Google Git
Sign in
aomedia / avm / refs/heads/main / . / av1 / encoder / trellis_quant.c
blob: 1aa1f80b8e552aaab2a60048f14f3494bdb7866b [file] [log] [blame] [edit]
/*
* Copyright (c) 2024, Alliance for Open Media. All rights reserved
*
* This source code is subject to the terms of the BSD 3-Clause Clear License
* and the Alliance for Open Media Patent License 1.0. If the BSD 3-Clause Clear
* License was not distributed with this source code in the LICENSE file, you
* can obtain it at aomedia.org/license/software-license/bsd-3-c-c/. If the
* Alliance for Open Media Patent License 1.0 was not distributed with this
* source code in the PATENTS file, you can obtain it at
* aomedia.org/license/patent-license/.
*/
#include "av1/encoder/trellis_quant.h"
#include "av1/encoder/encodetxb.h"
#include "aom_ports/mem.h"
#include "av1/common/blockd.h"
#include "av1/common/cost.h"
#include "av1/common/idct.h"
#include "av1/common/pred_common.h"
#include "av1/common/reconintra.h"
#include "av1/common/scan.h"
#include "av1/encoder/bitstream.h"
#include "av1/encoder/encodeframe.h"
#include "av1/encoder/rdopt.h"
#include "av1/encoder/tokenize.h"
typedef struct {
uint8_t *base;
int bufsize;
int idx;
} tcq_levels_t;
static void tcq_levels_init(tcq_levels_t *lev, uint8_t *mem_tcq, int bufsize) {
lev->base = mem_tcq;
lev->idx = 0;
lev->bufsize = bufsize;
}
static void tcq_levels_swap(tcq_levels_t *lev) { lev->idx ^= 1; }
static uint8_t *tcq_levels_prev(const tcq_levels_t *lev, int st) {
return &lev->base[(2 * st + lev->idx) * lev->bufsize];
}
static uint8_t *tcq_levels_cur(const tcq_levels_t *lev, int st) {
return &lev->base[(2 * st + !lev->idx) * lev->bufsize];
}
static AOM_INLINE void init_tcq_decision(tcq_node_t *decision) {
static const tcq_node_t def = { INT64_MAX >> 10, 0, -1, -2 };
for (int state = 0; state < TCQ_N_STATES; state++) {
memcpy(&decision[state], &def, sizeof(def));
}
}
// Initialize coeff neighbor magnitudes to zero before starting trellis
// optimization for a coeff block.
// Initialize previous state storage to identity mapping.
// As trellis decisions are made, magnitudes and
// state transitions will be updated.
static AOM_INLINE void init_tcq_ctx(struct tcq_ctx_t *tcq_ctx) {
static const int8_t init_st[4][TCQ_MAX_STATES] = {
{ 0, 1, 2, 3, 4, 5, 6, 7 },
{ 0, 1, 2, 3, 4, 5, 6, 7 },
{ 0, 1, 2, 3, 4, 5, 6, 7 },
{ 0, 1, 2, 3, 4, 5, 6, 7 },
};
memset(&tcq_ctx->mag_base, 0, sizeof(tcq_ctx->mag_base));
memset(&tcq_ctx->mag_mid, 0, sizeof(tcq_ctx->mag_mid));
memset(&tcq_ctx->ctx, 0, sizeof(tcq_ctx->ctx));
memset(&tcq_ctx->lev_new, 0, sizeof(tcq_ctx->lev_new));
int n_elem = sizeof(tcq_ctx->prev_st) / sizeof(tcq_ctx->prev_st[0]);
for (int i = 0; i < n_elem; i += 4) {
memcpy(tcq_ctx->prev_st[i], init_st, sizeof(init_st));
}
}
// Update context buffer for the current node
static AOM_INLINE void set_levels_buf(int prevId, int absLevel, uint8_t *levels,
const int16_t *scan, const int eob_minus1,
const int scan_pos, const int bwl,
const int sharpness) {
if (prevId == -2) {
return;
}
// update current abs level
levels[get_padded_idx(scan[scan_pos], bwl)] = AOMMIN(absLevel, INT8_MAX);
// check current node is a new start position? if so, set all previous
// position to 0. prevId == -1 means a new start, prevId == -2 ?
bool new_eob = prevId < 0 && scan_pos + 1 <= eob_minus1 && sharpness == 0;
if (new_eob) {
for (int si = scan_pos + 1; si <= eob_minus1; si++) {
levels[get_padded_idx(scan[si], bwl)] = 0;
}
}
}
static AOM_FORCE_INLINE int get_dqv(const int32_t *dequant, int coeff_idx,
const qm_val_t *iqmatrix) {
int dqv = dequant[!!coeff_idx];
if (iqmatrix != NULL)
dqv =
((iqmatrix[coeff_idx] * dqv) + (1 << (AOM_QM_BITS - 1))) >> AOM_QM_BITS;
return dqv;
}
static AOM_FORCE_INLINE int64_t get_coeff_dist(tran_low_t tcoeff,
tran_low_t dqcoeff, int shift) {
const int64_t diff = (tcoeff - dqcoeff) * (1 << shift);
const int64_t error = diff * diff;
return error;
}
static INLINE int get_coeff_cost_eob(int ci, tran_low_t abs_qc, int sign,
int coeff_ctx, int dc_sign_ctx,
const LV_MAP_COEFF_COST *txb_costs,
int bwl, TX_CLASS tx_class, int32_t t_sign,
int plane) {
int cost = 0;
const int row = ci >> bwl;
const int col = ci - (row << bwl);
int limits = get_lf_limits(row, col, tx_class, plane);
const int(*base_lf_eob_cost_ptr)[LF_BASE_SYMBOLS - 1] =
plane > 0 ? txb_costs->base_lf_eob_cost_uv : txb_costs->base_lf_eob_cost;
const int(*base_eob_cost_ptr)[3] =
plane > 0 ? txb_costs->base_eob_cost_uv : txb_costs->base_eob_cost;
cost += limits ? base_lf_eob_cost_ptr[coeff_ctx]
[AOMMIN(abs_qc, LF_BASE_SYMBOLS - 1) - 1]
: base_eob_cost_ptr[coeff_ctx][AOMMIN(abs_qc, 3) - 1];
if (abs_qc != 0) {
const int dc_ph_group = 0; // PH disabled
const bool dc_2dtx = (ci == 0);
const bool dc_hor = (col == 0) && tx_class == TX_CLASS_HORIZ;
const bool dc_ver = (row == 0) && tx_class == TX_CLASS_VERT;
if (dc_2dtx || dc_hor || dc_ver) {
if (plane == AOM_PLANE_V)
cost += txb_costs->v_dc_sign_cost[t_sign][dc_sign_ctx][sign];
else
cost += txb_costs->dc_sign_cost[dc_ph_group][dc_sign_ctx][sign];
} else {
cost += av1_cost_literal(1);
}
if (plane > 0) {
if (limits) {
if (abs_qc > LF_NUM_BASE_LEVELS) {
cost += get_br_lf_cost_tcq_uv(abs_qc);
}
} else {
if (abs_qc > NUM_BASE_LEVELS) {
int br_ctx = 0; /* get_br_ctx_eob_chroma */
cost += get_br_cost_tcq(abs_qc, txb_costs->lps_cost_uv[br_ctx]);
}
}
} else {
if (limits) {
if (abs_qc > LF_NUM_BASE_LEVELS) {
int br_ctx = get_br_ctx_lf_eob(ci, tx_class);
cost += get_br_lf_cost_tcq(abs_qc, txb_costs->lps_lf_cost[br_ctx]);
}
} else {
if (abs_qc > NUM_BASE_LEVELS) {
int br_ctx = 0; /* get_br_ctx_eob */
cost += get_br_cost_tcq(abs_qc, txb_costs->lps_cost[br_ctx]);
}
}
}
}
return cost;
}
static INLINE int get_coeff_cost_def(tran_low_t abs_qc, int coeff_ctx,
int diag_ctx, int plane,
const LV_MAP_COEFF_COST *txb_costs,
int q_i, int t_sign, int sign) {
(void)t_sign;
(void)sign;
int base_ctx = get_base_diag_ctx(diag_ctx) + get_base_ctx(coeff_ctx);
int mid_ctx = get_mid_ctx(coeff_ctx);
