Make some functions static to reduce lib size.
libaom.so size was reduced:
Before: 7125616 bytes
After: 7117360 bytes
Change-Id: Ia797f75544f8f021cafd7159bd12127d73325c38
diff --git a/av1/encoder/arm/neon/error_neon.c b/av1/encoder/arm/neon/error_neon.c
deleted file mode 100644
index fe5233f..0000000
--- a/av1/encoder/arm/neon/error_neon.c
+++ /dev/null
@@ -1,42 +0,0 @@
-/*
- * Copyright (c) 2016, Alliance for Open Media. All rights reserved
- *
- * This source code is subject to the terms of the BSD 2 Clause License and
- * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
- * was not distributed with this source code in the LICENSE file, you can
- * obtain it at www.aomedia.org/license/software. If the Alliance for Open
- * Media Patent License 1.0 was not distributed with this source code in the
- * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
- */
-
-#include <arm_neon.h>
-#include <assert.h>
-
-#include "./av1_rtcd.h"
-
-int64_t av1_block_error_fp_neon(const int16_t *coeff, const int16_t *dqcoeff,
- int block_size) {
- int64x2_t error = vdupq_n_s64(0);
-
- assert(block_size >= 8);
- assert((block_size % 8) == 0);
-
- do {
- const int16x8_t c = vld1q_s16(coeff);
- const int16x8_t d = vld1q_s16(dqcoeff);
- const int16x8_t diff = vsubq_s16(c, d);
- const int16x4_t diff_lo = vget_low_s16(diff);
- const int16x4_t diff_hi = vget_high_s16(diff);
- // diff is 15-bits, the squares 30, so we can store 2 in 31-bits before
- // accumulating them in 64-bits.
- const int32x4_t err0 = vmull_s16(diff_lo, diff_lo);
- const int32x4_t err1 = vmlal_s16(err0, diff_hi, diff_hi);
- const int64x2_t err2 = vaddl_s32(vget_low_s32(err1), vget_high_s32(err1));
- error = vaddq_s64(error, err2);
- coeff += 8;
- dqcoeff += 8;
- block_size -= 8;
- } while (block_size != 0);
-
- return vgetq_lane_s64(error, 0) + vgetq_lane_s64(error, 1);
-}
diff --git a/av1/encoder/encodeframe.c b/av1/encoder/encodeframe.c
index 316c066..6f9aabd 100644
--- a/av1/encoder/encodeframe.c
+++ b/av1/encoder/encodeframe.c
@@ -1999,6 +1999,68 @@
}
}
+// Checks to see if a super block is on a horizontal image edge.
+// In most cases this is the "real" edge unless there are formatting
+// bars embedded in the stream.
+static int active_h_edge(const AV1_COMP *cpi, int mi_row, int mi_step) {
+ int top_edge = 0;
+ int bottom_edge = cpi->common.mi_rows;
+ int is_active_h_edge = 0;
+
+ // For two pass account for any formatting bars detected.
+ if (cpi->oxcf.pass == 2) {
+ const TWO_PASS *const twopass = &cpi->twopass;
+
+ // The inactive region is specified in MBs not mi units.
+ // The image edge is in the following MB row.
+ top_edge += (int)(twopass->this_frame_stats.inactive_zone_rows * 2);
+
+ bottom_edge -= (int)(twopass->this_frame_stats.inactive_zone_rows * 2);
+ bottom_edge = AOMMAX(top_edge, bottom_edge);
+ }
+
+ if (((top_edge >= mi_row) && (top_edge < (mi_row + mi_step))) ||
+ ((bottom_edge >= mi_row) && (bottom_edge < (mi_row + mi_step)))) {
+ is_active_h_edge = 1;
+ }
+ return is_active_h_edge;
+}
+
+// Checks to see if a super block is on a vertical image edge.
+// In most cases this is the "real" edge unless there are formatting
+// bars embedded in the stream.
+static int active_v_edge(const AV1_COMP *cpi, int mi_col, int mi_step) {
+ int left_edge = 0;
+ int right_edge = cpi->common.mi_cols;
+ int is_active_v_edge = 0;
+
+ // For two pass account for any formatting bars detected.
+ if (cpi->oxcf.pass == 2) {
+ const TWO_PASS *const twopass = &cpi->twopass;
+
+ // The inactive region is specified in MBs not mi units.
