dynamic mcs selection for msg4

This commit is contained in:
francescomani
2022-01-06 15:04:42 +01:00
parent 7f20ad6c6e
commit 10e4ce2dce
@@ -1597,12 +1597,15 @@ void nr_generate_Msg4(module_id_t module_idP, int CC_id, frame_t frameP, sub_fra
uint8_t tb_scaling = 0;
uint16_t *vrb_map = cc[CC_id].vrb_map;
do {
rbSize++;
if(rbSize < BWPSize)
rbSize++;
else
mcsIndex++;
LOG_D(NR_MAC,"Calling nr_compute_tbs with N_PRB_DMRS %d, N_DMRS_SLOT %d\n",N_PRB_DMRS,N_DMRS_SLOT);
harq->tb_size = nr_compute_tbs(nr_get_Qm_dl(mcsIndex, mcsTableIdx),
nr_get_code_rate_dl(mcsIndex, mcsTableIdx),
rbSize, nrOfSymbols, N_PRB_DMRS * N_DMRS_SLOT, 0, tb_scaling,1) >> 3;
} while (rbSize < BWPSize && harq->tb_size < ra->mac_pdu_length);
nr_get_code_rate_dl(mcsIndex, mcsTableIdx),
rbSize, nrOfSymbols, N_PRB_DMRS * N_DMRS_SLOT, 0, tb_scaling,1) >> 3;
} while (harq->tb_size < ra->mac_pdu_length);
int i = 0;
while ((i < rbSize) && (rbStart + rbSize <= BWPSize)) {