Commit d145d8d6 authored by Robert Schmidt's avatar Robert Schmidt

Consider retransmissions in the preprocessor

parent 0ef332ef
...@@ -513,45 +513,79 @@ void nr_simple_dlsch_preprocessor(module_id_t module_id, ...@@ -513,45 +513,79 @@ void nr_simple_dlsch_preprocessor(module_id_t module_id,
module_id, UE_id, frame, slot, num_slots_per_tdd, &sched_ctrl->pucch_sched_idx); module_id, UE_id, frame, slot, num_slots_per_tdd, &sched_ctrl->pucch_sched_idx);
AssertFatal(sched_ctrl->pucch_sched_idx >= 0, "no uplink slot for PUCCH found!\n"); AssertFatal(sched_ctrl->pucch_sched_idx >= 0, "no uplink slot for PUCCH found!\n");
// Time-domain allocation
sched_ctrl->time_domain_allocation = 2;
// Freq-demain allocation
uint8_t *vrb_map = RC.nrmac[module_id]->common_channels[CC_id].vrb_map; uint8_t *vrb_map = RC.nrmac[module_id]->common_channels[CC_id].vrb_map;
const int current_harq_pid = sched_ctrl->current_harq_pid;
NR_UE_harq_t *harq = &sched_ctrl->harq_processes[current_harq_pid];
NR_UE_ret_info_t *retInfo = &sched_ctrl->retInfo[current_harq_pid];
const uint16_t bwpSize = NRRIV2BW(sched_ctrl->active_bwp->bwp_Common->genericParameters.locationAndBandwidth, 275); const uint16_t bwpSize = NRRIV2BW(sched_ctrl->active_bwp->bwp_Common->genericParameters.locationAndBandwidth, 275);
int rbStart = NRRIV2PRBOFFSET(sched_ctrl->active_bwp->bwp_Common->genericParameters.locationAndBandwidth, 275); int rbStart = NRRIV2PRBOFFSET(sched_ctrl->active_bwp->bwp_Common->genericParameters.locationAndBandwidth, 275);
while (rbStart < bwpSize && vrb_map[rbStart]) rbStart++;
int rbSize = 1; if (harq->round != 0) { /* retransmission */
while (rbStart + rbSize < bwpSize && !vrb_map[rbStart + rbSize]) rbSize++; sched_ctrl->time_domain_allocation = retInfo->time_domain_allocation;
DevAssert(rbSize >= 3); /* just ensure we have at least 3 RBs */
sched_ctrl->rbSize = rbSize; /* ensure that there is a free place for RB allocation */
sched_ctrl->rbStart = rbStart; int rbSize = 0;
while (rbSize < retInfo->rbSize) {
rbStart += rbSize; /* last iteration rbSize was not enough, skip it */
rbSize = 0;
while (rbStart < bwpSize && vrb_map[rbStart]) rbStart++;
if (rbStart >= bwpSize) {
LOG_E(MAC,
"cannot allocate retransmission for UE %d/RNTI %04x: no resources\n",
UE_id,
rnti);
return;
}
while (rbStart + rbSize < bwpSize
&& !vrb_map[rbStart + rbSize]
&& rbSize < retInfo->rbSize)
rbSize++;
}
sched_ctrl->rbSize = retInfo->rbSize;
sched_ctrl->rbStart = rbStart;
/* MCS etc: just reuse from previous scheduling opportunity */
sched_ctrl->mcsTableIdx = retInfo->mcsTableIdx;
sched_ctrl->mcs = retInfo->mcs;
sched_ctrl->numDmrsCdmGrpsNoData = retInfo->numDmrsCdmGrpsNoData;
} else {
// Time-domain allocation
sched_ctrl->time_domain_allocation = 2;
// Freq-demain allocation
while (rbStart < bwpSize && vrb_map[rbStart]) rbStart++;
int rbSize = 1;
while (rbStart + rbSize < bwpSize && !vrb_map[rbStart + rbSize]) rbSize++;
DevAssert(rbSize >= 3); /* just ensure we have at least 3 RBs */
sched_ctrl->rbSize = rbSize;
sched_ctrl->rbStart = rbStart;
// modulation scheme
sched_ctrl->mcsTableIdx = 0;
sched_ctrl->mcs = 9;
sched_ctrl->numDmrsCdmGrpsNoData = 1;
uint8_t N_PRB_DMRS =
getN_PRB_DMRS(sched_ctrl->active_bwp, sched_ctrl->numDmrsCdmGrpsNoData);
int nrOfSymbols = getNrOfSymbols(sched_ctrl->active_bwp,
sched_ctrl->time_domain_allocation);
const uint32_t TBS =
nr_compute_tbs(nr_get_Qm_dl(sched_ctrl->mcs, sched_ctrl->mcsTableIdx),
nr_get_code_rate_dl(sched_ctrl->mcs, sched_ctrl->mcsTableIdx),
sched_ctrl->rbSize,
nrOfSymbols,
N_PRB_DMRS, // FIXME // This should be multiplied by the
// number of dmrs symbols
0 /* N_PRB_oh, 0 for initialBWP */,
0 /* tb_scaling */,
1 /* nrOfLayers */)
>> 3;
AssertFatal(TBS != 0, "TBS is zero but requested %d RBs!\n", sched_ctrl->rbSize);
}
/* mark the corresponding RBs as used */ /* mark the corresponding RBs as used */
for (int rb = 0; rb < sched_ctrl->rbSize; rb++) for (int rb = 0; rb < sched_ctrl->rbSize; rb++)
vrb_map[rb + sched_ctrl->rbStart] = 1; vrb_map[rb + sched_ctrl->rbStart] = 1;
// modulation scheme
sched_ctrl->mcsTableIdx = 0;
sched_ctrl->mcs = 9;
sched_ctrl->numDmrsCdmGrpsNoData = 1;
uint8_t N_PRB_DMRS =
getN_PRB_DMRS(sched_ctrl->active_bwp, sched_ctrl->numDmrsCdmGrpsNoData);
int nrOfSymbols = getNrOfSymbols(sched_ctrl->active_bwp,
sched_ctrl->time_domain_allocation);
const uint32_t TBS =
nr_compute_tbs(nr_get_Qm_dl(sched_ctrl->mcs, sched_ctrl->mcsTableIdx),
nr_get_code_rate_dl(sched_ctrl->mcs, sched_ctrl->mcsTableIdx),
sched_ctrl->rbSize,
nrOfSymbols,
N_PRB_DMRS, // FIXME // This should be multiplied by the
// number of dmrs symbols
0 /* N_PRB_oh, 0 for initialBWP */,
0 /* tb_scaling */,
1 /* nrOfLayers */)
>> 3;
AssertFatal(TBS != 0, "TBS is zero but requested %d RBs!\n", sched_ctrl->rbSize);
} }
void nr_schedule_ue_spec(module_id_t module_id, void nr_schedule_ue_spec(module_id_t module_id,
......
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