From e4d3bf9f4135057bac709eee1c1c4f899e7e88e9 Mon Sep 17 00:00:00 2001 From: frtabu Date: Tue, 23 Nov 2021 14:55:18 +0100 Subject: [PATCH 01/23] Fixes after MR 1257 1225: sanitize address error introduced in ldpctest when using alternatives ldpc libraries and LOG flooding introduced in nrUE. Also improved openCL ldpc library build --- cmake_targets/CMakeLists.txt | 5 + common/config/config_load_configmodule.c | 9 +- .../CODING/nrLDPC_decoder/nrLDPC_decoder_CL.c | 67 ++-- .../nrLDPC_decoder_kernels_CL.cl | 302 ------------------ openair1/PHY/CODING/nrLDPC_load.c | 4 +- openair1/PHY/INIT/nr_parms.c | 4 +- .../NR_UE_TRANSPORT/nr_dlsch_demodulation.c | 46 +-- .../PHY/NR_UE_TRANSPORT/nr_initial_sync.c | 4 +- openair1/PHY/TOOLS/time_meas.c | 17 - openair1/PHY/TOOLS/time_meas.h | 19 +- 10 files changed, 92 insertions(+), 385 deletions(-) delete mode 100644 openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_decoder_kernels_CL.cl diff --git a/cmake_targets/CMakeLists.txt b/cmake_targets/CMakeLists.txt index f513ac0563..af3baa2584 100644 --- a/cmake_targets/CMakeLists.txt +++ b/cmake_targets/CMakeLists.txt @@ -1531,6 +1531,10 @@ set(PHY_LDPC_CL_SRC ${OPENAIR1_DIR}/PHY/CODING/nrLDPC_decoder/nrLDPC_decoder_CL.c ${OPENAIR1_DIR}/PHY/CODING/nrLDPC_encoder/ldpc_encoder_optim8segmulti.c ) +add_custom_target( nrLDPC_decoder_kernels_CL + COMMAND gcc ${OPENAIR1_DIR}/PHY/CODING/nrLDPC_decoder/nrLDPC_decoder_kernels_CL.c -dD -DNRLDPC_KERNEL_SOURCE -E -o ${CMAKE_CURRENT_BINARY_DIR}/nrLDPC_decoder_kernels_CL.clc + SOURCES ${OPENAIR1_DIR}/PHY/CODING/nrLDPC_decoder/nrLDPC_decoder_kernels_CL.c +) set(PHY_NR_CODINGIF ${OPENAIR1_DIR}/PHY/CODING/nrLDPC_load.c; @@ -1541,6 +1545,7 @@ add_library(ldpc_optim MODULE ${PHY_LDPC_OPTIM_SRC} ) add_library(ldpc_optim8seg MODULE ${PHY_LDPC_OPTIM8SEG_SRC} ) add_library(ldpc_cl MODULE ${PHY_LDPC_CL_SRC} ) target_link_libraries(ldpc_cl OpenCL) +add_dependencies(ldpc_cl nrLDPC_decoder_kernels_CL) if (CUDA_FOUND) cuda_add_library(ldpc_cuda MODULE ${PHY_LDPC_CUDA_SRC} ) diff --git a/common/config/config_load_configmodule.c b/common/config/config_load_configmodule.c index 740c6e240b..f98b1db167 100644 --- a/common/config/config_load_configmodule.c +++ b/common/config/config_load_configmodule.c @@ -219,11 +219,9 @@ configmodule_interface_t *load_configmodule(int argc, cfgparam = getenv("OAI_CONFIGMODULE"); } - /* default different for UE and softmodem because UE doesn't use config file*/ + /* default different for UE and softmodem because UE may run without config file */ /* and -O option is not mandatory for UE */ /* phy simulators behave as UE */ - /* test of exec name would better be replaced by a parameter to the l */ - /* oad_configmodule function */ if (cfgparam == NULL) { tmpflags = tmpflags | CONFIG_NOOOPT; @@ -249,9 +247,14 @@ configmodule_interface_t *load_configmodule(int argc, } cfgptr = calloc(sizeof(configmodule_interface_t),1); + /* argv_info is used to memorize command line options which have been recognized */ + /* and to detect unrecognized command line options which might have been specified */ cfgptr->argv_info = calloc(sizeof(int32_t), argc); + /* argv[0] is the exec name, always Ok */ cfgptr->argv_info[0] |= CONFIG_CMDLINEOPT_PROCESSED; + /* when OoptIdx is >0, -O option has been detected at position OoptIdx + * we must memorize arv[OoptIdx is Ok */ if (OoptIdx >= 0) { cfgptr->argv_info[OoptIdx] |= CONFIG_CMDLINEOPT_PROCESSED; cfgptr->argv_info[OoptIdx+1] |= CONFIG_CMDLINEOPT_PROCESSED; diff --git a/openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_decoder_CL.c b/openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_decoder_CL.c index 58e6b507af..c60286ab3a 100644 --- a/openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_decoder_CL.c +++ b/openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_decoder_CL.c @@ -25,15 +25,7 @@ * \version 1.0 * \note initial implem - translation of cuda version */ -/* uses HW component id for log messages ( --log_config.hw_log_level ) */ -#include -#include -#include -#include -#include "PHY/CODING/nrLDPC_decoder/nrLDPC_types.h" -#include "PHY/CODING/nrLDPC_decoder/nrLDPCdecoder_defs.h" -#include "assertions.h" -#include "common/utils/LOG/log.h" + #define MAX_ITERATION 2 #define MC 1 @@ -41,16 +33,29 @@ #define MAX_OCLDEV 10 #define MAX_OCLRUNTIME 5 - -#define CLSETKERNELARG(A,B,C,D) \ -rt=clSetKernelArg(A,B,C,D) ;\ -AssertFatal(rt == CL_SUCCESS, "Error %d setting kernel argument index %d\n" , (int)rt, B); - typedef struct{ char x; char y; short value; } h_element; + +#ifdef NRLDPC_KERNEL_SOURCE +#include "nrLDPC_decoder_kernels_CL.c" +#else +/* uses HW component id for log messages ( --log_config.hw_log_level ) */ +#include +#include +#include +#include +#include "PHY/CODING/nrLDPC_decoder/nrLDPC_types.h" +#include "PHY/CODING/nrLDPC_decoder/nrLDPCdecoder_defs.h" +#include "assertions.h" +#include "common/utils/LOG/log.h" + +#define CLSETKERNELARG(A,B,C,D) \ +rt=clSetKernelArg(A,B,C,D) ;\ +AssertFatal(rt == CL_SUCCESS, "Error %d setting kernel argument index %d\n" , (int)rt, B); + #include "../nrLDPC_decoder_LYC/bgs/BG1_compact_in_C.h" typedef struct{ @@ -213,20 +218,31 @@ void get_CompilErr(cl_program program, int pltf) { } -size_t load_source(char **source_str) { - int MAX_SOURCE_SIZE=(500*132); +size_t load_source(char **source_str, char *filename) { FILE *fp; + struct stat st ; size_t source_size; - - fp = fopen("nrLDPC_decoder_kernels_CL.cl", "r"); - AssertFatal(fp,"failed to open cl source: %s\n",strerror(errno)); + char *src= NULL; - *source_str = (char*)malloc(MAX_SOURCE_SIZE); - source_size = fread( *source_str, 1, MAX_SOURCE_SIZE, fp); - fclose( fp ); - return source_size; + if (filename == NULL) { + src = "nrLDPC_decoder_kernels_CL.clc"; + } else { + src = filename; + } + fp = fopen(src, "r"); + AssertFatal(fp,"failed to open cl source %s: %s\n",src,strerror(errno)); + + fstat(fileno(fp), &st); + source_size = st.st_size; + *source_str = (char*)malloc(source_size); + source_size = fread( *source_str, 1, source_size, fp); + fclose( fp ); + LOG_I(HW,"Loaded kernel sources from %s %u bytes\n", (filename==NULL)?"embedded cl code":src,(unsigned int)source_size ); + return source_size; } + + /* from here: entry points in decoder shared lib */ int ldpc_autoinit(void) { // called by the library loader cl_platform_id platforms[10]; @@ -284,11 +300,11 @@ int ldpc_autoinit(void) { // called by the library loader ocl.runtime[i].dev_tmp = clCreateBuffer(ocl.runtime[i].context, CL_MEM_READ_ONLY|CL_MEM_HOST_WRITE_ONLY, 68*384, NULL, (cl_int *)&rt); AssertFatal(rt == CL_SUCCESS, "Error %d creating buffer dev_tmp for platform %i \n" , (int)rt, i); char *source_str; - size_t source_size=load_source(&source_str); + size_t source_size=load_source(&source_str,"nrLDPC_decoder_kernels_CL.clc"); cl_program program = clCreateProgramWithSource(ocl.runtime[i].context, 1, (const char **)&source_str, (const size_t *)&source_size, (cl_int *)&rt); AssertFatal(rt == CL_SUCCESS, "Error %d creating program for platform %i \n" , (int)rt, i); - rt = clBuildProgram(program, ocl.runtime[i].num_devices,ocl.runtime[i].devices, NULL, NULL, NULL); + rt = clBuildProgram(program, ocl.runtime[i].num_devices,ocl.runtime[i].devices, NULL /* compile options */, NULL, NULL); if (rt == CL_BUILD_PROGRAM_FAILURE) { get_CompilErr(program,i); } @@ -434,3 +450,4 @@ int32_t nrLDPC_decod(t_nrLDPC_dec_params* p_decParams, int8_t* p_llr, int8_t* p_ return MAX_ITERATION; } +#endif //NRLDPC_KERNEL_SOURCE diff --git a/openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_decoder_kernels_CL.cl b/openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_decoder_kernels_CL.cl deleted file mode 100644 index d5e2fa5695..0000000000 --- a/openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_decoder_kernels_CL.cl +++ /dev/null @@ -1,302 +0,0 @@ -/* - * 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_decoder_kernels_CL.cl -* \brief kernel functions for ldpc decoder accelerated via openCL -* \author Francois TABURET -* \date 2021 -* \version 1.0 -* \company Nokia BellLabs France -* \email: francois.taburet@nokia-bell-labs.com -* \note initial implem - translation of cuda version -* \warning -*/ -#define define MAX_ITERATION 2 -#define MC 1 - -#define INT32_MAX 2147483647 - -typedef struct{ - char x; - char y; - short value; -} h_element; - -//__global char dev_dt [46*68*384]; -//__local char *dev_t; -//__global char dev_llr[68*384]; -//__global unsigned char dev_tmp[68*384]; - - - - -//__constant h_element dev_h_compact1[46*19] = {}; // used in kernel 1 -//__constant h_element dev_h_compact2[68*30] = {}; // used in kernel 2 - -// __device__ __constantant__ h_element dev_h_compact1[46*19]; // used in kernel 1 -// __device__ __constantant__ h_element dev_h_compact2[68*30]; // used in kernel 2 - -// row and col element count -__constant char h_ele_row_bg1_count[46] = { - 19, 19, 19, 19, 3, 8, 9, 7, 10, 9, - 7, 8, 7, 6, 7, 7, 6, 6, 6, 6, - 6, 6, 5, 5, 6, 5, 5, 4, 5, 5, - 5, 5, 5, 5, 5, 5, 5, 4, 5, 5, - 4, 5, 4, 5, 5, 4}; -__constant char h_ele_col_bg1_count[68] = { - 30, 28, 7, 11, 9, 4, 8, 12, 8, 7, - 12, 10, 12, 11, 10, 7, 10, 10, 13, 7, - 8, 11, 12, 5, 6, 6, 1, 1, 1, 1, - 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, - 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, - 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, - 1, 1, 1, 1, 1, 1, 1, 1}; -__constant char h_ele_row_bg2_count[42] = { - 8, 10, 8, 10, 4, 6, 6, 6, 4, 5, - 5, 5, 4, 5, 5, 4, 5, 5, 4, 4, - 4, 4, 3, 4, 4, 3, 5, 3, 4, 3, - 5, 3, 4, 4, 4, 4, 4, 3, 4, 4, - 4, 4}; -__constant char h_ele_col_bg2_count[52] = { - 22, 23, 10, 5, 5, 14, 7, 13, 6, 8, - 9, 16, 9, 12, 1, 1, 1, 1, 1, 1, - 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, - 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, - 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, - 1, 1}; - - - -// Kernel 1 -__kernel void ldpc_cnp_kernel_1st_iter( __global char * dev_llr, __global char * dev_dt, __local h_element *dev_h_compact1, int BG, int row, int col, int Zc) -{ -// int iMCW = blockIdx.y; // codeword id -// int iBlkRow = blockIdx.x; // block row in h_base -// int iSubRow = threadIdx.x; // row index in sub_block of h_base -// if(blockIdx.x == 0 && threadIdx.x == 1) printf("cnp %d\n", threadIdx.x); - int iMCW = get_group_id(1); // codeword id - int iBlkRow = get_group_id(0); // block row in h_base - int iBlkCol; // block col in h_base - int iSubRow = get_local_id(0);; // row index in sub_block of h_base - int iCol; // overall col index in h_base - int offsetR; - int shift_t; - -// For 2-min algorithm. - int Q_sign = 0; - int sq; - int Q, Q_abs; - int R_temp; - - int sign = 1; - int rmin1 = INT32_MAX; - int rmin2 = INT32_MAX; - char idx_min = 0; - - h_element h_element_t; - int s = (BG==1)? h_ele_row_bg1_count[iBlkRow]:h_ele_row_bg2_count[iBlkRow]; - offsetR = (iMCW * row*col*Zc) + iBlkRow * Zc + iSubRow; // row*col*Zc = size of dev_dt -// if(blockIdx.x == 0 && threadIdx.x == 1) printf("s: %d, offset %d\n", s, offsetR); -// The 1st recursion - for(int i = 0; i < s; i++) // loop through all the ZxZ sub-blocks in a row - { - h_element_t = dev_h_compact1[i*row+iBlkRow]; // compact_col == row - - iBlkCol = h_element_t.y; - shift_t = h_element_t.value; - - shift_t = (iSubRow + shift_t) % Zc; - iCol = (iMCW * col*Zc) + iBlkCol * Zc + shift_t; // col*Zc = size of llr - Q = dev_llr[iCol]; - Q_abs = (Q>0)? Q : -Q; - sq = Q < 0; -// if(blockIdx.x == 0 && threadIdx.x == 1) printf("i %d, icol %d, Q: %d\n", i, iCol, Q); - // quick version - sign = sign * (1 - sq * 2); - Q_sign |= sq << i; - - if (Q_abs < rmin1){ - rmin2 = rmin1; - rmin1 = Q_abs; - idx_min = i; - } else if (Q_abs < rmin2){ - rmin2 = Q_abs; - } - } - -// if(blockIdx.x == 0 && threadIdx.x == 1)printf("min1 %d, min2 %d, min1_idx %d\n", rmin1, rmin2, idx_min); - -// The 2nd recursion - for(int i = 0; i < s; i++){ - // v0: Best performance so far. 0.75f is the value of alpha. - sq = 1 - 2 * ((Q_sign >> i) & 0x01); - R_temp = 0.75f * sign * sq * (i != idx_min ? rmin1 : rmin2); - // write results to global memory - h_element_t = dev_h_compact1[i*row+iBlkRow]; - int addr_temp = offsetR + h_element_t.y * row * Zc; - dev_dt[addr_temp] = R_temp; -// if(blockIdx.x == 0 && threadIdx.x == 1)printf("R_temp %d, temp_addr %d\n", R_temp, addr_temp); - } -} - -// Kernel_1 -__kernel void ldpc_cnp_kernel( __global char * dev_llr, __global char * dev_dt, __local h_element *dev_h_compact1, int BG, int row, int col, int Zc) -{ -// if(blockIdx.x == 0 && threadIdx.x == 1) printf("cnp\n"); -// int iMCW = blockIdx.y; -// int iBlkRow = blockIdx.x; // block row in h_base // block col in h_base -// int iSubRow = threadIdx.x; // row index in sub_block of h_base - int iMCW = get_group_id(1); - int iBlkRow = get_group_id(0); // block row in h_base - int iBlkCol; // block col in h_base - int iSubRow = get_local_id(0);; // row index in sub_block of h_base - int