += 8) - t_left[i] = !!*(const uint64_t *)&left[i]; - break; - case TX_16X4: - for (i = 0; i < num_4x4_w; i += 8) - t_above[i] = !!*(const uint64_t *)&above[i]; - for (i = 0; i < num_4x4_h; i += 2) - t_left[i] = !!*(const uint16_t *)&left[i]; - break; - case TX_8X32: - for (i = 0; i < num_4x4_w; i += 4) - t_above[i] = !!*(const uint32_t *)&above[i]; - for (i = 0; i < num_4x4_h; i += 16) - t_left[i] = - !!(*(const uint64_t *)&left[i] | *(const uint64_t *)&left[i + 8]); - break; - case TX_32X8: - for (i = 0; i < num_4x4_w; i += 16) - t_above[i] = - !!(*(const uint64_t *)&above[i] | *(const uint64_t *)&above[i + 8]); - for (i = 0; i < num_4x4_h; i += 4) - t_left[i] = !!*(const uint32_t *)&left[i]; - break; -#endif - - default: assert(0 && "Invalid transform size."); break; - } - return; -#endif // CONFIG_CHROMA_2X2 - - switch (tx_size) { - case TX_4X4: - memcpy(t_above, above, sizeof(ENTROPY_CONTEXT) * num_4x4_w); - memcpy(t_left, left, sizeof(ENTROPY_CONTEXT) * num_4x4_h); - break; - case TX_8X8: - for (i = 0; i < num_4x4_w; i += 2) - t_above[i] = !!*(const uint16_t *)&above[i]; - for (i = 0; i < num_4x4_h; i += 2) - t_left[i] = !!*(const uint16_t *)&left[i]; - break; - case TX_16X16: - for (i = 0; i < num_4x4_w; i += 4) - t_above[i] = !!*(const uint32_t *)&above[i]; - for (i = 0; i < num_4x4_h; i += 4) - t_left[i] = !!*(const uint32_t *)&left[i]; - break; - case TX_32X32: - for (i = 0; i < num_4x4_w; i += 8) - t_above[i] = !!*(const uint64_t *)&above[i]; - for (i = 0; i < num_4x4_h; i += 8) - t_left[i] = !!*(const uint64_t *)&left[i]; - break; -#if CONFIG_TX64X64 - case TX_32X64: - for (i = 0; i < num_4x4_w; i += 8) - t_above[i] = !!*(const uint64_t *)&above[i]; - for (i = 0; i < num_4x4_h; i += 16) - t_left[i] = - !!(*(const uint64_t *)&left[i] | *(const uint64_t *)&left[i + 8]); - break; - case TX_64X32: - for (i = 0; i < num_4x4_w; i += 16) - t_above[i] = - !!(*(const uint64_t *)&above[i] | *(const uint64_t *)&above[i + 8]); - for (i = 0; i < num_4x4_h; i += 8) - t_left[i] = !!*(const uint64_t *)&left[i]; - break; - case TX_64X64: - for (i = 0; i < num_4x4_w; i += 16) - t_above[i] = - !!(*(const uint64_t *)&above[i] | *(const uint64_t *)&above[i + 8]); - for (i = 0; i < num_4x4_h; i += 16) - t_left[i] = - !!(*(const uint64_t *)&left[i] | *(const uint64_t *)&left[i + 8]); - break; -#endif // CONFIG_TX64X64 - case TX_4X8: - memcpy(t_above, above, sizeof(ENTROPY_CONTEXT) * num_4x4_w); - for (i = 0; i < num_4x4_h; i += 2) - t_left[i] = !!*(const uint16_t *)&left[i]; - break; - case TX_8X4: - for (i = 0; i < num_4x4_w; i += 2) - t_above[i] = !!*(const uint16_t *)&above[i]; - memcpy(t_left, left, sizeof(ENTROPY_CONTEXT) * num_4x4_h); - break; - case TX_8X16: - for (i = 0; i < num_4x4_w; i += 2) - t_above[i] = !!*(const uint16_t *)&above[i]; - for (i = 0; i < num_4x4_h; i += 4) - t_left[i] = !!*(const uint32_t *)&left[i]; - break; - case TX_16X8: - for (i = 0; i < num_4x4_w; i += 4) - t_above[i] = !!*(const uint32_t *)&above[i]; - for (i = 0; i < num_4x4_h; i += 2) - t_left[i] = !!*(const uint16_t *)&left[i]; - break; - case TX_16X32: - for (i = 0; i < num_4x4_w; i += 4) - t_above[i] = !!*(const uint32_t *)&above[i]; - for (i = 0; i < num_4x4_h; i += 8) - t_left[i] = !!*(const uint64_t *)&left[i]; - break; - case TX_32X16: - for (i = 0; i < num_4x4_w; i += 8) - t_above[i] = !!