From 9c63e887e675d2bf5934c82a92f289ed33f0c8a4 Mon Sep 17 00:00:00 2001
From: Jim Bankoski <jimbankoski@google.com>
Date: Mon, 3 Mar 2014 11:49:13 -0800
Subject: [PATCH] cost_coefs approximation speed up
This patch adds a new speed feature which doesn't do the rather
expensive entropy context lookup or save to the table, while
doing costing.
The speed up on desktop36p.y4m is around 10% other clips much less.
On the RTC test set this was + 1% in overall datarate.
Change-Id: Ia5144bbf45270671e7be9c8e4055369909e2f738
---
vp9/encoder/vp9_onyx_if.c | 5 +++
vp9/encoder/vp9_onyx_int.h | 4 +++
vp9/encoder/vp9_rdopt.c | 70 ++++++++++++++++++++++++--------------
3 files changed, 53 insertions(+), 26 deletions(-)
diff --git a/vp9/encoder/vp9_onyx_if.c b/vp9/encoder/vp9_onyx_if.c
index d7bac6f0bc..db0fd80fa1 100644
--- a/vp9/encoder/vp9_onyx_if.c
+++ b/vp9/encoder/vp9_onyx_if.c
@@ -662,6 +662,7 @@ static void set_good_speed_feature(VP9_COMMON *cm,
sf->use_lp32x32fdct = 1;
sf->subpel_iters_per_step = 1;
sf->use_fast_coef_updates = 2;
+ sf->use_fast_coef_costing = 1;
sf->adaptive_rd_thresh = 4;
sf->mode_skip_start = 6;
@@ -698,6 +699,7 @@ static void set_good_speed_feature(VP9_COMMON *cm,
sf->use_lp32x32fdct = 1;
sf->subpel_iters_per_step = 1;
sf->use_fast_coef_updates = 2;
+ sf->use_fast_coef_costing = 1;
sf->adaptive_rd_thresh = 4;
sf->mode_skip_start = 6;
@@ -730,6 +732,7 @@ static void set_good_speed_feature(VP9_COMMON *cm,
sf->intra_uv_mode_mask[i] = INTRA_DC_ONLY;
}
sf->use_fast_coef_updates = 2;
+ sf->use_fast_coef_costing = 1;
sf->adaptive_rd_thresh = 4;
sf->mode_skip_start = 6;
}
@@ -742,6 +745,7 @@ static void set_rt_speed_feature(VP9_COMMON *cm,
sf->adaptive_rd_thresh = 1;
sf->recode_loop = ((speed < 1) ? ALLOW_RECODE : ALLOW_RECODE_KFMAXBW);
sf->encode_breakout_thresh = 1;
+ sf->use_fast_coef_costing = 1;
if (speed == 1) {
sf->use_square_partition_only = !frame_is_intra_only(cm);
@@ -923,6 +927,7 @@ void vp9_set_speed_features(VP9_COMP *cpi) {
sf->use_uv_intra_rd_estimate = 0;
sf->use_fast_lpf_pick = 0;
sf->use_fast_coef_updates = 0;
+ sf->use_fast_coef_costing = 0;
sf->mode_skip_start = MAX_MODES; // Mode index at which mode skip mask set
sf->use_nonrd_pick_mode = 0;
sf->encode_breakout_thresh = 0;
diff --git a/vp9/encoder/vp9_onyx_int.h b/vp9/encoder/vp9_onyx_int.h
index 019cb13880..f1e5e3a68a 100644
--- a/vp9/encoder/vp9_onyx_int.h
+++ b/vp9/encoder/vp9_onyx_int.h
@@ -420,6 +420,10 @@ typedef struct {
// A binary mask indicating if NEARESTMV, NEARMV, ZEROMV, NEWMV
// modes are disabled in order from LSB to MSB for each BLOCK_SIZE.
int disable_inter_mode_mask[BLOCK_SIZES];
+
+ // This feature controls whether we do the expensive context update and
+ // calculation in the rd coefficient costing loop.
