ldpctest memory allocations

memory allocations for CUDA targets (harq->b) for gNB transmitter

CMakeLists.txt

@@ -170,7 +170,7 @@ endif()
 if (NOT CROSS_COMPILE)
   set(C_FLAGS_PROCESSOR "${C_FLAGS_PROCESSOR} -march=native")
 else ()
-  set(C_FLAGS_PROCESSOR "${C_FLAGS_PROCESSOR} -march=armv8-a")
+  set(C_FLAGS_PROCESSOR "${C_FLAGS_PROCESSOR} -march=armv8.2-a")
 endif()
 
 # add autotools definitions that were maybe used!
@@ -188,6 +188,10 @@ if(CMAKE_C_COMPILER_ID STREQUAL "Clang" OR CMAKE_CXX_COMPILER_ID STREQUAL "Clang
   set(commonOpts "${commonOpts} -Wno-unused-command-line-argument")
 endif()
 
+if (ENABLE_CUDA)
+  set(commonOpts "${commonOpts} -DENABLE_CUDA")
+endif()
+
 set(CMAKE_C_FLAGS
   "${C_FLAGS_PROCESSOR} ${commonOpts} -std=gnu11 -funroll-loops ${CMAKE_C_FLAGS}")
 set(CMAKE_CXX_FLAGS
@@ -987,6 +991,8 @@ set(PHY_SRC_UE
   ${OPENAIR1_DIR}/PHY/NR_TRANSPORT/srs_rx.c
   ${OPENAIR1_DIR}/PHY/NR_TRANSPORT/nr_uci_tools_common.c
   )
+  set(PHY_NR_SRC_CU
+     )
   set(PHY_NR_UE_SRC
   ${OPENAIR1_DIR}/PHY/INIT/nr_parms.c
   ${OPENAIR1_DIR}/PHY/MODULATION/nr_modulation.c
@@ -1064,7 +1070,17 @@ target_include_directories(PHY_UE PRIVATE ${blas_INCLUDE_DIRS} ${lapacke_INCLUDE
 add_library(PHY_NR_COMMON ${PHY_NR_SRC_COMMON})
 target_link_libraries(PHY_NR_COMMON PUBLIC UTIL)
 
-add_library(PHY_NR ${PHY_NR_SRC})
+if (ENABLE_CUDA)
+  find_package(CUDA REQUIRED)
+  if (NOT CUDA_FOUND)
+    message(FATAL_ERROR "no CUDA found")
+  endif()
+  set(CUDA_NVCC_FLAGS "${CUDA_NVCC_FLAGS};-arch=sm_87;")
+  set(CUDA_VERBOS_BUILD_ON)
+  cuda_add_library(PHY_NR ${PHY_NR_SRC} ${PHY_NR_SRC_CU})
+else()
+  add_library(PHY_NR ${PHY_NR_SRC})
+endif()
 target_link_libraries(PHY_NR nr_phy_common nr_common nr_fapi_p5)
 
 add_library(PHY_NR_NO_AVX_256 ${PHY_NR_SRC})
@@ -1980,13 +1996,21 @@ target_link_libraries(smallblocktest PRIVATE
 )
 
 
+#if (ENABLE_CUDA)
+#  find_package(CUDA REQUIRED)
+#  if (NOT CUDA_FOUND)
+#    message(FATAL_ERROR "no CUDA found")
+#  endif()
+#  cuda_add_library(ldpctest)
+#else()
+#endif()
 add_executable(ldpctest
   ${OPENAIR1_DIR}/PHY/CODING/TESTBENCH/ldpctest.c
   ${OPENAIR1_DIR}/PHY/CODING/nrLDPC_load.c
 )
 target_link_libraries(ldpctest PRIVATE
   -Wl,--start-group UTIL SIMU PHY_COMMON PHY_NR_COMMON -Wl,--end-group
-  m pthread dl shlib_loader ${T_LIB}
+  m pthread dl shlib_loader ${T_LIB} ${CUDA_LIBRARIES}
   # link 'check_crc' to make it resolved in the LDPC coding libraries
   # 'check_crc' is not used in ldpctest so it is not linked in the executable by default
   # --whole-archive links 'check_crc' in the executable even though it is note used, see 'man ld'

ci-scripts/ci_infra.yaml

@@ -455,3 +455,93 @@ l2sim4g_ext_dn:
   Host: "%%current_host%%"
   NetworkScript: docker exec l2sim4g-trf-gen ip a show dev eth0
   CmdPrefix: docker exec l2sim4g-trf-gen
+
+rfsim5g_multiue1:
+  Host: "%%current_host%%"
+  NetworkScript: docker exec rfsim5g-oai-nr-ue ip a show dev oaitun_ue1
+  CmdPrefix: docker exec rfsim5g-oai-nr-ue
+  IF: oaitun_ue1
+  AttachScript: docker start rfsim5g-oai-nr-ue
+  DetachScript: docker stop rfsim5g-oai-nr-ue
+  MTU: 1500
+
+rfsim5g_multiue2:
+  Host: "%%current_host%%"
+  NetworkScript: docker exec rfsim5g-oai-nr-ue ip a show dev oaitun_ue2
+  CmdPrefix: docker exec rfsim5g-oai-nr-ue
+  IF: oaitun_ue2
+  AttachScript: true
+  DetachScript: true
+  MTU: 1500
+
+rfsim5g_multiue3:
+  Host: "%%current_host%%"
+  NetworkScript: docker exec rfsim5g-oai-nr-ue ip a show dev oaitun_ue3
+  CmdPrefix: docker exec rfsim5g-oai-nr-ue
+  IF: oaitun_ue3
+  AttachScript: "true"
+  DetachScript: "true"
+  MTU: 1500
+
+rfsim5g_multiue4:
+  Host: "%%current_host%%"
+  NetworkScript: docker exec rfsim5g-oai-nr-ue ip a show dev oaitun_ue4
+  CmdPrefix: docker exec rfsim5g-oai-nr-ue
+  IF: oaitun_ue4
+  AttachScript: "true"
+  DetachScript: "true"
+  MTU: 1500
+
+rfsim5g_multiue5:
+  Host: "%%current_host%%"
+  NetworkScript: docker exec rfsim5g-oai-nr-ue ip a show dev oaitun_ue5
+  CmdPrefix: docker exec rfsim5g-oai-nr-ue
+  IF: oaitun_ue5
+  AttachScript: "true"
+  DetachScript: "true"
+  MTU: 1500
+
+rfsim5g_multiue6:
+  Host: "%%current_host%%"
+  NetworkScript: docker exec rfsim5g-oai-nr-ue ip a show dev oaitun_ue6
+  CmdPrefix: docker exec rfsim5g-oai-nr-ue
+  IF: oaitun_ue6
+  AttachScript: "true"
+  DetachScript: "true"
+  MTU: 1500
+
+rfsim5g_multiue7:
+  Host: "%%current_host%%"
+  NetworkScript: docker exec rfsim5g-oai-nr-ue ip a show dev oaitun_ue7
+  CmdPrefix: docker exec rfsim5g-oai-nr-ue
+  IF: oaitun_ue7
+  AttachScript: "true"
+  DetachScript: "true"
+  MTU: 1500
+
+rfsim5g_multiue8:
+  Host: "%%current_host%%"
+  NetworkScript: docker exec rfsim5g-oai-nr-ue ip a show dev oaitun_ue8
+  CmdPrefix: docker exec rfsim5g-oai-nr-ue
+  IF: oaitun_ue8
+  AttachScript: "true"
+  DetachScript: "true"
+  MTU: 1500
+
+rfsim5g_multiue9:
+  Host: "%%current_host%%"
+  NetworkScript: docker exec rfsim5g-oai-nr-ue ip a show dev oaitun_ue9
+  CmdPrefix: docker exec rfsim5g-oai-nr-ue
+  IF: oaitun_ue9
+  AttachScript: "true"
+  DetachScript: "true"
+  MTU: 1500
+
+rfsim5g_multiue10:
+  Host: "%%current_host%%"
+  NetworkScript: docker exec rfsim5g-oai-nr-ue ip a show dev oaitun_ue10
+  CmdPrefix: docker exec rfsim5g-oai-nr-ue
+  IF: oaitun_ue10
+  AttachScript: "true"
+  DetachScript: "true"
+  MTU: 1500

ci-scripts/cls_containerize.py

@@ -389,13 +389,13 @@ class Containerize():
 			elif image != 'ran-build':
 				cmd.run(f'sed -i -e "s#ran-build:latest#ran-build:{imageTag}#" docker/Dockerfile.{pattern}{self.dockerfileprefix}')
 			if image == 'oai-gnb-aerial':
-				cmd.run('cp -f /opt/nvidia-ipc/nvipc.src.2025.05.20.tar.gz .')
+				cmd.run('cp -f /opt/nvidia-ipc/nvipc_src.2025.08.27.tar.gz .')
 			logfile = f'{lSourcePath}/cmake_targets/log/{name}.docker.log'
 			ret = cmd.run(f'{self.cli} build {self.cliBuildOptions} --target {image} --tag {name}:{imageTag} --file docker/Dockerfile.{pattern}{self.dockerfileprefix} {option} . > {logfile} 2>&1', timeout=1200)
 			t = (name, archiveArtifact(cmd, ctx, logfile))
 			log_files.append(t)
 			if image == 'oai-gnb-aerial':
-				cmd.run('rm -f nvipc.src.2025.05.20.tar.gz')
+				cmd.run('rm -f nvipc_src.2025.08.27.tar.gz')
 			# check the status of the build
 			ret = cmd.run(f"{self.cli} image inspect --format=\'Size = {{{{.Size}}}} bytes\' {name}:{imageTag}")
 			if ret.returncode != 0:

ci-scripts/conf_files/gnb.sa.band78.106prb.n310.7ds2u.conf

@@ -112,7 +112,7 @@ gNBs =
 # 0 = neither, 1= group hopping, 2=sequence hopping
         pucchGroupHopping                                            = 0;
         hoppingId                                                    = 40;
-        p0_nominal                                                   = -70;
+        p0_nominal                                                   = -80;
 
       ssb_PositionsInBurst_Bitmap                                    = 1;
 
@@ -175,7 +175,7 @@ MACRLCs = (
   num_cc           = 1;
   tr_s_preference  = "local_L1";
   tr_n_preference  = "local_RRC";
-  pusch_TargetSNRx10 = 180;
+  pusch_TargetSNRx10 = 150;
   pucch_TargetSNRx10 = 200;
 }
 );
@@ -184,7 +184,7 @@ L1s = (
 {
   num_cc = 1;
   tr_n_preference = "local_mac";
-  prach_dtx_threshold = 120;
+  prach_dtx_threshold = 130;
   #pucch0_dtx_threshold = 150;
 }
 );
@@ -198,7 +198,7 @@ RUs = (
   att_rx         = 0;
   bands          = [78];
   max_pdschReferenceSignalPower = -27;
-  max_rxgain                    = 65;
+  max_rxgain                    = 70;
   eNB_instances  = [0];
   sf_extension = 0
   sdr_addrs = "addr=192.168.80.52,clock_source=internal,time_source=internal"

ci-scripts/mbim_scripts/stop_quectel_mbim.sh

@@ -1,3 +1,7 @@
 set -x
 
 sudo mbimcli -p -d /dev/cdc-wdm0 --set-radio-state=off
+
+IF=wwan0
+sudo ip link set ${IF} down
+sudo ip addr flush dev ${IF}

ci-scripts/xml_files/container_5g_rfsim_multiue.xml

@@ -0,0 +1,210 @@
+<!--
+
+ 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
+
+-->
+<testCaseList>
+        <htmlTabRef>rfsim-5gnr-tdd-multiue</htmlTabRef>
+        <htmlTabName>Monolithic SA TDD gNB multiue</htmlTabName>
+        <htmlTabIcon>wrench</htmlTabIcon>
+        <TestCaseRequestedList>
+ 111111
+ 800813
+ 000001
+ 000002
+ 000003
+ 000008
+ 020001
+ 020002
+ 020003
+ 020004
+ 020105
+ 444445
+ 030001
+ 030002
+ 444444
+ 100001
+ 222222
+        </TestCaseRequestedList>
+        <TestCaseExclusionList></TestCaseExclusionList>
+
+        <testCase id="111111">
+                <class>Pull_Local_Registry</class>
+                <desc>Pull Images from Local Registry</desc>
+                <svr_id>0</svr_id>
+                <images>oai-gnb-asan oai-nr-ue</images>
+        </testCase>
+        <testCase id="800813">
+                <class>Create_Workspace</class>
+                <desc>Create new Workspace</desc>
+                <eNB_instance>0</eNB_instance>
+                <eNB_serverId>0</eNB_serverId>
+        </testCase>
+        <testCase id="000001">
+                <class>Deploy_Object</class>
+                <desc>Deploy OAI 5G CoreNetwork</desc>
+                <yaml_path>ci-scripts/yaml_files/5g_rfsimulator_multiue</yaml_path>
+                <services>mysql oai-amf oai-smf oai-upf oai-ext-dn</services>
+                <eNB_instance>0</eNB_instance>
+                <eNB_serverId>0</eNB_serverId>
+        </testCase>
+
+        <testCase id="000002">
+                <class>Deploy_Object</class>
+                <desc>Deploy OAI 5G gNB RF sim SA</desc>
+                <yaml_path>ci-scripts/yaml_files/5g_rfsimulator_multiue</yaml_path>
+                <services>oai-gnb</services>
+                <eNB_instance>0</eNB_instance>
+                <eNB_serverId>0</eNB_serverId>
+        </testCase>
+
+        <testCase id="000003">
+                <class>Deploy_Object</class>
+                <desc>Deploy OAI 5G NR-UE RF sim SA</desc>
+                <yaml_path>ci-scripts/yaml_files/5g_rfsimulator_multiue</yaml_path>
+                <services>oai-nr-ue</services>
+                <eNB_instance>0</eNB_instance>
+                <eNB_serverId>0</eNB_serverId>
+        </testCase>
+
+        <testCase id="000008">
+                <class>Attach_UE</class>
+                <desc>Attach OAI UE (Wait for IP)</desc>
+                <id>rfsim5g_multiue1 rfsim5g_multiue2 rfsim5g_multiue3 rfsim5g_multiue4 rfsim5g_multiue5 rfsim5g_multiue6 rfsim5g_multiue7 rfsim5g_multiue8 rfsim5g_multiue9 rfsim5g_multiue10</id>
+                <nodes>acamas acamas acamas acamas acamas acamas acamas acamas acamas acamas</nodes>
+        </testCase>
+
+        <testCase id="020001">
+                <class>Ping</class>
+                <desc>Ping ext-dn from NR-UE</desc>
+                <id>rfsim5g_multiue1</id>
+                <nodes>acamas</nodes>
+                <svr_id>rfsim5g_ext_dn</svr_id>
+                <svr_node>acamas</svr_node>
+                <ping_args>-c 20 -i 0.25</ping_args>
+                <ping_packetloss_threshold>5</ping_packetloss_threshold>
+        </testCase>
+
+        <testCase id="020002">
+                <class>Ping</class>
+                <desc>Ping NR-UE from ext-dn</desc>
+                <id>rfsim5g_ext_dn</id>
+                <nodes>acamas</nodes>
+                <svr_id>rfsim5g_multiue1</svr_id>
+                <svr_node>acamas</svr_node>
+                <ping_args>-c 20 -i 0.25</ping_args>
+                <ping_packetloss_threshold>5</ping_packetloss_threshold>
+        </testCase>
+
+        <testCase id="020003">
+                <class>Ping</class>
+                <desc>Ping ext-dn from Second NR-UE</desc>
+                <id>rfsim5g_multiue2</id>
+                <nodes>acamas</nodes>
+                <svr_id>rfsim5g_ext_dn</svr_id>
+                <svr_node>acamas</svr_node>
+                <ping_args>-c 20 -i 0.25</ping_args>
+                <ping_packetloss_threshold>5</ping_packetloss_threshold>
+        </testCase>
+
+        <testCase id="020004">
+                <class>Ping</class>
+                <desc>Ping Second NR-UE from ext-dn</desc>
+                <id>rfsim5g_ext_dn</id>
+                <nodes>acamas</nodes>
+                <svr_id>rfsim5g_multiue2</svr_id>
+                <svr_node>acamas</svr_node>
+                <ping_args>-c 20 -i 0.25</ping_args>
+                <ping_packetloss_threshold>5</ping_packetloss_threshold>
+        </testCase>
+
+        <testCase id="020005">
+                <class>Ping</class>
+                <desc>Ping ext-dn from both UEs</desc>
+                <id>rfsim5g_multiue1 rfsim5g_multiue2</id>
+                <nodes>acamas acamas</nodes>
+                <svr_id>rfsim5g_ext_dn</svr_id>
+                <svr_node>acamas</svr_node>
+                <ping_args>-c 20 -i 0.25</ping_args>
+                <ping_packetloss_threshold>5</ping_packetloss_threshold>
+        </testCase>
+
+        <testCase id="020105">
+                <class>Ping</class>
+                <desc>Ping ext-dn from all UEs</desc>
+                <id>rfsim5g_multiue1 rfsim5g_multiue2 rfsim5g_multiue3 rfsim5g_multiue4 rfsim5g_multiue5 rfsim5g_multiue6 rfsim5g_multiue7 rfsim5g_multiue8 rfsim5g_multiue9 rfsim5g_multiue10</id>
+                <nodes>acamas acamas acamas acamas acamas acamas acamas acamas acamas acamas</nodes>
+                <svr_id>rfsim5g_ext_dn</svr_id>
+                <svr_node>acamas</svr_node>
+                <ping_args>-c 20 -i 0.25 -s 1016</ping_args>
+                <ping_packetloss_threshold>5</ping_packetloss_threshold>
+        </testCase>
+
+        <testCase id="030001">
+                <class>Iperf</class>
+                <desc>Iperf (DL/3Mbps/UDP)(30 sec)</desc>
+                <iperf_args>-u -b 3M -t 20 -R</iperf_args>
+                <id>rfsim5g_multiue1</id>
+                <nodes>acamas</nodes>
+                <svr_id>rfsim5g_ext_dn</svr_id>
+                <svr_node>acamas</svr_node>
+                <iperf_packetloss_threshold>5</iperf_packetloss_threshold>
+                <iperf_bitrate_threshold>90</iperf_bitrate_threshold>
+        </testCase>
+
+        <testCase id="030002">
+                <class>Iperf</class>
+                <desc>Iperf (UL/1Mbps/UDP)(30 sec)</desc>
+                <iperf_args>-u -b 1M -t 20</iperf_args>
+                <id>rfsim5g_multiue1</id>
+                <nodes>acamas</nodes>
+                <svr_id>rfsim5g_ext_dn</svr_id>
+                <svr_node>acamas</svr_node>
+                <iperf_packetloss_threshold>5</iperf_packetloss_threshold>
+                <iperf_bitrate_threshold>90</iperf_bitrate_threshold>
+        </testCase>
+
+        <testCase id="444444">
+                <class>Detach_UE</class>
+                <desc>Detach OAI UEs</desc>
+                <id>rfsim5g_multiue1</id>
+                <nodes>acamas</nodes>
+        </testCase>
+
+        <testCase id="100001">
+                <class>Undeploy_Object</class>
+                <always_exec>true</always_exec>
+                <desc>Undeploy all OAI 5G stack</desc>
+                <yaml_path>ci-scripts/yaml_files/5g_rfsimulator_multiue</yaml_path>
+                <d_retx_th>1,0,0,0</d_retx_th>
+                <u_retx_th>1,0,0,0</u_retx_th>
+                <eNB_instance>0</eNB_instance>
+                <eNB_serverId>0</eNB_serverId>
+        </testCase>
+
+        <testCase id="222222">
+                <class>Clean_Test_Server_Images</class>
+                <always_exec>true</always_exec>
+                <desc>Clean Test Images on Test Server</desc>
+                <svr_id>0</svr_id>
+                <images>oai-gnb-asan oai-nr-ue</images>
+        </testCase>
+
+</testCaseList>

ci-scripts/xml_files/container_sa_n310_nrue.xml

@@ -109,8 +109,8 @@
 
   <testCase id="000100">
     <class>Iperf</class>
-    <desc>iperf (DL/30Mbps/UDP)(30 sec)</desc>
-    <iperf_args>-u -b 30M -t 30 -i 1 -R</iperf_args>
+    <desc>iperf (DL/60Mbps/UDP)(30 sec)</desc>
+    <iperf_args>-u -b 60M -t 30 -i 1 -R</iperf_args>
     <id>oai_ue_caracal</id>
     <svr_id>oc-cn5g</svr_id>
     <iperf_packetloss_threshold>25</iperf_packetloss_threshold>

ci-scripts/yaml_files/5g_f1_rfsimulator/docker-compose.yaml

@@ -24,7 +24,7 @@ services:
                 ipv4_address: 192.168.71.131
     oai-amf:
         container_name: "rfsim5g-oai-amf"
-        image: oaisoftwarealliance/oai-amf:v2.1.9
+        image: oaisoftwarealliance/oai-amf:v2.1.10
         environment:
             - TZ=Europe/paris
         volumes:
@@ -36,7 +36,7 @@ services:
                 ipv4_address: 192.168.71.132
     oai-smf:
         container_name: "rfsim5g-oai-smf"
-        image: oaisoftwarealliance/oai-smf:v2.1.9
+        image: oaisoftwarealliance/oai-smf:v2.1.10
         environment:
             - TZ=Europe/Paris
         volumes:
@@ -48,7 +48,7 @@ services:
                 ipv4_address: 192.168.71.133
     oai-upf:
         container_name: "rfsim5g-oai-upf"
-        image: oaisoftwarealliance/oai-upf:v2.1.9
+        image: oaisoftwarealliance/oai-upf:v2.1.10
         init: true
         environment:
             - TZ=Europe/Paris

ci-scripts/yaml_files/5g_fdd_rfsimulator/docker-compose.yaml

@@ -24,7 +24,7 @@ services:
                 ipv4_address: 192.168.71.131
     oai-amf:
         container_name: "rfsim5g-oai-amf"
-        image: oaisoftwarealliance/oai-amf:v2.1.9
+        image: oaisoftwarealliance/oai-amf:v2.1.10
         environment:
             - TZ=Europe/paris
         volumes:
@@ -36,7 +36,7 @@ services:
                 ipv4_address: 192.168.71.132
     oai-smf:
         container_name: "rfsim5g-oai-smf"
-        image: oaisoftwarealliance/oai-smf:v2.1.9
+        image: oaisoftwarealliance/oai-smf:v2.1.10
         environment:
             - TZ=Europe/Paris
         volumes:
@@ -48,7 +48,7 @@ services:
                 ipv4_address: 192.168.71.133
     oai-upf:
         container_name: "rfsim5g-oai-upf"
-        image: oaisoftwarealliance/oai-upf:v2.1.9
+        image: oaisoftwarealliance/oai-upf:v2.1.10
         init: true
         environment:
             - TZ=Europe/Paris

