mirror of
https://github.com/ultravideo/uvg266.git
synced 2024-11-30 12:44:07 +00:00
Fix data race
This commit is contained in:
parent
7a49cb60fb
commit
35d7e0d88b
|
@ -49,6 +49,9 @@ kvz_rc_data * kvz_get_rc_data(const encoder_control_t * const encoder) {
|
||||||
|
|
||||||
if (data == NULL) return NULL;
|
if (data == NULL) return NULL;
|
||||||
if (pthread_mutex_init(&data->ck_frame_lock, NULL) != 0) return NULL;
|
if (pthread_mutex_init(&data->ck_frame_lock, NULL) != 0) return NULL;
|
||||||
|
for (int (i) = 0; (i) < KVZ_MAX_GOP_LAYERS; ++(i)) {
|
||||||
|
if (pthread_rwlock_init(&data->ck_ctu_lock[i], NULL) != 0) return NULL;
|
||||||
|
}
|
||||||
|
|
||||||
const int num_lcus = encoder->in.width_in_lcu * encoder->in.height_in_lcu;
|
const int num_lcus = encoder->in.width_in_lcu * encoder->in.height_in_lcu;
|
||||||
|
|
||||||
|
@ -88,6 +91,9 @@ void kvz_free_rc_data() {
|
||||||
if (data == NULL) return;
|
if (data == NULL) return;
|
||||||
|
|
||||||
pthread_mutex_destroy(&data->ck_frame_lock);
|
pthread_mutex_destroy(&data->ck_frame_lock);
|
||||||
|
for (int i = 0; i < KVZ_MAX_GOP_LAYERS; ++i) {
|
||||||
|
pthread_rwlock_destroy(&data->ck_ctu_lock[i]);
|
||||||
|
}
|
||||||
|
|
||||||
if (data->intra_bpp) FREE_POINTER(data->intra_bpp);
|
if (data->intra_bpp) FREE_POINTER(data->intra_bpp);
|
||||||
if (data->intra_dis) FREE_POINTER(data->intra_dis);
|
if (data->intra_dis) FREE_POINTER(data->intra_dis);
|
||||||
|
@ -507,6 +513,7 @@ void kvz_estimate_pic_lambda(encoder_state_t * const state) {
|
||||||
temp_lambda = est_lambda;
|
temp_lambda = est_lambda;
|
||||||
double taylor_e3;
|
double taylor_e3;
|
||||||
int iteration_number = 0;
|
int iteration_number = 0;
|
||||||
|
pthread_rwlock_rdlock(&state->frame->new_ratecontrol->ck_ctu_lock[layer]);
|
||||||
do {
|
do {
|
||||||
taylor_e3 = 0.0;
|
taylor_e3 = 0.0;
|
||||||
best_lambda = temp_lambda = solve_cubic_equation(state->frame, 0, ctu_count, layer, temp_lambda, bits);
|
best_lambda = temp_lambda = solve_cubic_equation(state->frame, 0, ctu_count, layer, temp_lambda, bits);
|
||||||
|
@ -520,6 +527,7 @@ void kvz_estimate_pic_lambda(encoder_state_t * const state) {
|
||||||
iteration_number++;
|
iteration_number++;
|
||||||
}
|
}
|
||||||
while (fabs(taylor_e3 - bits) > 0.01 && iteration_number <= 11);
|
while (fabs(taylor_e3 - bits) > 0.01 && iteration_number <= 11);
|
||||||
|
pthread_rwlock_unlock(&state->frame->new_ratecontrol->ck_ctu_lock[layer]);
|
||||||
}
|
}
|
||||||
total_weight = calculate_weights(state, layer, ctu_count, best_lambda);
|
total_weight = calculate_weights(state, layer, ctu_count, best_lambda);
|
||||||
state->frame->remaining_weight = bits;
|
state->frame->remaining_weight = bits;
|
||||||
|
@ -590,13 +598,13 @@ static double get_ctu_bits(encoder_state_t * const state, vector2d_t pos) {
|
||||||