*(const uint64_t *)&above[i]; - for (i = 0; i < num_4x4_h; i += 4) - t_left[i] = !!*(const uint32_t *)&left[i]; - break; -#if CONFIG_RECT_TX_EXT && (CONFIG_EXT_TX || CONFIG_VAR_TX) - case TX_4X16: - memcpy(t_above, above, sizeof(ENTROPY_CONTEXT) * num_4x4_w); - for (i = 0; i < num_4x4_h; i += 4) - t_left[i] = !!*(const uint32_t *)&left[i]; - break; - case TX_16X4: - for (i = 0; i < num_4x4_w; i += 4) - t_above[i] = !!*(const uint32_t *)&above[i]; - memcpy(t_left, left, sizeof(ENTROPY_CONTEXT) * num_4x4_h); - break; - case TX_8X32: - for (i = 0; i < num_4x4_w; i += 2) - t_above[i] = !!*(const uint16_t *)&above[i]; - for (i = 0; i < num_4x4_h; i += 8) - t_left[i] = !!*(const uint64_t *)&left[i]; - break; - case TX_32X8: - for (i = 0; i < num_4x4_w; i += 8) - t_above[i] = !!*(const uint64_t *)&above[i]; - for (i = 0; i < num_4x4_h; i += 2) - t_left[i] = !!*(const uint16_t *)&left[i]; - break; -#endif - default: assert(0 && "Invalid transform size."); break; - } } -void av1_get_entropy_contexts(BLOCK_SIZE bsize, TX_SIZE tx_size, +void av1_get_entropy_contexts(BLOCK_SIZE bsize, const struct macroblockd_plane *pd, - ENTROPY_CONTEXT t_above[2 * MAX_MIB_SIZE], - ENTROPY_CONTEXT t_left[2 * MAX_MIB_SIZE]) { -#if CONFIG_CHROMA_SUB8X8 + ENTROPY_CONTEXT t_above[MAX_MIB_SIZE], + ENTROPY_CONTEXT t_left[MAX_MIB_SIZE]) { const BLOCK_SIZE plane_bsize = - AOMMAX(BLOCK_4X4, get_plane_block_size(bsize, pd)); -#else - const BLOCK_SIZE plane_bsize = get_plane_block_size(bsize, pd); -#endif - get_entropy_contexts_plane(plane_bsize, tx_size, pd, t_above, t_left); + get_plane_block_size(bsize, pd->subsampling_x, pd->subsampling_y); + get_entropy_contexts_plane(plane_bsize, pd, t_above, t_left); } void av1_mv_pred(const AV1_COMP *cpi, MACROBLOCK *x, uint8_t *ref_y_buffer, int ref_y_stride, int ref_frame, BLOCK_SIZE block_size) { int i; int zero_seen = 0; - int best_index = 0; int best_sad = INT_MAX; int this_sad = INT_MAX; int max_mv = 0; @@ -1129,11 +683,15 @@ void av1_mv_pred(const AV1_COMP *cpi, MACROBLOCK *x, uint8_t *ref_y_buffer, uint8_t *ref_y_ptr; MV pred_mv[MAX_MV_REF_CANDIDATES + 1]; int num_mv_refs = 0; - - pred_mv[num_mv_refs++] = x->mbmi_ext->ref_mvs[ref_frame][0].as_mv; - if (x->mbmi_ext->ref_mvs[ref_frame][0].as_int != - x->mbmi_ext->ref_mvs[ref_frame][1].as_int) { - pred_mv[num_mv_refs++] = x->mbmi_ext->ref_mvs[ref_frame][1].as_mv; + const MV_REFERENCE_FRAME ref_frames[2] = { ref_frame, NONE_FRAME }; + const int_mv ref_mv = + av1_get_ref_mv_from_stack(0, ref_frames, 0, x->mbmi_ext); + const int_mv ref_mv1 = + av1_get_ref_mv_from_stack(0, ref_frames, 1, x->mbmi_ext); + + pred_mv[num_mv_refs++] = ref_mv.as_mv; + if (ref_mv.as_int != ref_mv1.as_int) { + pred_mv[num_mv_refs++] = ref_mv1.as_mv; } if (cpi->sf.adaptive_motion_search && block_size < x->max_partition_size) pred_mv[num_mv_refs++] = x->pred_mv[ref_frame]; @@ -1158,12 +716,10 @@ void av1_mv_pred(const AV1_COMP *cpi, MACROBLOCK *x, uint8_t *ref_y_buffer, // Note if it is the best so far. if (this_sad < best_sad) { best_sad = this_sad; - best_index = i; } } // Note the index of the mv that worked best in the reference list. - x->mv_best_ref_index[ref_frame] = best_index; x->max_mv_context[ref_frame] = max_mv; x->pred_mv_sad[ref_frame] = best_sad; } @@ -1172,7 +728,8 @@ void av1_setup_pred_block(const MACROBLOCKD *xd, struct buf_2d dst[MAX_MB_PLANE], const YV12_BUFFER_CONFIG *src, int mi_row, int mi_col, const struct scale_factors *scale, - const struct scale_factors *scale_uv) { + const struct scale_factors *scale_uv, + const int num_planes) { int i; dst[0].buf = src->y_buffer; @@ -1181,8 +738,8 @@ void av1_setup_pred_block(const MACROBLOCKD *xd, dst[2].buf = src->v_buffer; dst[1].stride = dst[2].stride = src->uv_stride; - for (i = 0; i < MAX_MB_PLANE; ++i) { - setup_pred_plane(dst + i, xd->mi[0]->mbmi.sb_type, dst[i].buf, + for (i = 0; i < num_planes; ++i) { + setup_pred_plane(dst + i, xd->mi[0]->sb_type, dst[i].buf, i ? src->uv_crop_width : src->y_crop_width, i ? src->uv_crop_height : src->y_crop_height, dst[i].stride, mi_row, mi_col, i ? scale_uv : scale, @@ -1192,7 +749,7 @@ void av1_setup_pred_block(const MACROBLOCKD *xd, int av1_raster_block_offset(BLOCK_SIZE plane_bsize, int raster_block, int stride) { - const int bw = b_width_log2_lookup[plane_bsize]; + const int bw = mi_size_wide_log2[plane_bsize]; const int y = 4 * (raster_block >> bw); const int x = 4 * (raster_block & ((1 << bw) - 1)); return y * stride + x; @@ -1214,43 +771,24 @@ YV12_BUFFER_CONFIG *av1_get_scaled_ref_frame(const AV1_COMP *cpi, : NULL; } -#if CONFIG_DUAL_FILTER int av1_get_switchable_rate(const AV1_COMMON *const cm, MACROBLOCK *x, const MACROBLOCKD *xd) { if (cm->interp_filter == SWITCHABLE) { - const MB_MODE_INFO *const mbmi = &xd->mi[0]->mbmi; + const MB_MODE_INFO *const mbmi = xd->mi[0]; int inter_filter_cost = 0; int dir; for (dir = 0; dir < 2; ++dir) { - if (has_subpel_mv_component(xd->mi[0], xd, dir) || - (mbmi->ref_frame[1] > INTRA_FRAME && - has_subpel_mv_component(xd->mi[0], xd, dir + 2))) { - const int ctx = av1_get_pred_context_switchable_interp(xd, dir); - const InterpFilter filter = - av1_extract_interp_filter(mbmi->interp_filters, dir); - inter_filter_cost += x->switchable_interp_costs[ctx][filter]; - } + const int ctx = av1_get_pred_context_switchable_interp(xd, dir); + const InterpFilter filter = + av1_extract_interp_filter(mbmi->interp_filters, dir); + inter_filter_cost += x->switchable_interp_costs[ctx][filter]; } return SWITCHABLE_INTERP_RATE_FACTOR * inter_filter_cost; } else { return 0; } } -#else -int av1_get_switchable_rate(const AV1_COMMON *const cm, MACROBLOCK *x, - const MACROBLOCKD *xd) { - if (cm->interp_filter == SWITCHABLE) { - const MB_MODE_INFO *const mbmi = &xd->mi[0]->mbmi; - const int ctx = av1_get_pred_context_switchable_interp(xd); - const InterpFilter filter = - av1_extract_interp_filter(mbmi->interp_filters, 0); - return SWITCHABLE_INTERP_RATE_FACTOR * - x->switchable_interp_costs[ctx][filter]; - } - return 0; -} -#endif void av1_set_rd_speed_thresholds(AV1_COMP *cpi) { int i; @@ -1262,22 +800,18 @@ void av1_set_rd_speed_thresholds(AV1_COMP *cpi) { if (sf->adaptive_rd_thresh) { rd->thresh_mult[THR_NEARESTMV] = 300; -#if CONFIG_EXT_REFS rd->thresh_mult[THR_NEARESTL2] = 300; rd->thresh_mult[THR_NEARESTL3] = 300; rd->thresh_mult[THR_NEARESTB] = 300; rd->thresh_mult[THR_NEARESTA2] = 300; -#endif // CONFIG_EXT_REFS rd->thresh_mult[THR_NEARESTA] = 300; rd->thresh_mult[THR_NEARESTG] = 300; } else { rd->thresh_mult[THR_NEARESTMV] = 0; -#if CONFIG_EXT_REFS