ci-scripts/yaml_files/5g_rfsimulator/README.md

@@ -37,9 +37,9 @@ Now pull images.
 
 ```bash
 $ docker pull mysql:8.0
-$ docker pull oaisoftwarealliance/oai-amf:v2.1.9
-$ docker pull oaisoftwarealliance/oai-smf:v2.1.9
-$ docker pull oaisoftwarealliance/oai-upf:v2.1.9
+$ docker pull oaisoftwarealliance/oai-amf:v2.1.10
+$ docker pull oaisoftwarealliance/oai-smf:v2.1.10
+$ docker pull oaisoftwarealliance/oai-upf:v2.1.10
 $ docker pull oaisoftwarealliance/trf-gen-cn5g:focal
 
 $ docker pull oaisoftwarealliance/oai-gnb:develop
@@ -52,7 +52,7 @@ $ docker logout
 
 **CAUTION: 2023/01/27 with the release `v1.5.0` of the `CN5G`, the previous version was not compatible any-more.**
 
-**This new version is working only with the `v2.1.9` of the `CN5G`.**
+**This new version is working only with `v2.1.9` or later of the `CN5G`.**
 
 # 2. Deploy containers #
 

ci-scripts/yaml_files/5g_rfsimulator/docker-compose.yaml

@@ -24,7 +24,7 @@ services:
                 ipv4_address: 192.168.71.131
     oai-amf:
         container_name: "rfsim5g-oai-amf"
-        image: oaisoftwarealliance/oai-amf:v2.1.9
+        image: oaisoftwarealliance/oai-amf:v2.1.10
         environment:
             - TZ=Europe/paris
         volumes:
@@ -36,7 +36,7 @@ services:
                 ipv4_address: 192.168.71.132
     oai-smf:
         container_name: "rfsim5g-oai-smf"
-        image: oaisoftwarealliance/oai-smf:v2.1.9
+        image: oaisoftwarealliance/oai-smf:v2.1.10
         environment:
             - TZ=Europe/Paris
         volumes:
@@ -48,7 +48,7 @@ services:
                 ipv4_address: 192.168.71.133
     oai-upf:
         container_name: "rfsim5g-oai-upf"
-        image: oaisoftwarealliance/oai-upf:v2.1.9
+        image: oaisoftwarealliance/oai-upf:v2.1.10
         init: true
         environment:
             - TZ=Europe/Paris

ci-scripts/yaml_files/5g_rfsimulator_24prb/docker-compose.yaml

@@ -24,7 +24,7 @@ services:
                 ipv4_address: 192.168.71.131
     oai-amf:
         container_name: "rfsim5g-oai-amf"
-        image: oaisoftwarealliance/oai-amf:v2.1.9
+        image: oaisoftwarealliance/oai-amf:v2.1.10
         environment:
             - TZ=Europe/paris
         volumes:
@@ -36,7 +36,7 @@ services:
                 ipv4_address: 192.168.71.132
     oai-smf:
         container_name: "rfsim5g-oai-smf"
-        image: oaisoftwarealliance/oai-smf:v2.1.9
+        image: oaisoftwarealliance/oai-smf:v2.1.10
         environment:
             - TZ=Europe/Paris
         volumes:
@@ -48,7 +48,7 @@ services:
                 ipv4_address: 192.168.71.133
     oai-upf:
         container_name: "rfsim5g-oai-upf"
-        image: oaisoftwarealliance/oai-upf:v2.1.9
+        image: oaisoftwarealliance/oai-upf:v2.1.10
         init: true
         environment:
             - TZ=Europe/Paris

ci-scripts/yaml_files/5g_rfsimulator_2x2/docker-compose.yaml

@@ -24,7 +24,7 @@ services:
                 ipv4_address: 192.168.71.131
     oai-amf:
         container_name: "rfsim5g-oai-amf"
-        image: oaisoftwarealliance/oai-amf:v2.1.9
+        image: oaisoftwarealliance/oai-amf:v2.1.10
         environment:
             - TZ=Europe/paris
         volumes:
@@ -36,7 +36,7 @@ services:
                 ipv4_address: 192.168.71.132
     oai-smf:
         container_name: "rfsim5g-oai-smf"
-        image: oaisoftwarealliance/oai-smf:v2.1.9
+        image: oaisoftwarealliance/oai-smf:v2.1.10
         environment:
             - TZ=Europe/Paris
         volumes:
@@ -48,7 +48,7 @@ services:
                 ipv4_address: 192.168.71.133
     oai-upf:
         container_name: "rfsim5g-oai-upf"
-        image: oaisoftwarealliance/oai-upf:v2.1.9
+        image: oaisoftwarealliance/oai-upf:v2.1.10
         init: true
         environment:
             - TZ=Europe/Paris

ci-scripts/yaml_files/5g_rfsimulator_e1/docker-compose.yaml

@@ -24,7 +24,7 @@ services:
                 ipv4_address: 192.168.71.131
     oai-amf:
         container_name: "rfsim5g-oai-amf"
-        image: oaisoftwarealliance/oai-amf:v2.1.9
+        image: oaisoftwarealliance/oai-amf:v2.1.10
         environment:
             - TZ=Europe/paris
         volumes:
@@ -36,7 +36,7 @@ services:
                 ipv4_address: 192.168.71.132
     oai-smf:
         container_name: "rfsim5g-oai-smf"
-        image: oaisoftwarealliance/oai-smf:v2.1.9
+        image: oaisoftwarealliance/oai-smf:v2.1.10
         environment:
             - TZ=Europe/Paris
         volumes:
@@ -48,7 +48,7 @@ services:
                 ipv4_address: 192.168.71.133
     oai-upf:
         container_name: "rfsim5g-oai-upf"
-        image: oaisoftwarealliance/oai-upf:v2.1.9
+        image: oaisoftwarealliance/oai-upf:v2.1.10
         init: true
         environment:
             - TZ=Europe/Paris

ci-scripts/yaml_files/5g_rfsimulator_fr2_32prb/docker-compose.yaml

@@ -24,7 +24,7 @@ services:
                 ipv4_address: 192.168.71.131
     oai-amf:
         container_name: "rfsim5g-oai-amf"
-        image: oaisoftwarealliance/oai-amf:v2.1.9
+        image: oaisoftwarealliance/oai-amf:v2.1.10
         environment:
             - TZ=Europe/paris
         volumes:
@@ -36,7 +36,7 @@ services:
                 ipv4_address: 192.168.71.132
     oai-smf:
         container_name: "rfsim5g-oai-smf"
-        image: oaisoftwarealliance/oai-smf:v2.1.9
+        image: oaisoftwarealliance/oai-smf:v2.1.10
         environment:
             - TZ=Europe/Paris
         volumes:
@@ -48,7 +48,7 @@ services:
                 ipv4_address: 192.168.71.133
     oai-upf:
         container_name: "rfsim5g-oai-upf"
-        image: oaisoftwarealliance/oai-upf:v2.1.9
+        image: oaisoftwarealliance/oai-upf:v2.1.10
         init: true
         environment:
             - TZ=Europe/Paris

ci-scripts/yaml_files/5g_rfsimulator_multiue/docker-compose.yaml

@@ -0,0 +1,146 @@
+services:
+    mysql:
+        container_name: "rfsim5g-mysql"
+        image: mysql:8.0
+        volumes:
+            - ../5g_rfsimulator/oai_db.sql:/docker-entrypoint-initdb.d/oai_db.sql
+            - ../5g_rfsimulator/mysql-healthcheck.sh:/tmp/mysql-healthcheck.sh
+        environment:
+            - TZ=Europe/Paris
+            - MYSQL_DATABASE=oai_db
+            - MYSQL_USER=test
+            - MYSQL_PASSWORD=test
+            - MYSQL_ROOT_PASSWORD=linux
+        healthcheck:
+            test: /bin/bash -c "/tmp/mysql-healthcheck.sh"
+            interval: 10s
+            timeout: 5s
+            retries: 30
+        networks:
+            public_net:
+                ipv4_address: 192.168.71.131
+    oai-amf:
+        container_name: "rfsim5g-oai-amf"
+        image: oaisoftwarealliance/oai-amf:v2.1.10
+        environment:
+            - TZ=Europe/paris
+        volumes:
+            - ../5g_rfsimulator/mini_nonrf_config.yaml:/openair-amf/etc/config.yaml
+        depends_on:
+            - mysql
+        networks:
+            public_net:
+                ipv4_address: 192.168.71.132
+    oai-smf:
+        container_name: "rfsim5g-oai-smf"
+        image: oaisoftwarealliance/oai-smf:v2.1.10
+        environment:
+            - TZ=Europe/Paris
+        volumes:
+            - ../5g_rfsimulator/mini_nonrf_config.yaml:/openair-smf/etc/config.yaml
+        depends_on:
+            - oai-amf
+        networks:
+            public_net:
+                ipv4_address: 192.168.71.133
+    oai-upf:
+        container_name: "rfsim5g-oai-upf"
+        image: oaisoftwarealliance/oai-upf:v2.1.10
+        environment:
+            - TZ=Europe/Paris
+        volumes:
+            - ../5g_rfsimulator/mini_nonrf_config.yaml:/openair-upf/etc/config.yaml
+        depends_on:
+            - oai-smf
+        cap_add:
+            - NET_ADMIN
+            - SYS_ADMIN
+        cap_drop:
+            - ALL
+        privileged: true
+        networks:
+            public_net:
+                ipv4_address: 192.168.71.134
+                interface_name: eth0
+            traffic_net:
+                ipv4_address: 192.168.72.134
+                interface_name: eth1
+    oai-ext-dn:
+        privileged: true
+        container_name: rfsim5g-oai-ext-dn
+        image: oaisoftwarealliance/trf-gen-cn5g:focal
+        entrypoint: /bin/bash -c \
+              "iptables -t nat -A POSTROUTING -o eth0 -j MASQUERADE;"\
+              "ip route add 12.1.1.0/24 via 192.168.72.134 dev eth0; sleep infinity"
+        depends_on:
+            - oai-upf
+        networks:
+            traffic_net:
+                ipv4_address: 192.168.72.135
+        healthcheck:
+            test: /bin/bash -c "ping -c 2 192.168.72.134"
+            interval: 10s
+            timeout: 5s
+            retries: 5
+    oai-gnb:
+        image: ${REGISTRY:-oaisoftwarealliance}/${GNB_IMG:-oai-gnb}:${TAG:-develop}
+        container_name: rfsim5g-oai-gnb
+        cap_drop:
+            - ALL
+        environment:
+            USE_ADDITIONAL_OPTIONS: -E --rfsim --log_config.global_log_options level,nocolor,time
+            ASAN_OPTIONS: detect_leaks=0
+        depends_on:
+            - oai-ext-dn
+        networks:
+            public_net:
+                ipv4_address: 192.168.71.140
+        volumes:
+            - ../../conf_files/gnb.sa.band78.106prb.rfsim.yaml:/opt/oai-gnb/etc/gnb.yaml
+        healthcheck:
+            test: /bin/bash -c "pgrep nr-softmodem"
+            interval: 10s
+            timeout: 5s
+            retries: 5
+    oai-nr-ue:
+        image: ${REGISTRY:-oaisoftwarealliance}/${NRUE_IMG:-oai-nr-ue}:${TAG:-develop}
+        container_name: rfsim5g-oai-nr-ue
+        cap_drop:
+            - ALL
+        cap_add:
+            - NET_ADMIN  # for interface bringup
+            - NET_RAW    # for ping
+        environment:
+            USE_ADDITIONAL_OPTIONS: -E --rfsim -r 106 --numerology 1 --uicc0.imsi 208990100001100 -C 3319680000 --rfsimulator.serveraddr 192.168.71.140 --log_config.global_log_options level,nocolor,time --num-ues 10
+        depends_on:
+            - oai-gnb
+        networks:
+            public_net:
+                ipv4_address: 192.168.71.150
+        devices:
+             - /dev/net/tun:/dev/net/tun
+        volumes:
+            - ./nrue.uicc.conf:/opt/oai-nr-ue/etc/nr-ue.conf
+        healthcheck:
+            test: /bin/bash -c "pgrep nr-uesoftmodem"
+            interval: 10s
+            timeout: 5s
+            retries: 5
+
+networks:
+    public_net:
+        driver: bridge
+        name: rfsim5g-oai-public-net
+        ipam:
+            config:
+                - subnet: 192.168.71.128/26
+        driver_opts:
+            com.docker.network.bridge.name: "rfsim5g-public"
+    traffic_net:
+        driver: bridge
+        name: rfsim5g-oai-traffic-net
+        ipam:
+            config:
+                - subnet: 192.168.72.128/26
+        driver_opts:
+            com.docker.network.bridge.name: "rfsim5g-traffic"

ci-scripts/yaml_files/5g_rfsimulator_multiue/nrue.uicc.conf

@@ -0,0 +1,75 @@
+uicc0 = {
+  imsi = "208990100001100";
+  key = "fec86ba6eb707ed08905757b1bb44b8f";
+  opc= "C42449363BBAD02B66D16BC975D77CC1";
+  dnn= "oai";
+  nssai_sst=1;
+}
+uicc1 = {
+  imsi = "208990100001101";
+  key = "fec86ba6eb707ed08905757b1bb44b8f";
+  opc= "C42449363BBAD02B66D16BC975D77CC1";
+  dnn= "oai";
+  nssai_sst=1;
+}
+uicc2 = {
+  imsi = "208990100001102";
+  key = "fec86ba6eb707ed08905757b1bb44b8f";
+  opc= "C42449363BBAD02B66D16BC975D77CC1";
+  dnn= "oai";
+  nssai_sst=1;
+}
+uicc3 = {
+  imsi = "208990100001103";
+  key = "fec86ba6eb707ed08905757b1bb44b8f";
+  opc= "C42449363BBAD02B66D16BC975D77CC1";
+  dnn= "oai";
+  nssai_sst=1;
+}
+uicc4 = {
+  imsi = "208990100001104";
+  key = "fec86ba6eb707ed08905757b1bb44b8f";
+  opc= "C42449363BBAD02B66D16BC975D77CC1";
+  dnn= "oai";
+  nssai_sst=1;
+}
+uicc5 = {
+  imsi = "208990100001105";
+  key = "fec86ba6eb707ed08905757b1bb44b8f";
+  opc= "C42449363BBAD02B66D16BC975D77CC1";
+  dnn= "oai";
+  nssai_sst=1;
+}
+uicc6 = {
+  imsi = "208990100001106";
+  key = "fec86ba6eb707ed08905757b1bb44b8f";
+  opc= "C42449363BBAD02B66D16BC975D77CC1";
+  dnn= "oai";
+  nssai_sst=1;
+}
+uicc7 = {
+  imsi = "208990100001107";
+  key = "fec86ba6eb707ed08905757b1bb44b8f";
+  opc= "C42449363BBAD02B66D16BC975D77CC1";
+  dnn= "oai";
+  nssai_sst=1;
+}
+uicc8 = {
+  imsi = "208990100001108";
+  key = "fec86ba6eb707ed08905757b1bb44b8f";
+  opc= "C42449363BBAD02B66D16BC975D77CC1";
+  dnn= "oai";
+  nssai_sst=1;
+}
+uicc9 = {
+  imsi = "208990100001109";
+  key = "fec86ba6eb707ed08905757b1bb44b8f";
+  opc= "C42449363BBAD02B66D16BC975D77CC1";
+  dnn= "oai";
+  nssai_sst=1;
+}
+
+
+
+thread-pool: "-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1"
+

ci-scripts/yaml_files/5g_rfsimulator_ntn_geo/docker-compose.yaml

@@ -24,7 +24,7 @@ services:
                 ipv4_address: 192.168.71.131
     oai-amf:
         container_name: "rfsim5g-oai-amf"
-        image: oaisoftwarealliance/oai-amf:v2.1.9
+        image: oaisoftwarealliance/oai-amf:v2.1.10
         environment:
             - TZ=Europe/paris
         volumes:
@@ -36,7 +36,7 @@ services:
                 ipv4_address: 192.168.71.132
     oai-smf:
         container_name: "rfsim5g-oai-smf"
-        image: oaisoftwarealliance/oai-smf:v2.1.9
+        image: oaisoftwarealliance/oai-smf:v2.1.10
         environment:
             - TZ=Europe/Paris
         volumes:
@@ -48,7 +48,7 @@ services:
                 ipv4_address: 192.168.71.133
     oai-upf:
         container_name: "rfsim5g-oai-upf"
-        image: oaisoftwarealliance/oai-upf:v2.1.9
+        image: oaisoftwarealliance/oai-upf:v2.1.10
         init: true
         environment:
             - TZ=Europe/Paris

ci-scripts/yaml_files/5g_rfsimulator_ntn_leo/docker-compose.yaml

@@ -23,7 +23,7 @@ services:
                 ipv4_address: 192.168.71.131
     oai-amf:
         container_name: "rfsim5g-oai-amf"
-        image: oaisoftwarealliance/oai-amf:v2.1.9
+        image: oaisoftwarealliance/oai-amf:v2.1.10
         environment:
             - TZ=Europe/paris
         volumes:
@@ -35,7 +35,7 @@ services:
                 ipv4_address: 192.168.71.132
     oai-smf:
         container_name: "rfsim5g-oai-smf"
-        image: oaisoftwarealliance/oai-smf:v2.1.9
+        image: oaisoftwarealliance/oai-smf:v2.1.10
         environment:
             - TZ=Europe/Paris
         volumes:
@@ -47,7 +47,7 @@ services:
                 ipv4_address: 192.168.71.133
     oai-upf:
         container_name: "rfsim5g-oai-upf"
-        image: oaisoftwarealliance/oai-upf:v2.1.9
+        image: oaisoftwarealliance/oai-upf:v2.1.10
         environment:
             - TZ=Europe/Paris
         volumes:

ci-scripts/yaml_files/5g_rfsimulator_u0_25prb/docker-compose.yaml

@@ -24,7 +24,7 @@ services:
                 ipv4_address: 192.168.71.131
     oai-amf:
         container_name: "rfsim5g-oai-amf"
-        image: oaisoftwarealliance/oai-amf:v2.1.9
+        image: oaisoftwarealliance/oai-amf:v2.1.10
         environment:
             - TZ=Europe/paris
         volumes:
@@ -36,7 +36,7 @@ services:
                 ipv4_address: 192.168.71.132
     oai-smf:
         container_name: "rfsim5g-oai-smf"
-        image: oaisoftwarealliance/oai-smf:v2.1.9
+        image: oaisoftwarealliance/oai-smf:v2.1.10
         environment:
             - TZ=Europe/Paris
         volumes:
@@ -48,7 +48,7 @@ services:
                 ipv4_address: 192.168.71.133
     oai-upf:
         container_name: "rfsim5g-oai-upf"
-        image: oaisoftwarealliance/oai-upf:v2.1.9
+        image: oaisoftwarealliance/oai-upf:v2.1.10
         init: true
         environment:
             - TZ=Europe/Paris

ci-scripts/yaml_files/sa_gnb_aerial/docker-compose.yaml

@@ -23,7 +23,7 @@ services:
       - ../../../cmake_targets/share:/opt/cuBB/share
     userns_mode: host
     ipc: "shareable"
-    image: cubb-build:25-1
+    image: cubb-build:25-2
     environment:
       - cuBB_SDK=/opt/nvidia/cuBB
     command: bash -c "sudo rm -rf /tmp/phy.log && sudo chmod +x /opt/nvidia/cuBB/aerial_l1_entrypoint.sh && /opt/nvidia/cuBB/aerial_l1_entrypoint.sh"

cmake_targets/build_oai

@@ -135,6 +135,8 @@ Options:
   CC=/usr/bin/clang CXX=/usr/bin/clang++ ./build_oai ... --sanitize-memory
 --sanitize-thread | -fsanitize=thread
   Enable the thread sanitizer on all targets
+--use-cuda
+  Tell build to use NVCC for appropriate .cu files and cuda_runtime functions  
 -h | --help
    Print this help"
 }
@@ -386,6 +388,9 @@ function main() {
         --sanitize-thread | -fsanitize=thread)
             CMAKE_CMD="$CMAKE_CMD -DSANITIZE_THREAD=True"
             shift;;
+	--use-cuda)
+	    CMAKE_CMD="$CMAKE_CMD -DENABLE_CUDA=True"
+	    shift;;
        --trace-asn1c-enc-dec)
             CMAKE_CMD="$CMAKE_CMD -DTRACE_ASN1C_ENC_DEC=ON"
             echo_info "Enabling asn1c internal traces via OAI logging system"

common/utils/T/.gitignore

@@ -3,6 +3,7 @@
 genids
 tracer/enb
 tracer/gnb
+tracer/gnb_mac
 tracer/extract_config
 tracer/record
 tracer/replay

common/utils/T/T_messages.txt

@@ -92,9 +92,13 @@ ID = UE_PHY_INITIATE_RA_PROCEDURE
     GROUP = ALL:PHY:UE:WIRESHARK
     FORMAT = int,frame : int,subframe_or_slot : int,preamble : int,power
 ID = GNB_PHY_DL_TICK
-    DESC = gNodeB downlink tick - one tick per ms at start of downlink processing
+    DESC = gNodeB downlink tick - one tick per slot at start of downlink processing
     GROUP = ALL:PHY:GRAPHIC:GNB
-    FORMAT = int,gNB_ID : int,frame : int,subframe
+    FORMAT = int,gNB_ID : int,frame : int,slot
+ID = GNB_PHY_UL_TICK
+    DESC = gNodeB uplink tick - one tick per slot at start of uplink processing
+    GROUP = ALL:PHY:GRAPHIC:GNB
+    FORMAT = int,gNB_ID : int,frame : int,slot
 ID = ENB_PHY_MIB
     DESC = MIB data
     GROUP = ALL:PHY:ENB:WIRESHARK
@@ -221,6 +225,22 @@ ID = NRUE_MAC_DL_RAR_PDU_WITH_DATA
     DESC = NR MAC downlink PDU for RAR
     GROUP = ALL:MAC:UE:WIRESHARK
     FORMAT = int,rnti : int,frame : int,slot : buffer,data
+ID = GNB_MAC_DL
+    DESC = NR MAC downlink scheduler decision for an UE
+    GROUP = ALL:MAC:GNB
+    FORMAT = int,rnti : int,frame : int,slot : int,mcs : int,tbs
+ID = GNB_MAC_LCID_DL
+    DESC = NR MAC downlink scheduler decision for an UE for a specific LCID
+    GROUP = ALL:MAC:GNB
+    FORMAT = int,rnti : int,frame : int,slot : int,lcid : int,data_size : int,tx_list_occupancy
+ID = GNB_MAC_UL
+    DESC = NR MAC uplink scheduler decision for an UE
+    GROUP = ALL:MAC:GNB
+    FORMAT = int,rnti : int,frame : int,slot : int,mcs : int,tbs
+ID = GNB_MAC_LCID_UL
+    DESC = NR MAC uplink traffic received for an UE for a specific LCID
+    GROUP = ALL:MAC:GNB
+    FORMAT = int,rnti : int,frame : int,slot : int,lcid : int,data_size
 
 #RLC logs
 ID = ENB_RLC_DL

common/utils/T/tracer/CMakeLists.txt

@@ -62,9 +62,13 @@ if(T_TRACER_GUI)
                         tracer_logger tracer_view tracer_events)
   target_link_libraries(gnb PRIVATE png)
 
-  add_executable(ue ue.c)
-  target_link_libraries(ue PRIVATE tracer_utils tracer_filter tracer_gui
+  add_executable(gnb_mac gnb_mac.c)
+  target_link_libraries(gnb_mac PRIVATE tracer_utils tracer_filter tracer_gui
                         tracer_logger tracer_view tracer_events)
+
+  add_executable(ue ue.c)
+  target_link_libraries(ue PRIVATE tracer_utils tracer_view tracer_gui
+                        tracer_logger tracer_filter tracer_events)
   target_link_libraries(ue PRIVATE png)
 
   add_executable(vcd vcd.c)
@@ -77,5 +81,5 @@ if(T_TRACER_GUI)
 
   add_subdirectory(gui)
 
-  add_dependencies(T_tools enb gnb ue vcd to_vcd)
+  add_dependencies(T_tools enb gnb gnb_mac ue vcd to_vcd)
 endif()

common/utils/T/tracer/Makefile

@@ -8,7 +8,7 @@ XLIBS=-lX11 -lpng -lXft
 
 all: record replay extract_config textlog enb ue vcd macpdu2wireshark \
      extract_input_subframe extract_output_subframe to_vcd extract multi \
-     gnb t_tracer_app_gnb t_tracer_app_ue
+     gnb t_tracer_app_gnb t_tracer_app_ue gnb_mac
 
 record: utils.o record.o database.o configuration.o
 	$(CC) $(CFLAGS) -o record $^ $(LIBS)
@@ -65,14 +65,18 @@ gnb: utils.o gnb.o database.o event.o handler.o configuration.o \
          filter/filter.a
 	$(CC) $(CFLAGS) -o gnb $^ $(LIBS) $(XLIBS)
 
-t_tracer_app_gnb: utils.o t_tracer_app_gnb.o database.o event.o handler.o configuration.o \
-         logger/logger.a filter/filter.a
+t_tracer_app_gnb: utils.o t_tracer_app_gnb.o database.o event.o handler.o \
+         configuration.o logger/logger.a filter/filter.a
 	$(CC) $(CFLAGS) -o t_tracer_app_gnb $^ $(LIBS) $(XLIBS)
 
-t_tracer_app_ue: utils.o t_tracer_app_ue.o database.o event.o handler.o configuration.o \
-         logger/logger.a filter/filter.a
+t_tracer_app_ue: utils.o t_tracer_app_ue.o database.o event.o handler.o \
+         configuration.o logger/logger.a filter/filter.a
 	$(CC) $(CFLAGS) -o t_tracer_app_ue $^ $(LIBS) $(XLIBS)
 
+gnb_mac: gnb_mac.o gui/gui.a view/view.a logger/logger.a filter/filter.a \
+         utils.o database.o configuration.o handler.o event.o
+	$(CC) $(CFLAGS) -o gnb_mac $^ $(LIBS) $(XLIBS)
+
 multi.o: ../incgen/T_IDs.h
 
 ../incgen/T_IDs.h: ../T_messages.txt
@@ -98,7 +102,8 @@ filter/filter.a:
 clean:
 	rm -f *.o core tracer_remote textlog enb ue vcd record replay
 	rm -f extract_config macpdu2wireshark extract_input_subframe
-	rm -f extract_output_subframe to_vcd extract multi gnb t_tracer_app_gnb t_tracer_app_ue
+	rm -f extract_output_subframe to_vcd extract multi gnb
+	rm -f t_tracer_app_gnb t_tracer_app_ue gnb_mac
 	cd gui && $(MAKE) clean
 	cd view && $(MAKE) clean
 	cd logger && $(MAKE) clean

common/utils/T/tracer/enb.c

@@ -340,12 +340,12 @@ static void enb_main_gui(enb_gui *e, gui *g, event_handler *h, void *database,
   xy_plot_set_range(g, w, 0, 1024*10, -10, 80);
   l = new_ttilog(h, database,
       "ENB_PHY_PUCCH_1_ENERGY", "frame", "subframe", "threshold", 0);
-  v = new_view_tti(10, g, w, new_color(g, "#ff0000"));
+  v = new_view_tti(10, g, w, new_color(g, "#ff0000"), 10);
   logger_add_view(l, v);
   e->pucch1_energy_ue_threshold_logger = l;
   l = new_ttilog(h, database,
       "ENB_PHY_PUCCH_1_ENERGY", "frame", "subframe", "energy", 1);
-  v = new_view_tti(10, g, w, new_color(g, "#0c0c72"));
+  v = new_view_tti(10, g, w, new_color(g, "#0c0c72"), 10);
   logger_add_view(l, v);
   e->pucch1_energy_ue_energy_logger = l;
 
@@ -368,7 +368,7 @@ static void enb_main_gui(enb_gui *e, gui *g, event_handler *h, void *database,
   widget_add_child(g, line, w, -1);
   l = new_ticked_ttilog(h, database, "ENB_PHY_DL_TICK", "frame", "subframe",
       "ENB_PHY_DLSCH_UE_DCI", "mcs", 0, -1);
-  v = new_view_tti(10, g, w, new_color(g, "#0c0c72"));
+  v = new_view_tti(10, g, w, new_color(g, "#0c0c72"), 10);
   logger_add_view(l, v);
   e->dl_mcs_logger = l;
 
@@ -379,7 +379,7 @@ static void enb_main_gui(enb_gui *e, gui *g, event_handler *h, void *database,
   widget_add_child(g, line, w, -1);
   l = new_ticked_ttilog(h, database, "ENB_PHY_DL_TICK", "frame", "subframe",
       "ENB_PHY_ULSCH_UE_DCI", "mcs", 0, -1);
-  v = new_view_tti(10, g, w, new_color(g, "#0c0c72"));
+  v = new_view_tti(10, g, w, new_color(g, "#0c0c72"), 10);
   logger_add_view(l, v);
   e->ul_mcs_logger = l;
 
@@ -394,8 +394,8 @@ static void enb_main_gui(enb_gui *e, gui *g, event_handler *h, void *database,
   widget_add_child(g, col, w, -1);
   container_set_child_growable(g, col, w, 1);
   l = new_throughputlog(h, database, "ENB_PHY_DL_TICK", "frame", "subframe",
-      "ENB_PHY_DLSCH_UE_DCI", "TBS");
-  v = new_view_scrolltti(10, g, w, new_color(g, "#0c0c72"), w2);
+      "ENB_PHY_DLSCH_UE_DCI", "TBS", 10);
+  v = new_view_scrolltti(10, g, w, new_color(g, "#0c0c72"), w2, 10);
   logger_add_view(l, v);
 
   /* UE x UL PHY (truly: DCI) throughput */
@@ -409,8 +409,8 @@ static void enb_main_gui(enb_gui *e, gui *g, event_handler *h, void *database,
   widget_add_child(g, col, w, -1);
   container_set_child_growable(g, col, w, 1);
   l = new_throughputlog(h, database, "ENB_PHY_DL_TICK", "frame", "subframe",
-      "ENB_PHY_ULSCH_UE_DCI", "TBS");
-  v = new_view_scrolltti(10, g, w, new_color(g, "#0c0c72"), w2);
+      "ENB_PHY_ULSCH_UE_DCI", "TBS", 10);
+  v = new_view_scrolltti(10, g, w, new_color(g, "#0c0c72"), w2, 10);
   logger_add_view(l, v);
 
   /* downlink/uplink UE DCIs */

common/utils/T/tracer/gnb_mac.c

@@ -0,0 +1,332 @@
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <unistd.h>
+
+#include "utils.h"
+#include "gui/gui.h"
+#include "view/view.h"
+#include "logger/logger.h"
+#include "filter/filter.h"
+#include "database.h"
+#include "configuration.h"
+#include "handler.h"
+
+typedef struct {
+  int used;
+} ue_list_item;
+
+typedef struct {
+  widget *container;
+} gnb_gui;
+
+typedef struct {
+  int socket;
+  gui *g;
+  gnb_gui *gg;
+  event_handler *h;
+  void *database;
+  ue_list_item ues[65536];
+  int rnti_arg;
+} gnb_data;
+
+void *gui_thread(void *_g)
+{
+  gui *g = _g;
+  gui_loop(g);
+  return 0;
+}
+
+void add_ue(gnb_data *gd, int rnti, int lcid)
+{
+  event_handler *h = gd->h;
+  void *database = gd->database;
+  gnb_gui *gg = gd->gg;
+  gui *g = gd->g;
+  widget *line;
+  widget *w, *w2, *col;
+  view *v;
+  logger *l;
+
+  gd->ues[rnti].used = 1;
+
+  line = new_container(g, HORIZONTAL);
+  widget_add_child(g, gg->container, line, -1);
+
+  /* label */
+  char s[512];
+  sprintf(s, "UE %d", rnti);
+  widget *label = new_label(g, s);
+  widget_add_child(g, line, label, -1);
+
+  /* DL MCS */
+  w = new_xy_plot(g, 128, 55, "DL mcs", 20);
+  xy_plot_set_range(g, w, 0, 1024*20, -2, 30);
+  widget_add_child(g, line, w, -1);
+  l = new_ticked_ttilog(h, database, "GNB_PHY_DL_TICK", "frame", "slot",
+      "GNB_MAC_DL", "mcs", 0, -1);
+  v = new_view_tti(10, g, w, new_color(g, "#0c0c72"), 20);
+  logger_add_view(l, v);
+  logger_set_filter(l,
+      filter_eq(
+        filter_evarg(database, "GNB_MAC_DL", "rnti"),
+        filter_int(rnti)));
+
+  /* DL TBS */
+  col = new_container(g, VERTICAL);
+  widget_add_child(g, line, col, -1);
+  w = new_xy_plot(g, 70, 10, "DL tbs", 35);
+  w2 = new_textarea(g, 70, 11, 64);
+  xy_plot_set_range(g, w, 0, 1024*20, 0, 1000);
+  xy_plot_set_tick_type(g, w, XY_PLOT_SCROLL_TICK);
+  widget_add_child(g, col, w2, -1);
+  widget_add_child(g, col, w, -1);
+  container_set_child_growable(g, col, w, 1);
+  l = new_ticked_ttilog(h, database, "GNB_PHY_DL_TICK", "frame", "slot",
+      "GNB_MAC_DL", "tbs", 0, -1);
+  v = new_view_tti(10, g, w, new_color(g, "#0c0c72"), 20);
+  view_tti_enable_automax(v, w2);
+  logger_add_view(l, v);
+  logger_set_filter(l,
+      filter_eq(
+        filter_evarg(database, "GNB_MAC_DL", "rnti"),
+        filter_int(rnti)));
+
+  /* DL lcid throughput */
+  col = new_container(g, VERTICAL);
+  widget_add_child(g, line, col, -1);
+  sprintf(s, "DL lcid %d", lcid);
+  w = new_xy_plot(g, 70, 10, s, 35);
+  w2 = new_textarea(g, 70, 11, 64);
+  xy_plot_set_range(g, w, 0, 1000, 0, 100000);
+  xy_plot_set_tick_type(g, w, XY_PLOT_SCROLL_TICK);
+  widget_add_child(g, col, w2, -1);
+  widget_add_child(g, col, w, -1);
+  container_set_child_growable(g, col, w, 1);
+  l = new_throughputlog(h, database, "GNB_PHY_DL_TICK", "frame", "slot",
+      "GNB_MAC_LCID_DL", "data_size", 20);
+  v = new_view_scrolltti(10, g, w, new_color(g, "#0c0c72"), w2, 20);
+  logger_add_view(l, v);
+  logger_set_filter(l,
+      filter_and(
+        filter_eq(
+          filter_evarg(database, "GNB_MAC_LCID_DL", "rnti"),
+          filter_int(rnti)),
+        filter_eq(
+          filter_evarg(database, "GNB_MAC_LCID_DL", "lcid"),
+          filter_int(lcid))));
+
+  /* UL MCS */
+  w = new_xy_plot(g, 128, 55, "UL mcs", 20);
+  xy_plot_set_range(g, w, 0, 1024*20, -2, 30);
+  widget_add_child(g, line, w, -1);
+  /* GNB_PHY_DL_TICK (and not UL) because GNB_MAC_UL is from DCI */
+  l = new_ticked_ttilog(h, database, "GNB_PHY_DL_TICK", "frame", "slot",
+      "GNB_MAC_UL", "mcs", 0, -1);
+  v = new_view_tti(10, g, w, new_color(g, "#0c0c72"), 20);
+  logger_add_view(l, v);
+  logger_set_filter(l,
+      filter_eq(
+        filter_evarg(database, "GNB_MAC_UL", "rnti"),
+        filter_int(rnti)));
+
+  /* UL TBS */
+  col = new_container(g, VERTICAL);
+  widget_add_child(g, line, col, -1);
+  w = new_xy_plot(g, 70, 10, "UL tbs", 35);
+  w2 = new_textarea(g, 70, 11, 64);
+  xy_plot_set_range(g, w, 0, 1024*20, 0, 1000);
+  xy_plot_set_tick_type(g, w, XY_PLOT_SCROLL_TICK);
+  widget_add_child(g, col, w2, -1);
+  widget_add_child(g, col, w, -1);
+  container_set_child_growable(g, col, w, 1);
+  /* GNB_PHY_DL_TICK (and not UL) because GNB_MAC_UL is from DCI */
+  l = new_ticked_ttilog(h, database, "GNB_PHY_DL_TICK", "frame", "slot",
+      "GNB_MAC_UL", "tbs", 0, -1);
+  v = new_view_tti(10, g, w, new_color(g, "#0c0c72"), 20);
+  view_tti_enable_automax(v, w2);
+  logger_add_view(l, v);
+  logger_set_filter(l,
+      filter_eq(
+        filter_evarg(database, "GNB_MAC_UL", "rnti"),
+        filter_int(rnti)));
+
+  /* UL lcid throughput */
+  col = new_container(g, VERTICAL);
+  widget_add_child(g, line, col, -1);
+  sprintf(s, "UL lcid %d", lcid);
+  w = new_xy_plot(g, 70, 10, s, 35);
+  w2 = new_textarea(g, 70, 11, 64);
+  xy_plot_set_range(g, w, 0, 1000, 0, 100000);
+  xy_plot_set_tick_type(g, w, XY_PLOT_SCROLL_TICK);
+  widget_add_child(g, col, w2, -1);
+  widget_add_child(g, col, w, -1);
+  container_set_child_growable(g, col, w, 1);
+  l = new_throughputlog(h, database, "GNB_PHY_UL_TICK", "frame", "slot",
+      "GNB_MAC_LCID_UL", "data_size", 20);
+  v = new_view_scrolltti(10, g, w, new_color(g, "#0c0c72"), w2, 20);
+  logger_add_view(l, v);
+  logger_set_filter(l,
+      filter_and(
+        filter_eq(
+          filter_evarg(database, "GNB_MAC_LCID_UL", "rnti"),
+          filter_int(rnti)),
+        filter_eq(
+          filter_evarg(database, "GNB_MAC_LCID_UL", "lcid"),
+          filter_int(lcid))));
+
+  /* lcid RLC tx buffer occupancy */
+  col = new_container(g, VERTICAL);
+  widget_add_child(g, line, col, -1);
+  sprintf(s, "rlc tx %d", lcid);
+  w = new_xy_plot(g, 70, 10, s, 35);
+  w2 = new_textarea(g, 70, 11, 64);
+  xy_plot_set_range(g, w, 0, 1024*20, 0, 1000);
+  xy_plot_set_tick_type(g, w, XY_PLOT_SCROLL_TICK);
+  widget_add_child(g, col, w2, -1);
+  widget_add_child(g, col, w, -1);
+  container_set_child_growable(g, col, w, 1);
+  l = new_ticked_ttilog(h, database, "GNB_PHY_DL_TICK", "frame", "slot",
+      "GNB_MAC_LCID_DL", "tx_list_occupancy", 0, -1);
+  v = new_view_tti(10, g, w, new_color(g, "#0c0c72"), 20);
+  view_tti_enable_automax(v, w2);
+  logger_add_view(l, v);
+  logger_set_filter(l,
+      filter_and(
+        filter_eq(
+          filter_evarg(database, "GNB_MAC_LCID_DL", "rnti"),
+          filter_int(rnti)),
+        filter_eq(
+          filter_evarg(database, "GNB_MAC_LCID_DL", "lcid"),
+          filter_int(lcid))));
+}
+
+void build_gui(gnb_gui *gg, gui *g)
+{
+  widget *main_window;
+
+  main_window = new_toplevel_window(g, 900, 1000, "gNB MAC tracer");
+  gg->container = new_container(g, VERTICAL);
+  widget_add_child(g, main_window, gg->container, -1);
+}
+
+void usage(void)
+{
+  printf(
+"options:\n"
+"    -d <database file>        this option is mandatory\n"
+"    -ip <host>                connect to given IP address (default %s)\n"
+"    -p <port>                 connect to given port (default %d)\n",
+  DEFAULT_REMOTE_IP,
+  DEFAULT_REMOTE_PORT
+  );
+  exit(1);
+}
+
+void check_new_ue(void *_gd, event e)
+{
+  gnb_data *gd = _gd;
+  int rnti = e.e[gd->rnti_arg].i;
+  if (!gd->ues[rnti].used)
+    add_ue(gd, rnti, 4);
+}
+
+int main(int n, char **v)
+{
+  gui *g;
+  gnb_gui gg;
+  gnb_data *gd;
+  char *database_filename = NULL;
+  void *database;
+  char *ip = DEFAULT_REMOTE_IP;
+  int port = DEFAULT_REMOTE_PORT;
+  event_handler *h;
+  int *is_on;
+  int number_of_events;
+  int i;
+
+  for (i = 1; i < n; i++) {
+    if (!strcmp(v[i], "-h") || !strcmp(v[i], "--help")) usage();
+    if (!strcmp(v[i], "-d")) { if (i > n-2) usage(); database_filename = v[++i]; continue; }
+    if (!strcmp(v[i], "-ip")) { if (i > n-2) usage(); ip = v[++i]; continue; }
+    if (!strcmp(v[i], "-p")) { if (i > n-2) usage(); port = atoi(v[++i]); continue; }
+    usage();
+  }
+
+  if (database_filename == NULL) {
+    printf("ERROR: provide a database file (-d)\n");
+    exit(1);
+  }
+
+  database = parse_database(database_filename);
+
+  load_config_file(database_filename);
+
+  number_of_events = number_of_ids(database);
+  is_on = calloc(number_of_events, sizeof(int));
+  if (is_on == NULL) abort();
+
+  on_off(database, "GNB_PHY_DL_TICK", is_on, 1);
+  on_off(database, "GNB_MAC_DL", is_on, 1);
+  on_off(database, "GNB_MAC_LCID_DL", is_on, 1);
+  on_off(database, "GNB_PHY_UL_TICK", is_on, 1);
+  on_off(database, "GNB_MAC_UL", is_on, 1);
+  on_off(database, "GNB_MAC_LCID_UL", is_on, 1);
+
+  gd = calloc(1, sizeof(*gd)); if (!gd) abort();
+
+  h = new_handler(database);
+
+  g = gui_init();
+  new_thread(gui_thread, g);
+
+  build_gui(&gg, g);
+
+  gd->g = g;
+  gd->gg = &gg;
+  gd->h = h;
+  gd->database = database;
+
+  /* get rnti_arg from event GNB_MAC_DL */
+  int event_id = event_id_from_name(database, "GNB_MAC_DL");
+  database_event_format f = get_format(database, event_id);
+  gd->rnti_arg = -1;
+  for (i = 0; i < f.count; i++) {
+    if (!strcmp(f.name[i], "rnti")) gd->rnti_arg = i;
+  }
+  if (gd->rnti_arg == -1) {
+    printf("fatal: event 'GNB_MAC_DL' does not have argument 'rnti'\n");
+    exit(1);
+  }
+  if (strcmp(f.type[gd->rnti_arg], "int") != 0) {
+    printf("fatal: argument 'rnti' of event 'GNB_MAC_DL' is not 'int'\n");
+    exit(1);
+  }
+
+  register_handler_function(h, event_id, check_new_ue, gd);
+
+  OBUF ebuf = {.osize = 0, .omaxsize = 0, .obuf = NULL};
+
+  gd->socket = -1;
+
+restart:
+  clear_remote_config();
+  if (gd->socket != -1) close(gd->socket);
+  gd->socket = connect_to(ip, port);
+
+  /* send the first message - activate selected traces */
+  int t = 1;
+  if (socket_send(gd->socket, &t, 1) == -1 ||
+      socket_send(gd->socket, &number_of_events, sizeof(int)) == -1 ||
+      socket_send(gd->socket, is_on, number_of_events * sizeof(int)) == -1)
+    goto restart;
+
+  while (1) {
+    event e;
+    e = get_event(gd->socket, &ebuf, database);
+    if (e.type == -1) goto restart;
+    handle_event(h, e);
+  }
+
+  return 0;
+}

common/utils/T/tracer/gui/gui.h

@@ -54,6 +54,8 @@ void container_set_child_growable(gui *_gui, widget *_this,
 int xy_plot_new_plot(gui *gui, widget *this, int color);
 void xy_plot_set_range(gui *gui, widget *this,
     float xmin, float xmax, float ymin, float ymax);
+void xy_plot_get_range(gui *gui, widget *this,
+    float *xmin, float *xmax, float *ymin, float *ymax);
 void xy_plot_set_points(gui *gui, widget *this,
     int plot, int npoints, float *x, float *y);
 void xy_plot_get_dimensions(gui *gui, widget *this, int *width, int *height);

common/utils/T/tracer/gui/xy_plot.c

@@ -280,6 +280,22 @@ void xy_plot_set_range(gui *_gui, widget *_this,
   gunlock(g);
 }
 
+void xy_plot_get_range(gui *_gui, widget *_this,
+    float *xmin, float *xmax, float *ymin, float *ymax)
+{
+  struct gui *g = _gui;
+  struct xy_plot_widget *this = _this;
+
+  glock(g);
+
+  *xmin = this->xmin;
+  *xmax = this->xmax;
+  *ymin = this->ymin;
+  *ymax = this->ymax;
+
+  gunlock(g);
+}
+
 void xy_plot_set_points(gui *_gui, widget *_this, int plot,
     int npoints, float *x, float *y)
 {

common/utils/T/tracer/logger/logger.h

@@ -15,8 +15,8 @@ logger *new_ticked_ttilog(void *event_handler, void *database,
     char *event_name, char *data_varname,
     int convert_to_dB, float empty_value);
 logger *new_throughputlog(void *event_handler, void *database,
-    char *tick_event_name, char *frame_varname, char *subframe_varname,
-    char *event_name, char *data_varname);
+    char *tick_event_name, char *frame_varname, char *tick_varname,
+    char *event_name, char *data_varname, int ticks_per_frame);
 logger *new_timelog(void *event_handler, void *database, char *event_name);
 logger *new_ticklog(void *event_handler, void *database,
     char *event_name, char *frame_name, char *subframe_name);

common/utils/T/tracer/logger/throughputlog.c

@@ -13,13 +13,14 @@ struct throughputlog {
   struct logger common;
   void *database;
   int tick_frame_arg;
-  int tick_subframe_arg;
+  int tick_tick_arg;
   int data_frame_arg;
-  int data_subframe_arg;
+  int data_tick_arg;
   int data_arg;
   int last_tick_frame;
-  int last_tick_subframe;
-  unsigned long bits[1000];
+  int last_tick_tick;
+  int ticks_per_frame;
+  unsigned long *bits;
   unsigned long total;
   int insert_point;
   char *tick_event_name;
@@ -32,16 +33,16 @@ static void _event(void *p, event e)
 {
   struct throughputlog *l = p;
   int frame;
-  int subframe;
+  int tick;
   unsigned long value;
 
   if (l->common.filter != NULL && filter_eval(l->common.filter, e) == 0)
     return;
 
   frame = e.e[l->data_frame_arg].i;
-  subframe = e.e[l->data_subframe_arg].i;
+  tick = e.e[l->data_tick_arg].i;
 
-  if (frame != l->last_tick_frame || subframe != l->last_tick_subframe) {
+  if (frame != l->last_tick_frame || tick != l->last_tick_tick) {
     printf("WARNING: %s:%d: data comes without previous tick!\n",
            __FILE__, __LINE__);
     return;
@@ -62,28 +63,34 @@ static void _tick_event(void *p, event e)
   struct throughputlog *l = p;
   int i;
   int frame;
-  int subframe;
+  int tick;
 
   if (l->tick_filter != NULL && filter_eval(l->tick_filter, e) == 0)
     return;
 
   frame = e.e[l->tick_frame_arg].i;
-  subframe = e.e[l->tick_subframe_arg].i;
+  tick = e.e[l->tick_tick_arg].i;
 
   for (i = 0; i < l->common.vsize; i++)
-    l->common.v[i]->append(l->common.v[i], frame, subframe, (double)l->total);
+    l->common.v[i]->append(l->common.v[i], frame, tick, (double)l->total);
 
-  l->insert_point = (l->insert_point + 1) % 1000;
-  l->total -= l->bits[l->insert_point];
-  l->bits[l->insert_point] = 0;
+  while (l->last_tick_frame != frame || l->last_tick_tick != tick) {
+    l->insert_point = (l->insert_point + 1) % (100 * l->ticks_per_frame);
+    l->total -= l->bits[l->insert_point];
+    l->bits[l->insert_point] = 0;
+    l->last_tick_tick++;
+    if (l->last_tick_tick == l->ticks_per_frame) {
+      l->last_tick_tick = 0;
+      l->last_tick_frame++;
+      l->last_tick_frame %= 1024;
+    }
+  }
 
-  l->last_tick_frame = frame;
-  l->last_tick_subframe = subframe;
 }
 
 logger *new_throughputlog(event_handler *h, void *database,
-    char *tick_event_name, char *frame_varname, char *subframe_varname,
-    char *event_name, char *data_varname)
+    char *tick_event_name, char *frame_varname, char *tick_varname,
+    char *event_name, char *data_varname, int ticks_per_frame)
 {
   struct throughputlog *ret;
   int event_id;
@@ -106,21 +113,21 @@ logger *new_throughputlog(event_handler *h, void *database,
 
   f = get_format(database, event_id);
 
-  /* look for frame and subframe */
+  /* look for frame and tick */
   ret->tick_frame_arg = -1;
-  ret->tick_subframe_arg = -1;
+  ret->tick_tick_arg = -1;
   for (i = 0; i < f.count; i++) {
     if (!strcmp(f.name[i], frame_varname)) ret->tick_frame_arg = i;
-    if (!strcmp(f.name[i], subframe_varname)) ret->tick_subframe_arg = i;
+    if (!strcmp(f.name[i], tick_varname)) ret->tick_tick_arg = i;
   }
   if (ret->tick_frame_arg == -1) {
     printf("%s:%d: frame argument '%s' not found in event '%s'\n",
         __FILE__, __LINE__, frame_varname, event_name);
     abort();
   }
-  if (ret->tick_subframe_arg == -1) {
-    printf("%s:%d: subframe argument '%s' not found in event '%s'\n",
-        __FILE__, __LINE__, subframe_varname, event_name);
+  if (ret->tick_tick_arg == -1) {
+    printf("%s:%d: tick argument '%s' not found in event '%s'\n",
+        __FILE__, __LINE__, tick_varname, event_name);
     abort();
   }
   if (strcmp(f.type[ret->tick_frame_arg], "int") != 0) {
@@ -128,9 +135,9 @@ logger *new_throughputlog(event_handler *h, void *database,
         __FILE__, __LINE__, frame_varname);
     abort();
   }
-  if (strcmp(f.type[ret->tick_subframe_arg], "int") != 0) {
+  if (strcmp(f.type[ret->tick_tick_arg], "int") != 0) {
     printf("%s:%d: argument '%s' has wrong type (should be 'int')\n",
-        __FILE__, __LINE__, subframe_varname);
+        __FILE__, __LINE__, tick_varname);
     abort();
   }
 
@@ -141,13 +148,13 @@ logger *new_throughputlog(event_handler *h, void *database,
 
   f = get_format(database, event_id);
 
-  /* look for frame, subframe and data args */
+  /* look for frame, tick and data args */
   ret->data_frame_arg = -1;
-  ret->data_subframe_arg = -1;
+  ret->data_tick_arg = -1;
   ret->data_arg = -1;
   for (i = 0; i < f.count; i++) {
     if (!strcmp(f.name[i], frame_varname)) ret->data_frame_arg = i;
-    if (!strcmp(f.name[i], subframe_varname)) ret->data_subframe_arg = i;
+    if (!strcmp(f.name[i], tick_varname)) ret->data_tick_arg = i;
     if (!strcmp(f.name[i], data_varname)) ret->data_arg = i;
   }
   if (ret->data_frame_arg == -1) {
@@ -155,9 +162,9 @@ logger *new_throughputlog(event_handler *h, void *database,
         __FILE__, __LINE__, frame_varname, event_name);
     abort();
   }
-  if (ret->data_subframe_arg == -1) {
-    printf("%s:%d: subframe argument '%s' not found in event '%s'\n",
-        __FILE__, __LINE__, subframe_varname, event_name);
+  if (ret->data_tick_arg == -1) {
+    printf("%s:%d: tick argument '%s' not found in event '%s'\n",
+        __FILE__, __LINE__, tick_varname, event_name);
     abort();
   }
   if (ret->data_arg == -1) {
@@ -170,9 +177,9 @@ logger *new_throughputlog(event_handler *h, void *database,
         __FILE__, __LINE__, frame_varname);
     abort();
   }
-  if (strcmp(f.type[ret->data_subframe_arg], "int") != 0) {
+  if (strcmp(f.type[ret->data_tick_arg], "int") != 0) {
     printf("%s:%d: argument '%s' has wrong type (should be 'int')\n",
-        __FILE__, __LINE__, subframe_varname);
+        __FILE__, __LINE__, tick_varname);
     abort();
   }
   if (strcmp(f.type[ret->data_arg], "int") != 0 &&
@@ -183,6 +190,10 @@ logger *new_throughputlog(event_handler *h, void *database,
     abort();
   }
 
+  ret->ticks_per_frame = ticks_per_frame;
+  ret->bits = calloc(100 * ticks_per_frame, sizeof(unsigned long));
+  if (ret->bits == NULL) abort();
+
   return ret;
 }
 

common/utils/T/tracer/ue.c

@@ -250,15 +250,15 @@ static void ue_main_gui(ue_gui *e, gui *g, event_handler *h, void *database,
   widget_add_child(g, line, w, -1);
   xy_plot_set_range(g, w, 0, 1024*10, -130, 35);
   /*l = new_ticked_ttilog(h, database,"UE_PHY_DL_TICK", "frame", "subframe","UE_PHY_MEAS", "rssi", 0, -1);
-  v = new_view_tti(10, g, w, new_color(g, "#720c0c"));
+  v = new_view_tti(10, g, w, new_color(g, "#720c0c"), 10);
   logger_add_view(l, v);
   e->phy_meas_logger = l;*/
   l = new_ticked_ttilog(h, database,"UE_PHY_DL_TICK", "frame", "subframe","UE_PHY_MEAS", "rsrp", 0, -1);
-  v = new_view_tti(10, g, w, new_color(g, "#0c0c72"));
+  v = new_view_tti(10, g, w, new_color(g, "#0c0c72"), 10);
   logger_add_view(l, v);
   e->phy_meas_logger = l;
   l = new_ticked_ttilog(h, database,"UE_PHY_DL_TICK", "frame", "subframe","UE_PHY_MEAS", "snr", 0, -1);
-  v = new_view_tti(10, g, w, new_color(g, "#0c720c"));
+  v = new_view_tti(10, g, w, new_color(g, "#0c720c"), 10);
   logger_add_view(l, v);
   e->phy_meas_logger = l;
 
@@ -277,7 +277,7 @@ static void ue_main_gui(ue_gui *e, gui *g, event_handler *h, void *database,
   widget_add_child(g, line, w, -1);
   l = new_ticked_ttilog(h, database, "UE_PHY_DL_TICK", "frame", "subframe",
       "UE_PHY_DLSCH_UE_DCI", "mcs", 0, -1);
-  v = new_view_tti(10, g, w, new_color(g, "#0c0c72"));
+  v = new_view_tti(10, g, w, new_color(g, "#0c0c72"), 10);
   logger_add_view(l, v);
   e->dl_mcs_logger = l;
 
@@ -288,7 +288,7 @@ static void ue_main_gui(ue_gui *e, gui *g, event_handler *h, void *database,
   widget_add_child(g, line, w, -1);
   l = new_ticked_ttilog(h, database, "UE_PHY_DL_TICK", "frame", "subframe",
       "UE_PHY_ULSCH_UE_DCI", "mcs", 0, -1);
-  v = new_view_tti(10, g, w, new_color(g, "#0c0c72"));
+  v = new_view_tti(10, g, w, new_color(g, "#0c0c72"), 10);
   logger_add_view(l, v);
   e->ul_mcs_logger = l;
 
@@ -301,12 +301,12 @@ static void ue_main_gui(ue_gui *e, gui *g, event_handler *h, void *database,
   xy_plot_set_range(g, w, 0, 1024*10, -30, 50);
   l = new_ttilog(h, database,
       "UE_PHY_PUSCH_TX_POWER", "frame", "subframe", "p0_pusch", 0);
-  v = new_view_tti(10, g, w, new_color(g, "#0c0c72"));
+  v = new_view_tti(10, g, w, new_color(g, "#0c0c72"), 10);
   logger_add_view(l, v);
   e->pusch_power_logger = l;
   l = new_ttilog(h, database,
       "UE_PHY_PUSCH_TX_POWER", "frame", "subframe", "ampl", 1);
-  v = new_view_tti(10, g, w, new_color(g, "#720c0c"));
+  v = new_view_tti(10, g, w, new_color(g, "#720c0c"), 10);
   logger_add_view(l, v);
   e->pusch_ampl_logger = l;
 
@@ -319,12 +319,12 @@ static void ue_main_gui(ue_gui *e, gui *g, event_handler *h, void *database,
   xy_plot_set_range(g, w, 0, 1024*10, -30, 50);
   l = new_ttilog(h, database,
       "UE_PHY_PUCCH_TX_POWER", "frame", "subframe", "p0_pucch", 0);
-  v = new_view_tti(10, g, w, new_color(g, "#0c0c72"));
+  v = new_view_tti(10, g, w, new_color(g, "#0c0c72"), 10);
   logger_add_view(l, v);
   e->pucch_power_logger = l;
   l = new_ttilog(h, database,
       "UE_PHY_PUCCH_TX_POWER", "frame", "subframe", "ampl", 1);
-  v = new_view_tti(10, g, w, new_color(g, "#720c0c"));
+  v = new_view_tti(10, g, w, new_color(g, "#720c0c"), 10);
   logger_add_view(l, v);
   e->pucch_ampl_logger = l;
 
@@ -335,22 +335,22 @@ static void ue_main_gui(ue_gui *e, gui *g, event_handler *h, void *database,
   xy_plot_set_range(g, w, 0, 1024*10, -10, 80);
   l = new_ttilog(h, database,
       "UE_PHY_PDSCH_ENERGY", "frame", "subframe", "pdsch_ch_level00", 1);
-  v = new_view_tti(10, g, w, new_color(g, "#ff0000"));
+  v = new_view_tti(10, g, w, new_color(g, "#ff0000"), 10);
   logger_add_view(l, v);
   e->pdsch_energy_logger = l;
   l = new_ttilog(h, database,
       "UE_PHY_PDSCH_ENERGY", "frame", "subframe", "pdsch_ch_level01", 1);
-  v = new_view_tti(10, g, w, new_color(g, "#00ff00"));
+  v = new_view_tti(10, g, w, new_color(g, "#00ff00"), 10);
   logger_add_view(l, v);
   e->pdsch_energy_logger = l;
   l = new_ttilog(h, database,
       "UE_PHY_PDSCH_ENERGY", "frame", "subframe", "pdsch_ch_level10", 1);
-  v = new_view_tti(10, g, w, new_color(g, "#0f0f0f"));
+  v = new_view_tti(10, g, w, new_color(g, "#0f0f0f"), 10);
   logger_add_view(l, v);
   e->pdsch_energy_logger = l;
   l = new_ttilog(h, database,
       "UE_PHY_PDSCH_ENERGY", "frame", "subframe", "pdsch_ch_level11", 1);
-  v = new_view_tti(10, g, w, new_color(g, "#0000ff"));
+  v = new_view_tti(10, g, w, new_color(g, "#0000ff"), 10);
   logger_add_view(l, v);
   e->pdsch_energy_logger = l;
 
@@ -361,22 +361,22 @@ static void ue_main_gui(ue_gui *e, gui *g, event_handler *h, void *database,
   xy_plot_set_range(g, w, 0, 1024*10, -10, 80);
   l = new_ttilog(h, database,
       "UE_PHY_PDCCH_ENERGY", "frame", "subframe", "pdcch_ch_level00", 1);
-  v = new_view_tti(10, g, w, new_color(g, "#ff0000"));
+  v = new_view_tti(10, g, w, new_color(g, "#ff0000"), 10);
   logger_add_view(l, v);
   e->pdcch_energy_logger = l;
   l = new_ttilog(h, database,
       "UE_PHY_PDCCH_ENERGY", "frame", "subframe", "pdcch_ch_level01", 1);
-  v = new_view_tti(10, g, w, new_color(g, "#00ff00"));
+  v = new_view_tti(10, g, w, new_color(g, "#00ff00"), 10);
   logger_add_view(l, v);
   e->pdcch_energy_logger = l;
   l = new_ttilog(h, database,
       "UE_PHY_PDCCH_ENERGY", "frame", "subframe", "pdcch_ch_level10", 1);
-  v = new_view_tti(10, g, w, new_color(g, "#0f0f0f"));
+  v = new_view_tti(10, g, w, new_color(g, "#0f0f0f"), 10);
   logger_add_view(l, v);
   e->pdcch_energy_logger = l;
   l = new_ttilog(h, database,
       "UE_PHY_PDCCH_ENERGY", "frame", "subframe", "pdcch_ch_level11", 1);
-  v = new_view_tti(10, g, w, new_color(g, "#0000ff"));
+  v = new_view_tti(10, g, w, new_color(g, "#0000ff"), 10);
   logger_add_view(l, v);
   e->pdcch_energy_logger = l;
 

common/utils/T/tracer/view/scrolltti.c

@@ -15,11 +15,14 @@ struct scrolltti {
   int plot;
   float refresh_rate;
   pthread_mutex_t lock;
-  unsigned long data[1000];
-  unsigned long total;  /* sum data[0..999] to have smoother value printed */
-  float xout[1000];
-  float yout[1000];
+  int ticks_per_frame;
+  unsigned long *data;
+  unsigned long total;  /* sum data[0..n] to have smoother value printed */
+  float *xout;
+  float *yout;
   int insert_point;
+  int next_tick_frame;
+  int next_tick_tick;
 };
 
 /* this array is used to get Y range 1000, 2000, 5000, 10000, ... */
@@ -38,21 +41,21 @@ static void *scrolltti_thread(void *_this)
     /* TODO: optimize */
     p = this->insert_point;
     max = 0;
-    for (i = 0; i < 1000; i++) {
+    for (i = 0; i < this->ticks_per_frame * 100; i++) {
       this->xout[i] = i;
       this->yout[i] = this->data[p];
       if (this->data[p] > max) max = this->data[p];
-      p = (p + 1) % 1000;
+      p = (p + 1) % (this->ticks_per_frame * 100);
     }
-    bps(o, this->total/1000., "b/s");
+    bps(o, this->total/(this->ticks_per_frame * 100.), "b/s");
     textarea_set_text(this->g, this->throughput_textarea, o);
     /* for Y range we want 1000, 2000, 5000, 10000, 20000, 50000, etc. */
     if (max < 1000) max = 1000;
     mlog = pow(10, floor(log10(max)));
     max = tolog[(int)ceil(max/mlog)] * mlog;
-    xy_plot_set_range(this->g, this->w, 0, 1000, 0, max);
+    xy_plot_set_range(this->g, this->w, 0, this->ticks_per_frame * 100, 0, max);
     xy_plot_set_points(this->g, this->w, this->plot,
-        1000, this->xout, this->yout);
+        this->ticks_per_frame * 100, this->xout, this->yout);
     if (pthread_mutex_unlock(&this->lock)) abort();
     sleepms(1000/this->refresh_rate);
   }
@@ -65,20 +68,40 @@ static void clear(view *this)
   /* TODO */
 }
 
-static void append(view *_this, int frame, int subframe, double value)
+static void insert(struct scrolltti *this, double value)
+{
+  this->total -= this->data[this->insert_point];
+  this->data[this->insert_point] = value;
+  this->total += this->data[this->insert_point];
+  this->insert_point = (this->insert_point + 1) % (this->ticks_per_frame * 100);
+}
+
+static void next_tick(struct scrolltti *this)
+{
+  this->next_tick_tick++;
+  if (this->next_tick_tick == this->ticks_per_frame) {
+    this->next_tick_tick = 0;
+    this->next_tick_frame++;
+    this->next_tick_frame %= 1024;
+  }
+}
+
+static void append(view *_this, int frame, int tick, double value)
 {
   struct scrolltti *this = (struct scrolltti *)_this;
 
   if (pthread_mutex_lock(&this->lock)) abort();
-  this->total -= this->data[this->insert_point];
-  this->data[this->insert_point] = value;
-  this->total += this->data[this->insert_point];
-  this->insert_point = (this->insert_point + 1) % 1000;
+  while (this->next_tick_frame != frame || this->next_tick_tick != tick) {
+    insert(this, 0);
+    next_tick(this);
+  }
+  insert(this, value);
+  next_tick(this);
   if (pthread_mutex_unlock(&this->lock)) abort();
 }
 
 view *new_view_scrolltti(float refresh_rate, gui *g, widget *w, int color,
-    widget *throughput_textarea)
+    widget *throughput_textarea, int ticks_per_frame)
 {
   struct scrolltti *ret = calloc(1, sizeof(struct scrolltti));
   if (ret == NULL) abort();
@@ -92,6 +115,14 @@ view *new_view_scrolltti(float refresh_rate, gui *g, widget *w, int color,
   ret->throughput_textarea = throughput_textarea;
   ret->plot = xy_plot_new_plot(g, w, color);
 
+  ret->ticks_per_frame = ticks_per_frame;
+  ret->data = calloc(ticks_per_frame * 100, sizeof(unsigned long));
+  if (ret->data == NULL) abort();
+  ret->xout = calloc(ticks_per_frame * 100, sizeof(float));
+  if (ret->xout == NULL) abort();
+  ret->yout = calloc(ticks_per_frame * 100, sizeof(float));
+  if (ret->yout == NULL) abort();
+
   if (pthread_mutex_init(&ret->lock, NULL)) abort();
 
   new_thread(scrolltti_thread, ret);

common/utils/T/tracer/view/tti.c

@@ -5,31 +5,36 @@
 #include <pthread.h>
 #include <stdarg.h>
 #include <string.h>
+#include <math.h>
 
 struct tti {
   view common;
   gui *g;
   widget *w;
+  int automax;
+  widget *w2;
   int plot;
   float refresh_rate;
   pthread_mutex_t lock;
-  float data[1024*10];
-  int valid[1024*10];
-  float xout[1024*10];
-  float yout[1024*10];
+  float *data;
+  int *valid;
+  float *xout;
+  float *yout;
   int last_insert_point;
+  int ticks_per_frame;
 };
 
-static int far_enough(int i, int last_insert, int plot_width)
+static int far_enough(int i, int last_insert, int plot_width,
+    int ticks_per_frame)
 {
   int p1;
   int p2;
   int hole_size_px;
   int hole_size_tti;
   hole_size_px = 10;
-  hole_size_tti = 10240 * hole_size_px / plot_width;
+  hole_size_tti = 1024 * ticks_per_frame * hole_size_px / plot_width;
   p1 = last_insert;
-  p2 = (last_insert + hole_size_tti) % (1024*10);
+  p2 = (last_insert + hole_size_tti) % (1024 * ticks_per_frame);
   if (p1 < p2) {
     return !(i > p1 && i < p2);
   }
@@ -48,15 +53,31 @@ static void *tti_thread(void *_this)
     if (pthread_mutex_lock(&this->lock)) abort();
     xy_plot_get_dimensions(this->g, this->w, &plot_width, &plot_height);
     length = 0;
+    double max = 0;
     /* TODO: optimize */
-    for (i = 0; i < 1024*10; i++)
+    for (i = 0; i < 1024 * this->ticks_per_frame; i++)
       /* do not take points too close after last insertion point */
       if (this->valid[i] &&
-          far_enough(i, this->last_insert_point, plot_width)) {
+          far_enough(i, this->last_insert_point, plot_width,
+                     this->ticks_per_frame)) {
         this->xout[length] = i;
         this->yout[length] = this->data[i];
+        if (this->data[i] > max) max = this->data[i];
         length++;
       }
+    if (this->automax) {
+      char o[128];
+      sprintf(o, "%d", (int)max);
+      textarea_set_text(this->g, this->w2, o);
+      /* for Y range we want 10, 20, 50, 100, 200, 500, etc. */
+      if (max < 10) max = 10;
+      double mlog = pow(10, floor(log10(max)));
+      static int tolog[11] = { -1, 1, 2, 5, 5, 5, 10, 10, 10, 10, 10 };
+      max = tolog[(int)ceil(max/mlog)] * mlog;
+      float xmin, xmax, ymin, ymax;
+      xy_plot_get_range(this->g, this->w, &xmin, &xmax, &ymin, &ymax);
+      xy_plot_set_range(this->g, this->w, xmin, xmax, 0, max);
+    }
     xy_plot_set_points(this->g, this->w, this->plot,
         length, this->xout, this->yout);
     if (pthread_mutex_unlock(&this->lock)) abort();
@@ -71,11 +92,11 @@ static void clear(view *this)
   /* TODO */
 }
 
-static void append(view *_this, int frame, int subframe, double value)
+static void append(view *_this, int frame, int tick, double value)
 {
   struct tti *this = (struct tti *)_this;
   int i;
-  int index = frame * 10 + subframe;
+  int index = frame * this->ticks_per_frame + tick;
 
   if (pthread_mutex_lock(&this->lock)) abort();
 
@@ -84,10 +105,10 @@ static void append(view *_this, int frame, int subframe, double value)
    * this may be wrong if delay between two append is
    * greater than 1024 frames (something like that)
    */
-  i = (this->last_insert_point + 1) % (1024*10);
+  i = (this->last_insert_point + 1) % (1024 * this->ticks_per_frame);
   while (i != index) {
     this->valid[i] = 0;
-    i = (i + 1) % (1024*10);
+    i = (i + 1) % (1024 * this->ticks_per_frame);
   }
 
   this->data[index] = value;
@@ -98,7 +119,8 @@ static void append(view *_this, int frame, int subframe, double value)
   if (pthread_mutex_unlock(&this->lock)) abort();
 }
 
-view *new_view_tti(float refresh_rate, gui *g, widget *w, int color)
+view *new_view_tti(float refresh_rate, gui *g, widget *w, int color,
+    int ticks_per_frame)
 {
   struct tti *ret = calloc(1, sizeof(struct tti));
   if (ret == NULL) abort();
@@ -115,7 +137,24 @@ view *new_view_tti(float refresh_rate, gui *g, widget *w, int color)
 
   if (pthread_mutex_init(&ret->lock, NULL)) abort();
 
+  ret->ticks_per_frame = ticks_per_frame;
+  ret->data = calloc(ticks_per_frame * 1024, sizeof(float));
+  if (ret->data == NULL) abort();
+  ret->valid = calloc(ticks_per_frame * 1024, sizeof(int));
+  if (ret->valid == NULL) abort();
+  ret->xout = calloc(ticks_per_frame * 1024, sizeof(float));
+  if (ret->xout == NULL) abort();
+  ret->yout = calloc(ticks_per_frame * 1024, sizeof(float));
+  if (ret->yout == NULL) abort();
+
   new_thread(tti_thread, ret);
 
   return (view *)ret;
 }
+
+void view_tti_enable_automax(view *_tti, widget *w2)
+{
+  struct tti *tti = (struct tti *)_tti;
+  tti->automax = 1;
+  tti->w2 = w2;
+}

common/utils/T/tracer/view/view.h

@@ -17,15 +17,16 @@ view *new_view_stdout(void);
 view *new_view_textlist(int maxsize, float refresh_rate, gui *g, widget *w);
 view *new_view_xy(int length, float refresh_rate, gui *g, widget *w,
     int color, enum xy_mode mode);
-view *new_view_tti(float refresh_rate, gui *g, widget *w,
-    int color);
+view *new_view_tti(float refresh_rate, gui *g, widget *w, int color,
+    int ticks_per_frame);
 view *new_view_scrolltti(float refresh_rate, gui *g, widget *w,
-    int color, widget *throughput_label);
+    int color, widget *throughput_label, int ticks_per_frame);
 view *new_view_time(int number_of_seconds, float refresh_rate,
     gui *g, widget *w);
 view *new_subview_time(view *time, int line, int color, int size);
 view *new_view_ticktime(float refresh_rate, gui *g, widget *w);
 view *new_subview_ticktime(view *ticktime, int line, int color, int size);
 void ticktime_set_tick(view *ticktime, void *logger);
+void view_tti_enable_automax(view *tti, widget *w2);
 
 #endif /* _VIEW_H_ */

common/utils/nr/nr_common.c

@@ -1465,13 +1465,13 @@ frequency_range_t get_freq_range_from_band(uint16_t band)
   return band <= 256 ? FR1 : FR2;
 }
 
-float get_beta_dmrs_pusch(int num_cdm_groups_no_data, pusch_dmrs_type_t dmrs_type)
+float get_beta_dmrs(int num_cdm_groups_no_data, bool is_type2)
 {
   float beta_dmrs_pusch = 1.0;
   if (num_cdm_groups_no_data == 2) {
     beta_dmrs_pusch = powf(10.0, 3.0 / 20.0);
   } else if (num_cdm_groups_no_data == 3) {
-    if (dmrs_type == pusch_dmrs_type2)
+    if (is_type2)
       beta_dmrs_pusch = powf(10.0, 4.77 / 20.0);
   }
   return beta_dmrs_pusch;

common/utils/nr/nr_common.h

@@ -320,15 +320,15 @@ frequency_range_t get_freq_range_from_arfcn(uint32_t arfcn);
 frequency_range_t get_freq_range_from_band(uint16_t band);
 
 /**
- * @brief Calculates the scaling factor for the ratio of PUSCH EPRE to DMRS EPRE.
+ * @brief Calculates the scaling factor for the ratio of PUSCH/PDSCH EPRE to DMRS EPRE.
  *
  * @param num_cdm_groups_no_data The number of CDM groups without data.
- * @param dmrs_type The DMRS type.
- * @return The calculated beta scaling factor for the ratio of PUSCH EPRE to DMRS EPRE.
+ * @param is_type2 true if calculating for DMRS configuration type 2
+ * @return The calculated beta scaling factor for the ratio of PUSCH/PDSCH EPRE to DMRS EPRE.
  *
- * @note The values are derived from TS 38.214 Table 6.2.2-1.
+ * @note The values are the same for PUSCH and PDSCH and are derived from TS 38.214 Table 6.2.2-1./4.1-1
  */
-float get_beta_dmrs_pusch(int num_cdm_groups_no_data, pusch_dmrs_type_t dmrs_type);
+float get_beta_dmrs(int num_cdm_groups_no_data, bool is_type2);
 
 #define CEILIDIV(a,b) ((a+b-1)/b)
 #define ROUNDIDIV(a,b) (((a<<1)+b)/(b<<1))

common/utils/telnetsrv/CMakeLists.txt

@@ -62,7 +62,7 @@ add_dependencies(telnetsrv telnetsrv_ci)
 
 message(STATUS "Add CI specific telnet functions for nrUE in telnetsrv_ciUE.so")
 add_library(telnetsrv_ciUE MODULE telnetsrv_ciUE.c)
-target_link_libraries(telnetsrv_ciUE PRIVATE asn1_nr_rrc_hdrs asn1_lte_rrc_hdrs)
+target_link_libraries(telnetsrv_ciUE PRIVATE asn1_nr_rrc_hdrs asn1_lte_rrc_hdrs UTIL)
 add_dependencies(telnetsrv telnetsrv_ciUE)
 
 message(STATUS "Add bearer specific telnet functions in libtelnetsrv_bearer.so")

doc/Aerial_FAPI_Split_Tutorial.md

@@ -230,10 +230,10 @@ After preparing the L1 software, the container needs to be committed to create a
 In this file the RU MAC address needs to be specified before commiting the image.
 
 ```bash
-~$ docker commit nv-cubb cubb-build:25-1
+~$ docker commit nv-cubb cubb-build:25-2
 ~$ docker image ls
 ..
-cubb-build                                    25-1                                           824156e0334c   2 weeks ago    23.9GB
+cubb-build                                    25-2                                           824156e0334c   2 weeks ago    23.9GB
 -..
 ```
 

doc/ORAN_FHI7.2_Tutorial.md

@@ -1235,7 +1235,13 @@ You should be able to connect a UE now.
 
 
 # OAI Management Plane
-We support Configuration Management in OAI gNB, where gNB configures CU-planes, interfaces, TX/RX antennas, and TX/RX carriers for the RU.
+In OAI gNB, we support:
+* Configuration Management: interface(s) creation, configuration of RU CU-planes, Tx/Rx antennas, and Tx/Rx carriers.
+* Performance Management: activation/deactivation of available RU performance measurements and its notification reception with 10s periodicity:
+```bash
+[HW]   [MPLANE] [PM: "192.168.80.9"][RX_ON_TIME  773428][RX_EARLY       0][RX_LATE       0][RX_TOTAL  878881][RX_ON_TIME_C  104744][RX_EARLY_C       0][RX_LATE_C       0][TX_TOTAL  435922]
+```
+
 The reference specifications:
 * `O-RAN.WG4.MP.0-v05.00`
 * `O-RAN.WG4.MP-YANGs-v04.00`
@@ -1375,7 +1381,7 @@ Once the mplane service is successfully enabled on the RU, two new users are bei
 oranbenetel:x:1000:1000::/home/oranbenetel:/bin/sh
 oranext:x:1001:1001::/home/oranext:/bin/sh
 ```
-Create `oranbenetel` home directory:
+OAI gNB requires the `sudo` access group for NETCONF session. In the case of Benetel O-RUs, the corresponding user is `oranbenetel`. Therefore, please create its home directory:
 ```bash
 mkdir /home/oranbenetel && chown oranbenetel:oranbenetel /home/oranbenetel
 ```
@@ -1393,7 +1399,7 @@ The reference gNB configuration file for one Benetel RAN550:
 The reference DU configuration file for two Benetel RAN650:
 [gnb-du.sa.band77.273prb.fhi72.8x8-benetel650_650-mplane.conf](../targets/PROJECTS/GENERIC-NR-5GC/CONF/gnb-du.sa.band77.273prb.fhi72.8x8-benetel650_650-mplane.conf)
 
-In order to run gNB/DU with M-plane, we need to modify the `fhi_72` section in the configuration file.
+In order to run gNB/DU with M-plane, we need to modify Tx gain `att_tx` in RU section, as well as the `fhi_72` section in the configuration file.
 Example for one RU:
 ```bash
 fhi_72 = {
@@ -1404,6 +1410,7 @@ fhi_72 = {
   du_key_pair = ("<path-to>/.ssh/id_rsa.pub", "<path-to>/.ssh/id_rsa");
   du_addr = ("00:11:22:33:44:66", "00:11:22:33:44:67"); # only one needed if one VF configured
   vlan_tag = (9, 9); # only one needed if one VF configured
+  ru_username = ("oranbenetel");
   ru_ip_addr = ("192.168.80.9");
   fh_config = ({
     T1a_cp_dl = (419, 470);
@@ -1423,6 +1430,7 @@ fhi_72 = {
   du_key_pair = ("/home/oaicicd/.ssh/id_rsa.pub", "/home/oaicicd/.ssh/id_rsa");
   du_addr = ("00:11:22:33:44:66", "00:11:22:33:44:67", "00:11:22:33:44:68", "00:11:22:33:44:69"); # only two needed if two VFs configured
   vlan_tag = (9, 9, 11, 11); # only two needed if two VFs configured
+  ru_username = ("oranbenetel", "oranbenetel");
   ru_ip_addr = ("192.168.80.9", "192.168.80.10");
   fh_config = (
 # RAN550 #1
@@ -1451,6 +1459,7 @@ fhi_72 = {
   * `du_key_pair`: ssh public and private keys to authenticate RU with NETCONF
   * `du_addr`: DU MAC address(es) to create CU-plane interface(s) in the RU
   * `vlan_tag`: VLAN U and C plane tags to create CU-plane interface(s) in the RU
+  * `ru_username`: Username with `sudo` access to connect to the RU via M-plane
   * `ru_ip_addr`: RU IP address to connect to the RU via M-plane
   * `dpdk_mem_size`: [*]
   * `dpdk_iova_mode`: [*]
@@ -1557,7 +1566,9 @@ sequenceDiagram
   
   note over du: DU configures xran
   
-  du->ru: 12. DU and RU exchange packets
+  du->ru: 12. DU activates Performance Measurements
+
+  du->ru: 13. DU and RU exchange packets
 ```
 
