Skip to content

O
OpenXG-RAN
Commit 192be604 authored by Florian Kaltenberger's avatar Florian Kaltenberger
Browse files
Options

further optim and fixes

parent c2d0eb7d
Hide whitespace changes
Inline Side-by-side
Showing with 29 additions and 22 deletions
openair1/PHY/LTE_TRANSPORT/dlsch_coding.c
View file @ 192be604
...@@ -61,6 +61,8 @@ ...@@ -61,6 +61,8 @@
extern int codingw; extern int codingw;
uint8_t segments_per_worker[9][3] = {{1,0,0},{2,0,0},{3,0,0},{2,4,0},{2,5,0},{3,6,0},{2,4,7},{2,5,8},{3,6,9}};
void free_eNB_dlsch(LTE_eNB_DLSCH_t *dlsch) void free_eNB_dlsch(LTE_eNB_DLSCH_t *dlsch)
{ {
int i; int i;
...@@ -288,7 +290,8 @@ int dlsch_encoding_2threads0(te_params *tep) { ...@@ -288,7 +290,8 @@ int dlsch_encoding_2threads0(te_params *tep) {
unsigned short iind; unsigned short iind;
unsigned short nb_rb = dlsch->harq_processes[harq_pid]->nb_rb; unsigned short nb_rb = dlsch->harq_processes[harq_pid]->nb_rb;
unsigned int Kr=0,Kr_bytes,r,r_offset=0,Kr_int=0; unsigned int Kr=0,Kr_bytes,r,r0,r_offset=0,Kr_int=0;
uint8_t *c_tmp[MAX_NUM_DLSCH_SEGMENTS];
uint8_t *d_tmp[MAX_NUM_DLSCH_SEGMENTS]; uint8_t *d_tmp[MAX_NUM_DLSCH_SEGMENTS];
// unsigned short m=dlsch->harq_processes[harq_pid]->mcs; // unsigned short m=dlsch->harq_processes[harq_pid]->mcs;
...@@ -298,7 +301,7 @@ int dlsch_encoding_2threads0(te_params *tep) { ...@@ -298,7 +301,7 @@ int dlsch_encoding_2threads0(te_params *tep) {
if (dlsch->harq_processes[harq_pid]->round == 0) { // this is a new packet if (dlsch->harq_processes[harq_pid]->round == 0) { // this is a new packet
#ifdef TD_DECODING #ifdef TD_DECODING
for (r=(dlsch->harq_processes[harq_pid]->C/(total_worker+1))*current_worker; r<(dlsch->harq_processes[harq_pid]->C/(total_worker+1))*(current_worker+1); r++) { for (r0=0,r=segments_per_worker[dlsch->harq_processes[harq_pid]->C-1][current_worker]; r<segments_per_worker[dlsch->harq_processes[harq_pid]->C][current_worker+1]; r0++,r++) {
if (r<dlsch->harq_processes[harq_pid]->Cminus) if (r<dlsch->harq_processes[harq_pid]->Cminus)
Kr = dlsch->harq_processes[harq_pid]->Kminus; Kr = dlsch->harq_processes[harq_pid]->Kminus;
...@@ -345,14 +348,15 @@ int dlsch_encoding_2threads0(te_params *tep) { ...@@ -345,14 +348,15 @@ int dlsch_encoding_2threads0(te_params *tep) {
else else
Kr_int = Kr; Kr_int = Kr;
for (r=(dlsch->harq_processes[harq_pid]->C/(total_worker+1))*current_worker; r<(dlsch->harq_processes[harq_pid]->C/(total_worker+1))*(current_worker+1); r++) { for (r0=0,r=segments_per_worker[dlsch->harq_processes[harq_pid]->C-1][current_worker]; r<segments_per_worker[dlsch->harq_processes[harq_pid]->C][current_worker+1]; r0++,r++) {
memset(dlsch->harq_processes[harq_pid]->d[r],0,(96+12+3+3*8448)*sizeof(uint8_t)); memset(dlsch->harq_processes[harq_pid]->d[r],0,(96+12+3+3*8448)*sizeof(uint8_t));
d_tmp[r] = &dlsch->harq_processes[harq_pid]->d[r][96]; c_tmp[r0] = &dlsch->harq_processes[harq_pid]->c[r][0];
d_tmp[r0] = &dlsch->harq_processes[harq_pid]->d[r][96];
} }
ldpc_encoder_optim_8seg(dlsch->harq_processes[harq_pid]->c,d_tmp,Kr,1,3,dlsch->harq_processes[harq_pid]->C,NULL,NULL,NULL,NULL); ldpc_encoder_optim_8seg(c_tmp,d_tmp,Kr,1,3,r0,NULL,NULL,NULL,NULL);
