Commit 91d223bd authored by Ronald S. Bultje's avatar Ronald S. Bultje
Browse files

Some minor optimizations for cost_coeffs().

Cycle timings for first 3 frames of bus (speed 0) at 1500kbps:
4x4: 298 -> 234 cycles
8x8: 1227 -> 878 cycles
16x16: 23426 -> 18134 cycles
32x32: 4906 -> 3664 cycles

Total encode time of first 50 frames of bus @ 1500kbps (speed 0) goes
from 3min0.7 to 2min51.6 seconds, i.e. 5.3% faster.

Change-Id: I68a0e1b530b0563b84a67342cca4b45146077e95
parent af660715
......@@ -133,8 +133,7 @@ struct macroblock {
unsigned char *active_ptr;
// note that token_costs is the cost when eob node is skipped
vp9_coeff_count token_costs[TX_SIZE_MAX_SB][BLOCK_TYPES];
vp9_coeff_count token_costs_noskip[TX_SIZE_MAX_SB][BLOCK_TYPES];
vp9_coeff_count token_costs[TX_SIZE_MAX_SB][BLOCK_TYPES][2];
int optimize;
......
......@@ -223,11 +223,11 @@ static void optimize_b(VP9_COMMON *const cm, MACROBLOCK *mb,
pt = trellis_get_coeff_context(scan, nb, i, t0, token_cache,
pad, default_eob);
rate0 +=
mb->token_costs_noskip[tx_size][type][ref][band][pt]
[tokens[next][0].token];
mb->token_costs[tx_size][type][ref][0][band][pt]
[tokens[next][0].token];
rate1 +=
mb->token_costs_noskip[tx_size][type][ref][band][pt]
[tokens[next][1].token];
mb->token_costs[tx_size][type][ref][0][band][pt]
[tokens[next][1].token];
}
UPDATE_RD_COST();
/* And pick the best. */
......@@ -275,22 +275,14 @@ static void optimize_b(VP9_COMMON *const cm, MACROBLOCK *mb,
if (t0 != DCT_EOB_TOKEN) {
pt = trellis_get_coeff_context(scan, nb, i, t0, token_cache,
pad, default_eob);
if (!x)
rate0 += mb->token_costs[tx_size][type][ref][band][pt][
tokens[next][0].token];
else
rate0 += mb->token_costs_noskip[tx_size][type][ref][band][pt][
tokens[next][0].token];
rate0 += mb->token_costs[tx_size][type][ref][!x][band][pt]
[tokens[next][0].token];
}
if (t1 != DCT_EOB_TOKEN) {
pt = trellis_get_coeff_context(scan, nb, i, t1, token_cache,
pad, default_eob);
if (!x)
rate1 += mb->token_costs[tx_size][type][ref][band][pt][
tokens[next][1].token];
else
rate1 += mb->token_costs_noskip[tx_size][type][ref][band][pt][
tokens[next][1].token];
rate1 += mb->token_costs[tx_size][type][ref][!x][band][pt]
[tokens[next][1].token];
}
}
......@@ -322,12 +314,12 @@ static void optimize_b(VP9_COMMON *const cm, MACROBLOCK *mb,
/* Update the cost of each path if we're past the EOB token. */
if (t0 != DCT_EOB_TOKEN) {
tokens[next][0].rate +=
mb->token_costs[tx_size][type][ref][band][0][t0];
mb->token_costs[tx_size][type][ref][1][band][0][t0];
tokens[next][0].token = ZERO_TOKEN;
}
if (t1 != DCT_EOB_TOKEN) {
tokens[next][1].rate +=
mb->token_costs[tx_size][type][ref][band][0][t1];
mb->token_costs[tx_size][type][ref][1][band][0][t1];
tokens[next][1].token = ZERO_TOKEN;
}
/* Don't update next, because we didn't add a new node. */
......@@ -343,8 +335,8 @@ static void optimize_b(VP9_COMMON *const cm, MACROBLOCK *mb,
error1 = tokens[next][1].error;
t0 = tokens[next][0].token;
t1 = tokens[next][1].token;
rate0 += mb->token_costs_noskip[tx_size][type][ref][band][pt][t0];
rate1 += mb->token_costs_noskip[tx_size][type][ref][band][pt][t1];
rate0 += mb->token_costs[tx_size][type][ref][0][band][pt][t0];
rate1 += mb->token_costs[tx_size][type][ref][0][band][pt][t1];
UPDATE_RD_COST();
best = rd_cost1 < rd_cost0;
final_eob = i0 - 1;
......
......@@ -116,8 +116,7 @@ static int rd_thresh_block_size_factor[BLOCK_SIZE_TYPES] =
#define MAX_RD_THRESH_FREQ_FACT 32
#define MAX_RD_THRESH_FREQ_INC 1
static void fill_token_costs(vp9_coeff_count (*c)[BLOCK_TYPES],
vp9_coeff_count (*cnoskip)[BLOCK_TYPES],
static void fill_token_costs(vp9_coeff_count (*c)[BLOCK_TYPES][2],
vp9_coeff_probs_model (*p)[BLOCK_TYPES]) {
int i, j, k, l;
TX_SIZE t;
......@@ -128,18 +127,18 @@ static void fill_token_costs(vp9_coeff_count (*c)[BLOCK_TYPES],