 
@@ -1596,7 +1607,9 @@ sequenceDiagram
     Band sector ID 0
     CC ID 0
     RU port ID 0
+    max Tx gain 24.0
 [HW]   [MPLANE] Successfully retrieved all the U-plane info - interface name, TX/RX carrier names, and TX/RX endpoint names.
+[HW]   [MPLANE] Successfully retreived all performance measurement names.
 [HW]   [MPLANE] RPC request to RU "192.168.80.9" = <get-schema> for module "ietf-yang-metadata".
 [HW]   [MPLANE] RPC request to RU "192.168.80.9" = <get-schema> for module "yang".
 [HW]   [MPLANE] RPC request to RU "192.168.80.9" = <get-schema> for module "ietf-inet-types".
@@ -1636,6 +1649,59 @@ sequenceDiagram
     <vlan-id xmlns="urn:o-ran:interfaces:1.0">9</vlan-id>
   </interface>
 </interfaces>
+<performance-measurement-objects xmlns="urn:o-ran:performance-management:1.0">
+  <rx-window-measurement-interval>10</rx-window-measurement-interval>
+  <tx-measurement-interval>10</tx-measurement-interval>
+  <notification-interval>10</notification-interval>
+  <rx-window-measurement-objects>
+    <measurement-object>RX_ON_TIME</measurement-object>
+    <active>false</active>
+    <object-unit>RU</object-unit>
+    <report-info>COUNT</report-info>
+  </rx-window-measurement-objects>
+  <rx-window-measurement-objects>
+    <measurement-object>RX_EARLY</measurement-object>
+    <active>false</active>
+    <object-unit>RU</object-unit>
+    <report-info>COUNT</report-info>
+  </rx-window-measurement-objects>
+  <rx-window-measurement-objects>
+    <measurement-object>RX_LATE</measurement-object>
+    <active>false</active>
+    <object-unit>RU</object-unit>
+    <report-info>COUNT</report-info>
+  </rx-window-measurement-objects>
+  <rx-window-measurement-objects>
+    <measurement-object>RX_TOTAL</measurement-object>
+    <active>false</active>
+    <object-unit>RU</object-unit>
+    <report-info>COUNT</report-info>
+  </rx-window-measurement-objects>
+  <rx-window-measurement-objects>
+    <measurement-object>RX_ON_TIME_C</measurement-object>
+    <active>false</active>
+    <object-unit>RU</object-unit>
+    <report-info>COUNT</report-info>
+  </rx-window-measurement-objects>
+  <rx-window-measurement-objects>
+    <measurement-object>RX_EARLY_C</measurement-object>
+    <active>false</active>
+    <object-unit>RU</object-unit>
+    <report-info>COUNT</report-info>
+  </rx-window-measurement-objects>
+  <rx-window-measurement-objects>
+    <measurement-object>RX_LATE_C</measurement-object>
+    <active>false</active>
+    <object-unit>RU</object-unit>
+    <report-info>COUNT</report-info>
+  </rx-window-measurement-objects>
+  <tx-measurement-objects>
+    <measurement-object>TX_TOTAL</measurement-object>
+    <active>false</active>
+    <object-unit>RU</object-unit>
+    <report-info>COUNT</report-info>
+  </tx-measurement-objects>
+</performance-measurement-objects>
 <processing-elements xmlns="urn:o-ran:processing-element:1.0">
   <transport-session-type>ETH-INTERFACE</transport-session-type>
   <ru-elements>
@@ -1729,8 +1795,10 @@ sequenceDiagram
       <iq-bitwidth>9</iq-bitwidth>
       <compression-type>STATIC</compression-type>
     </compression>
-    <cp-length>0</cp-length>
-    <cp-length-other>0</cp-length-other>
+    <frame-structure>193</frame-structure>
+    <cp-type>NORMAL</cp-type>
+    <cp-length>352</cp-length>
+    <cp-length-other>288</cp-length-other>
     <offset-to-absolute-frequency-center>0</offset-to-absolute-frequency-center>
     <e-axcid>
       <o-du-port-bitmask>61440</o-du-port-bitmask>
@@ -1746,8 +1814,10 @@ sequenceDiagram
       <iq-bitwidth>9</iq-bitwidth>
       <compression-type>STATIC</compression-type>
     </compression>
-    <cp-length>0</cp-length>
-    <cp-length-other>0</cp-length-other>
+    <frame-structure>193</frame-structure>
+    <cp-type>NORMAL</cp-type>
+    <cp-length>352</cp-length>
+    <cp-length-other>288</cp-length-other>
     <offset-to-absolute-frequency-center>0</offset-to-absolute-frequency-center>
     <e-axcid>
       <o-du-port-bitmask>61440</o-du-port-bitmask>
@@ -1763,8 +1833,10 @@ sequenceDiagram
       <iq-bitwidth>9</iq-bitwidth>
       <compression-type>STATIC</compression-type>
     </compression>
-    <cp-length>0</cp-length>
-    <cp-length-other>0</cp-length-other>
+    <frame-structure>193</frame-structure>
+    <cp-type>NORMAL</cp-type>
+    <cp-length>352</cp-length>
+    <cp-length-other>288</cp-length-other>
     <offset-to-absolute-frequency-center>0</offset-to-absolute-frequency-center>
     <e-axcid>
       <o-du-port-bitmask>61440</o-du-port-bitmask>
@@ -1780,8 +1852,10 @@ sequenceDiagram
       <iq-bitwidth>9</iq-bitwidth>
       <compression-type>STATIC</compression-type>
     </compression>
-    <cp-length>0</cp-length>
-    <cp-length-other>0</cp-length-other>
+    <frame-structure>193</frame-structure>
+    <cp-type>NORMAL</cp-type>
+    <cp-length>352</cp-length>
+    <cp-length-other>288</cp-length-other>
     <offset-to-absolute-frequency-center>0</offset-to-absolute-frequency-center>
     <e-axcid>
       <o-du-port-bitmask>61440</o-du-port-bitmask>
@@ -1797,9 +1871,15 @@ sequenceDiagram
       <iq-bitwidth>9</iq-bitwidth>
       <compression-type>STATIC</compression-type>
     </compression>
-    <cp-length>0</cp-length>
-    <cp-length-other>0</cp-length-other>
+    <frame-structure>193</frame-structure>
+    <cp-type>NORMAL</cp-type>
+    <cp-length>352</cp-length>
+    <cp-length-other>288</cp-length-other>
     <offset-to-absolute-frequency-center>0</offset-to-absolute-frequency-center>
+    <ul-fft-sampling-offsets>
+      <scs>KHZ_30</scs>
+      <ul-fft-sampling-offset>0</ul-fft-sampling-offset>
+    </ul-fft-sampling-offsets>
     <e-axcid>
       <o-du-port-bitmask>61440</o-du-port-bitmask>
       <band-sector-bitmask>3840</band-sector-bitmask>
@@ -1807,6 +1887,7 @@ sequenceDiagram
       <ru-port-bitmask>15</ru-port-bitmask>
       <eaxc-id>0</eaxc-id>
     </e-axcid>
+    <non-time-managed-delay-enabled>true</non-time-managed-delay-enabled>
   </low-level-rx-endpoints>
   <low-level-rx-endpoints>
     <name>LowLevelRxPrachEndpoint0</name>
@@ -1814,9 +1895,15 @@ sequenceDiagram
       <iq-bitwidth>9</iq-bitwidth>
       <compression-type>STATIC</compression-type>
     </compression>
-    <cp-length>0</cp-length>
-    <cp-length-other>0</cp-length-other>
+    <frame-structure>193</frame-structure>
+    <cp-type>NORMAL</cp-type>
+    <cp-length>352</cp-length>
+    <cp-length-other>288</cp-length-other>
     <offset-to-absolute-frequency-center>0</offset-to-absolute-frequency-center>
+    <ul-fft-sampling-offsets>
+      <scs>KHZ_30</scs>
+      <ul-fft-sampling-offset>0</ul-fft-sampling-offset>
+    </ul-fft-sampling-offsets>
     <e-axcid>
       <o-du-port-bitmask>61440</o-du-port-bitmask>
       <band-sector-bitmask>3840</band-sector-bitmask>
@@ -1824,6 +1911,7 @@ sequenceDiagram
       <ru-port-bitmask>15</ru-port-bitmask>
       <eaxc-id>4</eaxc-id>
     </e-axcid>
+    <non-time-managed-delay-enabled>true</non-time-managed-delay-enabled>
   </low-level-rx-endpoints>
   <low-level-rx-endpoints>
     <name>LowLevelRxEndpoint1</name>
@@ -1831,9 +1919,15 @@ sequenceDiagram
       <iq-bitwidth>9</iq-bitwidth>
       <compression-type>STATIC</compression-type>
     </compression>
-    <cp-length>0</cp-length>
-    <cp-length-other>0</cp-length-other>
+    <frame-structure>193</frame-structure>
+    <cp-type>NORMAL</cp-type>
+    <cp-length>352</cp-length>
+    <cp-length-other>288</cp-length-other>
     <offset-to-absolute-frequency-center>0</offset-to-absolute-frequency-center>
+    <ul-fft-sampling-offsets>
+      <scs>KHZ_30</scs>
+      <ul-fft-sampling-offset>0</ul-fft-sampling-offset>
+    </ul-fft-sampling-offsets>
     <e-axcid>
       <o-du-port-bitmask>61440</o-du-port-bitmask>
       <band-sector-bitmask>3840</band-sector-bitmask>
@@ -1841,6 +1935,7 @@ sequenceDiagram
       <ru-port-bitmask>15</ru-port-bitmask>
       <eaxc-id>1</eaxc-id>
     </e-axcid>
+    <non-time-managed-delay-enabled>true</non-time-managed-delay-enabled>
   </low-level-rx-endpoints>
   <low-level-rx-endpoints>
     <name>LowLevelRxPrachEndpoint1</name>
@@ -1848,9 +1943,15 @@ sequenceDiagram
       <iq-bitwidth>9</iq-bitwidth>
       <compression-type>STATIC</compression-type>
     </compression>
-    <cp-length>0</cp-length>
-    <cp-length-other>0</cp-length-other>
+    <frame-structure>193</frame-structure>
+    <cp-type>NORMAL</cp-type>
+    <cp-length>352</cp-length>
+    <cp-length-other>288</cp-length-other>
     <offset-to-absolute-frequency-center>0</offset-to-absolute-frequency-center>
+    <ul-fft-sampling-offsets>
+      <scs>KHZ_30</scs>
+      <ul-fft-sampling-offset>0</ul-fft-sampling-offset>
+    </ul-fft-sampling-offsets>
     <e-axcid>
       <o-du-port-bitmask>61440</o-du-port-bitmask>
       <band-sector-bitmask>3840</band-sector-bitmask>
@@ -1858,6 +1959,7 @@ sequenceDiagram
       <ru-port-bitmask>15</ru-port-bitmask>
       <eaxc-id>5</eaxc-id>
     </e-axcid>
+    <non-time-managed-delay-enabled>true</non-time-managed-delay-enabled>
   </low-level-rx-endpoints>
   <low-level-rx-endpoints>
     <name>LowLevelRxEndpoint2</name>
@@ -1865,9 +1967,15 @@ sequenceDiagram
       <iq-bitwidth>9</iq-bitwidth>
       <compression-type>STATIC</compression-type>
     </compression>
-    <cp-length>0</cp-length>
-    <cp-length-other>0</cp-length-other>
+    <frame-structure>193</frame-structure>
+    <cp-type>NORMAL</cp-type>
+    <cp-length>352</cp-length>
+    <cp-length-other>288</cp-length-other>
     <offset-to-absolute-frequency-center>0</offset-to-absolute-frequency-center>
+    <ul-fft-sampling-offsets>
+      <scs>KHZ_30</scs>
+      <ul-fft-sampling-offset>0</ul-fft-sampling-offset>
+    </ul-fft-sampling-offsets>
     <e-axcid>
       <o-du-port-bitmask>61440</o-du-port-bitmask>
       <band-sector-bitmask>3840</band-sector-bitmask>
@@ -1875,6 +1983,7 @@ sequenceDiagram
       <ru-port-bitmask>15</ru-port-bitmask>
       <eaxc-id>2</eaxc-id>
     </e-axcid>
+    <non-time-managed-delay-enabled>true</non-time-managed-delay-enabled>
   </low-level-rx-endpoints>
   <low-level-rx-endpoints>
     <name>LowLevelRxPrachEndpoint2</name>
@@ -1882,9 +1991,15 @@ sequenceDiagram
       <iq-bitwidth>9</iq-bitwidth>
       <compression-type>STATIC</compression-type>
     </compression>
-    <cp-length>0</cp-length>
-    <cp-length-other>0</cp-length-other>
+    <frame-structure>193</frame-structure>
+    <cp-type>NORMAL</cp-type>
+    <cp-length>352</cp-length>
+    <cp-length-other>288</cp-length-other>
     <offset-to-absolute-frequency-center>0</offset-to-absolute-frequency-center>
+    <ul-fft-sampling-offsets>
+      <scs>KHZ_30</scs>
+      <ul-fft-sampling-offset>0</ul-fft-sampling-offset>
+    </ul-fft-sampling-offsets>
     <e-axcid>
       <o-du-port-bitmask>61440</o-du-port-bitmask>
       <band-sector-bitmask>3840</band-sector-bitmask>
@@ -1892,6 +2007,7 @@ sequenceDiagram
       <ru-port-bitmask>15</ru-port-bitmask>
       <eaxc-id>6</eaxc-id>
     </e-axcid>
+    <non-time-managed-delay-enabled>true</non-time-managed-delay-enabled>
   </low-level-rx-endpoints>
   <low-level-rx-endpoints>
     <name>LowLevelRxEndpoint3</name>
@@ -1899,9 +2015,15 @@ sequenceDiagram
       <iq-bitwidth>9</iq-bitwidth>
       <compression-type>STATIC</compression-type>
     </compression>
-    <cp-length>0</cp-length>
-    <cp-length-other>0</cp-length-other>
+    <frame-structure>193</frame-structure>
+    <cp-type>NORMAL</cp-type>
+    <cp-length>352</cp-length>
+    <cp-length-other>288</cp-length-other>
     <offset-to-absolute-frequency-center>0</offset-to-absolute-frequency-center>
+    <ul-fft-sampling-offsets>
+      <scs>KHZ_30</scs>
+      <ul-fft-sampling-offset>0</ul-fft-sampling-offset>
+    </ul-fft-sampling-offsets>
     <e-axcid>
       <o-du-port-bitmask>61440</o-du-port-bitmask>
       <band-sector-bitmask>3840</band-sector-bitmask>
@@ -1909,6 +2031,7 @@ sequenceDiagram
       <ru-port-bitmask>15</ru-port-bitmask>
       <eaxc-id>3</eaxc-id>
     </e-axcid>
+    <non-time-managed-delay-enabled>true</non-time-managed-delay-enabled>
   </low-level-rx-endpoints>
   <low-level-rx-endpoints>
     <name>LowLevelRxPrachEndpoint3</name>
@@ -1916,9 +2039,15 @@ sequenceDiagram
       <iq-bitwidth>9</iq-bitwidth>
       <compression-type>STATIC</compression-type>
     </compression>
-    <cp-length>0</cp-length>
-    <cp-length-other>0</cp-length-other>
+    <frame-structure>193</frame-structure>
+    <cp-type>NORMAL</cp-type>
+    <cp-length>352</cp-length>
+    <cp-length-other>288</cp-length-other>
     <offset-to-absolute-frequency-center>0</offset-to-absolute-frequency-center>
+    <ul-fft-sampling-offsets>
+      <scs>KHZ_30</scs>
+      <ul-fft-sampling-offset>0</ul-fft-sampling-offset>
+    </ul-fft-sampling-offsets>
     <e-axcid>
       <o-du-port-bitmask>61440</o-du-port-bitmask>
       <band-sector-bitmask>3840</band-sector-bitmask>
@@ -1926,6 +2055,7 @@ sequenceDiagram
       <ru-port-bitmask>15</ru-port-bitmask>
       <eaxc-id>7</eaxc-id>
     </e-axcid>
+    <non-time-managed-delay-enabled>true</non-time-managed-delay-enabled>
   </low-level-rx-endpoints>
   <tx-array-carriers>
     <name>TxArray0</name>
@@ -1933,7 +2063,7 @@ sequenceDiagram
     <center-of-channel-bandwidth>3950400000</center-of-channel-bandwidth>
     <channel-bandwidth>100000000</channel-bandwidth>
     <active>ACTIVE</active>
-    <gain>0.0</gain>
+    <gain>24.0</gain>
     <downlink-radio-frame-offset>0</downlink-radio-frame-offset>
     <downlink-sfn-offset>0</downlink-sfn-offset>
   </tx-array-carriers>
@@ -1957,7 +2087,7 @@ sequenceDiagram
 [HW]   [MPLANE] Successfully validated candidate datastore for RU "192.168.80.9".
 [HW]   [MPLANE] RPC request to RU "192.168.80.9" = <commit> candidate datastore.
 [HW]   [MPLANE] RPC reply = OK.
-[HW]   [MPLANE] Successfully commited CU-planes configuration into running datastore for RU "192.168.80.9".
+[HW]   [MPLANE] Successfully commited configuration into running datastore for RU "192.168.80.9".
 [HW]   [MPLANE] Usage state = "idle" for RU "192.168.80.9".
 [HW]   [MPLANE] Received notification from RU "192.168.80.9" at (2025-03-29T12:40:23.049085102+00:00)
 {
@@ -2025,9 +2155,71 @@ sequenceDiagram
         "operation": "create"
       },
 ...
-}
-
 [HW]   [MPLANE] RU "192.168.80.9" is now ready.
+[HW]   [MPLANE] RPC request to RU "192.168.80.9" = <edit-config>:
+<performance-measurement-objects xmlns="urn:o-ran:performance-management:1.0">
+  <rx-window-measurement-interval>10</rx-window-measurement-interval>
+  <tx-measurement-interval>10</tx-measurement-interval>
+  <notification-interval>10</notification-interval>
+  <rx-window-measurement-objects>
+    <measurement-object>RX_ON_TIME</measurement-object>
+    <active>true</active>
+    <object-unit>RU</object-unit>
+    <report-info>COUNT</report-info>
+  </rx-window-measurement-objects>
+  <rx-window-measurement-objects>
+    <measurement-object>RX_EARLY</measurement-object>
+    <active>true</active>
+    <object-unit>RU</object-unit>
+    <report-info>COUNT</report-info>
+  </rx-window-measurement-objects>
+  <rx-window-measurement-objects>
+    <measurement-object>RX_LATE</measurement-object>
+    <active>true</active>
+    <object-unit>RU</object-unit>
+    <report-info>COUNT</report-info>
+  </rx-window-measurement-objects>
+  <rx-window-measurement-objects>
+    <measurement-object>RX_TOTAL</measurement-object>
+    <active>true</active>
+    <object-unit>RU</object-unit>
+    <report-info>COUNT</report-info>
+  </rx-window-measurement-objects>
+  <rx-window-measurement-objects>
+    <measurement-object>RX_ON_TIME_C</measurement-object>
+    <active>true</active>
+    <object-unit>RU</object-unit>
+    <report-info>COUNT</report-info>
+  </rx-window-measurement-objects>
+  <rx-window-measurement-objects>
+    <measurement-object>RX_EARLY_C</measurement-object>
+    <active>true</active>
+    <object-unit>RU</object-unit>
+    <report-info>COUNT</report-info>
+  </rx-window-measurement-objects>
+  <rx-window-measurement-objects>
+    <measurement-object>RX_LATE_C</measurement-object>
+    <active>true</active>
+    <object-unit>RU</object-unit>
+    <report-info>COUNT</report-info>
+  </rx-window-measurement-objects>
+  <tx-measurement-objects>
+    <measurement-object>TX_TOTAL</measurement-object>
+    <active>true</active>
+    <object-unit>RU</object-unit>
+    <report-info>COUNT</report-info>
+  </tx-measurement-objects>
+</performance-measurement-objects>
+
+[HW]   [MPLANE] RPC reply = OK.
+[HW]   [MPLANE] Successfully edited the candidate datastore for RU "192.168.80.9".
+[HW]   [MPLANE] RPC request to RU "192.168.80.9" = <validate> candidate datastore.
+[HW]   [MPLANE] RPC reply = OK.
+[HW]   [MPLANE] Successfully validated candidate datastore for RU "192.168.80.9".
+[HW]   [MPLANE] RPC request to RU "192.168.80.9" = <commit> candidate datastore.
+[HW]   [MPLANE] RPC reply = OK.
+[HW]   [MPLANE] Successfully commited configuration into running datastore for RU "192.168.80.9".
+[HW]   [MPLANE] Sucessfully activated PM after start-up procedure for RU "192.168.80.9".
 ```
 </details>
 

doc/nfapi.md

@@ -171,7 +171,7 @@ After installing WLS, you can run the build command as shown below:
 
 #### How to run OAI PNF with OAI VNF
 
-Refer to the above steps in [Quickstart](.#quickstart), but run the PNF first as it is the WLS "master".
+Refer to the above steps in [Quickstart](#quickstart), but run the PNF first as it is the WLS "master".
 
 #### How to run OAI PNF with OSC/Radisys O-DU
 

docker/Dockerfile.build.ubuntu.cross-arm64

@@ -37,20 +37,11 @@ RUN /bin/sh oaienv && \
     mkdir -p log ran_build/build ran_build/build-cross && \
     cd ran_build/build && \
     cmake ../../.. -GNinja && \
-    echo "====== Start of log for host-tools ======" && \
-    ninja ldpc_generators generate_T > ../../log/host-tools.txt 2>&1 && \
-    echo "====== End of log for host-tools ======" && \
+    ninja ldpc_generators generate_T && \
     cd ../build-cross/ && \
     # install missing libyaml-cpp-dev for arm64
     apt install -y libyaml-cpp-dev:arm64 && \
     cmake ../../.. -GNinja -DENABLE_TELNETSRV=ON \
       -DCMAKE_TOOLCHAIN_FILE=../../../cmake_targets/cross-arm.cmake -DNATIVE_DIR=../build \
       -DCMAKE_C_FLAGS="-Werror" -DCMAKE_CXX_FLAGS="-Werror" && \
-    echo "====== Start of log for cross-build executables ======" && \
-    # making always passing so we can analyze the all.txt file
-    ninja -k10 dlsim ulsim ldpctest polartest smallblocktest nr_pbchsim nr_dlschsim nr_ulschsim nr_dlsim nr_ulsim nr_pucchsim nr_prachsim lte-softmodem nr-softmodem nr-cuup oairu lte-uesoftmodem nr-uesoftmodem params_libconfig coding rfsimulator telnetsrv > ../../log/all.txt 2>&1; \
-    # displaying errors and warnings
-    grep -E -A3 "warning:|error:" ../../log/all.txt || true && \
-    # will fail only if string "errors:" is found
-    grep -E "error:" -vqz ../../log/all.txt && \
-    echo "====== End of log for cross-build executables ======"
+    ninja -k10 dlsim ulsim ldpctest polartest smallblocktest nr_pbchsim nr_dlschsim nr_ulschsim nr_dlsim nr_ulsim nr_pucchsim nr_prachsim lte-softmodem nr-softmodem nr-cuup oairu lte-uesoftmodem nr-uesoftmodem params_libconfig coding rfsimulator telnetsrv

docker/Dockerfile.gNB.aerial.ubuntu

@@ -45,7 +45,7 @@ WORKDIR /oai-ran
 COPY . .
 
 RUN /bin/sh oaienv && \
-    tar -xvzf nvipc*  && \
+    tar -xvzf nvipc_src.*.tar.gz  && \
     cd nvipc_src.* && \
     rm -rf build && mkdir build && cd build && \
     cmake .. -DNVIPC_DPDK_ENABLE=OFF -DNVIPC_DOCA_ENABLE=OFF -DNVIPC_CUDA_ENABLE=OFF -DENABLE_SLT_RSP=ON && \

docker/Dockerfile.gNB.aerial.ubuntu.sanitize-address

@@ -45,7 +45,7 @@ WORKDIR /oai-ran
 COPY . .
 
 RUN /bin/sh oaienv && \
-    tar -xvzf nvipc* && \
+    tar -xvzf nvipc_src.*.tar.gz  && \
     cd nvipc_src.* && \
     rm -rf build && mkdir build && cd build && \
     cmake .. -DNVIPC_DPDK_ENABLE=OFF -DNVIPC_DOCA_ENABLE=OFF -DNVIPC_CUDA_ENABLE=OFF -DENABLE_SLT_RSP=ON && \

executables/nr-gnb.c

@@ -174,6 +174,8 @@ static void rx_func(processingData_L1_t *info)
   int slot_rx = info->slot_rx;
   nfapi_nr_config_request_scf_t *cfg = &gNB->gNB_config;
 
+  T(T_GNB_PHY_UL_TICK, T_INT(gNB->Mod_id), T_INT(frame_rx), T_INT(slot_rx));
+
   // RX processing
   int rx_slot_type = nr_slot_select(cfg, frame_rx, slot_rx);
   if (rx_slot_type == NR_UPLINK_SLOT || rx_slot_type == NR_MIXED_SLOT) {

executables/nr-ru.c

@@ -878,11 +878,12 @@ static void fill_rf_config(RU_t *ru, char *rf_config_file)
   }
 }
 
-static void fill_split7_2_config(split7_config_t *split7, const nfapi_nr_config_request_scf_t *config, int slots_per_frame, uint16_t ofdm_symbol_size)
+static void fill_split7_2_config(split7_config_t *split7, const nfapi_nr_config_request_scf_t *config, const NR_DL_FRAME_PARMS *fp)
 {
   const nfapi_nr_prach_config_t *prach_config = &config->prach_config;
   const nfapi_nr_tdd_table_t *tdd_table = &config->tdd_table;
   const nfapi_nr_cell_config_t *cell_config = &config->cell_config;
+  const nfapi_nr_carrier_config_t *carrier_config = &config->carrier_config;
 
   DevAssert(prach_config->prach_ConfigurationIndex.tl.tag == NFAPI_NR_CONFIG_PRACH_CONFIG_INDEX_TAG);
   split7->prach_index = prach_config->prach_ConfigurationIndex.value;
@@ -893,7 +894,7 @@ static void fill_split7_2_config(split7_config_t *split7, const nfapi_nr_config_
   if (cell_config->frame_duplex_type.value == 1 /* TDD */) {
     DevAssert(tdd_table->tdd_period.tl.tag == NFAPI_NR_CONFIG_TDD_PERIOD_TAG);
     int nb_periods_per_frame = get_nb_periods_per_frame(tdd_table->tdd_period.value);
-    split7->n_tdd_period = slots_per_frame / nb_periods_per_frame;
+    split7->n_tdd_period = fp->slots_per_frame / nb_periods_per_frame;
     for (int slot = 0; slot < split7->n_tdd_period; ++slot) {
       for (int sym = 0; sym < 14; ++sym) {
         split7->slot_dirs[slot].sym_dir[sym] = tdd_table->max_tdd_periodicity_list[slot].max_num_of_symbol_per_slot_list[sym].slot_config.value;
@@ -901,7 +902,15 @@ static void fill_split7_2_config(split7_config_t *split7, const nfapi_nr_config_
     }
   }
 
-  split7->fftSize = log2(ofdm_symbol_size);
+  split7->fftSize = log2(fp->ofdm_symbol_size);
+
+  // M-plane related parameters
+  for (size_t i = 0; i < 5 ; i++) {
+    split7->dl_k0[i] = carrier_config->dl_k0[i].value;
+    split7->ul_k0[i] = carrier_config->ul_k0[i].value;
+  }
+  split7->cp_prefix0 = fp->nb_prefix_samples0;
+  split7->cp_prefix_other = fp->nb_prefix_samples;
 }
 
 /* this function maps the RU tx and rx buffers to the available rf chains.
@@ -1044,7 +1053,7 @@ void *ru_thread(void *param)
   nr_dump_frame_parms(fp);
   nr_phy_init_RU(ru);
   fill_rf_config(ru, ru->rf_config_file);
-  fill_split7_2_config(&ru->openair0_cfg.split7, &ru->config, fp->slots_per_frame, fp->ofdm_symbol_size);
+  fill_split7_2_config(&ru->openair0_cfg.split7, &ru->config, fp);
 
   // Start IF device if any
   if (ru->nr_start_if) {

executables/nr-softmodem-common.h

@@ -96,6 +96,7 @@
 #define CONFIG_HLP_TELN          "Start embedded telnet server \n"
 #define CONFIG_HLP_SNR           "Set average SNR in dB (for --siml1 option)\n"
 #define CONFIG_HLP_NOS1          "Disable s1 interface\n"
+#define CONFIG_HLP_USEGPU        "Use GPU for processing offload"
 
 /*--------------------------------------------------------------------------------------------------------------------------------*/
 /*                                            command line parameters for LOG utility                                             */
@@ -126,7 +127,6 @@ extern int32_t uplink_frequency_offset[MAX_NUM_CCs][4];
 
 extern int oai_exit;
 
-extern openair0_config_t openair0_cfg[MAX_CARDS];
 extern pthread_cond_t sync_cond;
 extern pthread_mutex_t sync_mutex;
 extern int sync_var;

executables/nr-softmodem.c

@@ -103,6 +103,7 @@ unsigned int mmapped_dma=0;
 uint64_t downlink_frequency[MAX_NUM_CCs][4];
 int32_t uplink_frequency_offset[MAX_NUM_CCs][4];
 char *uecap_file;
+uint32_t use_gpu;
 
 runmode_t mode = normal_txrx;
 

executables/nr-softmodem.h

@@ -27,6 +27,7 @@
   {"U" ,                    CONFIG_HLP_ULBM_PHYTEST,   0,                .u64ptr=&ulsch_slot_bitmap,          .defintval=0,                     TYPE_UINT64, 0},        \
   {"usrp-tx-thread-config", CONFIG_HLP_USRP_THREAD,    0,                .iptr=&usrp_tx_thread,               .defstrval=0,                     TYPE_INT,    0},        \
   {"uecap_file",            CONFIG_HLP_UECAP_FILE,     0,                .strptr=&uecap_file,                 .defstrval="./uecap_ports1.xml",  TYPE_STRING, 0},        \
+  {"use_gpu",            CONFIG_HLP_USE_GPU,     0,                .uptr=&use_gpu,                 .defintval=0,  TYPE_UINT, 0},        \
 }
 // clang-format on
 
@@ -39,7 +40,7 @@ extern uint32_t target_ul_bw;
 extern uint64_t dlsch_slot_bitmap;
 extern uint64_t ulsch_slot_bitmap;
 extern char *uecap_file;
-
+extern uint32_t use_gpu;
 // In nr-gnb.c
 extern void init_gNB();
 extern void stop_gNB(int);

executables/nr-ue.c

@@ -414,7 +414,7 @@ static void UE_synch(void *arg) {
   syncData_t *syncD = (syncData_t *)arg;
   PHY_VARS_NR_UE *UE = syncD->UE;
   UE->is_synchronized = 0;
-  openair0_config_t *cfg0 = &openair0_cfg[UE->rf_map.card];
+  openair0_config_t *cfg0 = &UE->openair0_cfg[UE->rf_map.card];
 
   if (UE->target_Nid_cell != -1) {
     LOG_W(NR_PHY, "Starting re-sync detection for target Nid_cell %i\n", UE->target_Nid_cell);
@@ -523,7 +523,7 @@ static void RU_write(nr_rxtx_thread_data_t *rxtxD, bool sl_tx_action, c16_t **tx
   radio_tx_burst_flag_t flags = TX_BURST_INVALID;
 
   if (UE->received_config_request) {
-    if (openair0_cfg[0].duplex_mode == duplex_mode_FDD || get_softmodem_params()->continuous_tx) {
+    if (UE->openair0_cfg[0].duplex_mode == duplex_mode_FDD || get_softmodem_params()->continuous_tx) {
       flags = TX_BURST_MIDDLE;
     // In case of Sidelink, USRP write needed only in case transmission
     // needs to be done in this slot and not based on tdd ULDL configuration.
@@ -696,8 +696,8 @@ static int handle_sync_req_from_mac(PHY_VARS_NR_UE *UE)
     if (dl_CarrierFreq != fp->dl_CarrierFreq || ul_CarrierFreq != fp->ul_CarrierFreq) {
       fp->dl_CarrierFreq = dl_CarrierFreq;
       fp->ul_CarrierFreq = ul_CarrierFreq;
-      nr_rf_card_config_freq(&openair0_cfg[UE->rf_map.card], ul_CarrierFreq, dl_CarrierFreq, 0);
-      UE->rfdevice.trx_set_freq_func(&UE->rfdevice, &openair0_cfg[0]);
+      nr_rf_card_config_freq(&UE->openair0_cfg[UE->rf_map.card], ul_CarrierFreq, dl_CarrierFreq, 0);
+      UE->rfdevice.trx_set_freq_func(&UE->rfdevice, &UE->openair0_cfg[UE->rf_map.card]);
       init_symbol_rotation(fp);
     }
 
@@ -731,12 +731,14 @@ static int UE_dl_preprocessing(PHY_VARS_NR_UE *UE,
     fp = &UE->SL_UE_PHY_PARAMS.sl_frame_params;
 
   // process what RRC thread sent to MAC
-  MessageDef *msg = NULL;
   do {
-    itti_poll_msg(TASK_MAC_UE, &msg);
-    if (msg)
-      process_msg_rcc_to_mac(msg);
-  } while (msg);
+    notifiedFIFO_elt_t *elt = pollNotifiedFIFO(&get_mac_inst(UE->Mod_id)->input_nf);
+    if (!elt) {
+      break;
+    }
+    process_msg_rcc_to_mac(NotifiedFifoData(elt), UE->Mod_id);
+    delNotifiedFIFO_elt(elt);
+  } while (true);
 
   if (UE->if_inst)
     UE->if_inst->slot_indication(UE->Mod_id, false);
@@ -966,7 +968,7 @@ void *UE_thread(void *arg)
   void *rxp[NB_ANTENNAS_RX];
   enum stream_status_e stream_status = STREAM_STATUS_UNSYNC;
   fapi_nr_config_request_t *cfg = &UE->nrUE_config;
-  int tmp = openair0_device_load(&(UE->rfdevice), &openair0_cfg[0]);
+  int tmp = openair0_device_load(&(UE->rfdevice), &UE->openair0_cfg[0]);
   AssertFatal(tmp == 0, "Could not load the device\n");
   NR_DL_FRAME_PARMS *fp = &UE->frame_parms;
   sl_nr_phy_config_request_t *sl_cfg = NULL;
@@ -1032,12 +1034,10 @@ void *UE_thread(void *arg)
             decoded_frame_rx = UE->SL_UE_PHY_PARAMS.sync_params.DFN;
           else {
             // We must wait the RRC layer decoded the MIB and sent us the frame number
-            MessageDef *msg = NULL;
-            itti_receive_msg(TASK_MAC_UE, &msg);
-            if (msg)
-              process_msg_rcc_to_mac(msg);
-            else
-              LOG_E(PHY, "It seems we arbort while trying to sync\n");
+            notifiedFIFO_elt_t *elt = pullNotifiedFIFO(&mac->input_nf);
+            AssertFatal(elt != NULL, "fifo error while waiting for MIB");
+            process_msg_rcc_to_mac(NotifiedFifoData(elt), UE->Mod_id);
+            delNotifiedFIFO_elt(elt);
             decoded_frame_rx = mac->mib_frame;
           }
           LOG_A(PHY,
@@ -1284,20 +1284,19 @@ void *UE_thread(void *arg)
 
 void init_NR_UE(int nb_inst, char *uecap_file, char *reconfig_file, char *rbconfig_file)
 {
-  NR_UE_RRC_INST_t *rrc_inst = nr_rrc_init_ue(uecap_file, nb_inst, get_nrUE_params()->nb_antennas_tx);
-  NR_UE_MAC_INST_t *mac_inst = nr_l2_init_ue(nb_inst);
-  AssertFatal(mac_inst, "Couldn't allocate MAC module\n");
+  for (int instance_id = 0; instance_id < nb_inst; instance_id++) {
+    NR_UE_RRC_INST_t* rrc = nr_rrc_init_ue(uecap_file, instance_id, get_nrUE_params()->nb_antennas_tx);
+    NR_UE_MAC_INST_t *mac = nr_l2_init_ue(instance_id);
 
-  for (int i = 0; i < nb_inst; i++) {
-    NR_UE_MAC_INST_t *mac = get_mac_inst(i);
-    mac->if_module = nr_ue_if_module_init(i);
+    nr_rrc_set_mac_queue(instance_id, &mac->input_nf);
+    mac->if_module = nr_ue_if_module_init(instance_id);
     AssertFatal(mac->if_module, "can not initialize IF module\n");
     if (!IS_SA_MODE(get_softmodem_params()) && !get_softmodem_params()->sl_mode) {
-      init_nsa_message(&rrc_inst[i], reconfig_file, rbconfig_file);
-      nr_rlc_activate_srb0(mac_inst[i].crnti, NULL, send_srb0_rrc);
+      init_nsa_message(rrc, reconfig_file, rbconfig_file);
+      nr_rlc_activate_srb0(mac->crnti, NULL, send_srb0_rrc);
     }
     //TODO: Move this call to RRC
-    start_sidelink((&rrc_inst[i])->ue_id);
+    start_sidelink(instance_id);
   }
 }
 

executables/nr-uesoftmodem.c

@@ -121,8 +121,6 @@ int32_t         uplink_frequency_offset[MAX_NUM_CCs][4];
 uint64_t        sidelink_frequency[MAX_NUM_CCs][4];
 
 // UE and OAI config variables
-
-openair0_config_t openair0_cfg[MAX_CARDS];
 int16_t           node_synch_ref[MAX_NUM_CCs];
 int               otg_enabled;
 double            cpuf;
@@ -249,41 +247,42 @@ void set_options(int CC_id, PHY_VARS_NR_UE *UE){
 
 }
 
-static void init_openair0()
+void init_openair0(PHY_VARS_NR_UE *ue)
 {
   int card;
   int freq_off = 0;
-  NR_DL_FRAME_PARMS *frame_parms = &PHY_vars_UE_g[0][0]->frame_parms;
+  NR_DL_FRAME_PARMS *frame_parms = &ue->frame_parms;
   bool is_sidelink = (get_softmodem_params()->sl_mode) ? true : false;
   if (is_sidelink)
-    frame_parms = &PHY_vars_UE_g[0][0]->SL_UE_PHY_PARAMS.sl_frame_params;
+    frame_parms = &ue->SL_UE_PHY_PARAMS.sl_frame_params;
 
   for (card=0; card<MAX_CARDS; card++) {
+    openair0_config_t *cfg = &ue->openair0_cfg[card];
     uint64_t dl_carrier, ul_carrier;
-    openair0_cfg[card].configFilename    = NULL;
-    openair0_cfg[card].sample_rate       = frame_parms->samples_per_subframe * 1e3;
-    openair0_cfg[card].samples_per_frame = frame_parms->samples_per_frame;
+    cfg->configFilename    = NULL;
+    cfg->sample_rate       = frame_parms->samples_per_subframe * 1e3;
+    cfg->samples_per_frame = frame_parms->samples_per_frame;
 
     if (frame_parms->frame_type==TDD)
-      openair0_cfg[card].duplex_mode = duplex_mode_TDD;
+      cfg->duplex_mode = duplex_mode_TDD;
     else
-      openair0_cfg[card].duplex_mode = duplex_mode_FDD;
+      cfg->duplex_mode = duplex_mode_FDD;
 
-    openair0_cfg[card].Mod_id = 0;
-    openair0_cfg[card].num_rb_dl = frame_parms->N_RB_DL;
-    openair0_cfg[card].clock_source = get_softmodem_params()->clock_source;
-    openair0_cfg[card].time_source = get_softmodem_params()->timing_source;
-    openair0_cfg[card].tune_offset = get_softmodem_params()->tune_offset;
-    openair0_cfg[card].tx_num_channels = min(4, frame_parms->nb_antennas_tx);
-    openair0_cfg[card].rx_num_channels = min(4, frame_parms->nb_antennas_rx);
+    cfg->Mod_id = 0;
+    cfg->num_rb_dl = frame_parms->N_RB_DL;
+    cfg->clock_source = get_softmodem_params()->clock_source;
+    cfg->time_source = get_softmodem_params()->timing_source;
+    cfg->tune_offset = get_softmodem_params()->tune_offset;
+    cfg->tx_num_channels = min(4, frame_parms->nb_antennas_tx);
+    cfg->rx_num_channels = min(4, frame_parms->nb_antennas_rx);
 
     LOG_I(PHY,
           "HW: Configuring card %d, sample_rate %f, tx/rx num_channels %d/%d, duplex_mode %s\n",
           card,
-          openair0_cfg[card].sample_rate,
-          openair0_cfg[card].tx_num_channels,
-          openair0_cfg[card].rx_num_channels,
-          duplex_mode_txt[openair0_cfg[card].duplex_mode]);
+          cfg->sample_rate,
+          cfg->tx_num_channels,
+          cfg->rx_num_channels,
+          duplex_mode_txt[cfg->duplex_mode]);
 
     if (is_sidelink) {
       dl_carrier = frame_parms->dl_CarrierFreq;
@@ -291,16 +290,18 @@ static void init_openair0()
     } else
       nr_get_carrier_frequencies(PHY_vars_UE_g[0][0], &dl_carrier, &ul_carrier);
 
-    nr_rf_card_config_freq(&openair0_cfg[card], ul_carrier, dl_carrier, freq_off);
+    nr_rf_card_config_freq(cfg, ul_carrier, dl_carrier, freq_off);
 
-    nr_rf_card_config_gain(&openair0_cfg[card], rx_gain_off);
+    nr_rf_card_config_gain(cfg, rx_gain_off);
 
-    openair0_cfg[card].configFilename = get_softmodem_params()->rf_config_file;
-
-    if (get_nrUE_params()->usrp_args) openair0_cfg[card].sdr_addrs = get_nrUE_params()->usrp_args;
-    if (get_nrUE_params()->tx_subdev) openair0_cfg[card].tx_subdev = get_nrUE_params()->tx_subdev;
-    if (get_nrUE_params()->rx_subdev) openair0_cfg[card].rx_subdev = get_nrUE_params()->rx_subdev;
+    cfg->configFilename = get_softmodem_params()->rf_config_file;
 
+    if (get_nrUE_params()->usrp_args)
+      cfg->sdr_addrs = get_nrUE_params()->usrp_args;
+    if (get_nrUE_params()->tx_subdev)
+      cfg->tx_subdev = get_nrUE_params()->tx_subdev;
+    if (get_nrUE_params()->rx_subdev)
+      cfg->rx_subdev = get_nrUE_params()->rx_subdev;
   }
 }
 
@@ -326,9 +327,11 @@ static void trigger_stop(int sig)
 static void trigger_deregistration(int sig)
 {
   if (!stop_immediately && IS_SA_MODE(get_softmodem_params())) {
-    MessageDef *msg = itti_alloc_new_message(TASK_NAS_NRUE, 0, NAS_DEREGISTRATION_REQ);
-    NAS_DEREGISTRATION_REQ(msg).cause = AS_DETACH;
-    itti_send_msg_to_task(TASK_NAS_NRUE, 0, msg);
+    for (int ue_inst = 0; ue_inst < NB_UE_INST; ue_inst++) {
+      MessageDef *msg = itti_alloc_new_message(TASK_NAS_NRUE, ue_inst, NAS_DEREGISTRATION_REQ);
+      NAS_DEREGISTRATION_REQ(msg).cause = AS_DETACH;
+      itti_send_msg_to_task(TASK_NAS_NRUE, ue_inst, msg);
+    }
     stop_immediately = true;
     static const char m[] = "Press ^C again to trigger immediate shutdown\n";
     __attribute__((unused)) int unused = write(STDOUT_FILENO, m, sizeof(m) - 1);
@@ -386,7 +389,6 @@ int main(int argc, char **argv)
   }
   //set_softmodem_sighandler();
   CONFIG_SETRTFLAG(CONFIG_NOEXITONHELP);
-  memset(openair0_cfg,0,sizeof(openair0_config_t)*MAX_CARDS);
   memset(tx_max_power,0,sizeof(int)*MAX_NUM_CCs);
   // initialize logging
   logInit();
@@ -425,12 +427,11 @@ int main(int argc, char **argv)
   char *pckg = strdup(OAI_PACKAGE_VERSION);
   LOG_I(HW, "Version: %s\n", pckg);
 
-  PHY_vars_UE_g = malloc(sizeof(*PHY_vars_UE_g) * NB_UE_INST);
+  PHY_vars_UE_g = calloc_or_fail(NB_UE_INST, sizeof(*PHY_vars_UE_g));
   for (int inst = 0; inst < NB_UE_INST; inst++) {
-    PHY_vars_UE_g[inst] = malloc(sizeof(*PHY_vars_UE_g[inst]) * MAX_NUM_CCs);
+    PHY_vars_UE_g[inst] = calloc_or_fail(MAX_NUM_CCs, sizeof(*PHY_vars_UE_g[inst]));
     for (int CC_id = 0; CC_id < MAX_NUM_CCs; CC_id++) {
-      PHY_vars_UE_g[inst][CC_id] = malloc(sizeof(*PHY_vars_UE_g[inst][CC_id]));
-      memset(PHY_vars_UE_g[inst][CC_id], 0, sizeof(*PHY_vars_UE_g[inst][CC_id]));
+      PHY_vars_UE_g[inst][CC_id] = calloc_or_fail(1, sizeof(*PHY_vars_UE_g[inst][CC_id]));
       // All instances use the same coding interface
       PHY_vars_UE_g[inst][CC_id]->nrLDPC_coding_interface = nrLDPC_coding_interface;
     }
@@ -498,12 +499,14 @@ int main(int argc, char **argv)
                                     get_softmodem_params()->numerology,
                                     nr_band);
         } else {
-	  MessageDef *msg = NULL;
-	  do {
-	    itti_poll_msg(TASK_MAC_UE, &msg);
-	    if (msg)
-	      process_msg_rcc_to_mac(msg);
-	  } while (msg);
+          do {
+            notifiedFIFO_elt_t *elt = pollNotifiedFIFO(&mac->input_nf);
+            if (!elt) {
+              break;
+            }
+            process_msg_rcc_to_mac(NotifiedFifoData(elt), inst);
+            delNotifiedFIFO_elt(elt);
+          } while (true);
           fapi_nr_config_request_t *nrUE_config = &UE[CC_id]->nrUE_config;
           nr_init_frame_parms_ue(&UE[CC_id]->frame_parms, nrUE_config, mac->nr_band);
         }
@@ -538,10 +541,10 @@ int main(int argc, char **argv)
                                     get_nrUE_params()->ofdm_offset_divisor);
           sl_ue_phy_init(UE[CC_id]);
         }
+        init_openair0(UE[CC_id]);
       }
     }
 
-    init_openair0();
     lock_memory_to_ram();
 
     if (IS_SOFTMODEM_DOSCOPE) {

executables/nr-uesoftmodem.h

@@ -5,6 +5,7 @@
 #include "common/utils/threadPool/thread-pool.h"
 #include "PHY/defs_nr_UE.h"
 
+#define MAX_NUM_NR_UE_INST 32
 extern int NB_UE_INST;
 extern uint16_t ue_id_g;
 

nfapi/open-nFAPI/fapi/src/nr_fapi_p7.c

@@ -2308,7 +2308,7 @@ uint8_t unpack_nr_srs_report_tlv_value(nfapi_srs_report_tlv_t *report_tlv, uint8
     }
   }
 #else
-  const uint16_t last_idx = ((report_tlv->length + 3) / 4) - 1;
+  const int16_t last_idx = ((report_tlv->length + 3) / 4) - 1;
   for (int i = 0; i < last_idx; i++) {
     if (!pull32(ppReadPackedMsg, &report_tlv->value[i], end)) {
       return 0;

openair1/PHY/CODING/CMakeLists.txt

@@ -13,6 +13,7 @@ target_link_libraries(coding PRIVATE log_headers)
 add_library(ldpc_orig MODULE
             nrLDPC_decoder/nrLDPC_decoder.c
             nrLDPC_encoder/ldpc_encoder.c
+            nrLDPC_encoder/ldpc_encoder32.c
 )
 set_target_properties(ldpc_orig PROPERTIES LIBRARY_OUTPUT_DIRECTORY ${CMAKE_BINARY_DIR})
 target_link_libraries(ldpc_orig PRIVATE ldpc_segment ldpc_gen_HEADERS)
@@ -20,6 +21,7 @@ target_link_libraries(ldpc_orig PRIVATE ldpc_segment ldpc_gen_HEADERS)
 add_library(ldpc MODULE
             nrLDPC_decoder/nrLDPC_decoder.c
             nrLDPC_encoder/ldpc_encoder_optim8segmulti.c
+            nrLDPC_encoder/ldpc_encoder32.c
 )