const int(*base_cost_ptr)[TCQ_CTXS][8] =
plane > 0 ? txb_costs->base_cost_uv : txb_costs->base_cost;
int cost = base_cost_ptr[base_ctx][q_i][AOMMIN(abs_qc, 3)];
if (abs_qc != 0) {
cost += av1_cost_literal(1);
if (abs_qc > NUM_BASE_LEVELS) {
if (plane == 0) {
cost += get_br_cost_tcq(abs_qc, txb_costs->lps_cost[mid_ctx]);
} else {
cost += get_br_cost_tcq(abs_qc, txb_costs->lps_cost_uv[mid_ctx]);
}
}
}
return cost;
}
static INLINE int get_coeff_cost_general(int ci, tran_low_t abs_qc, int sign,
int coeff_ctx, int mid_ctx,
int dc_sign_ctx,
const LV_MAP_COEFF_COST *txb_costs,
int bwl, TX_CLASS tx_class,
const int32_t *tmp_sign, int plane,
int limits, int q_i) {
int cost = 0;
const int(*base_lf_cost_ptr)[TCQ_CTXS][LF_BASE_SYMBOLS * 2] =
plane > 0 ? txb_costs->base_lf_cost_uv : txb_costs->base_lf_cost;
const int(*base_cost_ptr)[TCQ_CTXS][8] =
plane > 0 ? txb_costs->base_cost_uv : txb_costs->base_cost;
cost += limits ? base_lf_cost_ptr[coeff_ctx][q_i]
[AOMMIN(abs_qc, LF_BASE_SYMBOLS - 1)]
: base_cost_ptr[coeff_ctx][q_i][AOMMIN(abs_qc, 3)];
if (abs_qc != 0) {
const int dc_ph_group = 0; // PH disabled
const int row = ci >> bwl;
const int col = ci - (row << bwl);
const bool dc_2dtx = (ci == 0);
const bool dc_hor = (col == 0) && tx_class == TX_CLASS_HORIZ;
const bool dc_ver = (row == 0) && tx_class == TX_CLASS_VERT;
if (limits && (dc_2dtx || dc_hor || dc_ver)) {
if (plane == AOM_PLANE_V)
cost += txb_costs->v_dc_sign_cost[tmp_sign[ci]][dc_sign_ctx][sign];
else
cost += txb_costs->dc_sign_cost[dc_ph_group][dc_sign_ctx][sign];
} else {
cost += av1_cost_literal(1);
}
if (plane > 0) {
if (limits) {
if (abs_qc > LF_NUM_BASE_LEVELS) {
cost += get_br_lf_cost_tcq_uv(abs_qc);
}
} else {
if (abs_qc > NUM_BASE_LEVELS) {
cost += get_br_cost_tcq(abs_qc, txb_costs->lps_cost_uv[mid_ctx]);
}
}
} else {
if (limits) {
if (abs_qc > LF_NUM_BASE_LEVELS) {
cost += get_br_lf_cost_tcq(abs_qc, txb_costs->lps_lf_cost[mid_ctx]);
}
} else {
if (abs_qc > NUM_BASE_LEVELS) {
cost += get_br_cost_tcq(abs_qc, txb_costs->lps_cost[mid_ctx]);
}
}
}
}
return cost;
}
// Compare and update nodes info for current position, only used by pos eob - 1
static void update_node_eob(int64_t rdCost, int64_t distA, int64_t distB,
int64_t rdmult, int rateA, int rateB,
tran_low_t absA, tran_low_t absB, int limits,
int prev_rate, int prev_state,
tcq_node_t *decision_0, tcq_node_t *decision_1) {
(void)limits;
int parityA = 0;
int parityB = 1;
assert(parityA == tcq_parity(absA));
assert(parityB == tcq_parity(absB));
int64_t costA = rdCost + RDCOST(rdmult, rateA, distA);
int64_t costB = rdCost + RDCOST(rdmult, rateB, distB);
rateA += prev_rate;
rateB += prev_rate;
if (parityA) {
if (costA < decision_1->rdCost) {
decision_1->rdCost = costA;
decision_1->rate = rateA;
decision_1->prevId = prev_state;
decision_1->absLevel = absA;
}
} else {
if (costA < decision_0->rdCost) {
decision_0->rdCost = costA;
decision_0->rate = rateA;
decision_0->prevId = prev_state;
decision_0->absLevel = absA;
}
}
if (parityB) {
if (costB < decision_1->rdCost) {
decision_1->rdCost = costB;
decision_1->rate = rateB;
decision_1->prevId = prev_state;
decision_1->absLevel = absB;
}
} else {
if (costB < decision_0->rdCost) {
decision_0->rdCost = costB;
decision_0->rate = rateB;
decision_0->prevId = prev_state;
decision_0->absLevel = absB;
}
}
}
// Compare and update nodes info for current position
static void update_node_general(int64_t costA, int64_t costB, int64_t cost_zero,
int rateA, int rateB, int rate_zero,
tran_low_t absA, tran_low_t absB, int limits,
int prev_rate, int prev_state,
tcq_node_t *decision_0,
tcq_node_t *decision_1) {
assert(tcq_parity(absA) == 0);
assert(tcq_parity(absB) == 1);
(void)limits;
int even_bias = 1;
rateA += prev_rate;
rateB += prev_rate;
rate_zero += prev_rate;
if (cost_zero < costA && cost_zero < decision_0->rdCost + even_bias) {
decision_0->rdCost = cost_zero;
decision_0->rate = rate_zero;
decision_0->prevId = prev_state;
decision_0->absLevel = 0;
} else if (costA < decision_0->rdCost + even_bias) {
decision_0->rdCost = costA;
decision_0->rate = rateA;
decision_0->prevId = prev_state;
decision_0->absLevel = absA;
}
if (costB < decision_1->rdCost) {
decision_1->rdCost = costB;
decision_1->rate = rateB;
decision_1->prevId = prev_state;
decision_1->absLevel = absB;
}
}
// Evaluate NEW_EOB at the current position
static void decide_eob(int64_t costA, int64_t costB, int rateA, int rateB,
tran_low_t absA, tran_low_t absB, tcq_node_t *decision_0,
tcq_node_t *decision_1) {
if (costA < decision_0->rdCost) {
decision_0->rdCost = costA;
decision_0->rate = rateA;
decision_0->prevId = -1;
decision_0->absLevel = absA;
}
if (costB < decision_1->rdCost) {
decision_1->rdCost = costB;
decision_1->rate = rateB;
decision_1->prevId = -1;
decision_1->absLevel = absB;
}
}
// Populate the trellis from current position to next
void av1_decide_states_c(const struct tcq_node_t *prev,
const struct tcq_rate_t *rd,
const struct prequant_t *pq, int limits, int try_eob,
int64_t rdmult, struct tcq_node_t *decision) {
const int32_t *rate = rd->rate;
const int32_t *rate_zero = rd->rate_zero;
const int32_t *rate_eob = rd->rate_eob;
int64_t rdCost[2 * TCQ_MAX_STATES];
int64_t rdCost_zero[TCQ_MAX_STATES];
int64_t rdCost_eob[2];
// Init to 0 to avoid ASAN uninitialization warnings
memset(rdCost, 0, sizeof(rdCost));
memset(rdCost_zero, 0, sizeof(rdCost_zero));
init_tcq_decision(decision);
for (int i = 0; i < TCQ_N_STATES; i++) {
int a0 = tcq_quant(i);
int a1 = a0 + 2;
int64_t dist0 = pq->deltaDist[a0];
int64_t dist1 = pq->deltaDist[a1];
rdCost[2 * i] = prev[i].rdCost + RDCOST(rdmult, rate[2 * i], dist0);
rdCost[2 * i + 1] = prev[i].rdCost + RDCOST(rdmult, rate[2 * i + 1], dist1);
rdCost_zero[i] = prev[i].rdCost + RDCOST(rdmult, rate_zero[i], 0);
}
rdCost_eob[0] = RDCOST(rdmult, rate_eob[0], pq->deltaDist[0]);
rdCost_eob[1] = RDCOST(rdmult, rate_eob[1], pq->deltaDist[2]);
update_node_general(rdCost[0], rdCost[1], rdCost_zero[0], rate[0], rate[1],
