Add skip rd based gating for tx in compound type rd
Skip rd based gating is introduced for estimate_yrd_for_sb()
calls from compound type rd.
Encode Time Quality loss
cpu-used Reduction avg.psnr ovr.psnr ssim
3 0.507% 0.0023% 0.0027% 0.0159%
4 1.617% 0.0422% 0.0561% 0.0194%
5 2.961% 0.2044% 0.2238% 0.1039%
STATS_CHANGED
Change-Id: Iaefebd7f88ceeeb35f28a9249751957d7661780b
diff --git a/av1/encoder/compound_type.c b/av1/encoder/compound_type.c
index 14d30b0..83ecdee 100644
--- a/av1/encoder/compound_type.c
+++ b/av1/encoder/compound_type.c
@@ -20,7 +20,8 @@
typedef int64_t (*pick_interinter_mask_type)(
const AV1_COMP *const cpi, MACROBLOCK *x, const BLOCK_SIZE bsize,
const uint8_t *const p0, const uint8_t *const p1,
- const int16_t *const residual1, const int16_t *const diff10);
+ const int16_t *const residual1, const int16_t *const diff10,
+ uint64_t *best_sse);
// Checks if characteristics of search match
static INLINE int is_comp_rd_match(const AV1_COMP *const cpi,
@@ -184,7 +185,7 @@
const int16_t *const residual1,
const int16_t *const diff10,
int8_t *const best_wedge_sign,
- int8_t *const best_wedge_index) {
+ int8_t *const best_wedge_index, uint64_t *best_sse) {
const MACROBLOCKD *const xd = &x->e_mbd;
const struct buf_2d *const src = &x->plane[0].src;
const int bw = block_size_wide[bsize];
@@ -247,6 +248,7 @@
*best_wedge_index = wedge_index;
*best_wedge_sign = wedge_sign;
best_rd = rd;
+ *best_sse = sse;
}
}
@@ -255,13 +257,11 @@
}
// Choose the best wedge index the specified sign
-static int64_t pick_wedge_fixed_sign(const AV1_COMP *const cpi,
- const MACROBLOCK *const x,
- const BLOCK_SIZE bsize,
- const int16_t *const residual1,
- const int16_t *const diff10,
- const int8_t wedge_sign,
- int8_t *const best_wedge_index) {
+static int64_t pick_wedge_fixed_sign(
+ const AV1_COMP *const cpi, const MACROBLOCK *const x,
+ const BLOCK_SIZE bsize, const int16_t *const residual1,
+ const int16_t *const diff10, const int8_t wedge_sign,
+ int8_t *const best_wedge_index, uint64_t *best_sse) {
const MACROBLOCKD *const xd = &x->e_mbd;
const int bw = block_size_wide[bsize];
@@ -290,6 +290,7 @@
if (rd < best_rd) {
*best_wedge_index = wedge_index;
best_rd = rd;
+ *best_sse = sse;
}
}
return best_rd -
@@ -299,7 +300,8 @@
static int64_t pick_interinter_wedge(
const AV1_COMP *const cpi, MACROBLOCK *const x, const BLOCK_SIZE bsize,
const uint8_t *const p0, const uint8_t *const p1,
- const int16_t *const residual1, const int16_t *const diff10) {
+ const int16_t *const residual1, const int16_t *const diff10,
+ uint64_t *best_sse) {
MACROBLOCKD *const xd = &x->e_mbd;
MB_MODE_INFO *const mbmi = xd->mi[0];
const int bw = block_size_wide[bsize];
@@ -314,10 +316,10 @@
if (cpi->sf.inter_sf.fast_wedge_sign_estimate) {
wedge_sign = estimate_wedge_sign(cpi, x, bsize, p0, bw, p1, bw);
rd = pick_wedge_fixed_sign(cpi, x, bsize, residual1, diff10, wedge_sign,
- &wedge_index);
