ref: 2bd6fe08f8f6b84816e4369ae53725cf97a43c41
parent: 37501d687c509c348bc2bdbe75aa050e2ab0437c
author: Jingning Han <[email protected]>
date: Tue Jul 2 12:48:15 EDT 2013
Enable early termination in rd search This commit allows encoder to detect the cumulative rate-distortion cost per transformed block inside a partition. If the cumulative rd cost is already above the best rd value, it terminates the rest operations and continue to next prediction mode test. It reduces the runtime of bus at target bit-rate 2000 from 308 second to 266 second, i.e., about 13% speed-up at no performance penalty. Change-Id: I5f15a3d8955d97031d5653006027866a00654e7a
--- a/vp9/encoder/vp9_rdopt.c
+++ b/vp9/encoder/vp9_rdopt.c
@@ -705,7 +705,8 @@
args->bw * args->bh);
}
-
+// FIXME(jingning): need to make the rd test of chroma components consistent
+// with that of luma component. this function should be deprecated afterwards.
static int rdcost_plane(VP9_COMMON * const cm, MACROBLOCK *x, int plane,
BLOCK_SIZE_TYPE bsize, TX_SIZE tx_size) {MACROBLOCKD * const xd = &x->e_mbd;
@@ -713,7 +714,7 @@
const int bhl = b_height_log2(bsize) - xd->plane[plane].subsampling_y;
const int bw = 1 << bwl, bh = 1 << bhl;
struct rdcost_block_args args = { cm, x, { 0 }, { 0 }, tx_size, bw, bh,- 0, 0, 0, 0, 0 };
+ 0, 0, 0, INT64_MAX, 0 };
vpx_memcpy(&args.t_above, xd->plane[plane].above_context,
sizeof(ENTROPY_CONTEXT) * bw);
@@ -784,6 +785,16 @@
MACROBLOCKD *const xd = &x->e_mbd;
struct encode_b_args encode_args = {args->cm, x, NULL};+ if (args->skip)
+ return;
+ if (RDCOST(x->rdmult, x->rddiv, args->rate, args->dist) > args->best_rd) {+ args->skip = 1;
+ args->rate = INT_MAX;
+ args->dist = INT64_MAX;
+ args->sse = INT64_MAX;
+ return;
+ }
+
if (xd->mode_info_context->mbmi.ref_frame[0] == INTRA_FRAME)
encode_block_intra(plane, block, bsize, ss_txfrm_size, &encode_args);
else
@@ -796,6 +807,7 @@
static void super_block_yrd_for_txfm(VP9_COMMON *const cm, MACROBLOCK *x,
int *rate, int64_t *distortion,
int *skippable, int64_t *sse,
+ int64_t ref_best_rd,
BLOCK_SIZE_TYPE bsize, TX_SIZE tx_size) {MACROBLOCKD *const xd = &x->e_mbd;
struct macroblockd_plane *const pd = &xd->plane[0];
@@ -803,7 +815,7 @@
const int bhl = b_height_log2(bsize) - xd->plane[0].subsampling_y;
const int bw = 1 << bwl, bh = 1 << bhl;
struct rdcost_block_args args = { cm, x, { 0 }, { 0 }, tx_size, bw, bh,- 0, 0, 0, 0, 0 };
+ 0, 0, 0, ref_best_rd, 0 };
xd->mode_info_context->mbmi.txfm_size = tx_size;
vpx_memcpy(&args.t_above, pd->above_context, sizeof(ENTROPY_CONTEXT) * bw);
vpx_memcpy(&args.t_left, pd->left_context, sizeof(ENTROPY_CONTEXT) * bh);
@@ -812,7 +824,7 @@
*distortion = args.dist;
*rate = args.rate;