rate_zero[0], pq->absLevel[0], pq->absLevel[2], limits,
prev[0].rate, 0, &decision[0], &decision[4]);
update_node_general(rdCost[2], rdCost[3], rdCost_zero[1], rate[2], rate[3],
rate_zero[1], pq->absLevel[0], pq->absLevel[2], limits,
prev[1].rate, 1, &decision[4], &decision[0]);
update_node_general(rdCost[5], rdCost[4], rdCost_zero[2], rate[5], rate[4],
rate_zero[2], pq->absLevel[3], pq->absLevel[1], limits,
prev[2].rate, 2, &decision[1], &decision[5]);
update_node_general(rdCost[7], rdCost[6], rdCost_zero[3], rate[7], rate[6],
rate_zero[3], pq->absLevel[3], pq->absLevel[1], limits,
prev[3].rate, 3, &decision[5], &decision[1]);
update_node_general(rdCost[8], rdCost[9], rdCost_zero[4], rate[8], rate[9],
rate_zero[4], pq->absLevel[0], pq->absLevel[2], limits,
prev[4].rate, 4, &decision[6], &decision[2]);
update_node_general(rdCost[10], rdCost[11], rdCost_zero[5], rate[10],
rate[11], rate_zero[5], pq->absLevel[0], pq->absLevel[2],
limits, prev[5].rate, 5, &decision[2], &decision[6]);
update_node_general(rdCost[13], rdCost[12], rdCost_zero[6], rate[13],
rate[12], rate_zero[6], pq->absLevel[3], pq->absLevel[1],
limits, prev[6].rate, 6, &decision[7], &decision[3]);
update_node_general(rdCost[15], rdCost[14], rdCost_zero[7], rate[15],
rate[14], rate_zero[7], pq->absLevel[3], pq->absLevel[1],
limits, prev[7].rate, 7, &decision[3], &decision[7]);
if (try_eob) {
const int state0 = 0;
const int state1 = 4;
decide_eob(rdCost_eob[0], rdCost_eob[1], rate_eob[0], rate_eob[1],
pq->absLevel[0], pq->absLevel[2], &decision[state0],
&decision[state1]);
}
}
// Prepare 4 quant candidates for each coeff
// [0] and [2] are for Q0. One is even and the other is odd
// [1] and [3] are for Q1. One is even and the other is odd
void av1_pre_quant_c(tran_low_t tqc, struct prequant_t *pqData,
const int32_t *quant_ptr, int dqv, int log_scale,
int scan_pos) {
// calculate qIdx
const int shift = 16 - log_scale + QUANT_FP_BITS;
tran_low_t add = -((2 << shift) >> 1);
tran_low_t abs_tqc = abs(tqc);
tran_low_t qIdx = (int)AOMMAX(
1, AOMMIN(((1 << 16) - 1),
((int64_t)abs_tqc * quant_ptr[scan_pos != 0] + add) >> shift));
pqData->qIdx = qIdx;
const int64_t dist0 = get_coeff_dist(abs_tqc, 0, log_scale - 1);
int Idx_a = qIdx & 3;
tran_low_t dqca = (tran_low_t)ROUND_POWER_OF_TWO_64((tran_high_t)qIdx * dqv,
QUANT_TABLE_BITS) >>
log_scale;
pqData->absLevel[Idx_a] = (++qIdx) >> 1;
pqData->deltaDist[Idx_a] =
get_coeff_dist(abs_tqc, dqca, log_scale - 1) - dist0;
int Idx_b = qIdx & 3;
tran_low_t dqcb = (tran_low_t)ROUND_POWER_OF_TWO_64((tran_high_t)qIdx * dqv,
QUANT_TABLE_BITS) >>
log_scale;
pqData->absLevel[Idx_b] = (++qIdx) >> 1;
pqData->deltaDist[Idx_b] =
get_coeff_dist(abs_tqc, dqcb, log_scale - 1) - dist0;
int Idx_c = qIdx & 3;
tran_low_t dqcc = (tran_low_t)ROUND_POWER_OF_TWO_64((tran_high_t)qIdx * dqv,
QUANT_TABLE_BITS) >>
log_scale;
pqData->absLevel[Idx_c] = (++qIdx) >> 1;
pqData->deltaDist[Idx_c] =
get_coeff_dist(abs_tqc, dqcc, log_scale - 1) - dist0;
int Idx_d = qIdx & 3;
tran_low_t dqcd = (tran_low_t)ROUND_POWER_OF_TWO_64((tran_high_t)qIdx * dqv,
QUANT_TABLE_BITS) >>
log_scale;
pqData->absLevel[Idx_d] = (++qIdx) >> 1;
pqData->deltaDist[Idx_d] =
get_coeff_dist(abs_tqc, dqcd, log_scale - 1) - dist0;
}
static int get_coeff_cost(int ci, tran_low_t abs_qc, int sign, int coeff_ctx,
int mid_ctx, int dc_sign_ctx,
const LV_MAP_COEFF_COST *txb_costs, int bwl,
TX_CLASS tx_class, const int32_t *tmp_sign, int plane,
int limits, int q_i) {
return get_coeff_cost_general(ci, abs_qc, sign, coeff_ctx, mid_ctx,
dc_sign_ctx, txb_costs, bwl, tx_class, tmp_sign,
plane, limits, q_i);
}
// Update neighbor coeff magnitudes state for current diagonal.
void av1_update_nbr_diagonal_c(struct tcq_ctx_t *tcq_ctx, int row, int col,
int bwl) {
// Temp storage for next diagonal ctx, with padding.
uint8_t next_base_mag[32 + 8][TCQ_MAX_STATES];
uint8_t next_mid_mag[32 + 8][TCQ_MAX_STATES];
uint8_t(*next_base)[TCQ_MAX_STATES] = &next_base_mag[4];
uint8_t(*next_mid)[TCQ_MAX_STATES] = &next_mid_mag[4];
uint8_t(*mag_base)[TCQ_MAX_STATES] = &tcq_ctx->mag_base[4];
uint8_t(*mag_mid)[TCQ_MAX_STATES] = &tcq_ctx->mag_mid[4];
int idx_start = col;
int idx_end = 1 << bwl;
for (int st = 0; st < TCQ_MAX_STATES; st++) {
// Copy original coeff context from previous diagonal.
int orig_st = tcq_ctx->orig_st[st];
if (orig_st < 0) {
for (int i = 0; i < 32; i++) {
next_base[i][st] = 0;
next_mid[i][st] = 0;
}
} else {
for (int i = 0; i < 32; i++) {
next_base[i][st] = mag_base[i][orig_st];
next_mid[i][st] = mag_mid[i][orig_st];
}
}
}
memset(next_base_mag, 0, sizeof(next_base_mag[0]) * 4);
memset(next_mid_mag, 0, sizeof(next_mid_mag[0]) * 4);
memset(tcq_ctx->mag_base, 0, sizeof(tcq_ctx->mag_base));
memset(tcq_ctx->mag_mid, 0, sizeof(tcq_ctx->mag_mid));
int diag = row + col;
int max1 = diag < 5 ? 5 : 3;
int max2 = diag < 6 ? 5 : 3;
for (int st = 0; st < TCQ_MAX_STATES; st++) {
// Update neighbor magnitudes
int st1 = st;
for (int i = idx_start; st1 >= 0 && i < idx_end; i++) {
int lev = tcq_ctx->lev_new[i][st1];
st1 = tcq_ctx->prev_st[i][st1];
// Update base positions {1, 0}, {0, 1}
int base1 = AOMMIN(lev, max1);
next_base[i][st] += base1;
next_base[i - 1][st] += base1;
// Update mid positions {1, 0}, {0, 1}
next_mid[i][st] += lev;
next_mid[i - 1][st] += lev;
// Update base positions {2, 0}, {1. 1}, {0, 2}
int base2 = AOMMIN(lev, max2);
mag_base[i][st] += base2;
mag_base[i - 1][st] += base2;
mag_base[i - 2][st] += base2;
// Update mid position {1, 1}
mag_mid[i - 1][st] = lev;
}
}
// Calc next context info
memset(tcq_ctx->ctx, 0, sizeof(tcq_ctx->ctx));
static const int8_t max_tbl[6] = { 0, 8, 6, 4, 4, 4 };
int base_max = max_tbl[AOMMIN(diag, 5)];
for (int st = 0; st < TCQ_MAX_STATES; st++) {
for (int i = 0; i < 32; i++) {
int base_ctx = next_base[i][st];
int mid_ctx = next_mid[i][st];
base_ctx = AOMMIN((base_ctx + 1) >> 1, base_max);
mid_ctx = AOMMIN((mid_ctx + 1) >> 1, 6);
tcq_ctx->ctx[i][st] = (mid_ctx << 4) + base_ctx;
}
// Reset original state to prepare for next diagonal.
