Fix order of DLSCH deinterleaving and rate matching
On the gNB side tha data are first rate-matched and then interleaved. So on the UE side, the received LLRs have to be deinterleaved first before rate-matching. This significantly improved the performance, especially for MCS values > 3.
This commit is contained in:
@@ -467,33 +467,26 @@ void nr_interleaving_ldpc(uint32_t E, uint8_t Qm, uint8_t *e,uint8_t *f);
|
||||
|
||||
void nr_deinterleaving_ldpc(uint32_t E, uint8_t Qm, int16_t *e,int16_t *f);
|
||||
|
||||
uint32_t nr_rate_matching_ldpc(uint8_t Ilbrm,
|
||||
uint32_t Tbslbrm,
|
||||
uint8_t BG,
|
||||
uint16_t Z,
|
||||
uint32_t G,
|
||||
uint8_t *w,
|
||||
uint8_t *e,
|
||||
uint8_t C,
|
||||
uint8_t rvidx,
|
||||
uint8_t Qm,
|
||||
uint8_t Nl,
|
||||
uint8_t r);
|
||||
int nr_rate_matching_ldpc(uint8_t Ilbrm,
|
||||
uint32_t Tbslbrm,
|
||||
uint8_t BG,
|
||||
uint16_t Z,
|
||||
uint8_t *w,
|
||||
uint8_t *e,
|
||||
uint8_t C,
|
||||
uint8_t rvidx,
|
||||
uint32_t E);
|
||||
|
||||
int nr_rate_matching_ldpc_rx(uint8_t Ilbrm,
|
||||
uint32_t Tbslbrm,
|
||||
uint8_t BG,
|
||||
uint16_t Z,
|
||||
uint32_t G,
|
||||
uint32_t Tbslbrm,
|
||||
uint8_t BG,
|
||||
uint16_t Z,
|
||||
int16_t *w,
|
||||
int16_t *soft_input,
|
||||
uint8_t C,
|
||||
uint8_t rvidx,
|
||||
uint8_t clear,
|
||||
uint8_t Qm,
|
||||
uint8_t Nl,
|
||||
uint8_t r,
|
||||
uint32_t *E_out);
|
||||
uint32_t E);
|
||||
|
||||
decoder_if_t phy_threegpplte_turbo_decoder;
|
||||
decoder_if_t phy_threegpplte_turbo_decoder8;
|
||||
|
||||
Reference in New Issue
Block a user