moving scale inside tbs computation function
This commit is contained in:
@@ -463,8 +463,7 @@ uint32_t nr_compute_tbs(uint16_t Qm,
|
||||
uint16_t nb_symb_sch,
|
||||
uint16_t nb_dmrs_prb,
|
||||
uint16_t nb_rb_oh,
|
||||
uint8_t Nl,
|
||||
uint8_t scale);
|
||||
uint8_t Nl);
|
||||
|
||||
uint32_t nr_compute_tbslbrm(uint16_t table,
|
||||
uint16_t nb_rb,
|
||||
|
||||
Reference in New Issue
Block a user