rd->thresh_mult[THR_NEARESTL2] = 0; rd->thresh_mult[THR_NEARESTL3] = 0; rd->thresh_mult[THR_NEARESTB] = 0; rd->thresh_mult[THR_NEARESTA2] = 0; -#endif // CONFIG_EXT_REFS rd->thresh_mult[THR_NEARESTA] = 0; rd->thresh_mult[THR_NEARESTG] = 0; } @@ -1285,92 +819,35 @@ void av1_set_rd_speed_thresholds(AV1_COMP *cpi) { rd->thresh_mult[THR_DC] += 1000; rd->thresh_mult[THR_NEWMV] += 1000; -#if CONFIG_EXT_REFS rd->thresh_mult[THR_NEWL2] += 1000; rd->thresh_mult[THR_NEWL3] += 1000; rd->thresh_mult[THR_NEWB] += 1000; rd->thresh_mult[THR_NEWA2] = 1000; -#endif // CONFIG_EXT_REFS rd->thresh_mult[THR_NEWA] += 1000; rd->thresh_mult[THR_NEWG] += 1000; rd->thresh_mult[THR_NEARMV] += 1000; -#if CONFIG_EXT_REFS rd->thresh_mult[THR_NEARL2] += 1000; rd->thresh_mult[THR_NEARL3] += 1000; rd->thresh_mult[THR_NEARB] += 1000; rd->thresh_mult[THR_NEARA2] = 1000; -#endif // CONFIG_EXT_REFS rd->thresh_mult[THR_NEARA] += 1000; rd->thresh_mult[THR_NEARG] += 1000; - rd->thresh_mult[THR_ZEROMV] += 2000; -#if CONFIG_EXT_REFS - rd->thresh_mult[THR_ZEROL2] += 2000; - rd->thresh_mult[THR_ZEROL3] += 2000; - rd->thresh_mult[THR_ZEROB] += 2000; - rd->thresh_mult[THR_ZEROA2] = 2000; -#endif // CONFIG_EXT_REFS - rd->thresh_mult[THR_ZEROG] += 2000; - rd->thresh_mult[THR_ZEROA] += 2000; - - rd->thresh_mult[THR_TM] += 1000; - -#if CONFIG_COMPOUND_SINGLEREF - rd->thresh_mult[THR_SR_NEAREST_NEARMV] += 1200; -#if CONFIG_EXT_REFS - rd->thresh_mult[THR_SR_NEAREST_NEARL2] += 1200; - rd->thresh_mult[THR_SR_NEAREST_NEARL3] += 1200; - rd->thresh_mult[THR_SR_NEAREST_NEARB] += 1200; -#endif // CONFIG_EXT_REFS - rd->thresh_mult[THR_SR_NEAREST_NEARA] += 1200; - rd->thresh_mult[THR_SR_NEAREST_NEARG] += 1200; - - /* - rd->thresh_mult[THR_SR_NEAREST_NEWMV] += 1200; -#if CONFIG_EXT_REFS - rd->thresh_mult[THR_SR_NEAREST_NEWL2] += 1200; - rd->thresh_mult[THR_SR_NEAREST_NEWL3] += 1200; - rd->thresh_mult[THR_SR_NEAREST_NEWB] += 1200; -#endif // CONFIG_EXT_REFS - rd->thresh_mult[THR_SR_NEAREST_NEWA] += 1200; - rd->thresh_mult[THR_SR_NEAREST_NEWG] += 1200;*/ - - rd->thresh_mult[THR_SR_NEAR_NEWMV] += 1500; -#if CONFIG_EXT_REFS - rd->thresh_mult[THR_SR_NEAR_NEWL2] += 1500; - rd->thresh_mult[THR_SR_NEAR_NEWL3] += 1500; - rd->thresh_mult[THR_SR_NEAR_NEWB] += 1500; -#endif // CONFIG_EXT_REFS - rd->thresh_mult[THR_SR_NEAR_NEWA] += 1500; - rd->thresh_mult[THR_SR_NEAR_NEWG] += 1500; - - rd->thresh_mult[THR_SR_ZERO_NEWMV] += 2000; -#if CONFIG_EXT_REFS - rd->thresh_mult[THR_SR_ZERO_NEWL2] += 2000; - rd->thresh_mult[THR_SR_ZERO_NEWL3] += 2000; - rd->thresh_mult[THR_SR_ZERO_NEWB] += 2000; -#endif // CONFIG_EXT_REFS - rd->thresh_mult[THR_SR_ZERO_NEWA] += 2000; - rd->thresh_mult[THR_SR_ZERO_NEWG] += 2000; - - rd->thresh_mult[THR_SR_NEW_NEWMV] += 1700; -#if CONFIG_EXT_REFS - rd->thresh_mult[THR_SR_NEW_NEWL2] += 1700; - rd->thresh_mult[THR_SR_NEW_NEWL3] += 1700; - rd->thresh_mult[THR_SR_NEW_NEWB] += 1700; -#endif // CONFIG_EXT_REFS - rd->thresh_mult[THR_SR_NEW_NEWA] += 1700; - rd->thresh_mult[THR_SR_NEW_NEWG] += 1700; -#endif // CONFIG_COMPOUND_SINGLEREF + rd->thresh_mult[THR_GLOBALMV] += 2000; + rd->thresh_mult[THR_GLOBALL2] += 2000; + rd->thresh_mult[THR_GLOBALL3] += 2000; + rd->thresh_mult[THR_GLOBALB] += 2000; + rd->thresh_mult[THR_GLOBALA2] = 2000; + rd->thresh_mult[THR_GLOBALG] += 2000; + rd->thresh_mult[THR_GLOBALA] += 