+ // The image edge is in the following MB row.
+ left_edge += (int)(twopass->this_frame_stats.inactive_zone_cols * 2);
+
+ right_edge -= (int)(twopass->this_frame_stats.inactive_zone_cols * 2);
+ right_edge = AOMMAX(left_edge, right_edge);
+ }
+
+ if (((left_edge >= mi_col) && (left_edge < (mi_col + mi_step))) ||
+ ((right_edge >= mi_col) && (right_edge < (mi_col + mi_step)))) {
+ is_active_v_edge = 1;
+ }
+ return is_active_v_edge;
+}
+
+// Checks to see if a super block is at the edge of the active image.
+// In most cases this is the "real" edge unless there are formatting
+// bars embedded in the stream.
+static int active_edge_sb(const AV1_COMP *cpi, int mi_row, int mi_col) {
+ return active_h_edge(cpi, mi_row, cpi->common.seq_params.mib_size) ||
+ active_v_edge(cpi, mi_col, cpi->common.seq_params.mib_size);
+}
+
// Look at neighboring blocks and set a min and max partition size based on
// what they chose.
static void rd_auto_partition_range(AV1_COMP *cpi, const TileInfo *const tile,
@@ -2056,7 +2118,7 @@
// Test for blocks at the edge of the active image.
// This may be the actual edge of the image or where there are formatting
// bars.
- if (av1_active_edge_sb(cpi, mi_row, mi_col)) {
+ if (active_edge_sb(cpi, mi_row, mi_col)) {
min_size = BLOCK_4X4;
} else {
min_size = AOMMIN(cpi->sf.rd_auto_partition_min_limit, min_size);
@@ -2994,7 +3056,7 @@
// PARTITION_HORZ
if (partition_horz_allowed &&
- (do_rectangular_split || av1_active_h_edge(cpi, mi_row, mi_step))) {
+ (do_rectangular_split || active_h_edge(cpi, mi_row, mi_step))) {
av1_init_rd_stats(&sum_rdc);
subsize = get_partition_subsize(bsize, PARTITION_HORZ);
if (cpi->sf.adaptive_motion_search) load_pred_mv(x, ctx_none);
@@ -3078,7 +3140,7 @@
// PARTITION_VERT
if (partition_vert_allowed &&
- (do_rectangular_split || av1_active_v_edge(cpi, mi_col, mi_step))) {
+ (do_rectangular_split || active_v_edge(cpi, mi_col, mi_step))) {
av1_init_rd_stats(&sum_rdc);
subsize = get_partition_subsize(bsize, PARTITION_VERT);
@@ -3300,7 +3362,7 @@
pc_tree->partitioning == PARTITION_NONE);
}
if (partition_horz4_allowed && has_rows &&
- (do_rectangular_split || av1_active_h_edge(cpi, mi_row, mi_step))) {
+ (do_rectangular_split || active_h_edge(cpi, mi_row, mi_step))) {
av1_init_rd_stats(&sum_rdc);
const int quarter_step = mi_size_high[bsize] / 4;
PICK_MODE_CONTEXT *ctx_prev = ctx_none;
@@ -3344,7 +3406,7 @@
pc_tree->partitioning == PARTITION_NONE);
}
if (partition_vert4_allowed && has_cols &&
- (do_rectangular_split || av1_active_v_edge(cpi, mi_row, mi_step))) {
+ (do_rectangular_split || active_v_edge(cpi, mi_row, mi_step))) {
av1_init_rd_stats(&sum_rdc);
const int quarter_step = mi_size_wide[bsize] / 4;
PICK_MODE_CONTEXT *ctx_prev = ctx_none;
@@ -4658,60 +4720,6 @@
set_txfm_context(xd, max_tx_size, idy, idx);
}
-void av1_update_tx_type_count(const AV1_COMMON *cm, MACROBLOCKD *xd,
- int blk_row, int blk_col, int plane,
- TX_SIZE tx_size, FRAME_COUNTS *counts,
- uint8_t allow_update_cdf) {
- MB_MODE_INFO *mbmi = xd->mi[0];
- int is_inter = is_inter_block(mbmi);
- FRAME_CONTEXT *fc = xd->tile_ctx;
-#if !CONFIG_ENTROPY_STATS
- (void)counts;
-#endif // !CONFIG_ENTROPY_STATS
-
- // Only y plane's tx_type is updated
- if (plane > 0) return;
- TX_TYPE tx_type = av1_get_tx_type(PLANE_TYPE_Y, xd, blk_row, blk_col, tx_size,
- cm->reduced_tx_set_used);
- if (get_ext_tx_types(tx_size, is_inter, cm->reduced_tx_set_used) > 1 &&
- cm->base_qindex > 0 && !mbmi->skip &&
- !segfeature_active(&cm->seg, mbmi->segment_id, SEG_LVL_SKIP)) {
- const int eset = get_ext_tx_set(tx_size, is_inter, cm->reduced_tx_set_used);
- if (eset > 0) {
- const TxSetType tx_set_type =
- av1_get_ext_tx_set_type(tx_size, is_inter, cm->reduced_tx_set_used);
- if (is_inter) {
- if (allow_update_cdf) {
- update_cdf(fc->inter_ext_tx_cdf[eset][txsize_sqr_map[tx_size]],
- av1_ext_tx_ind[tx_set_type][tx_type],
- av1_num_ext_tx_set[tx_set_type]);
- }
-#if CONFIG_ENTROPY_STATS
- ++counts->inter_ext_tx[eset][txsize_sqr_map[tx_size]]
- [av1_ext_tx_ind[tx_set_type][tx_type]];
-#endif // CONFIG_ENTROPY_STATS
- } else {
- PREDICTION_MODE intra_dir;
- if (mbmi->filter_intra_mode_info.use_filter_intra)
- intra_dir = fimode_to_intradir[mbmi->filter_intra_mode_info
- .filter_intra_mode];
- else
- intra_dir = mbmi->mode;
-#if CONFIG_ENTROPY_STATS
- ++counts->intra_ext_tx[eset][txsize_sqr_map[tx_size]][intra_dir]
- [av1_ext_tx_ind[tx_set_type][tx_type]];
-#endif // CONFIG_ENTROPY_STATS
- if (allow_update_cdf) {
- update_cdf(
- fc->intra_ext_tx_cdf[eset][txsize_sqr_map[tx_size]][intra_dir],
- av1_ext_tx_ind[tx_set_type][tx_type],
- av1_num_ext_tx_set[tx_set_type]);
- }
- }
- }
- }
-}
-
static void encode_superblock(const AV1_COMP *const cpi, TileDataEnc *tile_data,
ThreadData *td, TOKENEXTRA **t, RUN_TYPE dry_run,
int mi_row, int mi_col, BLOCK_SIZE bsize,
diff --git a/av1/encoder/encodeframe.h b/av1/encoder/encodeframe.h
index 08ca65a..09589cc 100644
--- a/av1/encoder/encodeframe.h
+++ b/av1/encoder/encodeframe.h
@@ -35,10 +35,6 @@
void av1_encode_tile(struct AV1_COMP *cpi, struct ThreadData *td, int tile_row,
int tile_col);
-void av1_update_tx_type_count(const struct AV1Common *cm, MACROBLOCKD *xd,
- int blk_row, int blk_col, int plane,
- TX_SIZE tx_size, FRAME_COUNTS *counts,
- uint8_t allow_update_cdf);
#ifdef __cplusplus
} // extern "C"
#endif
diff --git a/av1/encoder/encodetxb.c b/av1/encoder/encodetxb.c
index 0655440..124e00e 100644
--- a/av1/encoder/encodetxb.c
+++ b/av1/encoder/encodetxb.c
@@ -650,6 +650,39 @@
}
}
+// TODO(angiebird): use this function whenever it's possible
+static int get_tx_type_cost(const AV1_COMMON *cm, const MACROBLOCK *x,
+ const MACROBLOCKD *xd, int plane, TX_SIZE tx_size,
+ TX_TYPE tx_type) {
+ if (plane > 0) return 0;
+
+ const TX_SIZE square_tx_size = txsize_sqr_map[tx_size];
+
+ const MB_MODE_INFO *mbmi = xd->mi[0];
+ const int is_inter = is_inter_block(mbmi);
