maki/openairinterface5g
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Enabled multi-threading for gNB L1 tx

Browse Source 
A second thread message is created for L1 tx processing and pushed into the threaPool.
The message contain the nfapi rel15 PDUs for SSB, PDCCH and PDSCH. The memory of these PDUs are reused for subsequent slots.
Need to fix the nfapi mode VNF PNF implementation to adapt to this memory scheme.
This commit is contained in:
Sakthivel Velumani
2021-08-05 21:29:53 +05:30
parent 4d05f4b288
commit cb68456357
10 changed files with 116 additions and 121 deletions
Download Patch File Download Diff File Expand all files Collapse all files
executables/nr-gnb.c
+16 -8
View File
@@ -113,15 +113,25 @@ time_stats_t softmodem_stats_rx_sf; // total rx time
void tx_func(void *param) {
processingData_L1_t *info = (processingData_L1_t *) param;
processingData_L1tx_t *info = (processingData_L1tx_t *) param;
PHY_VARS_gNB *gNB = info->gNB;
int frame_tx = info->frame_tx;
int slot_tx = info->slot_tx;
phy_procedures_gNB_TX(gNB, frame_tx,slot_tx, 1);
phy_procedures_gNB_TX(info,
frame_tx,
slot_tx,
1);
if ((frame_tx&127) == 0) dump_pdsch_stats(gNB);
// start FH TX processing
// If the later of the 2 L1 tx thread finishes first,
// we wait for the earlier one to finish and start the RU thread
// to avoid realtime issues with USRP
notifiedFIFO_elt_t *res;
res = pullTpool(gNB->resp_L1_tx, gNB->threadPool);
pushNotifiedFIFO(gNB->resp_L1_tx,res);
// Start RU TX processing.
res = pullTpool(gNB->resp_RU_tx, gNB->threadPool);
processingData_RU_t *syncMsg = (processingData_RU_t *)NotifiedFifoData(res);
syncMsg->frame_tx = frame_tx;
@@ -257,7 +267,6 @@ void rx_func(void *param) {
LOG_D(PHY,"%s() Exit proc[rx:%d%d tx:%d%d]\n", __FUNCTION__, frame_rx, slot_rx, frame_tx, slot_tx);
// Call the scheduler
start_meas(&gNB->ul_indication_stats);
pthread_mutex_lock(&gNB->UL_INFO_mutex);
gNB->UL_INFO.frame = frame_rx;
@@ -269,10 +278,9 @@ void rx_func(void *param) {
stop_meas(&gNB->ul_indication_stats);
notifiedFIFO_elt_t *res;
if (tx_slot_type == NR_DOWNLINK_SLOT || tx_slot_type == NR_MIXED_SLOT) {
res = pullTpool(gNB->resp_L1_tx, gNB->threadPool);
processingData_L1_t *syncMsg = (processingData_L1_t *)NotifiedFifoData(res);
processingData_L1tx_t *syncMsg = (processingData_L1tx_t *)NotifiedFifoData(res);
syncMsg->gNB = gNB;
syncMsg->frame_rx = frame_rx;
syncMsg->slot_rx = slot_rx;
@@ -327,7 +335,6 @@ static void *process_stats_thread(void *param) {
PHY_VARS_gNB *gNB = (PHY_VARS_gNB *)param;
reset_meas(&gNB->phy_proc_tx);
reset_meas(&gNB->dlsch_encoding_stats);
reset_meas(&gNB->phy_proc_rx);
reset_meas(&gNB->ul_indication_stats);
@@ -339,7 +346,8 @@ static void *process_stats_thread(void *param) {
while(!oai_exit)
{
sleep(1);
print_meas(&gNB->phy_proc_tx, "L1 Tx processing", NULL, NULL);
print_meas(gNB->phy_proc_tx_0, "L1 Tx processing thread 0", NULL, NULL);
print_meas(gNB->phy_proc_tx_1, "L1 Tx processing thread 1", NULL, NULL);
print_meas(&gNB->dlsch_encoding_stats, "DLSCH encoding", NULL, NULL);
print_meas(&gNB->phy_proc_rx, "L1 Rx processing", NULL, NULL);
print_meas(&gNB->ul_indication_stats, "UL Indication", NULL, NULL);
executables/nr-ru.c
+19 -4
View File
@@ -1270,12 +1270,27 @@ void *ru_thread( void *param ) {
wait_sync("ru_thread");
notifiedFIFO_elt_t *msg = newNotifiedFIFO_elt(sizeof(processingData_L1_t),0,gNB->resp_L1,rx_func);
notifiedFIFO_elt_t *msgL1Tx = newNotifiedFIFO_elt(sizeof(processingData_L1_t),0,gNB->resp_L1_tx,tx_func);
notifiedFIFO_elt_t *msgRUTx = newNotifiedFIFO_elt(sizeof(processingData_L1_t),0,gNB->resp_RU_tx,ru_tx_func);
processingData_L1_t *syncMsg;
notifiedFIFO_elt_t *res;
pushNotifiedFIFO(gNB->resp_L1,msg); // to unblock the process in the beginning
// we create 2 threads for L1 tx processing
notifiedFIFO_elt_t *msgL1Tx = newNotifiedFIFO_elt(sizeof(processingData_L1tx_t),0,gNB->resp_L1_tx,tx_func);
processingData_L1tx_t *msgDataTx = (processingData_L1tx_t *)NotifiedFifoData(msgL1Tx);
init_DLSCH_struct(gNB, msgDataTx);
reset_meas(&msgDataTx->phy_proc_tx);
gNB->phy_proc_tx_0 = &msgDataTx->phy_proc_tx;
pushNotifiedFIFO(gNB->resp_L1_tx,msgL1Tx); // to unblock the process in the beginning
msgL1Tx = newNotifiedFIFO_elt(sizeof(processingData_L1tx_t),0,gNB->resp_L1_tx,tx_func);
msgDataTx = (processingData_L1tx_t *)NotifiedFifoData(msgL1Tx);
init_DLSCH_struct(gNB, msgDataTx);
reset_meas(&msgTx->phy_proc_tx);
gNB->phy_proc_tx_1 = &msgDataTx->phy_proc_tx;
pushNotifiedFIFO(gNB->resp_L1_tx,msgL1Tx); // to unblock the process in the beginning
processingData_L1_t *syncMsg;
notifiedFIFO_elt_t *res;
notifiedFIFO_elt_t *msgRUTx = newNotifiedFIFO_elt(sizeof(processingData_L1_t),0,gNB->resp_RU_tx,ru_tx_func);
pushNotifiedFIFO(gNB->resp_RU_tx,msgRUTx); // to unblock the process in the beginning
if(!emulate_rf) {