iCol; // overall col index in h_base - int offsetR; - int shift_t; - -// For 2-min algorithm. - int Q_sign = 0; - int sq; - int Q, Q_abs; - int R_temp; - - int sign = 1; - int rmin1 = INT32_MAX; - int rmin2 = INT32_MAX; - char idx_min = 0; - - h_element h_element_t; - int s = (BG==1)? h_ele_row_bg1_count[iBlkRow]: h_ele_row_bg2_count[iBlkRow]; - offsetR = (iMCW *row*col*Zc) + iBlkRow * Zc + iSubRow; // row * col * Zc = size of dev_dt -// if(blockIdx.x == 0 && threadIdx.x == 1) printf("s: %d, offset %d\n", s, offsetR); -// The 1st recursion - for(int i = 0; i < s; i++) // loop through all the ZxZ sub-blocks in a row - { - h_element_t = dev_h_compact1[i*row+iBlkRow]; - - iBlkCol = h_element_t.y; - shift_t = h_element_t.value; - shift_t = (iSubRow + shift_t) % Zc; - iCol = iBlkCol * Zc + shift_t; - - R_temp = dev_dt[offsetR + iBlkCol * row * Zc]; - - Q = dev_llr[iMCW * (col*Zc) + iCol] - R_temp; - Q_abs = (Q>0)? Q : -Q; -// if(blockIdx.x == 0 && threadIdx.x == 1) printf("i %d, icol %d, Q: %d\n", i, iCol, Q); - sq = Q < 0; - sign = sign * (1 - sq * 2); - Q_sign |= sq << i; - - if (Q_abs < rmin1){ - rmin2 = rmin1; - rmin1 = Q_abs; - idx_min = i; - } else if (Q_abs < rmin2){ - rmin2 = Q_abs; - } - - } - -// if(blockIdx.x == 0 && threadIdx.x == 1)printf("min1 %d, min2 %d, min1_idx %d\n", rmin1, rmin2, idx_min); - -// The 2nd recursion - for(int i = 0; i < s; i ++){ - sq = 1 - 2 * ((Q_sign >> i) & 0x01); - R_temp = 0.75f * sign * sq * (i != idx_min ? rmin1 : rmin2); - - - // write results to global memory - h_element_t = dev_h_compact1[i*row+iBlkRow]; - int addr_temp = h_element_t.y * row * Zc + offsetR; - dev_dt[addr_temp] = R_temp; -// if(blockIdx.x == 0 && threadIdx.x == 1)printf("R_temp %d, temp_addr %d\n", R_temp, addr_temp); - } -} - -// Kernel 2: VNP processing -__kernel void -ldpc_vnp_kernel_normal(__global char * dev_llr, __global char * dev_dt, __global char * dev_const_llr, __local h_element *dev_h_compact2, int BG, int row, int col, int Zc) -{ -// int iMCW = blockIdx.y; -// int iBlkCol = blockIdx.x; -// int iSubCol = threadIdx.x; - int iMCW = get_group_id(1); - int iBlkCol = get_group_id(0); - int iBlkRow; - int iSubCol = get_local_id(0); - int iRow; - int iCol; - - int shift_t, sf; - int APP; - - h_element h_element_t; - - // update all the llr values - iCol = iBlkCol * Zc + iSubCol; - APP = dev_const_llr[iMCW *col*Zc + iCol]; - int offsetDt = iMCW *row*col*Zc + iBlkCol * row * Zc; - int s = (BG==1)? h_ele_col_bg1_count[iBlkCol]:h_ele_col_bg2_count[iBlkCol]; - - for(int i = 0; i < s; i++) - { - h_element_t = dev_h_compact2[i*col+iBlkCol]; - - shift_t = h_element_t.value%Zc; - iBlkRow = h_element_t.x; - - sf = iSubCol - shift_t; - sf = (sf + Zc) % Zc; - - iRow = iBlkRow * Zc + sf; - APP = APP + dev_dt[offsetDt + iRow]; - } - if(APP > SCHAR_MAX) APP = SCHAR_MAX; - if(APP < SCHAR_MIN) APP = SCHAR_MIN; - // write back to device global memory - dev_llr[iMCW *col*Zc + iCol] = APP; -} - - -__kernel void pack_decoded_bit(__global unsigned char * dev_llr, __global unsigned char * dev_tmp, int col, int Zc) -{ -// int iMCW = blockIdx.y; -// int btid = threadIdx.x; - unsigned char tmp[128]; - int iMCW = get_group_id(1); - int btid = get_local_id(0); - int tid = iMCW * col*Zc + get_group_id(0)*128 + btid; - tmp[btid] = 0; - - if(dev_llr[tid] < 0){ - tmp[btid] = 1 << (7-(btid&7)); - } -// __syncthreads(); - - if(btid < 16){ - dev_tmp[iMCW * col*Zc + get_group_id(0)*16+btid] = 0; - for(int i = 0; i < 8; i++){ - dev_tmp[iMCW * col*Zc + get_group_id(0)*16+btid] += tmp[btid*8+i]; - } - } -} - diff --git a/openair1/PHY/CODING/nrLDPC_load.c b/openair1/PHY/CODING/nrLDPC_load.c index 23e5032bad..3049070cd6 100644 --- a/openair1/PHY/CODING/nrLDPC_load.c +++ b/openair1/PHY/CODING/nrLDPC_load.c @@ -46,14 +46,14 @@ static loader_shlibfunc_t shlib_fdesc[3]; /* arguments used when called from phy simulators exec's which do not use the config module */ /* arg is used to initialize the config module so that the loader works as expected */ -char *arg[64]={"ldpctest","-O","cmdlineonly::dbgl0",NULL,NULL}; +char *arg[64]={"ldpctest",NULL}; int load_nrLDPClib(char *version) { char *ptr = (char*)config_get_if(); char libname[64]="ldpc"; if ( ptr==NULL ) {// phy simulators, config module possibly not loaded - load_configmodule(0,(char **)NULL,CONFIG_ENABLECMDLINEONLY) ; + load_configmodule(1,arg,CONFIG_ENABLECMDLINEONLY) ; logInit(); } shlib_fdesc[0].fname = "nrLDPC_decod"; diff --git a/openair1/PHY/INIT/nr_parms.c b/openair1/PHY/INIT/nr_parms.c index 347155ed3b..0c0d38ea0b 100644 --- a/openair1/PHY/INIT/nr_parms.c +++ b/openair1/PHY/INIT/nr_parms.c @@ -250,7 +250,7 @@ int nr_init_frame_parms(nfapi_nr_config_request_scf_t* cfg, int Ncp = NFAPI_CP_NORMAL; int mu = cfg->ssb_config.scs_common.value; -#if DISABLE_LOG_X +#if PHYSIM printf("Initializing frame parms for mu %d, N_RB %d, Ncp %d\n",mu, fp->N_RB_DL, Ncp); #else LOG_I(PHY,"Initializing frame parms for mu %d, N_RB %d, Ncp %d\n",mu, fp->N_RB_DL, Ncp); @@ -343,7 +343,7 @@ int nr_init_frame_parms_ue(NR_DL_FRAME_PARMS *fp, AssertFatal(fp->ul_CarrierFreq == (fp->dl_CarrierFreq + uplink_frequency_offset), "Disagreement in uplink frequency for band %d: ul_CarrierFreq = %lu Hz vs expected %lu Hz\n", fp->nr_band, fp->ul_CarrierFreq, fp->dl_CarrierFreq + uplink_frequency_offset); -#if DISABLE_LOG_X +#if PHYSIM printf("Initializing UE frame parms for mu %d, N_RB %d, Ncp %d\n",fp->numerology_index, fp->N_RB_DL, Ncp); #else LOG_I(PHY,"Initializing frame parms for mu %d, N_RB %d, Ncp %d\n",fp->numerology_index, fp->N_RB_DL, Ncp); diff --git a/openair1/PHY/NR_UE_TRANSPORT/nr_dlsch_demodulation.c b/openair1/PHY/NR_UE_TRANSPORT/nr_dlsch_demodulation.c index 0894ae9166..31b3f8896a 100644 --- a/openair1/PHY/NR_UE_TRANSPORT/nr_dlsch_demodulation.c +++ b/openair1/PHY/NR_UE_TRANSPORT/nr_dlsch_demodulation.c @@ -379,13 +379,10 @@ int nr_rx_pdsch(PHY_VARS_NR_UE *ue, len = (pilots==1)? ((config_type==NFAPI_NR_DMRS_TYPE1)?nb_rb*(12-6*dlsch0_harq->n_dmrs_cdm_groups): nb_rb*(12-4*dlsch0_harq->n_dmrs_cdm_groups)):(nb_rb*12); stop_meas(&ue->generic_stat_bis[proc->thread_id][slot]); -#if DISABLE_LOG_X - printf("[AbsSFN %u.%d] Slot%d Symbol %d type %d: Pilot/Data extraction %5.2f \n", - frame,nr_slot_rx,slot,symbol,type,ue->generic_stat_bis[proc->thread_id][slot].p_time/(cpuf*1000.0)); -#else - LOG_D(PHY, "[AbsSFN %u.%d] Slot%d Symbol %d type %d: Pilot/Data extraction %5.2f \n", - frame,nr_slot_rx,slot,symbol,type,ue->generic_stat_bis[proc->thread_id][slot].p_time/(cpuf*1000.0)); -#endif + if (cpumeas(CPUMEAS_GETSTATE)) + LOG_D(PHY, "[AbsSFN %u.