 set_target_properties(ldpc PROPERTIES LIBRARY_OUTPUT_DIRECTORY ${CMAKE_BINARY_DIR})
 target_link_libraries(ldpc PRIVATE ldpc_segment ldpc_gen_HEADERS)
@@ -42,22 +44,28 @@ if (ENABLE_LDPC_CUDA)
   if (NOT CUDA_FOUND)
     message(FATAL_ERROR "no CUDA found")
   endif()
-  SET(CUDA_NVCC_FLAG "${CUDA_NVCC_FLAGS};-arch=sm_60;")
+  SET(CUDA_NVCC_FLAGS "${CUDA_NVCC_FLAGS};-arch=sm_87;")
   SET(CUDA_VERBOSE_BUILD ON)
   cuda_add_library(ldpc_cuda MODULE
-                   nrLDPC_decoder_LYC/nrLDPC_decoder_LYC.cu
-                   nrLDPC_encoder/ldpc_encoder_optim8segmulti.c
+	  #                   nrLDPC_decoder_LYC/nrLDPC_decoder_LYC.cu
+                   nrLDPC_decoder/nrLDPC_decoder.c
+		   nrLDPC_encoder/ldpc_encoder_optim8segmulti.c
+		   nrLDPC_encoder/ldpc_encoder_cuda32.c
+		   nrLDPC_encoder/ldpc_BG1_Zc384_32bit.cu
+		   nrLDPC_encoder/ldpc_input.cu
                    # The slot coding layer cannot be linked with
                    # target_link_libraries like above
                    # because of cuda_add_library
                    # which already uses target_link_libraries
                    nrLDPC_coding/nrLDPC_coding_segment/nrLDPC_coding_segment_decoder.c
+                   nrLDPC_coding/nrLDPC_coding_segment/nrLDPC_coding_segment_encoder_cuda.c
                    nrLDPC_coding/nrLDPC_coding_segment/nrLDPC_coding_segment_encoder.c
                    nrLDPC_coding/nrLDPC_coding_segment/nr_rate_matching.c
   )
+  target_link_libraries(ldpc_cuda ldpc_gen_HEADERS ${T_LIB})
   set_target_properties(ldpc_cuda PROPERTIES CUDA_SEPARABLE_COMPILATION ON)
   set_target_properties(ldpc_cuda PROPERTIES LIBRARY_OUTPUT_DIRECTORY ${CMAKE_BINARY_DIR})
-
+  target_compile_definitions(ldpc_cuda PRIVATE ENABLE_CUDA)
   add_dependencies(ldpctest ldpc_cuda)
   add_dependencies(nr-softmodem ldpc_cuda)
   add_dependencies(nr-uesoftmodem ldpc_cuda)

openair1/PHY/CODING/TESTBENCH/ldpctest.c

@@ -34,6 +34,9 @@
 #include "openair1/PHY/defs_nr_common.h"
 #include "coding_unitary_defs.h"
 #include "common/utils/LOG/log.h"
+#ifdef ENABLE_CUDA
+#include <cuda_runtime.h>
+#endif
 
 #define MAX_BLOCK_LENGTH 8448
 
@@ -104,14 +107,16 @@ one_measurement_t test_ldpc(short max_iterations,
                             unsigned char qbits,
                             short Kprime,
                             unsigned int ntrials,
-                            int n_segments)
+                            int n_segments,
+	                    int gen_code,
+			    int use32bit)
 {
   one_measurement_t ret = {0};
   reset_meas(&ret.time_optim);
   reset_meas(&ret.time_decoder);
   // clock initiate
   // time_stats_t time,time_optim,tinput,tprep,tparity,toutput, time_decoder;
-  time_stats_t time, tinput, tprep, tparity, toutput;
+  time_stats_t time, tinput, tinput_memcpy, tprep, tparity, toutput;
   double n_iter_mean = 0;
   double n_iter_std = 0;
   int n_iter_max = 0;
@@ -119,14 +124,22 @@ one_measurement_t test_ldpc(short max_iterations,
   double sigma;
   sigma = 1.0 / sqrt(2 * SNR);
   cpu_meas_enabled = 1;
-  uint8_t *test_input[MAX_NUM_NR_DLSCH_SEGMENTS_PER_LAYER * NR_MAX_NB_LAYERS];
-  uint8_t estimated_output[MAX_NUM_DLSCH_SEGMENTS][Kprime];
+#ifdef ENABLE_CUDA
+  uint8_t **test_input,*test_input_p;
+#else
+  uint8_t *test_input[n_segments * NR_MAX_NB_LAYERS];
+#endif
+  uint8_t *channel_input[n_segments];
+  uint8_t estimated_output[n_segments][Kprime];
   memset(estimated_output, 0, sizeof(estimated_output));
-  uint8_t *channel_input[MAX_NUM_DLSCH_SEGMENTS];
   uint8_t *channel_input_optim;
+
   // double channel_output[68 * 384];
-  double modulated_input[MAX_NUM_DLSCH_SEGMENTS][68 * 384] = {0};
-  int8_t channel_output_fixed[MAX_NUM_DLSCH_SEGMENTS][68 * 384] = {0};
+  double modulated_input[n_segments][68 * 384];
+  memset(modulated_input,0,sizeof(modulated_input));
+  int8_t channel_output_fixed[n_segments][68 * 384];
+  memset(channel_output_fixed,0,sizeof(channel_output_fixed));
+
   short BG = 0, nrows = 0; //,ncols;
   int i1, Kb = 0;
   int R_ind = 0;
@@ -134,7 +147,7 @@ one_measurement_t test_ldpc(short max_iterations,
   int code_rate_vec[8] = {15, 13, 25, 12, 23, 34, 56, 89};
   // 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[MAX_NUM_DLSCH_SEGMENTS] = {0};
+  t_nrLDPC_dec_params decParams[n_segments];
 
   t_nrLDPC_time_stats decoder_profiler = {0};
 
@@ -142,6 +155,7 @@ one_measurement_t test_ldpc(short max_iterations,
 
   reset_meas(&time);
   reset_meas(&tinput);
+  reset_meas(&tinput_memcpy);
   reset_meas(&tprep);
   reset_meas(&tparity);
   reset_meas(&toutput);
@@ -259,18 +273,29 @@ one_measurement_t test_ldpc(short max_iterations,
   printf("To: %d\n", (Kb + nrows - no_punctured_columns) * Zc - removed_bit);
   printf("number of undecoded bits: %d\n", (Kb + nrows - no_punctured_columns - 2) * Zc - removed_bit);
 
+//  ldpc_toCompare.LDPCinit();
   // generate input block
-  for (int j = 0; j < MAX_NUM_DLSCH_SEGMENTS; j++) {
+#ifdef ENABLE_CUDA
+  cudaHostAlloc((void**)&test_input_p,n_segments*sizeof(uint8_t*),cudaHostAllocMapped);
+  test_input=(uint8_t **)test_input_p;
+  printf("test input %p\n",test_input);
+#endif
+  for (int j = 0; j < n_segments; j++) {
+#ifdef ENABLE_CUDA
+    cudaHostAlloc((void**)&test_input[j],((K + 7) & ~7) / 8,cudaHostAllocMapped);
+    printf("test input[%d] %p\n",j,test_input[j]);
+#else
     test_input[j] = malloc16(((K + 7) & ~7) / 8);
     memset(test_input[j], 0, ((K + 7) & ~7) / 8);
+#endif
     channel_input[j] = malloc16(68 * 384);
     memset(channel_input[j], 0, 68 * 384);
   }
-  channel_input_optim = malloc16(68 * 384);
-  memset(channel_input_optim, 0, 68 * 384);
+  channel_input_optim = malloc16(68 * 384 * sizeof(uint32_t));
+  if (use32bit ==0) memset(channel_input_optim, 0, 68 * 384 * sizeof(uint32_t));
 
   // Fill input segments with random values
-  for (int j = 0; j < MAX_NUM_DLSCH_SEGMENTS; j++) {
+  for (int j = 0; j < n_segments; j++) {
     int i = 0;
     for (i = 0; i < ((Kprime + 7) & ~7) / 8; i++)
       test_input[j][i] = (uint8_t)rand();
@@ -282,13 +307,17 @@ one_measurement_t test_ldpc(short max_iterations,
   }
 
   encoder_implemparams_t impp = {.Zc = Zc, .Kb = Kb, .BG = BG, .K = K};
-  impp.gen_code = 1;
-
+  impp.gen_code = gen_code;
+  impp.tparity = &tparity;
+  impp.tinput = &tinput;
+  impp.tinput_memcpy = &tinput_memcpy;
+  impp.toutput = &toutput;
   if (ntrials == 0)
     ldpc_orig.LDPCencoder(test_input, channel_input[0], &impp);
   impp.gen_code = 0;
   decode_abort_t dec_abort;
   init_abort(&dec_abort);
+  uint32_t **output32;
   for (int trial = 0; trial < ntrials; trial++) {
     unsigned int segment_bler = 0;
     //// encoder
@@ -301,26 +330,27 @@ one_measurement_t test_ldpc(short max_iterations,
     impp.n_segments = n_segments;
     start_meas(&ret.time_optim);
     impp.first_seg = 0;
-    ldpc_toCompare.LDPCencoder(test_input, channel_input_optim, &impp);
+    if (use32bit==0) ldpc_toCompare.LDPCencoder(test_input, channel_input_optim, &impp);
+    else output32=ldpc_toCompare.LDPCencoder32(test_input, &impp);
     stop_meas(&ret.time_optim);
 
     if (ntrials == 1)
       for (int j = 0; j < n_segments; j++)
         for (int i = 0; i < K + (nrows - no_punctured_columns) * Zc - removed_bit; i++) {
-          if (channel_input[j][i] != ((channel_input_optim[i] >> j) & 0x1)) {
-            printf("differ in seg %d pos %d (%u,%u)\n", j, i, channel_input[j][i], (channel_input_optim[i] >> j) & 0x1);
-            return ret;
-          }
+          if (((use32bit == 0) && (channel_input[j][i] != ((channel_input_optim[i] >> j) & 0x1))) ||                 ((use32bit == 1) && (channel_input[j][i] != (((output32[0][i] >> j) & 0x1))))) {
+               printf("differ in seg %d pos %d (%u,%u)\n", j, i, channel_input[j][i], (((uint32_t*)channel_input_optim)[i] >> j) & 0x1);
+               return ret;
+            }
         }
-
+    int bit;
     for (int j = 0; j < n_segments; j++) {
       for (int i = 2 * Zc; i < (Kb + nrows - no_punctured_columns) * Zc - removed_bit; i++) {
 #ifdef DEBUG_CODER
         if ((i & 0xf) == 0)
           printf("\ne %d..%d:    ", i, i + 15);
 #endif
-
-        if (((channel_input_optim[i - 2 * Zc] >> j) & 0x1) == 0)
+        bit = (use32bit==0) ? ((channel_input_optim[i - 2 * Zc] >> j) & 0x1) : ((output32[0][i - 2 * Zc] >> j) & 0x1);
+        if (bit == 0)
           modulated_input[j][i] = 1.0; /// sqrt(2);  //QPSK
         else
           modulated_input[j][i] = -1.0; /// sqrt(2);
@@ -345,7 +375,6 @@ one_measurement_t test_ldpc(short max_iterations,
       decParams[j].numMaxIter = max_iterations;
       decParams[j].outMode = nrLDPC_outMode_BIT;
       decParams[j].Kprime = Kprime;
-      ldpc_toCompare.LDPCinit();
     }
     for (int j = 0; j < n_segments; j++) {
       start_meas(&ret.time_decoder);
@@ -388,8 +417,12 @@ one_measurement_t test_ldpc(short max_iterations,
 
   ret.errors_bit_uncoded = ret.errors_bit_uncoded / (double)((Kb + nrows - no_punctured_columns - 2) * Zc - removed_bit);
 
-  for (int j = 0; j < MAX_NUM_DLSCH_SEGMENTS; j++) {
+  for (int j = 0; j < n_segments; j++) {
+#ifdef ENABLE_CUDA
+    cudaFree(test_input[j]);	  
+#else
     free(test_input[j]);
+#endif
     free(channel_input[j]);
   }
   free(channel_input_optim);
@@ -397,6 +430,7 @@ one_measurement_t test_ldpc(short max_iterations,
   print_meas(&time, "ldpc_encoder", NULL, NULL);
   print_meas(&ret.time_optim, "ldpc_encoder_optim", NULL, NULL);
   print_meas(&tinput, "ldpc_encoder_optim(input)", NULL, NULL);
+  print_meas(&tinput_memcpy, "ldpc_encoder_optim(input memcpy)", NULL, NULL);
   print_meas(&tprep, "ldpc_encoder_optim(prep)", NULL, NULL);
   print_meas(&tparity, "ldpc_encoder_optim(parity)", NULL, NULL);
   print_meas(&toutput, "ldpc_encoder_optim(output)", NULL, NULL);
@@ -437,19 +471,20 @@ int main(int argc, char *argv[])
 
   int n_trials = 1;
   double SNR_step = 0.1;
-
+  int gen_code = 1;
   randominit(0);
   int test_uncoded = 0;
   n_iter_stats_t dec_iter[400] = {0};
 
   short BG = 0, Zc;
+  int use32bit=0;
 
   if ((uniqCfg = load_configmodule(argc, argv, CONFIG_ENABLECMDLINEONLY)) == 0) {
     exit_fun("[LDPCTEST] Error, configuration module init failed\n");
   }
   logInit();
 
-  while ((c = getopt(argc, argv, "--:O:q:r:s:S:l:G:n:d:i:t:u:hv:")) != -1) {
+  while ((c = getopt(argc, argv, "--:O:q:r:s:S:l:Gn:d:i:t:u:hv:g:")) != -1) {
     /* ignore long options starting with '--', option '-O' and their arguments that are handled by configmodule */
     /* with this opstring getopt returns 1 for non-option arguments, refer to 'man 3 getopt' */
     if (c == 1 || c == '-' || c == 'O')
@@ -475,6 +510,7 @@ int main(int argc, char *argv[])
 
       case 'G':
         ldpc_version = "_cuda";
+	use32bit=1;
         break;
 
       case 'n':
@@ -503,6 +539,11 @@ int main(int argc, char *argv[])
       case 'v':
         ldpc_version = strdup(optarg);
         break;
+      case 'g':
+	gen_code = atoi(optarg);
+	AssertFatal(gen_code <= 4, "gen_code %d is not allowed\n",gen_code);
+        n_trials=0;	
+	break;
       case 'h':
       default:
         printf("CURRENTLY SUPPORTED CODE RATES: \n");
@@ -559,7 +600,9 @@ int main(int argc, char *argv[])
                                       qbits,
                                       Kprime, // block length bytes
                                       n_trials,
-                                      n_segments);
+                                      n_segments,
+				      gen_code,
+				      use32bit);
 
     decoded_errors[i] = res.errors;
     dec_iter[i] = res.dec_iter;

openair1/PHY/CODING/nrLDPC_coding/nrLDPC_coding_interface.h

@@ -152,6 +152,7 @@ typedef struct nrLDPC_segment_encoding_parameters_s{
   time_stats_t ts_interleave;
   time_stats_t ts_rate_match;
   time_stats_t ts_ldpc_encode;
+  time_stats_t ts_output;
 } nrLDPC_segment_encoding_parameters_t;
 
 /**
@@ -201,6 +202,7 @@ typedef struct nrLDPC_TB_encoding_parameters_s{
   uint32_t C;
   nrLDPC_segment_encoding_parameters_t *segments;
   unsigned char *output;
+  uint8_t **c_dev;
 } nrLDPC_TB_encoding_parameters_t;
 
 /**
@@ -226,10 +228,13 @@ typedef struct nrLDPC_slot_encoding_parameters_s{
   int nb_TBs;
   tpool_t *threadPool;
   time_stats_t *tinput;
+  time_stats_t *tinput_memcpy;
   time_stats_t *tprep;
   time_stats_t *tparity;
   time_stats_t *toutput;
+  time_stats_t *tconcat;
   nrLDPC_TB_encoding_parameters_t *TBs;
+  uint32_t use_gpu;
 } nrLDPC_slot_encoding_parameters_t;
 
 typedef int32_t(nrLDPC_coding_init_t)(void);

openair1/PHY/CODING/nrLDPC_coding/nrLDPC_coding_segment/nrLDPC_coding_segment_decoder.c

@@ -281,6 +281,11 @@ int nrLDPC_prepare_TB_decoding(nrLDPC_slot_decoding_parameters_t *nrLDPC_slot_de
 
 int32_t nrLDPC_coding_init(void)
 {
+  LOG_I(NR_PHY,"Initializing coding library\n");
+#ifdef ENABLE_CUDA
+  LOG_I(NR_PHY,"Calling cuda_support_init()\n");
+#endif
+  cuda_support_init();	
   return 0;
 }
 

openair1/PHY/CODING/nrLDPC_coding/nrLDPC_coding_segment/nrLDPC_coding_segment_encoder.c

@@ -367,9 +367,11 @@ static int nrLDPC_launch_TB_encoding(nrLDPC_slot_encoding_parameters_t *nrLDPC_s
   encoder_implemparams_t common_segment_params = {
     .n_segments = nrLDPC_TB_encoding_parameters->C,
     .tinput = nrLDPC_slot_encoding_parameters->tinput,
+    .tinput_memcpy = nrLDPC_slot_encoding_parameters->tinput_memcpy,
     .tprep = nrLDPC_slot_encoding_parameters->tprep,
     .tparity = nrLDPC_slot_encoding_parameters->tparity,
     .toutput = nrLDPC_slot_encoding_parameters->toutput,
+    .tconcat = nrLDPC_slot_encoding_parameters->tconcat,
     .Kb = nrLDPC_TB_encoding_parameters->Kb,
     .Zc = nrLDPC_TB_encoding_parameters->Z,
     .BG = nrLDPC_TB_encoding_parameters->BG,
@@ -406,16 +408,24 @@ int nrLDPC_coding_encoder(nrLDPC_slot_encoding_parameters_t *nrLDPC_slot_encodin
   for (int dlsch_id = 0; dlsch_id < nrLDPC_slot_encoding_parameters->nb_TBs; dlsch_id++) {
     // Compute number of tasks to encode TB
     nrLDPC_TB_encoding_parameters_t *nrLDPC_TB_encoding_parameters = &nrLDPC_slot_encoding_parameters->TBs[dlsch_id];
-    size_t n_seg = (nrLDPC_TB_encoding_parameters->C / 8 + ((nrLDPC_TB_encoding_parameters->C & 7) == 0 ? 0 : 1));
-    nbTasks += n_seg;
-
-    // Search for maximum E for sizing encoder output f and f2
-    for (int seg_id = 0; seg_id < nrLDPC_TB_encoding_parameters->C; seg_id++) {
-      uint32_t E = nrLDPC_TB_encoding_parameters->segments[seg_id].E;
-      Emax = E > Emax ? E : Emax;
+#if defined(ENABLE_CUDA)     
+    if (nrLDPC_slot_encoding_parameters->use_gpu && nrLDPC_TB_encoding_parameters->C > 8 && nrLDPC_TB_encoding_parameters->Z == 384) {
+        nrLDPC_coding_encoder32(nrLDPC_slot_encoding_parameters,nrLDPC_TB_encoding_parameters);
     }
-  }
+    else {
+#endif
+      size_t n_seg = (nrLDPC_TB_encoding_parameters->C / 8 + ((nrLDPC_TB_encoding_parameters->C & 7) == 0 ? 0 : 1));
+      nbTasks += n_seg;
 
+      // Search for maximum E for sizing encoder output f and f2
+      for (int seg_id = 0; seg_id < nrLDPC_TB_encoding_parameters->C; seg_id++) {
+        uint32_t E = nrLDPC_TB_encoding_parameters->segments[seg_id].E;
+        Emax = E > Emax ? E : Emax;
+      }
+#if defined(ENABLE_CUDA) 
+    }
+#endif
+  }
   // Create f and f2 to old encoding tasks outputs
   uint32_t Emax_ceil_mod = ceil_mod(Emax, 64);
   uint8_t f[nbTasks][Emax_ceil_mod] __attribute__((aligned(64)));
@@ -432,6 +442,10 @@ int nrLDPC_coding_encoder(nrLDPC_slot_encoding_parameters_t *nrLDPC_slot_encodin
     // For easier indexing we store the pointers to sub arrays of f and f2 in pointer arrays
     // Then a function to which we pass the pointer arrays can directly use f_2d[j] ans f2_2d[j]
     nrLDPC_TB_encoding_parameters_t *nrLDPC_TB_encoding_parameters = &nrLDPC_slot_encoding_parameters->TBs[dlsch_id];
+#if defined(ENABLE_CUDA) 
+    if (nrLDPC_slot_encoding_parameters->use_gpu && nrLDPC_TB_encoding_parameters->C > 8 && nrLDPC_TB_encoding_parameters->Z == 384) 
+      continue;
+#endif
     size_t n_seg = (nrLDPC_TB_encoding_parameters->C / 8 + ((nrLDPC_TB_encoding_parameters->C & 7) == 0 ? 0 : 1));
     uint8_t *f_2d[n_seg];
     uint8_t *f2_2d[n_seg];
@@ -449,16 +463,21 @@ int nrLDPC_coding_encoder(nrLDPC_slot_encoding_parameters_t *nrLDPC_slot_encodin
     completed_many_task_ans(&ans, nbTasks - nbEncode);
   }
   // Execute thread pool tasks
-  join_task_ans(&ans);
+  if (nbTasks > 0)  join_task_ans(&ans);
 
   // Write output
+  time_stats_t *tconcat = nrLDPC_slot_encoding_parameters->tconcat;
+  if(tconcat != NULL) start_meas(tconcat);
   nbTasks = 0;
   for (int dlsch_id = 0; dlsch_id < nrLDPC_slot_encoding_parameters->nb_TBs; dlsch_id++) {
     nrLDPC_TB_encoding_parameters_t *nrLDPC_TB_encoding_parameters = &nrLDPC_slot_encoding_parameters->TBs[dlsch_id];
     uint32_t C = nrLDPC_TB_encoding_parameters->C;
+#if defined(ENABLE_CUDA)
+    if (nrLDPC_slot_encoding_parameters->use_gpu && C > 8 && nrLDPC_TB_encoding_parameters->Z == 384) 
+      continue;
+#endif
     size_t n_seg = (C / 8 + ((C & 7) == 0 ? 0 : 1));
 
-    time_stats_t *toutput = nrLDPC_slot_encoding_parameters->toutput;
 
     for (int j = 0; j < n_seg; j++) {
       unsigned int macro_segment = j * 8;
@@ -478,7 +497,6 @@ int nrLDPC_coding_encoder(nrLDPC_slot_encoding_parameters_t *nrLDPC_slot_encodin
         }
       }
 
-      if(toutput != NULL) start_meas(toutput);
  
       uint32_t Eoffset=0;
       for (int s=0; s<macro_segment; s++)
@@ -494,10 +512,10 @@ int nrLDPC_coding_encoder(nrLDPC_slot_encoding_parameters_t *nrLDPC_slot_encodin
                         nrLDPC_TB_encoding_parameters->output,
                         Eoffset);
 
-      if(toutput != NULL) stop_meas(toutput);
     }
     nbTasks += n_seg;
   }
+  if(tconcat != NULL) stop_meas(tconcat);
 
   return 0;
 }

openair1/PHY/CODING/nrLDPC_coding/nrLDPC_coding_segment/nrLDPC_coding_segment_encoder_cuda.c

@@ -0,0 +1,688 @@
+/*
+ * 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 1.0  (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 PHY/CODING/nrLDPC_coding/nrLDPC_coding_segment/nrLDPC_coding_segment_encoder.c
+ * \brief Top-level routines for implementing LDPC encoding of transport channels
+ */
+
+#include "nr_rate_matching.h"
+#include "PHY/defs_gNB.h"
+#include "PHY/CODING/coding_extern.h"
+#include "PHY/CODING/coding_defs.h"
+#include "PHY/CODING/lte_interleaver_inline.h"
+#include "PHY/CODING/nrLDPC_coding/nrLDPC_coding_interface.h"
+#include "PHY/CODING/nrLDPC_extern.h"
+#include "PHY/NR_TRANSPORT/nr_transport_proto.h"
+#include "PHY/NR_TRANSPORT/nr_transport_common_proto.h"
+#include "PHY/NR_TRANSPORT/nr_dlsch.h"
+#include "SCHED_NR/sched_nr.h"
+#include "common/utils/LOG/vcd_signal_dumper.h"
+#include "common/utils/LOG/log.h"
+#include "common/utils/nr/nr_common.h"
+#include <openair2/UTIL/OPT/opt.h>
+
+#include <syscall.h>
+
+//#define DEBUG_LDPC_ENCODING
+//#define DEBUG_LDPC_ENCODING_FREE 1
+
+extern uint32_t **d_host;
+
+static void unpack_output(uint32_t *f,
+                         uint32_t E,
+                         uint32_t *f2,
+                         uint32_t E2,
+			 uint32_t E2_first_segment32,
+                         uint32_t E2_first_segment,
+                         uint32_t nb_segments,
+                         uint8_t *output) {
+
+  uint32_t s;
+ // int s0;
+  uint32_t *fp;
+  int foffset;
+  uint32_t *output_p = (uint32_t *)output;
+  //printf("unpack: E %d, E2 %d, E2_first_segment %d, E2_first_segment32 %d, nb_segments %d\n",E,E2,E2_first_segment,E2_first_segment32,nb_segments);
+
+  uint32_t bit_index = 0;
+#ifdef __AVX2__
+  simde__m256i shift0=simde_mm256_set_epi32(7,6,5,4,3,2,1,0);
+  simde__m256i shift1=simde_mm256_set_epi32(15,14,13,12,11,10,9,8);
+  simde__m256i shift2=simde_mm256_set_epi32(23,22,21,20,19,18,17,16);
+  simde__m256i shift3=simde_mm256_set_epi32(31,30,29,28,27,26,25,24);
+  simde__m256i vmask0=simde_mm256_set_epi32(0x80,0x40,0x20,0x10,0x8,0x4,0x2,0x1);
+  simde__m256i vmask1=simde_mm256_set_epi32(0x8000,0x4000,0x2000,0x1000,0x800,0x400,0x200,0x100);
+  simde__m256i vmask2=simde_mm256_set_epi32(0x800000,0x400000,0x200000,0x100000,0x80000,0x40000,0x20000,0x10000);
+  simde__m256i vmask3=simde_mm256_set_epi32(0x80000000,0x40000000,0x20000000,0x10000000,0x8000000,0x4000000,0x2000000,0x1000000);
+  int s2=0;
+  for (s = 0; s < E2_first_segment ; s++) { 
+    s2 = s&31;	  
+    foffset = (s>>5)*E;
+    fp = f+foffset;
+    int i;
+    if ((bit_index&31) == 0 ) {
+      for (i = 0; i < (E>>5)<<5; i+=32) {
+	simde__m256i f256 = simde_mm256_srli_epi32(*(simde__m256i*)&fp[i],s2);    
+  	simde__m256i cshift = simde_mm256_and_si256(simde_mm256_sllv_epi32(f256,shift0),vmask0);
+	f256 = simde_mm256_srli_epi32(*(simde__m256i*)&fp[i+8],s2);    
+  	cshift = simde_mm256_or_si256(simde_mm256_and_si256(simde_mm256_sllv_epi32(f256,shift1),vmask1),cshift);
+	f256 = simde_mm256_srli_epi32(*(simde__m256i*)&fp[i+16],s2);    
+  	cshift = simde_mm256_or_si256(simde_mm256_and_si256(simde_mm256_sllv_epi32(f256,shift2),vmask2),cshift);
+	f256 = simde_mm256_srli_epi32(*(simde__m256i*)&fp[i+24],s2);    
+  	cshift = simde_mm256_or_si256(simde_mm256_and_si256(simde_mm256_sllv_epi32(f256,shift3),vmask3),cshift);
+	*(output_p + (bit_index>>5))     = simde_mm256_extract_epi32(cshift,0) | 
+	                                   simde_mm256_extract_epi32(cshift,1) | 
+	                                   simde_mm256_extract_epi32(cshift,2) | 
+	                                   simde_mm256_extract_epi32(cshift,3) | 
+	                                   simde_mm256_extract_epi32(cshift,4) | 
+	                                   simde_mm256_extract_epi32(cshift,5) | 
+	                                   simde_mm256_extract_epi32(cshift,6) | 
+	                                   simde_mm256_extract_epi32(cshift,7);
+	bit_index+=32;
+      }
+      uint32_t Emod32=E&31;
+      if (Emod32 != 0) {
+	simde__m256i f256 = simde_mm256_srli_epi32(*(simde__m256i*)&fp[i],s2);    
+  	simde__m256i cshift = simde_mm256_and_si256(simde_mm256_sllv_epi32(f256,shift0),vmask0);
+	f256 = simde_mm256_srli_epi32(*(simde__m256i*)&fp[i+8],s2);    
+  	cshift = simde_mm256_or_si256(simde_mm256_and_si256(simde_mm256_sllv_epi32(f256,shift1),vmask1),cshift);
+	f256 = simde_mm256_srli_epi32(*(simde__m256i*)&fp[i+16],s2);    
+  	cshift = simde_mm256_or_si256(simde_mm256_and_si256(simde_mm256_sllv_epi32(f256,shift2),vmask2),cshift);
+	f256 = simde_mm256_srli_epi32(*(simde__m256i*)&fp[i+24],s2);    
+  	cshift = simde_mm256_or_si256(simde_mm256_and_si256(simde_mm256_sllv_epi32(f256,shift3),vmask3),cshift);
+	*(output_p + (bit_index>>5))     = (simde_mm256_extract_epi32(cshift,0) | 
+	                                   simde_mm256_extract_epi32(cshift,1) | 
+	                                   simde_mm256_extract_epi32(cshift,2) | 
+	                                   simde_mm256_extract_epi32(cshift,3) | 
+	                                   simde_mm256_extract_epi32(cshift,4) | 
+	                                   simde_mm256_extract_epi32(cshift,5) | 
+	                                   simde_mm256_extract_epi32(cshift,6) | 
+	                                   simde_mm256_extract_epi32(cshift,7))&((1<<Emod32)-1);
+        bit_index+=Emod32;
+      }
+    }
+    else {
+      for (i = 0; i < (E>>5)<<5; i+=32) {
+	simde__m256i f256 = simde_mm256_srli_epi32(*(simde__m256i*)&fp[i],s2);    
+  	simde__m256i cshift = simde_mm256_and_si256(simde_mm256_sllv_epi32(f256,shift0),vmask0);
+	f256 = simde_mm256_srli_epi32(*(simde__m256i*)&fp[i+8],s2);    
+  	cshift = simde_mm256_or_si256(simde_mm256_and_si256(simde_mm256_sllv_epi32(f256,shift1),vmask1),cshift);
+	f256 = simde_mm256_srli_epi32(*(simde__m256i*)&fp[i+16],s2);    
+  	cshift = simde_mm256_or_si256(simde_mm256_and_si256(simde_mm256_sllv_epi32(f256,shift2),vmask2),cshift);
+	f256 = simde_mm256_srli_epi32(*(simde__m256i*)&fp[i+24],s2);    
+  	cshift = simde_mm256_or_si256(simde_mm256_and_si256(simde_mm256_sllv_epi32(f256,shift3),vmask3),cshift);
+	uint32_t tmp    = simde_mm256_extract_epi32(cshift,0) | 
+	                  simde_mm256_extract_epi32(cshift,1) | 
+	                  simde_mm256_extract_epi32(cshift,2) | 
+	                  simde_mm256_extract_epi32(cshift,3) | 
+	                  simde_mm256_extract_epi32(cshift,4) | 
+	                  simde_mm256_extract_epi32(cshift,5) | 
+	                  simde_mm256_extract_epi32(cshift,6) | 
+	                  simde_mm256_extract_epi32(cshift,7);
+        *(output_p + (bit_index>>5))     |= (tmp<<(bit_index&31));
+        *(output_p + (bit_index>>5)+1)   |= (tmp>>(32-(bit_index&31)));					   
+	bit_index+=32;
+      }
+      uint32_t Emod32=E&31;
+      if (Emod32 != 0) {
+	simde__m256i f256 = simde_mm256_srli_epi32(*(simde__m256i*)&fp[i],s2);    
+  	simde__m256i cshift = simde_mm256_and_si256(simde_mm256_sllv_epi32(f256,shift0),vmask0);
+	f256 = simde_mm256_srli_epi32(*(simde__m256i*)&fp[i+8],s2);    
+  	cshift = simde_mm256_or_si256(simde_mm256_and_si256(simde_mm256_sllv_epi32(f256,shift1),vmask1),cshift);
+	f256 = simde_mm256_srli_epi32(*(simde__m256i*)&fp[i+16],s2);    
+  	cshift = simde_mm256_or_si256(simde_mm256_and_si256(simde_mm256_sllv_epi32(f256,shift2),vmask2),cshift);
+	f256 = simde_mm256_srli_epi32(*(simde__m256i*)&fp[i+24],s2);    
+  	cshift = simde_mm256_or_si256(simde_mm256_and_si256(simde_mm256_sllv_epi32(f256,shift3),vmask3),cshift);
+	uint32_t tmp = 	(simde_mm256_extract_epi32(cshift,0) | 
+	                 simde_mm256_extract_epi32(cshift,1) | 
+	                 simde_mm256_extract_epi32(cshift,2) | 
+	                 simde_mm256_extract_epi32(cshift,3) | 
+	                 simde_mm256_extract_epi32(cshift,4) | 
+	                 simde_mm256_extract_epi32(cshift,5) | 
+	                 simde_mm256_extract_epi32(cshift,6) | 
+	                 simde_mm256_extract_epi32(cshift,7))&((1<<Emod32)-1);
+	*(output_p + (bit_index>>5))     |= (tmp<<(bit_index&31));
+        *(output_p + (bit_index>>5)+1)   |= (tmp>>(32-(bit_index&31)));
+        bit_index+=Emod32;
+      }
+    }
+  }
+  for ( ; s < nb_segments ; s++){
+    s2 = s&31;	  
+    foffset = ((s>>5)-E2_first_segment32)*E2;
+    fp = f2+foffset;
+    int i;
+    if ((bit_index&31) == 0 ) {
+      for (i = 0; i < (E2>>5)<<5; i+=32) {
+	simde__m256i f256 = simde_mm256_srli_epi32(*(simde__m256i*)&fp[i],s2);    
+  	simde__m256i cshift = simde_mm256_and_si256(simde_mm256_sllv_epi32(f256,shift0),vmask0);
+	f256 = simde_mm256_srli_epi32(*(simde__m256i*)&fp[i+8],s2);    
+  	cshift = simde_mm256_or_si256(simde_mm256_and_si256(simde_mm256_sllv_epi32(f256,shift1),vmask1),cshift);
+	f256 = simde_mm256_srli_epi32(*(simde__m256i*)&fp[i+16],s2);    
+  	cshift = simde_mm256_or_si256(simde_mm256_and_si256(simde_mm256_sllv_epi32(f256,shift2),vmask2),cshift);
+	f256 = simde_mm256_srli_epi32(*(simde__m256i*)&fp[i+24],s2);    
+  	cshift = simde_mm256_or_si256(simde_mm256_and_si256(simde_mm256_sllv_epi32(f256,shift3),vmask3),cshift);
+	*(output_p + (bit_index>>5))     = simde_mm256_extract_epi32(cshift,0) | 
+	                                   simde_mm256_extract_epi32(cshift,1) | 
+	                                   simde_mm256_extract_epi32(cshift,2) | 
+	                                   simde_mm256_extract_epi32(cshift,3) | 
+	                                   simde_mm256_extract_epi32(cshift,4) | 
+	                                   simde_mm256_extract_epi32(cshift,5) | 
+	                                   simde_mm256_extract_epi32(cshift,6) | 
+	                                   simde_mm256_extract_epi32(cshift,7);
+	bit_index+=32;
+      }
+      uint32_t E2mod32=E2&31;
+      if (E2mod32 != 0) {
+	simde__m256i f256 = simde_mm256_srli_epi32(*(simde__m256i*)&fp[i],s2);    
+  	simde__m256i cshift = simde_mm256_and_si256(simde_mm256_sllv_epi32(f256,shift0),vmask0);
+	f256 = simde_mm256_srli_epi32(*(simde__m256i*)&fp[i+8],s2);    
+  	cshift = simde_mm256_or_si256(simde_mm256_and_si256(simde_mm256_sllv_epi32(f256,shift1),vmask1),cshift);
+	f256 = simde_mm256_srli_epi32(*(simde__m256i*)&fp[i+16],s2);    
+  	cshift = simde_mm256_or_si256(simde_mm256_and_si256(simde_mm256_sllv_epi32(f256,shift2),vmask2),cshift);
+	f256 = simde_mm256_srli_epi32(*(simde__m256i*)&fp[i+24],s2);    
+  	cshift = simde_mm256_or_si256(simde_mm256_and_si256(simde_mm256_sllv_epi32(f256,shift3),vmask3),cshift);
+	*(output_p + (bit_index>>5))     = (simde_mm256_extract_epi32(cshift,0) | 
+	                                   simde_mm256_extract_epi32(cshift,1) | 
+	                                   simde_mm256_extract_epi32(cshift,2) | 
+	                                   simde_mm256_extract_epi32(cshift,3) | 
+	                                   simde_mm256_extract_epi32(cshift,4) | 
+	                                   simde_mm256_extract_epi32(cshift,5) | 
+	                                   simde_mm256_extract_epi32(cshift,6) | 
+	                                   simde_mm256_extract_epi32(cshift,7))&((1<<E2mod32)-1);
+        bit_index+=E2mod32;
+      }
+    }
+    else {
+      for (i = 0; i < (E2>>5)<<5; i+=32) {
+	simde__m256i f256 = simde_mm256_srli_epi32(*(simde__m256i*)&fp[i],s2);    
+  	simde__m256i cshift = simde_mm256_and_si256(simde_mm256_sllv_epi32(f256,shift0),vmask0);
+	f256 = simde_mm256_srli_epi32(*(simde__m256i*)&fp[i+8],s2);    
+  	cshift = simde_mm256_or_si256(simde_mm256_and_si256(simde_mm256_sllv_epi32(f256,shift1),vmask1),cshift);
+	f256 = simde_mm256_srli_epi32(*(simde__m256i*)&fp[i+16],s2);    
+  	cshift = simde_mm256_or_si256(simde_mm256_and_si256(simde_mm256_sllv_epi32(f256,shift2),vmask2),cshift);
+	f256 = simde_mm256_srli_epi32(*(simde__m256i*)&fp[i+24],s2);    
+  	cshift = simde_mm256_or_si256(simde_mm256_and_si256(simde_mm256_sllv_epi32(f256,shift3),vmask3),cshift);
+	uint32_t tmp    = simde_mm256_extract_epi32(cshift,0) | 
+	                  simde_mm256_extract_epi32(cshift,1) | 
+	                  simde_mm256_extract_epi32(cshift,2) | 
+	                  simde_mm256_extract_epi32(cshift,3) | 
+	                  simde_mm256_extract_epi32(cshift,4) | 
+	                  simde_mm256_extract_epi32(cshift,5) | 
+	                  simde_mm256_extract_epi32(cshift,6) | 
+	                  simde_mm256_extract_epi32(cshift,7);
+        *(output_p + (bit_index>>5))     |= (tmp<<(bit_index&31));
+        *(output_p + (bit_index>>5)+1)   |= (tmp>>(32-(bit_index&31)));					   
+	bit_index+=32;
+      }
+      uint32_t E2mod32=E2&31;
+      if (E2mod32 != 0) {
+	simde__m256i f256 = simde_mm256_srli_epi32(*(simde__m256i*)&fp[i],s2);    
+  	simde__m256i cshift = simde_mm256_and_si256(simde_mm256_sllv_epi32(f256,shift0),vmask0);
+	f256 = simde_mm256_srli_epi32(*(simde__m256i*)&fp[i+8],s2);    
+  	cshift = simde_mm256_or_si256(simde_mm256_and_si256(simde_mm256_sllv_epi32(f256,shift1),vmask1),cshift);
+	f256 = simde_mm256_srli_epi32(*(simde__m256i*)&fp[i+16],s2);    
+  	cshift = simde_mm256_or_si256(simde_mm256_and_si256(simde_mm256_sllv_epi32(f256,shift2),vmask2),cshift);
+	f256 = simde_mm256_srli_epi32(*(simde__m256i*)&fp[i+24],s2);    
+  	cshift = simde_mm256_or_si256(simde_mm256_and_si256(simde_mm256_sllv_epi32(f256,shift3),vmask3),cshift);
+	uint32_t tmp = 	(simde_mm256_extract_epi32(cshift,0) | 
+	                 simde_mm256_extract_epi32(cshift,1) | 
+	                 simde_mm256_extract_epi32(cshift,2) | 
+	                 simde_mm256_extract_epi32(cshift,3) | 
+	                 simde_mm256_extract_epi32(cshift,4) | 
+	                 simde_mm256_extract_epi32(cshift,5) | 
+	                 simde_mm256_extract_epi32(cshift,6) | 
+	                 simde_mm256_extract_epi32(cshift,7))&((1<<E2mod32)-1);
+	*(output_p + (bit_index>>5))     |= (tmp<<(bit_index&31));
+        *(output_p + (bit_index>>5)+1)   |= (tmp>>(32-(bit_index&31)));
+        bit_index+=E2mod32;
+      }
+    }
+  }
+#elif defined(__aarch64__)
+  const int32_t ucShift0[32][4] = { {0,1,2,3}, {-1,0,1,2},{-2,-1,0,1}, {-3,-2,-1,0}, {-4,-3,-2,-1}, {-5,-4,-3,-2}, {-6,-5,-4,-3}, {-7,-6,-5,-4}, {-8,-7,-6,-5}, {-9,-8,-7,-6}, {-10,-9,-8,-7}, {-11,-10,-9,-8}, {-12,-11,-10,-9}, {-13,-12,-11,-10}, {-14,-13,-12,-11}, {-15,-14,-13,-12}, {-16,-15,-14,-13}, {-17,-16,-15,-14}, {-18,-17,-16,-15}, {-19,-18,-17,-16}, {-20,-19,-18,-17}, {-21,-20,-19,-18}, {-22,-21,-20,-19}, {-23,-22,-21,-20}, {-24,-23,-22,-21}, {-25,-24,-23,-22}, {-26,-25,-24,-23}, {-27,-26,-25,-24}, {-28,-27,-26,-25}, {-29,-28,-27,-26}, {-30,-29,-28,-27}, {-31,-30,-29,-28}}; 
+
+  const int32_t ucShift1[32][4] = { {4,5,6,7}, {3,4,5,6}, {2,3,4,5}, {1,2,3,4}, {0,1,2,3}, {-1,0,1,2},{-2,-1,0,1}, {-3,-2,-1,0}, {-4,-3,-2,-1}, {-5,-4,-3,-2}, {-6,-5,-4,-3}, {-7,-6,-5,-4}, {-8,-7,-6,-5}, {-9,-8,-7,-6}, {-10,-9,-8,-7}, {-11,-10,-9,-8}, {-12,-11,-10,-9}, {-13,-12,-11,-10}, {-14,-13,-12,-11}, {-15,-14,-13,-12}, {-16,-15,-14,-13}, {-17,-16,-15,-14}, {-18,-17,-16,-15}, {-19,-18,-17,-16}, {-20,-19,-18,-17}, {-21,-20,-19,-18}, {-22,-21,-20,-19}, {-23,-22,-21,-20}, {-24,-23,-22,-21}, {-25,-24,-23,-22}, {-26,-25,-24,-23}, {-27,-26,-25,-24}}; 
+
+  const int32_t ucShift2[32][4] = { {8,9,10,11},{7,8,9,10}, {6,7,8,9}, {5,6,7,8}, {4,5,6,7}, {3,4,5,6}, {2,3,4,5}, {1,2,3,4}, {0,1,2,3}, {-1,0,1,2},{-2,-1,0,1}, {-3,-2,-1,0}, {-4,-3,-2,-1}, {-5,-4,-3,-2}, {-6,-5,-4,-3}, {-7,-6,-5,-4}, {-8,-7,-6,-5}, {-9,-8,-7,-6}, {-10,-9,-8,-7}, {-11,-10,-9,-8}, {-12,-11,-10,-9}, {-13,-12,-11,-10}, {-14,-13,-12,-11}, {-15,-14,-13,-12}, {-16,-15,-14,-13}, {-17,-16,-15,-14}, {-18,-17,-16,-15}, {-19,-18,-17,-16}, {-20,-19,-18,-17}, {-21,-20,-19,-18}, {-22,-21,-20,-19},{-23,-22,-21,-20}}; 
+
+  const int32_t ucShift3[32][4] = { {12,13,14,15}, {11,12,13,14}, {10,11,12,13}, {9,10,11,12}, {8,9,10,11},{7,8,9,10}, {6,7,8,9}, {5,6,7,8}, {4,5,6,7}, {3,4,5,6}, {2,3,4,5}, {1,2,3,4}, {0,1,2,3}, {-1,0,1,2},{-2,-1,0,1}, {-3,-2,-1,0}, {-4,-3,-2,-1}, {-5,-4,-3,-2}, {-6,-5,-4,-3}, {-7,-6,-5,-4}, {-8,-7,-6,-5}, {-9,-8,-7,-6}, {-10,-9,-8,-7}, {-11,-10,-9,-8}, {-12,-11,-10,-9}, {-13,-12,-11,-10}, {-14,-13,-12,-11}, {-15,-14,-13,-12}, {-16,-15,-14,-13}, {-17,-16,-15,-14}, {-18,-17,-16,-15}, {-19,-18,-17,-16}}; 
+
+  const int32_t ucShift4[32][4] = { {16,17,18,19}, {15,16,17,18}, {14,15,16,17}, {13,14,15,16}, {12,13,14,15}, {11,12,13,14}, {10,11,12,13}, {9,10,11,12}, {8,9,10,11},{7,8,9,10}, {6,7,8,9}, {5,6,7,8}, {4,5,6,7}, {3,4,5,6}, {2,3,4,5}, {1,2,3,4}, {0,1,2,3}, {-1,0,1,2},{-2,-1,0,1}, {-3,-2,-1,0}, {-4,-3,-2,-1}, {-5,-4,-3,-2}, {-6,-5,-4,-3}, {-7,-6,-5,-4}, {-8,-7,-6,-5}, {-9,-8,-7,-6}, {-10,-9,-8,-7}, {-11,-10,-9,-8}, {-12,-11,-10,-9}, {-13,-12,-11,-10}, {-14,-13,-12,-11}, {-15,-14,-13,-12}}; 
+
+  const int32_t ucShift5[32][4] = { {20,21,22,23}, {19,20,21,22}, {18,19,20,21}, {17,18,19,20}, {16,17,18,19}, {15,16,17,18}, {14,15,16,17}, {13,14,15,16}, {12,13,14,15}, {11,12,13,14}, {10,11,12,13}, {9,10,11,12}, {8,9,10,11},{7,8,9,10}, {6,7,8,9}, {5,6,7,8}, {4,5,6,7}, {3,4,5,6}, {2,3,4,5}, {1,2,3,4}, {0,1,2,3}, {-1,0,1,2},{-2,-1,0,1}, {-3,-2,-1,0}, {-4,-3,-2,-1}, {-5,-4,-3,-2}, {-6,-5,-4,-3}, {-7,-6,-5,-4}, {-8,-7,-6,-5}, {-9,-8,-7,-6}, {-10,-9,-8,-7}, {-11,-10,-9,-8}}; 
+
+  const int32_t ucShift6[32][4] = { {24,25,26,27}, {23,24,25,26}, {22,23,24,25}, {21,22,23,24}, {20,21,22,23}, {19,20,21,22}, {18,19,20,21}, {17,18,19,20}, {16,17,18,19}, {15,16,17,18}, {14,15,16,17}, {13,14,15,16}, {12,13,14,15}, {11,12,13,14}, {10,11,12,13}, {9,10,11,12}, {8,9,10,11},{7,8,9,10}, {6,7,8,9}, {5,6,7,8}, {4,5,6,7}, {3,4,5,6}, {2,3,4,5}, {1,2,3,4}, {0,1,2,3}, {-1,0,1,2},{-2,-1,0,1}, {-3,-2,-1,0}, {-4,-3,-2,-1}, {-5,-4,-3,-2}, {-6,-5,-4,-3}, {-7,-6,-5,-4}}; 
+
+  const int32_t ucShift7[32][4] = { {28,29,30,31}, {27,28,29,30}, {26,27,28,29}, {25,26,27,28}, {24,25,26,27}, {23,24,25,26}, {22,23,24,25}, {21,22,23,24}, {20,21,22,23}, {19,20,21,22}, {18,19,20,21}, {17,18,19,20}, {16,17,18,19}, {15,16,17,18}, {14,15,16,17}, {13,14,15,16}, {12,13,14,15}, {11,12,13,14}, {10,11,12,13}, {9,10,11,12}, {8,9,10,11},{7,8,9,10}, {6,7,8,9}, {5,6,7,8}, {4,5,6,7}, {3,4,5,6}, {2,3,4,5}, {1,2,3,4}, {0,1,2,3}, {-1,0,1,2},{-2,-1,0,1}, {-3,-2,-1,0}}; 
+  const uint32_t __attribute__ ((aligned (16))) masks0[4] = {0x1,0x2,0x4,0x8};
+  const uint32_t __attribute__ ((aligned (16))) masks1[4] = {0x10,0x20,0x40,0x80};
+  const uint32_t __attribute__ ((aligned (16))) masks2[4] = {0x100,0x200,0x400,0x800};
+  const uint32_t __attribute__ ((aligned (16))) masks3[4] = {0x1000,0x2000,0x4000,0x8000};
+  const uint32_t __attribute__ ((aligned (16))) masks4[4] = {0x10000,0x20000,0x40000,0x80000};
+  const uint32_t __attribute__ ((aligned (16))) masks5[4] = {0x100000,0x200000,0x400000,0x800000};
+  const uint32_t __attribute__ ((aligned (16))) masks6[4] = {0x1000000,0x2000000,0x4000000,0x8000000};
+  const uint32_t __attribute__ ((aligned (16))) masks7[4] = {0x10000000,0x20000000,0x40000000,0x80000000};
+  int32x4_t vshift0[32],vshift1[32],vshift2[32],vshift3[32],vshift4[32],vshift5[32],vshift6[32],vshift7[32];
+  for (int n=0;n<32;n++) {
+	  vshift0[n] = vld1q_s32(ucShift0[n]);
+	  vshift1[n] = vld1q_s32(ucShift1[n]);
+	  vshift2[n] = vld1q_s32(ucShift2[n]);
+	  vshift3[n] = vld1q_s32(ucShift3[n]);
+	  vshift4[n] = vld1q_s32(ucShift4[n]);
+	  vshift5[n] = vld1q_s32(ucShift5[n]);
+	  vshift6[n] = vld1q_s32(ucShift6[n]);
+	  vshift7[n] = vld1q_s32(ucShift7[n]);
+  }
+  uint32x4_t vmask0  = vld1q_u32(masks0);
+  uint32x4_t vmask1  = vld1q_u32(masks1);
+  uint32x4_t vmask2  = vld1q_u32(masks2);
+  uint32x4_t vmask3  = vld1q_u32(masks3);
+  uint32x4_t vmask4  = vld1q_u32(masks4);
+  uint32x4_t vmask5  = vld1q_u32(masks5);
+  uint32x4_t vmask6  = vld1q_u32(masks6);
+  uint32x4_t vmask7  = vld1q_u32(masks7);
+  uint32_t output_tmp=0;
+  int s2=0;
+  for (s = 0; s < E2_first_segment ; s++) {
+    s2 = s&31;	  
+    foffset = (s>>5)*E;
+    fp = f+foffset;
+    int i;
+    if ((bit_index&31) == 0 ) {
+      for (i = 0; i < (E>>5)<<5; i+=32) {
+	uint32x4_t *fp128 = (uint32x4_t*)&fp[i];    
+  	uint32x4_t cshift = vandq_u32(vshlq_u32(fp128[0],vshift0[s2]),vmask0);
+  	cshift = vorrq_u32(cshift,vandq_u32(vshlq_u32(fp128[1],vshift1[s2]),vmask1));
+  	cshift = vorrq_u32(cshift,vandq_u32(vshlq_u32(fp128[2],vshift2[s2]),vmask2));
+  	cshift = vorrq_u32(cshift,vandq_u32(vshlq_u32(fp128[3],vshift3[s2]),vmask3));
+  	cshift = vorrq_u32(cshift,vandq_u32(vshlq_u32(fp128[4],vshift4[s2]),vmask4));
+  	cshift = vorrq_u32(cshift,vandq_u32(vshlq_u32(fp128[5],vshift5[s2]),vmask5));
+  	cshift = vorrq_u32(cshift,vandq_u32(vshlq_u32(fp128[6],vshift6[s2]),vmask6));
+  	cshift = vorrq_u32(cshift,vandq_u32(vshlq_u32(fp128[7],vshift7[s2]),vmask7));
+
+	*(output_p + (bit_index>>5))     = vaddvq_u32(cshift);
+	bit_index+=32;
+      }
+      uint32_t Emod32=E&31;
+      if (Emod32 != 0) {
+        uint32x4_t *fp128 = (uint32x4_t*)&fp[i];    
+  	uint32x4_t cshift = vandq_u32(vshlq_u32(fp128[0],vshift0[s2]),vmask0);