+ &wedge_index, best_sse);
} else {
rd = pick_wedge(cpi, x, bsize, p0, residual1, diff10, &wedge_sign,
- &wedge_index);
+ &wedge_index, best_sse);
}
mbmi->interinter_comp.wedge_sign = wedge_sign;
@@ -330,7 +332,8 @@
const uint8_t *const p0,
const uint8_t *const p1,
const int16_t *const residual1,
- const int16_t *const diff10) {
+ const int16_t *const diff10,
+ uint64_t *best_sse) {
MACROBLOCKD *const xd = &x->e_mbd;
MB_MODE_INFO *const mbmi = xd->mi[0];
const int bw = block_size_wide[bsize];
@@ -368,6 +371,7 @@
if (rd0 < best_rd) {
best_mask_type = cur_mask_type;
best_rd = rd0;
+ *best_sse = sse;
}
}
mbmi->interinter_comp.mask_type = best_mask_type;
@@ -407,8 +411,9 @@
aom_subtract_block(bh, bw, diff10, bw, p1, bw, p0, bw);
#endif
int8_t wedge_index = -1;
- int64_t rd =
- pick_wedge_fixed_sign(cpi, x, bsize, residual1, diff10, 0, &wedge_index);
+ uint64_t sse;
+ int64_t rd = pick_wedge_fixed_sign(cpi, x, bsize, residual1, diff10, 0,
+ &wedge_index, &sse);
mbmi->interintra_wedge_index = wedge_index;
return rd;
@@ -989,7 +994,7 @@
int mode_rate, int64_t rd_thresh, int *calc_pred_masked_compound,
int32_t *comp_rate, int64_t *comp_dist, int32_t *comp_model_rate,
int64_t *comp_model_dist, const int64_t comp_best_model_rd,
- int64_t *const comp_model_rd_cur, int *comp_rs2) {
+ int64_t *const comp_model_rd_cur, int *comp_rs2, int64_t ref_skip_rd) {
const AV1_COMMON *const cm = &cpi->common;
MACROBLOCKD *xd = &x->e_mbd;
MB_MODE_INFO *const mbmi = xd->mi[0];
@@ -1030,10 +1035,13 @@
// Function pointer to pick the appropriate mask
// compound_type == COMPOUND_WEDGE, calls pick_interinter_wedge()
// compound_type == COMPOUND_DIFFWTD, calls pick_interinter_seg()
+ uint64_t cur_sse = UINT64_MAX;
best_rd_cur = pick_interinter_mask[compound_type - COMPOUND_WEDGE](
- cpi, x, bsize, *preds0, *preds1, residual1, diff10);
+ cpi, x, bsize, *preds0, *preds1, residual1, diff10, &cur_sse);
*rs2 += get_interinter_compound_mask_rate(x, mbmi);
best_rd_cur += RDCOST(x->rdmult, *rs2 + rate_mv, 0);
+ assert(cur_sse != UINT64_MAX);
+ int64_t skip_rd_cur = RDCOST(x->rdmult, *rs2 + rate_mv, cur_sse);
// Although the true rate_mv might be different after motion search, but it
// is unlikely to be the best mode considering the transform rd cost and other
@@ -1044,6 +1052,18 @@
return INT64_MAX;
}
+ // Check if the mode is good enough based on skip rd
+ // TODO(nithya): Handle wedge_newmv_search if extending for lower speed
+ // setting
+ if (cpi->sf.inter_sf.txfm_rd_gate_level) {
+ int eval_txfm = check_txfm_eval(x, bsize, ref_skip_rd, skip_rd_cur,
+ cpi->sf.inter_sf.txfm_rd_gate_level, 1);
+ if (!eval_txfm) {
+ *comp_model_rd_cur = INT64_MAX;
+ return INT64_MAX;
+ }
+ }
+
// Compute cost if matching record not found, else, reuse data
if (comp_rate[compound_type] == INT_MAX) {
// Check whether new MV search for wedge is to be done
@@ -1081,7 +1101,7 @@
tmp_rd = pick_interinter_mask[compound_type - COMPOUND_WEDGE](