*sse = args.sse;
- *skippable = vp9_sby_is_skippable(xd, bsize);
+ *skippable = vp9_sby_is_skippable(xd, bsize) && (!args.skip);
}
static void choose_largest_txfm_size(VP9_COMP *cpi, MACROBLOCK *x,
@@ -839,7 +851,7 @@
mbmi->txfm_size = TX_4X4;
}
super_block_yrd_for_txfm(cm, x, rate, distortion, skip,
- &sse[mbmi->txfm_size], bs,
+ &sse[mbmi->txfm_size], INT64_MAX, bs,
mbmi->txfm_size);
cpi->txfm_stepdown_count[0]++;
}
@@ -864,6 +876,8 @@
for (n = TX_4X4; n <= max_txfm_size; n++) {r[n][1] = r[n][0];
+ if (r[n][0] == INT_MAX)
+ continue;
for (m = 0; m <= n - (n == max_txfm_size); m++) {if (m == n)
r[n][1] += vp9_cost_zero(tx_probs[m]);
@@ -877,6 +891,10 @@
s1 = vp9_cost_bit(skip_prob, 1);
for (n = TX_4X4; n <= max_txfm_size; n++) {+ if (d[n] == INT64_MAX) {+ rd[n][0] = rd[n][1] = INT64_MAX;
+ continue;
+ }
if (s[n]) {rd[n][0] = rd[n][1] = RDCOST(x->rdmult, x->rddiv, s1, d[n]);
} else {@@ -1020,7 +1038,8 @@
// Actually encode using the chosen mode if a model was used, but do not
// update the r, d costs
super_block_yrd_for_txfm(cm, x, rate, distortion, skip,
- &sse[mbmi->txfm_size], bs, mbmi->txfm_size);
+ &sse[mbmi->txfm_size], INT64_MAX,
+ bs, mbmi->txfm_size);
} else {*distortion = d[mbmi->txfm_size];
*rate = r[mbmi->txfm_size][cm->txfm_mode == TX_MODE_SELECT];
@@ -1046,7 +1065,8 @@
static void super_block_yrd(VP9_COMP *cpi,
MACROBLOCK *x, int *rate, int64_t *distortion,
int *skip, int64_t *psse, BLOCK_SIZE_TYPE bs,
- int64_t txfm_cache[NB_TXFM_MODES]) {+ int64_t txfm_cache[NB_TXFM_MODES],
+ int64_t ref_best_rd) {VP9_COMMON *const cm = &cpi->common;
int r[TX_SIZE_MAX_SB][2], s[TX_SIZE_MAX_SB];
int64_t d[TX_SIZE_MAX_SB], sse[TX_SIZE_MAX_SB];
@@ -1076,7 +1096,8 @@
&r[TX_32X32][0], &d[TX_32X32], &s[TX_32X32]);
} else {super_block_yrd_for_txfm(cm, x, &r[TX_32X32][0], &d[TX_32X32],
- &s[TX_32X32], &sse[TX_32X32], bs, TX_32X32);
+ &s[TX_32X32], &sse[TX_32X32], INT64_MAX,
+ bs, TX_32X32);
}
}
if (bs >= BLOCK_SIZE_MB16X16) {@@ -1085,7 +1106,8 @@
&r[TX_16X16][0], &d[TX_16X16], &s[TX_16X16]);
} else {super_block_yrd_for_txfm(cm, x, &r[TX_16X16][0], &d[TX_16X16],
- &s[TX_16X16], &sse[TX_16X16], bs, TX_16X16);
+ &s[TX_16X16], &sse[TX_16X16], INT64_MAX,
+ bs, TX_16X16);
}
}
if (model_used[TX_8X8]) {@@ -1093,7 +1115,7 @@
&r[TX_8X8][0], &d[TX_8X8], &s[TX_8X8]);
} else {super_block_yrd_for_txfm(cm, x, &r[TX_8X8][0], &d[TX_8X8], &s[TX_8X8],
- &sse[TX_8X8], bs, TX_8X8);
+ &sse[TX_8X8], INT64_MAX, bs, TX_8X8);
}
if (model_used[TX_4X4]) {model_rd_for_sb_y_tx(cpi, bs, TX_4X4, x, xd,
@@ -1100,7 +1122,7 @@
&r[TX_4X4][0], &d[TX_4X4], &s[TX_4X4]);
} else {super_block_yrd_for_txfm(cm, x, &r[TX_4X4][0], &d[TX_4X4], &s[TX_4X4],
- &sse[TX_4X4], bs, TX_4X4);