+  	cshift = vorrq_u32(cshift,vandq_u32(vshlq_u32(fp128[1],vshift1[s2]),vmask1));
+  	cshift = vorrq_u32(cshift,vandq_u32(vshlq_u32(fp128[2],vshift2[s2]),vmask2));
+  	cshift = vorrq_u32(cshift,vandq_u32(vshlq_u32(fp128[3],vshift3[s2]),vmask3));
+  	cshift = vorrq_u32(cshift,vandq_u32(vshlq_u32(fp128[4],vshift4[s2]),vmask4));
+  	cshift = vorrq_u32(cshift,vandq_u32(vshlq_u32(fp128[5],vshift5[s2]),vmask5));
+  	cshift = vorrq_u32(cshift,vandq_u32(vshlq_u32(fp128[6],vshift6[s2]),vmask6));
+  	cshift = vorrq_u32(cshift,vandq_u32(vshlq_u32(fp128[7],vshift7[s2]),vmask7));
+
+	*(output_p + (bit_index>>5))     = vaddvq_u32(cshift)&((1<<Emod32)-1);
+        bit_index+=Emod32;
+      }
+    }
+    else {
+      for (i = 0; i < (E>>5)<<5; i+=32) {
+	uint32x4_t *fp128 = (uint32x4_t*)&fp[i];    
+  	uint32x4_t cshift = vandq_u32(vshlq_u32(fp128[0],vshift0[s2]),vmask0);
+  	cshift = vorrq_u32(cshift,vandq_u32(vshlq_u32(fp128[1],vshift1[s2]),vmask1));
+  	cshift = vorrq_u32(cshift,vandq_u32(vshlq_u32(fp128[2],vshift2[s2]),vmask2));
+  	cshift = vorrq_u32(cshift,vandq_u32(vshlq_u32(fp128[3],vshift3[s2]),vmask3));
+  	cshift = vorrq_u32(cshift,vandq_u32(vshlq_u32(fp128[4],vshift4[s2]),vmask4));
+  	cshift = vorrq_u32(cshift,vandq_u32(vshlq_u32(fp128[5],vshift5[s2]),vmask5));
+  	cshift = vorrq_u32(cshift,vandq_u32(vshlq_u32(fp128[6],vshift6[s2]),vmask6));
+  	cshift = vorrq_u32(cshift,vandq_u32(vshlq_u32(fp128[7],vshift7[s2]),vmask7));
+	/*
+        uint32_t tmp = vaddvq_u32(cshift);
+	*(output_p + (bit_index>>5))     |= (tmp<<(bit_index&31));
+	*(output_p + (bit_index>>5)+1)   |= (tmp>>(32-(bit_index&31)));
+	*/
+        uint64_t tmp = (uint64_t)vaddvq_u32(cshift);
+	*(uint64_t*)(output_p + (bit_index>>5))     |= (tmp<<(bit_index&31));
+	bit_index+=32;
+      }
+      uint32_t Emod32=E&31;
+      if (Emod32 != 0) {
+        uint32x4_t *fp128 = (uint32x4_t*)&fp[i];    
+  	uint32x4_t cshift = vandq_u32(vshlq_u32(fp128[0],vshift0[s2]),vmask0);
+  	cshift = vorrq_u32(cshift,vandq_u32(vshlq_u32(fp128[1],vshift1[s2]),vmask1));
+  	cshift = vorrq_u32(cshift,vandq_u32(vshlq_u32(fp128[2],vshift2[s2]),vmask2));
+  	cshift = vorrq_u32(cshift,vandq_u32(vshlq_u32(fp128[3],vshift3[s2]),vmask3));
+  	cshift = vorrq_u32(cshift,vandq_u32(vshlq_u32(fp128[4],vshift4[s2]),vmask4));
+  	cshift = vorrq_u32(cshift,vandq_u32(vshlq_u32(fp128[5],vshift5[s2]),vmask5));
+  	cshift = vorrq_u32(cshift,vandq_u32(vshlq_u32(fp128[6],vshift6[s2]),vmask6));
+  	cshift = vorrq_u32(cshift,vandq_u32(vshlq_u32(fp128[7],vshift7[s2]),vmask7));
+        uint32_t tmp = vaddvq_u32(cshift);
+        tmp&=((1<<Emod32)-1);
+        *(output_p + (bit_index>>5))     |= (tmp<<(bit_index&31));
+	*(output_p + (bit_index>>5)+1)   |= (tmp>>(32-(bit_index&31)));
+        bit_index+=Emod32;
+      }
+    }
+  }
+//  s0 = s;
+  for ( ; s < nb_segments ; s++){
+    s2 = s&31;	  
+    foffset = ((s>>5)-E2_first_segment32)*E2;
+    fp = f2+foffset;
+    int i;
+    if ((bit_index&31) == 0 ) {
+      for (i = 0; i < (E2>>5)<<5; i+=32) {
+	uint32x4_t *fp128 = (uint32x4_t*)&fp[i];    
+  	uint32x4_t cshift = vandq_u32(vshlq_u32(fp128[0],vshift0[s2]),vmask0);
+  	cshift = vorrq_u32(cshift,vandq_u32(vshlq_u32(fp128[1],vshift1[s2]),vmask1));
+  	cshift = vorrq_u32(cshift,vandq_u32(vshlq_u32(fp128[2],vshift2[s2]),vmask2));
+  	cshift = vorrq_u32(cshift,vandq_u32(vshlq_u32(fp128[3],vshift3[s2]),vmask3));
+  	cshift = vorrq_u32(cshift,vandq_u32(vshlq_u32(fp128[4],vshift4[s2]),vmask4));
+  	cshift = vorrq_u32(cshift,vandq_u32(vshlq_u32(fp128[5],vshift5[s2]),vmask5));
+  	cshift = vorrq_u32(cshift,vandq_u32(vshlq_u32(fp128[6],vshift6[s2]),vmask6));
+  	cshift = vorrq_u32(cshift,vandq_u32(vshlq_u32(fp128[7],vshift7[s2]),vmask7));
+	*(output_p + (bit_index>>5))     = vaddvq_u32(cshift);
+	bit_index+=32;
+      }
+      uint32_t E2mod32=E2&31;
+      if (E2mod32 != 0) {
+        uint32x4_t *fp128 = (uint32x4_t*)&fp[i];    
+        uint32x4_t cshift = vandq_u32(vshlq_u32(fp128[0],vshift0[s2]),vmask0);
+  	cshift = vorrq_u32(cshift,vandq_u32(vshlq_u32(fp128[1],vshift1[s2]),vmask1));
+  	cshift = vorrq_u32(cshift,vandq_u32(vshlq_u32(fp128[2],vshift2[s2]),vmask2));
+  	cshift = vorrq_u32(cshift,vandq_u32(vshlq_u32(fp128[3],vshift3[s2]),vmask3));
+  	cshift = vorrq_u32(cshift,vandq_u32(vshlq_u32(fp128[4],vshift4[s2]),vmask4));
+  	cshift = vorrq_u32(cshift,vandq_u32(vshlq_u32(fp128[5],vshift5[s2]),vmask5));
+  	cshift = vorrq_u32(cshift,vandq_u32(vshlq_u32(fp128[6],vshift6[s2]),vmask6));
+  	cshift = vorrq_u32(cshift,vandq_u32(vshlq_u32(fp128[7],vshift7[s2]),vmask7));
+        *(output_p + (bit_index>>5))     = vaddvq_u32(cshift)&((1<<E2mod32)-1);
+        bit_index+=E2mod32;
+      }
+    }
+    else {
+      for (i = 0; i < (E2>>5)<<5; i+=32) {
+	uint32x4_t *fp128 = (uint32x4_t*)&fp[i];    
+        uint32x4_t cshift = vandq_u32(vshlq_u32(fp128[0],vshift0[s2]),vmask0);
+  	cshift = vorrq_u32(cshift,vandq_u32(vshlq_u32(fp128[1],vshift1[s2]),vmask1));
+  	cshift = vorrq_u32(cshift,vandq_u32(vshlq_u32(fp128[2],vshift2[s2]),vmask2));
+  	cshift = vorrq_u32(cshift,vandq_u32(vshlq_u32(fp128[3],vshift3[s2]),vmask3));
+  	cshift = vorrq_u32(cshift,vandq_u32(vshlq_u32(fp128[4],vshift4[s2]),vmask4));
+  	cshift = vorrq_u32(cshift,vandq_u32(vshlq_u32(fp128[5],vshift5[s2]),vmask5));
+  	cshift = vorrq_u32(cshift,vandq_u32(vshlq_u32(fp128[6],vshift6[s2]),vmask6));
+  	cshift = vorrq_u32(cshift,vandq_u32(vshlq_u32(fp128[7],vshift7[s2]),vmask7));
+        uint32_t tmp = vaddvq_u32(cshift);
+	*(output_p + (bit_index>>5))     |= (tmp<<(bit_index&31));
+	*(output_p + (bit_index>>5)+1)   |= (tmp>>(32-(bit_index&31)));
+	bit_index+=32;
+      }
+      uint32_t E2mod32=E2&31;
+      if (E2mod32 != 0) {
+        uint32x4_t *fp128 = (uint32x4_t*)&fp[i];    
+        uint32x4_t cshift = vandq_u32(vshlq_u32(fp128[0],vshift0[s2]),vmask0);
+  	cshift = vorrq_u32(cshift,vandq_u32(vshlq_u32(fp128[1],vshift1[s2]),vmask1));
+  	cshift = vorrq_u32(cshift,vandq_u32(vshlq_u32(fp128[2],vshift2[s2]),vmask2));
+  	cshift = vorrq_u32(cshift,vandq_u32(vshlq_u32(fp128[3],vshift3[s2]),vmask3));
+  	cshift = vorrq_u32(cshift,vandq_u32(vshlq_u32(fp128[4],vshift4[s2]),vmask4));
+  	cshift = vorrq_u32(cshift,vandq_u32(vshlq_u32(fp128[5],vshift5[s2]),vmask5));
+  	cshift = vorrq_u32(cshift,vandq_u32(vshlq_u32(fp128[6],vshift6[s2]),vmask6));
+        uint32_t tmp = vaddvq_u32(cshift);
+        tmp&=((1<<E2mod32)-1);
+        *(output_p + (bit_index>>5))     |= (tmp<<(bit_index&31));
+	*(output_p + (bit_index>>5)+1)   |= (tmp>>(32-(bit_index&31)));
+        bit_index+=E2mod32;
+      }
+    }
+  }
+#else // non SIMD version
+  unsigned int segpos,s2;	
+  for (s = 0; s < E2_first_segment ; s++) {
+    foffset = (s>>5)*E;
+    fp = f+foffset;
+    s2=s&31;
+    segpos = (1<<s2);
+  //  printf("E %d s %d: foffset %d, s2 %d, segpos %u\n",E,s,foffset,s2,segpos);
+    for (int i = 0; i < E; i++) {
+      output_p[bit_index>>5]|=((fp[i] & segpos)!=0)<<(bit_index&31); 
+      //printf("bit_index %d, output_p[%d] %x\n",bit_index, bit_index>>5,output_p[bit_index>>5]);
+      bit_index++;
+    }
+  }
+  for ( ; s < nb_segments ; s++){
+    foffset = ((s>>5)-E2_first_segment32)*E2;
+    fp = f2+foffset;
+    s2=s&31;
+    segpos = (1<<s2);
+//    printf("E2 %d s %d: foffset %d, s2 %d, segpos %u\n",E2,s,foffset,s2,segpos);
+    for (int i = 0; i < E2; i++) {
+      output_p[bit_index>>5]|=((fp[i] & segpos)!=0)<<(bit_index&31); 
+      bit_index++;
+    }
+  }
+#endif  
+}
+/**
+ * \typedef ldpc8blocks_args_t
+ * \struct ldpc8blocks_args_s
+ * \brief Arguments of an encoding task
+ * encode up to 8 code blocks
+ * \var nrLDPC_TB_encoding_parameters TB encoding parameters as defined in the coding library interface
+ * \var impp encoder implementation specific parameters for the task
+ * \var f first interleaver output to be filled by the task
+ * \var f2 second interleaver output to be filled by the task
+ * in case of a shift of E in the code blocks group processed by the task
+ */
+
+static void ldpcnblocks(nrLDPC_TB_encoding_parameters_t *nrLDPC_TB_encoding_parameters, encoder_implemparams_t impp)
+{
+ 
+  uint8_t mod_order = nrLDPC_TB_encoding_parameters->Qm;
+  uint16_t nb_rb = nrLDPC_TB_encoding_parameters->nb_rb;
+  uint32_t A = nrLDPC_TB_encoding_parameters->A;
+
+  unsigned int G = nrLDPC_TB_encoding_parameters->G;
+  LOG_D(PHY, "dlsch coding A %d K %d G %d (nb_rb %d, mod_order %d)\n", A, impp.K, G, nb_rb, (int)mod_order);
+
+  // nrLDPC_encoder output is in "d"
+  // let's make this interface happy!
+//  uint32_t d[4][68*384];
+  uint8_t *c[nrLDPC_TB_encoding_parameters->C];
+
+  if (!nrLDPC_TB_encoding_parameters->c_dev)  
+    for (int r = 0; r < nrLDPC_TB_encoding_parameters->C; r++)
+      c[r] = nrLDPC_TB_encoding_parameters->segments[r].c;
+  start_meas(&nrLDPC_TB_encoding_parameters->segments[impp.first_seg].ts_ldpc_encode);
+  LDPCencoder32(nrLDPC_TB_encoding_parameters->c_dev ? nrLDPC_TB_encoding_parameters->c_dev : c, &impp);
+  stop_meas(&nrLDPC_TB_encoding_parameters->segments[impp.first_seg].ts_ldpc_encode);
+  // Compute where to place in output buffer that is concatenation of all segments
+
+#ifdef DEBUG_LDPC_ENCODING
+  LOG_D(PHY, "rvidx in encoding = %d\n", nrLDPC_TB_encoding_parameters->rv_index);
+#endif
+  const uint32_t E = nrLDPC_TB_encoding_parameters->segments[0].E;
+  uint32_t E2=E;
+  uint32_t Emax = E;
+  int n_seg   = nrLDPC_TB_encoding_parameters->C>>5;
+  int n_seg2  = n_seg;
+  if ((nrLDPC_TB_encoding_parameters->C & 31) > 0) n_seg2++;
+  int r_shift = n_seg2; 
+  int r_shift2 = nrLDPC_TB_encoding_parameters->C;
+  for (int s=0;s<nrLDPC_TB_encoding_parameters->C;s++) {
+      //printf("segment %d E %d\n",s,nrLDPC_TB_encoding_parameters->segments[s].E);	  
+      if (nrLDPC_TB_encoding_parameters->segments[s].E != E) {
+	 E2=nrLDPC_TB_encoding_parameters->segments[s].E;
+         if(E2 > Emax)
+           Emax = E2;
+	 r_shift = s>>5;
+	 r_shift2 = s;
+	// printf("r_shift %d, r_shift2 %d\n",r_shift,r_shift2);
+         break;
+      }	 
+  }    
+
+  LOG_D(NR_PHY,
+        "Rate Matching, Code segment %d...%d r_shift %d n_seg2 %d (coded bits (G) %u, E %d, E2 %d Filler bits %d, Filler offset %d mod_order %d, nb_rb "
+          "%d,nrOfLayer %d)...\n",
+        0,
+        impp.n_segments-1,
+	r_shift,
+	n_seg2,
+        G,
+        E,E2,
+        impp.F,
+        impp.K - impp.F - 2 * impp.Zc,
+        mod_order,
+        nb_rb,
+        nrLDPC_TB_encoding_parameters->nb_layers);
+/*
+  printf("Rate Matching, Code segment 0..%d r_shift %d r_shift2 %d n_seg2 %d (coded bits (G) %u, E %d, E2 %d Filler bits %d, Filler offset %d mod_order %d, nb_rb "
+          "%d,nrOfLayer %d)...\n",
+        impp.n_segments-1,
+	r_shift,
+	r_shift2,
+	n_seg2,
+        G,
+        E,E2,
+        impp.F,
+        impp.K - impp.F - 2 * impp.Zc,
+        mod_order,
+        nb_rb,
+        nrLDPC_TB_encoding_parameters->nb_layers);
+*/
+
+  uint32_t Tbslbrm = nrLDPC_TB_encoding_parameters->tbslbrm;
+
+  uint32_t e[E*(r_shift+1)];
+  uint32_t e2[E2*(n_seg2-r_shift)];
+  uint32_t f[E*(r_shift+1)] __attribute__ ((aligned (64)));
+  uint32_t f2[E2*(n_seg2-r_shift)] __attribute__ ((aligned (64)));
+
+  // Interleaver outputs are stored in the output arrays
+  uint8_t *output = nrLDPC_TB_encoding_parameters->output;
+
+  start_meas(&nrLDPC_TB_encoding_parameters->segments[0].ts_rate_match);
+  memset(e,0,sizeof(e));
+  memset(f,0,sizeof(f));
+  if (1/*r_shift < n_seg2*/) { 
+    memset(e2,0,sizeof(e2));
+    memset(f2,0,sizeof(f2));
+  }
+
+  for (int r=0;r<n_seg2;r++) {
+    if (r<=r_shift)	  
+      nr_rate_matching_ldpc32(Tbslbrm,
+                              impp.BG,
+                              impp.Zc,
+                              d_host[r],
+                              e+(r*E),
+                              impp.n_segments,
+                              impp.F,
+                              impp.K - impp.F - 2 * impp.Zc,
+                              nrLDPC_TB_encoding_parameters->rv_index,
+                              E);
+    if (r>=r_shift)	  
+      nr_rate_matching_ldpc32(Tbslbrm,
+                              impp.BG,
+                              impp.Zc,
+                              d_host[r],
+                              e2+((r-r_shift)*E2),
+                              impp.n_segments,
+                              impp.F,
+                              impp.K - impp.F - 2 * impp.Zc,
+                              nrLDPC_TB_encoding_parameters->rv_index,
+                              E2);
+   /* 
+    if (r==(n_seg2-1)) {
+	    for (int i=0;i<16;i++) printf("rm: %x %x\n",d[n_seg2-1][i],e2[((n_seg2-1)*E2)+i]);
+    }
+    */
+  }
+  stop_meas(&nrLDPC_TB_encoding_parameters->segments[0].ts_rate_match);
+  if (impp.K - impp.F - 2 * impp.Zc > E) {
+    LOG_E(PHY,
+          "dlsch coding A %d  Kr %d G %d (nb_rb %d, mod_order %d)\n",
+          A,
+          impp.K,
+          G,
+          nb_rb,
+          (int)mod_order);
+
+    LOG_E(NR_PHY,
+          "Rate Matching, Code segments 0..%d (coded bits (G) %u, E %d, Kr %d, Filler bits %d, Filler offset %d mod_order %d, "
+          "nb_rb %d)...\n",
+          impp.n_segments,
+          G,
+          E,
+          impp.K,
+          impp.F,
+          impp.K - impp.F - 2 * impp.Zc,
+          mod_order,
+          nb_rb);
+  }
+  
+  //printf("interleaving r_shift %d, n_seg2 %d\n",r_shift,n_seg2);
+  start_meas(&nrLDPC_TB_encoding_parameters->segments[0].ts_interleave);
+  
+  for (int r=0;r<=r_shift;r++)
+    nr_interleaving_ldpc32(E,
+                           mod_order,
+                           e+E*r,
+                           f+E*r);
+
+  for (int r=r_shift;r<n_seg2;r++)
+    nr_interleaving_ldpc32(E2,
+                           mod_order,
+                           e2+E2*(r-r_shift),
+                           f2+E2*(r-r_shift));
+/*
+  for (int i=0;i<16;i++) printf("intl (f offset %d): %x %x\n",(n_seg2-1)*E2,e2[((n_seg2-1)*E2)+i],f2[((n_seg2-1)*E2)+i]);
+  printf("-------------------\n");
+  for (int i=E2-16;i<E2;i++) printf("intl (f offset %d): %x %x\n",(n_seg2-1)*E2,e2[((n_seg2-1)*E2)+i],f2[((n_seg2-1)*E2)+i]);
+  */
+  stop_meas(&nrLDPC_TB_encoding_parameters->segments[0].ts_interleave);
+
+  if (impp.tconcat) start_meas(impp.tconcat);
+  unpack_output(f,E,f2,E2,r_shift,r_shift2,nrLDPC_TB_encoding_parameters->C,output);
+  if (impp.tconcat) stop_meas(impp.tconcat);
+
+}
+
+int nrLDPC_coding_encoder32(nrLDPC_slot_encoding_parameters_t *nrLDPC_slot_encoding_parameters, nrLDPC_TB_encoding_parameters_t *nrLDPC_TB_encoding_parameters)
+{
+
+
+    encoder_implemparams_t common_segment_params = {
+      .n_segments = nrLDPC_TB_encoding_parameters->C,
+      .tinput = nrLDPC_slot_encoding_parameters->tinput,
+      .tinput_memcpy = nrLDPC_slot_encoding_parameters->tinput_memcpy,
+      .tprep = nrLDPC_slot_encoding_parameters->tprep,
+      .tparity = nrLDPC_slot_encoding_parameters->tparity,
+      .toutput = nrLDPC_slot_encoding_parameters->toutput,
+      .tconcat = nrLDPC_slot_encoding_parameters->tconcat,
+      .Kb = nrLDPC_TB_encoding_parameters->Kb,
+      .Zc = nrLDPC_TB_encoding_parameters->Z,
+      .BG = nrLDPC_TB_encoding_parameters->BG,
+      .output = nrLDPC_TB_encoding_parameters->output, 
+      .K = nrLDPC_TB_encoding_parameters->K,
+      .F = nrLDPC_TB_encoding_parameters->F,
+    };
+
+    LOG_D(NR_PHY,"Calling ldpcnblocks (C %d, Z %d, K %d)\n",common_segment_params.n_segments,common_segment_params.Zc,common_segment_params.K);
+    ldpcnblocks(nrLDPC_TB_encoding_parameters, common_segment_params);
+
+
+
+  return 0;
+}

openair1/PHY/CODING/nrLDPC_coding/nrLDPC_coding_segment/nr_rate_matching.c

@@ -24,7 +24,8 @@
 #include "common/utils/LOG/log.h"
 
 // #define RM_DEBUG 1
-
+#define USE128BIT
+ 
 static const uint8_t index_k0[2][4] = {{0, 17, 33, 56}, {0, 13, 25, 43}};
 
 void nr_interleaving_ldpc(uint32_t E, uint8_t Qm, uint8_t *e, uint8_t *f)
@@ -359,14 +360,15 @@ void nr_interleaving_ldpc(uint32_t E, uint8_t Qm, uint8_t *e, uint8_t *f)
       f = (uint8_t *)f_512;
 #endif
 #ifdef USE128BIT
-      e0_128 = (simde__m128i *)e0;
-      e1_128 = (simde__m128i *)e1;
-      e2_128 = (simde__m128i *)e2;
-      e3_128 = (simde__m128i *)e3;
-      e4_128 = (simde__m128i *)e4;
-      e5_128 = (simde__m128i *)e5;
-      e6_128 = (simde__m128i *)e6;
-      e7_128 = (simde__m128i *)e7;
+      simde__m128i *e0_128 = (simde__m128i *)e0;
+      simde__m128i *e1_128 = (simde__m128i *)e1;
+      simde__m128i *e2_128 = (simde__m128i *)e2;
+      simde__m128i *e3_128 = (simde__m128i *)e3;
+      simde__m128i *e4_128 = (simde__m128i *)e4;
+      simde__m128i *e5_128 = (simde__m128i *)e5;
+      simde__m128i *e6_128 = (simde__m128i *)e6;
+      simde__m128i *e7_128 = (simde__m128i *)e7;
+      simde__m128i *f_128  = (simde__m128i *)f;
       for (; i < (EQm & ~15); i += 16) {
         simde__m128i e0j = simde_mm_loadu_si128(e0_128++);
         simde__m128i e1j = simde_mm_loadu_si128(e1_128++);
@@ -444,6 +446,225 @@ void nr_interleaving_ldpc(uint32_t E, uint8_t Qm, uint8_t *e, uint8_t *f)
   }
 }
 
+void nr_interleaving_ldpc32(uint32_t E, uint8_t Qm, uint32_t *e, uint32_t *f)
+{
+  const uint32_t EQm = E / Qm;
+  memset(f, 0, E * sizeof(uint32_t));
+  switch(Qm) {
+    case 2: {
+      uint32_t *e0 = e;
+      uint32_t *e1 = e0 + EQm;
+      int i = 0;
+      for (; i < EQm; i++) {
+        *f++ = *e0++;
+        *f++ = *e1++;
+      }
+    } break;
+    case 4: {
+      uint32_t *e0 = e;
+      uint32_t *e1 = e0 + EQm;
+      uint32_t *e2 = e1 + EQm;
+      uint32_t *e3 = e2 + EQm;
+      int i = 0;
+      for (; i < EQm; i++) {
+        *f++ = *e0++;
+        *f++ = *e1++;
+        *f++ = *e2++;
+        *f++ = *e3++;
+      }
+    } break;
+    case 6: {
+      uint32_t *e0 = e;
+      uint32_t *e1 = e0 + EQm;
+      uint32_t *e2 = e1 + EQm;
+      uint32_t *e3 = e2 + EQm;
+      uint32_t *e4 = e3 + EQm;
+      uint32_t *e5 = e4 + EQm;
+      int i = 0;
+      simde__m128i *e0_128 = (simde__m128i *)e0;
+      simde__m128i *e1_128 = (simde__m128i *)e1;
+      simde__m128i *e2_128 = (simde__m128i *)e2;
+      simde__m128i *e3_128 = (simde__m128i *)e3;
+      simde__m128i *e4_128 = (simde__m128i *)e4;
+      simde__m128i *e5_128 = (simde__m128i *)e5;
+      simde__m128i *f128 = (simde__m128i *)f;
+      for (; i < (EQm & ~3); i += 4) {
+        simde__m128i e0j = simde_mm_loadu_si128(e0_128++);
+        simde__m128i e1j = simde_mm_loadu_si128(e1_128++);
+        simde__m128i e2j = simde_mm_loadu_si128(e2_128++);
+        simde__m128i e3j = simde_mm_loadu_si128(e3_128++);
+        simde__m128i e4j = simde_mm_loadu_si128(e4_128++);
+        simde__m128i e5j = simde_mm_loadu_si128(e5_128++);
+
+        simde__m128i tmp0 = simde_mm_unpacklo_epi32(e0j, e1j);   // e0(i) e1(i) e0(i+1) e1(i+1) 
+        simde__m128i tmp1 = simde_mm_unpacklo_epi32(e2j, e3j);   // e2(i) e3(i) e2(i+1) e3(i+1) 
+        simde__m128i tmp2 = simde_mm_unpacklo_epi32(e4j, e5j);   // e4(i) e5(i) e4(i+1) e5(i+1) 
+								 
+        simde_mm_storeu_si128(f128++,simde_mm_unpacklo_epi64(tmp0, tmp1));                        // e0(i) e1(i) e2(i) e3(i) 
+        simde_mm_storeu_si128(f128++,simde_mm_unpacklo_epi64(tmp2,simde_mm_unpackhi_epi64(tmp0,tmp0))); // e4(i) e5(i) e0(i+1) e1(i+1) 
+	simde_mm_storeu_si128(f128++,simde_mm_unpackhi_epi64(tmp1,tmp2));                         // e2(i+1) e3(i+1) e4(i+1) e5(i+1)
+        tmp0 = simde_mm_unpackhi_epi32(e0j, e1j);   // e0(i+2) e1(i+2) e0(i+3) e1(i+3) 
+        tmp1 = simde_mm_unpackhi_epi32(e2j, e3j);   // e2(i+2) e3(i+2) e2(i+3) e3(i+3) 
+        tmp2 = simde_mm_unpackhi_epi32(e4j, e5j);   // e4(i+2) e5(i+2) e4(i+3) e5(i+3) 
+								 
+        simde_mm_storeu_si128(f128++,simde_mm_unpacklo_epi64(tmp0, tmp1));                        // e0(i+2) e1(i+2) e2(i+2) e3(i+2) 
+        simde_mm_storeu_si128(f128++,simde_mm_unpacklo_epi64(tmp2,simde_mm_unpackhi_epi64(tmp0,tmp0))); // e4(i+2) e5(i+2) e0(i+3) e1(i+3) 
+	simde_mm_storeu_si128(f128++,simde_mm_unpackhi_epi64(tmp1,tmp2));                         // e2(i+3) e3(i+3) e4(i+3) e5(i+3)
+      }
+      e0 = (uint32_t *)e0_128;
+      e1 = (uint32_t *)e1_128;
+      e2 = (uint32_t *)e2_128;
+      e3 = (uint32_t *)e3_128;
+      e4 = (uint32_t *)e4_128;
+      e5 = (uint32_t *)e5_128;
+      f  = (uint32_t *)f128;
+      for (; i < EQm; i++) {
+        *f++ = *e0++;
+        *f++ = *e1++;
+        *f++ = *e2++;
+        *f++ = *e3++;
+        *f++ = *e4++;
+        *f++ = *e5++;
+      }
+    } break;
+    case 8: {
+      uint32_t *e0 = e;
+      uint32_t *e1 = e0 + EQm;
+      uint32_t *e2 = e1 + EQm;
+      uint32_t *e3 = e2 + EQm;
+      uint32_t *e4 = e3 + EQm;
+      uint32_t *e5 = e4 + EQm;
+      uint32_t *e6 = e5 + EQm;
+      uint32_t *e7 = e6 + EQm;
+      
+      int i = 0;
+#if 0 /*def __AVX2__, this needs to be fixed, revert to 128-bit below*/
+      simde__m256i *e0_256 = (simde__m256i *)e0;
+      simde__m256i *e1_256 = (simde__m256i *)e1;
+      simde__m256i *e2_256 = (simde__m256i *)e2;
+      simde__m256i *e3_256 = (simde__m256i *)e3;
+      simde__m256i *e4_256 = (simde__m256i *)e4;
+      simde__m256i *e5_256 = (simde__m256i *)e5;
+      simde__m256i *e6_256 = (simde__m256i *)e6;
+      simde__m256i *e7_256 = (simde__m256i *)e7;
+      simde__m256i *f256   = (simde__m256i *)f;
+
+      for (; i < (EQm & ~7); i += 8) {
+        simde__m256i e0j = simde_mm256_loadu_si256(e0_256++);
+        simde__m256i e1j = simde_mm256_loadu_si256(e1_256++);
+        simde__m256i e2j = simde_mm256_loadu_si256(e2_256++);
+        simde__m256i e3j = simde_mm256_loadu_si256(e3_256++);
+        simde__m256i e4j = simde_mm256_loadu_si256(e4_256++);
+        simde__m256i e5j = simde_mm256_loadu_si256(e5_256++);
+        simde__m256i e6j = simde_mm256_loadu_si256(e6_256++);
+        simde__m256i e7j = simde_mm256_loadu_si256(e7_256++);
+
+        simde__m256i tmp0 = simde_mm256_unpacklo_epi32(e0j, e1j);   // e0(i) e1(i) e0(i+1) e1(i+1) e0(i+4) e1(i+4) e0(i+5) e1(i+5) 
+        simde__m256i tmp1 = simde_mm256_unpacklo_epi32(e2j, e3j);   // e2(i) e3(i) e2(i+1) e3(i+1) e2(i+4) e3(i+4) e2(i+5) e3(i+5)
+        simde__m256i tmp2 = simde_mm256_unpacklo_epi32(e4j, e5j);   // e4(i) e5(i) e4(i+1) e5(i+1) e4(i+4) e5(i+4) e4(i+5) e5(i+5) 
+        simde__m256i tmp3 = simde_mm256_unpacklo_epi32(e6j, e7j);   // e6(i) e7(i) e6(i+1) e7(i+1) e6(i+4) e7(i+4) e6(i+5) e7(i+5)
+								    //
+        simde__m256i tmp4 = simde_mm256_unpacklo_epi64(tmp0, tmp1);   // e0(i) e1(i) e2(i) e3(i) e0(i+4) e1(i+4) e2(i+4) e3(i+4) 
+        simde__m256i tmp5 = simde_mm256_unpacklo_epi64(tmp2, tmp3);   // e4(i) e5(i) e6(i) e7(i) e4(i+4) e5(i+4) e6(i+4) e7(i+4) 
+								 
+        simde_mm256_storeu_si256(f256++,_mm256_permute2x128_si256(tmp4,tmp5,0x20)); // e0(i) e1(i) e2(i) e3(i) e4(i) e5(i) e6(i) e7(i) 
+										    
+        simde__m256i tmp6 = simde_mm256_unpackhi_epi64(tmp0, tmp1);   // e0(i+1) e1(i+1) e2(i+1) e3(i+1) e0(i+5) e1(i+5) e2(i+5) e3(i+5) 
+        simde__m256i tmp7 = simde_mm256_unpackhi_epi64(tmp2, tmp3);   // e4(i+1) e5(i+1) e6(i+1) e7(i+1) e4(i+5) e5(i+5) e6(i+5) e7(i+5) 
+        simde_mm256_storeu_si256(f256++,_mm256_permute2x128_si256(tmp6,tmp7,0x20)); // e0(i+1) e1(i+1) e2(i+1) e3(i+1) e4(i+1) e5(i+1) e6(i+1) e7(i+1) 
+										    //
+        tmp0 = simde_mm256_unpackhi_epi32(e0j, e1j);   // e0(i+2) e1(i+2) e0(i+3) e1(i+3) e0(i+6) e1(i+6) e0(i+7) e1(i+7) 
+        tmp1 = simde_mm256_unpackhi_epi32(e2j, e3j);   // e2(i+2) e3(i+2) e2(i+3) e3(i+3) e2(i+6) e3(i+6) e2(i+7) e3(i+7)
+        tmp2 = simde_mm256_unpackhi_epi32(e4j, e5j);   // e4(i+2) e5(i+2) e4(i+3) e5(i+3) e4(i+6) e5(i+6) e4(i+7) e5(i+7) 
+        tmp3 = simde_mm256_unpackhi_epi32(e6j, e7j);   // e6(i+2) e7(i+2) e6(i+3) e7(i+3) e6(i+6) e7(i+6) e6(i+7) e7(i+7)
+								    //
+        simde__m256i tmp8 = simde_mm256_unpacklo_epi64(tmp0, tmp1);   // e0(i+2) e1(i+2) e2(i+2) e3(i+2) e0(i+6) e1(i+6) e2(i+6) e3(i+6) 
+        simde__m256i tmp9 = simde_mm256_unpacklo_epi64(tmp2, tmp3);   // e4(i+2) e5(i+2) e6(i+2) e7(i+2) e4(i+6) e5(i+6) e6(i+6) e7(i+6) 
+								 
+        simde_mm256_storeu_si256(f256++,_mm256_permute2x128_si256(tmp8,tmp9,0x20)); // e0(i+2) e1(i+2) e2(i+2) e3(i+2) e4(i+2) e5(i+2) e6(i+2) e7(i+2) 
+        simde__m256i tmp10 = simde_mm256_unpackhi_epi64(tmp0, tmp1);   // e0(i+3) e1(i+3) e2(i+3) e3(i+3) e0(i+7) e1(i+7) e2(i+7) e3(i+7) 
+        simde__m256i tmp11 = simde_mm256_unpackhi_epi64(tmp2, tmp3);   // e4(i+3) e5(i+3) e6(i+3) e7(i+3) e4(i+7) e5(i+7) e6(i+7) e7(i+7) 
+        simde_mm256_storeu_si256(f256++,_mm256_permute2x128_si256(tmp10,tmp11,0x20)); // e0(i+3) e1(i+3) e2(i+3) e3(i+3) e4(i+3) e5(i+3) e6(i+3) e7(i+3) 
+        simde_mm256_storeu_si256(f256++,_mm256_permute2x128_si256(tmp4,tmp5,0x31)); // e0(i+4) e1(i+4) e2(i+4) e3(i+4) e4(i+4) e5(i+4) e6(i+4) e7(i+4) 
+        simde_mm256_storeu_si256(f256++,_mm256_permute2x128_si256(tmp6,tmp7,0x31)); // e0(i+5) e1(i+5) e2(i+5) e3(i+5) e4(i+5) e5(i+5) e6(i+5) e7(i+5) 
+        simde_mm256_storeu_si256(f256++,_mm256_permute2x128_si256(tmp8,tmp9,0x31)); // e0(i+6) e1(i+6) e2(i+6) e3(i+6) e4(i+6) e5(i+6) e6(i+6) e7(i+6) 
+        simde_mm256_storeu_si256(f256++,_mm256_permute2x128_si256(tmp10,tmp11,0x31)); // e0(i+7) e1(i+7) e2(i+7) e3(i+7) e4(i+7) e5(i+7) e6(i+7) e7(i+7) 
+      }
+      e0 = (uint32_t *)e0_256;
+      e1 = (uint32_t *)e1_256;
+      e2 = (uint32_t *)e2_256;
+      e3 = (uint32_t *)e3_256;
+      e4 = (uint32_t *)e4_256;
+      e5 = (uint32_t *)e5_256;
+      e6 = (uint32_t *)e6_256;
+      e7 = (uint32_t *)e7_256;
+      f = (uint32_t *)f256;
+#endif
+      simde__m128i *e0_128 = (simde__m128i *)e0;
+      simde__m128i *e1_128 = (simde__m128i *)e1;
+      simde__m128i *e2_128 = (simde__m128i *)e2;
+      simde__m128i *e3_128 = (simde__m128i *)e3;
+      simde__m128i *e4_128 = (simde__m128i *)e4;
+      simde__m128i *e5_128 = (simde__m128i *)e5;
+      simde__m128i *e6_128 = (simde__m128i *)e6;
+      simde__m128i *e7_128 = (simde__m128i *)e7;
+      simde__m128i *f128   = (simde__m128i *)f;
+
+      for (; i < (EQm & ~3); i += 4) {
+        simde__m128i e0j = simde_mm_loadu_si128(e0_128++);
+        simde__m128i e1j = simde_mm_loadu_si128(e1_128++);
+        simde__m128i e2j = simde_mm_loadu_si128(e2_128++);
+        simde__m128i e3j = simde_mm_loadu_si128(e3_128++);
+        simde__m128i e4j = simde_mm_loadu_si128(e4_128++);
+        simde__m128i e5j = simde_mm_loadu_si128(e5_128++);
+        simde__m128i e6j = simde_mm_loadu_si128(e6_128++);
+        simde__m128i e7j = simde_mm_loadu_si128(e7_128++);
+
+        simde__m128i tmp0 = simde_mm_unpacklo_epi32(e0j, e1j);   // e0(i) e1(i) e0(i+1) e1(i+1) 
+        simde__m128i tmp1 = simde_mm_unpacklo_epi32(e2j, e3j);   // e2(i) e3(i) e2(i+1) e3(i+1) 
+        simde__m128i tmp2 = simde_mm_unpacklo_epi32(e4j, e5j);   // e4(i) e5(i) e4(i+1) e5(i+1) 
+        simde__m128i tmp3 = simde_mm_unpacklo_epi32(e6j, e7j);   // e6(i) e7(i) e6(i+1) e7(i+1)
+								 
+        simde_mm_storeu_si128(f128++,simde_mm_unpacklo_epi64(tmp0, tmp1)); // e0(i) e1(i) e2(i) e3(i) 
+        simde_mm_storeu_si128(f128++,simde_mm_unpacklo_epi64(tmp2, tmp3)); // e4(i) e5(i) e6(i) e7(i)
+	simde_mm_storeu_si128(f128++,simde_mm_unpackhi_epi64(tmp0, tmp1));  // e0(i+1) e1(i+1) e2(i+1) e3(i+1) 
+        simde_mm_storeu_si128(f128++,simde_mm_unpackhi_epi64(tmp2, tmp3)); // e4(i+1) e5(i+1) e6(i+1) e7(i+1)
+									   //
+        tmp0 = simde_mm_unpackhi_epi32(e0j, e1j);   // e0(i+2) e1(i+2) e0(i+3) e1(i+3) 
+        tmp1 = simde_mm_unpackhi_epi32(e2j, e3j);   // e2(i+2) e3(i+2) e2(i+3) e3(i+3) 
+        tmp2 = simde_mm_unpackhi_epi32(e4j, e5j);   // e4(i+2) e5(i+2) e4(i+3) e5(i+3) 
+        tmp3 = simde_mm_unpackhi_epi32(e6j, e7j);   // e6(i+2) e7(i+2) e6(i+3) e7(i+3)
+								 
+        simde_mm_storeu_si128(f128++,simde_mm_unpacklo_epi64(tmp0, tmp1)); // e0(i+2) e1(i+2) e2(i+2) e3(i+2) 
+        simde_mm_storeu_si128(f128++,simde_mm_unpacklo_epi64(tmp2, tmp3)); // e4(i+2) e5(i+2) e6(i+2) e7(i+2)
+	simde_mm_storeu_si128(f128++,simde_mm_unpackhi_epi64(tmp0, tmp1)); // e0(i+3) e1(i+3) e2(i+3) e3(i+3) 
+        simde_mm_storeu_si128(f128++,simde_mm_unpackhi_epi64(tmp2, tmp3)); //e4(i+3) e5(i+3) e6(i+3) e7(i+3)
+      }
+      e0 = (uint32_t *)e0_128;
+      e1 = (uint32_t *)e1_128;
+      e2 = (uint32_t *)e2_128;
+      e3 = (uint32_t *)e3_128;
+      e4 = (uint32_t *)e4_128;
+      e5 = (uint32_t *)e5_128;
+      e6 = (uint32_t *)e6_128;
+      e7 = (uint32_t *)e7_128;
+      f = (uint32_t *)f128;
+      
+      for (; i < EQm; i++) {
+        *f++ = *e0++;
+        *f++ = *e1++;
+        *f++ = *e2++;
+        *f++ = *e3++;
+        *f++ = *e4++;
+        *f++ = *e5++;
+        *f++ = *e6++;
+        *f++ = *e7++;
+      }
+    } break;
+    default:
+      AssertFatal(false, "Should be here!\n");
+  }
+}
 void nr_deinterleaving_ldpc(uint32_t E, uint8_t Qm, int16_t *e, int16_t *f)
 {
   switch (Qm) {
@@ -610,6 +831,103 @@ int nr_rate_matching_ldpc(uint32_t Tbslbrm,
   return 0;
 }
 
+int nr_rate_matching_ldpc32(uint32_t Tbslbrm,
+                            uint8_t BG,
+                            uint16_t Z,
+                            uint32_t *d,
+                            uint32_t *e,
+                            uint8_t C,
+                            uint32_t F,
+                            uint32_t Foffset,
+                            uint8_t rvidx,
+                            uint32_t E)
+{
+  if (C == 0) {
+    LOG_E(PHY, "nr_rate_matching: invalid parameter C %d\n", C);
+    return -1;
+  }
+
+  //Bit selection
+  uint32_t N = (BG == 1) ? (66 * Z) : (50 * Z);
+  uint32_t Ncb;
+  if (Tbslbrm == 0)
+    Ncb = N;
+  else {
+    uint32_t Nref = 3 * Tbslbrm / (2 * C); //R_LBRM = 2/3
+    Ncb = min(N, Nref);
+  }
+
+  uint32_t ind = (index_k0[BG - 1][rvidx] * Ncb / N) * Z;
+
+#ifdef RM_DEBUG
+  printf("nr_rate_matching_ldpc: E %u, F %u, Foffset %u, k0 %u, Ncb %u, rvidx %d, Tbslbrm %u\n",
+         E,
+         F,
+         Foffset,
+         ind,
+         Ncb,
+         rvidx,
+         Tbslbrm);
+#endif
+
+  if (Foffset > E) {
+    LOG_E(PHY,
+          "nr_rate_matching: invalid parameters (Foffset %d > E %d) F %d, k0 %d, Ncb %d, rvidx %d, Tbslbrm %d\n",
+          Foffset,
+          E,
+          F,
+          ind,
+          Ncb,
+          rvidx,
+          Tbslbrm);
+    return -1;
+  }
+  if (Foffset > Ncb) {
+    LOG_E(PHY, "nr_rate_matching: invalid parameters (Foffset %d > Ncb %d)\n", Foffset, Ncb);
+    return -1;
+  }
+
+  if (ind >= Foffset && ind < (F + Foffset))
+    ind = F + Foffset;
+
+  uint32_t k = 0;
+  if (ind < Foffset) { // case where we have some bits before the filler and the rest after
+    memcpy((void *)e, (void *)(d + ind), (Foffset - ind)<<2);
+
+    if (E + F <= Ncb - ind) { // E+F doesn't contain all coded bits
+      memcpy((void *)(e + Foffset - ind), (void *)(d + Foffset + F), (E - Foffset + ind)<<2);
+      k = E;
+    } else {
+      memcpy((void *)(e + Foffset - ind), (void *)(d + Foffset + F), (Ncb - Foffset - F)<<2);
+      k = Ncb - F - ind;
+    }
+  } else {
+    if (E <= Ncb - ind) { // E+F doesn't contain all coded bits
+      memcpy((void *)(e), (void *)(d + ind), E<<2);
+      k = E;
+    } else {
+      memcpy((void *)(e), (void *)(d + ind), (Ncb - ind)<<2);
+      k = Ncb - ind;
+    }
+  }
+
+  while (k < E) { // case where we do repetitions (low mcs)
+    for (ind = 0; (ind < Ncb) && (k < E); ind++) {
+#ifdef RM_DEBUG
+      printf("RM_TX k%u Ind: %u (%d)\n", k, ind, d[ind]);
+#endif
+
+      if (ind == Foffset)
+        ind = F + Foffset; // skip filler bits
+
+      e[k++] = d[ind];
+
+    }
+  }
+
+  return 0;
+}
+
 int nr_rate_matching_ldpc_rx(uint32_t Tbslbrm,
                              uint8_t BG,
                              uint16_t Z,

openair1/PHY/CODING/nrLDPC_coding/nrLDPC_coding_segment/nr_rate_matching.h

@@ -33,6 +33,15 @@
  */
 void nr_interleaving_ldpc(uint32_t E, uint8_t Qm, uint8_t *e, uint8_t *f);
 
+/**
+ * \brief interleave a code segment after encoding and rate matching (32bit)
+ * \param E size of the code segment in bits
+ * \param Qm modulation order
+ * \param e input rate matched segment
+ * \param f output interleaved segment
+ */
+
+void nr_interleaving_ldpc32(uint32_t E, uint8_t Qm, uint32_t *e, uint32_t *f);
 /**
  * \brief deinterleave a code segment before RX rate matching and decoding
  * \param E size of the code segment in bits
@@ -66,6 +75,30 @@ int nr_rate_matching_ldpc(uint32_t Tbslbrm,
                           uint8_t rvidx,
                           uint32_t E);
 
+/**
+ * \brief rate match a code segment after encoding (32bit)
+ * \Tbslbrm Transport Block size LBRM
+ * \param BG LDPC base graph number
+ * \param Z segment lifting size
+ * \param d input encoded segment
+ * \param e output rate matched segment
+ * \param C number of segments in the Transport Block
+ * \param F number of filler bits in the segment
+ * \param Foffset offset of the filler bits in the segment
+ * \param rvidx redundancy version index
+ * \param E size of the code segment in bits
+ */
+int nr_rate_matching_ldpc32(uint32_t Tbslbrm,
+                            uint8_t BG,
+                            uint16_t Z,
+                            uint32_t *d,
+                            uint32_t *e,
+                            uint8_t C,
+                            uint32_t F,
+                            uint32_t Foffset,
+                            uint8_t rvidx,
+                            uint32_t E);
+
 /**
  * \brief rate match a code segment before decoding
  * \Tbslbrm Transport Block size LBRM

openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_decoder.c

@@ -161,6 +161,7 @@ static inline uint32_t nrLDPC_decoder_core(int8_t* p_llr,
 
 int32_t LDPCinit()
 {
+  cuda_support_init();
   return 0;
 }
 

openair1/PHY/CODING/nrLDPC_defs.h

@@ -55,9 +55,11 @@ typedef struct {
   unsigned int first_seg; // optim8segmulti
   unsigned char gen_code; // orig
   time_stats_t *tinput;
+  time_stats_t *tinput_memcpy;
   time_stats_t *tprep;
   time_stats_t *tparity;
   time_stats_t *toutput;
+  time_stats_t *tconcat;
   /// Size in bits of the code segments
   uint32_t K;
   /// Number of lifting sizes to fit the payload
@@ -91,5 +93,7 @@ typedef int32_t(LDPC_decoderfunc_t)(t_nrLDPC_dec_params *p_decParams,
                                     t_nrLDPC_time_stats *time_stats,
                                     decode_abort_t *ab);
 typedef int32_t(LDPC_encoderfunc_t)(uint8_t **, uint8_t *, encoder_implemparams_t *);
+typedef uint32_t**(LDPC_encoderfunc32_t)(uint8_t **, encoder_implemparams_t *);
 
+void cuda_support_init(void);
 #endif

openair1/PHY/CODING/nrLDPC_encoder/ldpc_BG1_Zc384_32bit.cu

@@ -0,0 +1,214 @@
+#include <stdio.h>
+#include <stdint.h>
+#include <cuda_runtime.h>
+// generated code for Zc=384, byte encoding
+__global__ void ldpc_BG1_Zc384_worker(uint32_t *c[4],uint32_t *d[4]) {
+  uint32_t *c32=c[blockIdx.x];
+  uint32_t *d32=d[blockIdx.x]+8448-2*384;
+
+  int i2 = threadIdx.x;
+  int i1 = blockIdx.y;
+  if (i2 < 384) {
+    c32+=i2;
+    d32+=i2;
+    switch(i1) {
+    case 0:
+      d32[0]=c32[307]^c32[76]^c32[205]^c32[276]^c32[787]^c32[1018]^c32[855]^c32[1586]^c32[1612]^c32[1864]^c32[2673]^c32[2377]^c32[2304]^c32[3360]^c32[3404]^c32[3347]^c32[4021]^c32[3984]^c32[4096]^c32[4824]^c32[4769]^c32[4807]^c32[5707]^c32[5643]^c32[5529]^c32[6475]^c32[6304]^c32[6200]^c32[7229]^c32[7090]^c32[6975]^c32[7968]^c32[7809]^c32[7812]^c32[8557]^c32[8743]^c32[8753]^c32[9233]^c32[9558]^c32[9447]^c32[10341]^c32[10184]^c32[10325]^c32[10969]^c32[10840]^c32[10964]^c32[11735]^c32[11619]^c32[11573]^c32[12394]^c32[12642]^c32[12592]^c32[13170]^c32[13187]^c32[13356]^c32[14066]^c32[14064]^c32[14095]^c32[14772]^c32[14923]^c32[14797]^c32[15690]^c32[15373]^c32[15399]^c32[16474]^c32[16240]^c32[16485];
+
+       break;
+    case 1:
+      d32[384]=c32[307]^c32[308]^c32[77]^c32[206]^c32[277]^c32[787]^c32[788]^c32[1019]^c32[856]^c32[1586]^c32[1587]^c32[1613]^c32[1865]^c32[2673]^c32[2674]^c32[2378]^c32[2305]^c32[3361]^c32[3405]^c32[3348]^c32[4021]^c32[4022]^c32[3985]^c32[4097]^c32[4824]^c32[4825]^c32[4770]^c32[4808]^c32[5708]^c32[5644]^c32[5530]^c32[6476]^c32[6305]^c32[6201]^c32[7229]^c32[7230]^c32[7091]^c32[6976]^c32[7968]^c32[7969]^c32[7810]^c32[7813]^c32[8557]^c32[8558]^c32[8744]^c32[8754]^c32[9233]^c32[9234]^c32[9559]^c32[9448]^c32[10341]^c32[10342]^c32[10185]^c32[10326]^c32[10970]^c32[10841]^c32[10965]^c32[11735]^c32[11736]^c32[11620]^c32[11574]^c32[12394]^c32[12395]^c32[12643]^c32[12593]^c32[13171]^c32[13188]^c32[13357]^c32[14066]^c32[14067]^c32[14065]^c32[14096]^c32[14772]^c32[14773]^c32[14924]^c32[14798]^c32[15690]^c32[15691]^c32[15374]^c32[15400]^c32[16474]^c32[16475]^c32[16241]^c32[16486];
+
+       break;
+    case 2:
+      d32[768]=c32[308]^c32[77]^c32[205]^c32[206]^c32[276]^c32[277]^c32[788]^c32[1018]^c32[1019]^c32[855]^c32[856]^c32[1587]^c32[1613]^c32[1864]^c32[1865]^c32[2674]^c32[2378]^c32[2304]^c32[2305]^c32[3361]^c32[3404]^c32[3405]^c32[3347]^c32[3348]^c32[4022]^c32[3985]^c32[4096]^c32[4097]^c32[4825]^c32[4769]^c32[4770]^c32[4807]^c32[4808]^c32[5708]^c32[5643]^c32[5644]^c32[5529]^c32[5530]^c32[6476]^c32[6304]^c32[6305]^c32[6200]^c32[6201]^c32[7230]^c32[7091]^c32[6975]^c32[6976]^c32[7969]^c32[7809]^c32[7810]^c32[7812]^c32[7813]^c32[8558]^c32[8744]^c32[8753]^c32[8754]^c32[9234]^c32[9559]^c32[9447]^c32[9448]^c32[10342]^c32[10184]^c32[10185]^c32[10325]^c32[10326]^c32[10970]^c32[10840]^c32[10841]^c32[10964]^c32[10965]^c32[11736]^c32[11620]^c32[11573]^c32[11574]^c32[12395]^c32[12643]^c32[12592]^c32[12593]^c32[13171]^c32[13187]^c32[13188]^c32[13356]^c32[13357]^c32[14067]^c32[14064]^c32[14065]^c32[14095]^c32[14096]^c32[14773]^c32[14924]^c32[14797]^c32[14798]^c32[15691]^c32[15373]^c32[15374]^c32[15399]^c32[15400]^c32[16475]^c32[16241]^c32[16485]^c32[16486];
+
+       break;
+    case 3:
+      d32[1152]=c32[308]^c32[77]^c32[206]^c32[276]^c32[277]^c32[788]^c32[1019]^c32[855]^c32[856]^c32[1587]^c32[1613]^c32[1865]^c32[2674]^c32[2378]^c32[2304]^c32[2305]^c32[3361]^c32[3405]^c32[3347]^c32[3348]^c32[4022]^c32[3985]^c32[4097]^c32[4825]^c32[4770]^c32[4807]^c32[4808]^c32[5708]^c32[5644]^c32[5529]^c32[5530]^c32[6476]^c32[6305]^c32[6200]^c32[6201]^c32[7230]^c32[7091]^c32[6976]^c32[7969]^c32[7810]^c32[7812]^c32[7813]^c32[8558]^c32[8744]^c32[8753]^c32[8754]^c32[9234]^c32[9559]^c32[9447]^c32[9448]^c32[10342]^c32[10185]^c32[10325]^c32[10326]^c32[10970]^c32[10841]^c32[10964]^c32[10965]^c32[11736]^c32[11620]^c32[11574]^c32[12395]^c32[12643]^c32[12592]^c32[12593]^c32[13171]^c32[13188]^c32[13356]^c32[13357]^c32[14067]^c32[14065]^c32[14095]^c32[14096]^c32[14773]^c32[14924]^c32[14798]^c32[15691]^c32[15374]^c32[15399]^c32[15400]^c32[16475]^c32[16241]^c32[16485]^c32[16486];
+
+       break;
+    case 4:
