fixing txdataF allocation in NR_gNB_DLSCH_t
This commit is contained in:
committed by
Thomas Schlichter
parent
566e603671
commit
714452baef
@@ -151,11 +151,11 @@ NR_gNB_DLSCH_t *new_gNB_dlsch(unsigned char Kmimo,
|
||||
}
|
||||
}
|
||||
|
||||
dlsch->txdataF[layer] = (int32_t *)malloc16((NR_MAX_PDSCH_ENCODED_LENGTH>>1)*sizeof(int32_t*));
|
||||
dlsch->txdataF[layer] = (int32_t *)malloc16((NR_MAX_PDSCH_ENCODED_LENGTH/NR_MAX_NB_LAYERS)*sizeof(int32_t)); // NR_MAX_NB_LAYERS is already included in NR_MAX_PDSCH_ENCODED_LENGTH
|
||||
}
|
||||
|
||||
for (int q=0; q<NR_MAX_NB_CODEWORDS; q++)
|
||||
dlsch->mod_symbs[q] = (int32_t *)malloc16((NR_MAX_PDSCH_ENCODED_LENGTH>>1)*sizeof(int32_t*));
|
||||
dlsch->mod_symbs[q] = (int32_t *)malloc16(NR_MAX_PDSCH_ENCODED_LENGTH*sizeof(int32_t));
|
||||
|
||||
dlsch->calib_dl_ch_estimates = (int32_t**)malloc16(64*sizeof(int32_t*));
|
||||
for (aa=0; aa<64; aa++) {
|
||||
|
||||
Reference in New Issue
Block a user