for (r=(dlsch->harq_processes[harq_pid]->C/(total_worker+1))*current_worker; r<(dlsch->harq_processes[harq_pid]->C/(total_worker+1))*(current_worker+1); r++) { for (r0=0,r=segments_per_worker[dlsch->harq_processes[harq_pid]->C-1][current_worker]; r<segments_per_worker[dlsch->harq_processes[harq_pid]->C][current_worker+1]; r0++,r++) {
dlsch->harq_processes[harq_pid]->RTC[r] = dlsch->harq_processes[harq_pid]->RTC[r] =
sub_block_interleaving_turbo((Kr_int), sub_block_interleaving_turbo((Kr_int),
&dlsch->harq_processes[harq_pid]->d[r][96], &dlsch->harq_processes[harq_pid]->d[r][96],
...@@ -365,19 +369,19 @@ int dlsch_encoding_2threads0(te_params *tep) { ...@@ -365,19 +369,19 @@ int dlsch_encoding_2threads0(te_params *tep) {
// Fill in the "e"-sequence from 36-212, V8.6 2009-03, p. 16-17 (for each "e") and concatenate the // Fill in the "e"-sequence from 36-212, V8.6 2009-03, p. 16-17 (for each "e") and concatenate the
// outputs for each code segment, see Section 5.1.5 p.20 // outputs for each code segment, see Section 5.1.5 p.20
for (r=0,r_offset=0; r<(dlsch->harq_processes[harq_pid]->C/(total_worker+1))*(current_worker+1); r++) { for (r=0,r_offset=0; r<segments_per_worker[dlsch->harq_processes[harq_pid]->C][current_worker+1]; r++) {
if(r<(dlsch->harq_processes[harq_pid]->C/(total_worker+1))*(current_worker)){ if(r<segments_per_worker[dlsch->harq_processes[harq_pid]->C][current_worker]){
int Nl=dlsch->harq_processes[harq_pid]->Nl; int Nl=dlsch->harq_processes[harq_pid]->Nl;
int Qm=dlsch->harq_processes[harq_pid]->Qm; int Qm=dlsch->harq_processes[harq_pid]->Qm;
int C = dlsch->harq_processes[harq_pid]->C; int C = dlsch->harq_processes[harq_pid]->C;
int Gp = G/Nl/Qm; int Gp = G/Nl/Qm;
int GpmodC = Gp%C; int GpmodC = Gp%C;
if (r < (C-(GpmodC))) if (r < (C-(GpmodC)))
r_offset += Nl*Qm * (Gp/C); r_offset += Nl*Qm * (Gp/C);
else else
r_offset += Nl*Qm * ((GpmodC==0?0:1) + (Gp/C)); r_offset += Nl*Qm * ((GpmodC==0?0:1) + (Gp/C));
} }
else{ else{
r_offset += lte_rate_matching_turbo(dlsch->harq_processes[harq_pid]->RTC[r], r_offset += lte_rate_matching_turbo(dlsch->harq_processes[harq_pid]->RTC[r],
G, //G G, //G
dlsch->harq_processes[harq_pid]->w[r], dlsch->harq_processes[harq_pid]->w[r],
...@@ -467,7 +471,8 @@ int dlsch_encoding_2threads(PHY_VARS_eNB *eNB, ...@@ -467,7 +471,8 @@ int dlsch_encoding_2threads(PHY_VARS_eNB *eNB,
unsigned short nb_rb = dlsch->harq_processes[harq_pid]->nb_rb; unsigned short nb_rb = dlsch->harq_processes[harq_pid]->nb_rb;
unsigned int A,Z; unsigned int A,Z;
unsigned char mod_order; unsigned char mod_order;
unsigned int Kr=0,Kr_bytes,r,r_offset=0,Kr_int=0; unsigned int Kr=0,Kr_bytes,r,r0,r_offset=0,Kr_int=0;
uint8_t *c_tmp[MAX_NUM_DLSCH_SEGMENTS];
uint8_t *d_tmp[MAX_NUM_DLSCH_SEGMENTS]; uint8_t *d_tmp[MAX_NUM_DLSCH_SEGMENTS];
// unsigned short m=dlsch->harq_processes[harq_pid]->mcs; // unsigned short m=dlsch->harq_processes[harq_pid]->mcs;
...@@ -552,7 +557,7 @@ int dlsch_encoding_2threads(PHY_VARS_eNB *eNB, ...@@ -552,7 +557,7 @@ int dlsch_encoding_2threads(PHY_VARS_eNB *eNB,
#ifdef TD_DECODING #ifdef TD_DECODING
start_meas(te_main_stats); start_meas(te_main_stats);
for (r=(dlsch->harq_processes[harq_pid]->C/(worker_num+1))*worker_num; r<dlsch->harq_processes[harq_pid]->C; r++) { for (r=segments_per_worker[dlsch->harq_processes[harq_pid]->C][worker_num-1]; r<dlsch->harq_processes[harq_pid]->C; r++) {
if (r<dlsch->harq_processes[harq_pid]->Cminus) if (r<dlsch->harq_processes[harq_pid]->Cminus)