2000; + + rd->thresh_mult[THR_PAETH] += 1000; rd->thresh_mult[THR_COMP_NEAREST_NEARESTLA] += 1000; -#if CONFIG_EXT_REFS rd->thresh_mult[THR_COMP_NEAREST_NEARESTL2A] += 1000; rd->thresh_mult[THR_COMP_NEAREST_NEARESTL3A] += 1000; -#endif // CONFIG_EXT_REFS rd->thresh_mult[THR_COMP_NEAREST_NEARESTGA] += 1000; -#if CONFIG_EXT_REFS rd->thresh_mult[THR_COMP_NEAREST_NEARESTLB] += 1000; rd->thresh_mult[THR_COMP_NEAREST_NEARESTL2B] += 1000; rd->thresh_mult[THR_COMP_NEAREST_NEARESTL3B] += 1000; @@ -1380,13 +857,10 @@ void av1_set_rd_speed_thresholds(AV1_COMP *cpi) { rd->thresh_mult[THR_COMP_NEAREST_NEARESTL3A2] += 1000; rd->thresh_mult[THR_COMP_NEAREST_NEARESTGA2] += 1000; -#if CONFIG_EXT_COMP_REFS - rd->thresh_mult[THR_COMP_NEAREST_NEARESTLL2] += 1000; - rd->thresh_mult[THR_COMP_NEAREST_NEARESTLL3] += 1000; - rd->thresh_mult[THR_COMP_NEAREST_NEARESTLG] += 1000; - rd->thresh_mult[THR_COMP_NEAREST_NEARESTBA] += 1000; -#endif // CONFIG_EXT_COMP_REFS -#endif // CONFIG_EXT_REFS + rd->thresh_mult[THR_COMP_NEAREST_NEARESTLL2] += 2000; + rd->thresh_mult[THR_COMP_NEAREST_NEARESTLL3] += 2000; + rd->thresh_mult[THR_COMP_NEAREST_NEARESTLG] += 2000; + rd->thresh_mult[THR_COMP_NEAREST_NEARESTBA] += 2000; rd->thresh_mult[THR_COMP_NEAR_NEARLA] += 1200; rd->thresh_mult[THR_COMP_NEAREST_NEWLA] += 1500; @@ -1394,16 +868,15 @@ void av1_set_rd_speed_thresholds(AV1_COMP *cpi) { rd->thresh_mult[THR_COMP_NEAR_NEWLA] += 1700; rd->thresh_mult[THR_COMP_NEW_NEARLA] += 1700; rd->thresh_mult[THR_COMP_NEW_NEWLA] += 2000; - rd->thresh_mult[THR_COMP_ZERO_ZEROLA] += 2500; + rd->thresh_mult[THR_COMP_GLOBAL_GLOBALLA] += 2500; -#if CONFIG_EXT_REFS rd->thresh_mult[THR_COMP_NEAR_NEARL2A] += 1200; rd->thresh_mult[THR_COMP_NEAREST_NEWL2A] += 1500; rd->thresh_mult[THR_COMP_NEW_NEARESTL2A] += 1500; rd->thresh_mult[THR_COMP_NEAR_NEWL2A] += 1700; rd->thresh_mult[THR_COMP_NEW_NEARL2A] += 1700; rd->thresh_mult[THR_COMP_NEW_NEWL2A] += 2000; - rd->thresh_mult[THR_COMP_ZERO_ZEROL2A] += 2500; + rd->thresh_mult[THR_COMP_GLOBAL_GLOBALL2A] += 2500; rd->thresh_mult[THR_COMP_NEAR_NEARL3A] += 1200; rd->thresh_mult[THR_COMP_NEAREST_NEWL3A] += 1500; @@ -1411,8 +884,7 @@ void av1_set_rd_speed_thresholds(AV1_COMP *cpi) { rd->thresh_mult[THR_COMP_NEAR_NEWL3A] += 1700; rd->thresh_mult[THR_COMP_NEW_NEARL3A] += 1700; rd->thresh_mult[THR_COMP_NEW_NEWL3A] += 2000; - rd->thresh_mult[THR_COMP_ZERO_ZEROL3A] += 2500; -#endif // CONFIG_EXT_REFS + rd->thresh_mult[THR_COMP_GLOBAL_GLOBALL3A] += 2500; rd->thresh_mult[THR_COMP_NEAR_NEARGA] += 1200; rd->thresh_mult[THR_COMP_NEAREST_NEWGA] += 1500; @@ -1420,16 +892,15 @@ void av1_set_rd_speed_thresholds(AV1_COMP *cpi) { rd->thresh_mult[THR_COMP_NEAR_NEWGA] += 1700; rd->thresh_mult[THR_COMP_NEW_NEARGA] += 1700; rd->thresh_mult[THR_COMP_NEW_NEWGA] += 2000; - rd->thresh_mult[THR_COMP_ZERO_ZEROGA] += 2500; + rd->thresh_mult[THR_COMP_GLOBAL_GLOBALGA] += 2500; -#if CONFIG_EXT_REFS rd->thresh_mult[THR_COMP_NEAR_NEARLB] += 1200; rd->thresh_mult[THR_COMP_NEAREST_NEWLB] += 1500; rd->thresh_mult[THR_COMP_NEW_NEARESTLB] += 1500; rd->thresh_mult[THR_COMP_NEAR_NEWLB] += 1700; rd->thresh_mult[THR_COMP_NEW_NEARLB] += 1700; rd->thresh_mult[THR_COMP_NEW_NEWLB] += 2000; - rd->thresh_mult[THR_COMP_ZERO_ZEROLB] += 2500; + rd->thresh_mult[THR_COMP_GLOBAL_GLOBALLB] += 