tcq_ctx->orig_st[st] = st;
}
}
// Process the first position
void trellis_first_pos(const tcq_param_t *p, int scan_pos,
tcq_levels_t *tcq_lev, tcq_ctx_t *tcq_ctx,
tcq_node_t *trellis) {
int plane = p->plane;
TX_SIZE tx_size = p->tx_size;
TX_CLASS tx_class = p->tx_class;
int log_scale = p->log_scale;
int64_t rdmult = p->rdmult;
int dc_sign_ctx = p->dc_sign_ctx;
const int16_t *scan = p->scan;
const int32_t *tmp_sign = p->tmp_sign;
const tran_low_t *qcoeff = p->qcoeff;
const tran_low_t *tcoeff = p->tcoeff;
const int32_t *quant = p->quant;
const int32_t *dequant = p->dequant;
const qm_val_t *iqmatrix = p->iqmatrix;
const uint16_t *block_eob_rate = p->block_eob_rate;
const LV_MAP_COEFF_COST *txb_costs = p->txb_costs;
const int bwl = get_txb_bwl(tx_size);
const int height = get_txb_high(tx_size);
int blk_pos = scan[scan_pos];
tcq_node_t *decision = &trellis[scan_pos << TCQ_N_STATES_LOG];
prequant_t pqData;
int tempdqv = get_dqv(dequant, scan[scan_pos], iqmatrix);
av1_pre_quant(tcoeff[blk_pos], &pqData, quant, tempdqv, log_scale, scan_pos);
// init state
init_tcq_decision(decision);
const int row = blk_pos >> bwl;
const int col = blk_pos - (row << bwl);
int limits = get_lf_limits(row, col, tx_class, plane);
// calculate rate distortion
// try to quantize first coeff to nzcoeff
int coeff_ctx = get_lower_levels_ctx_eob(bwl, height, scan_pos);
int eob_rate = block_eob_rate[scan_pos];
int t_sign = tmp_sign[blk_pos];
int rate_Q0_a =
get_coeff_cost_eob(blk_pos, pqData.absLevel[0], (qcoeff[blk_pos] < 0),
coeff_ctx, dc_sign_ctx, txb_costs, bwl, tx_class,
t_sign, plane) +
eob_rate;
int rate_Q0_b =
get_coeff_cost_eob(blk_pos, pqData.absLevel[2], (qcoeff[blk_pos] < 0),
coeff_ctx, dc_sign_ctx, txb_costs, bwl, tx_class,
t_sign, plane) +
eob_rate;
const int state0 = 0;
const int state1 = 4;
update_node_eob(0, pqData.deltaDist[0], pqData.deltaDist[2], rdmult,
rate_Q0_a, rate_Q0_b, pqData.absLevel[0], pqData.absLevel[2],
limits, 0, -1, &decision[state0], &decision[state1]);
if (tcq_ctx) {
for (int i = 0; i < TCQ_MAX_STATES; i++) {
tcq_ctx->orig_st[i] = i;
tcq_ctx->prev_st[col][i] = -1;
tcq_ctx->lev_new[col][i] = 0;
}
tcq_ctx->lev_new[col][state0] =
AOMMIN(decision[state0].absLevel, MAX_VAL_BR_CTX);
tcq_ctx->lev_new[col][state1] =
AOMMIN(decision[state1].absLevel, MAX_VAL_BR_CTX);
if ((col == 0 && row != 0) || row == height - 1) {
av1_update_nbr_diagonal(tcq_ctx, row, col, bwl);
}
}
if (tcq_lev) {
uint8_t *levels0 = tcq_levels_cur(tcq_lev, state0);
uint8_t *levels1 = tcq_levels_cur(tcq_lev, state1);
set_levels_buf(decision[state0].prevId, decision[state0].absLevel, levels0,
scan, scan_pos, scan_pos, bwl, 0);
set_levels_buf(decision[state1].prevId, decision[state1].absLevel, levels1,
scan, scan_pos, scan_pos, bwl, 0);
}
}
void av1_get_rate_dist_def_luma_c(const struct tcq_param_t *p,
const struct prequant_t *pq,
const struct tcq_coeff_ctx_t *coeff_ctx,
int blk_pos, int diag_ctx, int eob_rate,
struct tcq_rate_t *rd) {
const LV_MAP_COEFF_COST *txb_costs = p->txb_costs;
TX_CLASS tx_class = p->tx_class;
int bwl = p->bwl;
const int plane = 0;
const int t_sign = 0;
const int sign = 0;
const int dc_sign_ctx = 0;
const tran_low_t *absLevel = pq->absLevel;
for (int i = 0; i < TCQ_N_STATES; i++) {
int q_i = tcq_quant(i);
int a0 = q_i;
int a1 = a0 + 2;
int base_ctx =
get_base_diag_ctx(diag_ctx) + get_base_ctx(coeff_ctx->coef[i]);
int cost0 = get_coeff_cost_def(absLevel[a0], coeff_ctx->coef[i], diag_ctx,
plane, txb_costs, q_i, t_sign, sign);
int cost1 = get_coeff_cost_def(absLevel[a1], coeff_ctx->coef[i], diag_ctx,
plane, txb_costs, q_i, t_sign, sign);
rd->rate_zero[i] = txb_costs->base_cost[base_ctx][q_i][0];
rd->rate[2 * i] = cost0;
rd->rate[2 * i + 1] = cost1;
}
rd->rate_eob[0] =
eob_rate + get_coeff_cost_eob(blk_pos, absLevel[0], sign,
coeff_ctx->coef_eob, dc_sign_ctx, txb_costs,
bwl, tx_class, t_sign, plane);
rd->rate_eob[1] =
eob_rate + get_coeff_cost_eob(blk_pos, absLevel[2], sign,
coeff_ctx->coef_eob, dc_sign_ctx, txb_costs,
bwl, tx_class, t_sign, plane);
}
void av1_get_rate_dist_def_chroma_c(const struct LV_MAP_COEFF_COST *txb_costs,
const struct prequant_t *pq,
const struct tcq_coeff_ctx_t *coeff_ctx,
int blk_pos, int bwl, TX_CLASS tx_class,
int diag_ctx, int eob_rate, int plane,
int t_sign, int sign,
struct tcq_rate_t *rd) {
const tran_low_t *absLevel = pq->absLevel;
const int dc_sign_ctx = 0;
for (int i = 0; i < TCQ_N_STATES; i++) {
int q_i = tcq_quant(i);
int a0 = q_i;
int a1 = a0 + 2;
int base_ctx = (diag_ctx & 15) + (coeff_ctx->coef[i] & 15);
int cost0 = get_coeff_cost_def(absLevel[a0], coeff_ctx->coef[i], diag_ctx,
plane, txb_costs, q_i, t_sign, sign);
int cost1 = get_coeff_cost_def(absLevel[a1], coeff_ctx->coef[i], diag_ctx,
plane, txb_costs, q_i, t_sign, sign);
rd->rate_zero[i] = txb_costs->base_cost_uv[base_ctx][q_i][0];
rd->rate[2 * i] = cost0;
rd->rate[2 * i + 1] = cost1;
}
rd->rate_eob[0] =
eob_rate + get_coeff_cost_eob(blk_pos, absLevel[0], sign,
coeff_ctx->coef_eob, dc_sign_ctx, txb_costs,
bwl, tx_class, t_sign, plane);
rd->rate_eob[1] =
eob_rate + get_coeff_cost_eob(blk_pos, absLevel[2], sign,
coeff_ctx->coef_eob, dc_sign_ctx, txb_costs,
bwl, tx_class, t_sign, plane);
}
void av1_get_rate_dist_lf_luma_c(const struct tcq_param_t *p,
const struct prequant_t *pq,
const struct tcq_coeff_ctx_t *coeff_ctx,
int blk_pos, int diag_ctx, int eob_rate,