%d] Slot%d Symbol %d type %d: Pilot/Data extraction %5.2f \n", + frame,nr_slot_rx,slot,symbol,type,ue->generic_stat_bis[proc->thread_id][slot].p_time/(cpuf*1000.0)); + start_meas(&ue->generic_stat_bis[proc->thread_id][slot]); n_tx = dlsch0_harq->Nl; @@ -402,11 +399,9 @@ int nr_rx_pdsch(PHY_VARS_NR_UE *ue, nb_rb_pdsch); stop_meas(&ue->generic_stat_bis[proc->thread_id][slot]); -#if DISABLE_LOG_X - printf("[AbsSFN %u.%d] Slot%d Symbol %d: Channel Scale %5.2f \n",frame,nr_slot_rx,slot,symbol,ue->generic_stat_bis[proc->thread_id][slot].p_time/(cpuf*1000.0)); -#else + if (cpumeas(CPUMEAS_GETSTATE)) LOG_D(PHY, "[AbsSFN %u.%d] Slot%d Symbol %d: Channel Scale %5.2f \n",frame,nr_slot_rx,slot,symbol,ue->generic_stat_bis[proc->thread_id][slot].p_time/(cpuf*1000.0)); -#endif + start_meas(&ue->generic_stat_bis[proc->thread_id][slot]); if (first_symbol_flag==1) { if (beamforming_mode==0){ @@ -463,11 +458,8 @@ int nr_rx_pdsch(PHY_VARS_NR_UE *ue, #endif stop_meas(&ue->generic_stat_bis[proc->thread_id][slot]); -#if DISABLE_LOG_X - printf("[AbsSFN %u.%d] Slot%d Symbol %d first_symbol_flag %d: Channel Level %5.2f \n",frame,nr_slot_rx,slot,symbol,first_symbol_flag,ue->generic_stat_bis[proc->thread_id][slot].p_time/(cpuf*1000.0)); -#else - LOG_D(PHY, "[AbsSFN %u.%d] Slot%d Symbol %d first_symbol_flag %d: Channel Level %5.2f \n",frame,nr_slot_rx,slot,symbol,first_symbol_flag,ue->generic_stat_bis[proc->thread_id][slot].p_time/(cpuf*1000.0)); -#endif + if (cpumeas(CPUMEAS_GETSTATE)) + LOG_D(PHY, "[AbsSFN %u.%d] Slot%d Symbol %d first_symbol_flag %d: Channel Level %5.2f \n",frame,nr_slot_rx,slot,symbol,first_symbol_flag,ue->generic_stat_bis[proc->thread_id][slot].p_time/(cpuf*1000.0)); start_meas(&ue->generic_stat_bis[proc->thread_id][slot]); // Now channel compensation @@ -515,11 +507,9 @@ int nr_rx_pdsch(PHY_VARS_NR_UE *ue, } stop_meas(&ue->generic_stat_bis[proc->thread_id][slot]); -#if DISABLE_LOG_X - printf("[AbsSFN %u.%d] Slot%d Symbol %d log2_maxh %d channel_level %d: Channel Comp %5.2f \n", frame, nr_slot_rx, slot, symbol, pdsch_vars[gNB_id]->log2_maxh, proc->channel_level, ue->generic_stat_bis[proc->thread_id][slot].p_time/(cpuf*1000.0)); -#else - LOG_D(PHY, "[AbsSFN %u.%d] Slot%d Symbol %d log2_maxh %d channel_level %d: Channel Comp %5.2f \n", frame, nr_slot_rx, slot, symbol, pdsch_vars[gNB_id]->log2_maxh, proc->channel_level, ue->generic_stat_bis[proc->thread_id][slot].p_time/(cpuf*1000.0)); -#endif + if (cpumeas(CPUMEAS_GETSTATE)) + LOG_D(PHY, "[AbsSFN %u.%d] Slot%d Symbol %d log2_maxh %d channel_level %d: Channel Comp %5.2f \n", frame, nr_slot_rx, slot, symbol, pdsch_vars[gNB_id]->log2_maxh, proc->channel_level, ue->generic_stat_bis[proc->thread_id][slot].p_time/(cpuf*1000.0)); + start_meas(&ue->generic_stat_bis[proc->thread_id][slot]); if (frame_parms->nb_antennas_rx > 1) { @@ -574,11 +564,8 @@ int nr_rx_pdsch(PHY_VARS_NR_UE *ue, } stop_meas(&ue->generic_stat_bis[proc->thread_id][slot]); -#if DISABLE_LOG_X - printf("[AbsSFN %u.%d] Slot%d Symbol %d: Channel Combine %5.2f \n",frame,nr_slot_rx,slot,symbol,ue->generic_stat_bis[proc->thread_id][slot].p_time/(cpuf*1000.0)); -#else - LOG_D(PHY, "[AbsSFN %u.%d] Slot%d Symbol %d: Channel Combine %5.2f \n",frame,nr_slot_rx,slot,symbol,ue->generic_stat_bis[proc->thread_id][slot].p_time/(cpuf*1000.0)); -#endif + if (cpumeas(CPUMEAS_GETSTATE)) + LOG_D(PHY, "[AbsSFN %u.%d] Slot%d Symbol %d: Channel Combine %5.2f \n",frame,nr_slot_rx,slot,symbol,ue->generic_stat_bis[proc->thread_id][slot].p_time/(cpuf*1000.0)); start_meas(&ue->generic_stat_bis[proc->thread_id][slot]); /* Store the valid DL RE's */ pdsch_vars[gNB_id]->dl_valid_re[symbol-1] = len; @@ -657,11 +644,8 @@ int nr_rx_pdsch(PHY_VARS_NR_UE *ue, } stop_meas(&ue->generic_stat_bis[proc->thread_id][slot]); -#if DISABLE_LOG_X - printf("[AbsSFN %u.%d] Slot%d Symbol %d: LLR Computation %5.2f \n",frame,nr_slot_rx,slot,symbol,ue->generic_stat_bis[proc->thread_id][slot].p_time/(cpuf*1000.0)); -#else - LOG_D(PHY, "[AbsSFN %u.%d] Slot%d Symbol %d: LLR Computation %5.2f \n",frame,nr_slot_rx,slot,symbol,ue->generic_stat_bis[proc->thread_id][slot].p_time/(cpuf*1000.0)); -#endif + if (cpumeas(CPUMEAS_GETSTATE)) + LOG_D(PHY, "[AbsSFN %u.%d] Slot%d Symbol %d: LLR Computation %5.2f \n",frame,nr_slot_rx,slot,symbol,ue->generic_stat_bis[proc->thread_id][slot].p_time/(cpuf*1000.0)); // Please keep it: useful for debugging #ifdef DEBUG_PDSCH_RX diff --git a/openair1/PHY/NR_UE_TRANSPORT/nr_initial_sync.c b/openair1/PHY/NR_UE_TRANSPORT/nr_initial_sync.c index 767edf9cc6..44e103e90e 100644 --- a/openair1/PHY/NR_UE_TRANSPORT/nr_initial_sync.c +++ b/openair1/PHY/NR_UE_TRANSPORT/nr_initial_sync.c @@ -353,7 +353,7 @@ int nr_initial_sync(UE_nr_rxtx_proc_t *proc, if( (abs(ue->common_vars.freq_offset) > 150) && (ret == 0) ) { ret=-1; -#if DISABLE_LOG_X +#if PHYSIM printf("Ignore MIB with high freq offset [%d Hz] estimation \n",ue->common_vars.freq_offset); #else LOG_E(HW, "Ignore MIB with high freq offset [%d Hz] estimation \n",ue->common_vars.freq_offset); @@ -363,7 +363,7 @@ int nr_initial_sync(UE_nr_rxtx_proc_t *proc, if (ret==0) { // PBCH found so indicate sync to higher layers and configure frame parameters //#ifdef DEBUG_INITIAL_SYNCH -#if DISABLE_LOG_X +#if PHYSIM printf("[UE%d] In synch, rx_offset %d samples\n",ue->Mod_id, ue->rx_offset); #else LOG_I(PHY, "[UE%d] In synch, rx_offset %d samples\n",ue->Mod_id, ue->rx_offset); diff --git a/openair1/PHY/TOOLS/time_meas.c b/openair1/PHY/TOOLS/time_meas.c index 1b62687f7e..353f170434 100644 --- a/openair1/PHY/TOOLS/time_meas.c +++ b/openair1/PHY/TOOLS/time_meas.c @@ -51,24 +51,7 @@ double get_cpu_freq_GHz(void) return cpu_freq_GHz; } -int cpumeas(int action) -{ - switch (action) { - case CPUMEAS_ENABLE: - opp_enabled = 1; - break; - case CPUMEAS_DISABLE: - opp_enabled = 0; - break; - - case CPUMEAS_GETSTATE: - default: - break; - } - - return opp_enabled; -} void print_meas_now(time_stats_t *ts, const char *name, diff --git a/openair1/PHY/TOOLS/time_meas.h b/openair1/PHY/TOOLS/time_meas.h index bb2bb3d276..56d8299470 100644 --- a/openair1/PHY/TOOLS/time_meas.h +++ b/openair1/PHY/TOOLS/time_meas.h @@ -115,7 +115,24 @@ static inline uint32_t rdtsc_oai(void) { #define CPUMEAS_DISABLE 0 #define CPUMEAS_ENABLE 1 #define CPUMEAS_GETSTATE 2 -int cpumeas(int action); +static inline int cpumeas(int action) { + switch (action) { + case CPUMEAS_ENABLE: + opp_enabled = 1; + break; + + case CPUMEAS_DISABLE: + opp_enabled = 0; + break; + + case CPUMEAS_GETSTATE: + default: + break; + } + + return opp_enabled; +} + static inline void start_meas(time_stats_t *ts) { if (opp_enabled) { if (ts->meas_flag==0) { From 25e20db71fb0bbd268fec16c710de08cb3e4a86e Mon Sep 17 00:00:00 2001 From: frtabu Date: Tue, 23 Nov 2021 17:24:44 +0100 Subject: [PATCH 02/23] ldpc openCL doc update --- openair1/PHY/CODING/DOC/LDPCImplementation.md | 44 ++++++++++++++----- 1 file changed, 32 insertions(+), 12 deletions(-) diff --git a/openair1/PHY/CODING/DOC/LDPCImplementation.md b/openair1/PHY/CODING/DOC/LDPCImplementation.md index c5411540b6..a82c204074 100644 --- a/openair1/PHY/CODING/DOC/LDPCImplementation.md +++ b/openair1/PHY/CODING/DOC/LDPCImplementation.md @@ -24,12 +24,33 @@ loading `libldpc_cl.so` instead of `libldpc.so`: `make ldpc_cl` -`cp ../../../openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_decoder_kernels_CL.cl` +This command creates the `libldpc_cl.so` shared library. To perform this build successfully, only the OpenCL header `(/usr/include/CL/opencl.h)` and library `(/usr/lib/x86_64-linux-gnu/libOpenCL.so)`are required, they implement OpenCL API support which is not hardware dependent. + +``` +Scanning dependencies of target nrLDPC_decoder_kernels_CL +Built target nrLDPC_decoder_kernels_CL +Scanning dependencies of target ldpc_cl +Building C object CMakeFiles/ldpc_cl.dir/usr/local/oai/oai-develop/openairinterface5g/openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_decoder_CL.c.o +In file included from /usr/include/CL/cl.h:32, + from /usr/include/CL/opencl.h:38, + from /usr/local/oai/oai-develop/openairinterface5g/openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_decoder_CL.c:49: +/usr/include/CL/cl_version.h:34:9: note: #pragma message: cl_version.h: CL_TARGET_OPENCL_VERSION is not defined. Defaulting to 220 (OpenCL 2.2) + #pragma message("cl_version.h: CL_TARGET_OPENCL_VERSION is not defined. Defaulting to 220 (OpenCL 2.2)") + ^~~~~~~ + +Building C object CMakeFiles/ldpc_cl.dir/usr/local/oai/oai-develop/openairinterface5g/openair1/PHY/CODING/nrLDPC_encoder/ldpc_encoder_optim8segmulti.c.o +Linking C shared module libldpc_cl.so +Built target ldpc_cl + +``` + +At runtime, to successfully use hardware acceleration via OpenCL, you need to install vendor specific packages which deliver the required drivers and tools to make use of their GPU (Nvidia, Intel...) , fpga (Xilinx, Intel) or CPU (Intel, AMD, ARM...) through OpenCL. `./nr-softmodem -O libconfig:gnb.band78.sa.fr1.106PRB.usrpb210.conf:dbgl5 --rfsim --rfsimulator.serveraddr server --sa --log_config.gtpu_log_level info --loader.ldpc.shlibversion _cl` -``` [LOADER] library libldpc_cl.so successfully loaded +``` ------------------------------------------------ +[LOADER] library libldpc_cl.so successfully loaded [HW] Platform 0, OpenCL profile FULL_PROFILE [HW] Platform 0, OpenCL version OpenCL 2.1 LINUX [HW] Device 0 is available @@ -62,8 +83,9 @@ loading `libldpc_cl.so` instead of `libldpc.so`: `./nr-uesoftmodem -r 106 --numerology 1 --band 78 -C 3619200000 --rfsim --sa -O libconfig:/usr/local/oai/conf/nrue_sim.conf:dbgl5 --nokrnmod --loader.ldpc.shlibversion _cl --log_config.hw_log_level info` -```[CONFIG] shlibversion set to _cl from command line +``` ............................................................ +[CONFIG] shlibversion set to _cl from command line [CONFIG] loader.ldpc 1 options set from command line [LOADER] library libldpc_cl.so successfully loaded [HW] Platform 0, OpenCL profile FULL_PROFILE @@ -94,7 +116,6 @@ loading `libldpc_cl.so` instead of `libldpc.so`: [HW] Device 0, max Work Items size for dimension: 1 512 [HW] Device 0, max Work Items size for dimension: 2 512 ------------------------------------------------------------ -​``` ``` A mechanism to select ldpc implementation is also available in the `ldpctest` phy simulator via the `-v`option, which can be used to specify the version of the ldpc shared library to be used. @@ -103,7 +124,9 @@ A mechanism to select ldpc implementation is also available in the `ldpctest` ph Loading libldpc_cuda.so, the cuda implementation of the ldpc decoder: -```$ ./ldpctest -v _cuda +``` +$ ./ldpctest -v _cuda +ldpctest -v _cuda Initializing random number generator, seed 0 block length 8448: n_trials 1: @@ -119,18 +142,16 @@ log init done [CONFIG] loader.ldpc: 1/2 parameters successfully set [LOADER] library libldpc_cuda.so successfully loaded ................................... -​``` ``` + Loading libldpc_cl.so, the opencl implementation of the ldpc decoder: `make ldpc_cl` -`cp ../../../openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_decoder_kernels_CL.cl` -`./ldpctest -v _cl` - -```$ ./ldpctest -v _cl +``` +$ ./ldpctest -v _cl Initializing random number generator, seed 0 block length 8448: n_trials 1: @@ -164,7 +185,6 @@ log init done [HW] Device 0, max Work Items size for dimension: 1 512 [HW] Device 0, max Work Items size for dimension: 2 512 ................................ -​``` ``` @@ -174,6 +194,6 @@ Libraries implementing the LDPC algorithms must be named `libldpc<_version>.so`, `libldpc_cuda.so`has been tested with the `ldpctest` executable, usage from the softmodem's has to be tested. -`libldpc_cl`is under development. +`libldpc_cl.so`is under development. [oai Wikis home](https://gitlab.eurecom.fr/oai/openairinterface5g/wikis/home) From 683072f4133c45306b1ed051e16b3e4e42bb152a Mon Sep 17 00:00:00 2001 From: frtabu Date: Wed, 24 Nov 2021 11:00:05 +0100 Subject: [PATCH 03/23] Fix time measurment LOGS in phy_procedures_nr_UE.c --- openair1/SCHED_NR_UE/phy_procedures_nr_ue.c | 49 +++++++++++---------- 1 file changed, 26 insertions(+), 23 deletions(-) diff --git a/openair1/SCHED_NR_UE/phy_procedures_nr_ue.c b/openair1/SCHED_NR_UE/phy_procedures_nr_ue.c index 850b41d2fb..3b4c3657df 100644 --- a/openair1/SCHED_NR_UE/phy_procedures_nr_ue.c +++ b/openair1/SCHED_NR_UE/phy_procedures_nr_ue.c @@ -844,7 +844,8 @@ int nr_ue_pdsch_procedures(PHY_VARS_NR_UE *ue, UE_nr_rxtx_proc_t *proc, int gNB_ } else AssertFatal(1==0,"Not RA_PDSCH, SI_PDSCH or PDSCH\n"); stop_meas(&ue->dlsch_llr_stats_parallelization[proc->thread_id][slot]); - LOG_D(PHY, "[AbsSFN %d.%d] LLR Computation Symbol %d %5.2f \n",frame_rx,nr_slot_rx,m,ue->dlsch_llr_stats_parallelization[proc->thread_id][slot].p_time/(cpuf*1000.0)); + if (cpumeas(CPUMEAS_GETSTATE)) + LOG_D(PHY, "[AbsSFN %d.