target_bits = MAX(target_bits + state->frame->cur_pic_target_bits - state->frame->cur_frame_bits_coded - (int)total_weight, 10);
|
target_bits = MAX(target_bits + state->frame->cur_pic_target_bits - state->frame->cur_frame_bits_coded - (int)total_weight, 10);
|
||||||
pthread_mutex_unlock(&state->frame->rc_lock);
|
pthread_mutex_unlock(&state->frame->rc_lock);
|
||||||
|
|
||||||
|
pthread_rwlock_rdlock(&state->frame->new_ratecontrol->ck_ctu_lock[layer]);
|
||||||
//just similar with the process at frame level, details can refer to the function TEncRCPic::kvz_estimate_pic_lambda
|
//just similar with the process at frame level, details can refer to the function TEncRCPic::kvz_estimate_pic_lambda
|
||||||
do {
|
do {
|
||||||
taylor_e3 = 0.0;
|
taylor_e3 = 0.0;
|
||||||
best_lambda = solve_cubic_equation(state->frame, index, last_ctu, layer, temp_lambda, target_bits);
|
best_lambda = solve_cubic_equation(state->frame, index, last_ctu, layer, temp_lambda, target_bits);
|
||||||
temp_lambda = best_lambda;
|
temp_lambda = best_lambda;
|
||||||
for (int i = index; i < last_ctu; i++) {
|
for (int i = index; i < last_ctu; i++) {
|
||||||
|
|
||||||
double CLCU = state->frame->new_ratecontrol->c_para[layer][i];
|
double CLCU = state->frame->new_ratecontrol->c_para[layer][i];
|
||||||
double KLCU = state->frame->new_ratecontrol->k_para[layer][i];
|
double KLCU = state->frame->new_ratecontrol->k_para[layer][i];
|
||||||
double a = -CLCU * KLCU / pow((double)state->frame->lcu_stats[i].pixels, KLCU - 1.0);
|
double a = -CLCU * KLCU / pow((double)state->frame->lcu_stats[i].pixels, KLCU - 1.0);
|
||||||
|
@ -608,6 +616,7 @@ static double get_ctu_bits(encoder_state_t * const state, vector2d_t pos) {
|
||||||
|
|
||||||
double c_ctu = state->frame->new_ratecontrol->c_para[layer][index];
|
double c_ctu = state->frame->new_ratecontrol->c_para[layer][index];
|
||||||
double k_ctu = state->frame->new_ratecontrol->k_para[layer][index];
|
double k_ctu = state->frame->new_ratecontrol->k_para[layer][index];
|
||||||
|
pthread_rwlock_unlock(&state->frame->new_ratecontrol->ck_ctu_lock[layer]);
|
||||||
double a = -c_ctu * k_ctu / pow(((double)state->frame->lcu_stats[index].pixels), k_ctu - 1.0);
|
double a = -c_ctu * k_ctu / pow(((double)state->frame->lcu_stats[index].pixels), k_ctu - 1.0);
|
||||||
double b = -1.0 / (k_ctu - 1.0);
|
double b = -1.0 / (k_ctu - 1.0);
|
||||||
|
|
||||||
|
@ -646,9 +655,11 @@ void kvz_set_ctu_qp_lambda(encoder_state_t * const state, vector2d_t pos) {
|
||||||
beta = state->frame->rc_beta;
|
beta = state->frame->rc_beta;
|
||||||
}
|
}
|
||||||
else {
|
else {
|
||||||
|
pthread_rwlock_rdlock(&state->frame->new_ratecontrol->ck_ctu_lock[layer]);
|
||||||
alpha = -state->frame->new_ratecontrol->c_para[layer][index] *
|
alpha = -state->frame->new_ratecontrol->c_para[layer][index] *
|
||||||
state->frame->new_ratecontrol->k_para[layer][index];