int coeff_sign, struct tcq_rate_t *rd) {
const tran_low_t *absLevel = pq->absLevel;
const LV_MAP_COEFF_COST *txb_costs = p->txb_costs;
const int32_t *tmp_sign = p->tmp_sign;
TX_CLASS tx_class = p->tx_class;
int bwl = p->bwl;
int dc_sign_ctx = p->dc_sign_ctx;
int t_sign = tmp_sign[blk_pos];
int plane = 0;
for (int i = 0; i < TCQ_N_STATES; i++) {
int q_i = tcq_quant(i);
int a0 = q_i;
int a1 = a0 + 2;
int base_ctx =
get_base_diag_ctx(diag_ctx) + get_base_ctx(coeff_ctx->coef[i]);
int mid_ctx = get_mid_diag_ctx(diag_ctx) + get_mid_ctx(coeff_ctx->coef[i]);
int cost0 = get_coeff_cost(blk_pos, absLevel[a0], coeff_sign, base_ctx,
mid_ctx, dc_sign_ctx, txb_costs, bwl, tx_class,
tmp_sign, plane, 1, q_i);
int cost1 = get_coeff_cost(blk_pos, absLevel[a1], coeff_sign, base_ctx,
mid_ctx, dc_sign_ctx, txb_costs, bwl, tx_class,
tmp_sign, plane, 1, q_i);
rd->rate_zero[i] = txb_costs->base_lf_cost[base_ctx][q_i][0];
rd->rate[2 * i] = cost0;
rd->rate[2 * i + 1] = cost1;
}
rd->rate_eob[0] =
eob_rate + get_coeff_cost_eob(blk_pos, absLevel[0], coeff_sign,
coeff_ctx->coef_eob, dc_sign_ctx, txb_costs,
bwl, tx_class, t_sign, plane);
rd->rate_eob[1] =
eob_rate + get_coeff_cost_eob(blk_pos, absLevel[2], coeff_sign,
coeff_ctx->coef_eob, dc_sign_ctx, txb_costs,
bwl, tx_class, t_sign, plane);
}
void av1_get_rate_dist_lf_chroma_c(const struct LV_MAP_COEFF_COST *txb_costs,
const struct prequant_t *pq,
const struct tcq_coeff_ctx_t *coeff_ctx,
int blk_pos, int diag_ctx, int eob_rate,
int dc_sign_ctx, const int32_t *tmp_sign,
int bwl, TX_CLASS tx_class, int plane,
int coeff_sign, struct tcq_rate_t *rd) {
const tran_low_t *absLevel = pq->absLevel;
int t_sign = tmp_sign[blk_pos];
for (int i = 0; i < TCQ_N_STATES; i++) {
int q_i = tcq_quant(i);
int a0 = q_i;
int a1 = a0 + 2;
int base_ctx =
get_base_diag_ctx(diag_ctx) + get_base_ctx(coeff_ctx->coef[i]);
int mid_ctx = get_mid_diag_ctx(diag_ctx) + get_mid_ctx(coeff_ctx->coef[i]);
int cost0 = get_coeff_cost(blk_pos, absLevel[a0], coeff_sign, base_ctx,
mid_ctx, dc_sign_ctx, txb_costs, bwl, tx_class,
tmp_sign, plane, 1, q_i);
int cost1 = get_coeff_cost(blk_pos, absLevel[a1], coeff_sign, base_ctx,
mid_ctx, dc_sign_ctx, txb_costs, bwl, tx_class,
tmp_sign, plane, 1, q_i);
rd->rate_zero[i] = txb_costs->base_lf_cost_uv[base_ctx][q_i][0];
rd->rate[2 * i] = cost0;
rd->rate[2 * i + 1] = cost1;
}
rd->rate_eob[0] =
eob_rate + get_coeff_cost_eob(blk_pos, absLevel[0], coeff_sign,
coeff_ctx->coef_eob, dc_sign_ctx, txb_costs,
bwl, tx_class, t_sign, plane);
rd->rate_eob[1] =
eob_rate + get_coeff_cost_eob(blk_pos, absLevel[2], coeff_sign,
coeff_ctx->coef_eob, dc_sign_ctx, txb_costs,
bwl, tx_class, t_sign, plane);
}
void av1_update_states_c(const tcq_node_t *decision, int col,
struct tcq_ctx_t *tcq_ctx) {
int8_t tmp_orig_st[TCQ_N_STATES];
memcpy(tmp_orig_st, tcq_ctx->orig_st, sizeof(tcq_ctx->orig_st));
for (int i = 0; i < TCQ_N_STATES; i++) {
int prevId = decision[i].prevId;
int absLevel = decision[i].absLevel;
tcq_ctx->lev_new[col][i] = AOMMIN(absLevel, MAX_VAL_BR_CTX);
tcq_ctx->prev_st[col][i] = prevId;
if (prevId < 0) {
tcq_ctx->orig_st[i] = -1;
} else {
tcq_ctx->orig_st[i] = tmp_orig_st[prevId];
}
}
}
void av1_get_coeff_ctx_c(const struct tcq_ctx_t *tcq_ctx, int col,
struct tcq_coeff_ctx_t *coeff_ctx) {
for (int i = 0; i < TCQ_N_STATES; i++) {
int orig_st = tcq_ctx->orig_st[i];
coeff_ctx->coef[i] = orig_st == -1 ? 0 : tcq_ctx->ctx[col][orig_st];
}
}
// Get diagonal context for 2D luma block
static AOM_INLINE int get_diag_ctx(int lf, int blk_pos, int scan_pos, int bwl) {
int diag_ctx;
if (lf) {
diag_ctx = get_nz_map_ctx_from_stats_lf(0, blk_pos, bwl, TX_CLASS_2D);
if (scan_pos > 0) {
diag_ctx += 7 << 8;
}
} else {
diag_ctx = get_nz_map_ctx_from_stats(0, blk_pos, bwl, TX_CLASS_2D, 0);
}
return diag_ctx;
}
// TCQ 8-state for a 2D luma block.
static void trellis_loop_diagonal_st8(const tcq_param_t *p, int scan_hi,
int scan_lo, tcq_ctx_t *tcq_ctx,
tcq_node_t *trellis) {
int plane = p->plane;
int log_scale = p->log_scale;
int try_eob = p->sharpness == 0;
int64_t rdmult = p->rdmult;
const int16_t *scan = p->scan;
const tran_low_t *tcoeff = p->tcoeff;
const int32_t *quant = p->quant;
const int32_t *dequant = p->dequant;
const qm_val_t *iqmatrix = p->iqmatrix;
const uint16_t *block_eob_rate = p->block_eob_rate;
int bwl = p->bwl;
int height = p->txb_height;
assert(plane == 0);
assert(p->tx_class == TX_CLASS_2D);
(void)plane;
int dc_coeff_sign = tcoeff[0] < 0;
int blk_pos_inc = (1 << bwl) - 1;
int blk_pos, row, col;
// Handle default region.
while (scan_hi >= 10) {
blk_pos = scan[scan_hi];
row = blk_pos >> bwl;
col = blk_pos - (row << bwl);
int inc = AOMMIN(height - 1 - row, col);
scan_lo = scan_hi - inc;
int lf = 0;
int diag_ctx = get_diag_ctx(lf, blk_pos, scan_lo, bwl);
assert(scan_lo >= 0);
for (int scan_pos = scan_hi; scan_pos >= scan_lo; scan_pos--) {
tcq_node_t *decision = &trellis[scan_pos << TCQ_N_STATES_LOG];
tcq_node_t *prev_decision = &decision[TCQ_N_STATES];
prequant_t pqData;
int tempdqv = get_dqv(dequant, scan[scan_pos], iqmatrix);
av1_pre_quant(tcoeff[blk_pos], &pqData, quant, tempdqv, log_scale,
scan_pos);
// Get coeff contexts
tcq_coeff_ctx_t coeff_ctx;
av1_get_coeff_ctx(tcq_ctx, col, &coeff_ctx);
coeff_ctx.coef_eob = get_lower_levels_ctx_eob(bwl, height, scan_pos);
int eob_rate = block_eob_rate[scan_pos];
// Calculate rate and distortion.