+ if (get_ext_tx_types(tx_size, is_inter, cm->reduced_tx_set_used) > 1 &&
+ !xd->lossless[xd->mi[0]->segment_id]) {
+ const int ext_tx_set =
+ get_ext_tx_set(tx_size, is_inter, cm->reduced_tx_set_used);
+ if (is_inter) {
+ if (ext_tx_set > 0)
+ return x->inter_tx_type_costs[ext_tx_set][square_tx_size][tx_type];
+ } else {
+ if (ext_tx_set > 0) {
+ PREDICTION_MODE intra_dir;
+ if (mbmi->filter_intra_mode_info.use_filter_intra)
+ intra_dir = fimode_to_intradir[mbmi->filter_intra_mode_info
+ .filter_intra_mode];
+ else
+ intra_dir = mbmi->mode;
+ return x->intra_tx_type_costs[ext_tx_set][square_tx_size][intra_dir]
+ [tx_type];
+ }
+ }
+ }
+ return 0;
+}
+
static AOM_FORCE_INLINE int warehouse_efficients_txb(
const AV1_COMMON *const cm, const MACROBLOCK *x, const int plane,
const int block, const TX_SIZE tx_size, const TXB_CTX *const txb_ctx,
@@ -674,7 +707,7 @@
av1_txb_init_levels(qcoeff, width, height, levels);
- cost += av1_tx_type_cost(cm, x, xd, plane, tx_size, tx_type);
+ cost += get_tx_type_cost(cm, x, xd, plane, tx_size, tx_type);
cost += get_eob_cost(eob, eob_costs, coeff_costs, tx_class);
@@ -1638,7 +1671,7 @@
levels);
}
- const int tx_type_cost = av1_tx_type_cost(cm, x, xd, plane, tx_size, tx_type);
+ const int tx_type_cost = get_tx_type_cost(cm, x, xd, plane, tx_size, tx_type);
if (eob == 0)
accu_rate += skip_cost;
else
@@ -1695,7 +1728,7 @@
? pd->seg_iqmatrix[mbmi->segment_id][qm_tx_size]
: cm->giqmatrix[NUM_QM_LEVELS - 1][0][qm_tx_size];
assert(width == (1 << bwl));
- const int tx_type_cost = av1_tx_type_cost(cm, x, xd, plane, tx_size, tx_type);
+ const int tx_type_cost = get_tx_type_cost(cm, x, xd, plane, tx_size, tx_type);
TxbInfo txb_info = {
qcoeff, levels, dqcoeff, tcoeff, dequant, shift,
tx_size, txs_ctx, tx_type, bwl, width, height,
@@ -1764,6 +1797,60 @@
blk_row);
}
+static void update_tx_type_count(const AV1_COMMON *cm, MACROBLOCKD *xd,
+ int blk_row, int blk_col, int plane,
+ TX_SIZE tx_size, FRAME_COUNTS *counts,
+ uint8_t allow_update_cdf) {
+ MB_MODE_INFO *mbmi = xd->mi[0];
+ int is_inter = is_inter_block(mbmi);
+ FRAME_CONTEXT *fc = xd->tile_ctx;
+#if !CONFIG_ENTROPY_STATS
+ (void)counts;
+#endif // !CONFIG_ENTROPY_STATS
+
+ // Only y plane's tx_type is updated
+ if (plane > 0) return;
+ TX_TYPE tx_type = av1_get_tx_type(PLANE_TYPE_Y, xd, blk_row, blk_col, tx_size,
+ cm->reduced_tx_set_used);
+ if (get_ext_tx_types(tx_size, is_inter, cm->reduced_tx_set_used) > 1 &&
+ cm->base_qindex > 0 && !mbmi->skip &&
+ !segfeature_active(&cm->seg, mbmi->segment_id, SEG_LVL_SKIP)) {
+ const int eset = get_ext_tx_set(tx_size, is_inter, cm->reduced_tx_set_used);
+ if (eset > 0) {
+ const TxSetType tx_set_type =
+ av1_get_ext_tx_set_type(tx_size, is_inter, cm->reduced_tx_set_used);
+ if (is_inter) {
+ if (allow_update_cdf) {
+ update_cdf(fc->inter_ext_tx_cdf[eset][txsize_sqr_map[tx_size]],
+ av1_ext_tx_ind[tx_set_type][tx_type],
+ av1_num_ext_tx_set[tx_set_type]);
+ }
+#if CONFIG_ENTROPY_STATS
+ ++counts->inter_ext_tx[eset][txsize_sqr_map[tx_size]]
+ [av1_ext_tx_ind[tx_set_type][tx_type]];
+#endif // CONFIG_ENTROPY_STATS
+ } else {
+ PREDICTION_MODE intra_dir;
+ if (mbmi->filter_intra_mode_info.use_filter_intra)
+ intra_dir = fimode_to_intradir[mbmi->filter_intra_mode_info
+ .filter_intra_mode];
+ else
+ intra_dir = mbmi->mode;
+#if CONFIG_ENTROPY_STATS
+ ++counts->intra_ext_tx[eset][txsize_sqr_map[tx_size]][intra_dir]
+ [av1_ext_tx_ind[tx_set_type][tx_type]];
+#endif // CONFIG_ENTROPY_STATS
+ if (allow_update_cdf) {
+ update_cdf(
+ fc->intra_ext_tx_cdf[eset][txsize_sqr_map[tx_size]][intra_dir],
+ av1_ext_tx_ind[tx_set_type][tx_type],
+ av1_num_ext_tx_set[tx_set_type]);
+ }
+ }
+ }
+ }
+}
+
void av1_update_and_record_txb_context(int plane, int block, int blk_row,
int blk_col, BLOCK_SIZE plane_bsize,
TX_SIZE tx_size, void *arg) {
@@ -1815,8 +1902,8 @@
uint8_t levels_buf[TX_PAD_2D];
uint8_t *const levels = set_levels(levels_buf, width);
av1_txb_init_levels(tcoeff, width, height, levels);
- av1_update_tx_type_count(cm, xd, blk_row, blk_col, plane, tx_size, td->counts,
- allow_update_cdf);
+ update_tx_type_count(cm, xd, blk_row, blk_col, plane, tx_size, td->counts,
+ allow_update_cdf);
const PLANE_TYPE plane_type = pd->plane_type;
const TX_TYPE tx_type = av1_get_tx_type(plane_type, xd, blk_row, blk_col,
diff --git a/av1/encoder/rdopt.c b/av1/encoder/rdopt.c
index c3b2a9d..29b471a 100644
--- a/av1/encoder/rdopt.c
+++ b/av1/encoder/rdopt.c
@@ -676,10 +676,9 @@
#define FAST_EXT_TX_CORR_MARGIN 0.5
#define FAST_EXT_TX_EDST_MARGIN 0.3
-int inter_block_yrd(const AV1_COMP *cpi, MACROBLOCK *x, RD_STATS *rd_stats,
- BLOCK_SIZE bsize, int64_t ref_best_rd, int fast);
-int inter_block_uvrd(const AV1_COMP *cpi, MACROBLOCK *x, RD_STATS *rd_stats,
- BLOCK_SIZE bsize, int64_t ref_best_rd, int fast);
+static int inter_block_yrd(const AV1_COMP *cpi, MACROBLOCK *x,
+ RD_STATS *rd_stats, BLOCK_SIZE bsize,
+ int64_t ref_best_rd, int fast);
static unsigned pixel_dist_visible_only(
const AV1_COMP *const cpi, const MACROBLOCK *x, const uint8_t *src,
@@ -1866,19 +1865,6 @@
return error;
}
-int64_t av1_block_error_fp_c(const int16_t *coeff, const int16_t *dqcoeff,
- int block_size) {
- int i;
- int64_t error = 0;
-
- for (i = 0; i < block_size; i++) {
- const int diff = coeff[i] - dqcoeff[i];
- error += diff * diff;
- }
-
- return error;
-}
-
int64_t av1_highbd_block_error_c(const tran_low_t *coeff,
const tran_low_t *dqcoeff, intptr_t block_size,
int64_t *ssz, int bd) {
@@ -2029,9 +2015,9 @@
return n;
}
-void av1_inverse_transform_block_facade(MACROBLOCKD *xd, int plane, int block,
- int blk_row, int blk_col, int eob,
- int reduced_tx_set) {
+static void inverse_transform_block_facade(MACROBLOCKD *xd, int plane,
+ int block, int blk_row, int blk_col,
+ int eob, int reduced_tx_set) {
struct macroblockd_plane *const pd = &xd->plane[plane];
tran_low_t *dqcoeff = BLOCK_OFFSET(pd->dqcoeff, block);