+ &sse[TX_4X4], INT64_MAX, bs, TX_4X4);
}
choose_txfm_size_from_modelrd(cpi, x, r, rate, d, distortion, s,
skip, sse, bs, model_used);
@@ -1107,14 +1129,16 @@
} else {if (bs >= BLOCK_SIZE_SB32X32)
super_block_yrd_for_txfm(cm, x, &r[TX_32X32][0], &d[TX_32X32],
- &s[TX_32X32], &sse[TX_32X32], bs, TX_32X32);
+ &s[TX_32X32], &sse[TX_32X32], ref_best_rd,
+ bs, TX_32X32);
if (bs >= BLOCK_SIZE_MB16X16)
super_block_yrd_for_txfm(cm, x, &r[TX_16X16][0], &d[TX_16X16],
- &s[TX_16X16], &sse[TX_16X16], bs, TX_16X16);
+ &s[TX_16X16], &sse[TX_16X16], ref_best_rd,
+ bs, TX_16X16);
super_block_yrd_for_txfm(cm, x, &r[TX_8X8][0], &d[TX_8X8], &s[TX_8X8],
- &sse[TX_8X8], bs, TX_8X8);
+ &sse[TX_8X8], ref_best_rd, bs, TX_8X8);
super_block_yrd_for_txfm(cm, x, &r[TX_4X4][0], &d[TX_4X4], &s[TX_4X4],
- &sse[TX_4X4], bs, TX_4X4);
+ &sse[TX_4X4], ref_best_rd, bs, TX_4X4);
choose_txfm_size_from_rd(cpi, x, r, rate, d, distortion, s,
skip, txfm_cache, bs);
}
@@ -1394,8 +1418,11 @@
x->e_mbd.mode_info_context->mbmi.mode = mode;
super_block_yrd(cpi, x, &this_rate_tokenonly, &this_distortion, &s, NULL,
- bsize, local_txfm_cache);
+ bsize, local_txfm_cache, best_rd);
+ if (this_rate_tokenonly == INT_MAX)
+ continue;
+
this_rate = this_rate_tokenonly + bmode_costs[mode];
this_rd = RDCOST(x->rdmult, x->rddiv, this_rate, this_distortion);
@@ -1409,7 +1436,7 @@
*skippable = s;
}
- if (cpi->sf.tx_size_search_method == USE_FULL_RD) {+ if (cpi->sf.tx_size_search_method == USE_FULL_RD && this_rd < INT64_MAX) { for (i = 0; i < NB_TXFM_MODES; i++) {int64_t adj_rd = this_rd + local_txfm_cache[i] -
local_txfm_cache[cpi->common.txfm_mode];
@@ -2525,7 +2552,7 @@
int_mv *frame_mv,
int mi_row, int mi_col,
int_mv single_newmv[MAX_REF_FRAMES],
- int64_t *psse) {+ int64_t *psse, int64_t ref_best_rd) {VP9_COMMON *cm = &cpi->common;
MACROBLOCKD *xd = &x->e_mbd;
MB_MODE_INFO *mbmi = &xd->mode_info_context->mbmi;
@@ -2758,8 +2785,14 @@
// Y cost and distortion
super_block_yrd(cpi, x, rate_y, distortion_y, &skippable_y, psse,
- bsize, txfm_cache);
+ bsize, txfm_cache, ref_best_rd);
+ if (*rate_y == INT_MAX) {+ *rate2 = INT_MAX;
+ *distortion = INT64_MAX;
+ return INT64_MAX;
+ }
+
*rate2 += *rate_y;
*distortion += *distortion_y;
@@ -3166,8 +3199,11 @@
continue;
}
super_block_yrd(cpi, x, &rate_y, &distortion_y, &skippable, NULL,
- bsize, txfm_cache);
+ bsize, txfm_cache, best_rd);
+ if (rate_y == INT_MAX)
+ continue;
+
uv_tx = mbmi->txfm_size;
if (bsize < BLOCK_SIZE_MB16X16 && uv_tx == TX_8X8)
uv_tx = TX_4X4;
@@ -3319,7 +3355,7 @@
&mode_excluded, &disable_skip,
&tmp_best_filter, frame_mv[this_mode],
mi_row, mi_col,
- single_newmv, &total_sse);
+ single_newmv, &total_sse, best_rd);
if (this_rd == INT64_MAX)
continue;
}