tcq_rate_t rd;
av1_get_rate_dist_def_luma(p, &pqData, &coeff_ctx, blk_pos, diag_ctx,
eob_rate, &rd);
av1_decide_states(prev_decision, &rd, &pqData, lf, try_eob, rdmult,
decision);
av1_update_states(decision, col, tcq_ctx);
blk_pos += blk_pos_inc;
col--;
row++;
}
av1_update_nbr_diagonal(tcq_ctx, row - 1, col + 1, bwl);
scan_hi = scan_lo - 1;
}
// Handle LF region.
while (scan_hi >= 0) {
blk_pos = scan[scan_hi];
row = blk_pos >> bwl;
col = blk_pos - (row << bwl);
int inc = AOMMIN(height - 1 - row, col);
scan_lo = scan_hi - inc;
int lf = 1;
int diag_ctx = get_diag_ctx(lf, blk_pos, scan_lo, bwl);
assert(scan_lo >= 0);
for (int scan_pos = scan_hi; scan_pos >= scan_lo; scan_pos--) {
tcq_node_t *decision = &trellis[scan_pos << TCQ_N_STATES_LOG];
tcq_node_t *prev_decision = &decision[TCQ_N_STATES];
prequant_t pqData;
int tempdqv = get_dqv(dequant, scan[scan_pos], iqmatrix);
av1_pre_quant(tcoeff[blk_pos], &pqData, quant, tempdqv, log_scale,
scan_pos);
// Get coeff contexts
tcq_coeff_ctx_t coeff_ctx;
av1_get_coeff_ctx(tcq_ctx, col, &coeff_ctx);
coeff_ctx.coef_eob = get_lower_levels_ctx_eob(bwl, height, scan_pos);
int eob_rate = block_eob_rate[scan_pos];
// Calculate rate and distortion.
tcq_rate_t rd;
av1_get_rate_dist_lf_luma(p, &pqData, &coeff_ctx, blk_pos, diag_ctx,
eob_rate, dc_coeff_sign, &rd);
av1_decide_states(prev_decision, &rd, &pqData, lf, try_eob, rdmult,
decision);
av1_update_states(decision, col, tcq_ctx);
blk_pos += blk_pos_inc;
col--;
row++;
}
if (scan_hi != 0) {
av1_update_nbr_diagonal(tcq_ctx, row - 1, col + 1, bwl);
}
scan_hi = scan_lo - 1;
}
}
// General TCQ 8-state, used by non 2D Luma
void trellis_loop(const tcq_param_t *p, int first_scan_pos, int scan_hi,
int scan_lo, tcq_levels_t *tcq_lev, tcq_node_t *trellis) {
int plane = p->plane;
TX_CLASS tx_class = p->tx_class;
int log_scale = p->log_scale;
int sharpness = p->sharpness;
int try_eob = sharpness == 0;
int64_t rdmult = p->rdmult;
int dc_sign_ctx = p->dc_sign_ctx;
const int16_t *scan = p->scan;
const int32_t *tmp_sign = p->tmp_sign;
const tran_low_t *tcoeff = p->tcoeff;
const int32_t *quant = p->quant;
const int32_t *dequant = p->dequant;
const qm_val_t *iqmatrix = p->iqmatrix;
const uint16_t *block_eob_rate = p->block_eob_rate;
const LV_MAP_COEFF_COST *txb_costs = p->txb_costs;
const int bwl = p->bwl;
const int height = p->txb_height;
for (int scan_pos = scan_hi; scan_pos >= scan_lo; scan_pos--) {
tcq_levels_swap(tcq_lev);
uint8_t *levels[TCQ_MAX_STATES];
uint8_t *prev_levels[TCQ_MAX_STATES];
for (int i = 0; i < TCQ_N_STATES; i++) {
prev_levels[i] = tcq_levels_prev(tcq_lev, i);
levels[i] = tcq_levels_cur(tcq_lev, i);
}
int blk_pos = scan[scan_pos];
int row = blk_pos >> bwl;
int col = blk_pos - (row << bwl);
int limits = get_lf_limits(row, col, tx_class, plane);
tcq_node_t *decision = &trellis[scan_pos << TCQ_N_STATES_LOG];
tcq_node_t *prd = &decision[TCQ_N_STATES];
prequant_t pqData;
int tempdqv = get_dqv(dequant, scan[scan_pos], iqmatrix);
av1_pre_quant(tcoeff[blk_pos], &pqData, quant, tempdqv, log_scale,
scan_pos);
// init state
init_tcq_decision(decision);
const int coeff_sign = tcoeff[blk_pos] < 0;
// calculate contexts
tcq_coeff_ctx_t coeff_ctx;
int eob_ctx = get_lower_levels_ctx_eob(bwl, height, scan_pos);
int eob_rate = block_eob_rate[scan_pos];
coeff_ctx.coef_eob = eob_ctx;
tcq_rate_t rd;
// Calculate contexts and rate distortion
if (limits) {
if (plane == 0) {
int base_diag_ctx =
get_nz_map_ctx_from_stats_lf(0, blk_pos, bwl, tx_class);
int mid_diag_ctx = 7 * (tx_class == TX_CLASS_2D ? blk_pos > 0
: tx_class == TX_CLASS_HORIZ ? col == 0
: row == 0);
for (int i = 0; i < TCQ_N_STATES; i++) {
int base_ctx =
get_lower_levels_lf_ctx(prev_levels[i], blk_pos, bwl, tx_class);
int br_ctx = get_br_lf_ctx(prev_levels[i], blk_pos, bwl, tx_class);
br_ctx -= mid_diag_ctx;
coeff_ctx.coef[i] = base_ctx - base_diag_ctx + (br_ctx << 4);
}
int diag_ctx = base_diag_ctx + (mid_diag_ctx << 8);
av1_get_rate_dist_lf_luma(p, &pqData, &coeff_ctx, blk_pos, diag_ctx,
eob_rate, coeff_sign, &rd);
} else {
int diag_ctx = get_nz_map_ctx_from_stats_lf_chroma(0, tx_class, plane);
for (int i = 0; i < TCQ_N_STATES; i++) {
int base_ctx = get_lower_levels_lf_ctx_chroma(prev_levels[i], blk_pos,
bwl, tx_class, plane);
coeff_ctx.coef[i] = base_ctx - diag_ctx;
}
av1_get_rate_dist_lf_chroma(txb_costs, &pqData, &coeff_ctx, blk_pos,
diag_ctx, eob_rate, dc_sign_ctx, tmp_sign,
bwl, tx_class, plane, coeff_sign, &rd);
}
} else {
if (plane == 0) {
int diag_ctx = get_nz_map_ctx_from_stats(0, blk_pos, bwl, tx_class, 0);
for (int i = 0; i < TCQ_N_STATES; i++) {
int base_ctx = get_lower_levels_ctx(prev_levels[i], blk_pos, bwl,
tx_class, plane);
int br_ctx = get_br_ctx(prev_levels[i], blk_pos, bwl, tx_class);
coeff_ctx.coef[i] = base_ctx - diag_ctx + (br_ctx << 4);
}
av1_get_rate_dist_def_luma(p, &pqData, &coeff_ctx, blk_pos, diag_ctx,
eob_rate, &rd);
} else {
int diag_ctx =
get_nz_map_ctx_from_stats_chroma(0, blk_pos, tx_class, plane);
for (int i = 0; i < TCQ_N_STATES; i++) {
int base_ctx = get_lower_levels_ctx_chroma(prev_levels[i], blk_pos,
bwl, tx_class, plane);
int br_ctx =
get_br_ctx_chroma(prev_levels[i], blk_pos, bwl, tx_class);
coeff_ctx.coef[i] = base_ctx - diag_ctx + (br_ctx << 4);
}
av1_get_rate_dist_def_chroma(txb_costs, &pqData, &coeff_ctx, blk_pos,
bwl, tx_class, diag_ctx, eob_rate, plane,
tmp_sign[blk_pos], coeff_sign, &rd);
}
}
av1_decide_states(prd, &rd, &pqData, limits, try_eob, rdmult, decision);
// copy corresponding context from previous level buffer
for (int state = 0; state < TCQ_N_STATES && scan_pos != first_scan_pos;
state++) {
int prevId = decision[state].prevId;
if (prevId >= 0)
memcpy(levels[state], prev_levels[prevId],
sizeof(uint8_t) * tcq_lev->bufsize);
}
// update levels_buf
for (int state = 0; state < TCQ_N_STATES && scan_pos != 0; state++) {
set_levels_buf(decision[state].prevId, decision[state].absLevel,
levels[state], scan, first_scan_pos, scan_pos, bwl,
sharpness);
}
}
}
// Pre-calculate eob bits (rate) for each EOB candidate position from 1
// to the initial eob location. Store rate in array block_eob_rate[],
// starting with index.