+      d32[1536]=c32[332]^c32[949];
+
+       break;
+    case 5:
+      d32[1920]=c32[80]^c32[233]^c32[362]^c32[49]^c32[195]^c32[944]^c32[791]^c32[1012]^c32[782]^c32[1743]^c32[1769]^c32[1637]^c32[2446]^c32[2534]^c32[2461]^c32[2419]^c32[3133]^c32[3177]^c32[3120]^c32[4178]^c32[4141]^c32[3869]^c32[4981]^c32[4926]^c32[4964]^c32[5480]^c32[5416]^c32[5686]^c32[6248]^c32[6461]^c32[6357]^c32[7002]^c32[7247]^c32[7132]^c32[7741]^c32[7966]^c32[7969]^c32[8714]^c32[8516]^c32[8526]^c32[9390]^c32[9331]^c32[9220]^c32[9382]^c32[10114]^c32[10341]^c32[10098]^c32[11126]^c32[10997]^c32[11121]^c32[11892]^c32[11776]^c32[11730]^c32[12551]^c32[12415]^c32[12365]^c32[12529]^c32[13327]^c32[13344]^c32[13129]^c32[13839]^c32[13837]^c32[13868]^c32[14929]^c32[14696]^c32[14954]^c32[15463]^c32[15530]^c32[15556]^c32[16247]^c32[16397]^c32[16258]^c32[16179];
+
+       break;
+    case 6:
+      d32[2304]=c32[278]^c32[4865]^c32[7681]^c32[8799]^c32[10076]^c32[13309]^c32[13842]^c32[15585];
+
+       break;
+    case 7:
+      d32[2688]=c32[9]^c32[830]^c32[3388]^c32[5709]^c32[6434]^c32[10866];
+
+       break;
+    case 8:
+      d32[3072]=c32[374]^c32[94]^c32[143]^c32[247]^c32[272]^c32[375]^c32[376]^c32[343]^c32[62]^c32[63]^c32[307]^c32[854]^c32[958]^c32[1085]^c32[804]^c32[805]^c32[922]^c32[1025]^c32[1026]^c32[947]^c32[1653]^c32[1757]^c32[1679]^c32[1783]^c32[1547]^c32[1650]^c32[1651]^c32[2356]^c32[2460]^c32[2444]^c32[2548]^c32[2371]^c32[2474]^c32[2475]^c32[2469]^c32[3427]^c32[3147]^c32[3087]^c32[3190]^c32[3191]^c32[3414]^c32[3133]^c32[3134]^c32[4088]^c32[4192]^c32[4051]^c32[4155]^c32[4163]^c32[3882]^c32[3883]^c32[4891]^c32[4611]^c32[4836]^c32[4939]^c32[4940]^c32[4874]^c32[4977]^c32[4978]^c32[5390]^c32[5494]^c32[5710]^c32[5429]^c32[5430]^c32[5596]^c32[5699]^c32[5700]^c32[6158]^c32[6262]^c32[6371]^c32[6474]^c32[6475]^c32[6267]^c32[6370]^c32[6371]^c32[6912]^c32[7016]^c32[7157]^c32[7261]^c32[7042]^c32[7145]^c32[7146]^c32[8035]^c32[7755]^c32[7876]^c32[7979]^c32[7980]^c32[7879]^c32[7982]^c32[7983]^c32[8624]^c32[8728]^c32[8810]^c32[8530]^c32[8820]^c32[8539]^c32[8540]^c32[9300]^c32[9404]^c32[9241]^c32[9345]^c32[9514]^c32[9233]^c32[9234]^c32[9234]^c32[10024]^c32[10128]^c32[10251]^c32[10354]^c32[10355]^c32[10008]^c32[10111]^c32[10112]^c32[11036]^c32[10756]^c32[10907]^c32[11010]^c32[11011]^c32[11031]^c32[11134]^c32[11135]^c32[11802]^c32[11522]^c32[11686]^c32[11790]^c32[11640]^c32[11743]^c32[11744]^c32[12461]^c32[12565]^c32[12325]^c32[12429]^c32[12659]^c32[12378]^c32[12379]^c32[12327]^c32[13237]^c32[13341]^c32[13254]^c32[13357]^c32[13358]^c32[13423]^c32[13142]^c32[13143]^c32[14133]^c32[13853]^c32[14131]^c32[13850]^c32[13851]^c32[14162]^c32[13881]^c32[13882]^c32[14839]^c32[14943]^c32[14606]^c32[14710]^c32[14864]^c32[14967]^c32[14968]^c32[14816]^c32[15373]^c32[15477]^c32[15440]^c32[15543]^c32[15544]^c32[15466]^c32[15569]^c32[15570]^c32[16157]^c32[16261]^c32[16307]^c32[16411]^c32[16168]^c32[16271]^c32[16272]^c32[16496];
+
+       break;
+    case 9:
+      d32[3456]=c32[366]^c32[1000]^c32[8001]^c32[8581]^c32[10041]^c32[13359]^c32[13887]^c32[15442];
+
+       break;
+    case 10:
+      d32[3840]=c32[869]^c32[1875]^c32[3346]^c32[5487]^c32[6527]^c32[11106];
+
+       break;
+    case 11:
+      d32[4224]=c32[257]^c32[38]^c32[39]^c32[26]^c32[192]^c32[155]^c32[321]^c32[226]^c32[8]^c32[48]^c32[1121]^c32[902]^c32[903]^c32[968]^c32[1134]^c32[805]^c32[971]^c32[870]^c32[1536]^c32[1701]^c32[1702]^c32[1562]^c32[1728]^c32[1814]^c32[1596]^c32[2623]^c32[2404]^c32[2405]^c32[2327]^c32[2493]^c32[2638]^c32[2420]^c32[3310]^c32[3092]^c32[3354]^c32[3136]^c32[3297]^c32[3079]^c32[3971]^c32[4136]^c32[4137]^c32[3934]^c32[4100]^c32[4046]^c32[4212]^c32[4774]^c32[4939]^c32[4940]^c32[4719]^c32[4885]^c32[4757]^c32[4923]^c32[5657]^c32[5439]^c32[5593]^c32[5759]^c32[5479]^c32[5645]^c32[6425]^c32[6207]^c32[6254]^c32[6420]^c32[6150]^c32[6316]^c32[7179]^c32[6960]^c32[6961]^c32[7040]^c32[7206]^c32[6925]^c32[7091]^c32[7918]^c32[7699]^c32[7700]^c32[7759]^c32[7925]^c32[7762]^c32[7928]^c32[8507]^c32[8672]^c32[8673]^c32[8693]^c32[8475]^c32[8703]^c32[8485]^c32[9567]^c32[9348]^c32[9349]^c32[9508]^c32[9290]^c32[9397]^c32[9563]^c32[9224]^c32[10291]^c32[10072]^c32[10073]^c32[10134]^c32[10300]^c32[10275]^c32[10057]^c32[10919]^c32[11085]^c32[10790]^c32[10956]^c32[10914]^c32[11080]^c32[11685]^c32[11850]^c32[11851]^c32[11569]^c32[11735]^c32[11523]^c32[11689]^c32[12344]^c32[12509]^c32[12510]^c32[12592]^c32[12374]^c32[12542]^c32[12324]^c32[12335]^c32[13120]^c32[13286]^c32[13137]^c32[13303]^c32[13306]^c32[13088]^c32[14016]^c32[14181]^c32[14182]^c32[14014]^c32[14180]^c32[14045]^c32[13827]^c32[14722]^c32[14887]^c32[14888]^c32[14873]^c32[14655]^c32[14747]^c32[14913]^c32[15640]^c32[15421]^c32[15422]^c32[15707]^c32[15489]^c32[15733]^c32[15515]^c32[16424]^c32[16205]^c32[16206]^c32[16190]^c32[16356]^c32[16435]^c32[16217]^c32[16316];
+
+       break;
+    case 12:
+      d32[4608]=c32[77]^c32[954]^c32[7854]^c32[8680]^c32[10034]^c32[13898];
+
+       break;
+    case 13:
+      d32[4992]=c32[293]^c32[294]^c32[63]^c32[192]^c32[263]^c32[313]^c32[773]^c32[774]^c32[1005]^c32[842]^c32[1572]^c32[1573]^c32[1599]^c32[1851]^c32[2659]^c32[2660]^c32[2364]^c32[2675]^c32[2481]^c32[3347]^c32[3391]^c32[3334]^c32[4007]^c32[4008]^c32[3971]^c32[4083]^c32[4810]^c32[4811]^c32[4756]^c32[4794]^c32[5694]^c32[5630]^c32[5516]^c32[5642]^c32[6462]^c32[6291]^c32[6187]^c32[7215]^c32[7216]^c32[7077]^c32[6962]^c32[7954]^c32[7955]^c32[7796]^c32[7799]^c32[8543]^c32[8544]^c32[8730]^c32[8740]^c32[9219]^c32[9220]^c32[9545]^c32[9434]^c32[10327]^c32[10328]^c32[10171]^c32[10312]^c32[10956]^c32[10827]^c32[10951]^c32[11721]^c32[11722]^c32[11606]^c32[11560]^c32[12380]^c32[12381]^c32[12629]^c32[12579]^c32[13157]^c32[13174]^c32[13343]^c32[14052]^c32[14053]^c32[14051]^c32[14082]^c32[14758]^c32[14759]^c32[14910]^c32[14784]^c32[15676]^c32[15677]^c32[15360]^c32[15386]^c32[15475]^c32[16460]^c32[16461]^c32[16227]^c32[16472];
+
+       break;
+    case 14:
+      d32[5376]=c32[142]^c32[9464]^c32[11657]^c32[12377]^c32[13403]^c32[16140];
+
+       break;
+    case 15:
+      d32[5760]=c32[193]^c32[346]^c32[91]^c32[161]^c32[162]^c32[241]^c32[1057]^c32[904]^c32[1124]^c32[1125]^c32[770]^c32[1856]^c32[1882]^c32[1750]^c32[2559]^c32[2647]^c32[2573]^c32[2574]^c32[3246]^c32[3290]^c32[3232]^c32[3233]^c32[3907]^c32[3870]^c32[3982]^c32[4710]^c32[4655]^c32[4692]^c32[4693]^c32[5593]^c32[5529]^c32[5414]^c32[5415]^c32[6361]^c32[6190]^c32[6469]^c32[6470]^c32[7115]^c32[6976]^c32[7245]^c32[7854]^c32[7695]^c32[7697]^c32[7698]^c32[7890]^c32[8827]^c32[8629]^c32[8638]^c32[8639]^c32[9503]^c32[9444]^c32[9332]^c32[9333]^c32[10227]^c32[10070]^c32[10210]^c32[10211]^c32[10302]^c32[10855]^c32[11110]^c32[10849]^c32[10850]^c32[11621]^c32[11889]^c32[11843]^c32[12664]^c32[12528]^c32[12477]^c32[12478]^c32[13056]^c32[13073]^c32[13241]^c32[13242]^c32[13952]^c32[13950]^c32[13980]^c32[13981]^c32[13879]^c32[14658]^c32[14809]^c32[14683]^c32[15576]^c32[15643]^c32[15668]^c32[15669]^c32[16360]^c32[16510]^c32[16370]^c32[16371];
+
+       break;
+    case 16:
+      d32[6144]=c32[364]^c32[133]^c32[262]^c32[333]^c32[844]^c32[1075]^c32[912]^c32[781]^c32[1643]^c32[1669]^c32[1537]^c32[2346]^c32[2434]^c32[2361]^c32[2642]^c32[3417]^c32[3077]^c32[3404]^c32[4078]^c32[4041]^c32[4153]^c32[4881]^c32[4826]^c32[4864]^c32[5380]^c32[5700]^c32[5586]^c32[6148]^c32[6361]^c32[6257]^c32[7286]^c32[7147]^c32[7032]^c32[8025]^c32[7866]^c32[7869]^c32[8614]^c32[8800]^c32[8810]^c32[8505]^c32[9290]^c32[9231]^c32[9504]^c32[10014]^c32[10241]^c32[9998]^c32[11026]^c32[10897]^c32[11021]^c32[11792]^c32[11676]^c32[11630]^c32[12451]^c32[12315]^c32[12649]^c32[13227]^c32[13244]^c32[13413]^c32[14123]^c32[14121]^c32[14152]^c32[14829]^c32[14596]^c32[14854]^c32[15363]^c32[15430]^c32[15456]^c32[15649]^c32[16147]^c32[16297]^c32[16158];
+
+       break;
+    case 17:
+      d32[6528]=c32[260]^c32[11055]^c32[12369]^c32[13414]^c32[16503];
+
+       break;
+    case 18:
+      d32[6912]=c32[898]^c32[9379]^c32[10264]^c32[13956]^c32[14596];
+
+       break;
+    case 19:
+      d32[7296]=c32[145]^c32[981]^c32[5720]^c32[6386]^c32[7877];
+
+       break;
+    case 20:
+      d32[7680]=c32[366]^c32[135]^c32[264]^c32[335]^c32[187]^c32[846]^c32[1077]^c32[914]^c32[1645]^c32[1671]^c32[1539]^c32[2348]^c32[2436]^c32[2363]^c32[2510]^c32[3419]^c32[3079]^c32[3406]^c32[4080]^c32[4043]^c32[4155]^c32[4883]^c32[4828]^c32[4866]^c32[5382]^c32[5702]^c32[5588]^c32[6150]^c32[6363]^c32[6259]^c32[7288]^c32[7149]^c32[7034]^c32[7176]^c32[8027]^c32[7868]^c32[7871]^c32[8616]^c32[8802]^c32[8812]^c32[8789]^c32[9292]^c32[9233]^c32[9506]^c32[10016]^c32[10243]^c32[10000]^c32[11028]^c32[10899]^c32[11023]^c32[11794]^c32[11678]^c32[11632]^c32[12453]^c32[12317]^c32[12651]^c32[13229]^c32[13246]^c32[13415]^c32[14125]^c32[14123]^c32[14154]^c32[14831]^c32[14598]^c32[14856]^c32[15365]^c32[15432]^c32[15458]^c32[16149]^c32[16299]^c32[16160];
+
+       break;
+    case 21:
+      d32[8064]=c32[973]^c32[3942]^c32[12616]^c32[15573]^c32[16225];
+
+       break;
+    case 22:
+      d32[8448]=c32[30]^c32[9227]^c32[10217]^c32[13078];
+
+       break;
+    case 23:
+      d32[8832]=c32[792]^c32[1625]^c32[7741]^c32[13851];
+
+       break;
+    case 24:
+      d32[9216]=c32[157]^c32[310]^c32[55]^c32[126]^c32[298]^c32[1021]^c32[868]^c32[1089]^c32[1820]^c32[1846]^c32[1714]^c32[2523]^c32[2611]^c32[2538]^c32[2462]^c32[3210]^c32[3254]^c32[3197]^c32[3307]^c32[3871]^c32[4218]^c32[3946]^c32[4674]^c32[4619]^c32[4657]^c32[5557]^c32[5493]^c32[5379]^c32[6325]^c32[6154]^c32[6434]^c32[7079]^c32[6940]^c32[7209]^c32[7818]^c32[8043]^c32[8046]^c32[8791]^c32[8593]^c32[8603]^c32[8787]^c32[9467]^c32[9408]^c32[9297]^c32[10191]^c32[10034]^c32[10175]^c32[10819]^c32[11074]^c32[10814]^c32[11585]^c32[11853]^c32[11807]^c32[12628]^c32[12492]^c32[12442]^c32[13404]^c32[13421]^c32[13206]^c32[13916]^c32[13914]^c32[13945]^c32[14622]^c32[14773]^c32[14647]^c32[15540]^c32[15607]^c32[15633]^c32[16324]^c32[16474]^c32[16335];
+
+       break;
+    case 25:
+      d32[9600]=c32[840]^c32[4625]^c32[5759]^c32[11064];
+
+       break;
+    case 26:
+      d32[9984]=c32[71]^c32[1617]^c32[3148]^c32[11656];
+
+       break;
+    case 27:
+      d32[10368]=c32[962]^c32[4802]^c32[6245];
+
+       break;
+    case 28:
+      d32[10752]=c32[222]^c32[3091]^c32[14836]^c32[16402];
+
+       break;
+    case 29:
+      d32[11136]=c32[2]^c32[155]^c32[284]^c32[354]^c32[355]^c32[866]^c32[1097]^c32[933]^c32[934]^c32[1020]^c32[1665]^c32[1691]^c32[1559]^c32[2368]^c32[2456]^c32[2382]^c32[2383]^c32[3439]^c32[3099]^c32[3425]^c32[3426]^c32[4100]^c32[4063]^c32[4175]^c32[4903]^c32[4848]^c32[4885]^c32[4886]^c32[5402]^c32[5722]^c32[5607]^c32[5608]^c32[6170]^c32[6383]^c32[6278]^c32[6279]^c32[6924]^c32[7169]^c32[7054]^c32[8047]^c32[7888]^c32[7890]^c32[7891]^c32[8636]^c32[8822]^c32[8831]^c32[8448]^c32[9312]^c32[9253]^c32[9525]^c32[9526]^c32[10036]^c32[10263]^c32[10019]^c32[10020]^c32[11048]^c32[10919]^c32[11042]^c32[11043]^c32[10757]^c32[11814]^c32[11698]^c32[11652]^c32[12473]^c32[12337]^c32[12670]^c32[12671]^c32[13249]^c32[13266]^c32[13434]^c32[13435]^c32[14145]^c32[14143]^c32[14173]^c32[14174]^c32[13971]^c32[14851]^c32[14618]^c32[14876]^c32[15385]^c32[15452]^c32[15477]^c32[15478]^c32[16169]^c32[16319]^c32[16179]^c32[16180];
+
+       break;
+    case 30:
+      d32[11520]=c32[14]^c32[167]^c32[295]^c32[296]^c32[366]^c32[367]^c32[159]^c32[878]^c32[1108]^c32[1109]^c32[945]^c32[946]^c32[1677]^c32[1703]^c32[1570]^c32[1571]^c32[2380]^c32[2468]^c32[2394]^c32[2395]^c32[3451]^c32[3110]^c32[3111]^c32[3437]^c32[3438]^c32[4112]^c32[4075]^c32[4186]^c32[4187]^c32[4915]^c32[4859]^c32[4860]^c32[4897]^c32[4898]^c32[5414]^c32[5733]^c32[5734]^c32[5619]^c32[5620]^c32[6182]^c32[6394]^c32[6395]^c32[6290]^c32[6291]^c32[6936]^c32[7181]^c32[7065]^c32[7066]^c32[8059]^c32[7899]^c32[7900]^c32[7902]^c32[7903]^c32[7909]^c32[8648]^c32[8450]^c32[8459]^c32[8460]^c32[9324]^c32[9265]^c32[9537]^c32[9538]^c32[10048]^c32[10274]^c32[10275]^c32[10031]^c32[10032]^c32[10244]^c32[11060]^c32[10930]^c32[10931]^c32[11054]^c32[11055]^c32[11826]^c32[11710]^c32[11663]^c32[11664]^c32[12485]^c32[12349]^c32[12298]^c32[12299]^c32[13261]^c32[13277]^c32[13278]^c32[13062]^c32[13063]^c32[14157]^c32[14154]^c32[14155]^c32[14185]^c32[14186]^c32[14863]^c32[14630]^c32[14887]^c32[14888]^c32[15397]^c32[15463]^c32[15464]^c32[15489]^c32[15490]^c32[16181]^c32[16331]^c32[16191]^c32[16192];
+
+       break;
+    case 31:
+      d32[11904]=c32[181]^c32[180]^c32[334]^c32[333]^c32[79]^c32[78]^c32[150]^c32[148]^c32[149]^c32[1045]^c32[1044]^c32[892]^c32[891]^c32[1113]^c32[1111]^c32[1112]^c32[868]^c32[1844]^c32[1843]^c32[1870]^c32[1869]^c32[1738]^c32[1737]^c32[2547]^c32[2546]^c32[2635]^c32[2634]^c32[2562]^c32[2560]^c32[2561]^c32[3234]^c32[3233]^c32[3278]^c32[3277]^c32[3221]^c32[3219]^c32[3220]^c32[3895]^c32[3894]^c32[3858]^c32[3857]^c32[3970]^c32[3969]^c32[4698]^c32[4697]^c32[4643]^c32[4642]^c32[4681]^c32[4679]^c32[4680]^c32[5581]^c32[5580]^c32[5517]^c32[5516]^c32[5403]^c32[5401]^c32[5402]^c32[5591]^c32[6349]^c32[6348]^c32[6178]^c32[6177]^c32[6458]^c32[6456]^c32[6457]^c32[7103]^c32[7102]^c32[6964]^c32[6963]^c32[7233]^c32[7232]^c32[7842]^c32[7841]^c32[7683]^c32[7682]^c32[7686]^c32[7684]^c32[7685]^c32[8815]^c32[8814]^c32[8617]^c32[8616]^c32[8627]^c32[8625]^c32[8626]^c32[9491]^c32[9490]^c32[9432]^c32[9431]^c32[9321]^c32[9319]^c32[9320]^c32[10215]^c32[10214]^c32[10058]^c32[10057]^c32[10199]^c32[10197]^c32[10198]^c32[10843]^c32[10842]^c32[11098]^c32[11097]^c32[10838]^c32[10836]^c32[10837]^c32[11609]^c32[11608]^c32[11877]^c32[11876]^c32[11831]^c32[11830]^c32[12652]^c32[12651]^c32[12516]^c32[12515]^c32[12466]^c32[12464]^c32[12465]^c32[13428]^c32[13427]^c32[13061]^c32[13060]^c32[13230]^c32[13228]^c32[13229]^c32[13940]^c32[13939]^c32[13938]^c32[13937]^c32[13969]^c32[13967]^c32[13968]^c32[14646]^c32[14645]^c32[14797]^c32[14796]^c32[14671]^c32[14670]^c32[15564]^c32[15563]^c32[15631]^c32[15630]^c32[15657]^c32[15655]^c32[15656]^c32[16348]^c32[16347]^c32[16498]^c32[16497]^c32[16359]^c32[16357]^c32[16358];
+
+       break;
+    case 32:
+      d32[12288]=c32[211]^c32[364]^c32[108]^c32[109]^c32[179]^c32[180]^c32[102]^c32[1075]^c32[921]^c32[922]^c32[1142]^c32[1143]^c32[1874]^c32[1900]^c32[1767]^c32[1768]^c32[2577]^c32[2665]^c32[2591]^c32[2592]^c32[3264]^c32[3307]^c32[3308]^c32[3250]^c32[3251]^c32[3925]^c32[3888]^c32[3999]^c32[4000]^c32[4728]^c32[4672]^c32[4673]^c32[4710]^c32[4711]^c32[5611]^c32[5546]^c32[5547]^c32[5432]^c32[5433]^c32[6379]^c32[6207]^c32[6208]^c32[6487]^c32[6488]^c32[7133]^c32[6994]^c32[7262]^c32[7263]^c32[7872]^c32[7712]^c32[7713]^c32[7715]^c32[7716]^c32[8461]^c32[8647]^c32[8656]^c32[8657]^c32[9521]^c32[9462]^c32[9350]^c32[9351]^c32[9417]^c32[10245]^c32[10087]^c32[10088]^c32[10228]^c32[10229]^c32[10873]^c32[11127]^c32[11128]^c32[10867]^c32[10868]^c32[10927]^c32[11639]^c32[11523]^c32[11860]^c32[11861]^c32[12298]^c32[12546]^c32[12495]^c32[12496]^c32[13074]^c32[13090]^c32[13091]^c32[13259]^c32[13260]^c32[13970]^c32[13967]^c32[13968]^c32[13998]^c32[13999]^c32[14676]^c32[14827]^c32[14700]^c32[14701]^c32[15594]^c32[15660]^c32[15661]^c32[15686]^c32[15687]^c32[16378]^c32[16144]^c32[16388]^c32[16389];
+
+       break;
+    case 33:
+      d32[12672]=c32[1091]^c32[1544]^c32[8809]^c32[16233];
+
+       break;
+    case 34:
+      d32[13056]=c32[230]^c32[5524]^c32[11722]^c32[13368];
+
+       break;
+    case 35:
+      d32[13440]=c32[189]^c32[342]^c32[87]^c32[158]^c32[1053]^c32[900]^c32[1121]^c32[1088]^c32[1852]^c32[1878]^c32[1746]^c32[2555]^c32[2643]^c32[2570]^c32[3242]^c32[3286]^c32[3229]^c32[3903]^c32[3866]^c32[3978]^c32[4706]^c32[4651]^c32[4689]^c32[4943]^c32[5589]^c32[5525]^c32[5411]^c32[6357]^c32[6186]^c32[6466]^c32[7111]^c32[6972]^c32[7241]^c32[7850]^c32[7691]^c32[7694]^c32[8823]^c32[8625]^c32[8635]^c32[9499]^c32[9440]^c32[9329]^c32[9218]^c32[10223]^c32[10066]^c32[10207]^c32[10851]^c32[11106]^c32[10846]^c32[11617]^c32[11885]^c32[11839]^c32[12660]^c32[12524]^c32[12474]^c32[13436]^c32[13069]^c32[13238]^c32[13948]^c32[13946]^c32[13977]^c32[14654]^c32[14805]^c32[14679]^c32[15572]^c32[15639]^c32[15665]^c32[16356]^c32[16506]^c32[16367];
+
+       break;
+    case 36:
+      d32[13824]=c32[210]^c32[11065]^c32[11817]^c32[13845];
+
+       break;
+    case 37:
+      d32[14208]=c32[38]^c32[39]^c32[192]^c32[321]^c32[8]^c32[902]^c32[903]^c32[1134]^c32[971]^c32[1037]^c32[1701]^c32[1702]^c32[1728]^c32[1596]^c32[2404]^c32[2405]^c32[2493]^c32[2420]^c32[3092]^c32[3136]^c32[3079]^c32[4136]^c32[4137]^c32[4100]^c32[4212]^c32[4939]^c32[4940]^c32[4885]^c32[4923]^c32[5439]^c32[5759]^c32[5645]^c32[6207]^c32[6420]^c32[6316]^c32[6960]^c32[6961]^c32[7206]^c32[7091]^c32[7699]^c32[7700]^c32[7925]^c32[7928]^c32[8672]^c32[8673]^c32[8475]^c32[8485]^c32[9348]^c32[9349]^c32[9290]^c32[9563]^c32[10072]^c32[10073]^c32[10300]^c32[10057]^c32[10066]^c32[11085]^c32[10956]^c32[11080]^c32[11850]^c32[11851]^c32[11735]^c32[11689]^c32[12509]^c32[12510]^c32[12374]^c32[12324]^c32[13286]^c32[13303]^c32[13088]^c32[14181]^c32[14182]^c32[14180]^c32[13827]^c32[14887]^c32[14888]^c32[14655]^c32[14913]^c32[15421]^c32[15422]^c32[15489]^c32[15515]^c32[16205]^c32[16206]^c32[16356]^c32[16217];
+
+       break;
+    case 38:
+      d32[14592]=c32[185]^c32[7089]^c32[7969]^c32[9430];
+
+       break;
+    case 39:
+      d32[14976]=c32[1026]^c32[2397]^c32[5722]^c32[14889];
+
+       break;
+    case 40:
+      d32[15360]=c32[175]^c32[6181]^c32[13368];
+
+       break;
+    case 41:
+      d32[15744]=c32[820]^c32[2618]^c32[7051]^c32[14112];
+
+       break;
+    case 42:
+      d32[16128]=c32[142]^c32[295]^c32[39]^c32[40]^c32[110]^c32[111]^c32[113]^c32[1006]^c32[852]^c32[853]^c32[1073]^c32[1074]^c32[1805]^c32[1831]^c32[1698]^c32[1699]^c32[2508]^c32[2596]^c32[2522]^c32[2523]^c32[3195]^c32[3238]^c32[3239]^c32[3181]^c32[3182]^c32[3086]^c32[3856]^c32[4203]^c32[3930]^c32[3931]^c32[4659]^c32[4987]^c32[4988]^c32[4641]^c32[4642]^c32[5542]^c32[5477]^c32[5478]^c32[5747]^c32[5748]^c32[6310]^c32[6522]^c32[6523]^c32[6418]^c32[6419]^c32[7064]^c32[6925]^c32[7193]^c32[7194]^c32[7803]^c32[8027]^c32[8028]^c32[8030]^c32[8031]^c32[8776]^c32[8578]^c32[8587]^c32[8588]^c32[9452]^c32[9393]^c32[9281]^c32[9282]^c32[10176]^c32[10018]^c32[10019]^c32[10159]^c32[10160]^c32[10804]^c32[11058]^c32[11059]^c32[10798]^c32[10799]^c32[11570]^c32[11838]^c32[11791]^c32[11792]^c32[12613]^c32[12477]^c32[12426]^c32[12427]^c32[13389]^c32[13405]^c32[13406]^c32[13190]^c32[13191]^c32[13901]^c32[13898]^c32[13899]^c32[13929]^c32[13930]^c32[14607]^c32[14758]^c32[14631]^c32[14632]^c32[15525]^c32[15591]^c32[15592]^c32[15617]^c32[15618]^c32[16309]^c32[16459]^c32[16319]^c32[16320];
+
+       break;
+    case 43:
+      d32[16512]=c32[92]^c32[245]^c32[374]^c32[60]^c32[61]^c32[956]^c32[803]^c32[1023]^c32[1024]^c32[881]^c32[1755]^c32[1781]^c32[1649]^c32[2458]^c32[2546]^c32[2472]^c32[2473]^c32[3145]^c32[3189]^c32[3131]^c32[3132]^c32[4190]^c32[4153]^c32[3881]^c32[4609]^c32[4938]^c32[4975]^c32[4976]^c32[5492]^c32[5428]^c32[5697]^c32[5698]^c32[6260]^c32[6473]^c32[6368]^c32[6369]^c32[7014]^c32[7259]^c32[7144]^c32[7753]^c32[7978]^c32[7980]^c32[7981]^c32[8726]^c32[8528]^c32[8537]^c32[8538]^c32[9402]^c32[9343]^c32[9231]^c32[9232]^c32[10126]^c32[10353]^c32[10109]^c32[10110]^c32[10754]^c32[11009]^c32[11132]^c32[11133]^c32[11520]^c32[11788]^c32[11742]^c32[12563]^c32[12427]^c32[12376]^c32[12377]^c32[12420]^c32[13339]^c32[13356]^c32[13140]^c32[13141]^c32[13851]^c32[13849]^c32[13879]^c32[13880]^c32[13938]^c32[14941]^c32[14708]^c32[14966]^c32[15475]^c32[15542]^c32[15567]^c32[15568]^c32[16259]^c32[16409]^c32[16269]^c32[16270];
+
+       break;
+    case 44:
+      d32[16896]=c32[197]^c32[350]^c32[95]^c32[166]^c32[80]^c32[1061]^c32[908]^c32[1129]^c32[1860]^c32[1886]^c32[1754]^c32[2563]^c32[2651]^c32[2578]^c32[3250]^c32[3294]^c32[3237]^c32[3911]^c32[3874]^c32[3986]^c32[4714]^c32[4659]^c32[4697]^c32[5597]^c32[5533]^c32[5419]^c32[5454]^c32[6365]^c32[6194]^c32[6474]^c32[7119]^c32[6980]^c32[7249]^c32[7075]^c32[7858]^c32[7699]^c32[7702]^c32[8831]^c32[8633]^c32[8643]^c32[9507]^c32[9448]^c32[9337]^c32[10231]^c32[10074]^c32[10215]^c32[10859]^c32[11114]^c32[10854]^c32[11625]^c32[11893]^c32[11847]^c32[12668]^c32[12532]^c32[12482]^c32[13060]^c32[13077]^c32[13246]^c32[13956]^c32[13954]^c32[13985]^c32[14662]^c32[14813]^c32[14687]^c32[15580]^c32[15647]^c32[15673]^c32[16364]^c32[16130]^c32[16375];
+
+       break;
+    case 45:
+      d32[17280]=c32[903]^c32[4757]^c32[7695];
+
+       break;
+    
+    }
+  }
+}
+extern "C" int ldpc_BG1_Zc384_cuda32(uint32_t *c[4],uint32_t *d[4],int n_inputs) { 
+
+ dim3 numblocks(n_inputs,46);
+ ldpc_BG1_Zc384_worker<<<numblocks,384>>>(c,d);
+ cudaError_t err=cudaPeekAtLastError();
+ if (err!=cudaSuccess) {
+    printf("cuda error: %s (c %p, d %p)\n",cudaGetErrorString(err),c,d);
+    exit(-1);
+ }
+ cudaDeviceSynchronize();
+ return(0);
+}

openair1/PHY/CODING/nrLDPC_encoder/ldpc_encode_parity_check.c

@@ -36,13 +36,27 @@
 #include "assertions.h"
 #include "common/utils/LOG/log.h"
 
-
+#define USE_PERMUTEX
+#ifdef __aarch64__
+#define USE_ALIGNR
+#endif
 #ifdef __AVX512F__
+#if defined(__AVX512VBMI__) && defined(USE_PERMUTEX)
+// For AVX512 machines, use an AVX512 version of the encoder for Zc=384 only for now. This is used almost exclusively for
+// high-throughput cases. The version with "permutex2var" instruction uses less memory (i.e. 1/64th of the memory to store the
+// input), but uses more reads instead of creating 64 shifts of the input with memcpy
+#include "ldpc384_simd512_permutex_byte.c"
+#else
 #include "ldpc384_simd512_byte.c"
+#endif
 #else
 #include "ldpc384_byte.c"
 #endif
+#ifdef USE_ALIGNR
+#include "ldpc384_alignr_byte_128.c"
+#else
 #include "ldpc384_byte_128.c"
+#endif
 #include "ldpc352_byte.c"
 #include "ldpc352_byte_128.c"
 #include "ldpc320_byte.c"
@@ -89,16 +103,24 @@
 #include "ldpc_BG2_Zc80_byte.c"
 #include "ldpc_BG2_Zc72_byte.c"
 
-static void encode_parity_check_part_optim(uint8_t *cc, uint8_t *d, short BG,short Zc,short Kb, int simd_size, int ncols)
+
+static void encode_parity_check_part_optim(uint8_t *cc, uint8_t *d, short BG,short Zc,short Kb, int simd_size, int ncols,time_stats_t *tinput_memcpy)
 {
   unsigned char c[2 * 22 * Zc * simd_size] __attribute__((aligned(64))); //double size matrix of c
+  if (tinput_memcpy) start_meas(tinput_memcpy);
   for (int i1 = 0; i1 < ncols; i1++)   {
     memcpy(&c[2 * i1 * Zc], &cc[i1 * Zc], Zc * sizeof(unsigned char));
     memcpy(&c[(2 * i1 + 1) * Zc], &cc[i1 * Zc], Zc * sizeof(unsigned char));
   }
-  for (int i1 = 1; i1 < simd_size; i1++) {
-    memcpy(&c[(2 * ncols * Zc * i1)], &c[i1], (2 * ncols * Zc * sizeof(unsigned char)) - i1);
+#if (defined(USE_PERMUTEX) && defined(__AVX512VBMI__)) || defined(USE_ALIGNR)
+  if (Zc < 384 || BG == 2)
+#endif
+  {
+    for (int i1 = 1; i1 < simd_size; i1++) {
+      memcpy(&c[(2 * ncols * Zc * i1)], &c[i1], (2 * ncols * Zc * sizeof(unsigned char)) - i1);
+    }
   }
+  if (tinput_memcpy) stop_meas(tinput_memcpy);
   if (BG == 1) {
     switch (Zc) {
       case 176:

openair1/PHY/CODING/nrLDPC_encoder/ldpc_encode_parity_check_cuda.c

@@ -0,0 +1,94 @@
+/*
+ * 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 ldpc_encode_parity_check.c
+ * \brief Parity check function used by ldpc encoders
+ * \author Florian Kaltenberger, Raymond Knopp, Kien le Trung (Eurecom)
+ * \email openair_tech@eurecom.fr
+ * \date 27-03-2018
+ * \version 1.0
+ * \note
+ * \warning
+ */
+
+#include <stdlib.h>
+#include <math.h>
+#include <stdio.h>
+#include <string.h>
+#include "assertions.h"
+#include "common/utils/LOG/log.h"
+#include <cuda_runtime.h>
+
+
+
+int ldpc_BG1_Zc384_cuda32(uint32_t **c,uint32_t **d,int n_inputs);
+
+
+void encode_parity_check_part_cuda(uint32_t **c, uint32_t **d, short BG,short Zc,short Kb, int ncols, int n_inputs)
+{
+  
+  if (BG == 1) {
+    switch (Zc) {
+      case 176:
+      case 192:
+      case 208:
+      case 224:
+      case 240:
+      case 256:
+      case 288:
+      case 320:
+      case 352:
+	AssertFatal(1==0,"BG %d Zc %d not supported yet for CUDA\n",BG, Zc);
+        break;
+      case 384:
+	ldpc_BG1_Zc384_cuda32(c, d, n_inputs);
+        break;
+      default:
+        AssertFatal(false, "BG %d Zc %d is not supported yet\n", BG, Zc);
+    }
+  } else if (BG == 2) {
+    switch (Zc) {
+      case 72:
+      case 80:
+      case 88:
+      case 96:
+      case 104:
+      case 112:
+      case 120:
+      case 128:
+      case 144:
+      case 160:
+      case 176:
+      case 192:
+      case 208:
+      case 224:
+      case 240:
+      case 256:
+      case 288:
+      case 320:
+      case 352:
+      case 384:
+      default:
+        AssertFatal(false , "BG %d Zc %d is not supported yet\n", BG, Zc);
+    }
+  } else
+    AssertFatal(false, "BG %d is not supported\n", BG);
+}

openair1/PHY/CODING/nrLDPC_encoder/ldpc_encoder.c

@@ -58,11 +58,10 @@ int LDPCencoder(unsigned char **inputArray, unsigned char *outputArray, encoder_
   int no_punctured_columns, removed_bit;
   int nind=0;
   int indlist[1000];
-  int indlist2[1000];
 
   const short *Gen_shift_values = choose_generator_matrix(BG, Zc);
   if (Gen_shift_values==NULL) {
-    printf("ldpc_encoder_orig: could not find generator matrix\n");
+    printf("ldpc_encoder_orig: could not find generator matrix BG %d Zc %d\n",BG,Zc);
     return(-1);
   }
 
@@ -93,32 +92,35 @@ int LDPCencoder(unsigned char **inputArray, unsigned char *outputArray, encoder_
 
   // parity check part
 
-  if (gen_code>=1)
+  if (gen_code==1 || gen_code==2 || gen_code==3)
   {
     char fname[100];
     sprintf(fname,"ldpc_BG%d_Zc%d_byte.c",BG,Zc);
     FILE *fd=fopen(fname,"w");
     AssertFatal(fd!=NULL,"cannot open %s\n",fname);
-    sprintf(fname,"ldpc_BG%d_Zc%d_16bit.c",BG,Zc);
-    FILE *fd2=fopen(fname,"w");
-    AssertFatal(fd2!=NULL,"cannot open %s\n",fname);
+
 
     int shift;
     char data_type[100];
     char xor_command[100];
     int mask;
-
-
+    char permutex_command[100];
+    int use_permutex=0;
+    char alignr_command[100];
+    int use_alignr=0;
 
 
     fprintf(fd,"#include \"PHY/sse_intrin.h\"\n");
-    fprintf(fd2,"#include \"PHY/sse_intrin.h\"\n");
+
 
     if (gen_code == 1 && (Zc&63)==0) {
       shift=6;
       mask=63;
       strcpy(data_type,"__m512i");
       strcpy(xor_command,"_mm512_xor_si512");
+      strcpy(permutex_command,"_mm512_permutex2var_epi8");
+      strcpy(alignr_command,"_mm512_alignr_epi8");
+      use_alignr=1;
     }
     else if (gen_code == 1 && (Zc&31)==0) {
       shift=5; // AVX2 - 256-bit SIMD
@@ -126,14 +128,15 @@ int LDPCencoder(unsigned char **inputArray, unsigned char *outputArray, encoder_
       strcpy(data_type,"simde__m256i");
       strcpy(xor_command,"simde_mm256_xor_si256");
     }
-    else if ((Zc&15)==0) {
+    else if ((gen_code <=2) && (Zc&15)==0) {
       shift=4; // SSE4 - 128-bit SIMD
       mask=15;
       strcpy(data_type,"simde__m128i");
       strcpy(xor_command,"simde_mm_xor_si128");
-
+      strcpy(alignr_command,"simde_mm_alignr_epi8");
+      use_alignr=1;
     }
-    else if ((Zc&7)==0) {
+    else if ((gen_code <=2) && (Zc&7)==0) {
       shift=3; // MMX  - 64-bit SIMD
       mask=7;
       strcpy(data_type,"simde__m64");
@@ -145,21 +148,30 @@ int LDPCencoder(unsigned char **inputArray, unsigned char *outputArray, encoder_
       strcpy(data_type,"uint8_t");
       strcpy(xor_command,"scalar_xor");
       fprintf(fd,"#define scalar_xor(a,b) ((a)^(b))\n");
-      fprintf(fd2,"#define scalar_xor(a,b) ((a)^(b))\n");
+
     }
     fprintf(fd,"// generated code for Zc=%d, byte encoding\n",Zc);
-    fprintf(fd2,"// generated code for Zc=%d, 16bit encoding\n",Zc);
+    if (use_permutex==1) {
+	 fprintf(fd,"  const uint8_t abshift_bytes[%d][%d] __attribute__((aligned(64)))= {\n",1+mask,1+mask); 
+         for (int i=0;i<=mask;i++) {
+	    fprintf(fd,"{");
+	    for (int j=0;j<=mask;j++) {
+              fprintf(fd,"%d",i+j);
+	      if (j<mask) fprintf(fd,",");
+	      else fprintf(fd,"}");
+	    }
+	    if (i<mask) fprintf(fd,",\n");
+	    else fprintf(fd,"};\n");
+      	}
+	fprintf(fd,"static inline __m512i abshift_get(int r)\n {\n");
+        fprintf(fd,"   return * (const __m512i *) abshift_bytes[r];\n}");
+    }
     fprintf(fd,"static inline void ldpc_BG%d_Zc%d_byte(uint8_t *c,uint8_t *d) {\n",BG,Zc);
-    fprintf(fd2,"static inline void ldpc_BG%d_Zc%d_16bit(uint16_t *c,uint16_t *d) {\n",BG,Zc);
     fprintf(fd,"  %s *csimd=(%s *)c,*dsimd=(%s *)d;\n\n",data_type,data_type,data_type);
-    fprintf(fd2,"  %s *csimd=(%s *)c,*dsimd=(%s *)d;\n\n",data_type,data_type,data_type);
     fprintf(fd,"  %s *c2,*d2;\n\n",data_type);
-    fprintf(fd2,"  %s *c2,*d2;\n\n",data_type);
     fprintf(fd,"  int i2;\n");
-    fprintf(fd2,"  int i2;\n");
     fprintf(fd,"  for (i2=0; i2<%d; i2++) {\n",Zc>>shift);
-    if (shift > 0)
-      fprintf(fd2,"  for (i2=0; i2<%d; i2++) {\n",Zc>>(shift-1));
+
     for (i2=0; i2 < 1; i2++)
     {
       //t=Kb*Zc+i2;
@@ -169,53 +181,122 @@ int LDPCencoder(unsigned char **inputArray, unsigned char *outputArray, encoder_
 
       fprintf(fd,"     c2=&csimd[i2];\n");
       fprintf(fd,"     d2=&dsimd[i2];\n");
-      fprintf(fd2,"     c2=&csimd[i2];\n");
-      fprintf(fd2,"     d2=&dsimd[i2];\n");
 
       for (i1 = 0; i1 < nrows; i1++)
-
       {
         fprintf(fd,"\n//row: %d\n",i1);
-        fprintf(fd2,"\n//row: %d\n",i1);
-        AssertFatal(shift > 0 , "The result of the right shift is undefined because the right operand is negative\n");
+        AssertFatal(shift >= 0 , "The result of the right shift is undefined because the right operand is negative\n");
 	fprintf(fd,"     d2[%d]=",(Zc*i1)>>shift);
-	fprintf(fd2,"     d2[%d]=",(Zc*i1)>>(shift-1));
 
         nind=0;
 
         for (i3=0; i3 < ncols; i3++)
         {
           temp_prime=i1 * ncols + i3;
-
-
 	  for (i4=0; i4 < no_shift_values[temp_prime]; i4++)
-	    {
-	          
+  	  {
 	      var=(int)((i3*Zc + (Gen_shift_values[ pointer_shift_values[temp_prime]+i4 ]+1)%Zc)/Zc);
 	      int index =var*2*Zc + (i3*Zc + (Gen_shift_values[ pointer_shift_values[temp_prime]+i4 ]+1)%Zc) % Zc;
 	      printf("var %d, i3 %d, i4 %d, index %d, shift %d, Zc %d, pointer_shift_values[%d] %d gen_shift_value %d\n",var,i3,i4,index,shift,Zc,temp_prime,pointer_shift_values[temp_prime],Gen_shift_values[pointer_shift_values[temp_prime]]);
-	      indlist[nind] = ((index&mask)*((2*Zc*ncols)>>shift)/* *Kb */)+(index>>shift);
-	      printf("indlist[%d] %d, index&mask %d, index>>shift %d\n",nind,indlist[nind],index&mask,index>>shift);
-	      indlist2[nind++] = ((index&(mask>>1))*((2*Zc*ncols)>>(shift-1))*Kb)+(index>>(shift-1));
-	    }
-	  
-
-        }
+   	      printf("indlist[%d] %d, index&mask %d, index>>shift %d\n",nind,indlist[nind],index&mask,index>>shift);
+	      if (use_permutex==1 || use_alignr==1) {
+                indlist[nind++] = index;
+	      }
+	      else {
+	        indlist[nind] = ((index&mask)*((2*Zc*ncols)>>shift)/* *Kb */)+(index>>shift);
+	      }
+	  }//i4
+        }//i3
 	for (i4=0;i4<nind-1;i4++) {
-	  fprintf(fd,"%s(c2[%d],",xor_command,indlist[i4]);
-	  fprintf(fd2,"%s(c2[%d],",xor_command,indlist2[i4]);
+           if (use_permutex==0 && use_alignr==0) {
+  	      fprintf(fd,"%s(c2[%d],",xor_command,indlist[i4]);
+	   }
+	   else {
+              if ((indlist[i4]&mask) == 0) {
+  	        fprintf(fd,"%s(c2[%d],",xor_command,indlist[i4]>>shift);
+	      }
+	      else if (use_permutex==1) {
+                fprintf(fd,"%s(%s(c2[%d],abshift_get(%d),c2[%d]),",xor_command,permutex_command,(indlist[i4]>>shift),indlist[i4]&mask,(indlist[i4]>>shift)+1);
+	      }
+	      else if (use_alignr==1) {
+                fprintf(fd,"%s(%s(c2[%d],c2[%d],%d),",xor_command,alignr_command,(indlist[i4]>>shift)+1,(indlist[i4]>>shift),indlist[i4]&mask);
+	      }
+	   }
+	} //i4
+	if (use_permutex==0 && use_alignr==0)
+	  fprintf(fd,"c2[%d]",indlist[i4]);
+	else {
+          if ((indlist[i4]&mask) == 0) {
+  	     fprintf(fd,"c2[%d]",indlist[i4]>>shift);
+	  }
+	  else if (use_permutex==1){
+             fprintf(fd,"%s(c2[%d],abshift_get(%d),c2[%d])",permutex_command,(indlist[i4]>>shift),indlist[i4]&mask,(indlist[i4]>>shift)+1);
+	  }
+	  else if (use_alignr==1) {
+             fprintf(fd,"%s(c2[%d],c2[%d],%d)",alignr_command,(indlist[i4]>>shift)+1,(indlist[i4]>>shift),indlist[i4]&mask);
+
+	  }
 	}
-	fprintf(fd,"c2[%d]",indlist[i4]);
-	fprintf(fd2,"c2[%d]",indlist2[i4]);
-	for (i4=0;i4<nind-1;i4++) { fprintf(fd,")"); fprintf(fd2,")"); }
-	fprintf(fd,";\n");
-  fprintf(fd2, ";\n");
-      }
-      fprintf(fd,"  }\n}\n");
-      fprintf(fd2,"  }\n}\n");
-    }
+	for (i4=0;i4<nind-1;i4++) fprintf(fd,")"); 
+  	fprintf(fd,";\n");
+       } // i1 
+       fprintf(fd,"  }\n}\n");
+    } // i2
+    fclose(fd);
+  }
+  else if (gen_code == 4) { // CUDA
+    char fname[100];
+    sprintf(fname,"ldpc_BG%d_Zc%d_32bit.cu",BG,Zc);
+    FILE *fd=fopen(fname,"w");
+    AssertFatal(fd!=NULL,"cannot open %s\n",fname);
+    printf("Writing to %s\n",fname);
+    fprintf(fd,"#include <stdio.h>\n#include <stdint.h>\n#include <cuda_runtime.h>\n");
+
+    fprintf(fd,"// generated code for Zc=%d, byte encoding\n",Zc);
+    fprintf(fd,"__global__ void ldpc_BG%d_Zc%d_worker(uint32_t *c[4],uint32_t *d[4]) {\n",BG,Zc);
+    fprintf(fd,"  uint32_t *c32=c[blockIdx.x];\n  uint32_t *d32=d[blockIdx.x];\n\n");
+    fprintf(fd,"  int i2 = threadIdx.x;\n");
+    fprintf(fd,"  int i1 = blockIdx.y;\n");
+    fprintf(fd,"  if (i2 < %d) {\n",Zc);
+    fprintf(fd,"    c32+=i2;\n");
+    fprintf(fd,"    d32+=i2;\n");
+    fprintf(fd,"    switch(i1) {\n");
+
+    for (int i1=0;i1<nrows;i1++) {
+      nind = 0;
+      fprintf(fd,"    case %d:\n",i1);
+      fprintf(fd,"      d32[%d]=",(Zc*i1));
+      for (i3=0; i3 < ncols; i3++)
+      {
+          temp_prime=i1 * ncols + i3;
+	  for (i4=0; i4 < no_shift_values[temp_prime]; i4++)
+  	  {
+	     var=(int)((i3*Zc + (Gen_shift_values[ pointer_shift_values[temp_prime]+i4 ]+1)%Zc)/Zc);
+  	     int index =var*2*Zc + (i3*Zc + (Gen_shift_values[ pointer_shift_values[temp_prime]+i4 ]+1)%Zc) % Zc;
+	     printf("var %d, i3 %d, i4 %d, index %d (index mod 2Zc) %d, Zc %d, pointer_shift_values[%d] %d gen_shift_value %d offset %d\n",var,i3,i4,index,index%(2*Zc),Zc,temp_prime,pointer_shift_values[temp_prime],Gen_shift_values[pointer_shift_values[temp_prime]],(i3*Zc + (Gen_shift_values[ pointer_shift_values[temp_prime]+i4 ]+1)%Zc) % Zc);
+	     if (index%(2*Zc) >= Zc) printf("***********************\n");
+   	     indlist[nind] = index;
+	     printf("indlist[%d] %d, index %d\n",nind,indlist[nind],index);
+	     nind++;
+   	  } //i4
+      } // i3
+      for (i4=0;i4<nind-1;i4++) {
+         fprintf(fd,"c32[%d]^",indlist[i4]);
+      } //i4
+      fprintf(fd,"c32[%d];\n\n",indlist[i4]);
+      fprintf(fd,"       break;\n");
+    }// i1
+    fprintf(fd,"     }\n");
+    fprintf(fd,"  }\n");
+    fprintf(fd,"}\n");
+
+    fprintf(fd,"extern \"C\" int ldpc_BG%d_Zc%d_cuda32(uint32_t *c[4],uint32_t *d[4],int n_inputs) { \n",BG,Zc);
+    fprintf(fd,"dim3 numblocks(n_inputs,%d);\n",nrows);
+    fprintf(fd,"ldpc_BG%d_Zc%d_worker<<<numblocks,%d>>>(c,d);\n",BG,Zc,Zc);
+    fprintf(fd," cudaDeviceSynchronize();\n");
+    fprintf(fd,"  return(0);\n");
+    fprintf(fd,"}\n");
     fclose(fd);
-    fclose(fd2);
   }
   else if(gen_code==0)
   {
@@ -241,10 +322,12 @@ int LDPCencoder(unsigned char **inputArray, unsigned char *outputArray, encoder_
 
           for (i4 = 0; i4 < no_shift_values[temp_prime]; i4++) {
             channel_temp = channel_temp ^ c[i3 * Zc + Gen_shift_values[pointer_shift_values[temp_prime] + i4]];
+	    //if (i1==0) printf("index %d\n",i3 * Zc + Gen_shift_values[pointer_shift_values[temp_prime] + i4]);
           }
         }
 
         d[i2+i1*Zc]=channel_temp;
+	//if (i1==0) printf("reference: d[%d] %d c[%d] %d\n",i2,d[i2],i2,c[i2]);
         // output[t+i1*Zc]=channel_temp;
       }
     }

openair1/PHY/CODING/nrLDPC_encoder/ldpc_encoder32.c

@@ -0,0 +1,52 @@
+/*
+ * 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 ldpc_encoder.c
+ * \brief Defines the LDPC encoder
+ * \author Florian Kaltenberger, Raymond Knopp, Kien le Trung (Eurecom)
+ * \email openair_tech@eurecom.fr
+ * \date 27-03-2018
+ * \version 1.0
+ * \note
+ * \warning
+ */
+
+
+
+#include <stdlib.h>
+#include <math.h>
+#include <stdio.h>
+#include <string.h>
+#include "defs.h"
+#include "assertions.h"
+#include "openair1/PHY/CODING/nrLDPC_defs.h"
+#include "openair1/PHY/CODING/nrLDPC_extern.h"
+#include "ldpc_generate_coefficient.c"
+
+
+void cuda_support_init() {
+   return;
+}
+uint32_t **LDPCencoder32(uint8_t **input,encoder_implemparams_t *impp)
+{
+	AssertFatal(1==0,"Should not be getting here\n");
+}
+

openair1/PHY/CODING/nrLDPC_encoder/ldpc_encoder_cuda32.c

@@ -0,0 +1,791 @@
+/*
+ * 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 ldpc_encoder_optim8segmulti.c
+ * \brief Defines the optimized LDPC encoder
+ * \author Florian Kaltenberger, Raymond Knopp, Kien le Trung (Eurecom)
+ * \email openair_tech@eurecom.fr
+ * \date 27-03-2018
+ * \version 1.0
+ * \note
+ * \warning
+ */
+
+#include <stdlib.h>
+#include <math.h>
+#include <stdio.h>
+#include <string.h>
+#include "assertions.h"
+#include "common/utils/LOG/log.h"
+#include "time_meas.h"
+#include "openair1/PHY/CODING/nrLDPC_defs.h"
+#include "PHY/sse_intrin.h"
+#include "openair1/PHY/CODING/nrLDPC_extern.h"
+
+#include "ldpc_generate_coefficient.c"
+
+#include <cuda_runtime.h>
+
+//#define DEBUG_LDPC 1 
+
+#include "ldpc_encode_parity_check_cuda.c"
+uint32_t *c_dev;
+uint32_t **c_host;
+uint32_t *c_devh[4];
+uint32_t *d_dev;
+uint32_t **d_host;
+uint32_t *d_devh[4];
+uint32_t *cc_dev;
+uint32_t **cc_host;
+uint32_t *cc_devh[4];
+uint32_t *input_dev;
+uint32_t **input_host;
+uint32_t *input_devh[128];
+int managed = 0, concurrent = 0, uva = 0, pageable = 0, pageable_uses_host = 0, register_host = 0;
+
+#define USE_GPU_CIRCCOPY 1
+#define USE_GPU_FOR_INPUT 1
+
+void cuda_support_init() {
+
+
+    int dev = 0;
+    struct cudaDeviceProp prop;
+    cudaGetDeviceProperties(&prop, dev);
+
+
+    cudaDeviceGetAttribute(&managed, cudaDevAttrManagedMemory, dev);
+    cudaDeviceGetAttribute(&concurrent, cudaDevAttrConcurrentManagedAccess, dev);
+    cudaDeviceGetAttribute(&uva, cudaDevAttrUnifiedAddressing, dev);
+    cudaDeviceGetAttribute(&pageable, cudaDevAttrPageableMemoryAccess, dev);
+    cudaDeviceGetAttribute(&pageable_uses_host, cudaDevAttrPageableMemoryAccessUsesHostPageTables, dev);
+    cudaDeviceGetAttribute(&register_host, cudaDevAttrHostRegisterSupported,dev);
+
+    LOG_I(NR_PHY,"Device: %s (cc %d.%d)\n", prop.name, prop.major, prop.minor);
+    LOG_I(NR_PHY,"Unified Virtual Addressing (UVA): %s\n", uva ? "YES" : "NO");
+    LOG_I(NR_PHY,"Managed (Unified) Memory:        %s\n", managed ? "YES" : "NO");
+    LOG_I(NR_PHY,"Concurrent managed access:       %s\n", concurrent ? "YES" : "NO");
+    LOG_I(NR_PHY,"Pageable memory access:          %s\n", pageable ? "YES" : "NO");
+    LOG_I(NR_PHY,"Uses host page tables:           %s\n", pageable_uses_host ? "YES" : "NO");
+    LOG_I(NR_PHY,"Host Register supported:         %s\n", register_host ? "YES" : "NO");
+
+  // initialize input and output memory
+  if (!pageable && !register_host) {
+    cudaError_t err=cudaMalloc((void **)&c_dev,4*sizeof(uint32_t*));
+    AssertFatal(err == cudaSuccess,"CUDA Error (c_dev): %s\n", cudaGetErrorString(err));
+    err=cudaHostAlloc((void **)&c_host,4*sizeof(uint32_t*),cudaHostAllocDefault);
+    AssertFatal(err == cudaSuccess,"CUDA Error (c_host): %s\n", cudaGetErrorString(err));
+    for (int i=0;i<4;i++) {
+      err=cudaMalloc((void**)&c_devh[i],2*22*384*sizeof(uint32_t));
+      AssertFatal(err == cudaSuccess,"CUDA Error (c_devh[%d]): %s\n", i,cudaGetErrorString(err));
+      err=cudaHostAlloc((void**)&c_host[i],2*22*384*sizeof(uint32_t),cudaHostAllocDefault);
+      AssertFatal(err == cudaSuccess,"CUDA Error (chost[%d]): %s\n", i,cudaGetErrorString(err));
+    }
+    err = cudaMemcpy(c_dev,c_devh,4*sizeof(uint32_t*),cudaMemcpyHostToDevice);
+    AssertFatal(err == cudaSuccess,"CUDA Error (memcpy c_devh -> c_dev): %s\n", cudaGetErrorString(err));
+    err=cudaMalloc((void**)&d_dev,4*sizeof(uint32_t*));
+    AssertFatal(err == cudaSuccess,"CUDA Error: %s\n", cudaGetErrorString(err));
+    err=cudaHostAlloc((void **)&d_host,4*sizeof(uint32_t*),cudaHostAllocDefault);
+    AssertFatal(err == cudaSuccess,"CUDA Error (d_host): %s\n", cudaGetErrorString(err));
+    for (int i=0;i<4;i++) {
+      err=cudaMalloc((void**)&d_devh[i],68*384*sizeof(uint32_t));
+      AssertFatal(err == cudaSuccess,"CUDA Error (d_devh[%d]: %s\n", i,cudaGetErrorString(err));
+      err=cudaHostAlloc((void**)&d_host[i],68*384*sizeof(uint32_t),cudaHostAllocDefault);
+      AssertFatal(err == cudaSuccess,"CUDA Error (d_host[%d]): %s\n", i,cudaGetErrorString(err));
+    }
+    err=cudaMemcpy(d_dev,d_devh,4*sizeof(uint32_t*),cudaMemcpyHostToDevice);
+    AssertFatal(err == cudaSuccess,"CUDA Error (memcpy d_devh -> d_dev): %s\n", cudaGetErrorString(err));
+    err=cudaMalloc((void**)&cc_dev,4*sizeof(uint32_t*));
+    AssertFatal(err == cudaSuccess,"CUDA Error: %s\n", cudaGetErrorString(err));
+    err=cudaHostAlloc((void **)&cc_host,4*sizeof(uint32_t*),cudaHostAllocDefault);
+    AssertFatal(err == cudaSuccess,"CUDA Error (cc_host): %s\n", cudaGetErrorString(err));
+    for (int i=0;i<4;i++) {
+      err=cudaMalloc((void**)&cc_devh[i],68*384*sizeof(uint32_t));
+      AssertFatal(err == cudaSuccess,"CUDA Error (cc_devh[%d]: %s\n", i,cudaGetErrorString(err));
+      err=cudaHostAlloc((void**)&cc_host[i],68*384*sizeof(uint32_t),cudaHostAllocDefault);
+      AssertFatal(err == cudaSuccess,"CUDA Error (cc_host[%d]): %s\n", i,cudaGetErrorString(err));
+    }
+    err=cudaMemcpy(cc_dev,cc_devh,4*sizeof(uint32_t*),cudaMemcpyHostToDevice);
+    AssertFatal(err == cudaSuccess,"CUDA Error (memcpy cc_devh -> d_dev): %s\n", cudaGetErrorString(err));
+    err=cudaMalloc((void**)&input_dev,128*sizeof(uint8_t*));
+    AssertFatal(err == cudaSuccess,"CUDA Error: %s\n", cudaGetErrorString(err));
+    err=cudaHostAlloc((void **)&input_host,128*sizeof(uint8_t*),cudaHostAllocDefault);
+    AssertFatal(err == cudaSuccess,"CUDA Error (cc_host): %s\n", cudaGetErrorString(err));
+    for (int i=0;i<128;i++) {
+      err=cudaMalloc((void**)&input_devh[i],(8448/8)*sizeof(uint8_t));
+      AssertFatal(err == cudaSuccess,"CUDA Error (input_devh[%d]: %s\n", i,cudaGetErrorString(err));
+      err=cudaHostAlloc((void**)&input_host[i],(8448/8)*sizeof(uint8_t),cudaHostAllocDefault);
+      AssertFatal(err == cudaSuccess,"CUDA Error (input_host[%d]): %s\n", i,cudaGetErrorString(err));
+    }
+    err=cudaMemcpy(input_dev,input_devh,128*sizeof(uint8_t*),cudaMemcpyHostToDevice);
+    AssertFatal(err == cudaSuccess,"CUDA Error (memcpy cc_devh -> d_dev): %s\n", cudaGetErrorString(err));
+  }
+  else {
+    LOG_I(NR_PHY,"Allocating c,d,cc arrays for CPU/GPU shared-memory\n");
+    cudaError_t err=cudaHostAlloc((void **)&c_host,4*sizeof(uint32_t*),cudaHostAllocMapped|cudaHostAllocPortable);
+    AssertFatal(err == cudaSuccess,"CUDA Error (c_host): %s\n", cudaGetErrorString(err));
+    err = cudaHostGetDevicePointer((void**)&c_dev, c_host, 0);
+    AssertFatal(err == cudaSuccess,"CUDA Error (c_dev): %s\n", cudaGetErrorString(err));
+    LOG_I(NR_PHY,"c_host %p, c_dev %p\n",c_host,c_dev);
+    for (int i=0;i<4;i++) {
+      err=cudaHostAlloc((void**)&c_host[i],2*22*384*sizeof(uint32_t),cudaHostAllocMapped);
+      AssertFatal(err == cudaSuccess,"CUDA Error (c_host[%d]): %s\n", i,cudaGetErrorString(err));
+      err = cudaHostGetDevicePointer((void**)&c_devh[i], c_host[i], 0);
+      AssertFatal(err == cudaSuccess,"CUDA Error (c_devh[%d]): %s\n", i,cudaGetErrorString(err));
+    }
+    err=cudaMemcpy(c_dev,c_devh,4*sizeof(uint32_t*),cudaMemcpyHostToDevice);
+    AssertFatal(err == cudaSuccess,"CUDA Error (memcpy c_devh -> c_dev): %s\n", cudaGetErrorString(err));
+    err=cudaHostAlloc((void **)&d_host,4*sizeof(uint32_t*),cudaHostAllocMapped);
+    AssertFatal(err == cudaSuccess,"CUDA Error (d_host): %s\n", cudaGetErrorString(err));
+    err=cudaHostGetDevicePointer((void**)&d_dev, d_host, 0);
+    AssertFatal(err == cudaSuccess,"CUDA Error cudaHostGetDevicePointer(d_dev): %s\n", cudaGetErrorString(err));
+    LOG_I(NR_PHY,"d_host %p, d_dev %p\n",d_host,d_dev);
+    for (int i=0;i<4;i++) {
+      err=cudaHostAlloc((void**)&d_host[i],68*384*sizeof(uint32_t),cudaHostAllocMapped);
+      AssertFatal(err == cudaSuccess,"CUDA Error (d_host[%d]): %s\n", i,cudaGetErrorString(err));
+      err=cudaHostGetDevicePointer((void**)&d_devh[i], d_host[i], 0);
+      AssertFatal(err == cudaSuccess,"CUDA Error (cudaHostGetDevicePointer) d_devh[%d]: %s\n", i,cudaGetErrorString(err));
+    }
+    err=cudaMemcpy(d_dev,d_devh,4*sizeof(uint32_t*),cudaMemcpyHostToDevice);
+    AssertFatal(err == cudaSuccess,"CUDA Error (memcpy d_devh -> d_dev): %s\n", cudaGetErrorString(err));
+    err=cudaHostAlloc((void **)&cc_host,4*sizeof(uint32_t*),cudaHostAllocMapped);
+    AssertFatal(err == cudaSuccess,"CUDA Error (cc_host): %s\n", cudaGetErrorString(err));
+    err=cudaHostGetDevicePointer((void**)&cc_dev, cc_host, 0);
+    AssertFatal(err == cudaSuccess,"CUDA Error cudaHostGetDevicePointer(cc_dev): %s\n", cudaGetErrorString(err));
+    LOG_I(NR_PHY,"cc_host %p, cc_dev %p\n",cc_host,cc_dev);
+    for (int i=0;i<4;i++) {
+      err=cudaHostAlloc((void**)&cc_host[i],68*384*sizeof(uint32_t),cudaHostAllocMapped);
+      AssertFatal(err == cudaSuccess,"CUDA Error (cc_host[%d]): %s\n", i,cudaGetErrorString(err));
+      err=cudaHostGetDevicePointer((void**)&cc_devh[i], cc_host[i], 0);
+      AssertFatal(err == cudaSuccess,"CUDA Error (cudaHostGetDevicePointer) cc_devh[%d]: %s\n", i,cudaGetErrorString(err));
+    }
+    err=cudaMemcpy(cc_dev,cc_devh,4*sizeof(uint32_t*),cudaMemcpyHostToDevice);
+    AssertFatal(err == cudaSuccess,"CUDA Error (memcpy d_devh -> d_dev): %s\n", cudaGetErrorString(err));
+    err=cudaHostAlloc((void **)&input_host,128*sizeof(uint8_t*),cudaHostAllocMapped);
+    AssertFatal(err == cudaSuccess,"CUDA Error (input_host): %s\n", cudaGetErrorString(err));
+    err=cudaHostGetDevicePointer((void**)&input_dev, input_host, 0);
+    AssertFatal(err == cudaSuccess,"CUDA Error cudaHostGetDevicePointer(cc_host): %s\n", cudaGetErrorString(err));
+    LOG_I(NR_PHY,"input_host %p, input_dev %p\n",input_host,input_dev);
+    /*
+    for (int i=0;i<128;i++) {
+      err=cudaHostAlloc((void**)&input_host[i],(8448/8)*sizeof(uint8_t),cudaHostAllocMapped);
+      AssertFatal(err == cudaSuccess,"CUDA Error (input_host[%d]): %s\n", i,cudaGetErrorString(err));
+      err=cudaHostGetDevicePointer((void**)&input_devh[i], input_host[i], 0);
+      AssertFatal(err == cudaSuccess,"CUDA Error (cudaHostGetDevicePointer) input_devh[%d]: %s\n", i,cudaGetErrorString(err));
+    }
+    err=cudaMemcpy(input_dev,input_devh,128*sizeof(uint8_t*),cudaMemcpyHostToDevice);
+    AssertFatal(err == cudaSuccess,"CUDA Error (memcpy input_devh -> input_dev): %s\n", cudaGetErrorString(err));
+    */
+  }
+}
+
+uint32_t **LDPCencoder32(uint8_t **input, encoder_implemparams_t *impp)
+{
+  //set_log(PHY, 4);
+
+  int Zc = impp->Zc;
+  int Kb = impp->Kb;
+  short block_length = impp->K;
+  short BG = impp->BG;
+  int nrows=0,ncols=0;
+  int rate=3;
+  int no_punctured_columns,removed_bit;
+
+  if(impp->tinput != NULL) start_meas(impp->tinput);
+  //determine number of bits in codeword
+  if (BG==1)
+    {
+      nrows=46; //parity check bits
+      ncols=22; //info bits
+      rate=3;
+    }
+    else if (BG==2)
+    {
+      nrows=42; //parity check bits
+      ncols=10; // info bits
+      rate=5;
+    }
+
+#ifdef DEBUG_LDPC
+  LOG_I(PHY,"ldpc_encoder_cuda32: BG %d, Zc %d, Kb %d, block_length %d, segments %d\n",BG,Zc,Kb,block_length,impp->n_segments);
+  LOG_I(PHY,"ldpc_encoder_cuda32: PDU (seg 0) %x %x %x %x\n",input[0][0],input[0][1],input[0][2],input[0][3]);
+#endif
+
+  AssertFatal(Zc > 0, "no valid Zc found for block length %d\n", block_length);
+
+  int n_inputs = (impp->n_segments/32)+(((impp->n_segments&31) > 0) ? 1: 0);
+//  uint32_t  cc[4][22*Zc]; //padded input, unpacked, max size
+
+  // calculate number of punctured bits
+  no_punctured_columns=(int)((nrows-2)*Zc+block_length-block_length*rate)/Zc;
+  removed_bit=(nrows-no_punctured_columns-2) * Zc+block_length-(int)(block_length*rate);
+  // clear input
+  /*
+  for (int i=0;i<n_inputs;i++) { 
+    if (!pageable && !register_host)
+      cudaMemset(cc_devh[i],0,22*Zc*sizeof(uint32_t));
+    else
+      memset(cc_host[i],0,22*Zc*sizeof(uint32_t));
+  }*/
+
+
+#if 0
+  // unoptimized version of input processing
+  unsigned int temp;
+  for (int i_dword=0; i_dword < block_length; i_dword++) {
+    unsigned int i = i_dword;
+    for (int j = 0; j < impp->n_segments; j++) {
+
+      temp = (input[j][i/8]&(128>>(i&7)))>>(7-(i&7));
+      cc[j>>5][i] |= (temp << (j&31));
+    }
+  }
+#else
+//  uint32_t *ccp[4];
+//  for (int s=0;s<n_inputs;s++) ccp[s]=cc[s];  
+#ifdef USE_GPU_FOR_INPUT
+  if (!pageable && !register_host) {
+    for (int r=0;r<impp->n_segments;r++) {
+        cudaMemcpy(input_devh[r],input[r],block_length>>3,cudaMemcpyHostToDevice);
+    }
+  }
+  ldpc_input(pageable||register_host? input : input_dev,(uint32_t**)cc_dev,block_length,impp->n_segments);
+//  for (int i=0;i<16;i++) printf("i %d: cc[0] %x\n",i,cc[0][i]);
+#else  
+#ifndef __aarch64__
+  simde__m256i andmask = simde_mm256_set_epi32(0x1,0x2,0x4,0x8,0x10,0x20,0x40,0x80);  // every 8 bits -> 8 bytes, pattern repeats.
+  simde__m256i zero256 = simde_mm256_setzero_si256();
+  simde__m256i masks[32];
+  register simde__m256i c256;
+  for (int i=0;i<32;i++) {
+    masks[i] = simde_mm256_set1_epi32(1<<i);
+  }
+
+  int i;
+  for (i=0; i < block_length; i += 8) {
+    unsigned int i8 = i >> 3;
+    for (int j0=0;j0<n_inputs;j0++) {
+      c256 = simde_mm256_and_si256(simde_mm256_cmpeq_epi32(simde_mm256_andnot_si256(simde_mm256_set1_epi32(input[j0<<5][i8]),andmask),zero256),masks[0]);
+      int j2 = 32;
+      if (j0==(n_inputs-1)) j2= impp->n_segments-(j0<<5);
+      for (int j=1; j < j2; j++) {
+        c256 = simde_mm256_or_si256(simde_mm256_and_si256(simde_mm256_cmpeq_epi32(simde_mm256_andnot_si256(simde_mm256_set1_epi32(input[(j0<<5)+j][i8]),andmask),zero256),masks[j]),c256);
+      }
+      ((simde__m256i *)cc_host[j0])[i8] = c256;
+    }
+  }
+#else
+  int i2=0;
+  const int32_t ucShifta[32][4] = { 
+	  {-7,-6,-5,-4}, // 0  
+	  {-6,-5,-4,-3}, // 1
+	  {-5,-4,-3,-2}, // 2 
+	  {-4,-3,-2,-1}, // 3 
+	  {-3,-2,-1,0}, // 4 
+	  {-2,-1,0,1}, // 5
+	  {-1,0,1,2}, // 6
+	  {0,1,2,3}, // 7
+	  {1,2,3,4}, // 8
+	  {2,3,4,5}, // 9
+	  {3,4,5,6}, // 10
+	  {4,5,6,7}, // 11
+	  {5,6,7,8}, // 12
+	  {6,7,8,9}, // 13
+	  {7,8,9,10}, // 14
+	  {8,9,10,11}, // 15
+	  {9,10,11,12}, // 16
+	  {10,11,12,13}, // 17
+	  {11,12,13,14}, // 18
+	  {12,13,14,15}, // 19
+	  {13,14,15,16}, // 20
+	  {14,15,16,17}, // 21
+	  {15,16,17,18}, // 22
+	  {16,17,18,19}, // 23
+	  {17,18,19,20}, // 24
+	  {18,19,20,21}, // 25
+	  {19,20,21,22}, // 26
+	  {20,21,22,23}, // 27
+	  {21,22,23,24}, // 28
+	  {22,23,24,25}, // 29
+	  {23,24,25,26}, // 30
+	  {24,25,26,27}}; // 31
+  const int32_t ucShiftb[32][4] = { 
+	  {-3,-2,-1,0}, // 0 
+	  {-2,-1,0,1}, // 1
+	  {-1,0,1,2}, // 2
+	  {0,1,2,3}, // 3
+	  {1,2,3,4}, // 4
+	  {2,3,4,5}, // 5
+	  {3,4,5,6}, // 6 
+	  {4,5,6,7}, // 7
+	  {5,6,7,8}, // 8
+	  {6,7,8,9}, // 9
+	  {7,8,9,10}, // 10
+	  {8,9,10,11}, // 11
+	  {9,10,11,12}, // 12
+	  {10,11,12,13}, // 13
+	  {11,12,13,14}, // 14
+	  {12,13,14,15}, // 15
+	  {13,14,15,16}, // 16
+	  {14,15,16,17}, // 17
+	  {15,16,17,18}, // 18
+	  {16,17,18,19}, // 19
+	  {17,18,19,20}, // 20
+	  {18,19,20,21}, // 21
+	  {19,20,21,22}, // 22
+	  {20,21,22,23}, // 23
+	  {21,22,23,24}, // 24
+	  {22,23,24,25}, // 25
+	  {23,24,25,26}, // 26
+	  {24,25,26,27}, // 27
+	  {25,26,27,28}, // 28
+	  {26,27,28,29}, // 29
+	  {27,28,29,30}, // 30
+	  {28,29,30,31}}; // 31
+			      
+  const int32_t ucShiftc[32][4] = { 
+	  {-15,-14,-13,-12}, // 0
+	  {-14,-13,-12,-11}, // 1
+	  {-13,-12,-11,-10}, // 2
+          {-12,-11,-10,-9}, // 3
+	  {-11,-10,-9,-8}, // 4
+	  {-10,-9,-8,-7}, // 5
+	  {-9,-8,-7,-6}, // 6
+	  {-8,-7,-6,-5}, // 7
+	  {-7,-6,-5,-4}, // 8  
+	  {-6,-5,-4,-3}, // 9
+	  {-5,-4,-3,-2}, // 10 
+	  {-4,-3,-2,-1}, // 11 
+	  {-3,-2,-1,0}, // 12 
+	  {-2,-1,0,1}, // 13
+	  {-1,0,1,2}, // 14
+	  {0,1,2,3}, // 15
+	  {1,2,3,4}, // 16
+	  {2,3,4,5}, // 17
+	  {3,4,5,6}, // 18
+	  {4,5,6,7}, // 19
+	  {5,6,7,8}, // 20
+	  {6,7,8,9}, // 21
+	  {7,8,9,10}, // 22
+	  {8,9,10,11}, // 23
+	  {9,10,11,12}, // 24
+	  {10,11,12,13}, // 25
+	  {11,12,13,14}, // 26
+	  {12,13,14,15}, // 27
+	  {13,14,15,16}, // 28
+	  {14,15,16,17}, // 29
+	  {15,16,17,18}, // 30
+	  {16,17,18,19}}; // 31
+  const int32_t ucShiftd[32][4] = { 
+	  {-11,-10,-9,-8}, // 0
+	  {-10,-9,-8,-7}, // 1
+	  {-9,-8,-7,-6}, // 2
+	  {-8,-7,-6,-5}, // 3
+	  {-7,-6,-5,-4}, // 4  
+	  {-6,-5,-4,-3}, // 5
+	  {-5,-4,-3,-2}, // 6
+	  {-4,-3,-2,-1}, // 7 
+	  {-3,-2,-1,0}, // 8
+	  {-2,-1,0,1}, // 9
+	  {-1,0,1,2}, // 10
+	  {0,1,2,3}, // 11 
+	  {1,2,3,4}, // 12
+	  {2,3,4,5}, // 13
+	  {3,4,5,6}, // 14 
+	  {4,5,6,7}, // 15
+	  {5,6,7,8}, // 16
+	  {6,7,8,9}, // 17
+	  {7,8,9,10}, // 18
+	  {8,9,10,11}, // 19
+	  {9,10,11,12}, // 20
+	  {10,11,12,13}, // 21
+	  {11,12,13,14}, // 22
+	  {12,13,14,15}, // 23
+	  {13,14,15,16}, // 24
+	  {14,15,16,17}, // 25
+	  {15,16,17,18}, // 26
+	  {16,17,18,19}, // 27
+	  {17,18,19,20}, // 28
+	  {18,19,20,21}, // 29
+	  {19,20,21,22}, // 30
+	  {20,21,22,23}};// 31
+			  
+  const int32_t ucShifte[32][4] = { 
+	  {-23,-22,-21,-20}, // 0 
+	  {-22,-21,-20,-19}, // 1
+	  {-21,-20,-19,-18}, // 2
+	  {-20,-19,-18,-17}, // 3
+	  {-19,-18,-17,-16}, // 4 
+	  {-18,-17,-16,-15}, // 5
+	  {-17,-16,-15,-14}, // 6
+	  {-16,-15,-14,-13}, // 7
+	  {-15,-14,-13,-12}, // 8
+	  {-14,-13,-12,-11}, // 9
+	  {-13,-12,-11,-10}, // 10
+          {-12,-11,-10,-9}, // 11
+	  {-11,-10,-9,-8}, // 12
+	  {-10,-9,-8,-7}, // 13
+	  {-9,-8,-7,-6}, // 14 
+	  {-8,-7,-6,-5}, // 15
+	  {-7,-6,-5,-4}, // 16 
+	  {-6,-5,-4,-3}, // 17
+	  {-5,-4,-3,-2}, // 18 
+	  {-4,-3,-2,-1}, // 19 
+	  {-3,-2,-1,0}, // 20 
+	  {-2,-1,0,1}, // 21
+	  {-1,0,1,2}, // 22
+	  {0,1,2,3}, // 23
+	  {1,2,3,4}, // 24
+	  {2,3,4,5}, // 25
+	  {3,4,5,6}, // 26
+	  {4,5,6,7}, // 27
+	  {5,6,7,8}, // 28
+	  {6,7,8,9}, // 29
+	  {7,8,9,10}, // 30
+	  {8,9,10,11}}; // 31 
+			  
+  const int32_t ucShiftf[32][4] = { 
+	  {-19,-18,-17,-16}, // 0 
+	  {-18,-17,-16,-15}, // 1
+	  {-17,-16,-15,-14}, // 2
+	  {-16,-15,-14,-13}, // 3
+	  {-15,-14,-13,-12}, // 4
+	  {-14,-13,-12,-11}, // 5
+	  {-13,-12,-11,-10}, // 6
+          {-12,-11,-10,-9}, // 7
+	  {-11,-10,-9,-8}, // 8
+	  {-10,-9,-8,-7}, // 9  
+	  {-9,-8,-7,-6}, // 10 
+	  {-8,-7,-6,-5}, // 11
+	  {-7,-6,-5,-4}, // 12 
+	  {-6,-5,-4,-3}, // 13
+	  {-5,-4,-3,-2}, // 14
+	  {-4,-3,-2,-1}, // 15 
+	  {-3,-2,-1,0}, // 16
+	  {-2,-1,0,1}, // 17
+	  {-1,0,1,2}, // 18
+	  {0,1,2,3}, // 19
+	  {1,2,3,4}, // 20
+	  {2,3,4,5}, // 21
+	  {3,4,5,6}, // 22 
+	  {4,5,6,7}, // 23
+	  {5,6,7,8}, // 24
+	  {6,7,8,9}, // 25
+	  {7,8,9,10}, // 26
+	  {8,9,10,11}, // 27
+	  {9,10,11,12}, // 28
+	  {10,11,12,13}, // 29
+	  {11,12,13,14}, // 30
+	  {12,13,14,15}}; // 31
+			  
+  const int32_t ucShiftg[32][4] = { 
+	  {-31,-30,-29,-28},
+	  {-30,-29,-28,-27},
+	  {-29,-28,-27,-26},
+	  {-28,-27,-26,-25},
+	  {-27,-26,-25,-24},
+	  {-26,-25,-24,-23},
+	  {-25,-24,-23,-22},
+	  {-24,-23,-22,-21},
+	  {-23,-22,-21,-20}, // 0 
+	  {-22,-21,-20,-19}, // 1
+	  {-21,-20,-19,-18}, // 2
+	  {-20,-19,-18,-17}, // 3
+	  {-19,-18,-17,-16}, // 4 
+	  {-18,-17,-16,-15}, // 5
+	  {-17,-16,-15,-14}, // 6
+	  {-16,-15,-14,-13}, // 7
+	  {-15,-14,-13,-12}, // 8
+	  {-14,-13,-12,-11}, // 9
+	  {-13,-12,-11,-10}, // 10
+          {-12,-11,-10,-9}, // 11
+	  {-11,-10,-9,-8}, // 12
+	  {-10,-9,-8,-7}, // 13
+	  {-9,-8,-7,-6}, // 14 
+	  {-8,-7,-6,-5}, // 15
+	  {-7,-6,-5,-4}, // 16 
+	  {-6,-5,-4,-3}, // 17
+	  {-5,-4,-3,-2}, // 18 
+	  {-4,-3,-2,-1}, // 19 
+	  {-3,-2,-1,0}, // 20 
+	  {-2,-1,0,1}, // 21
+	  {-1,0,1,2}, // 22
+	  {0,1,2,3}}; // 23
+			  
+  const int32_t ucShifth[32][4] = { 
+	  {-27,-26,-25,-24}, // 0
+	  {-26,-25,-24,-23}, // 1
+	  {-25,-24,-23,-22}, // 2
+	  {-24,-23,-22,-21}, // 3
+	  {-23,-22,-21,-20}, // 4 
+	  {-22,-21,-20,-19}, // 5
+	  {-21,-20,-19,-18}, // 6
+	  {-20,-19,-18,-17}, // 7
+	  {-19,-18,-17,-16}, // 8 
+	  {-18,-17,-16,-15}, // 9
+	  {-17,-16,-15,-14}, // 10
+	  {-16,-15,-14,-13}, // 11
+	  {-15,-14,-13,-12}, // 12
+	  {-14,-13,-12,-11}, // 13
+	  {-13,-12,-11,-10}, // 14
+          {-12,-11,-10,-9}, // 15
+	  {-11,-10,-9,-8}, // 16
+	  {-10,-9,-8,-7}, // 17 
+	  {-9,-8,-7,-6}, // 18 
+	  {-8,-7,-6,-5}, // 19
+	  {-7,-6,-5,-4}, // 20 
+	  {-6,-5,-4,-3}, // 21
+	  {-5,-4,-3,-2}, // 22
+	  {-4,-3,-2,-1}, // 23 
+	  {-3,-2,-1,0}, // 24 
+	  {-2,-1,0,1}, // 25
+	  {-1,0,1,2}, // 26
+	  {0,1,2,3}, // 27
+	  {1,2,3,4}, // 28
+	  {2,3,4,5}, // 29
+	  {3,4,5,6}, // 30
+          {4,5,6,7}}; // 31 
+  const uint32_t __attribute__ ((aligned (16))) masksa[4] = {0x80,0x40,0x20,0x10};
+  const uint32_t __attribute__ ((aligned (16))) masksb[4] = {0x8,0x4,0x2,0x1};
+  const uint32_t __attribute__ ((aligned (16))) masksc[4] = {0x8000,0x4000,0x2000,0x1000};
+  const uint32_t __attribute__ ((aligned (16))) masksd[4] = {0x800,0x400,0x200,0x100};
+  const uint32_t __attribute__ ((aligned (16))) maskse[4] = {0x800000,0x400000,0x200000,0x100000};
+  const uint32_t __attribute__ ((aligned (16))) masksf[4] = {0x80000,0x40000,0x20000,0x10000};
+  const uint32_t __attribute__ ((aligned (16))) masksg[4] = {0x80000000,0x40000000,0x20000000,0x10000000};
+  const uint32_t __attribute__ ((aligned (16))) masksh[4] = {0x8000000,0x4000000,0x2000000,0x1000000};
+  int32x4_t vshifta[32],vshiftb[32],vshiftc[32],vshiftd[32],vshifte[32],vshiftf[32],vshiftg[32],vshifth[32];
+  uint32x4_t vmasksa  = vld1q_u32(masksa);
+  uint32x4_t vmasksb  = vld1q_u32(masksb);
+  uint32x4_t vmasksc  = vld1q_u32(masksc);
+  uint32x4_t vmasksd  = vld1q_u32(masksd);
+  uint32x4_t vmaskse  = vld1q_u32(maskse);
+  uint32x4_t vmasksf  = vld1q_u32(masksf);
+  uint32x4_t vmasksg  = vld1q_u32(masksg);
+  uint32x4_t vmasksh  = vld1q_u32(masksh);
+  uint32x4_t in;
+
+  for (int n=0;n<32;n++) { 
+    vshifta[n] = vld1q_s32(ucShifta[n]);
+    vshiftb[n] = vld1q_s32(ucShiftb[n]);
+    vshiftc[n] = vld1q_s32(ucShiftc[n]);
+    vshiftd[n] = vld1q_s32(ucShiftd[n]);
+    vshifte[n] = vld1q_s32(ucShifte[n]);
+    vshiftf[n] = vld1q_s32(ucShiftf[n]);
+    vshiftg[n] = vld1q_s32(ucShiftg[n]);
+    vshifth[n] = vld1q_s32(ucShifth[n]);
+  }
+  i2=0;
+  int j0=0,j1=0,j2=0,j3=0;
+  if (impp->n_segments <= 32) {
+	  j0=impp->n_segments;
+  }
+  else if (impp->n_segments <= 64) {
+     j0=32;
+     j1=impp->n_segments-32;
+  }
+  else if (impp->n_segments <= 96) {
+     j0=32; j1=32;
+     j2=impp->n_segments-64;
+  }
+  else if (impp->n_segments <= 128) {
+     j0=32; j1=32; j2=32;
+     j3=impp->n_segments-96;
+  }
+  uint32x4_t *ccp,cc0,cc1,cc2,cc3,cc4,cc5,cc6,cc7;
+  for (int i=0; i < (block_length>>5); i++,i2+=8) {
+    in = vdupq_n_u32(((uint32_t*)input[0])[i]);
+    cc0 = vshlq_u32(vandq_u32(in,vmasksa),vshifta[0]);
+    cc1 = vshlq_u32(vandq_u32(in,vmasksb),vshiftb[0]);
+    cc2 = vshlq_u32(vandq_u32(in,vmasksc),vshiftc[0]);
+    cc3 = vshlq_u32(vandq_u32(in,vmasksd),vshiftd[0]);
+    cc4 = vshlq_u32(vandq_u32(in,vmaskse),vshifte[0]);
+    cc5 = vshlq_u32(vandq_u32(in,vmasksf),vshiftf[0]);
+    cc6 = vshlq_u32(vandq_u32(in,vmasksg),vshiftg[0]);
+    cc7 = vshlq_u32(vandq_u32(in,vmasksh),vshifth[0]);
+    for (int j = 1; j < j0; j++) {
+      in = vdupq_n_u32(((uint32_t*)input[j])[i]);
+      cc0 = vorrq_u32(cc0,vshlq_u32(vandq_u32(in,vmasksa),vshifta[j]));
+      cc1 = vorrq_u32(cc1,vshlq_u32(vandq_u32(in,vmasksb),vshiftb[j]));
+      cc2 = vorrq_u32(cc2,vshlq_u32(vandq_u32(in,vmasksc),vshiftc[j]));
+      cc3 = vorrq_u32(cc3,vshlq_u32(vandq_u32(in,vmasksd),vshiftd[j]));
+      cc4 = vorrq_u32(cc4,vshlq_u32(vandq_u32(in,vmaskse),vshifte[j]));
+      cc5 = vorrq_u32(cc5,vshlq_u32(vandq_u32(in,vmasksf),vshiftf[j]));
+      cc6 = vorrq_u32(cc6,vshlq_u32(vandq_u32(in,vmasksg),vshiftg[j]));
+      cc7 = vorrq_u32(cc7,vshlq_u32(vandq_u32(in,vmasksh),vshifth[j]));
+    }
+    ccp=&((uint32x4_t *)cc_host[0])[i2];
+    ccp[0] = cc0;
+    ccp[1] = cc1;
+    ccp[2] = cc2;
+    ccp[3] = cc3;
+    ccp[4] = cc4;
+    ccp[5] = cc5;
+    ccp[6] = cc6;
+    ccp[7] = cc7;
+    if (j1>0) {
+      in = vdupq_n_u32(((uint32_t*)input[32])[i]);
+      cc0 = vshlq_u32(vandq_u32(in,vmasksa),vshifta[0]);
+      cc1 = vshlq_u32(vandq_u32(in,vmasksb),vshiftb[0]);
+      cc2 = vshlq_u32(vandq_u32(in,vmasksc),vshiftc[0]);
+      cc3 = vshlq_u32(vandq_u32(in,vmasksd),vshiftd[0]);
+      cc4 = vshlq_u32(vandq_u32(in,vmaskse),vshifte[0]);
+      cc5 = vshlq_u32(vandq_u32(in,vmasksf),vshiftf[0]);
+      cc6 = vshlq_u32(vandq_u32(in,vmasksg),vshiftg[0]);
+      cc7 = vshlq_u32(vandq_u32(in,vmasksh),vshifth[0]);
+      for (int j = 1; j < j1; j++) {
+        in = vdupq_n_u32(((uint32_t*)input[32+j])[i]);
+        cc0 = vorrq_u32(cc0,vshlq_u32(vandq_u32(in,vmasksa),vshifta[j]));
+        cc1 = vorrq_u32(cc1,vshlq_u32(vandq_u32(in,vmasksb),vshiftb[j]));
+        cc2 = vorrq_u32(cc2,vshlq_u32(vandq_u32(in,vmasksc),vshiftc[j]));
+        cc3 = vorrq_u32(cc3,vshlq_u32(vandq_u32(in,vmasksd),vshiftd[j]));
+        cc4 = vorrq_u32(cc4,vshlq_u32(vandq_u32(in,vmaskse),vshifte[j]));
+        cc5 = vorrq_u32(cc5,vshlq_u32(vandq_u32(in,vmasksf),vshiftf[j]));
+        cc6 = vorrq_u32(cc6,vshlq_u32(vandq_u32(in,vmasksg),vshiftg[j]));
+        cc7 = vorrq_u32(cc7,vshlq_u32(vandq_u32(in,vmasksh),vshifth[j]));
+      }
+      ccp=&((uint32x4_t *)cc_host[1])[i2];
+      ccp[0] = cc0;
+      ccp[1] = cc1;
+      ccp[2] = cc2;
+      ccp[3] = cc3;
+      ccp[4] = cc4;
+      ccp[5] = cc5;
+      ccp[6] = cc6;
+      ccp[7] = cc7;
+    }
+    if (j2>0) {
+      in = vdupq_n_u32(((uint32_t*)input[64])[i]);
+      cc0 = vshlq_u32(vandq_u32(in,vmasksa),vshifta[0]);
+      cc1 = vshlq_u32(vandq_u32(in,vmasksb),vshiftb[0]);
+      cc2 = vshlq_u32(vandq_u32(in,vmasksc),vshiftc[0]);
+      cc3 = vshlq_u32(vandq_u32(in,vmasksd),vshiftd[0]);
+      cc4 = vshlq_u32(vandq_u32(in,vmaskse),vshifte[0]);
+      cc5 = vshlq_u32(vandq_u32(in,vmasksf),vshiftf[0]);
+      cc6 = vshlq_u32(vandq_u32(in,vmasksg),vshiftg[0]);
+      cc7 = vshlq_u32(vandq_u32(in,vmasksh),vshifth[0]);
+      for (int j = 1; j < j2; j++) {
+        in = vdupq_n_u32(((uint32_t*)input[64+j])[i]);
+        cc0 = vorrq_u32(cc0,vshlq_u32(vandq_u32(in,vmasksa),vshifta[j]));
+        cc1 = vorrq_u32(cc1,vshlq_u32(vandq_u32(in,vmasksb),vshiftb[j]));
+        cc2 = vorrq_u32(cc2,vshlq_u32(vandq_u32(in,vmasksc),vshiftc[j]));
+        cc3 = vorrq_u32(cc3,vshlq_u32(vandq_u32(in,vmasksd),vshiftd[j]));
+        cc4 = vorrq_u32(cc4,vshlq_u32(vandq_u32(in,vmaskse),vshifte[j]));
+        cc5 = vorrq_u32(cc5,vshlq_u32(vandq_u32(in,vmasksf),vshiftf[j]));
+        cc6 = vorrq_u32(cc6,vshlq_u32(vandq_u32(in,vmasksg),vshiftg[j]));
+        cc7 = vorrq_u32(cc7,vshlq_u32(vandq_u32(in,vmasksh),vshifth[j]));
+      }
+      ccp=&((uint32x4_t *)cc_host[2])[i2];
+      ccp[0] = cc0;
+      ccp[1] = cc1;
+      ccp[2] = cc2;
+      ccp[3] = cc3;
+      ccp[4] = cc4;
+      ccp[5] = cc5;
+      ccp[6] = cc6;
+      ccp[7] = cc7;
+    }
+    if (j3>0) {
+      in = vdupq_n_u32(((uint32_t*)input[96])[i]);
+      cc0 = vshlq_u32(vandq_u32(in,vmasksa),vshifta[0]);
+      cc1 = vshlq_u32(vandq_u32(in,vmasksb),vshiftb[0]);
+      cc2 = vshlq_u32(vandq_u32(in,vmasksc),vshiftc[0]);
+      cc3 = vshlq_u32(vandq_u32(in,vmasksd),vshiftd[0]);
+      cc4 = vshlq_u32(vandq_u32(in,vmaskse),vshifte[0]);
+      cc5 = vshlq_u32(vandq_u32(in,vmasksf),vshiftf[0]);
+      cc6 = vshlq_u32(vandq_u32(in,vmasksg),vshiftg[0]);
+      cc7 = vshlq_u32(vandq_u32(in,vmasksh),vshifth[0]);
+      for (int j = 1; j < j3; j++) {
+        in = vdupq_n_u32(((uint32_t*)input[96+j])[i]);
+        cc0 = vorrq_u32(cc0,vshlq_u32(vandq_u32(in,vmasksa),vshifta[j]));
+        cc1 = vorrq_u32(cc1,vshlq_u32(vandq_u32(in,vmasksb),vshiftb[j]));
+        cc2 = vorrq_u32(cc2,vshlq_u32(vandq_u32(in,vmasksc),vshiftc[j]));
+        cc3 = vorrq_u32(cc3,vshlq_u32(vandq_u32(in,vmasksd),vshiftd[j]));
+        cc4 = vorrq_u32(cc4,vshlq_u32(vandq_u32(in,vmaskse),vshifte[j]));
+        cc5 = vorrq_u32(cc5,vshlq_u32(vandq_u32(in,vmasksf),vshiftf[j]));
+        cc6 = vorrq_u32(cc6,vshlq_u32(vandq_u32(in,vmasksg),vshiftg[j]));
+        cc7 = vorrq_u32(cc7,vshlq_u32(vandq_u32(in,vmasksh),vshifth[j]));
+      }
+      ccp=&((uint32x4_t *)cc_host[3])[i2];
+      ccp[0] = cc0;
+      ccp[1] = cc1;
+      ccp[2] = cc2;
+      ccp[3] = cc3;
+      ccp[4] = cc4;
+      ccp[5] = cc5;
+      ccp[6] = cc6;
+      ccp[7] = cc7;
+    }
+  }
+#endif
+#endif
+#endif
+  if(impp->tinput != NULL) stop_meas(impp->tinput);
+  if(impp->tinput_memcpy != NULL) start_meas(impp->tinput_memcpy);
+#ifdef USE_GPU_CIRCCOPY
+  circcopy_c(cc_dev,c_dev,n_inputs);
+#else
+#ifdef USE_CUDA_MEMCPY 
+  for (int s=0;s<n_inputs;s++)  {
+    for (int i1 = 0; i1 < ncols; i1++)   {
+      cudaError_t err = cudaMemcpy(&c_devh[s][2 * i1 * Zc], &cc_devh[s][i1 * Zc], Zc * sizeof(uint32_t),cudaMemcpyDeviceToDevice);
+      AssertFatal(err == cudaSuccess, "1st memcpy cc_devh[%d] -> c_devh[%d] %p CUDA Error: %s\n", s, s, c_devh[s],cudaGetErrorString(err)); 			
+      err = cudaMemcpy(&c_devh[s][(2 * i1 + 1) * Zc], &cc_devh[s][i1 * Zc], Zc * sizeof(uint32_t),cudaMemcpyDeviceToDevice);
+      AssertFatal(err == cudaSuccess, "2nd memcpy cc_devh[%d] -> c_devh[%d] %p CUDA Error: %s\n", s, s, c_devh[s],cudaGetErrorString(err)); 			
+    }
+#else
+  for (int s=0;s<n_inputs;s++)  {
+    for (int i1 = 0; i1 < ncols; i1++)   {
+      memcpy(&c_host[s][2 * i1 * Zc], &cc_host[s][i1 * Zc], Zc * sizeof(uint32_t));
+      memcpy(&c_host[s][(2 * i1 + 1) * Zc], &cc_host[s][i1 * Zc], Zc * sizeof(uint32_t));
+    }
+    if (!pageable && !register_host) {
+      cudaError_t err = cudaMemcpy(c_devh[s],c_host[s],sizeof(uint32_t)*2*22*Zc,cudaMemcpyHostToDevice);
+      AssertFatal(err == cudaSuccess, "memcpy c_host[%d] -> c_devh[%d] %p CUDA Error: %s\n", s, s, c_devh[s],cudaGetErrorString(err)); 			
+    }
+  }
+#endif 
+#endif
+  if(impp->tinput != NULL) stop_meas(impp->tinput_memcpy);
+  if (BG==1 && Zc==384)  {
+    //parity check part
+    if(impp->tparity != NULL) start_meas(impp->tparity);
+    encode_parity_check_part_cuda((uint32_t**)c_dev, (uint32_t**)d_dev, BG, Zc, Kb, ncols,n_inputs);
+    if(impp->tparity != NULL) stop_meas(impp->tparity);
+  }
+  else {
+    AssertFatal(1==0,"Only BG1 Zc=384 for now\n");
+  }
+  if(impp->toutput != NULL) start_meas(impp->toutput);
+  for (int s=0;s<n_inputs;s++) {
+	  
+    if (!pageable && !register_host) {
+      cudaMemcpy(d_host[s],&cc_devh[s][2*Zc],sizeof(uint32_t)*(block_length-(2*Zc)),cudaMemcpyDeviceToHost);
+      cudaError_t err = cudaMemcpy(&d_host[s][block_length-(2*Zc)],d_devh[s]+block_length-(2*Zc),sizeof(uint32_t)*((nrows-no_punctured_columns) * Zc-removed_bit),cudaMemcpyDeviceToHost);
+      AssertFatal(err == cudaSuccess, "d_dev[%d] %p CUDA Error: %s\n", s, d_devh[s],cudaGetErrorString(err)); 			
+    }
+    else {
+       memcpy(d_host[s],&cc_host[s][2*Zc],sizeof(uint32_t)*(block_length-(2*Zc)));
+    }
+  }
+  if(impp->toutput != NULL) stop_meas(impp->toutput);
+  return d_host;
+}
+

openair1/PHY/CODING/nrLDPC_encoder/ldpc_encoder_optim8segmulti.c

@@ -43,6 +43,11 @@
 #include "ldpc_encode_parity_check.c"
 #include "ldpc_generate_coefficient.c"
 
+
+void ini_input32_luts() {
+	return;
+}
+
 int LDPCencoder(uint8_t **input, uint8_t *output, encoder_implemparams_t *impp)
 {
   //set_log(PHY, 4);
@@ -107,7 +112,6 @@ int LDPCencoder(uint8_t **input, uint8_t *output, encoder_implemparams_t *impp)
   //interleave up to 8 transport-block segements at a time
 
   unsigned int i_byte = 0;
-
 #if defined(__AVX512F__) && defined(__AVX512BW__) && defined(__AVX512VBMI__)
   const __m512i masks5[8] = { _mm512_set1_epi8(0x1), _mm512_set1_epi8(0x2),
                               _mm512_set1_epi8(0x4), _mm512_set1_epi8(0x8),
@@ -193,12 +197,9 @@ int LDPCencoder(uint8_t **input, uint8_t *output, encoder_implemparams_t *impp)
   if(impp->tinput != NULL) stop_meas(impp->tinput);
 
   if ((BG==1 && Zc>=176) || (BG==2 && Zc>=72)) {
-    // extend matrix
-    if(impp->tprep != NULL) start_meas(impp->tprep);
-    if(impp->tprep != NULL) stop_meas(impp->tprep);
     //parity check part
     if(impp->tparity != NULL) start_meas(impp->tparity);
-    encode_parity_check_part_optim(cc, dd, BG, Zc, Kb, simd_size, ncols);
+    encode_parity_check_part_optim(cc, dd, BG, Zc, Kb, simd_size, ncols,impp->tinput_memcpy);
     if(impp->tparity != NULL) stop_meas(impp->tparity);
   }
   else {
@@ -207,8 +208,10 @@ int LDPCencoder(uint8_t **input, uint8_t *output, encoder_implemparams_t *impp)
       return(-1);
     }
   }
+  if(impp->toutput != NULL) start_meas(impp->toutput);
   memcpy(output,&cc[2*Zc],(block_length-(2*Zc)));
   memcpy(output+block_length-(2*Zc),dd,((nrows-no_punctured_columns) * Zc-removed_bit));
+  if(impp->toutput != NULL) stop_meas(impp->toutput);
   return 0;
 }
 

openair1/PHY/CODING/nrLDPC_encoder/ldpc_input.cu

@@ -0,0 +1,175 @@
+#include <stdio.h>
+#include <stdint.h>
+#include <cuda_runtime.h>
+/*
+__device__ const uint32_t masks[4] = {0x80,0x8000,0x800000,0x80000000};
+__global__ void ldpc_input_worker(uint32_t **input,uint32_t *cc[4],int block_length,int nseg) {
+
+  int block_off = blockIdx.x*blockDim.x<<2;	
+  int i2 = threadIdx.x<<2;
+  uint32_t *out=cc[blockIdx.y] + block_off + i2;
+  int nseg0 = (blockIdx.y << 5);
+  int nseg1;
+  if ((nseg0 + 32) <= nseg) nseg1 = nseg0+32;
+  else nseg1 = nseg0 + (nseg&31);
+  int bit_offset = i2+block_off;
+  int uint32_offset = bit_offset>>5;  
+  uint32_t mask = masks[(bit_offset&31)>>3]; 
+  uint32_t mask0 = mask>>(bit_offset&7);bit_offset++;
+  uint32_t mask1 = mask>>(bit_offset&7);bit_offset++;
+  uint32_t mask2 = mask>>(bit_offset&7);bit_offset++;
+  uint32_t mask3 = mask>>(bit_offset&7);
+  uint32_t tmp,jmod;
+  uint32_t otmp0,otmp1,otmp2,otmp3;
+  if (bit_offset  < block_length) {
+      tmp=input[nseg0][uint32_offset];
+      otmp0 = ((tmp&mask0) > 0); 
+      otmp1 = ((tmp&mask1) > 0); 
+      otmp2 = ((tmp&mask2) > 0); 
+      otmp3 = ((tmp&mask3) > 0); 
+      for (int j=nseg0+1;j<nseg1;j++) {
+	 tmp=input[j][uint32_offset];
+	 jmod = j&31;
+         otmp0 |= (((tmp&mask0) > 0)<<jmod); 
+         otmp1 |= (((tmp&mask1) > 0)<<jmod); 
+         otmp2 |= (((tmp&mask2) > 0)<<jmod); 
+         otmp3 |= (((tmp&mask3) > 0)<<jmod); 
+      }	      
+      out[0]=otmp0;
+      out[1]=otmp1;
+      out[2]=otmp2;
+      out[3]=otmp3;
+  } 
+}
+*/
+
+#define ITERATIONS 1
+#define NTHREADS 384 
+__device__ uint32_t masks[32] = {
+	0x80,0x40,0x20,0x10,0x8,0x4,0x2,0x1,
+	0x8000,0x4000,0x2000,0x1000,0x800,0x400,0x200,0x100,
+	0x800000,0x400000,0x200000,0x100000,0x80000,0x40000,0x20000,0x10000,
+	0x80000000,0x40000000,0x20000000,0x10000000,0x8000000,0x4000000,0x2000000,0x1000000};
+__global__ void ldpc_input_worker(uint32_t **input,uint32_t *cc[4],int block_length,int nseg) {
+
+  int block_off = blockIdx.x*blockDim.x*ITERATIONS;	
+  int i2 = threadIdx.x*ITERATIONS;
+  uint32_t *out=cc[blockIdx.y] + block_off + i2;
+  int nseg0 = (blockIdx.y << 5);
+  int nseg1;
+  if ((nseg0 + 32) <= nseg) nseg1 = nseg0+32;
+  else nseg1 = nseg0 + (nseg&31);
+  int bit_offset = i2+block_off;
+  int uint32_offset = bit_offset>>5;  
+#if ITERATIONS==1
+  uint32_t mask0 = masks[bit_offset&31];
+#else
+  uint32_t *mask = &masks[bit_offset&31]; 
+  uint32_t mask0 = mask[0];
+#endif
+#if ITERATIONS==4 || ITERATIONS==8
+  uint32_t mask1 = mask[1];
+  uint32_t mask2 = mask[2];
+  uint32_t mask3 = mask[3];
+#endif
+#if ITERATIONS==8
+  uint32_t mask4 = mask[4];
+  uint32_t mask5 = mask[5];
+  uint32_t mask6 = mask[6];
+  uint32_t mask7 = mask[7];
+#endif
+  uint32_t tmp,jmod;
+  uint32_t otmp0;
+#if ITERATIONS==4 || ITERATIONS==8
+  uint32_t otmp1,otmp2,otmp3;
+#endif
+#if ITERATIONS==8
+  uint32_t otmp4,otmp5,otmp6,otmp7;
+#endif
+  if (bit_offset  < block_length) {
+      tmp=input[nseg0][uint32_offset];
+      otmp0 = ((tmp&mask0) > 0); 
+#if ITERATIONS==4 || ITERATIONS==8
+      otmp1 = ((tmp&mask1) > 0); 
+      otmp2 = ((tmp&mask2) > 0); 
+      otmp3 = ((tmp&mask3) > 0); 
+#endif
+#if ITERATIONS==8
+      otmp4 = ((tmp&mask4) > 0); 
+      otmp5 = ((tmp&mask5) > 0); 
+      otmp6 = ((tmp&mask6) > 0); 
+      otmp7 = ((tmp&mask7) > 0);
+#endif 
+      for (int j=nseg0+1;j<nseg1;j++) {
+	 tmp=input[j][uint32_offset];
+	 jmod = j&31;
+         otmp0 |= (((tmp&mask0) > 0)<<jmod); 
+#if ITERATIONS==4 || ITERATIONS==8
+         otmp1 |= (((tmp&mask1) > 0)<<jmod); 
+         otmp2 |= (((tmp&mask2) > 0)<<jmod); 
+         otmp3 |= (((tmp&mask3) > 0)<<jmod);
+#endif 
+#if ITERATIONS==8
+         otmp4 |= (((tmp&mask4) > 0)<<jmod); 
+         otmp5 |= (((tmp&mask5) > 0)<<jmod); 
+         otmp6 |= (((tmp&mask6) > 0)<<jmod); 
+         otmp7 |= (((tmp&mask7) > 0)<<jmod); 
+#endif
+      }	      
+      out[0]=otmp0;
+#if ITERATIONS==4 || ITERATIONS==8
+      out[1]=otmp1;
+      out[2]=otmp2;
+      out[3]=otmp3;
+#endif
+#if ITERATIONS==8
+      out[4]=otmp4;
+      out[5]=otmp5;
+      out[6]=otmp6;
+      out[7]=otmp7;
+#endif
+  } 
+}
+__global__ void circcopy_c_worker(uint32_t **cc,uint32_t **c) {
+
+  int s = blockIdx.x;	
+  int i1 = blockIdx.y;
+  int i  = threadIdx.x;
+  uint32_t tmp;
+  if (i<384) {
+    tmp = cc[s][(i1*384) + i];	  
+    c[s][2*i1*384 + i] = tmp;
+    c[s][(2*i1+1)*384 + i] = tmp; 
+  }
+}
+
+extern "C" int ldpc_input(uint32_t **input,uint32_t *cc[4],int block_length,int nseg) { 
+
+ int numb = block_length/(NTHREADS*ITERATIONS);
+ if ((block_length%(NTHREADS*ITERATIONS)) > 0) numb++;
+ int ns = nseg>>5;
+ if ((nseg&31)>0) ns++;
+
+ dim3 numblocks(numb,ns);
+ ldpc_input_worker<<<numblocks,NTHREADS>>>(input,cc,block_length,nseg);
+ cudaError_t err=cudaPeekAtLastError();
+ if (err!=cudaSuccess) {
+    printf("cuda error: %s (input %p, cc %p, block_length %d, nseg %d, numb %d, ns %d)\n",cudaGetErrorString(err),input,cc,block_length,nseg,numb,ns);
+    exit(-1);
+ }
+ cudaDeviceSynchronize();
+ return(0);
+}
+
+extern "C" int circcopy_c(uint32_t **cc,uint32_t **c,int n_inputs) { 
+
+ dim3 numblocks(n_inputs,22);
+ circcopy_c_worker<<<numblocks,384>>>(cc,c);
+ cudaError_t err=cudaPeekAtLastError();
+ if (err!=cudaSuccess) {
+    printf("cuda error: %s (cc %p, c %p, n_inputs %d)\n",cudaGetErrorString(err),cc,c,n_inputs);
+    exit(-1);
+ }
+ cudaDeviceSynchronize();
+ return(0);
+}

openair1/PHY/CODING/nrLDPC_extern.h

@@ -29,6 +29,7 @@ typedef struct ldpc_interface_s {
   LDPC_shutdownfunc_t *LDPCshutdown;
   LDPC_decoderfunc_t *LDPCdecoder;
   LDPC_encoderfunc_t *LDPCencoder;
+  LDPC_encoderfunc32_t *LDPCencoder32;
 } ldpc_interface_t;
 
 /* functions to load the LDPC shared lib, implemented in openair1/PHY/CODING/nrLDPC_load.c */
@@ -37,6 +38,7 @@ int free_LDPClib(ldpc_interface_t *ldpc_interface);
 
 LDPC_decoderfunc_t LDPCdecoder;
 LDPC_encoderfunc_t LDPCencoder;
+LDPC_encoderfunc32_t LDPCencoder32;
 
 // inline functions:
 #endif

openair1/PHY/CODING/nrLDPC_load.c

@@ -56,14 +56,17 @@ int load_LDPClib(char *version, ldpc_interface_t *itf)
   loader_shlibfunc_t shlib_fdesc[] = {{.fname = "LDPCinit"},
                                       {.fname = "LDPCshutdown"},
                                       {.fname = "LDPCdecoder"},
-                                      {.fname = "LDPCencoder"}};
+                                      {.fname = "LDPCencoder"},
+                                      {.fname = "LDPCencoder32"}};
   int ret;
   ret = load_module_version_shlib(libname, version, shlib_fdesc, sizeofArray(shlib_fdesc), NULL);
   AssertFatal((ret >= 0), "Error loading ldpc decoder");
+  LOG_I(NR_PHY,"loading LDPC version %s\n",version);
   itf->LDPCinit = (LDPC_initfunc_t *)shlib_fdesc[0].fptr;
   itf->LDPCshutdown = (LDPC_shutdownfunc_t *)shlib_fdesc[1].fptr;
   itf->LDPCdecoder = (LDPC_decoderfunc_t *)shlib_fdesc[2].fptr;
   itf->LDPCencoder = (LDPC_encoderfunc_t *)shlib_fdesc[3].fptr;
+  itf->LDPCencoder32 = (LDPC_encoderfunc32_t *)shlib_fdesc[4].fptr;
 
   AssertFatal(itf->LDPCinit() == 0, "error starting LDPC library %s %s\n", libname, version);
 

openair1/PHY/INIT/nr_init.c

@@ -21,6 +21,7 @@
 
 #include "executables/softmodem-common.h"
 #include "executables/nr-softmodem-common.h"
+#include "executables/nr-softmodem.h"
 #include "common/utils/nr/nr_common.h"
 #include "common/ran_context.h"
 #include "PHY/defs_gNB.h"
@@ -107,6 +108,8 @@ void phy_init_nr_gNB(PHY_VARS_gNB *gNB)
   NR_gNB_COMMON *const common_vars  = &gNB->common_vars;
   NR_gNB_PRACH *const prach_vars   = &gNB->prach_vars;
 
+  gNB->use_gpu = use_gpu;
+
   common_vars->analog_bf = cfg->analog_beamforming_ve.analog_bf_vendor_ext.value;
   LOG_I(PHY, "L1 configured with%s analog beamforming\n", common_vars->analog_bf ? "" : "out");
   if (common_vars->analog_bf) {
@@ -421,7 +424,7 @@ void init_DLSCH_struct(PHY_VARS_gNB *gNB, processingData_L1tx_t *msg)
     LOG_D(PHY, "Allocating Transport Channel Buffers for DLSCH %d/%d\n", i, gNB->max_nb_pdsch);
     msg->dlsch[i] = (NR_gNB_DLSCH_t *)malloc16(num_cw * sizeof(NR_gNB_DLSCH_t));
     for (int j = 0; j < num_cw; j++) {
-      msg->dlsch[i][j] = new_gNB_dlsch(fp, grid_size);