const PLANE_TYPE plane_type = get_plane_type(plane);
@@ -2068,11 +2054,11 @@
tx_size;
}
-void dist_block(const AV1_COMP *cpi, MACROBLOCK *x, int plane,
- BLOCK_SIZE plane_bsize, int block, int blk_row, int blk_col,
- TX_SIZE tx_size, int64_t *out_dist, int64_t *out_sse,
- OUTPUT_STATUS output_status,
- int use_transform_domain_distortion) {
+static void dist_block(const AV1_COMP *cpi, MACROBLOCK *x, int plane,
+ BLOCK_SIZE plane_bsize, int block, int blk_row,
+ int blk_col, TX_SIZE tx_size, int64_t *out_dist,
+ int64_t *out_sse, OUTPUT_STATUS output_status,
+ int use_transform_domain_distortion) {
MACROBLOCKD *const xd = &x->e_mbd;
const struct macroblock_plane *const p = &x->plane[plane];
#if CONFIG_DIST_8X8
@@ -2711,9 +2697,9 @@
}
}
- av1_inverse_transform_block_facade(xd, plane, block, blk_row, blk_col,
- x->plane[plane].eobs[block],
- cm->reduced_tx_set_used);
+ inverse_transform_block_facade(xd, plane, block, blk_row, blk_col,
+ x->plane[plane].eobs[block],
+ cm->reduced_tx_set_used);
// This may happen because of hash collision. The eob stored in the hash
// table is non-zero, but the real eob is zero. We need to make sure tx_type
@@ -2945,39 +2931,6 @@
}
}
-// TODO(angiebird): use this function whenever it's possible
-int av1_tx_type_cost(const AV1_COMMON *cm, const MACROBLOCK *x,
- const MACROBLOCKD *xd, int plane, TX_SIZE tx_size,
- TX_TYPE tx_type) {
- if (plane > 0) return 0;
-
- const TX_SIZE square_tx_size = txsize_sqr_map[tx_size];
-
- const MB_MODE_INFO *mbmi = xd->mi[0];
- const int is_inter = is_inter_block(mbmi);
- if (get_ext_tx_types(tx_size, is_inter, cm->reduced_tx_set_used) > 1 &&
- !xd->lossless[xd->mi[0]->segment_id]) {
- const int ext_tx_set =
- get_ext_tx_set(tx_size, is_inter, cm->reduced_tx_set_used);
- if (is_inter) {
- if (ext_tx_set > 0)
- return x->inter_tx_type_costs[ext_tx_set][square_tx_size][tx_type];
- } else {
- if (ext_tx_set > 0) {
- PREDICTION_MODE intra_dir;
- if (mbmi->filter_intra_mode_info.use_filter_intra)
- intra_dir = fimode_to_intradir[mbmi->filter_intra_mode_info
- .filter_intra_mode];
- else
- intra_dir = mbmi->mode;
- return x->intra_tx_type_costs[ext_tx_set][square_tx_size][intra_dir]
- [tx_type];
- }
- }
- }
- return 0;
-}
-
static int64_t txfm_yrd(const AV1_COMP *const cpi, MACROBLOCK *x,
RD_STATS *rd_stats, int64_t ref_best_rd, BLOCK_SIZE bs,
TX_SIZE tx_size) {
@@ -4673,8 +4626,9 @@
// Return value 0: early termination triggered, no valid rd cost available;
// 1: rd cost values are valid.
-int inter_block_yrd(const AV1_COMP *cpi, MACROBLOCK *x, RD_STATS *rd_stats,
- BLOCK_SIZE bsize, int64_t ref_best_rd, int fast) {
+static int inter_block_yrd(const AV1_COMP *cpi, MACROBLOCK *x,
+ RD_STATS *rd_stats, BLOCK_SIZE bsize,
+ int64_t ref_best_rd, int fast) {
MACROBLOCKD *const xd = &x->e_mbd;
int is_cost_valid = 1;
int64_t this_rd = 0;
@@ -5146,8 +5100,9 @@
// Return value 0: early termination triggered, no valid rd cost available;
// 1: rd cost values are valid.