+ int use_fast_coef_costing;
} SPEED_FEATURES;
typedef struct {
diff --git a/vp9/encoder/vp9_rdopt.c b/vp9/encoder/vp9_rdopt.c
index 76683dae40..3b2d78889d 100644
--- a/vp9/encoder/vp9_rdopt.c
+++ b/vp9/encoder/vp9_rdopt.c
@@ -72,6 +72,7 @@ struct rdcost_block_args {
int64_t this_rd;
int64_t best_rd;
int skip;
+ int use_fast_coef_costing;
const scan_order *so;
};
@@ -549,12 +550,12 @@ static const int16_t band_counts[TX_SIZES][8] = {
{ 1, 2, 3, 4, 11, 256 - 21, 0 },
{ 1, 2, 3, 4, 11, 1024 - 21, 0 },
};
-
static INLINE int cost_coeffs(MACROBLOCK *x,
int plane, int block,
ENTROPY_CONTEXT *A, ENTROPY_CONTEXT *L,
TX_SIZE tx_size,
- const int16_t *scan, const int16_t *nb) {
+ const int16_t *scan, const int16_t *nb,
+ int use_fast_coef_costing) {
MACROBLOCKD *const xd = &x->e_mbd;
MB_MODE_INFO *mbmi = &xd->mi_8x8[0]->mbmi;
struct macroblock_plane *p = &x->plane[plane];
@@ -568,7 +569,6 @@ static INLINE int cost_coeffs(MACROBLOCK *x,
uint8_t *p_tok = x->token_cache;
int pt = combine_entropy_contexts(*A, *L);
int c, cost;
-
// Check for consistency of tx_size with mode info
assert(type == PLANE_TYPE_Y ? mbmi->tx_size == tx_size
: get_uv_tx_size(mbmi) == tx_size);
@@ -594,9 +594,13 @@ static INLINE int cost_coeffs(MACROBLOCK *x,
v = qcoeff[rc];
t = vp9_dct_value_tokens_ptr[v].token;
- pt = get_coef_context(nb, p_tok, c);
- cost += (*token_costs)[!prev_t][pt][t] + vp9_dct_value_cost_ptr[v];
- p_tok[rc] = vp9_pt_energy_class[t];
+ if (use_fast_coef_costing) {
+ cost += (*token_costs)[!prev_t][!prev_t][t] + vp9_dct_value_cost_ptr[v];
+ } else {
+ pt = get_coef_context(nb, p_tok, c);
+ cost += (*token_costs)[!prev_t][pt][t] + vp9_dct_value_cost_ptr[v];
+ p_tok[rc] = vp9_pt_energy_class[t];
+ }
prev_t = t;
if (!--band_left) {
band_left = *band_count++;
@@ -606,8 +610,12 @@ static INLINE int cost_coeffs(MACROBLOCK *x,
// eob token
if (band_left) {
- pt = get_coef_context(nb, p_tok, c);
- cost += (*token_costs)[0][pt][EOB_TOKEN];
+ if (use_fast_coef_costing) {
+ cost += (*token_costs)[0][!prev_t][EOB_TOKEN];
+ } else {
+ pt = get_coef_context(nb, p_tok, c);
+ cost += (*token_costs)[0][pt][EOB_TOKEN];
+ }
}
}
@@ -616,7 +624,6 @@ static INLINE int cost_coeffs(MACROBLOCK *x,
return cost;
}
-
static void dist_block(int plane, int block, TX_SIZE tx_size,
struct rdcost_block_args* args) {
const int ss_txfrm_size = tx_size << 1;
@@ -648,7 +655,8 @@ static void rate_block(int plane, int block, BLOCK_SIZE plane_bsize,
args->rate = cost_coeffs(args->x, plane, block, args->t_above + x_idx,
args->t_left + y_idx, tx_size,
- args->so->scan, args->so->neighbors);
+ args->so->scan, args->so->neighbors,
+ args->use_fast_coef_costing);
}
static void block_rd_txfm(int plane, int block, BLOCK_SIZE plane_bsize,
@@ -732,12 +740,14 @@ static void txfm_rd_in_plane(MACROBLOCK *x,
int *rate, int64_t *distortion,
int *skippable, int64_t *sse,
int64_t ref_best_rd, int plane,
- BLOCK_SIZE bsize, TX_SIZE tx_size) {
+ BLOCK_SIZE bsize, TX_SIZE tx_size,
+ int use_fast_coef_casting) {
MACROBLOCKD *const xd = &x->e_mbd;
struct macroblockd_plane *const pd = &xd->plane[plane];
struct rdcost_block_args args = { 0 };
args.x = x;
args.best_rd = ref_best_rd;
+ args.use_fast_coef_costing = use_fast_coef_casting;
if (plane == 0)
xd->mi_8x8[0]->mbmi.tx_size = tx_size;
@@ -776,7 +786,7 @@ static void choose_largest_txfm_size(VP9_COMP *cpi, MACROBLOCK *x,
txfm_rd_in_plane(x, rate, distortion, skip,
&sse[mbmi->tx_size], ref_best_rd, 0, bs,
- mbmi->tx_size);
+ mbmi->tx_size, cpi->sf.use_fast_coef_costing);
cpi->tx_stepdown_count[0]++;
}
@@ -920,7 +930,8 @@ static void choose_txfm_size_from_modelrd(VP9_COMP *cpi, MACROBLOCK *x,
// Actually encode using the chosen mode if a model was used, but do not
// update the r, d costs
txfm_rd_in_plane(x, rate, distortion, skip,
- &sse[mbmi->tx_size], ref_best_rd, 0, bs, mbmi->tx_size);
+ &sse[mbmi->tx_size], ref_best_rd, 0, bs, mbmi->tx_size,
+ cpi->sf.use_fast_coef_costing);
if (max_tx_size == TX_32X32 && best_tx == TX_32X32) {
cpi->tx_stepdown_count[0]++;
@@ -968,7 +979,8 @@ static void inter_super_block_yrd(VP9_COMP *cpi, MACROBLOCK *x, int *rate,
for (tx_size = TX_4X4; tx_size <= max_tx_size; ++tx_size)
txfm_rd_in_plane(x, &r[tx_size][0], &d[tx_size],
&s[tx_size], &sse[tx_size],
- ref_best_rd, 0, bs, tx_size);
+ ref_best_rd, 0, bs, tx_size,
+ cpi->sf.use_fast_coef_costing);
choose_txfm_size_from_rd(cpi, x, r, rate, d, distortion, s,
skip, txfm_cache, bs);
}
@@ -997,7 +1009,8 @@ static void intra_super_block_yrd(VP9_COMP *cpi, MACROBLOCK *x, int *rate,
for (tx_size = TX_4X4; tx_size <= max_txsize_lookup[bs]; ++tx_size)
txfm_rd_in_plane(x, &r[tx_size][0], &d[tx_size],
&s[tx_size], &sse[tx_size],
- ref_best_rd, 0, bs, tx_size);
+ ref_best_rd, 0, bs, tx_size,
+ cpi->sf.use_fast_coef_costing);
choose_txfm_size_from_rd(cpi, x, r, rate, d, distortion, s,
skip, txfm_cache, bs);
}
@@ -1100,7 +1113,8 @@ static int64_t rd_pick_intra4x4block(VP9_COMP *cpi, MACROBLOCK *x, int ib,
vp9_fwht4x4(src_diff, coeff, 8);
vp9_regular_quantize_b_4x4(x, 0, block, so->scan, so->iscan);
ratey += cost_coeffs(x, 0, block, tempa + idx, templ + idy, TX_4X4,
- so->scan, so->neighbors);
+ so->scan, so->neighbors,
+ cpi->sf.use_fast_coef_costing);
if (RDCOST(x->rdmult, x->rddiv, ratey, distortion) >= best_rd)
goto next;
vp9_iwht4x4_add(BLOCK_OFFSET(pd->dqcoeff, block), dst, dst_stride,
@@ -1112,7 +1126,8 @@ static int64_t rd_pick_intra4x4block(VP9_COMP *cpi, MACROBLOCK *x, int ib,
vp9_fht4x4(src_diff, coeff, 8, tx_type);
vp9_regular_quantize_b_4x4(x, 0, block, so->scan, so->iscan);
ratey += cost_coeffs(x, 0, block, tempa + idx, templ + idy, TX_4X4,
- so->scan, so->neighbors);
+ so->scan, so->neighbors,
+ cpi->sf.use_fast_coef_costing);
distortion += vp9_block_error(coeff, BLOCK_OFFSET(pd->dqcoeff, block),
16, &unused) >> 2;
if (RDCOST(x->rdmult, x->rddiv, ratey, distortion) >= best_rd)
@@ -1296,7 +1311,7 @@ static int64_t rd_pick_intra_sby_mode(VP9_COMP *cpi, MACROBLOCK *x,
return best_rd;
}
-static void super_block_uvrd(MACROBLOCK *x,
+static void super_block_uvrd(const VP9_COMP *cpi, MACROBLOCK *x,
int *rate, int64_t *distortion, int *skippable,
int64_t *sse, BLOCK_SIZE bsize,
int64_t ref_best_rd) {