+      msg->dlsch[i][j] = new_gNB_dlsch(fp, grid_size, gNB->use_gpu);
     }
   }
 }
@@ -434,7 +437,7 @@ void reset_DLSCH_struct(const PHY_VARS_gNB *gNB, processingData_L1tx_t *msg)
   int num_cw = NR_MAX_NB_LAYERS > 4? 2:1;
   for (int i = 0; i < gNB->max_nb_pdsch; i++) {
     for (int j = 0; j < num_cw; j++) {
-      free_gNB_dlsch(&msg->dlsch[i][j], grid_size, fp);
+      free_gNB_dlsch(&msg->dlsch[i][j], grid_size, fp, gNB->use_gpu);
     }
     free(msg->dlsch[i]);
   }

openair1/PHY/NR_ESTIMATION/nr_ul_channel_estimation.c

@@ -499,8 +499,7 @@ int nr_pusch_channel_estimation(PHY_VARS_gNB *gNB,
                                          pusch_pdu->scid,
                                          Ns,
                                          symbol);
-    pusch_dmrs_type_t dmrs_type = pusch_pdu->dmrs_config_type == NFAPI_NR_DMRS_TYPE1 ? pusch_dmrs_type1 : pusch_dmrs_type2;
-    float beta_dmrs_pusch = get_beta_dmrs_pusch(pusch_pdu->num_dmrs_cdm_grps_no_data, dmrs_type);
+    float beta_dmrs_pusch = get_beta_dmrs(pusch_pdu->num_dmrs_cdm_grps_no_data, pusch_pdu->dmrs_config_type == pusch_dmrs_type2);
     int16_t dmrs_scaling = (1 / beta_dmrs_pusch) * (1 << 14);
     nr_pusch_dmrs_rx(gNB,
                      Ns,

openair1/PHY/NR_REFSIG/nr_dmrs_rx.c

@@ -117,7 +117,8 @@ int nr_pdsch_dmrs_rx(const PHY_VARS_NR_UE *ue,
                      unsigned short p,
                      unsigned char lp,
                      unsigned short nb_pdsch_rb,
-                     uint8_t config_type)
+                     uint8_t config_type,
+                     int16_t dmrs_scaling)
 {
   typedef int array_of_w[2];
   const array_of_w *wf = (config_type == NFAPI_NR_DMRS_TYPE1) ? wf1 : wf2;
@@ -131,6 +132,7 @@ int nr_pdsch_dmrs_rx(const PHY_VARS_NR_UE *ue,
       for (int i = 0; i < nb_pdsch_rb * ((config_type == NFAPI_NR_DMRS_TYPE1) ? 6 : 4); i++) {
         int w = (wf[p - 1000][i & 1]) * (wt[p - 1000][lp]);
         output[i] = get_modulated(nr_gold_pdsch, i, w == 1);
+        output[i] = c16mulRealShift(output[i], dmrs_scaling, 14);
 
 #ifdef DEBUG_PDSCH
         printf("nr_pdsch_dmrs_rx dmrs config type %d port %d nb_pdsch_rb %d\n", config_type, p, nb_pdsch_rb);

openair1/PHY/NR_REFSIG/refsig_defs_ue.h

@@ -49,7 +49,8 @@ int nr_pdsch_dmrs_rx(const PHY_VARS_NR_UE *ue,
                      unsigned short p,
                      unsigned char lp,
                      unsigned short nb_pdsch_rb,
-                     uint8_t config_type);
+                     uint8_t config_type,
+                     int16_t dmrs_scaling);
 
 void sl_generate_pss(SL_NR_UE_INIT_PARAMS_t *sl_init_params, uint8_t n_sl_id2, uint16_t scaling);
 void sl_generate_pss_ifft_samples(sl_nr_ue_phy_params_t *sl_ue_params, SL_NR_UE_INIT_PARAMS_t *sl_init_params);

openair1/PHY/NR_TRANSPORT/nr_dlsch.c

@@ -778,9 +778,11 @@ void nr_generate_pdsch(processingData_L1tx_t *msgTx, int frame, int slot)
   NR_DL_FRAME_PARMS *frame_parms = &gNB->frame_parms;
   time_stats_t *dlsch_encoding_stats = &gNB->dlsch_encoding_stats;
   time_stats_t *tinput = &gNB->tinput;
+  time_stats_t *tinput_memcpy = &gNB->tinput_memcpy;
   time_stats_t *tprep = &gNB->tprep;
   time_stats_t *tparity = &gNB->tparity;
   time_stats_t *toutput = &gNB->toutput;
+  time_stats_t *tconcat = &gNB->tconcat;
   time_stats_t *dlsch_rate_matching_stats = &gNB->dlsch_rate_matching_stats;
   time_stats_t *dlsch_interleaving_stats = &gNB->dlsch_interleaving_stats;
   time_stats_t *dlsch_segmentation_stats = &gNB->dlsch_segmentation_stats;
@@ -827,9 +829,8 @@ void nr_generate_pdsch(processingData_L1tx_t *msgTx, int frame, int slot)
   }
 
   unsigned char output[size_output >> 3] __attribute__((aligned(64)));
-  bzero(output, sizeof(output));
-
   start_meas(dlsch_encoding_stats);
+  bzero(output, sizeof(output));
   if (nr_dlsch_encoding(gNB,
                         msgTx,
                         frame,
@@ -837,15 +838,18 @@ void nr_generate_pdsch(processingData_L1tx_t *msgTx, int frame, int slot)
                         frame_parms,
                         output,
                         tinput,
+                        tinput_memcpy,
                         tprep,
                         tparity,
                         toutput,
+                        tconcat,
                         dlsch_rate_matching_stats,
                         dlsch_interleaving_stats,
                         dlsch_segmentation_stats)
       == -1) {
     return;
   }
+
   stop_meas(dlsch_encoding_stats);
 
   unsigned char *output_ptr = output;

openair1/PHY/NR_TRANSPORT/nr_dlsch.h

@@ -49,12 +49,14 @@ int nr_dlsch_encoding(PHY_VARS_gNB *gNB,
                       NR_DL_FRAME_PARMS *frame_parms,
                       unsigned char *output,
                       time_stats_t *tinput,
+                      time_stats_t *tinput_memcpy,
                       time_stats_t *tprep,
                       time_stats_t *tparity,
                       time_stats_t *toutput,
                       time_stats_t *dlsch_rate_matching_stats,
                       time_stats_t *dlsch_interleaving_stats,
-                      time_stats_t *dlsch_segmentation_stats);
+                      time_stats_t *dlsch_segmentation_stats,
+                      time_stats_t *dlsch_output_stats);
 
 void dump_pdsch_stats(FILE *fd,PHY_VARS_gNB *gNB);
 

openair1/PHY/NR_TRANSPORT/nr_dlsch_coding.c

@@ -39,10 +39,13 @@
 #include <syscall.h>
 #include <openair2/UTIL/OPT/opt.h>
 
+#ifdef ENABLE_CUDA
+#include <cuda_runtime.h>
+#endif
 // #define DEBUG_DLSCH_CODING
 // #define DEBUG_DLSCH_FREE 1
 
-void free_gNB_dlsch(NR_gNB_DLSCH_t *dlsch, uint16_t N_RB, const NR_DL_FRAME_PARMS *frame_parms)
+void free_gNB_dlsch(NR_gNB_DLSCH_t *dlsch, uint16_t N_RB, const NR_DL_FRAME_PARMS *frame_parms, int use_gpumem)
 {
   int max_layers = (frame_parms->nb_antennas_tx < NR_MAX_NB_LAYERS) ? frame_parms->nb_antennas_tx : NR_MAX_NB_LAYERS;
   uint16_t a_segments = MAX_NUM_NR_DLSCH_SEGMENTS_PER_LAYER * max_layers;
@@ -54,7 +57,12 @@ void free_gNB_dlsch(NR_gNB_DLSCH_t *dlsch, uint16_t N_RB, const NR_DL_FRAME_PARM
 
   NR_DL_gNB_HARQ_t *harq = &dlsch->harq_process;
   if (harq->b) {
-    free16(harq->b, a_segments * 1056);
+#ifdef ENABLE_CUDA	  
+    if (use_gpumem) 
+      cudaFreeHost(harq->b);
+    else	  
+#endif  
+      free16(harq->b, a_segments * 1056);
     harq->b = NULL;
   }
   if (harq->f) {
@@ -68,7 +76,7 @@ void free_gNB_dlsch(NR_gNB_DLSCH_t *dlsch, uint16_t N_RB, const NR_DL_FRAME_PARM
   free(harq->c);
 }
 
-NR_gNB_DLSCH_t new_gNB_dlsch(NR_DL_FRAME_PARMS *frame_parms, uint16_t N_RB)
+NR_gNB_DLSCH_t new_gNB_dlsch(NR_DL_FRAME_PARMS *frame_parms, uint16_t N_RB, int use_gpumem)
 {
   int max_layers = (frame_parms->nb_antennas_tx < NR_MAX_NB_LAYERS) ? frame_parms->nb_antennas_tx : NR_MAX_NB_LAYERS;
   uint16_t a_segments = MAX_NUM_NR_DLSCH_SEGMENTS_PER_LAYER * max_layers; // number of segments to be allocated
@@ -89,16 +97,35 @@ NR_gNB_DLSCH_t new_gNB_dlsch(NR_DL_FRAME_PARMS *frame_parms, uint16_t N_RB)
   bzero(harq->b, dlsch_bytes);
 
   harq->c = (uint8_t **)malloc16(a_segments * sizeof(uint8_t *));
+#ifdef ENABLE_CUDA
+  uint8_t *c_devh[a_segments];
+#endif
   for (int r = 0; r < a_segments; r++) {
     // account for filler in first segment and CRCs for multiple segment case
     // [hna] 8448 is the maximum CB size in NR
     //       68*348 = 68*(maximum size of Zc)
     //       In section 5.3.2 in 38.212, the for loop is up to N + 2*Zc (maximum size of N is 66*Zc, therefore 68*Zc)
-    harq->c[r] = malloc16(8448);
+#ifdef ENABLE_CUDA
+    if (use_gpumem) {
+      cudaHostAlloc((void**)&harq->c[r], 8448/8, cudaHostAllocMapped);
+      cudaError_t err=cudaHostAlloc((void**)&harq->c[r],(8448/8)*sizeof(uint8_t),cudaHostAllocMapped);
+      AssertFatal(err == cudaSuccess,"CUDA Error (harq->c[%d]): %s\n", r,cudaGetErrorString(err));
+      err=cudaHostGetDevicePointer((void**)&c_devh[r], harq->c[r], 0);
+      AssertFatal(err == cudaSuccess,"CUDA Error (cudaHostGetDevicePointer) harq->c_devh[%d]: %s\n", r,cudaGetErrorString(err));
+    }
+    else 
+#endif
+      harq->c[r] = malloc16(8448/8);
     AssertFatal(harq->c[r], "cannot allocate harq->c[%d]\n", r);
-    bzero(harq->c[r], 8448);
+    bzero(harq->c[r], 8448/8);
   }
-
+#ifdef ENABLE_CUDA
+  if (use_gpumem) {
+    cudaError_t err=cudaMalloc((void**)&harq->c_dev,a_segments*sizeof(uint8_t*));
+    err=cudaMemcpy(harq->c_dev,c_devh,a_segments*sizeof(uint8_t*),cudaMemcpyHostToDevice);
+    AssertFatal(err == cudaSuccess,"CUDA Error (memcpy c_devh -> input_dev): %s\n", cudaGetErrorString(err));
+  }
+#endif
   harq->f = malloc16(N_RB * NR_SYMBOLS_PER_SLOT * NR_NB_SC_PER_RB * 8 * NR_MAX_NB_LAYERS);
   AssertFatal(harq->f, "cannot allocate harq->f\n");
   bzero(harq->f, N_RB * NR_SYMBOLS_PER_SLOT * NR_NB_SC_PER_RB * 8 * NR_MAX_NB_LAYERS);
@@ -113,9 +140,11 @@ int nr_dlsch_encoding(PHY_VARS_gNB *gNB,
                       NR_DL_FRAME_PARMS *frame_parms,
                       unsigned char *output,
                       time_stats_t *tinput,
+                      time_stats_t *tinput_memcpy,
                       time_stats_t *tprep,
                       time_stats_t *tparity,
                       time_stats_t *toutput,
+		      time_stats_t *tconcat,
                       time_stats_t *dlsch_rate_matching_stats,
                       time_stats_t *dlsch_interleaving_stats,
                       time_stats_t *dlsch_segmentation_stats)
@@ -129,10 +158,13 @@ int nr_dlsch_encoding(PHY_VARS_gNB *gNB,
                                                        .nb_TBs = msgTx->num_pdsch_slot,
                                                        .threadPool = &gNB->threadPool,
                                                        .tinput = tinput,
+                                                       .tinput_memcpy = tinput_memcpy,
                                                        .tprep = tprep,
                                                        .tparity = tparity,
                                                        .toutput = toutput,
-                                                       .TBs = TBs};
+						       .tconcat = tconcat,
+                                                       .TBs = TBs,
+  						       .use_gpu = gNB->use_gpu};
 
   int num_segments = 0;
 
@@ -171,6 +203,7 @@ int nr_dlsch_encoding(PHY_VARS_gNB *gNB,
       phy_stats->dlsch_stats.current_Qm = rel15->qamModOrder[0];
     }
 
+    start_meas(dlsch_segmentation_stats);
     int max_bytes = MAX_NUM_NR_DLSCH_SEGMENTS_PER_LAYER * rel15->nrOfLayers * 1056;
     int B;
     if (A > NR_MAX_PDSCH_TBS) {
@@ -205,7 +238,6 @@ int nr_dlsch_encoding(PHY_VARS_gNB *gNB,
     TB_parameters->BG = rel15->maintenance_parms_v3.ldpcBaseGraph;
     TB_parameters->Z = harq->Z;
     TB_parameters->A = A;
-    start_meas(dlsch_segmentation_stats);
     TB_parameters->Kb = nr_segmentation(harq->b,
                                         harq->c,
                                         B,
@@ -277,7 +309,9 @@ int nr_dlsch_encoding(PHY_VARS_gNB *gNB,
       reset_meas(&segment_parameters->ts_rate_match);
       reset_meas(&segment_parameters->ts_ldpc_encode);
     }
-
+#ifdef ENABLE_CUDA
+    if (gNB->use_gpu) TB_parameters->c_dev = (uint8_t**)harq->c_dev;
+#endif  
     segments_offset += TB_parameters->C;
 
     /* output and its parts for each dlsch should be aligned on 64 bytes (or 8 * 64 bits)

openair1/PHY/NR_TRANSPORT/nr_transport_proto.h

@@ -106,9 +106,9 @@ void nr_generate_pbch(PHY_VARS_gNB *gNB,
 void nr_init_pbch_interleaver(uint8_t *interleaver);
 uint32_t nr_pbch_extra_byte_generation(int sfn, int n_hf, int ssb_index, int ssb_sc_offset, int Lmax);
 
-NR_gNB_DLSCH_t new_gNB_dlsch(NR_DL_FRAME_PARMS *frame_parms, uint16_t N_RB);
+NR_gNB_DLSCH_t new_gNB_dlsch(NR_DL_FRAME_PARMS *frame_parms, uint16_t N_RB,int use_gpu);
 
-void free_gNB_dlsch(NR_gNB_DLSCH_t *dlsch, uint16_t N_RB, const NR_DL_FRAME_PARMS *frame_parms);
+void free_gNB_dlsch(NR_gNB_DLSCH_t *dlsch, uint16_t N_RB, const NR_DL_FRAME_PARMS *frame_parms,int use_gpu);
 
 /** \brief This function is the top-level entry point to PUSCH demodulation, after frequency-domain transformation and channel estimation.  It performs
     - RB extraction (signal and channel estimates)

openair1/PHY/NR_UE_ESTIMATION/nr_dl_channel_estimation.c

@@ -19,6 +19,7 @@
  *      contact@openairinterface.org
  */
 
+#include "nr_common.h"
 #include <string.h>
 #include "SCHED_NR_UE/defs.h"
 #include "nr_estimation.h"
@@ -448,12 +449,15 @@ int nr_prs_channel_estimation(uint8_t gNB_id,
     peak_estimator(&chT_interpol[rxAnt][0], NR_PRS_IDFT_OVERSAMP_FACTOR * frame_params->ofdm_symbol_size, &prs_toa, &ch_pwr, mean_val);
 
     // adjusting the rx_gains for channel peak power
-    ch_pwr_dbm = 10 * log10(ch_pwr) + 30 - SQ15_SQUARED_NORM_FACTOR_DB - ((int)openair0_cfg[0].rx_gain[0] - (int)openair0_cfg[0].rx_gain_offset[0]) - dB_fixed(frame_params->ofdm_symbol_size);
+    ch_pwr_dbm = 10 * log10(ch_pwr) + 30 - SQ15_SQUARED_NORM_FACTOR_DB
+                 - ((int)ue->openair0_cfg[0].rx_gain[0] - (int)ue->openair0_cfg[0].rx_gain_offset[0])
+                 - dB_fixed(frame_params->ofdm_symbol_size);
 
-    prs_meas[rxAnt]->rsrp_dBm =
-        10 * log10(prs_meas[rxAnt]->rsrp) + 30 - SQ15_SQUARED_NORM_FACTOR_DB - ((int)openair0_cfg[0].rx_gain[0] - (int)openair0_cfg[0].rx_gain_offset[0]) - dB_fixed(ue->frame_parms.ofdm_symbol_size);
+    prs_meas[rxAnt]->rsrp_dBm = 10 * log10(prs_meas[rxAnt]->rsrp) + 30 - SQ15_SQUARED_NORM_FACTOR_DB
+                                - ((int)ue->openair0_cfg[0].rx_gain[0] - (int)ue->openair0_cfg[0].rx_gain_offset[0])
+                                - dB_fixed(ue->frame_parms.ofdm_symbol_size);
 
-    //prs measurements
+    // prs measurements
     prs_meas[rxAnt]->gNB_id     = gNB_id;
     prs_meas[rxAnt]->sfn        = proc->frame_rx;
     prs_meas[rxAnt]->slot       = proc->nr_slot_rx;
@@ -1258,6 +1262,7 @@ int nr_pdsch_channel_estimation(PHY_VARS_NR_UE *ue,
                                 unsigned short scrambling_id,
                                 unsigned short BWPStart,
                                 uint8_t config_type,
+                                int n_dmrs_cdm_groups,
                                 uint16_t rb_offset,
                                 unsigned short bwp_start_subcarrier,
                                 unsigned short nb_rb_pdsch,
@@ -1291,8 +1296,10 @@ int nr_pdsch_channel_estimation(PHY_VARS_NR_UE *ue,
 
   c16_t pilot[3280] __attribute__((aligned(16)));
   // Note: pilot returned by the following function is already the complex conjugate of the transmitted DMRS
+  float beta_dmrs_pdsch = get_beta_dmrs(n_dmrs_cdm_groups, config_type == NFAPI_NR_DMRS_TYPE2);
+  int16_t dmrs_scaling = (int16_t)((1 / beta_dmrs_pdsch) * (1 << 14));
   const uint32_t *gold = nr_gold_pdsch(fp->N_RB_DL, fp->symbols_per_slot, scrambling_id, nscid, slot, symbol);
-  nr_pdsch_dmrs_rx(ue, slot, gold, pilot, 1000 + p, 0, nb_rb_pdsch + rb_offset, config_type);
+  nr_pdsch_dmrs_rx(ue, slot, gold, pilot, 1000 + p, 0, nb_rb_pdsch + rb_offset, config_type, dmrs_scaling);
 
   delay_t delay = {0};
 

openair1/PHY/NR_UE_ESTIMATION/nr_estimation.h

@@ -87,6 +87,7 @@ int nr_pdsch_channel_estimation(PHY_VARS_NR_UE *ue,
                                 unsigned short scrambling_id,
                                 unsigned short BWPStart,
                                 uint8_t config_type,
+                                int n_dmrs_cdm_groups,
                                 uint16_t rb_offset,
                                 unsigned short bwp_start_subcarrier,
                                 unsigned short nb_rb_pdsch,
@@ -151,6 +152,7 @@ float_t get_nr_RSRP(module_id_t Mod_id,uint8_t CC_id,uint8_t gNB_index);
 int nr_sl_psbch_rsrp_measurements(sl_nr_ue_phy_params_t *sl_phy_params,
                                   NR_DL_FRAME_PARMS *fp,
                                   c16_t rxdataF[][fp->samples_per_slot_wCP],
-                                  bool use_SSS);
+                                  bool use_SSS,
+                                  openair0_config_t *openair0_cfg);
 /** @}*/
 #endif

openair1/PHY/NR_UE_ESTIMATION/nr_ue_measurements.c

@@ -157,7 +157,9 @@ void nr_ue_measurements(PHY_VARS_NR_UE *ue,
     ue->measurements.rx_power_avg_dB[gNB_id] = dB_fixed( ue->measurements.rx_power_avg[gNB_id]);
     ue->measurements.wideband_cqi_tot[gNB_id] = ue->measurements.rx_power_tot_dB[gNB_id] - ue->measurements.n0_power_tot_dB;
     ue->measurements.wideband_cqi_avg[gNB_id] = ue->measurements.rx_power_avg_dB[gNB_id] - dB_fixed(ue->measurements.n0_power_avg);
-    ue->measurements.rx_rssi_dBm[gNB_id] = ue->measurements.rx_power_avg_dB[gNB_id] + 30 - SQ15_SQUARED_NORM_FACTOR_DB - ((int)openair0_cfg[0].rx_gain[0] - (int)openair0_cfg[0].rx_gain_offset[0]) - dB_fixed(ue->frame_parms.ofdm_symbol_size);
+    ue->measurements.rx_rssi_dBm[gNB_id] = ue->measurements.rx_power_avg_dB[gNB_id] + 30 - SQ15_SQUARED_NORM_FACTOR_DB
+                                           - ((int)ue->openair0_cfg[0].rx_gain[0] - (int)ue->openair0_cfg[0].rx_gain_offset[0])
+                                           - dB_fixed(ue->frame_parms.ofdm_symbol_size);
 
     LOG_D(PHY, "[gNB %d] Slot %d, RSSI %d dB (%d dBm/RE), WBandCQI %d dB, rxPwrAvg %d, n0PwrAvg %d\n",
       gNB_id,
@@ -235,7 +237,7 @@ void nr_ue_ssb_rsrp_measurements(PHY_VARS_NR_UE *ue,
   uint32_t rsrp_avg = nr_ue_calculate_ssb_rsrp(fp, proc, rxdataF, symbol_offset, fp->ssb_start_subcarrier);
   float rsrp_db_per_re = 10 * log10(rsrp_avg);
 
-  openair0_config_t *cfg0 = &openair0_cfg[0];
+  openair0_config_t *cfg0 = &ue->openair0_cfg[0];
 
   ue->measurements.ssb_rsrp_dBm[ssb_index] = rsrp_db_per_re + 30 - SQ15_SQUARED_NORM_FACTOR_DB
                                              - ((int)cfg0->rx_gain[0] - (int)cfg0->rx_gain_offset[0])
@@ -271,8 +273,8 @@ void nr_ue_rrc_measurements(PHY_VARS_NR_UE *ue,
   const uint8_t k_length = 8;
   uint8_t l_sss = (ue->symbol_offset + 2) % ue->frame_parms.symbols_per_slot;
   unsigned int ssb_offset = ue->frame_parms.first_carrier_offset + ue->frame_parms.ssb_start_subcarrier;
-  double rx_gain = openair0_cfg[0].rx_gain[0];
-  double rx_gain_offset = openair0_cfg[0].rx_gain_offset[0];
+  double rx_gain = ue->openair0_cfg[0].rx_gain[0];
+  double rx_gain_offset = ue->openair0_cfg[0].rx_gain_offset[0];
 
   ue->measurements.n0_power_tot = 0;
 
@@ -344,7 +346,8 @@ void nr_ue_rrc_measurements(PHY_VARS_NR_UE *ue,
 int nr_sl_psbch_rsrp_measurements(sl_nr_ue_phy_params_t *sl_phy_params,
                                   NR_DL_FRAME_PARMS *fp,
                                   c16_t rxdataF[][fp->samples_per_slot_wCP],
-                                  bool use_SSS)
+                                  bool use_SSS,
+                                  openair0_config_t *openair0_cfg)
 {
   SL_NR_UE_PSBCH_t *psbch_rx = &sl_phy_params->psbch;
   uint8_t numsym = (fp->Ncp) ? SL_NR_NUM_SYMBOLS_SSB_EXT_CP : SL_NR_NUM_SYMBOLS_SSB_NORMAL_CP;

openair1/PHY/NR_UE_TRANSPORT/csi_rx.c

@@ -238,7 +238,7 @@ static int nr_get_csi_rs_signal(const PHY_VARS_NR_UE *ue,
 
   *rsrp = rsrp_sum/meas_count;
   *rsrp_dBm = dB_fixed(*rsrp) + 30 - SQ15_SQUARED_NORM_FACTOR_DB
-      - ((int)openair0_cfg[0].rx_gain[0] - (int)openair0_cfg[0].rx_gain_offset[0]) - dB_fixed(ue->frame_parms.ofdm_symbol_size);
+      - ((int)ue->openair0_cfg[0].rx_gain[0] - (int)ue->openair0_cfg[0].rx_gain_offset[0]) - dB_fixed(ue->frame_parms.ofdm_symbol_size);
 
 #ifdef NR_CSIRS_DEBUG
   LOG_I(NR_PHY, "RSRP = %i (%i dBm)\n", *rsrp, *rsrp_dBm);

openair1/PHY/NR_UE_TRANSPORT/nr_dlsch_demodulation.c

@@ -80,7 +80,7 @@ unsigned char offset_mumimo_llr_drange[29][3]={{8,8,8},{7,7,7},{7,7,7},{7,7,7},{
 /* compute the MMSE up to 4x4 matrices */
 static void nr_dlsch_mmse(uint32_t rx_size_symbol,
                           unsigned char n_rx,
-                          unsigned char n_tx, // number of layer
+                          unsigned char nl, // number of layer
                           int32_t rxdataF_comp[][n_rx][rx_size_symbol * NR_SYMBOLS_PER_SLOT],
                           c16_t dl_ch_mag[][n_rx][rx_size_symbol],
                           c16_t dl_ch_magb[][n_rx][rx_size_symbol],
@@ -203,7 +203,7 @@ void nr_dlsch_scale_channel(uint32_t rx_size_symbol,
                             uint32_t len,
                             unsigned short nb_rb);
 static void nr_dlsch_detection_mrc(uint32_t rx_size_symbol,
-                                   short n_tx,
+                                   short nl,
                                    short n_rx,
                                    int32_t rxdataF_comp[][n_rx][rx_size_symbol * NR_SYMBOLS_PER_SLOT],
                                    int ***rho,
@@ -1456,7 +1456,7 @@ static void nr_dlsch_mmse(uint32_t rx_size_symbol,
                           int length,
                           uint32_t noise_var)
 {
-  uint32_t nb_rb_0 = length / 12 + ((length % 12) ? 1 : 0);
+  uint32_t nb_rb_0 = (length + 11) / 12;
   c16_t determ_fin[12 * nb_rb_0] __attribute__((aligned(32)));
 
   ///Allocate H^*H matrix elements and sub elements
@@ -1513,7 +1513,7 @@ static void nr_dlsch_mmse(uint32_t rx_size_symbol,
                     determ_fin, // determin
                     nb_rb_0,
                     fp_flag, // fixed point flag
-                    shift - (fp_flag == 1 ? 2 : 0)); // the out put is Q15
+                    shift - (fp_flag == 1 ? 1 : 0)); // the out put is Q15
 
   // multiply Matrix inversion pf H_h_H by the rx signal vector
   c16_t outtemp[12 * nb_rb_0] __attribute__((aligned(32)));
@@ -1531,7 +1531,7 @@ static void nr_dlsch_mmse(uint32_t rx_size_symbol,
                            (c16_t *)(rxdataF_comp[ctx][0] + symbol * rx_size_symbol),
                            outtemp,
                            sizeofArray(outtemp),
-                           shift - (fp_flag == 1 ? 2 : 0));
+                           shift - (fp_flag == 1 ? 1 : 0));
       nr_a_sum_b(rxdataF_zforcing[rtx], outtemp, nb_rb_0); // a = a + b
     }
 #ifdef DEBUG_DLSCH_DEMOD
@@ -1582,13 +1582,9 @@ static void nr_dlsch_mmse(uint32_t rx_size_symbol,
       dl_ch_mag128b_0[0] = mmtmpD2;
       dl_ch_mag128r_0[0] = mmtmpD2;
 
-      dl_ch_mag128_0[0] = simde_mm_mulhi_epi16(dl_ch_mag128_0[0],QAM_amp128);
-      dl_ch_mag128_0[0] = simde_mm_slli_epi16(dl_ch_mag128_0[0],1);
-
-      dl_ch_mag128b_0[0] = simde_mm_mulhi_epi16(dl_ch_mag128b_0[0],QAM_amp128b);
-      dl_ch_mag128b_0[0] = simde_mm_slli_epi16(dl_ch_mag128b_0[0],1);
-      dl_ch_mag128r_0[0] = simde_mm_mulhi_epi16(dl_ch_mag128r_0[0],QAM_amp128r);
-      dl_ch_mag128r_0[0] = simde_mm_slli_epi16(dl_ch_mag128r_0[0],1);
+      dl_ch_mag128_0[0] = simde_mm_mulhrs_epi16(dl_ch_mag128_0[0], QAM_amp128);
+      dl_ch_mag128b_0[0] = simde_mm_mulhrs_epi16(dl_ch_mag128b_0[0],QAM_amp128b);
+      dl_ch_mag128r_0[0] = simde_mm_mulhrs_epi16(dl_ch_mag128r_0[0],QAM_amp128r);
 
       determ_fin_128 += 1;
       dl_ch_mag128_0 += 1;

openair1/PHY/NR_UE_TRANSPORT/nr_initial_sync_sl.c

@@ -518,7 +518,7 @@ nr_initial_sync_t sl_nr_slss_search(PHY_VARS_NR_UE *UE, UE_nr_rxtx_proc_t *proc,
                 sync_params->DFN,
                 sync_params->slot_offset);
 
-          UE->adjust_rxgain = nr_sl_psbch_rsrp_measurements(sl_ue, frame_parms, rxdataF, false);
+          UE->adjust_rxgain = nr_sl_psbch_rsrp_measurements(sl_ue, frame_parms, rxdataF, false, UE->openair0_cfg);
 
           UE->init_sync_frame = sync_params->remaining_frames;
           result.rx_offset = sync_params->rx_offset;

openair1/PHY/NR_UE_TRANSPORT/nr_ulsch_ue.c

@@ -91,8 +91,8 @@ The function pointers are set once before calling the mapping funcion for
 all symbols based on different parameters. Then the mapping is done for
 each symbol by calling the function pointers.
 */
-static void (*map_dmrs_ptr)(const unsigned int, const c16_t *, c16_t *);
-static void (*map_data_dmrs_ptr)(const unsigned int num_cdm_no_data, const c16_t *, c16_t *);
+typedef void (*map_dmrs_func_t)(const unsigned int, const c16_t *, c16_t *);
+typedef void (*map_data_dmrs_func_t)(const unsigned int, const c16_t *, c16_t *);
 
 /*
 The following set of functions map dmrs and/or data REs in one RB based on
@@ -191,7 +191,9 @@ static void map_over_dc(const unsigned int right_dc,
                         const c16_t **ptrs,
                         const c16_t **dmrs,
                         const c16_t **data,
-                        c16_t **out)
+                        c16_t **out,
+                        map_dmrs_func_t map_data_dmrs_ptr,
+                        map_dmrs_func_t map_dmrs_ptr)
 {
   // if first RE is DC no need to map in this function
   if (right_dc == 0)
@@ -277,7 +279,9 @@ static void map_current_symbol(const nr_phy_pxsch_params_t p,
                                const c16_t *dmrs_seq,
                                const c16_t *ptrs_seq,
                                const c16_t **data,
-                               c16_t *out)
+                               c16_t *out,
+                               map_dmrs_func_t map_dmrs_ptr,
+                               map_data_dmrs_func_t map_data_dmrs_ptr)
 {
   const unsigned int abs_start_rb = p.bwp_start + p.start_rb;
   const unsigned int start_sc = (p.first_sc_offset + abs_start_rb * NR_NB_SC_PER_RB) % p.fft_size;
@@ -297,7 +301,7 @@ static void map_current_symbol(const nr_phy_pxsch_params_t p,
         // map RB at DC
         if (rb_over_dc) {
           // if DC is in middle of RB, the following function handles it.
-          map_over_dc(rb_over_dc, n_cdm, p.fft_size, dmrs_per_rb, data_per_rb, p.delta, 0, NULL, &p_mod_dmrs, NULL, &out_tmp);
+          map_over_dc(rb_over_dc, n_cdm, p.fft_size, dmrs_per_rb, data_per_rb, p.delta, 0, NULL, &p_mod_dmrs, NULL, &out_tmp, map_data_dmrs_ptr, map_dmrs_ptr);
           continue;
         } else {
           // else just move the pointer and following function will map the rb
@@ -315,7 +319,7 @@ static void map_current_symbol(const nr_phy_pxsch_params_t p,
       for (unsigned int rb = 0; rb < p.nb_rb; rb++) {
         if (rb == dc_rb) {
           if (rb_over_dc) {
-            map_over_dc(rb_over_dc, n_cdm, p.fft_size, dmrs_per_rb, data_per_rb, p.delta, 0, NULL, &p_mod_dmrs, &data_tmp, &out_tmp);
+            map_over_dc(rb_over_dc, n_cdm, p.fft_size, dmrs_per_rb, data_per_rb, p.delta, 0, NULL, &p_mod_dmrs, &data_tmp, &out_tmp, map_data_dmrs_ptr, map_dmrs_ptr);
             continue;
           } else {
             out_tmp -= p.fft_size;
@@ -339,7 +343,7 @@ static void map_current_symbol(const nr_phy_pxsch_params_t p,
       if (rb < non_ptrs_rb || ptrs_idx_rb % p.K_ptrs) {
         if (rb == dc_rb) {
           if (rb_over_dc) {
-            map_over_dc(rb_over_dc, n_cdm, p.fft_size, 0, 0, p.delta, 0, NULL, NULL, &data_tmp, &out_tmp);
+            map_over_dc(rb_over_dc, n_cdm, p.fft_size, 0, 0, p.delta, 0, NULL, NULL, &data_tmp, &out_tmp, map_data_dmrs_ptr, map_dmrs_ptr);
             continue;
           } else {
             out_tmp -= p.fft_size;
@@ -351,7 +355,7 @@ static void map_current_symbol(const nr_phy_pxsch_params_t p,
       } else {
         if (rb == dc_rb) {
           if (rb_over_dc) {
-            map_over_dc(rb_over_dc, n_cdm, p.fft_size, 0, 0, p.delta, ptrs_idx_re, &p_mod_ptrs, NULL, &data_tmp, &out_tmp);
+            map_over_dc(rb_over_dc, n_cdm, p.fft_size, 0, 0, p.delta, ptrs_idx_re, &p_mod_ptrs, NULL, &data_tmp, &out_tmp, map_data_dmrs_ptr, map_dmrs_ptr);
             continue;
           } else {
             out_tmp -= p.fft_size;
@@ -370,7 +374,7 @@ static void map_current_symbol(const nr_phy_pxsch_params_t p,
     for (unsigned int rb = 0; rb < p.nb_rb; rb++) {
       if (rb == dc_rb) {
         if (rb_over_dc) {
-          map_over_dc(rb_over_dc, n_cdm, p.fft_size, 0, 0, p.delta, 0, NULL, NULL, &data_tmp, &out_tmp);
+          map_over_dc(rb_over_dc, n_cdm, p.fft_size, 0, 0, p.delta, 0, NULL, NULL, &data_tmp, &out_tmp, map_data_dmrs_ptr, map_dmrs_ptr);
           continue;
         } else {
           out_tmp -= p.fft_size;
@@ -396,7 +400,7 @@ static void dmrs_amp_mult(const uint32_t dmrs_port,
                           const pusch_dmrs_type_t dmrs_type,
                           const unsigned int num_cdm_groups_no_data)
 {
-  float beta_dmrs_pusch = get_beta_dmrs_pusch(num_cdm_groups_no_data, dmrs_type);
+  float beta_dmrs_pusch = get_beta_dmrs(num_cdm_groups_no_data, dmrs_type == pusch_dmrs_type2);
   /* short array that hold amplitude for k_prime = 0 and k_prime = 1 */
   int32_t alpha_dmrs[2] __attribute((aligned(16)));
   for (int_fast8_t i = 0; i < sizeofArray(alpha_dmrs); i++) {
@@ -420,6 +424,8 @@ static void map_symbols(const nr_phy_pxsch_params_t p,
                         c16_t *out)
 {
   // asign the function pointers
+  map_dmrs_func_t map_dmrs_ptr = NULL;
+  map_data_dmrs_func_t map_data_dmrs_ptr = NULL;
   if (p.dmrs_type == pusch_dmrs_type1) {
     map_dmrs_ptr = map_dmrs_type1_cdm1_rb;
     map_data_dmrs_ptr = (p.num_cdm_no_data == 1) ? map_data_dmrs_type1_cdm1_rb : NULL;
@@ -460,7 +466,9 @@ static void map_symbols(const nr_phy_pxsch_params_t p,
                        mod_dmrs_amp,
                        mod_ptrs_amp,
                        &cur_data, // increments every symbol
-                       out + l * p.fft_size);
+                       out + l * p.fft_size,
+                       map_dmrs_ptr,
+                       map_data_dmrs_ptr);
   }
 }
 

openair1/PHY/TOOLS/tools_defs.h

@@ -828,6 +828,47 @@ static inline void rotate_cpx_vector(const c16_t *const x, const c16_t *const al
     // log2_amp - increase the output amplitude by a factor 2^log2_amp (default is 0)
     //            WARNING: log2_amp>0 can cause overflow!!
 
+
+#ifdef __aarch64__
+    if (output_shift == 15) { // allows specific NEON instruction
+
+      int16x8_t ar = (int16x8_t)vdupq_n_s16(alpha->r);
+      int16x8_t ai = (int16x8_t)vdupq_n_s16(alpha->i);
+      int16x8_t *y_128 = (int16x8_t *)y;
+      int16x8_t *x_128 = (int16x8_t *)x;
+      for (uint32_t i = 0; i < (N >> 2); i++) {
+        // Split interleaved -> separate real/imag
+        int16x8_t br = vuzp1q_s16(x_128[i], x_128[i]);
+        int16x8_t bi = vuzp2q_s16(x_128[i], x_128[i]);
+
+        // Start with the two “diagonal” products using high-half, doubling, sat:
+        // x = round( (2*ar*br) / 2^16 ), y = round( (2*ar*bi) / 2^16 )
+        int16x8_t real = vqdmulhq_s16(ar, br);
+        int16x8_t imag = vqdmulhq_s16(ar, bi);
+
+        // real -= round( (2*ai*bi) / 2^16 )
+        real = vqrdmlshq_s16(real, ai, bi);
+
+        // imag += round( (2*ai*br) / 2^16 )
+        imag = vqrdmlahq_s16(imag, ai, br);
+
+        // Re-interleave [real, imag]
+        int16x8x2_t z = vzipq_s16(real, imag);
+
+        y_128[i] = z.val[0];
+        /*
+        printf("y : (%d %d) (%d %d) (%d %d) (%d %d)\n",
+                 vgetq_lane_s16(y_128[i],0),
+                 vgetq_lane_s16(y_128[i],1),
+                 vgetq_lane_s16(y_128[i],2),
+                 vgetq_lane_s16(y_128[i],3),
+                 vgetq_lane_s16(y_128[i],4),
+                 vgetq_lane_s16(y_128[i],5),
+                 vgetq_lane_s16(y_128[i],6),
+                 vgetq_lane_s16(y_128[i],7));*/
+      }
+    } else {
+#endif
     uint32_t i; // loop counter
 
     simd_q15_t *y_128, alpha_128;
@@ -859,7 +900,10 @@ static inline void rotate_cpx_vector(const c16_t *const x, const c16_t *const al
               shift));
       // print_ints("y_128[0]=", &y_128[0]);
     }
-#if defined(__x86__) || defined(__x86_64__)
+#ifdef __aarch64__
+    }
+#endif //__aarch64__
+#if defined(__x86_64__) || defined(__i386__)
   }
 #endif
 }

openair1/PHY/defs_gNB.h

@@ -62,6 +62,9 @@ typedef struct {
   uint8_t *b;
   /// Pointers to transport block segments
   uint8_t **c;
+#ifdef ENABLE_CUDA
+  uint8_t *c_dev;
+#endif
   /// Frame where current HARQ round was sent
   uint32_t frame;
   /// Interleaver outputs
@@ -504,9 +507,11 @@ typedef struct PHY_VARS_gNB_s {
   time_stats_t dlsch_resource_mapping_stats;
   time_stats_t dlsch_precoding_stats;
   time_stats_t tinput;
+  time_stats_t tinput_memcpy;
   time_stats_t tprep;
   time_stats_t tparity;
   time_stats_t toutput;
+  time_stats_t tconcat;
   
   time_stats_t dlsch_rate_matching_stats;
   time_stats_t dlsch_interleaving_stats;
@@ -556,6 +561,7 @@ typedef struct PHY_VARS_gNB_s {
   int L1_tx_thread_core;
   struct processingData_L1tx *msgDataTx;
   void *scopeData;
+  int use_gpu;
 } PHY_VARS_gNB;
 
 struct puschSymbolReqId {

openair1/PHY/defs_nr_UE.h

@@ -336,6 +336,7 @@ typedef struct {
 
 /// Top-level PHY Data Structure for UE
 typedef struct PHY_VARS_NR_UE_s {
+  openair0_config_t openair0_cfg[MAX_CARDS];
   /// \brief Module ID indicator for this instance
   uint8_t Mod_id;
   /// \brief Component carrier ID for this PHY instance

openair1/SCHED_NR_UE/phy_procedures_nr_ue.c

@@ -553,6 +553,7 @@ static int nr_ue_pdsch_procedures(PHY_VARS_NR_UE *ue,
                                       dlschCfg->dlDmrsScramblingId,
                                       dlschCfg->BWPStart,
                                       dlschCfg->dmrsConfigType,
+                                      dlschCfg->n_dmrs_cdm_groups,
                                       dlschCfg->rb_offset,
                                       ue->frame_parms.first_carrier_offset + (dlschCfg->BWPStart + dlschCfg->start_rb) * 12,
                                       dlschCfg->number_rbs,

openair1/SCHED_NR_UE/phy_procedures_nr_ue_sl.c

@@ -227,7 +227,7 @@ int psbch_pscch_processing(PHY_VARS_NR_UE *ue, const UE_nr_rxtx_proc_t *proc, nr
       sym = (sym == 0) ? 5 : sym + 1;
     }
 
-    ue->adjust_rxgain = nr_sl_psbch_rsrp_measurements(sl_phy_params, fp, rxdataF, false);
+    ue->adjust_rxgain = nr_sl_psbch_rsrp_measurements(sl_phy_params, fp, rxdataF, false, ue->openair0_cfg);
 
     LOG_D(NR_PHY, " ------  Decode SL-MIB: frame.slot %d.%d ------  \n", frame_rx % 1024, nr_slot_rx);
 

openair1/SIMULATION/NR_PHY/dlschsim.c

@@ -516,7 +516,7 @@ int main(int argc, char **argv)
         bzero(output, sizeof(output));
 	if (input_fd == NULL) {
     msgDataTx.num_pdsch_slot = 1;
-	  nr_dlsch_encoding(gNB, &msgDataTx, frame, slot, frame_parms, output, NULL, NULL, NULL, NULL, NULL, NULL, NULL);
+	  nr_dlsch_encoding(gNB, &msgDataTx, frame, slot, frame_parms, output, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL);
 	}
 
 	for (SNR = snr0; SNR < snr1 && !stop; SNR += snr_step) {

openair1/SIMULATION/NR_PHY/dlsim.c

@@ -182,6 +182,7 @@ extern void fix_scd(NR_ServingCellConfig_t *scd);// forward declaration
 /* specific dlsim DL preprocessor: uses rbStart/rbSize/mcs/nrOfLayers from command line of dlsim */
 int g_mcsIndex = -1, g_mcsTableIdx = 0, g_rbStart = -1, g_rbSize = -1, g_nrOfLayers = 1, g_pmi = 0;
 
+
 void nr_dlsim_preprocessor(module_id_t module_id, frame_t frame, slot_t slot)
 {
   NR_UE_info_t *UE_info = RC.nrmac[module_id]->UE_info.connected_ue_list[0];
@@ -296,8 +297,9 @@ void validate_input_pmi(nfapi_nr_config_request_scf_t *gNB_config,
               num_antenna_ports, pmi_pdu->num_ant_ports, pmi);
 }
 
-
 configmodule_interface_t *uniqCfg = NULL;
+extern uint32_t use_gpu;
+
 int main(int argc, char **argv)
 {
   stop = false;
@@ -393,7 +395,7 @@ int main(int argc, char **argv)
 
   FILE *scg_fd=NULL;
 
-  while ((c = getopt(argc, argv, "--:O:f:hA:p:f:g:i:n:s:S:t:v:x:y:z:o:H:M:N:F:GR:d:PI:L:a:b:e:m:w:T:U:q:X:Y:Z:")) != -1) {
+  while ((c = getopt(argc, argv, "--:O:f:hA:p:f:g:i:n:s:S:t:v:x:y:z:o:H:M:N:F:GR:d:PQI:L:a:b:e:m:w:T:U:q:X:Y:Z:")) != -1) {
 
     /* ignore long options starting with '--', option '-O' and their arguments that are handled by configmodule */
     /* with this opstring getopt returns 1 for non-option arguments, refer to 'man 3 getopt' */
@@ -513,7 +515,10 @@ int main(int argc, char **argv)
       print_perf=1;
       cpu_meas_enabled = 1;
       break;
-      
+     
+    case 'Q':
+      use_gpu=1;
+      break; 
     case 'I':
       max_ldpc_iterations = atoi(optarg);
       break;
@@ -924,8 +929,7 @@ printf("%d\n", slot);
 
   init_nr_ue_transport(UE);
 
-  nr_l2_init_ue(1);
-  UE_mac = get_mac_inst(0);
+  UE_mac = nr_l2_init_ue(0);
   ue_init_config_request(UE_mac, get_slots_per_frame_from_scs(mu));
 
   UE->if_inst = nr_ue_if_module_init(0);
@@ -1033,12 +1037,15 @@ printf("%d\n", slot);
     reset_meas(&gNB->dlsch_rate_matching_stats);
     reset_meas(&gNB->dlsch_segmentation_stats);
     reset_meas(&gNB->dlsch_modulation_stats);
+    reset_meas(&gNB->dlsch_precoding_stats);
     reset_meas(&gNB->dlsch_encoding_stats);
     reset_meas(&gNB->dci_generation_stats);
     reset_meas(&gNB->tinput);
+    reset_meas(&gNB->tinput_memcpy);
     reset_meas(&gNB->tprep);
     reset_meas(&gNB->tparity);
     reset_meas(&gNB->toutput);
+    reset_meas(&gNB->tconcat);
     reset_meas(&gNB->phase_comp_stats);
 
     uint32_t errors_scrambling[16] = {0};
@@ -1337,12 +1344,19 @@ printf("%d\n", slot);
       printStatIndent2(&gNB->dci_generation_stats, "DCI encoding time");
       printStatIndent2(&gNB->dlsch_encoding_stats,"DLSCH encoding time");
       printStatIndent3(&gNB->dlsch_segmentation_stats,"DLSCH segmentation time");
+      gNB->tinput.trials = gNB->dlsch_encoding_stats.trials;
       printStatIndent3(&gNB->tinput,"DLSCH LDPC input processing time");
-      printStatIndent3(&gNB->tprep,"DLSCH LDPC input preparation time");
+      gNB->tinput_memcpy.trials = gNB->dlsch_encoding_stats.trials;
+      printStatIndent3(&gNB->tinput_memcpy,"DLSCH LDPC input memcpy time");
+      gNB->tparity.trials = gNB->dlsch_encoding_stats.trials;
       printStatIndent3(&gNB->tparity,"DLSCH LDPC parity generation time");
+      gNB->toutput.trials = gNB->dlsch_encoding_stats.trials;
       printStatIndent3(&gNB->toutput,"DLSCH LDPC output generation time");
+      gNB->dlsch_rate_matching_stats.trials = gNB->dlsch_encoding_stats.trials;
       printStatIndent3(&gNB->dlsch_rate_matching_stats,"DLSCH Rate Matching time");
+      gNB->dlsch_interleaving_stats.trials = gNB->dlsch_encoding_stats.trials;
       printStatIndent3(&gNB->dlsch_interleaving_stats,  "DLSCH Interleaving time");
+      printStatIndent3(&gNB->tconcat,  "DLSCH Segment Concatenation time");
       printStatIndent2(&gNB->dlsch_modulation_stats,"DLSCH modulation time");
       printStatIndent2(&gNB->dlsch_scrambling_stats, "DLSCH scrambling time");
       printStatIndent2(&gNB->dlsch_precoding_stats,"DLSCH Mapping/Precoding time");

openair1/SIMULATION/NR_PHY/nr_unitary_common.c

@@ -32,6 +32,7 @@
 
 // Define signal handler to attempt graceful termination
 bool stop = false;
+uint32_t use_gpu=0;
 
 void sigint_handler(int arg)
 {