void av1_calc_block_eob_rate_c(struct macroblock *x, int plane, TX_SIZE tx_size,
int eob, uint16_t *block_eob_rate) {
const MACROBLOCKD *xd = &x->e_mbd;
const MB_MODE_INFO *mbmi = xd->mi[0];
const int is_inter = is_inter_block(mbmi, xd->tree_type);
const PLANE_TYPE plane_type = get_plane_type(plane);
const TX_SIZE txs_ctx = get_txsize_entropy_ctx(tx_size);
const CoeffCosts *coeff_costs = &x->coeff_costs;
const LV_MAP_COEFF_COST *txb_costs =
&coeff_costs->coeff_costs[txs_ctx][plane_type];
const int eob_multi_size = txsize_log2_minus4[tx_size];
const LV_MAP_EOB_COST *txb_eob_costs =
&coeff_costs->eob_costs[eob_multi_size][plane_type];
const int *tbl_eob_cost = txb_eob_costs->eob_cost[is_inter];
block_eob_rate[0] = tbl_eob_cost[0];
block_eob_rate[1] = tbl_eob_cost[1];
int scan_pos = 2;
int n_offset_bits = 0;
while (scan_pos < eob) {
int eob_pt_low = AOMMIN(2 + n_offset_bits, EOB_PT_INDEX_COUNT - 1);
int eob_pt_rate = tbl_eob_cost[eob_pt_low];
if (eob_multi_size == 4 && (eob_pt_low == EOB_PT_INDEX_COUNT - 1))
eob_pt_rate += av1_cost_literal(1);
else if (eob_multi_size > 4 && (eob_pt_low == EOB_PT_INDEX_COUNT - 1))
eob_pt_rate += av1_cost_literal(2);
for (int bit = 0; bit < 2; bit++) {
int eob_ctx = n_offset_bits;
int extra_bit_rate = txb_costs->eob_extra_cost[eob_ctx][bit];
int eob_rate =
eob_pt_rate + extra_bit_rate + av1_cost_literal(n_offset_bits);
for (int i = 0; i < (1 << n_offset_bits); i++) {
block_eob_rate[scan_pos++] = eob_rate;
}
}
n_offset_bits++;
}
}
// Determine the best quantization option for each coeff from DC to EOB
int av1_find_best_path_c(const struct tcq_node_t *trellis, const int16_t *scan,
const int32_t *dequant, const qm_val_t *iqmatrix,
const tran_low_t *tcoeff, int first_scan_pos,
int log_scale, tran_low_t *qcoeff, tran_low_t *dqcoeff,
int *min_rate, int64_t *min_cost) {
int64_t min_path_cost = INT64_MAX;
int trel_min_rate = 0;
int prev_id = -2;
for (int state = 0; state < TCQ_N_STATES; state++) {
const tcq_node_t *decision = &trellis[state];
if (decision->rdCost < min_path_cost) {
prev_id = state;
min_path_cost = decision->rdCost;
trel_min_rate = decision->rate;
}
}
// Backtrack to reconstruct qcoeff / dqcoeff blocks.
int scan_pos = 0;
if (!iqmatrix) {
int dqv = dequant[0];
int dqv_ac = dequant[1];
for (; prev_id >= 0; scan_pos++) {
const tcq_node_t *decision =
&trellis[(scan_pos << TCQ_N_STATES_LOG) + prev_id];
prev_id = decision->prevId;
int abs_level = decision->absLevel;
int blk_pos = scan[scan_pos];
int sign = -(tcoeff[blk_pos] < 0);
int q_i = prev_id >= 0 ? tcq_quant(prev_id) : 0;
int qc = (abs_level == 0) ? 0 : (2 * abs_level - q_i);
int dqc = (tran_low_t)ROUND_POWER_OF_TWO_64((tran_high_t)qc * dqv,
QUANT_TABLE_BITS) >>
log_scale;
qcoeff[blk_pos] = (abs_level ^ sign) - sign;
dqcoeff[blk_pos] = (dqc ^ sign) - sign;
dqv = dqv_ac;
}
} else {
for (; prev_id >= 0; scan_pos++) {
const tcq_node_t *decision =
&trellis[(scan_pos << TCQ_N_STATES_LOG) + prev_id];
prev_id = decision->prevId;
int abs_level = decision->absLevel;
int blk_pos = scan[scan_pos];
int sign = tcoeff[blk_pos] < 0;
qcoeff[blk_pos] = sign ? -abs_level : abs_level;
int dqv = get_dqv(dequant, blk_pos, iqmatrix);
int q_i = prev_id >= 0 ? tcq_quant(prev_id) : 0;
int qc = (abs_level == 0) ? 0 : (2 * abs_level - q_i);
int dqc = (tran_low_t)ROUND_POWER_OF_TWO_64((tran_high_t)qc * dqv,
QUANT_TABLE_BITS) >>
log_scale;
dqcoeff[blk_pos] = sign ? -dqc : dqc;
}
}
int eob = scan_pos;
for (; scan_pos <= first_scan_pos; scan_pos++) {
int blk_pos = scan[scan_pos];
qcoeff[blk_pos] = 0;
dqcoeff[blk_pos] = 0;
}
*min_rate = trel_min_rate;
*min_cost = min_path_cost;
return eob;
}
/*
Algorithem description:
Unlike regular scaler quantization, trellis coded quant has dependency on
already coded coeffs in the decoder side. In order to correctly build the
dependency, the encoder creates a trellis and implements dependency in it.
The encoder flow:
1. Start at the first candidate EOB position and proceed towards DC (coeff[0])
2. For each of the states, the RD cost is calculated for each state transition,
keeping the best option out of the 2 (or 3) incoming candidates at each state.
the state quantizer mapping and state transition table are shown below. In
addition, it checks NEW_EOB at each of the coeff position and the next state is
0/4 in this case.
3. After the last coeff is processed, pick the lowest RD cost out of states 0-3
and all_zero, back track until it reaches eob.
* State-quantizer mapping and state transition table *
---------------------------------------------------------
| Current state | Quantizer | Parity of Qk | Next state |
|--------------------------------------------------------
| | | Even | 0 |
| 0 | 0 |----------------------------
| | | Odd | 4 |
|--------------------------------------------------------
| | | Even | 4 |
| 1 | 0 |----------------------------
| | | Odd | 0 |
|--------------------------------------------------------
| | | Even | 1 |
| 2 | 1 |----------------------------
| | | Odd | 5 |
|--------------------------------------------------------
| | | Even | 5 |
| 3 | 1 |----------------------------
| | | Odd | 1 |
|--------------------------------------------------------
| | | Even | 6 |
| 4 | 0 |----------------------------
| | | Odd | 2 |
|--------------------------------------------------------
| | | Even | 2 |
| 5 | 0 |----------------------------
| | | Odd | 6 |
|--------------------------------------------------------
| | | Even | 7 |
| 6 | 1 |----------------------------
| | | Odd | 3 |
|--------------------------------------------------------
| | | Even | 3 |
| 7 | 1 |----------------------------
| | | Odd | 7 |
|--------------------------------------------------------
*/
int av1_trellis_quant(const struct AV1_COMP *cpi, MACROBLOCK *x, int plane,
int block, TX_SIZE tx_size, TX_TYPE tx_type,
CctxType cctx_type, const TXB_CTX *const txb_ctx,
int *rate_cost, int sharpness) {
MACROBLOCKD *xd = &x->e_mbd;
const struct macroblock_plane *p = &x->plane[plane];
const SCAN_ORDER *scan_order =
get_scan(tx_size, get_primary_tx_type(tx_type));
const int16_t *scan = scan_order->scan;
int eob = p->eobs[block];
const int32_t *dequant = p->dequant_QTX;
const int32_t *quant = p->quant_fp_QTX; // quant_QTX
const qm_val_t *iqmatrix =
av1_get_iqmatrix(&cpi->common.quant_params, xd, plane, tx_size, tx_type);
const int block_offset = BLOCK_OFFSET(block);
tran_low_t *qcoeff = p->qcoeff + block_offset;
tran_low_t *dqcoeff = p->dqcoeff + block_offset;
const tran_low_t *tcoeff = p->coeff + block_offset;
const CoeffCosts *coeff_costs = &x->coeff_costs;
// This function is not called if eob = 0.
