testing with 2x2 and 4x4 with N310, small modifications in nr_ru_procedures...

testing with 2x2 and 4x4 with N310, small modifications in nr_ru_procedures and usrp_lib.cpp for multiple antennas

executables/nr-ru.c

@@ -1397,6 +1397,7 @@ void *ru_thread( void *param ) {
       }
     }
 
+
     // At this point, all information for subframe has been received on FH interface
     res = pullTpool(gNB->resp_L1, gNB->threadPool);
     syncMsg = (processingData_L1_t *)NotifiedFifoData(res);
@@ -1409,6 +1410,7 @@ void *ru_thread( void *param ) {
     res->key = proc->tti_rx;
     pushTpool(gNB->threadPool, res);
 
+
   }
 
   printf( "Exiting ru_thread \n");

openair1/PHY/INIT/nr_init_ru.c

@@ -121,37 +121,39 @@ int nr_phy_init_RU(RU_t *ru) {
 		RC.nb_nr_L1_inst,NUMBER_OF_gNB_MAX);
 
     LOG_E(PHY,"[INIT] %s() RC.nb_nr_L1_inst:%d \n", __FUNCTION__, RC.nb_nr_L1_inst);
-    
-    int beam_count = 0;
-    if (ru->nb_tx>1) {//Enable beamforming when nb_tx > 1
-      for (p=0;p<ru->nb_log_antennas;p++) {
-        //if ((fp->L_ssb >> (63-p)) & 0x01)//64 bit-map with the MSB @2⁶³ corresponds to SSB ssb_index 0
-          beam_count++;
-      }
-      AssertFatal(ru->nb_bfw==(beam_count*ru->nb_tx),"Number of beam weights from config file is %d while the expected number is %d",ru->nb_bfw,(beam_count*ru->nb_tx));
-    
-      int l_ind = 0;
-      for (i=0; i<RC.nb_nr_L1_inst; i++) {
+   
+    if (ru->do_precoding == 1) { 
+      int beam_count = 0;
+      if (ru->nb_tx>1) {//Enable beamforming when nb_tx > 1
         for (p=0;p<ru->nb_log_antennas;p++) {
+          //if ((fp->L_ssb >> (63-p)) & 0x01)//64 bit-map with the MSB @2⁶³ corresponds to SSB ssb_index 0
+            beam_count++;
+        }
+        AssertFatal(ru->nb_bfw==(beam_count*ru->nb_tx),"Number of beam weights from config file is %d while the expected number is %d",ru->nb_bfw,(beam_count*ru->nb_tx));
+    
+        int l_ind = 0;
+        for (i=0; i<RC.nb_nr_L1_inst; i++) {
+          for (p=0;p<ru->nb_log_antennas;p++) {
           //if ((fp->L_ssb >> (63-p)) & 0x01)  {
-	    ru->beam_weights[i][p] = (int32_t **)malloc16_clear(ru->nb_tx*sizeof(int32_t*));
-	    for (j=0; j<ru->nb_tx; j++) {
-	      ru->beam_weights[i][p][j] = (int32_t *)malloc16_clear(fp->ofdm_symbol_size*sizeof(int32_t));
-              AssertFatal(ru->bw_list[i],"ru->bw_list[%d] is null\n",i);
-              for (re=0; re<fp->ofdm_symbol_size; re++) 
-		ru->beam_weights[i][p][j][re] = ru->bw_list[i][l_ind];
+	      ru->beam_weights[i][p] = (int32_t **)malloc16_clear(ru->nb_tx*sizeof(int32_t*));
+	      for (j=0; j<ru->nb_tx; j++) {
+	        ru->beam_weights[i][p][j] = (int32_t *)malloc16_clear(fp->ofdm_symbol_size*sizeof(int32_t));
+                AssertFatal(ru->bw_list[i],"ru->bw_list[%d] is null\n",i);
+                for (re=0; re<fp->ofdm_symbol_size; re++) 
+		  ru->beam_weights[i][p][j][re] = ru->bw_list[i][l_ind];
               //printf("Beam Weight %08x for beam %d and tx %d\n",ru->bw_list[i][l_ind],p,j);
-              l_ind++; 
-  	    } // for j
-	  //}
-        } // for p
-      } //for i
-    }
+                l_ind++; 
+  	      } // for j
+	    //}
+          } // for p
+        } //for i
+      }
 
-    ru->common.beam_id = (uint8_t**)malloc16_clear(ru->nb_tx*sizeof(uint8_t*));
-    for(i=0; i< ru->nb_tx; ++i) {
-      ru->common.beam_id[i] = (uint8_t*)malloc16_clear(fp->symbols_per_slot*fp->slots_per_frame*sizeof(uint8_t));
-      memset(ru->common.beam_id[i],255,fp->symbols_per_slot*fp->slots_per_frame);
+      ru->common.beam_id = (uint8_t**)malloc16_clear(ru->nb_tx*sizeof(uint8_t*));
+      for(i=0; i< ru->nb_tx; ++i) {
+        ru->common.beam_id[i] = (uint8_t*)malloc16_clear(fp->symbols_per_slot*fp->slots_per_frame*sizeof(uint8_t));
+        memset(ru->common.beam_id[i],255,fp->symbols_per_slot*fp->slots_per_frame);
+      }
     }
   } // !=IF5
 

openair1/SCHED_NR/nr_ru_procedures.c

@@ -305,15 +305,16 @@ static void *nr_feptx_thread(void *param) {
       ////////////precoding////////////
       VCD_SIGNAL_DUMPER_DUMP_FUNCTION_BY_NAME(VCD_SIGNAL_DUMPER_FUNCTIONS_PHY_PROCEDURES_RU_FEPTX_PREC+feptx->index+1 , 1);
       
-      start_meas(&ru->precoding_stats);
-
-      for(i=0; i<ru->nb_log_antennas; ++i) {
-	memcpy((void*) &ru->common.beam_id[i][slot*fp->symbols_per_slot+l],
-	       (void*) &ru->gNB_list[0]->common_vars.beam_id[i][slot*fp->symbols_per_slot+l],
-	       (fp->symbols_per_slot>>1)*sizeof(uint8_t));
+      if (aa==0 && l==0) start_meas(&ru->precoding_stats);
+
+      if (ru->do_precoding == 1) {
+        for(i=0; i<ru->nb_log_antennas; ++i) {
+	   memcpy((void*) &ru->common.beam_id[i][slot*fp->symbols_per_slot+l],
+	          (void*) &ru->gNB_list[0]->common_vars.beam_id[i][slot*fp->symbols_per_slot+l],
+	          (fp->symbols_per_slot>>1)*sizeof(uint8_t));
+         }
       }
 
-
       if (ru->nb_tx == 1 && ru->nb_log_antennas == 1) {
 	memcpy((void*)&ru->common.txdataF_BF[0][l*fp->ofdm_symbol_size],
 	       (void*)&ru->gNB_list[0]->common_vars.txdataF[0][txdataF_offset + l*fp->ofdm_symbol_size],
@@ -321,10 +322,9 @@ static void *nr_feptx_thread(void *param) {
       }
       else if (ru->do_precoding == 0) {
         int gNB_tx = ru->gNB_list[0]->frame_parms.nb_antennas_tx;
-        for (int aa=0;aa<ru->nb_tx;aa++) 
-           memcpy((void*)&ru->common.txdataF_BF[aa][l*fp->ofdm_symbol_size],
-                  (void*)&ru->gNB_list[0]->common_vars.txdataF[aa%gNB_tx][txdataF_offset + l*fp->ofdm_symbol_size],
-                  (fp->samples_per_slot_wCP>>1)*sizeof(int32_t));
+        memcpy((void*)&ru->common.txdataF_BF[aa][l*fp->ofdm_symbol_size],
+               (void*)&ru->gNB_list[0]->common_vars.txdataF[aa%gNB_tx][txdataF_offset + l*fp->ofdm_symbol_size],
+               (fp->samples_per_slot_wCP>>1)*sizeof(int32_t));
       }
       else {
         bw  = ru->beam_weights[0];
@@ -340,13 +340,13 @@ static void *nr_feptx_thread(void *param) {
                         txdataF_offset);//here
         }
       }
-      stop_meas(&ru->precoding_stats);
+      if (aa==0 && l==0) stop_meas(&ru->precoding_stats);
       VCD_SIGNAL_DUMPER_DUMP_FUNCTION_BY_NAME(VCD_SIGNAL_DUMPER_FUNCTIONS_PHY_PROCEDURES_RU_FEPTX_PREC+feptx->index+1 , 0);
 