-int inter_block_uvrd(const AV1_COMP *cpi, MACROBLOCK *x, RD_STATS *rd_stats,
- BLOCK_SIZE bsize, int64_t ref_best_rd, int fast) {
+static int inter_block_uvrd(const AV1_COMP *cpi, MACROBLOCK *x,
+ RD_STATS *rd_stats, BLOCK_SIZE bsize,
+ int64_t ref_best_rd, int fast) {
MACROBLOCKD *const xd = &x->e_mbd;
MB_MODE_INFO *const mbmi = xd->mi[0];
int plane;
@@ -8773,75 +8728,6 @@
ctx->mbmi_ext = *x->mbmi_ext;
}
-// Do we have an internal image edge (e.g. formatting bars).
-int av1_internal_image_edge(const AV1_COMP *cpi) {
- return (cpi->oxcf.pass == 2) &&
- ((cpi->twopass.this_frame_stats.inactive_zone_rows > 0) ||
- (cpi->twopass.this_frame_stats.inactive_zone_cols > 0));
-}
-
-// Checks to see if a super block is on a horizontal image edge.
-// In most cases this is the "real" edge unless there are formatting
-// bars embedded in the stream.
-int av1_active_h_edge(const AV1_COMP *cpi, int mi_row, int mi_step) {
- int top_edge = 0;
- int bottom_edge = cpi->common.mi_rows;
- int is_active_h_edge = 0;
-
- // For two pass account for any formatting bars detected.
- if (cpi->oxcf.pass == 2) {
- const TWO_PASS *const twopass = &cpi->twopass;
-
- // The inactive region is specified in MBs not mi units.
- // The image edge is in the following MB row.
- top_edge += (int)(twopass->this_frame_stats.inactive_zone_rows * 2);
-
- bottom_edge -= (int)(twopass->this_frame_stats.inactive_zone_rows * 2);
- bottom_edge = AOMMAX(top_edge, bottom_edge);
- }
-
- if (((top_edge >= mi_row) && (top_edge < (mi_row + mi_step))) ||
- ((bottom_edge >= mi_row) && (bottom_edge < (mi_row + mi_step)))) {
- is_active_h_edge = 1;
- }
- return is_active_h_edge;
-}
-
-// Checks to see if a super block is on a vertical image edge.
-// In most cases this is the "real" edge unless there are formatting
-// bars embedded in the stream.
-int av1_active_v_edge(const AV1_COMP *cpi, int mi_col, int mi_step) {
- int left_edge = 0;
- int right_edge = cpi->common.mi_cols;
- int is_active_v_edge = 0;
-
- // For two pass account for any formatting bars detected.
- if (cpi->oxcf.pass == 2) {
- const TWO_PASS *const twopass = &cpi->twopass;
-
- // The inactive region is specified in MBs not mi units.
- // The image edge is in the following MB row.
- left_edge += (int)(twopass->this_frame_stats.inactive_zone_cols * 2);
-
- right_edge -= (int)(twopass->this_frame_stats.inactive_zone_cols * 2);
- right_edge = AOMMAX(left_edge, right_edge);
- }
-
- if (((left_edge >= mi_col) && (left_edge < (mi_col + mi_step))) ||
- ((right_edge >= mi_col) && (right_edge < (mi_col + mi_step)))) {
- is_active_v_edge = 1;
- }
- return is_active_v_edge;
-}
-
-// Checks to see if a super block is at the edge of the active image.
-// In most cases this is the "real" edge unless there are formatting
-// bars embedded in the stream.
-int av1_active_edge_sb(const AV1_COMP *cpi, int mi_row, int mi_col) {
- return av1_active_h_edge(cpi, mi_row, cpi->common.seq_params.mib_size) ||
- av1_active_v_edge(cpi, mi_col, cpi->common.seq_params.mib_size);
-}
-
static void restore_uv_color_map(const AV1_COMP *const cpi, MACROBLOCK *x) {
MACROBLOCKD *const xd = &x->e_mbd;
MB_MODE_INFO *const mbmi = xd->mi[0];
diff --git a/av1/encoder/rdopt.h b/av1/encoder/rdopt.h
index f30822d..49117c5 100644
--- a/av1/encoder/rdopt.h
+++ b/av1/encoder/rdopt.h
@@ -130,15 +130,6 @@
struct macroblock *x, int mi_row, int mi_col, struct RD_STATS *rd_cost,
BLOCK_SIZE bsize, PICK_MODE_CONTEXT *ctx, int64_t best_rd_so_far);
-int av1_internal_image_edge(const struct AV1_COMP *cpi);
-int av1_active_h_edge(const struct AV1_COMP *cpi, int mi_row, int mi_step);
-int av1_active_v_edge(const struct AV1_COMP *cpi, int mi_col, int mi_step);
-int av1_active_edge_sb(const struct AV1_COMP *cpi, int mi_row, int mi_col);
-
-int av1_tx_type_cost(const AV1_COMMON *cm, const MACROBLOCK *x,
- const MACROBLOCKD *xd, int plane, TX_SIZE tx_size,
- TX_TYPE tx_type);
-
void av1_inverse_transform_block_facade(MACROBLOCKD *xd, int plane, int block,
int blk_row, int blk_col, int eob,
int reduced_tx_set);
diff --git a/av1/encoder/speed_features.c b/av1/encoder/speed_features.c
index f79c293..5b15ba8 100644
--- a/av1/encoder/speed_features.c
+++ b/av1/encoder/speed_features.c
@@ -78,6 +78,13 @@
}
}
+// Do we have an internal image edge (e.g. formatting bars).