|
state->frame->new_ratecontrol->k_para[layer][index];
|
||||||
beta = state->frame->new_ratecontrol->k_para[layer][index] - 1;
|
beta = state->frame->new_ratecontrol->k_para[layer][index] - 1;
|
||||||
|
pthread_rwlock_unlock(&state->frame->new_ratecontrol->ck_ctu_lock[layer]);
|
||||||
}
|
}
|
||||||
|
|
||||||
double est_lambda;
|
double est_lambda;
|
||||||
|
@ -666,7 +677,7 @@ void kvz_set_ctu_qp_lambda(encoder_state_t * const state, vector2d_t pos) {
|
||||||
}
|
}
|
||||||
else {
|
else {
|
||||||
// In case wpp is used the previous ctus may not be ready from above rows
|
// In case wpp is used the previous ctus may not be ready from above rows
|
||||||
const int ctu_limit = encoder->cfg.wpp ? pos.y * encoder->in.width_in_lcu : 0;
|
// const int ctu_limit = encoder->cfg.wpp ? pos.y * encoder->in.width_in_lcu : 0;
|
||||||
|
|
||||||
est_lambda = alpha * pow(bpp, beta);
|
est_lambda = alpha * pow(bpp, beta);
|
||||||
const double clip_lambda = state->frame->lambda;
|
const double clip_lambda = state->frame->lambda;
|
||||||
|
@ -848,9 +859,25 @@ void kvz_update_after_picture(encoder_state_t * const state) {
|
||||||
state->frame->new_ratecontrol->previous_lambdas[layer] = lambda;
|
state->frame->new_ratecontrol->previous_lambdas[layer] = lambda;
|
||||||
|
|
||||||
update_pic_ck(state, pic_bpp, total_distortion, lambda, layer);
|
update_pic_ck(state, pic_bpp, total_distortion, lambda, layer);
|
||||||
|
if (state->frame->num <= 4 || state->frame->is_irap){
|
||||||
|
for (int i = 1; i < 5; ++i) {
|
||||||
|
pthread_rwlock_wrlock(&state->frame->new_ratecontrol->ck_ctu_lock[i]);
|
||||||
|
}
|
||||||
|
}
|
||||||
|
else{
|
||||||
|
pthread_rwlock_wrlock(&state->frame->new_ratecontrol->ck_ctu_lock[layer]);
|
||||||
|
}
|
||||||
for(int i = 0; i < state->encoder_control->in.width_in_lcu * state->encoder_control->in.height_in_lcu; i++) {
|
for(int i = 0; i < state->encoder_control->in.width_in_lcu * state->encoder_control->in.height_in_lcu; i++) {
|
||||||
update_ck(state, i, layer);
|
update_ck(state, i, layer);
|
||||||
}
|
}
|
||||||
|
if (state->frame->num <= 4 || state->frame->is_irap){
|
||||||
|
for (int i = 1; i < 5; ++i) {
|
||||||
|
pthread_rwlock_unlock(&state->frame->new_ratecontrol->ck_ctu_lock[i]);
|
||||||
|
}
|
||||||
|
}
|
||||||
|
else{
|
||||||
|
pthread_rwlock_unlock(&state->frame->new_ratecontrol->ck_ctu_lock[layer]);
|
||||||
|
}
|
||||||
}
|
}
|
||||||
|
|
||||||
|
|
||||||
|
|
|
@ -46,6 +46,7 @@ typedef struct kvz_rc_data {
|
||||||
double intra_alpha;
|
double intra_alpha;
|
||||||
double intra_beta;
|
double intra_beta;
|
||||||
|
|
||||||
|
pthread_rwlock_t ck_ctu_lock[KVZ_MAX_GOP_LAYERS];
|
||||||
pthread_mutex_t ck_frame_lock;
|
pthread_mutex_t ck_frame_lock;
|
||||||
} kvz_rc_data;
|
} kvz_rc_data;
|
||||||
|
|
||||||
|
|
Loading…
Reference in a new issue