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

Make number of UL/DL actors in NR UE fully configurable

Browse Source 
Due to different processing requirements on different CPUs it might
be beneficial to enable the users to set the number of UL/DL actors.

Allow disabling UL/DL actor frameworks and processing inline or
setting the number of DL/UL threads with --num-ul-actors/num-dl-actors

This was tested to work with 0 UL/DL actors with rfsim.
This commit is contained in:
Bartosz Podrygajlo
2025-05-28 07:49:10 +02:00
parent b2c9a1d2b5
commit 56330fde7c
4 changed files with 32 additions and 18 deletions
Download Patch File Download Diff File Expand all files Collapse all files
executables/nr-ue.c
+13 -6
View File
@@ -106,9 +106,12 @@ static void *NRUE_phy_stub_standalone_pnf_task(void *arg);
static void start_process_slot_tx(void* arg) {
notifiedFIFO_elt_t *newTx = arg;
nr_rxtx_thread_data_t *curMsgTx = NotifiedFifoData(newTx);
int num_ul_actors_to_use =
get_nrUE_params()->num_ul_actors == 0 ? NUM_UL_ACTORS : min(get_nrUE_params()->num_ul_actors, NUM_UL_ACTORS);
pushNotifiedFIFO(&curMsgTx->UE->ul_actors[curMsgTx->proc.nr_slot_tx % num_ul_actors_to_use].fifo, newTx);
int num_ul_actors = get_nrUE_params()->num_ul_actors;
if (num_ul_actors > 0) {
pushNotifiedFIFO(&curMsgTx->UE->ul_actors[curMsgTx->proc.nr_slot_tx % num_ul_actors].fifo, newTx);
} else {
newTx->processingFunc(curMsgTx);
}
}
static size_t dump_L1_UE_meas_stats(PHY_VARS_NR_UE *ue, char *output, size_t max_len)
@@ -677,10 +680,10 @@ static int handle_sync_req_from_mac(PHY_VARS_NR_UE *UE)
/* Clearing UE harq while DL actors are active causes race condition.
So we let the current execution to complete here.*/
for (int i = 0; i < NUM_DL_ACTORS; i++) {
for (int i = 0; i < get_nrUE_params()->num_dl_actors; i++) {
flush_actor(UE->dl_actors + i);
}
for (int i = 0; i < NUM_UL_ACTORS; i++) {
for (int i = 0; i < get_nrUE_params()->num_ul_actors; i++) {
flush_actor(UE->ul_actors + i);
}
@@ -1183,7 +1186,11 @@ void *UE_thread(void *arg)
int ret = UE_dl_preprocessing(UE, &curMsgRx->proc, tx_wait_for_dlsch, &curMsgRx->phy_data, &stats_printed);
if (ret != INT_MAX)
shiftForNextFrame = ret;
pushNotifiedFIFO(&UE->dl_actors[curMsg.proc.nr_slot_rx % NUM_DL_ACTORS].fifo, newRx);
if (get_nrUE_params()->num_dl_actors > 0) {
pushNotifiedFIFO(&UE->dl_actors[curMsg.proc.nr_slot_rx % get_nrUE_params()->num_dl_actors].fifo, newRx);
} else {
newRx->processingFunc(curMsgRx);
}
// apply new NTN timing information
apply_ntn_config(UE, fp, slot_nr, &duration_rx_to_tx, &timing_advance, &ntn_koffset);