+static int has_internal_image_edge(const AV1_COMP *cpi) {
+ return (cpi->oxcf.pass == 2) &&
+ ((cpi->twopass.this_frame_stats.inactive_zone_rows > 0) ||
+ (cpi->twopass.this_frame_stats.inactive_zone_cols > 0));
+}
+
static void set_good_speed_feature_framesize_dependent(AV1_COMP *cpi,
SPEED_FEATURES *sf,
int speed) {
@@ -118,7 +125,7 @@
// Also if the image edge is internal to the coded area.
if ((speed >= 2) && (cpi->oxcf.pass == 2) &&
((cpi->twopass.fr_content_type == FC_GRAPHICS_ANIMATION) ||
- (av1_internal_image_edge(cpi)))) {
+ (has_internal_image_edge(cpi)))) {
sf->disable_split_mask = DISABLE_COMPOUND_SPLIT;
}
@@ -340,7 +347,7 @@
if (speed & PARTITION_SF) {
if ((cpi->twopass.fr_content_type == FC_GRAPHICS_ANIMATION) ||
- av1_internal_image_edge(cpi)) {
+ has_internal_image_edge(cpi)) {
sf->use_square_partition_only = !frame_is_boosted(cpi);
} else {
sf->use_square_partition_only = !frame_is_intra_only(cm);
diff --git a/av1/encoder/x86/error_sse2.asm b/av1/encoder/x86/error_sse2.asm
index 4680f1f..72e9e22 100644
--- a/av1/encoder/x86/error_sse2.asm
+++ b/av1/encoder/x86/error_sse2.asm
@@ -77,49 +77,3 @@
movd edx, m5
%endif
RET
-
-; Compute the sum of squared difference between two int16_t vectors.
-; int64_t av1_block_error_fp(int16_t *coeff, int16_t *dqcoeff,
-; intptr_t block_size)
-
-INIT_XMM sse2
-cglobal block_error_fp, 3, 3, 6, uqc, dqc, size
- pxor m4, m4 ; sse accumulator
- pxor m5, m5 ; dedicated zero register
- lea uqcq, [uqcq+sizeq*2]
- lea dqcq, [dqcq+sizeq*2]
- neg sizeq
-.loop:
- mova m2, [uqcq+sizeq*2]
- mova m0, [dqcq+sizeq*2]
- mova m3, [uqcq+sizeq*2+mmsize]
- mova m1, [dqcq+sizeq*2+mmsize]
- psubw m0, m2
- psubw m1, m3
- ; individual errors are max. 15bit+sign, so squares are 30bit, and
- ; thus the sum of 2 should fit in a 31bit integer (+ unused sign bit)
- pmaddwd m0, m0
- pmaddwd m1, m1
- ; accumulate in 64bit
- punpckldq m3, m0, m5
- punpckhdq m0, m5
- paddq m4, m3
- punpckldq m3, m1, m5
- paddq m4, m0
- punpckhdq m1, m5
- paddq m4, m3
- paddq m4, m1
- add sizeq, mmsize
- jl .loop
-
- ; accumulate horizontally and store in return value
- movhlps m5, m4
- paddq m4, m5
-%if ARCH_X86_64
- movq rax, m4
-%else
- pshufd m5, m4, 0x1
- movd eax, m4
- movd edx, m5
-%endif
- RET
