remove gnb dependence in common nr_modulation file
This commit is contained in:
@@ -7,7 +7,6 @@
|
||||
|
||||
#include <stdint.h>
|
||||
#include "PHY/defs_nr_common.h"
|
||||
#include "PHY/defs_gNB.h"
|
||||
|
||||
#define DMRS_MOD_ORDER 2
|
||||
/*Precoding matices: W[pmi][antenna_port][layer]*/
|
||||
@@ -127,7 +126,7 @@ c16_t nr_layer_precoder_cm(int n_layers,
|
||||
int symSz,
|
||||
c16_t datatx_F_precoding[n_layers][symSz],
|
||||
int ap,
|
||||
nfapi_nr_pm_pdu_t *pmi_pdu,
|
||||
c16_t weights[NR_MAX_NB_LAYERS][NR_MAX_CSI_PORTS],
|
||||
int offset);
|
||||
|
||||
/*! \brief Precoding with SIMDe, txdataF_precoded[] = prec_matrix[] * txdataF_res_mapped[]
|
||||
@@ -140,7 +139,7 @@ void nr_layer_precoder_simd(const int n_layers,
|
||||
const int symSz,
|
||||
const c16_t txdataF_res_mapped[n_layers][symSz],
|
||||
const int ant,
|
||||
const nfapi_nr_pm_pdu_t *pmi_pdu,
|
||||
c16_t weights[NR_MAX_NB_LAYERS][NR_MAX_CSI_PORTS],
|
||||
const int sc_offset,
|
||||
const int re_cnt,
|
||||
c16_t *txdataF_precoded);
|
||||
|
||||
Reference in New Issue
Block a user