Kr = dlsch->harq_processes[harq_pid]->Kminus; Kr = dlsch->harq_processes[harq_pid]->Kminus;
...@@ -604,17 +609,18 @@ int dlsch_encoding_2threads(PHY_VARS_eNB *eNB, ...@@ -604,17 +609,18 @@ int dlsch_encoding_2threads(PHY_VARS_eNB *eNB,
else else
Kr_int = Kr; Kr_int = Kr;
for (r=(dlsch->harq_processes[harq_pid]->C/(worker_num+1))*worker_num; r<dlsch->harq_processes[harq_pid]->C; r++) { for (r0=0,r=segments_per_worker[dlsch->harq_processes[harq_pid]->C][worker_num-1]; r<dlsch->harq_processes[harq_pid]->C; r++,r0++) {
memset(dlsch->harq_processes[harq_pid]->d[r],0,(96+12+3+3*8448)*sizeof(uint8_t)); memset(dlsch->harq_processes[harq_pid]->d[r],0,(96+12+3+3*8448)*sizeof(uint8_t));
d_tmp[r] = &dlsch->harq_processes[harq_pid]->d[r][96]; d_tmp[r0] = &dlsch->harq_processes[harq_pid]->c[r][0];
d_tmp[r0] = &dlsch->harq_processes[harq_pid]->d[r][96];
} }
start_meas(te_stats); start_meas(te_stats);
ldpc_encoder_optim_8seg(dlsch->harq_processes[harq_pid]->c,d_tmp,Kr,1,3,dlsch->harq_processes[harq_pid]->C,NULL,NULL,NULL,NULL); ldpc_encoder_optim_8seg(c_tmp,d_tmp,Kr,1,3,r0,NULL,NULL,NULL,NULL);
stop_meas(te_stats); stop_meas(te_stats);
start_meas(i_stats); start_meas(i_stats);
for (r=(dlsch->harq_processes[harq_pid]->C/(worker_num+1))*worker_num; r<dlsch->harq_processes[harq_pid]->C; r++) { for (r=segments_per_worker[dlsch->harq_processes[harq_pid]->C][worker_num-1]; r<dlsch->harq_processes[harq_pid]->C; r++) {
dlsch->harq_processes[harq_pid]->RTC[r] = dlsch->harq_processes[harq_pid]->RTC[r] =
sub_block_interleaving_turbo((Kr_int), sub_block_interleaving_turbo((Kr_int),
&dlsch->harq_processes[harq_pid]->d[r][96], &dlsch->harq_processes[harq_pid]->d[r][96],
...@@ -647,7 +653,7 @@ int dlsch_encoding_2threads(PHY_VARS_eNB *eNB, ...@@ -647,7 +653,7 @@ int dlsch_encoding_2threads(PHY_VARS_eNB *eNB,
for (r=0,r_offset=0; r<dlsch->harq_processes[harq_pid]->C; r++) { for (r=0,r_offset=0; r<dlsch->harq_processes[harq_pid]->C; r++) {
// get information for E for the segments that are handled by the worker thread // get information for E for the segments that are handled by the worker thread
if (r<(dlsch->harq_processes[harq_pid]->C/(worker_num+1))*worker_num) { if (r<segments_per_worker[dlsch->harq_processes[harq_pid]->C][worker_num-1]) {
int Nl=dlsch->harq_processes[harq_pid]->Nl; int Nl=dlsch->harq_processes[harq_pid]->Nl;
int Qm=dlsch->harq_processes[harq_pid]->Qm; int Qm=dlsch->harq_processes[harq_pid]->Qm;
int C = dlsch->harq_processes[harq_pid]->C; int C = dlsch->harq_processes[harq_pid]->C;
...@@ -738,7 +744,7 @@ int dlsch_encoding_all(PHY_VARS_eNB *eNB, ...@@ -738,7 +744,7 @@ int dlsch_encoding_all(PHY_VARS_eNB *eNB,
C = C+1; C = C+1;
} }
} }
/*
if(C >= 8 && get_nprocs()>=8 && codingw)//one main three worker if(C >= 8 && get_nprocs()>=8 && codingw)//one main three worker
{ {
encoding_return = encoding_return =
...@@ -757,7 +763,8 @@ int dlsch_encoding_all(PHY_VARS_eNB *eNB, ...@@ -757,7 +763,8 @@ int dlsch_encoding_all(PHY_VARS_eNB *eNB,
i_stats, i_stats,
3); 3);
} }
else if(C >= 6 && get_nprocs()>=6 && codingw)//one main two worker */
if(C >= 7 && get_nprocs()>=6 && codingw)//one main two worker
{ {
encoding_return = encoding_return =
dlsch_encoding_2threads(eNB, dlsch_encoding_2threads(eNB,
......
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