@@ -1323,7 +1338,8 @@ static void super_block_uvrd(MACROBLOCK *x,
for (plane = 1; plane < MAX_MB_PLANE; ++plane) {
txfm_rd_in_plane(x, &pnrate, &pndist, &pnskip, &pnsse,
- ref_best_rd, plane, bsize, uv_txfm_size);
+ ref_best_rd, plane, bsize, uv_txfm_size,
+ cpi->sf.use_fast_coef_costing);
if (pnrate == INT_MAX)
goto term;
*rate += pnrate;
@@ -1359,7 +1375,7 @@ static int64_t rd_pick_intra_sbuv_mode(VP9_COMP *cpi, MACROBLOCK *x,
xd->mi_8x8[0]->mbmi.uv_mode = mode;
- super_block_uvrd(x, &this_rate_tokenonly,
+ super_block_uvrd(cpi, x, &this_rate_tokenonly,
&this_distortion, &s, &this_sse, bsize, best_rd);
if (this_rate_tokenonly == INT_MAX)
continue;
@@ -1402,14 +1418,15 @@ static int64_t rd_pick_intra_sbuv_mode(VP9_COMP *cpi, MACROBLOCK *x,
return best_rd;
}
-static int64_t rd_sbuv_dcpred(const VP9_COMMON *cm, MACROBLOCK *x,
+static int64_t rd_sbuv_dcpred(const VP9_COMP *cpi, MACROBLOCK *x,
int *rate, int *rate_tokenonly,
int64_t *distortion, int *skippable,
BLOCK_SIZE bsize) {
+ const VP9_COMMON *cm = &cpi->common;
int64_t unused;
x->e_mbd.mi_8x8[0]->mbmi.uv_mode = DC_PRED;
- super_block_uvrd(x, rate_tokenonly, distortion,
+ super_block_uvrd(cpi, x, rate_tokenonly, distortion,
skippable, &unused, bsize, INT64_MAX);
*rate = *rate_tokenonly + x->intra_uv_mode_cost[cm->frame_type][DC_PRED];
return RDCOST(x->rdmult, x->rddiv, *rate, *distortion);
@@ -1425,7 +1442,7 @@ static void choose_intra_uv_mode(VP9_COMP *cpi, PICK_MODE_CONTEXT *ctx,
// Use an estimated rd for uv_intra based on DC_PRED if the
// appropriate speed flag is set.
if (cpi->sf.use_uv_intra_rd_estimate) {
- rd_sbuv_dcpred(&cpi->common, x, rate_uv, rate_uv_tokenonly, dist_uv,
+ rd_sbuv_dcpred(cpi, x, rate_uv, rate_uv_tokenonly, dist_uv,
skip_uv, bsize < BLOCK_8X8 ? BLOCK_8X8 : bsize);
// Else do a proper rd search for each possible transform size that may
// be considered in the main rd loop.
@@ -1588,7 +1605,8 @@ static int64_t encode_inter_mb_segment(VP9_COMP *cpi,
16, &ssz);
thissse += ssz;
thisrate += cost_coeffs(x, 0, k, ta + (k & 1), tl + (k >> 1), TX_4X4,
- so->scan, so->neighbors);
+ so->scan, so->neighbors,
+ cpi->sf.use_fast_coef_costing);
rd1 = RDCOST(x->rdmult, x->rddiv, thisrate, thisdistortion >> 2);
rd2 = RDCOST(x->rdmult, x->rddiv, 0, thissse >> 2);
rd = MIN(rd1, rd2);
@@ -3026,7 +3044,7 @@ static int64_t handle_inter_mode(VP9_COMP *cpi, MACROBLOCK *x,
rdcosty = RDCOST(x->rdmult, x->rddiv, *rate2, *distortion);
rdcosty = MIN(rdcosty, RDCOST(x->rdmult, x->rddiv, 0, *psse));
- super_block_uvrd(x, rate_uv, distortion_uv, &skippable_uv, &sseuv,
+ super_block_uvrd(cpi, x, rate_uv, distortion_uv, &skippable_uv, &sseuv,
bsize, ref_best_rd - rdcosty);
if (*rate_uv == INT_MAX) {
*rate2 = INT_MAX;
@@ -4145,7 +4163,7 @@ int64_t vp9_rd_pick_inter_mode_sub8x8(VP9_COMP *cpi, MACROBLOCK *x,
// then dont bother looking at UV
vp9_build_inter_predictors_sbuv(&x->e_mbd, mi_row, mi_col,
BLOCK_8X8);
- super_block_uvrd(x, &rate_uv, &distortion_uv, &uv_skippable,
+ super_block_uvrd(cpi, x, &rate_uv, &distortion_uv, &uv_skippable,
&uv_sse, BLOCK_8X8, tmp_best_rdu);
if (rate_uv == INT_MAX)
continue;
--
GitLab