for (l = 0; l < PREV_COEF_CONTEXTS; l++) {
vp9_prob probs[ENTROPY_NODES];
vp9_model_to_full_probs(p[t][i][j][k][l], probs);
vp9_cost_tokens((int *)cnoskip[t][i][j][k][l], probs,
vp9_cost_tokens((int *)c[t][i][j][0][k][l], probs,
vp9_coef_tree);
#if CONFIG_BALANCED_COEFTREE
// Replace the eob node prob with a very small value so that the
// cost approximately equals the cost without the eob node
probs[1] = 1;
vp9_cost_tokens((int *)c[t][i][j][k][l], probs, vp9_coef_tree);
vp9_cost_tokens((int *)c[t][i][j][1][k][l], probs, vp9_coef_tree);
#else
vp9_cost_tokens_skip((int *)c[t][i][j][k][l], probs,
vp9_cost_tokens_skip((int *)c[t][i][j][1][k][l], probs,
vp9_coef_tree);
assert(c[t][i][j][k][l][DCT_EOB_TOKEN] ==
cnoskip[t][i][j][k][l][DCT_EOB_TOKEN]);
assert(c[t][i][j][0][k][l][DCT_EOB_TOKEN] ==
c[t][i][j][1][k][l][DCT_EOB_TOKEN]);
#endif
}
}
......@@ -260,9 +259,7 @@ void vp9_initialize_rd_consts(VP9_COMP *cpi, int qindex) {
}
}
fill_token_costs(cpi->mb.token_costs,
cpi->mb.token_costs_noskip,
cpi->common.fc.coef_probs);
fill_token_costs(cpi->mb.token_costs, cpi->common.fc.coef_probs);
for (i = 0; i < NUM_PARTITION_CONTEXTS; i++)
vp9_cost_tokens(cpi->mb.partition_cost[i],
......@@ -310,18 +307,13 @@ static INLINE int cost_coeffs(VP9_COMMON *const cm, MACROBLOCK *mb,
int cost = 0, pad;
const int *scan, *nb;
const int eob = xd->plane[plane].eobs[block];
const int16_t *qcoeff_ptr = BLOCK_OFFSET(xd->plane[plane].qcoeff,
block, 16);
const int16_t *qcoeff_ptr = BLOCK_OFFSET(xd->plane[plane].qcoeff, block, 16);
const int ref = mbmi->ref_frame[0] != INTRA_FRAME;
unsigned int (*token_costs)[PREV_COEF_CONTEXTS][MAX_ENTROPY_TOKENS] =
mb->token_costs[tx_size][type][ref];
unsigned int (*token_costs)[COEF_BANDS][PREV_COEF_CONTEXTS]
[MAX_ENTROPY_TOKENS] = mb->token_costs[tx_size][type][ref];
ENTROPY_CONTEXT above_ec, left_ec;
TX_TYPE tx_type = DCT_DCT;
const int segment_id = xd->mode_info_context->mbmi.segment_id;
unsigned int (*token_costs_noskip)[PREV_COEF_CONTEXTS][MAX_ENTROPY_TOKENS] =
mb->token_costs_noskip[tx_size][type][ref];
int seg_eob, default_eob;
uint8_t token_cache[1024];
const uint8_t * band_translate;
......@@ -390,26 +382,38 @@ static INLINE int cost_coeffs(VP9_COMMON *const cm, MACROBLOCK *mb,
if (eob < seg_eob)
assert(qcoeff_ptr[scan[eob]] == 0);
{
for (c = 0; c < eob; c++) {
int v = qcoeff_ptr[scan[c]];
int t = vp9_dct_value_tokens_ptr[v].token;
if (eob == 0) {
// single eob token
cost += token_costs[0][0][pt][DCT_EOB_TOKEN];
} else {
int t, v, prev_rc = 0;
// dc token
v = qcoeff_ptr[0];
t = vp9_dct_value_tokens_ptr[v].token;
cost += token_costs[0][0][pt][t] + vp9_dct_value_cost_ptr[v];
token_cache[0] = vp9_pt_energy_class[t];
// ac tokens
for (c = 1; c < eob; c++) {
const int rc = scan[c];
int band = get_coef_band(band_translate, c);
if (c)
pt = vp9_get_coef_context(scan, nb, pad, token_cache, c, default_eob);
if (!c || token_cache[scan[c - 1]]) // do not skip eob
cost += token_costs_noskip[band][pt][t] + vp9_dct_value_cost_ptr[v];
else
cost += token_costs[band][pt][t] + vp9_dct_value_cost_ptr[v];
token_cache[scan[c]] = vp9_pt_energy_class[t];
v = qcoeff_ptr[rc];
t = vp9_dct_value_tokens_ptr[v].token;
pt = vp9_get_coef_context(scan, nb, pad, token_cache, c, default_eob);
// as an index at some level
cost += token_costs[!token_cache[prev_rc]][band][pt][t] +
vp9_dct_value_cost_ptr[v];
token_cache[rc] = vp9_pt_energy_class[t];
prev_rc = rc;
}
// eob token
if (c < seg_eob) {
if (c)
pt = vp9_get_coef_context(scan, nb, pad, token_cache, c, default_eob);
cost += mb->token_costs_noskip[tx_size][type][ref]
[get_coef_band(band_translate, c)]
[pt][DCT_EOB_TOKEN];
pt = vp9_get_coef_context(scan, nb, pad, token_cache, c, default_eob);
cost += token_costs[0][get_coef_band(band_translate, c)][pt]
[DCT_EOB_TOKEN];
}
}
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment