Merge remote-tracking branch 'origin/develop-nr' into nr-polar-encoder-optimizations

ci-scripts/conf_files/enb.band7.tm1.100PRB.usrpb210.conf

@@ -197,8 +197,8 @@ MACRLCs = (
 	tr_s_preference = "local_L1";
 	tr_n_preference = "local_RRC";
 	phy_test_mode = 0;
-        puSch10xSnr     =  200;
-        puCch10xSnr     =  200;
+        puSch10xSnr     =  160;
+        puCch10xSnr     =  160;
         }  
 );
 
@@ -227,7 +227,7 @@ RUs = (
 THREAD_STRUCT = (
   {
     #three config for level of parallelism "PARALLEL_SINGLE_THREAD", "PARALLEL_RU_L1_SPLIT", or "PARALLEL_RU_L1_TRX_SPLIT"
-    parallel_config    = "PARALLEL_RU_L1_TRX_SPLIT";
+    parallel_config    = "PARALLEL_SINGLE_THREAD";
     #two option for worker "WORKER_DISABLE" or "WORKER_ENABLE"
     worker_config      = "WORKER_ENABLE";
   }

ci-scripts/conf_files/enb.band7.tm1.25PRB.usrpb210.conf

@@ -197,8 +197,8 @@ MACRLCs = (
 	tr_s_preference = "local_L1";
 	tr_n_preference = "local_RRC";
 	phy_test_mode = 0;
-        puSch10xSnr     =  200;
-        puCch10xSnr     =  200;
+        puSch10xSnr     =  160;
+        puCch10xSnr     =  160;
         }  
 );
 
@@ -218,7 +218,7 @@ RUs = (
          att_rx         = 0;
          bands          = [7];
          max_pdschReferenceSignalPower = -27;
-         max_rxgain                    = 115;
+         max_rxgain                    = 125;
          eNB_instances  = [0];
 
     }
@@ -227,7 +227,7 @@ RUs = (
 THREAD_STRUCT = (
   {
     #three config for level of parallelism "PARALLEL_SINGLE_THREAD", "PARALLEL_RU_L1_SPLIT", or "PARALLEL_RU_L1_TRX_SPLIT"
-    parallel_config    = "PARALLEL_RU_L1_TRX_SPLIT";
+    parallel_config    = "PARALLEL_SINGLE_THREAD";
     #two option for worker "WORKER_DISABLE" or "WORKER_ENABLE"
     worker_config      = "WORKER_ENABLE";
   }

ci-scripts/conf_files/enb.band7.tm1.50PRB.usrpb210.conf

@@ -197,8 +197,8 @@ MACRLCs = (
 	tr_s_preference = "local_L1";
 	tr_n_preference = "local_RRC";
 	phy_test_mode = 0;
-        puSch10xSnr     =  200;
-        puCch10xSnr     =  200;
+        puSch10xSnr     =  160;
+        puCch10xSnr     =  160;
         }  
 );
 
@@ -218,7 +218,7 @@ RUs = (
          att_rx         = 0;
          bands          = [7];
          max_pdschReferenceSignalPower = -27;
-         max_rxgain                    = 115;
+         max_rxgain                    = 120;
          eNB_instances  = [0];
 
     }
@@ -227,7 +227,7 @@ RUs = (
 THREAD_STRUCT = (
   {
     #three config for level of parallelism "PARALLEL_SINGLE_THREAD", "PARALLEL_RU_L1_SPLIT", or "PARALLEL_RU_L1_TRX_SPLIT"
-    parallel_config    = "PARALLEL_RU_L1_TRX_SPLIT";
+    parallel_config    = "PARALLEL_SINGLE_THREAD";
     #two option for worker "WORKER_DISABLE" or "WORKER_ENABLE"
     worker_config      = "WORKER_ENABLE";
   }

ci-scripts/conf_files/rcc.band7.tm1.if4p5.lo.100PRB.usrpb210.conf

@@ -200,8 +200,8 @@ MACRLCs = (
 	tr_s_preference = "local_L1";
 	tr_n_preference = "local_RRC";
 	phy_test_mode = 0;
-        puSch10xSnr     =  200;
-        puCch10xSnr     =  200;
+        puSch10xSnr     =  160;
+        puCch10xSnr     =  160;
         }  
 );
 
@@ -234,7 +234,7 @@ RUs = (
 THREAD_STRUCT = (
   {
     #three config for level of parallelism "PARALLEL_SINGLE_THREAD", "PARALLEL_RU_L1_SPLIT", or "PARALLEL_RU_L1_TRX_SPLIT"
-    parallel_config    = "PARALLEL_RU_L1_TRX_SPLIT";
+    parallel_config    = "PARALLEL_SINGLE_THREAD";
     #two option for worker "WORKER_DISABLE" or "WORKER_ENABLE"
     worker_config      = "WORKER_ENABLE";
   }

ci-scripts/conf_files/rcc.band7.tm1.if4p5.lo.25PRB.usrpb210.conf

@@ -200,8 +200,8 @@ MACRLCs = (
 	tr_s_preference = "local_L1";
 	tr_n_preference = "local_RRC";
 	phy_test_mode = 0;
-        puSch10xSnr     =  200;
-        puCch10xSnr     =  200;
+        puSch10xSnr     =  160;
+        puCch10xSnr     =  160;
         }  
 );
 
@@ -234,7 +234,7 @@ RUs = (
 THREAD_STRUCT = (
   {
     #three config for level of parallelism "PARALLEL_SINGLE_THREAD", "PARALLEL_RU_L1_SPLIT", or "PARALLEL_RU_L1_TRX_SPLIT"
-    parallel_config    = "PARALLEL_RU_L1_TRX_SPLIT";
+    parallel_config    = "PARALLEL_SINGLE_THREAD";
     #two option for worker "WORKER_DISABLE" or "WORKER_ENABLE"
     worker_config      = "WORKER_ENABLE";
   }

ci-scripts/conf_files/rcc.band7.tm1.if4p5.lo.50PRB.usrpb210.conf

@@ -200,8 +200,8 @@ MACRLCs = (
 	tr_s_preference = "local_L1";
 	tr_n_preference = "local_RRC";
 	phy_test_mode = 0;
-        puSch10xSnr     =  200;
-        puCch10xSnr     =  200;
+        puSch10xSnr     =  160;
+        puCch10xSnr     =  160;
         }  
 );
 
@@ -234,7 +234,7 @@ RUs = (
 THREAD_STRUCT = (
   {
     #three config for level of parallelism "PARALLEL_SINGLE_THREAD", "PARALLEL_RU_L1_SPLIT", or "PARALLEL_RU_L1_TRX_SPLIT"
-    parallel_config    = "PARALLEL_RU_L1_TRX_SPLIT";
+    parallel_config    = "PARALLEL_SINGLE_THREAD";
     #two option for worker "WORKER_DISABLE" or "WORKER_ENABLE"
     worker_config      = "WORKER_ENABLE";
   }

ci-scripts/conf_files/rru.fdd.band7.conf

@@ -20,7 +20,7 @@ RUs = (
 THREAD_STRUCT = (
   {
     #three config for level of parallelism "PARALLEL_SINGLE_THREAD", "PARALLEL_RU_L1_SPLIT", or "PARALLEL_RU_L1_TRX_SPLIT"
-    parallel_config    = "PARALLEL_RU_L1_TRX_SPLIT";
+    parallel_config    = "PARALLEL_SINGLE_THREAD";
     #two option for worker "WORKER_DISABLE" or "WORKER_ENABLE"
     worker_config      = "WORKER_ENABLE";
   }

ci-scripts/main.py

@@ -1504,7 +1504,8 @@ class SSHConnection():
 			return ENB_PROCESS_ASSERTION
 		if foundRealTimeIssue:
 			logging.debug('\u001B[1;37;41m eNB faced real time issues! \u001B[0m')
-			return ENB_PROCESS_REALTIME_ISSUE
+			self.htmleNBFailureMsg += 'eNB faced real time issues!\n'
+			#return ENB_PROCESS_REALTIME_ISSUE
 		return 0
 
 	def TerminateeNB(self):

cmake_targets/build_oai

@@ -339,6 +339,10 @@ function main() {
             FORCE_DEADLINE_SCHEDULER_FLAG_USER="True"
             echo_info "Enabling the usage of deadline scheduler"
             shift 1;;
+       --enable-cpu-affinity)
+            CPU_AFFINITY_FLAG_USER="True"
+            echo_info "Enabling CPU Affinity (only valid when not using deadline scheduler)"
+            shift 1;;
        --disable-cpu-affinity)
             CPU_AFFINITY_FLAG_USER="False"
             echo_info "Disabling CPU Affinity (only valid when not using deadline scheduler)"

openair1/PHY/CODING/TESTBENCH/ldpctest.c

@@ -157,7 +157,9 @@ int test_ldpc(short No_iteration,
   //double code_rate_actual_vec[8] = {0.2, 0.33333, 0.4, 0.5, 0.66667, 0.73333, 0.81481, 0.88};
 
   t_nrLDPC_dec_params decParams;
-
+  t_nrLDPC_procBuf nrLDPC_procBuf;
+  t_nrLDPC_procBuf* p_nrLDPC_procBuf = &nrLDPC_procBuf;
+    
   t_nrLDPC_time_stats decoder_profiler;
   t_nrLDPC_time_stats* p_decoder_profiler =&decoder_profiler ;
 
@@ -210,6 +212,9 @@ int test_ldpc(short No_iteration,
   reset_meas(&decoder_profiler.llr2bit);
   //reset_meas(&decoder_profiler.total);
 
+  // Allocate LDPC decoder buffers
+  p_nrLDPC_procBuf = nrLDPC_init_mem();
+
   for (j=0;j<MAX_NUM_DLSCH_SEGMENTS;j++) {
     for (i=0; i<block_length/8; i++) {
       test_input[j][i]=(unsigned char) rand();
@@ -411,7 +416,7 @@ int test_ldpc(short No_iteration,
       // decoder supports BG2, Z=128 & 256
       //esimated_output=ldpc_decoder(channel_output_fixed, block_length, No_iteration, (double)((float)nom_rate/(float)denom_rate));
       ///nrLDPC_decoder(&decParams, channel_output_fixed, estimated_output, NULL);
-	n_iter = nrLDPC_decoder(&decParams, (int8_t*)channel_output_fixed[j], (int8_t*)estimated_output[j], p_decoder_profiler);
+          n_iter = nrLDPC_decoder(&decParams, (int8_t*)channel_output_fixed[j], (int8_t*)estimated_output[j], p_nrLDPC_procBuf, p_decoder_profiler);
       
 	stop_meas(time_decoder);
       }
@@ -486,6 +491,8 @@ int test_ldpc(short No_iteration,
   //free(channel_output_fixed);
   //free(estimated_output);
 
+  nrLDPC_free_mem(p_nrLDPC_procBuf);
+
   print_meas(&time,"ldpc_encoder",NULL,NULL);
   print_meas(time_optim,"ldpc_encoder_optim",NULL,NULL);
   print_meas(&tinput,"ldpc_encoder_optim(input)",NULL,NULL);

openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_bnProc.h

@@ -37,12 +37,17 @@
    \param p_lut Pointer to decoder LUTs
    \param Z Lifting size
 */
-static inline void nrLDPC_bnProcPc(t_nrLDPC_lut* p_lut, uint16_t Z)
+static inline void nrLDPC_bnProcPc(t_nrLDPC_lut* p_lut, t_nrLDPC_procBuf* p_procBuf, uint16_t Z)
 {
     const uint8_t*  lut_numBnInBnGroups = p_lut->numBnInBnGroups;
     const uint32_t* lut_startAddrBnGroups = p_lut->startAddrBnGroups;
     const uint16_t* lut_startAddrBnGroupsLlr = p_lut->startAddrBnGroupsLlr;
 
+    int8_t* bnProcBuf    = p_procBuf->bnProcBuf;
+    int8_t* bnProcBufRes = p_procBuf->bnProcBufRes;
+    int8_t* llrRes       = p_procBuf->llrRes;
+    int8_t* llrProcBuf   = p_procBuf->llrProcBuf;
+        
     __m128i* p_bnProcBuf;
     __m256i* p_bnProcBufRes;
     __m128i* p_llrProcBuf;
@@ -1676,7 +1681,7 @@ static inline void nrLDPC_bnProcPc(t_nrLDPC_lut* p_lut, uint16_t Z)
    \param p_lut Pointer to decoder LUTs
    \param Z Lifting size
 */
-static inline void nrLDPC_bnProc(t_nrLDPC_lut* p_lut, uint16_t Z)
+static inline void nrLDPC_bnProc(t_nrLDPC_lut* p_lut, t_nrLDPC_procBuf* p_procBuf, uint16_t Z)
 {
     // BN Processing calculating the values to send back to the CNs for next iteration
     // bnProcBufRes contains the sum of all edges to each BN at the start of each group
@@ -1685,6 +1690,10 @@ static inline void nrLDPC_bnProc(t_nrLDPC_lut* p_lut, uint16_t Z)
     const uint32_t* lut_startAddrBnGroups = p_lut->startAddrBnGroups;
     const uint16_t* lut_startAddrBnGroupsLlr = p_lut->startAddrBnGroupsLlr;
 
+    int8_t* bnProcBuf    = p_procBuf->bnProcBuf;
+    int8_t* bnProcBufRes = p_procBuf->bnProcBufRes;
+    int8_t* llrRes       = p_procBuf->llrRes;
+    
     __m256i* p_bnProcBuf;
     __m256i* p_bnProcBufRes;
     __m256i* p_llrRes;

openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_cnProc.h

@@ -36,11 +36,14 @@
    \param p_lut Pointer to decoder LUTs
    \param Z Lifting size
 */
-static inline void nrLDPC_cnProc_BG2(t_nrLDPC_lut* p_lut, uint16_t Z)
+static inline void nrLDPC_cnProc_BG2(t_nrLDPC_lut* p_lut, t_nrLDPC_procBuf* p_procBuf, uint16_t Z)
 {
     const uint8_t*  lut_numCnInCnGroups   = p_lut->numCnInCnGroups;
     const uint32_t* lut_startAddrCnGroups = p_lut->startAddrCnGroups;
 
+    int8_t* cnProcBuf    = p_procBuf->cnProcBuf;
+    int8_t* cnProcBufRes = p_procBuf->cnProcBufRes;
+    
     __m256i* p_cnProcBuf;
     __m256i* p_cnProcBufRes;
 
@@ -362,11 +365,14 @@ static inline void nrLDPC_cnProc_BG2(t_nrLDPC_lut* p_lut, uint16_t Z)
    \param p_lut Pointer to decoder LUTs
    \param Z Lifting size
 */
-static inline void nrLDPC_cnProc_BG1(t_nrLDPC_lut* p_lut, uint16_t Z)
+static inline void nrLDPC_cnProc_BG1(t_nrLDPC_lut* p_lut, t_nrLDPC_procBuf* p_procBuf, uint16_t Z)
 {
     const uint8_t*  lut_numCnInCnGroups   = p_lut->numCnInCnGroups;
     const uint32_t* lut_startAddrCnGroups = p_lut->startAddrCnGroups;
 
+    int8_t* cnProcBuf    = p_procBuf->cnProcBuf;
+    int8_t* cnProcBufRes = p_procBuf->cnProcBufRes;
+    
     __m256i* p_cnProcBuf;
     __m256i* p_cnProcBufRes;
 
@@ -858,11 +864,14 @@ static inline void nrLDPC_cnProc_BG1(t_nrLDPC_lut* p_lut, uint16_t Z)
    \param Z Lifting size
    \return 32-bit parity check indicator
 */
-static inline uint32_t nrLDPC_cnProcPc_BG1(t_nrLDPC_lut* p_lut, uint16_t Z)
+static inline uint32_t nrLDPC_cnProcPc_BG1(t_nrLDPC_lut* p_lut, t_nrLDPC_procBuf* p_procBuf, uint16_t Z)
 {
     const uint8_t*  lut_numCnInCnGroups   = p_lut->numCnInCnGroups;
     const uint32_t* lut_startAddrCnGroups = p_lut->startAddrCnGroups;
 
+    int8_t* cnProcBuf    = p_procBuf->cnProcBuf;
+    int8_t* cnProcBufRes = p_procBuf->cnProcBufRes;
+    
     __m256i* p_cnProcBuf;
     __m256i* p_cnProcBufRes;
 
@@ -1490,11 +1499,14 @@ static inline uint32_t nrLDPC_cnProcPc_BG1(t_nrLDPC_lut* p_lut, uint16_t Z)
    \param Z Lifting size
    \return 32-bit parity check indicator
 */
-static inline uint32_t nrLDPC_cnProcPc_BG2(t_nrLDPC_lut* p_lut, uint16_t Z)
+static inline uint32_t nrLDPC_cnProcPc_BG2(t_nrLDPC_lut* p_lut, t_nrLDPC_procBuf* p_procBuf, uint16_t Z)
 {
     const uint8_t*  lut_numCnInCnGroups   = p_lut->numCnInCnGroups;
     const uint32_t* lut_startAddrCnGroups = p_lut->startAddrCnGroups;
 
+    int8_t* cnProcBuf    = p_procBuf->cnProcBuf;
+    int8_t* cnProcBufRes = p_procBuf->cnProcBufRes;
+    
     __m256i* p_cnProcBuf;
     __m256i* p_cnProcBufRes;
 

openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_decoder.h

@@ -32,6 +32,7 @@
 #define __NR_LDPC_DECODER__H__
 
 #include "nrLDPC_types.h"
+#include "nrLDPC_init_mem.h"
 
 /**
    \brief LDPC decoder
@@ -40,6 +41,6 @@
    \param p_llrOut Output vector
    \param p_profiler LDPC profiler statistics
 */
-int32_t nrLDPC_decoder(t_nrLDPC_dec_params* p_decParams, int8_t* p_llr, int8_t* p_llrOut, t_nrLDPC_time_stats* p_profiler);
+int32_t nrLDPC_decoder(t_nrLDPC_dec_params* p_decParams, int8_t* p_llr, int8_t* p_llrOut, t_nrLDPC_procBuf* p_procBuf, t_nrLDPC_time_stats* p_profiler);
 
 #endif

openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_defs.h

@@ -100,23 +100,7 @@
 #define NR_LDPC_MAX_NUM_LLR 26112
 
 // ==============================================================================
-// GLOBAL VARIABLES
-
-// Aligned on 32 bytes = 256 bits for AVX2
-/** CN processing buffer */
-static int8_t cnProcBuf   [NR_LDPC_SIZE_CN_PROC_BUF] __attribute__ ((aligned(32)));
-/** Buffer for CN processing results */
-static int8_t cnProcBufRes[NR_LDPC_SIZE_CN_PROC_BUF] __attribute__ ((aligned(32)));
-
-/** BN processing buffer */
-static int8_t bnProcBuf   [NR_LDPC_SIZE_BN_PROC_BUF] __attribute__ ((aligned(32)));
-/** Buffer for BN processing results */
-static int8_t bnProcBufRes[NR_LDPC_SIZE_BN_PROC_BUF] __attribute__ ((aligned(32)));
-
-/** Buffer for LLR results */
-static int8_t llrRes    [NR_LDPC_MAX_NUM_LLR] __attribute__ ((aligned(32)));
-/** LLR processing buffer */
-static int8_t llrProcBuf[NR_LDPC_MAX_NUM_LLR] __attribute__ ((aligned(32)));
+// GLOBAL CONSTANT VARIABLES
 
 /** Start addresses for the cnProcBuf for each CN group in BG1*/
 static const uint32_t lut_startAddrCnGroups_BG1[NR_LDPC_NUM_CN_GROUPS_BG1] = {0, 1152, 8832, 43392, 61824, 75264, 81408, 88320, 92160};

openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_init_mem.h

+/*
+ * Licensed to the OpenAirInterface (OAI) Software Alliance under one or more
+ * contributor license agreements.  See the NOTICE file distributed with
+ * this work for additional information regarding copyright ownership.
+ * The OpenAirInterface Software Alliance licenses this file to You under
+ * the OAI Public License, Version 1.1  (the "License"); you may not use this file
+ * except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *      http://www.openairinterface.org/?page_id=698
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ *-------------------------------------------------------------------------------
+ * For more information about the OpenAirInterface (OAI) Software Alliance:
+ *      contact@openairinterface.org
+ */
+
+/*!\file nrLDPC_init_mem.h
+ * \brief Defines the function to initialize the LDPC decoder and sets correct LUTs.
+ * \author Sebastian Wagner (TCL Communications) Email: <mailto:sebastian.wagner@tcl.com>
+ * \date 07-12-2018
+ * \version 1.0
+ * \note
+ * \warning
+ */
+
+#ifndef __NR_LDPC_INIT_MEM__H__
+#define __NR_LDPC_INIT_MEM__H__
+
+#include <stdlib.h>
+#include "nrLDPC_defs.h"
+#include "nrLDPC_types.h"
+
+#ifndef malloc32_clear
+/**
+   \brief Allocates 32 byte aligned memory and initializes to zero
+   \param size Input size in bytes
+   \return Pointer to memory
+*/
+static inline void* malloc32_clear(size_t size)
+{
+    void* ptr = (void*) memalign(32, size+32);
+    memset(ptr, 0, size);
+    return ptr;
+}
+#endif
+
+/**
+   \brief Allocates and initializes the internal decoder processing buffers
+   \param p_decParams Pointer to decoder parameters
+   \param p_lut Pointer to decoder LUTs
+   \return Number of LLR values
+*/
+static inline t_nrLDPC_procBuf* nrLDPC_init_mem(void)
+{
+    t_nrLDPC_procBuf* p_procBuf = (t_nrLDPC_procBuf*) malloc32_clear(sizeof(t_nrLDPC_procBuf));
+
+    if (p_procBuf)
+    {
+        p_procBuf->cnProcBuf    = (int8_t*) malloc32_clear(NR_LDPC_SIZE_CN_PROC_BUF*sizeof(int8_t));
+        p_procBuf->cnProcBufRes = (int8_t*) malloc32_clear(NR_LDPC_SIZE_CN_PROC_BUF*sizeof(int8_t));
+        p_procBuf->bnProcBuf    = (int8_t*) malloc32_clear(NR_LDPC_SIZE_BN_PROC_BUF*sizeof(int8_t));
+        p_procBuf->bnProcBufRes = (int8_t*) malloc32_clear(NR_LDPC_SIZE_BN_PROC_BUF*sizeof(int8_t));
+        p_procBuf->llrRes       = (int8_t*) malloc32_clear(NR_LDPC_MAX_NUM_LLR     *sizeof(int8_t));
+        p_procBuf->llrProcBuf   = (int8_t*) malloc32_clear(NR_LDPC_MAX_NUM_LLR     *sizeof(int8_t));
+    }
+
+    return(p_procBuf);
+}
+
+static inline void nrLDPC_free_mem(t_nrLDPC_procBuf* p_procBuf)
+{
+    free(p_procBuf->cnProcBuf);
+    free(p_procBuf->cnProcBufRes);
+    free(p_procBuf->bnProcBuf);
+    free(p_procBuf->bnProcBufRes);
+    free(p_procBuf->llrRes);
+    free(p_procBuf->llrProcBuf);
+
+    free(p_procBuf);
+}
+#endif

openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_mPass.h

@@ -41,13 +41,15 @@
    \param Z Lifting size
    \param BG Base graph
 */
-static inline void nrLDPC_llr2llrProcBuf(t_nrLDPC_lut* p_lut, int8_t* llr, uint16_t Z, uint8_t BG)
+static inline void nrLDPC_llr2llrProcBuf(t_nrLDPC_lut* p_lut, int8_t* llr, t_nrLDPC_procBuf* p_procBuf, uint16_t Z, uint8_t BG)
 {
     const uint16_t* lut_llr2llrProcBuf = p_lut->llr2llrProcBuf;
     uint32_t i;
     const uint8_t numBn2CnG1 = p_lut->numBnInBnGroups[0];
     uint32_t colG1 = NR_LDPC_START_COL_PARITY_BG1*Z;
 
+    int8_t* llrProcBuf = p_procBuf->llrProcBuf;
+    
     if (BG == 2)
     {
         colG1 = NR_LDPC_START_COL_PARITY_BG2*Z;
@@ -74,11 +76,13 @@ static inline void nrLDPC_llr2llrProcBuf(t_nrLDPC_lut* p_lut, int8_t* llr, uint1
    \param Z Lifting size
    \param BG Base graph
 */
-static inline void nrLDPC_llr2CnProcBuf(t_nrLDPC_lut* p_lut, int8_t* llr, uint16_t numLLR, uint16_t Z, uint8_t BG)
+static inline void nrLDPC_llr2CnProcBuf(t_nrLDPC_lut* p_lut, int8_t* llr, t_nrLDPC_procBuf* p_procBuf, uint16_t numLLR, uint16_t Z, uint8_t BG)
 {
     const uint32_t* lut_llr2CnProcBuf = p_lut->llr2CnProcBuf;
     const uint8_t* lut_numEdgesPerBn = p_lut->numEdgesPerBn;
 
+    int8_t* cnProcBuf = p_procBuf->cnProcBuf;
+    
     int8_t curLLR;
     uint8_t numEdges;
     uint32_t i;
@@ -123,12 +127,15 @@ static inline void nrLDPC_llr2CnProcBuf(t_nrLDPC_lut* p_lut, int8_t* llr, uint16
    \param p_lut Pointer to decoder LUTs
    \param Z Lifting size
 */
-static inline void nrLDPC_cn2bnProcBuf(t_nrLDPC_lut* p_lut, uint16_t Z)
+static inline void nrLDPC_cn2bnProcBuf(t_nrLDPC_lut* p_lut, t_nrLDPC_procBuf* p_procBuf, uint16_t Z)
 {
     const uint32_t* lut_cn2bnProcBuf = p_lut->cn2bnProcBuf;
     const uint8_t*  lut_numCnInCnGroups = p_lut->numCnInCnGroups;
     const uint32_t* lut_startAddrCnGroups = p_lut->startAddrCnGroups;
 
+    int8_t* cnProcBufRes = p_procBuf->cnProcBufRes;
+    int8_t* bnProcBuf    = p_procBuf->bnProcBuf;
+    
     const uint32_t* p_lut_cn2bn;
     int8_t* p_cnProcBufRes;
     uint32_t bitOffsetInGroup;
@@ -245,12 +252,15 @@ static inline void nrLDPC_cn2bnProcBuf(t_nrLDPC_lut* p_lut, uint16_t Z)
    \param p_lut Pointer to decoder LUTs
    \param Z Lifting size
 */
-static inline void nrLDPC_cn2bnProcBuf_BG1(t_nrLDPC_lut* p_lut, uint16_t Z)
+static inline void nrLDPC_cn2bnProcBuf_BG1(t_nrLDPC_lut* p_lut, t_nrLDPC_procBuf* p_procBuf, uint16_t Z)
 {
     const uint32_t* lut_cn2bnProcBuf = p_lut->cn2bnProcBuf;
     const uint8_t*  lut_numCnInCnGroups = p_lut->numCnInCnGroups;
     const uint32_t* lut_startAddrCnGroups = p_lut->startAddrCnGroups;
 
+    int8_t* cnProcBufRes = p_procBuf->cnProcBufRes;
+    int8_t* bnProcBuf    = p_procBuf->bnProcBuf;
+    
     const uint32_t* p_lut_cn2bn;
     int8_t* p_cnProcBufRes;
     uint32_t bitOffsetInGroup;
@@ -418,12 +428,15 @@ static inline void nrLDPC_cn2bnProcBuf_BG1(t_nrLDPC_lut* p_lut, uint16_t Z)
    \param p_lut Pointer to decoder LUTs
    \param Z Lifting size
 */
-static inline void nrLDPC_bn2cnProcBuf(t_nrLDPC_lut* p_lut, uint16_t Z)
+static inline void nrLDPC_bn2cnProcBuf(t_nrLDPC_lut* p_lut, t_nrLDPC_procBuf* p_procBuf, uint16_t Z)
 {
     const uint32_t* lut_cn2bnProcBuf = p_lut->cn2bnProcBuf;
     const uint8_t*  lut_numCnInCnGroups = p_lut->numCnInCnGroups;
     const uint32_t* lut_startAddrCnGroups = p_lut->startAddrCnGroups;
 
+    int8_t* cnProcBuf = p_procBuf->cnProcBuf;
+    int8_t* bnProcBufRes = p_procBuf->bnProcBufRes;
+    
     int8_t* p_cnProcBuf;
     const uint32_t* p_lut_cn2bn;
     uint32_t bitOffsetInGroup;
@@ -543,12 +556,15 @@ static inline void nrLDPC_bn2cnProcBuf(t_nrLDPC_lut* p_lut, uint16_t Z)
    \param p_lut Pointer to decoder LUTs
    \param Z Lifting size
 */
-static inline void nrLDPC_bn2cnProcBuf_BG1(t_nrLDPC_lut* p_lut, uint16_t Z)
+static inline void nrLDPC_bn2cnProcBuf_BG1(t_nrLDPC_lut* p_lut, t_nrLDPC_procBuf* p_procBuf, uint16_t Z)
 {
     const uint32_t* lut_cn2bnProcBuf = p_lut->cn2bnProcBuf;
     const uint8_t*  lut_numCnInCnGroups = p_lut->numCnInCnGroups;
     const uint32_t* lut_startAddrCnGroups = p_lut->startAddrCnGroups;
 
+    int8_t* cnProcBuf    = p_procBuf->cnProcBuf;
+    int8_t* bnProcBufRes = p_procBuf->bnProcBufRes;
+    
     int8_t* p_cnProcBuf;
     const uint32_t* p_lut_cn2bn;
     uint32_t bitOffsetInGroup;
@@ -720,11 +736,13 @@ static inline void nrLDPC_bn2cnProcBuf_BG1(t_nrLDPC_lut* p_lut, uint16_t Z)
    \param llrOut Pointer to output LLRs
    \param numLLR Number of LLR values
 */
-static inline void nrLDPC_llrRes2llrOut(t_nrLDPC_lut* p_lut, int8_t* llrOut, uint16_t numLLR)
+static inline void nrLDPC_llrRes2llrOut(t_nrLDPC_lut* p_lut, int8_t* llrOut, t_nrLDPC_procBuf* p_procBuf, uint16_t numLLR)
 {
     const uint16_t* lut_llr2llrProcBuf = p_lut->llr2llrProcBuf;
     uint32_t i;
 
+    int8_t* llrRes = p_procBuf->llrRes;
+    
     for (i=0; i<numLLR; i++)
     {
         llrOut[i] = llrRes[lut_llr2llrProcBuf[i]];

openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_tools/nrLDPC_debug.h

@@ -87,38 +87,38 @@ static inline void nrLDPC_initFile(const char* fileName)
    \brief Writes data of predefined buffers to file
    \param buffer Enum of buffer name to write
 */
-static inline void nrLDPC_debug_writeBuffer2File(e_nrLDPC_buffers buffer)
+static inline void nrLDPC_debug_writeBuffer2File(e_nrLDPC_buffers buffer, t_nrLDPC_procBuf* p_procBuf)
 {
     switch (buffer)
     {
     case nrLDPC_buffers_LLR_PROC:
     {
-        nrLDPC_writeFile("llrProcBuf.txt", llrProcBuf, NR_LDPC_MAX_NUM_LLR);
+        nrLDPC_writeFile("llrProcBuf.txt", p_procBuf->llrProcBuf, NR_LDPC_MAX_NUM_LLR);
         break;
     }
     case nrLDPC_buffers_CN_PROC:
     {
-        nrLDPC_writeFile("cnProcBuf.txt", cnProcBuf, NR_LDPC_SIZE_CN_PROC_BUF);
+        nrLDPC_writeFile("cnProcBuf.txt", p_procBuf->cnProcBuf, NR_LDPC_SIZE_CN_PROC_BUF);
         break;
     }
     case nrLDPC_buffers_CN_PROC_RES:
     {
-        nrLDPC_writeFile("cnProcBufRes.txt", cnProcBufRes, NR_LDPC_SIZE_CN_PROC_BUF);
+        nrLDPC_writeFile("cnProcBufRes.txt", p_procBuf->cnProcBufRes, NR_LDPC_SIZE_CN_PROC_BUF);
         break;
     }
     case nrLDPC_buffers_BN_PROC:
     {
-        nrLDPC_writeFile("bnProcBuf.txt", bnProcBuf, NR_LDPC_SIZE_BN_PROC_BUF);
+        nrLDPC_writeFile("bnProcBuf.txt", p_procBuf->bnProcBuf, NR_LDPC_SIZE_BN_PROC_BUF);
         break;
     }
     case nrLDPC_buffers_BN_PROC_RES:
     {
-        nrLDPC_writeFile("bnProcBufRes.txt", bnProcBufRes, NR_LDPC_SIZE_BN_PROC_BUF);
+        nrLDPC_writeFile("bnProcBufRes.txt", p_procBuf->bnProcBufRes, NR_LDPC_SIZE_BN_PROC_BUF);
         break;
     }
     case nrLDPC_buffers_LLR_RES:
     {
-        nrLDPC_writeFile("llrRes.txt", llrRes, NR_LDPC_MAX_NUM_LLR);
+        nrLDPC_writeFile("llrRes.txt", p_procBuf->llrRes, NR_LDPC_MAX_NUM_LLR);
         break;
     }
     }

openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_types.h

@@ -88,4 +88,16 @@ typedef struct nrLDPC_time_stats {
     time_stats_t total; /**< Statistics for total processing time */
 } t_nrLDPC_time_stats;
 
+/**
+   Structure containing the processing buffers
+ */
+typedef struct nrLDPC_procBuf {
+    int8_t* cnProcBuf; /**< CN processing buffer */
+    int8_t* cnProcBufRes; /**< Buffer for CN processing results */
+    int8_t* bnProcBuf; /**< BN processing buffer */
+    int8_t* bnProcBufRes; /**< Buffer for BN processing results */
+    int8_t* llrRes; /**< Buffer for LLR results */
+    int8_t* llrProcBuf; /**< LLR processing buffer */
+} t_nrLDPC_procBuf;
+
 #endif

openair1/PHY/INIT/nr_init_ue.c

@@ -943,9 +943,7 @@ void init_nr_ue_transport(PHY_VARS_NR_UE *ue,int abstraction_flag) {
 void phy_init_nr_top(PHY_VARS_NR_UE *ue)
 {
   NR_DL_FRAME_PARMS *frame_parms = &ue->frame_parms;
-  NR_UE_DLSCH_t *dlsch0 = ue->dlsch[0][0][0];
-  dlsch0 =(NR_UE_DLSCH_t *)malloc16(sizeof(NR_UE_DLSCH_t));
-  
+    
   crcTableInit();
 
   init_dfts();

openair1/PHY/NR_TRANSPORT/nr_transport_common_proto.h

+
 /*
  * Licensed to the OpenAirInterface (OAI) Software Alliance under one or more
  * contributor license agreements.  See the NOTICE file distributed with
@@ -33,7 +34,7 @@
 #ifndef __NR_TRANSPORT_COMMON_PROTO__H__
 #define __NR_TRANSPORT_COMMON_PROTO__H__
 
-#define MAX_NUM_NR_DLSCH_SEGMENTS 32
+#define MAX_NUM_NR_DLSCH_SEGMENTS 16
 #define MAX_NUM_NR_ULSCH_SEGMENTS MAX_NUM_NR_DLSCH_SEGMENTS
 
 #define MAX_NR_DLSCH_PAYLOAD_BYTES (MAX_NUM_NR_DLSCH_SEGMENTS*1056)

openair1/PHY/NR_UE_TRANSPORT/nr_transport_ue.h

@@ -34,7 +34,7 @@
 #include <limits.h>
 #include "PHY/impl_defs_top.h"
 
-
+#include "PHY/CODING/nrLDPC_decoder/nrLDPC_types.h"
 
 //#include "PHY/defs_nr_UE.h"
 #include "../NR_TRANSPORT/nr_transport_common_proto.h"
@@ -267,6 +267,8 @@ typedef struct {
   double w_abs[MAX_NUM_NR_DLSCH_SEGMENTS][3*8448];
   /// soft bits for each received segment ("d"-sequence)(for definition see 36-212 V8.6 2009-03, p.15)
   int16_t *d[MAX_NUM_NR_DLSCH_SEGMENTS];
+  /// LDPC processing buffers
+  t_nrLDPC_procBuf* p_nrLDPC_procBuf[MAX_NUM_DLSCH_SEGMENTS];
   /// Number of code segments 
   uint32_t C;
   /// Number of bits in code segments
@@ -353,9 +355,7 @@ typedef struct {
   /// Maximum number of LDPC iterations
   uint8_t max_ldpc_iterations;
   /// number of iterations used in last turbo decoding
-  uint8_t last_iteration_cnt;
-  
-
+  uint8_t last_iteration_cnt;  
 } NR_UE_DLSCH_t;
 
 typedef enum {format0_0,

targets/ARCH/ADRV9371_ZC706/SYRTEM_IQ_QuickStart.readme

@@ -115,8 +115,16 @@ sudo cset shield --reset
 cd eur_oai_develop-nr/openairinterface5g/
 source oaienv
 
-sudo -E ./cmake_targets/build_oai -c --gNB -w ADRV9371_ZC706 -I
-sudo -E ./cmake_targets/build_oai -c --gNB -w ADRV9371_ZC706
+sudo -E ./cmake_targets/build_oai -c --gNB -w ADRV9371_ZC706 -I --enable-cpu-affinity
+
+/* CPU available shall be mapped on CPU2 to get_nprocs() number with hyperthreading disabled */
+/* because : */
+/* - cset shield leave CPU0 for system */
+/* - CPU1 is reserved for HwRxTx thread for I/Q acquisition ! */
+/* Note: "/usr/local/etc/syriq/cpu-irq.sh" shall be changed if more than 8 cores are used */
+/*   ie. "cset shield --force --kthread on -c 1-7" where "-c 1-7" reserves CPU1 to CPU7 for application (CPU0 is for system) */
+
+sudo -E ./cmake_targets/build_oai -c --gNB -w ADRV9371_ZC706 --enable-cpu-affinity
 
 
 

targets/COMMON/threads_t.h

@@ -2,19 +2,20 @@
 #define _THREADS_T_H_
 
 typedef struct threads_s {
-    int iq;
-    int one;
-    int two;
-    int three;
-    int slot1_proc_one;
-    int slot1_proc_two;
-    int slot1_proc_three;
-    int dlsch_td_one;
-	int dlsch_td_two;
-	int dlsch_td_three;
-	int dlsch_td1_one;
-	int dlsch_td1_two;
-	int dlsch_td1_three;
+  int main;
+  int sync;
+  int one;
+  int two;
+  int three;
+  int slot1_proc_one;
+  int slot1_proc_two;
+  int slot1_proc_three;
+  //int dlsch_td_one;
+  //int dlsch_td_two;
+  //int dlsch_td_three;
+  //int dlsch_td1_one;
+  //int dlsch_td1_two;
+  //int dlsch_td1_three;
 } threads_t;
 
 #endif /* _THREADS_T_H_ */

targets/RT/USER/lte-softmodem.c

@@ -253,7 +253,7 @@ extern char uecap_xer[1024];
 char uecap_xer_in=0;
 
 int oaisim_flag=0;
-threads_t threads= {-1,-1,-1,-1,-1,-1,-1};
+threads_t threads= {-1,-1,-1,-1,-1,-1,-1,-1};
 
 /* see file openair2/LAYER2/MAC/main.c for why abstraction_flag is needed
  * this is very hackish - find a proper solution
@@ -906,14 +906,16 @@ int main( int argc, char **argv )
   char cpu_affinity[1024];
   CPU_ZERO(&cpuset);
 #ifdef CPU_AFFINITY
+  int j;
   if (get_nprocs() > 2) {
-    CPU_SET(0, &cpuset);
+    for (j = 2; j < get_nprocs(); j++)
+      CPU_SET(j, &cpuset);
     s = pthread_setaffinity_np(pthread_self(), sizeof(cpu_set_t), &cpuset);
     if (s != 0) {
       perror( "pthread_setaffinity_np");
       exit_fun("Error setting processor affinity");
     }
-    LOG_I(HW, "Setting the affinity of main function to CPU 0, for device library to use CPU 0 only!\n");
+    LOG_I(HW, "Setting the affinity of main function to all CPUs, for device library to use CPU 0 only!\n");
   }
 #endif
   

targets/RT/USER/lte-softmodem.h

@@ -169,7 +169,8 @@
 {"clock",                   CONFIG_HLP_CLK,         0,                      uptr:&clock_source,                 defintval:0,                    TYPE_UINT,      0},                     \
 {"wait-for-sync",           NULL,                   PARAMFLAG_BOOL,         iptr:&wait_for_sync,                defintval:0,                    TYPE_INT,       0},                     \
 {"single-thread-enable",    CONFIG_HLP_NOSNGLT,     PARAMFLAG_BOOL,         iptr:&single_thread_flag,           defintval:0,                    TYPE_INT,       0},                     \
-{"threadIQ",                NULL,                   0,                      iptr:&(threads.iq),                 defintval:1,                    TYPE_INT,       0},                     \
+{"threadMain",                NULL,                   0,                      iptr:&(threads.main),                 defintval:1,                    TYPE_INT,       0},                     \
+{"threadSync",                NULL,                   0,                      iptr:&(threads.sync),                 defintval:1,                    TYPE_INT,       0},                     \
 {"threadOneSubframe",       NULL,                   0,                      iptr:&(threads.one),                defintval:1,                    TYPE_INT,       0},                     \
 {"threadTwoSubframe",       NULL,                   0,                      iptr:&(threads.two),                defintval:1,                    TYPE_INT,       0},                     \
 {"threadThreeSubframe",     NULL,                   0,                      iptr:&(threads.three),              defintval:1,                    TYPE_INT,       0},                     \

targets/RT/USER/lte-ue.c

@@ -456,8 +456,8 @@ static void *UE_thread_synch(void *arg)
 
   cpu_set_t cpuset;
   CPU_ZERO(&cpuset);
-  if ( threads.iq != -1 )
-    CPU_SET(threads.iq, &cpuset);
+  if ( threads.sync != -1 )
+    CPU_SET(threads.sync, &cpuset);
   // this thread priority must be lower that the main acquisition thread
   sprintf(threadname, "sync UE %d\n", UE->Mod_id);
   init_thread(100000, 500000, FIFO_PRIORITY-1, &cpuset, threadname);
@@ -1435,8 +1435,8 @@ void *UE_thread(void *arg) {
 
   cpu_set_t cpuset;
   CPU_ZERO(&cpuset);
-  if ( threads.iq != -1 )
-    CPU_SET(threads.iq, &cpuset);
+  if ( threads.main != -1 )
+    CPU_SET(threads.main, &cpuset);
   init_thread(100000, 500000, FIFO_PRIORITY, &cpuset,
 	      "UHD Threads");
 

targets/RT/USER/lte-uesoftmodem.c

@@ -260,7 +260,7 @@ extern char uecap_xer[1024];
 char uecap_xer_in=0;
 
 int oaisim_flag=0;
-threads_t threads= {-1,-1,-1,-1,-1,-1,-1};
+threads_t threads = {-1,-1,-1,-1,-1,-1,-1};
 
 /* see file openair2/LAYER2/MAC/main.c for why abstraction_flag is needed
  * this is very hackish - find a proper solution
@@ -1014,14 +1014,17 @@ printf("~~~~~~~~~~~~~~~~~~~~successfully get the parallel config[%d], worker con
   char cpu_affinity[1024];
   CPU_ZERO(&cpuset);
 #ifdef CPU_AFFINITY
+  int j;
   if (get_nprocs() > 2) {
-    CPU_SET(0, &cpuset);
+    for (j = 2; j < get_nprocs(); j++)
+      CPU_SET(j, &cpuset);
+    
     s = pthread_setaffinity_np(pthread_self(), sizeof(cpu_set_t), &cpuset);
     if (s != 0) {
       perror( "pthread_setaffinity_np");
       exit_fun("Error setting processor affinity");
     }
-    LOG_I(HW, "Setting the affinity of main function to CPU 0, for device library to use CPU 0 only!\n");
+    LOG_I(HW, "Setting the affinity of main function to all CPUs, for device library to use CPU 0 only!\n");
   }
 #endif
   

targets/RT/USER/nr-ru.c

@@ -1402,7 +1402,7 @@ static void* ru_thread( void* param ) {
   int                ret;
   int                slot = fp->slots_per_frame-1;
   int                frame    =1023; 
-  char               filename[40];
+  char               filename[40],threadname[40];
   int                print_frame = 8;
   int                i = 0;
 
@@ -1411,7 +1411,8 @@ static void* ru_thread( void* param ) {
 
 
   // set default return value
-  thread_top_init("ru_thread",0,870000,1000000,1000000);
+  sprintf(threadname,"ru_thread %d",ru->idx);
+  thread_top_init(threadname,0,870000,1000000,1000000);
 
   LOG_I(PHY,"Starting RU %d (%s,%s),\n",ru->idx,NB_functions[ru->function],NB_timing[ru->if_timing]);
 

targets/RT/USER/nr-softmodem.c

@@ -247,8 +247,6 @@ double cpuf;
 extern char uecap_xer[1024];
 char uecap_xer_in=0;
 
-threads_t threads= {-1,-1,-1,-1,-1,-1,-1};
-
 /* see file openair2/LAYER2/MAC/main.c for why abstraction_flag is needed
  * this is very hackish - find a proper solution
  */
@@ -1034,14 +1032,19 @@ int main( int argc, char **argv )
   char cpu_affinity[1024];
   CPU_ZERO(&cpuset);
 #ifdef CPU_AFFINITY
+  int j;
   if (get_nprocs() > 2) {
-    CPU_SET(0, &cpuset);
+    // CPU_SET(1, &cpuset);
+    for (j = 2; j < get_nprocs(); j++)
+    {
+      CPU_SET(j, &cpuset);
+    }
     s = pthread_setaffinity_np(pthread_self(), sizeof(cpu_set_t), &cpuset);
     if (s != 0) {
       perror( "pthread_setaffinity_np");
       exit_fun("Error setting processor affinity");
     }
-    LOG_I(HW, "Setting the affinity of main function to CPU 0, for device library to use CPU 0 only!\n");
+    LOG_I(HW, "Setting the affinity of main function to all CPUs, for device library to use CPU 0 only!\n");
   }
 #endif
   

targets/RT/USER/nr-softmodem.h

@@ -152,12 +152,6 @@
 {"external-clock",        	 CONFIG_HLP_EXCCLK,	PARAMFLAG_BOOL,   uptr:&clock_source,			defintval:0,			   TYPE_INT,	  0},			   \
 {"wait-for-sync",         	 NULL,  		PARAMFLAG_BOOL,   iptr:&wait_for_sync,  		defintval:0,			   TYPE_INT,	  0},			   \
 {"single-thread-disable", 	 CONFIG_HLP_NOSNGLT,	PARAMFLAG_BOOL,   iptr:&single_thread_flag,		defintval:1,			   TYPE_INT,	  0},			   \
-{"threadIQ",              	 NULL,  		0,		  iptr:&(threads.iq),			defintval:1,			   TYPE_INT,	  0},			   \
-{"threadOneSubframe",     	 NULL,  		0,		  iptr:&(threads.one),  		defintval:1,			   TYPE_INT,	  0},			   \
-{"threadTwoSubframe",    	 NULL,  		0,		  iptr:&(threads.two),  		defintval:1,			   TYPE_INT,	  0},			   \
-{"threadThreeSubframe",    	 NULL,  		0,		  iptr:&(threads.three),  		defintval:1,			   TYPE_INT,	  0},			   \
-{"threadSlot1ProcOne",     	 NULL,  		0,		  iptr:&(threads.slot1_proc_one),      	defintval:1,			   TYPE_INT,	  0},			   \
-{"threadSlot1ProcTwo",    	 NULL,  		0,		  iptr:&(threads.slot1_proc_two),      	defintval:1,			   TYPE_INT,	  0},			   \
 {"A" ,  		  	 CONFIG_HLP_TADV,	0,		  uptr:&timing_advance, 		defintval:0,			   TYPE_UINT,	  0},			   \
 {"C" ,  		  	 CONFIG_HLP_DLF,	0,		  uptr:&(downlink_frequency[0][0]),	defuintval:DEFAULT_DLF,  	   TYPE_UINT,	  0},			   \
 {"a" ,  		  	 CONFIG_HLP_CHOFF,	0,		  iptr:&chain_offset,			defintval:0,			   TYPE_INT,	  0},			   \

targets/RT/USER/nr-ue.c

@@ -341,8 +341,8 @@ static void *UE_thread_synch(void *arg) {
 
     cpu_set_t cpuset;
     CPU_ZERO(&cpuset);
-    if ( threads.iq != -1 )
-        CPU_SET(threads.iq, &cpuset);
+    if ( threads.sync != -1 )
+        CPU_SET(threads.sync, &cpuset);
     // this thread priority must be lower that the main acquisition thread
     sprintf(threadname, "sync UE %d", UE->Mod_id);
     init_thread(100000, 500000, FIFO_PRIORITY-1, &cpuset, threadname);
@@ -797,15 +797,15 @@ void *UE_thread(void *arg) {
 
     cpu_set_t cpuset;
     CPU_ZERO(&cpuset);
-    if ( threads.iq != -1 )
-        CPU_SET(threads.iq, &cpuset);
-    init_thread(100000, 500000, FIFO_PRIORITY, &cpuset,
-                "UHD Threads");
+    if ( threads.main != -1 )
+        CPU_SET(threads.main, &cpuset);
+    sprintf(threadname, "Main UE %d", UE->Mod_id);
+    init_thread(100000, 500000, FIFO_PRIORITY, &cpuset,threadname);
+    
     if ((oaisim_flag == 0) && (UE->mode !=loop_through_memory))
         AssertFatal(0== openair0_device_load(&(UE->rfdevice), &openair0_cfg[0]), "");
     UE->rfdevice.host_type = RAU_HOST;
-    sprintf(threadname, "Main UE %d", UE->Mod_id);
-    pthread_setname_np(pthread_self(), threadname);
+
     init_UE_threads(UE);
 
 #ifdef NAS_UE

targets/RT/USER/nr-uesoftmodem.c

@@ -268,7 +268,8 @@ char uecap_xer[1024],uecap_xer_in=0;
 int oaisim_flag=0;
 int emulate_rf = 0;
 
-threads_t threads= {-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1};
+threads_t threads = {-1,-1,-1,-1,-1,-1,-1,-1};
+int threads_offset = 0;
 
 char* usrp_args=NULL;
 char* usrp_clksrc=NULL;
@@ -1027,8 +1028,25 @@ int main( int argc, char **argv ) {
   //dump_frame_parms(frame_parms[0]);
   
   init_openair0();
-  
 
+  // default threads.main = 2
+  // if there are enough processors, all others on subsequent CPUs
+  if (threads_offset>0) {
+    if (threads_offset+8<=get_nprocs()) {
+      if (threads.main<0) threads.main=threads_offset+1;
+      if (threads.sync<0) threads.sync=threads_offset+2;
+      if (threads.one<0) threads.one=threads_offset+3;
+      if (threads.two<0) threads.two=threads_offset+4;
+      if (threads.three<0) threads.three=threads_offset+5;
+      if (threads.slot1_proc_one<0) threads.slot1_proc_one=threads_offset+6;
+      if (threads.slot1_proc_two<0) threads.slot1_proc_two=threads_offset+7;
+      if (threads.slot1_proc_three<0) threads.slot1_proc_three=threads_offset+8;
+    }
+    else {
+      LOG_E(HW,"Not enough CPUs available (nprocs (=%d) >= threadmain (=%d) + 10)\n",get_nprocs(),threads_offset);
+      exit_fun("Error setting thread affinity\n");
+    }
+  }
 
 #ifndef DEADLINE_SCHEDULER
 
@@ -1039,14 +1057,17 @@ int main( int argc, char **argv ) {
     char cpu_affinity[1024];
     CPU_ZERO(&cpuset);
 #ifdef CPU_AFFINITY
-    if (get_nprocs() > 2) {
-        CPU_SET(0, &cpuset);
+    int j;
+    if (get_nprocs()>=2) {
+      for (j = 2; j < get_nprocs(); j++)
+	CPU_SET(j, &cpuset);
+
         s = pthread_setaffinity_np(pthread_self(), sizeof(cpu_set_t), &cpuset);
         if (s != 0) {
             perror( "pthread_setaffinity_np");
             exit_fun("Error setting processor affinity");
         }
-        LOG_I(HW, "Setting the affinity of main function to CPU 0, for device library to use CPU 0 only!\n");
+        LOG_I(HW, "Setting the affinity of main function to all CPUs, for device library to use CPU 0 only!\n");
     }
 #endif
 

targets/RT/USER/nr-uesoftmodem.h

@@ -157,13 +157,16 @@
 {"external-clock",          CONFIG_HLP_EXCCLK,      PARAMFLAG_BOOL,         uptr:&clock_source,                 defintval:0,                    TYPE_INT,       0},                     \
 {"wait-for-sync",           NULL,                   PARAMFLAG_BOOL,         iptr:&wait_for_sync,                defintval:0,                    TYPE_INT,       0},                     \
 {"single-thread-disable",   CONFIG_HLP_NOSNGLT,     PARAMFLAG_BOOL,         iptr:&single_thread_flag,           defintval:1,                    TYPE_INT,       0},                     \
-{"threadIQ",                NULL,                   0,                      iptr:&(threads.iq),                 defintval:1,                    TYPE_INT,       0},                     \
-{"threadOneSubframe",       NULL,                   0,                      iptr:&(threads.one),                defintval:1,                    TYPE_INT,       0},                     \
-{"threadTwoSubframe",       NULL,                   0,                      iptr:&(threads.two),                defintval:1,                    TYPE_INT,       0},                     \
-{"threadThreeSubframe",     NULL,                   0,                      iptr:&(threads.three),              defintval:1,                    TYPE_INT,       0},                     \
-{"threadSlot1ProcOne",      NULL,                   0,                      iptr:&(threads.slot1_proc_one),     defintval:1,                    TYPE_INT,       0},                     \
-{"threadSlot1ProcTwo",      NULL,                   0,                      iptr:&(threads.slot1_proc_two),     defintval:1,                    TYPE_INT,       0},                     \
-{"nr_dlsch-demod-shift",    CONFIG_HLP_DLSHIFT,     0,                      iptr:(int32_t *)&nr_dlsch_demod_shift, defintval:0,                    TYPE_INT,       0},                     \
+{"threadoffset",            NULL,                   0,                      iptr:&(threads_offset),             defintval:0,                    TYPE_INT,       0},                     \
+{"threadMain",              NULL,                   0,                      iptr:&(threads.main),               defintval:-1,                    TYPE_INT,       0},                     \
+{"threadSync",              NULL,                   0,                      iptr:&(threads.sync),               defintval:-1,                    TYPE_INT,       0},                     \
+{"threadOneSubframe",       NULL,                   0,                      iptr:&(threads.one),                defintval:-1,                    TYPE_INT,       0},                     \
+{"threadTwoSubframe",       NULL,                   0,                      iptr:&(threads.two),                defintval:-1,                    TYPE_INT,       0},                     \
+{"threadThreeSubframe",     NULL,                   0,                      iptr:&(threads.three),              defintval:-1,                    TYPE_INT,       0},                     \
+{"threadSlot1ProcOne",      NULL,                   0,                      iptr:&(threads.slot1_proc_one),     defintval:-1,                    TYPE_INT,       0},                     \
+{"threadSlot1ProcTwo",      NULL,                   0,                      iptr:&(threads.slot1_proc_two),     defintval:-1,                    TYPE_INT,       0},                     \
+{"threadSlot1ProcThree",    NULL,                   0,                      iptr:&(threads.slot1_proc_three),   defintval:-1,                    TYPE_INT,       0},                     \
+{"nr-dlsch-demod-shift",    CONFIG_HLP_DLSHIFT,     0,                      iptr:(int32_t *)&nr_dlsch_demod_shift, defintval:0,                     TYPE_INT,       0},                     \
 {"A" ,                      CONFIG_HLP_TADV,        0,                      uptr:&timing_advance,               defintval:0,                    TYPE_UINT,      0},                     \
 {"C" ,                      CONFIG_HLP_DLF,         0,                      uptr:&(downlink_frequency[0][0]),   defuintval:2680000000,          TYPE_UINT,      0},                     \
 {"a" ,                      CONFIG_HLP_CHOFF,       0,                      iptr:&chain_offset,                 defintval:0,                    TYPE_INT,       0},                     \

targets/RT/USER/rt_wrapper.c

@@ -294,20 +294,20 @@ void thread_top_init(char *thread_name,
   char cpu_affinity[1024];
   cpu_set_t cpuset;
 
-  /* Set affinity mask to include CPUs 1 to MAX_CPUS */
+  /* Set affinity mask to include CPUs 2 to MAX_CPUS */
   /* CPU 0 is reserved for UHD threads */
   /* CPU 1 is reserved for all RX_TX threads */
-  /* Enable CPU Affinity only if number of CPUs >2 */
+  /* Enable CPU Affinity only if number of CPUs > 2 */
   CPU_ZERO(&cpuset);
 
 #ifdef CPU_AFFINITY
-  if (get_nprocs() > 2)
+  if (affinity == 0) {
+    LOG_W(HW,"thread_top_init() called with affinity==0, but overruled by #ifdef CPU_AFFINITY\n");
+  }
+  else if (get_nprocs() > 2)
   {
-    if (affinity == 0)
-      CPU_SET(0,&cpuset);
-    else
-      for (j = 1; j < get_nprocs(); j++)
-        CPU_SET(j, &cpuset);
+    for (j = 2; j < get_nprocs(); j++)
+      CPU_SET(j, &cpuset);
     s = pthread_setaffinity_np(pthread_self(), sizeof(cpu_set_t), &cpuset);
     if (s != 0)
     {
@@ -315,21 +315,29 @@ void thread_top_init(char *thread_name,
       exit_fun("Error setting processor affinity");
     }
   }
+#else //CPU_AFFINITY
+  if (affinity) {
+    LOG_W(HW,"thread_top_init() called with affinity>0, but overruled by #ifndef CPU_AFFINITY.\n");
+  }
 #endif //CPU_AFFINITY
 
   /* Check the actual affinity mask assigned to the thread */
   s = pthread_getaffinity_np(pthread_self(), sizeof(cpu_set_t), &cpuset);
-  if (s != 0) {
+  if (s != 0)
+  {
     perror( "pthread_getaffinity_np");
     exit_fun("Error getting processor affinity ");
   }
   memset(cpu_affinity,0,sizeof(cpu_affinity));
   for (j = 0; j < 1024; j++)
-    if (CPU_ISSET(j, &cpuset)) {  
+  {
+    if (CPU_ISSET(j, &cpuset))
+    {  
       char temp[1024];
       sprintf (temp, " CPU_%d", j);
       strcat(cpu_affinity, temp);
     }
+  }
 
   memset(&sparam, 0, sizeof(sparam));
   sparam.sched_priority = sched_get_priority_max(SCHED_FIFO);