       ////////////FEPTX////////////
-      start_meas(&ru->ofdm_mod_stats);
+      if (aa==0 && l==0) start_meas(&ru->ofdm_mod_stats);
       nr_feptx0(ru,slot,start,fp->symbols_per_slot>>1,aa);
-      stop_meas(&ru->ofdm_mod_stats);
+      if (aa==0 && l==0) stop_meas(&ru->ofdm_mod_stats);
 
       if (release_thread(&feptx->mutex_feptx,&feptx->instance_cnt_feptx,"NR feptx thread")<0) break;
 

targets/ARCH/USRP/USERSPACE/LIB/usrp_lib.cpp

@@ -385,14 +385,14 @@ static int trx_usrp_write(openair0_device *device,
 #if defined(__x86_64) || defined(__i386__)
   #ifdef __AVX2__
       nsamps2 = (nsamps+7)>>3;
-      __m256i buff_tx[8][nsamps2];
+      __m256i buff_tx[cc<2?2:cc][nsamps2];
   #else
     nsamps2 = (nsamps+3)>>2;
-    __m128i buff_tx[8][nsamps2];
+    __m128i buff_tx[cc<2?2:cc][nsamps2];
   #endif
 #elif defined(__arm__)
     nsamps2 = (nsamps+3)>>2;
-    int16x8_t buff_tx[8][nsamps2];
+    int16x8_t buff_tx[cc<2?2:cc][nsamps2];
 #else
 #error Unsupported CPU architecture, USRP device cannot be built
 #endif
@@ -522,14 +522,14 @@ void *trx_usrp_write_thread(void * arg){
     #if defined(__x86_64) || defined(__i386__)
       #ifdef __AVX2__
         nsamps2 = (nsamps+7)>>3;
-        __m256i buff_tx[8][nsamps2];
+        __m256i buff_tx[cc<2?2:cc][nsamps2];
       #else
         nsamps2 = (nsamps+3)>>2;
-        __m128i buff_tx[8][nsamps2];
+        __m128i buff_tx[cc<2?2:cc][nsamps2];
       #endif
     #elif defined(__arm__)
       nsamps2 = (nsamps+3)>>2;
-      int16x8_t buff_tx[8][nsamps2];
+      int16x8_t buff_tx[cc<2?2:cc][nsamps2];
     #else
     #error Unsupported CPU architecture, USRP device cannot be built
     #endif
@@ -623,14 +623,14 @@ static int trx_usrp_read(openair0_device *device, openair0_timestamp *ptimestamp
 #if defined(__x86_64) || defined(__i386__)
 #ifdef __AVX2__
   nsamps2 = (nsamps+7)>>3;
-  __m256i buff_tmp[2][nsamps2];
+  __m256i buff_tmp[cc<2 ? 2 : cc][nsamps2];
 #else
   nsamps2 = (nsamps+3)>>2;
-  __m128i buff_tmp[2][nsamps2];
+  __m128i buff_tmp[cc<2 ? 2 : cc][nsamps2];
 #endif
 #elif defined(__arm__)
   nsamps2 = (nsamps+3)>>2;
-  int16x8_t buff_tmp[2][nsamps2];
+  int16x8_t buff_tmp[cc<2 ? 2 : cc][nsamps2];
 #endif
 
     samples_received=0;