assert(eob > 0);
const AV1_COMMON *cm = &cpi->common;
const PLANE_TYPE plane_type = get_plane_type(plane);
const TX_SIZE txs_ctx = get_txsize_entropy_ctx(tx_size);
const TX_CLASS tx_class = tx_type_to_class[get_primary_tx_type(tx_type)];
const MB_MODE_INFO *mbmi = xd->mi[0];
const int bwl = get_txb_bwl(tx_size);
const int width = get_txb_wide(tx_size);
const int height = get_txb_high(tx_size);
assert(width == (1 << bwl));
const int is_inter = is_inter_block(mbmi, xd->tree_type);
const int bob_code = p->bobs[block];
const int is_fsc = ((
#if CONFIG_FSC_RES_HLS
cm->seq_params.enable_fsc &&
#endif
xd->mi[0]->fsc_mode[xd->tree_type == CHROMA_PART] &&
plane == PLANE_TYPE_Y) ||
use_inter_fsc(&cpi->common, plane, tx_type, is_inter));
const LV_MAP_COEFF_COST *txb_costs =
&coeff_costs->coeff_costs[txs_ctx][plane_type];
const int rshift =
(sharpness +
(cpi->oxcf.q_cfg.aq_mode == VARIANCE_AQ && mbmi->segment_id < 4
? 7 - mbmi->segment_id
: 2) +
(cpi->oxcf.q_cfg.aq_mode != VARIANCE_AQ &&
cpi->oxcf.q_cfg.deltaq_mode == DELTA_Q_PERCEPTUAL &&
cm->delta_q_info.delta_q_present_flag && x->sb_energy_level < 0
? (3 - x->sb_energy_level)
: 0));
int64_t rdmult = av1_compute_rdmult_for_plane(
x->rdmult, plane_rd_mult[is_inter][plane_type], xd->bd, rshift);
int si = eob - 1;
// populate trellis
assert(si < MAX_TRELLIS);
tcq_node_t trellis[MAX_TRELLIS * TCQ_MAX_STATES];
// Precalc block eob rate.
uint16_t block_eob_rate[MAX_TRELLIS];
av1_calc_block_eob_rate(x, plane, tx_size, eob, block_eob_rate);
// Collect TCQ related parameters.
tcq_param_t param;
int log_scale = av1_get_tx_scale(tx_size) + 1;
param.plane = plane;
param.bwl = bwl;
param.txb_height = height;
param.tx_size = tx_size;
param.tx_class = tx_class;
param.sharpness = sharpness;
param.rdmult = rdmult;
param.log_scale = log_scale;
param.dc_sign_ctx = txb_ctx->dc_sign_ctx;
param.scan = scan;
param.tmp_sign = xd->tmp_sign;
param.qcoeff = qcoeff;
param.tcoeff = tcoeff;
param.quant = quant;
param.dequant = dequant;
param.iqmatrix = iqmatrix;
param.block_eob_rate = block_eob_rate;
param.txb_costs = txb_costs;
// Start of TCQ
int first_scan_pos = si;
int scan_hi = first_scan_pos - 1;
int is_luma_2d = plane == 0 && tx_class == TX_CLASS_2D;
// Use faster path for 2D-luma blocks.
// Otherwise, use generic trellis code.
if (is_luma_2d) {
// Buffers for diagonal contexts.
tcq_ctx_t tcq_ctx;
init_tcq_ctx(&tcq_ctx);
// Process the first position
trellis_first_pos(&param, first_scan_pos, 0, &tcq_ctx, trellis);
// Speed-up version for 2D Luma by exploiting parallelism
// Process coeffs diagonal-by-diagonal.
if (scan_hi >= 0) {
trellis_loop_diagonal_st8(&param, scan_hi, 0, &tcq_ctx, trellis);
}
} else {
// Coeff level buffers.
int bufsize = (width + 4) * (height + 4) + TX_PAD_END;
int mem_tcq_sz = sizeof(uint8_t) * bufsize * (2 << TCQ_N_STATES_LOG);
uint8_t *mem_tcq = (uint8_t *)malloc(mem_tcq_sz);
if (!mem_tcq) {
exit(1);
}
if (eob > 1) {
memset(mem_tcq, 0, mem_tcq_sz);
}
tcq_levels_t tcq_lev;
tcq_levels_init(&tcq_lev, mem_tcq, bufsize);
// Process the first position
trellis_first_pos(&param, first_scan_pos, &tcq_lev, 0, trellis);
if (scan_hi >= 0) {
// Generic trellis loop
trellis_loop(&param, first_scan_pos, scan_hi, 0, &tcq_lev, trellis);
}
free(mem_tcq);
}
// find best path
int min_rate = 0;
int64_t min_path_cost = INT64_MAX;
eob = av1_find_best_path(trellis, scan, dequant, iqmatrix, tcoeff,
first_scan_pos, log_scale, qcoeff, dqcoeff,
&min_rate, &min_path_cost);
int txb_skip_ctx = txb_ctx->txb_skip_ctx;
int non_skip_cost = 0;
int skip_cost = 0;
if (plane == AOM_PLANE_V) {
txb_skip_ctx +=
(x->plane[AOM_PLANE_U].eobs[block] ? V_TXB_SKIP_CONTEXT_OFFSET : 0);
non_skip_cost = txb_costs->v_txb_skip_cost[txb_skip_ctx][0];
skip_cost = txb_costs->v_txb_skip_cost[txb_skip_ctx][1];
} else {
const int pred_mode_ctx =
(is_inter || mbmi->fsc_mode[xd->tree_type == CHROMA_PART]) ? 1 : 0;
non_skip_cost = txb_costs->txb_skip_cost[pred_mode_ctx][txb_skip_ctx][0];
skip_cost = txb_costs->txb_skip_cost[pred_mode_ctx][txb_skip_ctx][1];
}
int accu_rate = 0;
set_bob(x, plane, block, tx_size, tx_type);
if (eob == 0)
assert(0); // in current implementation, this could not happen.
else {
const int tx_type_cost = get_tx_type_cost(x, xd, plane, tx_size, tx_type,
cm->features.reduced_tx_set_used,
eob, bob_code, is_fsc);
int64_t rd_cost_skip = RDCOST(rdmult, skip_cost, 0);
accu_rate = non_skip_cost + tx_type_cost + min_rate;
int64_t rd_cost_coded =
min_path_cost +
(int64_t)RDCOST(rdmult, non_skip_cost + tx_type_cost, 0);
// skip block
if ((rd_cost_coded > rd_cost_skip) && sharpness == 0) {
for (int scan_idx = 0; scan_idx <= first_scan_pos; scan_idx++) {
int blk_idx = scan[scan_idx];
qcoeff[blk_idx] = 0;
dqcoeff[blk_idx] = 0;
}
accu_rate = skip_cost;
eob = 0;
}
}
p->eobs[block] = eob;
p->txb_entropy_ctx[block] =
av1_get_txb_entropy_context(qcoeff, scan_order, p->eobs[block]);
accu_rate += get_cctx_type_cost(cm, x, xd, plane, tx_size, block, cctx_type);
*rate_cost = accu_rate;
return eob;
}