cpi, x, bsize, *tmp_preds0, *tmp_preds1, tmp_buf.residual1,
- tmp_buf.diff10);
+ tmp_buf.diff10, &cur_sse);
// we can reuse rs2 here
tmp_rd += RDCOST(x->rdmult, *rs2 + *out_rate_mv, 0);
@@ -1166,7 +1186,8 @@
const BUFFER_SET *tmp_dst,
const CompoundTypeRdBuffers *buffers, int *rate_mv,
int64_t *rd, RD_STATS *rd_stats, int64_t ref_best_rd,
- int *is_luma_interp_done, int64_t rd_thresh) {
+ int64_t ref_skip_rd, int *is_luma_interp_done,
+ int64_t rd_thresh) {
const AV1_COMMON *cm = &cpi->common;
MACROBLOCKD *xd = &x->e_mbd;
MB_MODE_INFO *mbmi = xd->mi[0];
@@ -1275,6 +1296,7 @@
}
restore_dst_buf(xd, *tmp_dst, 1);
} else {
+ int64_t sse_y[COMPOUND_DISTWTD + 1];
// Calculate model_rd for COMPOUND_AVERAGE and COMPOUND_DISTWTD
for (int comp_type = COMPOUND_AVERAGE; comp_type <= COMPOUND_DISTWTD;
comp_type++) {
@@ -1287,6 +1309,7 @@
est_rate[comp_type] += masked_type_cost[comp_type];
comp_model_rate[comp_type] = est_rate[comp_type];
comp_model_dist[comp_type] = est_dist[comp_type];
+ sse_y[comp_type] = x->pred_sse[xd->mi[0]->ref_frame[0]];
if (comp_type == COMPOUND_AVERAGE) {
*is_luma_interp_done = 1;
restore_dst_buf(xd, *tmp_dst, 1);
@@ -1302,8 +1325,19 @@
RD_STATS est_rd_stats;
const int64_t mode_rd = RDCOST(x->rdmult, rs2 + *rate_mv, 0);
const int64_t tmp_rd_thresh = AOMMIN(*rd, rd_thresh) - mode_rd;
- const int64_t est_rd_ =
- estimate_yrd_for_sb(cpi, bsize, x, tmp_rd_thresh, &est_rd_stats);
+ int64_t est_rd_ = INT64_MAX;
+ int eval_txfm = 1;
+ // Check if the mode is good enough based on skip rd
+ if (cpi->sf.inter_sf.txfm_rd_gate_level) {
+ int64_t skip_rd = RDCOST(x->rdmult, rs2 + *rate_mv, sse_y[best_type]);
+ eval_txfm = check_txfm_eval(x, bsize, ref_skip_rd, skip_rd,
+ cpi->sf.inter_sf.txfm_rd_gate_level, 1);
+ }
+ // Evaluate further if skip rd is low enough
+ if (eval_txfm) {
+ est_rd_ =
+ estimate_yrd_for_sb(cpi, bsize, x, tmp_rd_thresh, &est_rd_stats);
+ }
if (est_rd_ != INT64_MAX) {
best_rd_cur = RDCOST(x->rdmult, rs2 + *rate_mv + est_rd_stats.rate,
@@ -1348,8 +1382,21 @@
// Compute RD cost for the current type
RD_STATS est_rd_stats;
const int64_t tmp_rd_thresh = AOMMIN(*rd, rd_thresh) - mode_rd;
- const int64_t est_rd =
- estimate_yrd_for_sb(cpi, bsize, x, tmp_rd_thresh, &est_rd_stats);
+ int64_t est_rd = INT64_MAX;
+ int eval_txfm = 1;
+ // Check if the mode is good enough based on skip rd
+ if (cpi->sf.inter_sf.txfm_rd_gate_level) {
+ int64_t sse_y = compute_sse_plane(x, xd, PLANE_TYPE_Y, bsize);
+ int64_t skip_rd = RDCOST(x->rdmult, rs2 + *rate_mv, sse_y);
+ eval_txfm = check_txfm_eval(x, bsize, ref_skip_rd, skip_rd,
+ cpi->sf.inter_sf.txfm_rd_gate_level, 1);
+ }
+ // Evaluate further if skip rd is low enough
+ if (eval_txfm) {
+ est_rd = estimate_yrd_for_sb(cpi, bsize, x, tmp_rd_thresh,
+ &est_rd_stats);
+ }
+
if (est_rd != INT64_MAX) {
best_rd_cur = RDCOST(x->rdmult, rs2 + *rate_mv + est_rd_stats.rate,
est_rd_stats.dist);
@@ -1393,7 +1440,8 @@
&tmp_rate_mv, preds0, preds1, buffers->residual1, buffers->diff10,
strides, rd_stats->rate, tmp_rd_thresh, &calc_pred_masked_compound,
comp_rate, comp_dist, comp_model_rate, comp_model_dist,
- best_type_stats.comp_best_model_rd, &comp_model_rd_cur, comp_rs2);
+ best_type_stats.comp_best_model_rd, &comp_model_rd_cur, comp_rs2,
+ ref_skip_rd);
}
}
// Update stats for best compound type
diff --git a/av1/encoder/compound_type.h b/av1/encoder/compound_type.h
index c1ef298..f2bd857 100644
--- a/av1/encoder/compound_type.h
+++ b/av1/encoder/compound_type.h
@@ -38,7 +38,8 @@
const BUFFER_SET *tmp_dst,
const CompoundTypeRdBuffers *buffers, int *rate_mv,
int64_t *rd, RD_STATS *rd_stats, int64_t ref_best_rd,
- int *is_luma_interp_done, int64_t rd_thresh);
+ int64_t ref_skip_rd, int *is_luma_interp_done,
+ int64_t rd_thresh);
#ifdef __cplusplus
} // extern "C"
diff --git a/av1/encoder/model_rd.h b/av1/encoder/model_rd.h
index 1ca636d..c353c8f 100644
--- a/av1/encoder/model_rd.h
+++ b/av1/encoder/model_rd.h
@@ -70,6 +70,21 @@
return sse;
}
+static AOM_INLINE int64_t compute_sse_plane(MACROBLOCK *x, MACROBLOCKD *xd,
+ int plane, const BLOCK_SIZE bsize) {
+ struct macroblockd_plane *const pd = &xd->plane[plane];
+ const BLOCK_SIZE plane_bsize =
+ get_plane_block_size(bsize, pd->subsampling_x, pd->subsampling_y);
+ int bw, bh;
+ const struct macroblock_plane *const p = &x->plane[plane];
+ get_txb_dimensions(xd, plane, plane_bsize, 0, 0, plane_bsize, NULL, NULL, &bw,
+ &bh);
+
+ int64_t sse = calculate_sse(xd, p, pd, bw, bh);
+
+ return sse;
+}
+
static AOM_INLINE void model_rd_from_sse(const AV1_COMP *const cpi,
const MACROBLOCK *const x,
BLOCK_SIZE plane_bsize, int plane,
diff --git a/av1/encoder/rdopt.c b/av1/encoder/rdopt.c
index 1bf7b81..4e19e24 100644
--- a/av1/encoder/rdopt.c
+++ b/av1/encoder/rdopt.c
@@ -1502,11 +1502,9 @@
rd_stats->rdcost = est_rd;
if (rd_stats->rdcost < *best_est_rd) {
*best_est_rd = rd_stats->rdcost;
- int64_t skip_rdy = INT64_MAX;
- if (cpi->sf.inter_sf.txfm_rd_gate_level) {
- skip_rdy = RDCOST(x->rdmult, mode_rate, sse_y);
- }
- ref_skip_rd[1] = skip_rdy;
+ ref_skip_rd[1] = cpi->sf.inter_sf.txfm_rd_gate_level
+ ? RDCOST(x->rdmult, mode_rate, sse_y)
+ : INT64_MAX;
}
if (cm->current_frame.reference_mode == SINGLE_REFERENCE) {
if (!is_comp_pred) {
@@ -1535,7 +1533,7 @@
skip_rd = RDCOST(x->rdmult, rd_stats->rate, curr_sse);
skip_rdy = RDCOST(x->rdmult, rd_stats->rate, sse_y);
int eval_txfm = check_txfm_eval(x, bsize, ref_skip_rd[0], skip_rd,
- cpi->sf.inter_sf.txfm_rd_gate_level);
+ cpi->sf.inter_sf.txfm_rd_gate_level, 0);
if (!eval_txfm) continue;
}
@@ -2491,7 +2489,7 @@
compmode_interinter_cost = av1_compound_type_rd(
cpi, x, bsize, cur_mv, mode_search_mask, masked_compound_used,
&orig_dst, &tmp_dst, rd_buffers, &rate_mv, &best_rd_compound,
- rd_stats, ref_best_rd, &is_luma_interp_done, rd_thresh);
+ rd_stats, ref_best_rd, skip_rd[1], &is_luma_interp_done, rd_thresh);
if (ref_best_rd < INT64_MAX &&
(best_rd_compound >> comp_type_rd_shift) * comp_type_rd_scale >
ref_best_rd) {
@@ -4633,7 +4631,7 @@
skip_rd = RDCOST(x->rdmult, mode_rate, curr_sse);
int eval_txfm =
check_txfm_eval(x, bsize, search_state.best_skip_rd[0], skip_rd,
- cpi->sf.inter_sf.txfm_rd_gate_level);
+ cpi->sf.inter_sf.txfm_rd_gate_level, 0);
if (!eval_txfm) continue;
}
diff --git a/av1/encoder/rdopt_utils.h b/av1/encoder/rdopt_utils.h
index 4a4344e..4d0b05e 100644
--- a/av1/encoder/rdopt_utils.h
+++ b/av1/encoder/rdopt_utils.h
@@ -329,21 +329,29 @@
static INLINE int check_txfm_eval(MACROBLOCK *const x, BLOCK_SIZE bsize,
int64_t best_skip_rd, int64_t skip_rd,
- int level) {
+ int level, int is_luma_only) {
int eval_txfm = 1;
// Derive aggressiveness factor for gating the transform search
- // Lower value indicates more aggresiveness. Be more conservative (high value)
- // for (i) low quantizers (ii) regions where prediction is poor
- const int scale[3] = { INT_MAX, 3, 2 };
+ // Lower value indicates more aggressiveness. Be more conservative (high
+ // value) for (i) low quantizers (ii) regions where prediction is poor
+ const int scale[4] = { INT_MAX, 3, 3, 2 };
+
int aggr_factor =
AOMMAX(1, ((MAXQ - x->qindex) * 2 + QINDEX_RANGE / 2) >> QINDEX_BITS);
if (best_skip_rd >
(x->source_variance << (num_pels_log2_lookup[bsize] + RDDIV_BITS)))
aggr_factor *= scale[level];
- int64_t rd_thresh = (best_skip_rd == INT64_MAX)
- ? best_skip_rd
- : (int64_t)(best_skip_rd * aggr_factor);
+ // Be more conservative for luma only cases (called from compound type rd)
+ // since best_skip_rd is computed after and skip_rd is computed (with 8-bit
+ // prediction signals blended for WEDGE/DIFFWTD rather than 16-bit) before
+ // interpolation filter search
+ const int luma_mul[4] = { INT_MAX, 16, 15, 11 };
+ int mul_factor = is_luma_only ? luma_mul[level] : 8;
+ int64_t rd_thresh =
+ (best_skip_rd == INT64_MAX)
+ ? best_skip_rd
+ : (int64_t)(best_skip_rd * aggr_factor * mul_factor >> 3);
if (skip_rd > rd_thresh) eval_txfm = 0;
return eval_txfm;
}
diff --git a/av1/encoder/speed_features.c b/av1/encoder/speed_features.c
index 24639d8..0024c66 100644
--- a/av1/encoder/speed_features.c
+++ b/av1/encoder/speed_features.c
@@ -510,7 +510,7 @@
sf->inter_sf.alt_ref_search_fp = 1;
sf->inter_sf.prune_ref_mv_idx_search = 1;
sf->inter_sf.txfm_rd_gate_level =
- (boosted || cm->allow_screen_content_tools) ? 0 : 1;
+ (boosted || cm->allow_screen_content_tools) ? 0 : 2;
sf->inter_sf.disable_smooth_interintra = 1;
@@ -577,7 +577,7 @@
sf->inter_sf.disable_obmc = 1;
sf->inter_sf.disable_onesided_comp = 1;
sf->inter_sf.txfm_rd_gate_level =
- (boosted || cm->allow_screen_content_tools) ? 0 : 2;
+ (boosted || cm->allow_screen_content_tools) ? 0 : 3;
sf->inter_sf.prune_inter_modes_if_skippable = 1;
sf->lpf_sf.lpf_pick = LPF_PICK_FROM_FULL_IMAGE_NON_DUAL;