2500; rd->thresh_mult[THR_COMP_NEAR_NEARL2B] += 1200; rd->thresh_mult[THR_COMP_NEAREST_NEWL2B] += 1500; @@ -1437,7 +908,7 @@ void av1_set_rd_speed_thresholds(AV1_COMP *cpi) { rd->thresh_mult[THR_COMP_NEAR_NEWL2B] += 1700; rd->thresh_mult[THR_COMP_NEW_NEARL2B] += 1700; rd->thresh_mult[THR_COMP_NEW_NEWL2B] += 2000; - rd->thresh_mult[THR_COMP_ZERO_ZEROL2B] += 2500; + rd->thresh_mult[THR_COMP_GLOBAL_GLOBALL2B] += 2500; rd->thresh_mult[THR_COMP_NEAR_NEARL3B] += 1200; rd->thresh_mult[THR_COMP_NEAREST_NEWL3B] += 1500; @@ -1445,7 +916,7 @@ void av1_set_rd_speed_thresholds(AV1_COMP *cpi) { rd->thresh_mult[THR_COMP_NEAR_NEWL3B] += 1700; rd->thresh_mult[THR_COMP_NEW_NEARL3B] += 1700; rd->thresh_mult[THR_COMP_NEW_NEWL3B] += 2000; - rd->thresh_mult[THR_COMP_ZERO_ZEROL3B] += 2500; + rd->thresh_mult[THR_COMP_GLOBAL_GLOBALL3B] += 2500; rd->thresh_mult[THR_COMP_NEAR_NEARGB] += 1200; rd->thresh_mult[THR_COMP_NEAREST_NEWGB] += 1500; @@ -1453,7 +924,7 @@ void av1_set_rd_speed_thresholds(AV1_COMP *cpi) { rd->thresh_mult[THR_COMP_NEAR_NEWGB] += 1700; rd->thresh_mult[THR_COMP_NEW_NEARGB] += 1700; rd->thresh_mult[THR_COMP_NEW_NEWGB] += 2000; - rd->thresh_mult[THR_COMP_ZERO_ZEROGB] += 2500; + rd->thresh_mult[THR_COMP_GLOBAL_GLOBALGB] += 2500; rd->thresh_mult[THR_COMP_NEAR_NEARLA2] += 1200; rd->thresh_mult[THR_COMP_NEAREST_NEWLA2] += 1500; @@ -1461,7 +932,7 @@ void av1_set_rd_speed_thresholds(AV1_COMP *cpi) { rd->thresh_mult[THR_COMP_NEAR_NEWLA2] += 1700; rd->thresh_mult[THR_COMP_NEW_NEARLA2] += 1700; rd->thresh_mult[THR_COMP_NEW_NEWLA2] += 2000; - rd->thresh_mult[THR_COMP_ZERO_ZEROLA2] += 2500; + rd->thresh_mult[THR_COMP_GLOBAL_GLOBALLA2] += 2500; rd->thresh_mult[THR_COMP_NEAR_NEARL2A2] += 1200; rd->thresh_mult[THR_COMP_NEAREST_NEWL2A2] += 1500; @@ -1469,7 +940,7 @@ void av1_set_rd_speed_thresholds(AV1_COMP *cpi) { rd->thresh_mult[THR_COMP_NEAR_NEWL2A2] += 1700; rd->thresh_mult[THR_COMP_NEW_NEARL2A2] += 1700; rd->thresh_mult[THR_COMP_NEW_NEWL2A2] += 2000; - rd->thresh_mult[THR_COMP_ZERO_ZEROL2A2] += 2500; + rd->thresh_mult[THR_COMP_GLOBAL_GLOBALL2A2] += 2500; rd->thresh_mult[THR_COMP_NEAR_NEARL3A2] += 1200; rd->thresh_mult[THR_COMP_NEAREST_NEWL3A2] += 1500; @@ -1477,7 +948,7 @@ void av1_set_rd_speed_thresholds(AV1_COMP *cpi) { rd->thresh_mult[THR_COMP_NEAR_NEWL3A2] += 1700; rd->thresh_mult[THR_COMP_NEW_NEARL3A2] += 1700; rd->thresh_mult[THR_COMP_NEW_NEWL3A2] += 2000; - rd->thresh_mult[THR_COMP_ZERO_ZEROL3A2] += 2500; + rd->thresh_mult[THR_COMP_GLOBAL_GLOBALL3A2] += 2500; rd->thresh_mult[THR_COMP_NEAR_NEARGA2] += 1200; rd->thresh_mult[THR_COMP_NEAREST_NEWGA2] += 1500; @@ -1485,124 +956,55 @@ void av1_set_rd_speed_thresholds(AV1_COMP *cpi) { rd->thresh_mult[THR_COMP_NEAR_NEWGA2] += 1700; rd->thresh_mult[THR_COMP_NEW_NEARGA2] += 1700; rd->thresh_mult[THR_COMP_NEW_NEWGA2] += 2000; - rd->thresh_mult[THR_COMP_ZERO_ZEROGA2] += 2500; - -#if CONFIG_EXT_COMP_REFS - rd->thresh_mult[THR_COMP_NEAR_NEARLL2] += 1200; - rd->thresh_mult[THR_COMP_NEAREST_NEWLL2] += 1500; - rd->thresh_mult[THR_COMP_NEW_NEARESTLL2] += 1500; - rd->thresh_mult[THR_COMP_NEAR_NEWLL2] += 1700; - rd->thresh_mult[THR_COMP_NEW_NEARLL2] += 1700; - rd->thresh_mult[THR_COMP_NEW_NEWLL2] += 2000; - rd->thresh_mult[THR_COMP_ZERO_ZEROLL2] += 2500; - - rd->thresh_mult[THR_COMP_NEAR_NEARLL3] += 