Improve winner mode processing of key frame
- Perform winner mode processing only when a mode is identified with less rd
than best_rd
- Disable rd based breakouts in winner mode processing
Instruction Count
cpu-used Reduction(%) Quality Loss(%)
2 0.00 +0.00
3 0.23 +0.04
4 0.55 +0.00
STATS_CHANGED
Change-Id: If820b88aa65ffdc1de7b05f2518d2dcda21a369b
diff --git a/av1/encoder/rdopt.c b/av1/encoder/rdopt.c
index 65f65f6..303f154 100644
--- a/av1/encoder/rdopt.c
+++ b/av1/encoder/rdopt.c
@@ -4228,8 +4228,8 @@
int *centroids, int n, uint16_t *color_cache, int n_cache,
MB_MODE_INFO *best_mbmi, uint8_t *best_palette_color_map, int64_t *best_rd,
int64_t *best_model_rd, int *rate, int *rate_tokenonly, int *rate_overhead,
- int64_t *distortion, int *skippable, PICK_MODE_CONTEXT *ctx,
- uint8_t *blk_skip) {
+ int64_t *distortion, int *skippable, int *beat_best_rd,
+ PICK_MODE_CONTEXT *ctx, uint8_t *blk_skip) {
optimize_palette_colors(color_cache, n_cache, n, 1, centroids);
int k = av1_remove_duplicates(centroids, n);
if (k < PALETTE_MIN_SIZE) {
@@ -4271,6 +4271,9 @@
}
if (this_rd < *best_rd) {
*best_rd = this_rd;
+ // Setting beat_best_rd flag because current mode rd is better than best_rd.
+ // This flag need to be updated only for palette evaluation in key frames
+ if (beat_best_rd) *beat_best_rd = 1;
memcpy(best_palette_color_map, color_map,
block_width * block_height * sizeof(color_map[0]));
*best_mbmi = *mbmi;
@@ -4288,7 +4291,7 @@
int mi_col, int dc_mode_cost, MB_MODE_INFO *best_mbmi,
uint8_t *best_palette_color_map, int64_t *best_rd, int64_t *best_model_rd,
int *rate, int *rate_tokenonly, int64_t *distortion, int *skippable,
- PICK_MODE_CONTEXT *ctx, uint8_t *best_blk_skip) {
+ int *beat_best_rd, PICK_MODE_CONTEXT *ctx, uint8_t *best_blk_skip) {
int rate_overhead = 0;
MACROBLOCKD *const xd = &x->e_mbd;
MB_MODE_INFO *const mbmi = xd->mi[0];
@@ -4375,8 +4378,8 @@
palette_rd_y(cpi, x, mbmi, bsize, mi_row, mi_col, dc_mode_cost, data,
centroids, n, color_cache, n_cache, best_mbmi,
best_palette_color_map, best_rd, best_model_rd, rate,
- rate_tokenonly, &rate_overhead, distortion, skippable, ctx,
- best_blk_skip);
+ rate_tokenonly, &rate_overhead, distortion, skippable,
+ beat_best_rd, ctx, best_blk_skip);
}
// K-means clustering.
@@ -4396,8 +4399,8 @@
palette_rd_y(cpi, x, mbmi, bsize, mi_row, mi_col, dc_mode_cost, data,
centroids, n, color_cache, n_cache, best_mbmi,
best_palette_color_map, best_rd, best_model_rd, rate,
- rate_tokenonly, &rate_overhead, distortion, skippable, ctx,
- best_blk_skip);
+ rate_tokenonly, &rate_overhead, distortion, skippable,
+ beat_best_rd, ctx, best_blk_skip);
}
}
@@ -4714,7 +4717,9 @@
MACROBLOCKD *const xd = &x->e_mbd;
MB_MODE_INFO *const mbmi = xd->mi[0];
RD_STATS rd_stats;
- super_block_yrd(cpi, x, &rd_stats, bsize, *best_rd);
+ // In order to improve txfm search avoid rd based breakouts during winner
+ // mode evaluation. Hence passing ref_best_rd as a maximum value
+ super_block_yrd(cpi, x, &rd_stats, bsize, INT64_MAX);
if (rd_stats.rate == INT_MAX) return;
int this_rate_tokenonly = rd_stats.rate;
if (!xd->lossless[mbmi->segment_id] && block_signals_txsize(mbmi->sb_type)) {
@@ -4754,6 +4759,8 @@
const int cols = block_size_wide[bsize];
int is_directional_mode;
uint8_t directional_mode_skip_mask[INTRA_MODES] = { 0 };
+ // Flag to check rd of any intra mode is better than best_rd passed to this
+ // function
int beat_best_rd = 0;
const int *bmode_costs;
PALETTE_MODE_INFO *const pmi = &mbmi->palette_mode_info;
@@ -4834,6 +4841,8 @@
if (this_rd < best_rd) {
best_mbmi = *mbmi;
best_rd = this_rd;
+ // Setting beat_best_rd flag because current mode rd is better than
+ // best_rd passed to this function
beat_best_rd = 1;
*rate = this_rate;
*rate_tokenonly = this_rate_tokenonly;
@@ -4848,7 +4857,7 @@
rd_pick_palette_intra_sby(
cpi, x, bsize, mi_row, mi_col, bmode_costs[DC_PRED], &best_mbmi,
best_palette_color_map, &best_rd, &best_model_rd, rate, rate_tokenonly,
- distortion, skippable, ctx, ctx->blk_skip);
+ distortion, skippable, &beat_best_rd, ctx, ctx->blk_skip);
}
if (beat_best_rd && av1_filter_intra_allowed_bsize(&cpi->common, bsize)) {
@@ -4858,7 +4867,10 @@
best_mbmi = *mbmi;
}
}
-
+ // No mode is identified with less rd value than best_rd passed to this
+ // function. In such cases winner mode processing is not necessary and return
+ // best_rd as INT64_MAX to indicate best mode is not identified
+ if (!beat_best_rd) return INT64_MAX;
// If previous searches use only the default tx type/no R-D optimization of
// quantized coeffs, do an extra search for the best tx type/better R-D
// optimization of quantized coeffs
@@ -11609,7 +11621,7 @@
rate_overhead_palette = rd_pick_palette_intra_sby(
cpi, x, bsize, mi_row, mi_col, intra_mode_cost[DC_PRED],
&best_mbmi_palette, best_palette_color_map, &best_rd_palette,
- &best_model_rd_palette, NULL, NULL, NULL, NULL, ctx, best_blk_skip);
+ &best_model_rd_palette, NULL, NULL, NULL, NULL, NULL, ctx, best_blk_skip);
if (pmi->palette_size[0] == 0) return;
memcpy(x->blk_skip, best_blk_skip,