1200; - rd->thresh_mult[THR_COMP_NEAREST_NEWLL3] += 1500; - rd->thresh_mult[THR_COMP_NEW_NEARESTLL3] += 1500; - rd->thresh_mult[THR_COMP_NEAR_NEWLL3] += 1700; - rd->thresh_mult[THR_COMP_NEW_NEARLL3] += 1700; - rd->thresh_mult[THR_COMP_NEW_NEWLL3] += 2000; - rd->thresh_mult[THR_COMP_ZERO_ZEROLL3] += 2500; - - rd->thresh_mult[THR_COMP_NEAR_NEARLG] += 1200; - rd->thresh_mult[THR_COMP_NEAREST_NEWLG] += 1500; - rd->thresh_mult[THR_COMP_NEW_NEARESTLG] += 1500; - rd->thresh_mult[THR_COMP_NEAR_NEWLG] += 1700; - rd->thresh_mult[THR_COMP_NEW_NEARLG] += 1700; - rd->thresh_mult[THR_COMP_NEW_NEWLG] += 2000; - rd->thresh_mult[THR_COMP_ZERO_ZEROLG] += 2500; - - rd->thresh_mult[THR_COMP_NEAR_NEARBA] += 1200; - rd->thresh_mult[THR_COMP_NEAREST_NEWBA] += 1500; - rd->thresh_mult[THR_COMP_NEW_NEARESTBA] += 1500; - rd->thresh_mult[THR_COMP_NEAR_NEWBA] += 1700; - rd->thresh_mult[THR_COMP_NEW_NEARBA] += 1700; - rd->thresh_mult[THR_COMP_NEW_NEWBA] += 2000; - rd->thresh_mult[THR_COMP_ZERO_ZEROBA] += 2500; -#endif // CONFIG_EXT_COMP_REFS -#endif // CONFIG_EXT_REFS + rd->thresh_mult[THR_COMP_GLOBAL_GLOBALGA2] += 2500; rd->thresh_mult[THR_H_PRED] += 2000; rd->thresh_mult[THR_V_PRED] += 2000; rd->thresh_mult[THR_D135_PRED] += 2500; - rd->thresh_mult[THR_D207_PRED] += 2500; - rd->thresh_mult[THR_D153_PRED] += 2500; - rd->thresh_mult[THR_D63_PRED] += 2500; - rd->thresh_mult[THR_D117_PRED] += 2500; + rd->thresh_mult[THR_D203_PRED] += 2500; + rd->thresh_mult[THR_D157_PRED] += 2500; + rd->thresh_mult[THR_D67_PRED] += 2500; + rd->thresh_mult[THR_D113_PRED] += 2500; rd->thresh_mult[THR_D45_PRED] += 2500; - rd->thresh_mult[THR_COMP_INTERINTRA_ZEROL] += 1500; - rd->thresh_mult[THR_COMP_INTERINTRA_NEARESTL] += 1500; - rd->thresh_mult[THR_COMP_INTERINTRA_NEARL] += 1500; - rd->thresh_mult[THR_COMP_INTERINTRA_NEWL] += 2000; - -#if CONFIG_EXT_REFS - rd->thresh_mult[THR_COMP_INTERINTRA_ZEROL2] += 1500; - rd->thresh_mult[THR_COMP_INTERINTRA_NEARESTL2] += 1500; - rd->thresh_mult[THR_COMP_INTERINTRA_NEARL2] += 1500; - rd->thresh_mult[THR_COMP_INTERINTRA_NEWL2] += 2000; - - rd->thresh_mult[THR_COMP_INTERINTRA_ZEROL3] += 1500; - rd->thresh_mult[THR_COMP_INTERINTRA_NEARESTL3] += 1500; - rd->thresh_mult[THR_COMP_INTERINTRA_NEARL3] += 1500; - rd->thresh_mult[THR_COMP_INTERINTRA_NEWL3] += 2000; -#endif // CONFIG_EXT_REFS - - rd->thresh_mult[THR_COMP_INTERINTRA_ZEROG] += 1500; - rd->thresh_mult[THR_COMP_INTERINTRA_NEARESTG] += 1500; - rd->thresh_mult[THR_COMP_INTERINTRA_NEARG] += 1500; - rd->thresh_mult[THR_COMP_INTERINTRA_NEWG] += 2000; - -#if CONFIG_EXT_REFS - rd->thresh_mult[THR_COMP_INTERINTRA_ZEROB] += 1500; - rd->thresh_mult[THR_COMP_INTERINTRA_NEARESTB] += 1500; - rd->thresh_mult[THR_COMP_INTERINTRA_NEARB] += 1500; - rd->thresh_mult[THR_COMP_INTERINTRA_NEWB] += 2000; - - rd->thresh_mult[THR_COMP_INTERINTRA_ZEROA2] += 1500; - rd->thresh_mult[THR_COMP_INTERINTRA_NEARESTA2] += 1500; - rd->thresh_mult[THR_COMP_INTERINTRA_NEARA2] += 1500; - rd->thresh_mult[THR_COMP_INTERINTRA_NEWA2] += 2000; -#endif // CONFIG_EXT_REFS - - rd->thresh_mult[THR_COMP_INTERINTRA_ZEROA] += 1500; - rd->thresh_mult[THR_COMP_INTERINTRA_NEARESTA] += 1500; - rd->thresh_mult[THR_COMP_INTERINTRA_NEARA] += 1500; - rd->thresh_mult[THR_COMP_INTERINTRA_NEWA] += 2000; + rd->thresh_mult[THR_COMP_NEAR_NEARLL2] += 1600; + rd->thresh_mult[THR_COMP_NEAREST_NEWLL2] += 2000; + rd->thresh_mult[THR_COMP_NEW_NEARESTLL2] += 2000; + rd->thresh_mult[THR_COMP_NEAR_NEWLL2] += 2200; + rd->thresh_mult[THR_COMP_NEW_NEARLL2] += 2200; + rd->thresh_mult[THR_COMP_NEW_NEWLL2] += 2400; + rd->thresh_mult[THR_COMP_GLOBAL_GLOBALLL2] += 3200; + + rd->thresh_mult[THR_COMP_NEAR_NEARLL3] += 1600; + rd->thresh_mult[THR_COMP_NEAREST_NEWLL3] += 2000; + rd->thresh_mult[THR_COMP_NEW_NEARESTLL3] += 2000; + rd->thresh_mult[THR_COMP_NEAR_NEWLL3] += 2200; + rd->thresh_mult[THR_COMP_NEW_NEARLL3] += 2200; + rd->thresh_mult[THR_COMP_NEW_NEWLL3] += 2400; + rd->thresh_mult[THR_COMP_GLOBAL_GLOBALLL3] += 3200; + + rd->thresh_mult[THR_COMP_NEAR_NEARLG] += 1600; + rd->thresh_mult[THR_COMP_NEAREST_NEWLG] += 2000; + rd->thresh_mult[THR_COMP_NEW_NEARESTLG] += 2000; + rd->thresh_mult[THR_COMP_NEAR_NEWLG] += 2200; + rd->thresh_mult[THR_COMP_NEW_NEARLG] += 2200; + rd->thresh_mult[THR_COMP_NEW_NEWLG] += 2400; + rd->thresh_mult[THR_COMP_GLOBAL_GLOBALLG] += 3200; + + rd->thresh_mult[THR_COMP_NEAR_NEARBA] += 1600; + rd->thresh_mult[THR_COMP_NEAREST_NEWBA] += 2000; + rd->thresh_mult[THR_COMP_NEW_NEARESTBA] += 2000; + rd->thresh_mult[THR_COMP_NEAR_NEWBA] += 2200; + rd->thresh_mult[THR_COMP_NEW_NEARBA] += 2200; + rd->thresh_mult[THR_COMP_NEW_NEWBA] += 2400; + rd->thresh_mult[THR_COMP_GLOBAL_GLOBALBA] += 3200; } void av1_set_rd_speed_thresholds_sub8x8(AV1_COMP *cpi) { - static const int thresh_mult[MAX_REFS] = { -#if CONFIG_EXT_REFS - 2500, - 2500, - 2500, - 2500, - 2500, - 2500, - 2500, - 4500, - 4500, - 4500, - 4500, - 4500, - 4500, - 4500, - 4500, - 4500, - 4500, - 4500, - 4500, - 2500 -#else // !CONFIG_EXT_REFS - 2500, - 2500, - 2500, - 4500, - 4500, - 2500 -#endif // CONFIG_EXT_REFS - }; + static const int thresh_mult[MAX_REFS] = { 2500, 2500, 2500, 2500, 2500, + 2500, 2500, 4500, 4500, 4500, + 4500, 4500, 4500, 4500, 4500, + 4500, 4500, 4500, 4500, 2500 }; RD_OPT *const rd = &cpi->rd; memcpy(rd->thresh_mult_sub8x8, thresh_mult, sizeof(thresh_mult)); } @@ -1611,15 +1013,12 @@ void av1_update_rd_thresh_fact(const AV1_COMMON *const cm, int (*factor_buf)[MAX_MODES], int rd_thresh, int bsize, int best_mode_index) { if (rd_thresh > 0) { -#if CONFIG_CB4X4 const int top_mode = MAX_MODES; -#else - const int top_mode = bsize < BLOCK_8X8 ? MAX_REFS : MAX_MODES; -#endif int mode; for (mode = 0; mode < top_mode; ++mode) { const BLOCK_SIZE min_size = AOMMAX(bsize - 1, BLOCK_4X4); - const BLOCK_SIZE max_size = AOMMIN(bsize + 2, (int)cm->sb_size); + const BLOCK_SIZE max_size = + AOMMIN(bsize + 2, (int)cm->seq_params.sb_size); BLOCK_SIZE bs; for (bs = min_size; bs <= max_size; ++bs) { int *const fact = &factor_buf[bs][mode]; @@ -1635,8 +1034,7 @@ void av1_update_rd_thresh_fact(const AV1_COMMON *const cm, int av1_get_intra_cost_penalty(int qindex, int qdelta, aom_bit_depth_t bit_depth) { - const int q = av1_dc_quant(qindex, qdelta, bit_depth); -#if CONFIG_HIGHBITDEPTH + const int q = av1_dc_quant_Q3(qindex, qdelta, bit_depth); switch (bit_depth) { case AOM_BITS_8: return 20 * q; case AOM_BITS_10: return 5 * q; @@ -1645,7 +1043,4 @@ int av1_get_intra_cost_penalty(int qindex, int qdelta, assert(0 && "bit_depth should be AOM_BITS_8, AOM_BITS_10 or AOM_BITS_12"); return -1; } -#else - return 20 * q; -#endif // CONFIG_HIGHBITDEPTH } |