%d] LLR Computation Symbol %d %5.2f \n",frame_rx,nr_slot_rx,m,ue->dlsch_llr_stats_parallelization[proc->thread_id][slot].p_time/(cpuf*1000.0)); if(first_symbol_flag) { proc->first_symbol_available = 1; @@ -1024,10 +1025,13 @@ bool nr_ue_dlsch_procedures(PHY_VARS_NR_UE *ue, stop_meas(&ue->dlsch_decoding_stats[proc->thread_id]); - LOG_D(PHY, " --> Unscrambling for CW0 %5.3f\n", - (ue->dlsch_unscrambling_stats.p_time)/(cpuf*1000.0)); - LOG_D(PHY, "AbsSubframe %d.%d --> LDPC Decoding for CW0 %5.3f\n", - frame_rx%1024, nr_slot_rx,(ue->dlsch_decoding_stats[proc->thread_id].p_time)/(cpuf*1000.0)); + + if (cpumeas(CPUMEAS_GETSTATE)) { + LOG_D(PHY, " --> Unscrambling for CW0 %5.3f\n", + (ue->dlsch_unscrambling_stats.p_time)/(cpuf*1000.0)); + LOG_D(PHY, "AbsSubframe %d.%d --> LDPC Decoding for CW0 %5.3f\n", + frame_rx%1024, nr_slot_rx,(ue->dlsch_decoding_stats[proc->thread_id].p_time)/(cpuf*1000.0)); + } if(is_cw1_active) { // start ldpc decode for CW 1 @@ -1079,16 +1083,13 @@ bool nr_ue_dlsch_procedures(PHY_VARS_NR_UE *ue, LOG_T(PHY,"CWW sequential dlsch decoding, ret1 = %d\n", ret1); } - - stop_meas(&ue->dlsch_decoding_stats[proc->thread_id]); - - LOG_D(PHY, " --> Unscrambling for CW1 %5.3f\n", - (ue->dlsch_unscrambling_stats.p_time)/(cpuf*1000.0)); - LOG_D(PHY, "AbsSubframe %d.%d --> ldpc Decoding for CW1 %5.3f\n", - frame_rx%1024, nr_slot_rx,(ue->dlsch_decoding_stats[proc->thread_id].p_time)/(cpuf*1000.0)); - LOG_D(PHY,"AbsSubframe %d.%d --> ldpc Decoding for CW1 %5.3f\n", - frame_rx%1024, nr_slot_rx,(ue->dlsch_decoding_stats[proc->thread_id].p_time)/(cpuf*1000.0)); - + stop_meas(&ue->dlsch_decoding_stats[proc->thread_id]); + if (cpumeas(CPUMEAS_GETSTATE)) { + LOG_D(PHY, " --> Unscrambling for CW1 %5.3f\n", + (ue->dlsch_unscrambling_stats.p_time)/(cpuf*1000.0)); + LOG_D(PHY, "AbsSubframe %d.%d --> ldpc Decoding for CW1 %5.3f\n", + frame_rx%1024, nr_slot_rx,(ue->dlsch_decoding_stats[proc->thread_id].p_time)/(cpuf*1000.0)); + } LOG_D(PHY, "harq_pid: %d, TBS expected dlsch1: %d \n", harq_pid, dlsch1->harq_processes[harq_pid]->TBS); } @@ -1394,8 +1395,8 @@ void *UE_thread_slot1_dl_processing(void *arg) { stop_meas(&ue->ue_front_end_per_slot_stat[proc->thread_id][1]); - LOG_D(PHY, "[AbsSFN %d.%d] Slot1: FFT + Channel Estimate + Pdsch Proc Slot0 %5.2f \n",frame_rx,nr_slot_rx,ue->ue_front_end_per_slot_stat[proc->thread_id][1].p_time/(cpuf*1000.0)); - + if (cpumeas(CPUMEAS_GETSTATE)) + LOG_D(PHY, "[AbsSFN %d.%d] Slot1: FFT + Channel Estimate + Pdsch Proc Slot0 %5.2f \n",frame_rx,nr_slot_rx,ue->ue_front_end_per_slot_stat[proc->thread_id][1].p_time/(cpuf*1000.0)); //wait until pdcch is decoded uint32_t wait = 0; @@ -1484,7 +1485,8 @@ void *UE_thread_slot1_dl_processing(void *arg) { //printf("Set available LLR slot1 to 1 AbsSubframe %d.%d \n",frame_rx,nr_slot_rx); stop_meas(&ue->pdsch_procedures_per_slot_stat[proc->thread_id][1]); - LOG_D(PHY, "[AbsSFN %d.%d] Slot1: LLR Computation %5.2f \n",frame_rx,nr_slot_rx,ue->pdsch_procedures_per_slot_stat[proc->thread_id][1].p_time/(cpuf*1000.0)); + if (cpumeas(CPUMEAS_GETSTATE)) + LOG_D(PHY, "[AbsSFN %d.%d] Slot1: LLR Computation %5.2f \n",frame_rx,nr_slot_rx,ue->pdsch_procedures_per_slot_stat[proc->thread_id][1].p_time/(cpuf*1000.0)); if (pthread_mutex_lock(&proc->mutex_slot1_dl_processing) != 0) { LOG_E( PHY, "[SCHED][UE] error locking mutex for UE RXTX\n" ); @@ -1853,9 +1855,10 @@ int phy_procedures_nrUE_RX(PHY_VARS_NR_UE *ue, dlsch_parallel); stop_meas(&ue->dlsch_procedures_stat[proc->thread_id]); - LOG_D(PHY, "[SFN %d] Slot1: Pdsch Proc %5.2f\n",nr_slot_rx,ue->pdsch_procedures_stat[proc->thread_id].p_time/(cpuf*1000.0)); - LOG_D(PHY, "[SFN %d] Slot0 Slot1: Dlsch Proc %5.2f\n",nr_slot_rx,ue->dlsch_procedures_stat[proc->thread_id].p_time/(cpuf*1000.0)); - + if (cpumeas(CPUMEAS_GETSTATE)) { + LOG_D(PHY, "[SFN %d] Slot1: Pdsch Proc %5.2f\n",nr_slot_rx,ue->pdsch_procedures_stat[proc->thread_id].p_time/(cpuf*1000.0)); + LOG_D(PHY, "[SFN %d] Slot0 Slot1: Dlsch Proc %5.2f\n",nr_slot_rx,ue->dlsch_procedures_stat[proc->thread_id].p_time/(cpuf*1000.0)); + } // deactivate dlsch once dlsch proc is done ue->dlsch[proc->thread_id][gNB_id][0]->active = 0; @@ -1939,8 +1942,8 @@ phy_procedures_emos_UE_RX(ue,slot,gNB_id); VCD_SIGNAL_DUMPER_DUMP_FUNCTION_BY_NAME(VCD_SIGNAL_DUMPER_FUNCTIONS_PHY_PROCEDURES_UE_RX, VCD_FUNCTION_OUT); stop_meas(&ue->phy_proc_rx[proc->thread_id]); -LOG_D(PHY, "------FULL RX PROC [SFN %d]: %5.2f ------\n",nr_slot_rx,ue->phy_proc_rx[proc->thread_id].p_time/(cpuf*1000.0)); - +if (cpumeas(CPUMEAS_GETSTATE)) + LOG_D(PHY, "------FULL RX PROC [SFN %d]: %5.2f ------\n",nr_slot_rx,ue->phy_proc_rx[proc->thread_id].p_time/(cpuf*1000.0)); //#endif //pdsch LOG_D(PHY," ****** end RX-Chain for AbsSubframe %d.%d ****** \n", frame_rx%1024, nr_slot_rx); From fefa42c889ac8d52e28a86b0ad27e3da090dfd33 Mon Sep 17 00:00:00 2001 From: Raymond Knopp Date: Fri, 3 Dec 2021 18:13:52 +0100 Subject: [PATCH 04/23] phy simulators remove measurments printf and check/allow log level config --- openair1/PHY/CODING/TESTBENCH/ldpctest.c | 18 ++++++++++++++---- .../NR_UE_TRANSPORT/nr_dlsch_demodulation.c | 6 +++--- openair1/SCHED_NR_UE/phy_procedures_nr_ue.c | 15 ++++++++------- openair1/SIMULATION/NR_PHY/dlsim.c | 1 + openair1/SIMULATION/NR_PHY/ulsim.c | 3 ++- 5 files changed, 28 insertions(+), 15 deletions(-) diff --git a/openair1/PHY/CODING/TESTBENCH/ldpctest.c b/openair1/PHY/CODING/TESTBENCH/ldpctest.c index aff707a1d5..371deba3f2 100644 --- a/openair1/PHY/CODING/TESTBENCH/ldpctest.c +++ b/openair1/PHY/CODING/TESTBENCH/ldpctest.c @@ -514,7 +514,7 @@ int main(int argc, char *argv[]) unsigned char qbits=8; unsigned int decoded_errors[10000]; // initiate the size of matrix equivalent to size of SNR int c,i=0, i1 = 0; - + int loglvl=OAILOG_WARNING; int n_trials = 1; double SNR_step = 0.1; @@ -525,8 +525,11 @@ int main(int argc, char *argv[]) n_iter_stats_t dec_iter[3]; short BG=0,Zc,Kb=0; - - while ((c = getopt (argc, argv, "q:r:s:S:l:G:n:d:i:t:u:hv:")) != -1) + if ( load_configmodule(argc,argv,CONFIG_ENABLECMDLINEONLY) == 0) { + exit_fun(" Error, configuration module init failed\n"); + } // must be done before specific options parsing to prevent errasing them + + while ((c = getopt (argc, argv, "q:r:s:S:l:L:G:n:d:i:t:u:hv:")) != -1) switch (c) { case 'q': @@ -544,6 +547,10 @@ int main(int argc, char *argv[]) case 'l': block_length = atoi(optarg); break; + + case 'L': + loglvl = atoi(optarg); + break; case 'G': ldpc_version="_cuda"; @@ -581,10 +588,11 @@ int main(int argc, char *argv[]) printf("BG1 (blocklength > 3840): 1/3, 2/3, 22/25 (8/9) \n"); printf("BG2 (blocklength <= 3840): 1/5, 1/3, 2/3 \n\n"); printf("-h This message\n"); + printf("-L \n"); printf("-q Quantization bits, Default: 8\n"); printf("-r Nominator rate, (1, 2, 22), Default: 1\n"); printf("-d Denominator rate, (3, 5, 25), Default: 1\n"); - printf("-l Block length (l > 3840 -> BG1, rest BG2 ), Default: 8448\n"); + printf("-l Block length (l > 3840 -> BG1, rest BG2 ), Default: 8448\n"); printf("-G give 1 to run cuda for LDPC, Default: 0\n"); printf("-n Number of simulation trials, Default: 1\n"); //printf("-M MCS2 for TB 2\n"); @@ -603,6 +611,8 @@ int main(int argc, char *argv[]) printf("n_trials %d: \n", n_trials); printf("SNR0 %f: \n", SNR0); + logInit(); + set_glog(loglvl); if (ldpc_version != NULL) load_nrLDPClib(ldpc_version); diff --git a/openair1/PHY/NR_UE_TRANSPORT/nr_dlsch_demodulation.c b/openair1/PHY/NR_UE_TRANSPORT/nr_dlsch_demodulation.c index 31b3f8896a..1e031d531c 100644 --- a/openair1/PHY/NR_UE_TRANSPORT/nr_dlsch_demodulation.c +++ b/openair1/PHY/NR_UE_TRANSPORT/nr_dlsch_demodulation.c @@ -399,8 +399,8 @@ int nr_rx_pdsch(PHY_VARS_NR_UE *ue, nb_rb_pdsch); stop_meas(&ue->generic_stat_bis[proc->thread_id][slot]); - if (cpumeas(CPUMEAS_GETSTATE)) - LOG_D(PHY, "[AbsSFN %u.%d] Slot%d Symbol %d: Channel Scale %5.2f \n",frame,nr_slot_rx,slot,symbol,ue->generic_stat_bis[proc->thread_id][slot].p_time/(cpuf*1000.0)); + if (cpumeas(CPUMEAS_GETSTATE)) + LOG_D(PHY, "[AbsSFN %u.%d] Slot%d Symbol %d: Channel Scale %5.2f \n",frame,nr_slot_rx,slot,symbol,ue->generic_stat_bis[proc->thread_id][slot].p_time/(cpuf*1000.0)); start_meas(&ue->generic_stat_bis[proc->thread_id][slot]); if (first_symbol_flag==1) { @@ -509,7 +509,6 @@ int nr_rx_pdsch(PHY_VARS_NR_UE *ue, stop_meas(&ue->generic_stat_bis[proc->thread_id][slot]); if (cpumeas(CPUMEAS_GETSTATE)) LOG_D(PHY, "[AbsSFN %u.%d] Slot%d Symbol %d log2_maxh %d channel_level %d: Channel Comp %5.2f \n", frame, nr_slot_rx, slot, symbol, pdsch_vars[gNB_id]->log2_maxh, proc->channel_level, ue->generic_stat_bis[proc->thread_id][slot].p_time/(cpuf*1000.0)); - start_meas(&ue->generic_stat_bis[proc->thread_id][slot]); if (frame_parms->nb_antennas_rx > 1) { @@ -566,6 +565,7 @@ int nr_rx_pdsch(PHY_VARS_NR_UE *ue, stop_meas(&ue->generic_stat_bis[proc->thread_id][slot]); if (cpumeas(CPUMEAS_GETSTATE)) LOG_D(PHY, "[AbsSFN %u.%d] Slot%d Symbol %d: Channel Combine %5.2f \n",frame,nr_slot_rx,slot,symbol,ue->generic_stat_bis[proc->thread_id][slot].p_time/(cpuf*1000.0)); + start_meas(&ue->generic_stat_bis[proc->thread_id][slot]); /* Store the valid DL RE's */ pdsch_vars[gNB_id]->dl_valid_re[symbol-1] = len; diff --git a/openair1/SCHED_NR_UE/phy_procedures_nr_ue.c b/openair1/SCHED_NR_UE/phy_procedures_nr_ue.c index 3b4c3657df..b65cb333f7 100644 --- a/openair1/SCHED_NR_UE/phy_procedures_nr_ue.c +++ b/openair1/SCHED_NR_UE/phy_procedures_nr_ue.c @@ -1025,7 +1025,6 @@ bool nr_ue_dlsch_procedures(PHY_VARS_NR_UE *ue, stop_meas(&ue->dlsch_decoding_stats[proc->thread_id]); - if (cpumeas(CPUMEAS_GETSTATE)) { LOG_D(PHY, " --> Unscrambling for CW0 %5.3f\n", (ue->dlsch_unscrambling_stats.p_time)/(cpuf*1000.0)); @@ -1083,12 +1082,12 @@ bool nr_ue_dlsch_procedures(PHY_VARS_NR_UE *ue, LOG_T(PHY,"CWW sequential dlsch decoding, ret1 = %d\n", ret1); } - stop_meas(&ue->dlsch_decoding_stats[proc->thread_id]); + stop_meas(&ue->dlsch_decoding_stats[proc->thread_id]); if (cpumeas(CPUMEAS_GETSTATE)) { - LOG_D(PHY, " --> Unscrambling for CW1 %5.3f\n", - (ue->dlsch_unscrambling_stats.p_time)/(cpuf*1000.0)); - LOG_D(PHY, "AbsSubframe %d.%d --> ldpc Decoding for CW1 %5.3f\n", - frame_rx%1024, nr_slot_rx,(ue->dlsch_decoding_stats[proc->thread_id].p_time)/(cpuf*1000.0)); + LOG_D(PHY, " --> Unscrambling for CW1 %5.3f\n", + (ue->dlsch_unscrambling_stats.p_time)/(cpuf*1000.0)); + LOG_D(PHY, "AbsSubframe %d.%d --> ldpc Decoding for CW1 %5.3f\n", + frame_rx%1024, nr_slot_rx,(ue->dlsch_decoding_stats[proc->thread_id].p_time)/(cpuf*1000.0)); } LOG_D(PHY, "harq_pid: %d, TBS expected dlsch1: %d \n", harq_pid, dlsch1->harq_processes[harq_pid]->TBS); } @@ -1859,7 +1858,6 @@ int phy_procedures_nrUE_RX(PHY_VARS_NR_UE *ue, LOG_D(PHY, "[SFN %d] Slot1: Pdsch Proc %5.2f\n",nr_slot_rx,ue->pdsch_procedures_stat[proc->thread_id].p_time/(cpuf*1000.0)); LOG_D(PHY, "[SFN %d] Slot0 Slot1: Dlsch Proc %5.2f\n",nr_slot_rx,ue->dlsch_procedures_stat[proc->thread_id].p_time/(cpuf*1000.0)); } - // deactivate dlsch once dlsch proc is done ue->dlsch[proc->thread_id][gNB_id][0]->active = 0; @@ -1933,6 +1931,8 @@ if (nr_slot_rx==9) { } stop_meas(&ue->generic_stat); +if (cpumeas(CPUMEAS_GETSTATE)) + LOG_D(PHY,"after tubo until end of Rx %5.2f \n",ue->generic_stat.p_time/(cpuf*1000.0)); #ifdef EMOS phy_procedures_emos_UE_RX(ue,slot,gNB_id); @@ -1944,6 +1944,7 @@ VCD_SIGNAL_DUMPER_DUMP_FUNCTION_BY_NAME(VCD_SIGNAL_DUMPER_FUNCTIONS_PHY_PROCEDUR stop_meas(&ue->phy_proc_rx[proc->thread_id]); if (cpumeas(CPUMEAS_GETSTATE)) LOG_D(PHY, "------FULL RX PROC [SFN %d]: %5.2f ------\n",nr_slot_rx,ue->phy_proc_rx[proc->thread_id].p_time/(cpuf*1000.0)); + //#endif //pdsch LOG_D(PHY," ****** end RX-Chain for AbsSubframe %d.%d ****** \n", frame_rx%1024, nr_slot_rx); diff --git a/openair1/SIMULATION/NR_PHY/dlsim.c b/openair1/SIMULATION/NR_PHY/dlsim.c index fa868a6896..80d1a512b1 100644 --- a/openair1/SIMULATION/NR_PHY/dlsim.c +++ b/openair1/SIMULATION/NR_PHY/dlsim.c @@ -645,6 +645,7 @@ int main(int argc, char **argv) printf("%s -h(elp) -p(extended_prefix) -N cell_id -f output_filename -F input_filename -g channel_model -n n_frames -t Delayspread -s snr0 -S snr1 -x transmission_mode -y TXant -z RXant -i Intefrence0 -j Interference1 -A interpolation_file -C(alibration offset dB) -N CellId\n", argv[0]); printf("-h This message\n"); + printf("-L \n"); //printf("-p Use extended prefix mode\n"); //printf("-d Use TDD\n"); printf("-n Number of frames to simulate\n"); diff --git a/openair1/SIMULATION/NR_PHY/ulsim.c b/openair1/SIMULATION/NR_PHY/ulsim.c index 97c753d455..9be5a18805 100644 --- a/openair1/SIMULATION/NR_PHY/ulsim.c +++ b/openair1/SIMULATION/NR_PHY/ulsim.c @@ -600,6 +600,7 @@ int main(int argc, char **argv) //printf("-C Generate Calibration information for Abstraction (effective SNR adjustment to remove Pe bias w.r.t. AWGN)\n"); printf("-F Input filename (.txt format) for RX conformance testing\n"); printf("-G Offset of samples to read from file (0 default)\n"); + printf("-L \n"); printf("-M Multiple SSB positions in burst\n"); printf("-N Nid_cell\n"); printf("-O oversampling factor (1,2,4,8,16)\n"); @@ -640,7 +641,7 @@ int main(int argc, char **argv) else if (N_RB_UL == 106) bandwidth = 40; else if (N_RB_UL == 32) bandwidth = 50; else { printf("Add N_RB_UL %d\n",N_RB_UL); exit(-1); } - + LOG_I( PHY,"++++++++++++++++++++++++++++++++++++++++++++++%i+++++++++++++++++++++++++++++++++++++++++",loglvl); if (openair0_cfg[0].threequarter_fs == 1) sampling_frequency*=.75; UE2gNB = new_channel_desc_scm(n_tx, n_rx, channel_model, From 883964c7daf2ebb34649d17baac123d8323fb226 Mon Sep 17 00:00:00 2001 From: Raymond Knopp Date: Tue, 7 Dec 2021 10:42:46 +0100 Subject: [PATCH 05/23] remove remaining printf under PHYSIM build option --- openair1/PHY/INIT/nr_parms.c | 11 +++-------- openair1/PHY/NR_UE_TRANSPORT/nr_initial_sync.c | 10 ++-------- 2 files changed, 5 insertions(+), 16 deletions(-) diff --git a/openair1/PHY/INIT/nr_parms.c b/openair1/PHY/INIT/nr_parms.c index 0c0d38ea0b..35bc28d9c1 100644 --- a/openair1/PHY/INIT/nr_parms.c +++ b/openair1/PHY/INIT/nr_parms.c @@ -250,11 +250,9 @@ int nr_init_frame_parms(nfapi_nr_config_request_scf_t* cfg, int Ncp = NFAPI_CP_NORMAL; int mu = cfg->ssb_config.scs_common.value; -#if PHYSIM - printf("Initializing frame parms for mu %d, N_RB %d, Ncp %d\n",mu, fp->N_RB_DL, Ncp); -#else + LOG_I(PHY,"Initializing frame parms for mu %d, N_RB %d, Ncp %d\n",mu, fp->N_RB_DL, Ncp); -#endif + if (Ncp == NFAPI_CP_EXTENDED) AssertFatal(mu == NR_MU_2,"Invalid cyclic prefix %d for numerology index %d\n", Ncp, mu); @@ -343,11 +341,8 @@ int nr_init_frame_parms_ue(NR_DL_FRAME_PARMS *fp, AssertFatal(fp->ul_CarrierFreq == (fp->dl_CarrierFreq + uplink_frequency_offset), "Disagreement in uplink frequency for band %d: ul_CarrierFreq = %lu Hz vs expected %lu Hz\n", fp->nr_band, fp->ul_CarrierFreq, fp->dl_CarrierFreq + uplink_frequency_offset); -#if PHYSIM - printf("Initializing UE frame parms for mu %d, N_RB %d, Ncp %d\n",fp->numerology_index, fp->N_RB_DL, Ncp); -#else LOG_I(PHY,"Initializing frame parms for mu %d, N_RB %d, Ncp %d\n",fp->numerology_index, fp->N_RB_DL, Ncp); -#endif + if (Ncp == NFAPI_CP_EXTENDED) AssertFatal(fp->numerology_index == NR_MU_2,"Invalid cyclic prefix %d for numerology index %d\n", Ncp, fp->numerology_index); diff --git a/openair1/PHY/NR_UE_TRANSPORT/nr_initial_sync.c b/openair1/PHY/NR_UE_TRANSPORT/nr_initial_sync.c index 44e103e90e..2963ad5010 100644 --- a/openair1/PHY/NR_UE_TRANSPORT/nr_initial_sync.c +++ b/openair1/PHY/NR_UE_TRANSPORT/nr_initial_sync.c @@ -353,21 +353,15 @@ int nr_initial_sync(UE_nr_rxtx_proc_t *proc, if( (abs(ue->common_vars.freq_offset) > 150) && (ret == 0) ) { ret=-1; -#if PHYSIM - printf("Ignore MIB with high freq offset [%d Hz] estimation \n",ue->common_vars.freq_offset); -#else LOG_E(HW, "Ignore MIB with high freq offset [%d Hz] estimation \n",ue->common_vars.freq_offset); -#endif }*/ if (ret==0) { // PBCH found so indicate sync to higher layers and configure frame parameters //#ifdef DEBUG_INITIAL_SYNCH -#if PHYSIM - printf("[UE%d] In synch, rx_offset %d samples\n",ue->Mod_id, ue->rx_offset); -#else + LOG_I(PHY, "[UE%d] In synch, rx_offset %d samples\n",ue->Mod_id, ue->rx_offset); -#endif + //#endif if (ue->UE_scan_carrier == 0) { From 52fc3ff5a54a61aa4391bd5839da42c93c1bbcf2 Mon Sep 17 00:00:00 2001 From: frtabu Date: Thu, 9 Dec 2021 14:41:06 +0100 Subject: [PATCH 06/23] Switch loglvl to ARNING by default as for other simulators --- openair1/SIMULATION/NR_PHY/dlsim.c | 2 +- openair1/SIMULATION/NR_PHY/ulsim.c | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-) diff --git a/openair1/SIMULATION/NR_PHY/dlsim.c b/openair1/SIMULATION/NR_PHY/dlsim.c index 80d1a512b1..9ba165f55f 100644 --- a/openair1/SIMULATION/NR_PHY/dlsim.c +++ b/openair1/SIMULATION/NR_PHY/dlsim.c @@ -400,7 +400,7 @@ int main(int argc, char **argv) NR_UE_MAC_INST_t *UE_mac; int cyclic_prefix_type = NFAPI_CP_NORMAL; int run_initial_sync=0; - int loglvl=OAILOG_INFO; + int loglvl=OAILOG_WARNING; //float target_error_rate = 0.01; int css_flag=0; diff --git a/openair1/SIMULATION/NR_PHY/ulsim.c b/openair1/SIMULATION/NR_PHY/ulsim.c index 9be5a18805..dcb1849ced 100644 --- a/openair1/SIMULATION/NR_PHY/ulsim.c +++ b/openair1/SIMULATION/NR_PHY/ulsim.c @@ -272,7 +272,7 @@ int main(int argc, char **argv) //unsigned char frame_type = 0; NR_DL_FRAME_PARMS *frame_parms; - int loglvl = OAILOG_INFO; + int loglvl = OAILOG_WARNING; //uint64_t SSB_positions=0x01; uint16_t nb_symb_sch = 12; int start_symbol = 0; From caf4035286cfc625b91e8c79874fbf24f135fa9c Mon Sep 17 00:00:00 2001 From: frtabu Date: Wed, 24 Nov 2021 11:00:05 +0100 Subject: [PATCH 07/23] Fix time measurment LOGS in phy_procedures_nr_UE.c --- openair1/SCHED_NR_UE/phy_procedures_nr_ue.c | 12 ++++++------ 1 file changed, 6 insertions(+), 6 deletions(-) diff --git a/openair1/SCHED_NR_UE/phy_procedures_nr_ue.c b/openair1/SCHED_NR_UE/phy_procedures_nr_ue.c index b65cb333f7..30480b89d9 100644 --- a/openair1/SCHED_NR_UE/phy_procedures_nr_ue.c +++ b/openair1/SCHED_NR_UE/phy_procedures_nr_ue.c @@ -1025,12 +1025,12 @@ bool nr_ue_dlsch_procedures(PHY_VARS_NR_UE *ue, stop_meas(&ue->dlsch_decoding_stats[proc->thread_id]); - if (cpumeas(CPUMEAS_GETSTATE)) { - LOG_D(PHY, " --> Unscrambling for CW0 %5.3f\n", - (ue->dlsch_unscrambling_stats.p_time)/(cpuf*1000.0)); - LOG_D(PHY, "AbsSubframe %d.%d --> LDPC Decoding for CW0 %5.3f\n", - frame_rx%1024, nr_slot_rx,(ue->dlsch_decoding_stats[proc->thread_id].p_time)/(cpuf*1000.0)); - } + if (cpumeas(CPUMEAS_GETSTATE)) { + LOG_D(PHY, " --> Unscrambling for CW0 %5.3f\n", + (ue->dlsch_unscrambling_stats.p_time)/(cpuf*1000.0)); + LOG_D(PHY, "AbsSubframe %d.%d --> LDPC Decoding for CW0 %5.3f\n", + frame_rx%1024, nr_slot_rx,(ue->dlsch_decoding_stats[proc->thread_id].p_time)/(cpuf*1000.0)); + } if(is_cw1_active) { // start ldpc decode for CW 1 From 4f7936565a23fc4ff0cff6793a3020f60d3c42fc Mon Sep 17 00:00:00 2001 From: Raymond Knopp Date: Fri, 3 Dec 2021 18:13:52 +0100 Subject: [PATCH 08/23] phy simulators remove measurments printf and check/allow log level config --- openair1/PHY/NR_UE_TRANSPORT/nr_dlsch_demodulation.c | 2 +- openair1/SCHED_NR_UE/phy_procedures_nr_ue.c | 1 + 2 files changed, 2 insertions(+), 1 deletion(-) diff --git a/openair1/PHY/NR_UE_TRANSPORT/nr_dlsch_demodulation.c b/openair1/PHY/NR_UE_TRANSPORT/nr_dlsch_demodulation.c index 1e031d531c..4412115d1e 100644 --- a/openair1/PHY/NR_UE_TRANSPORT/nr_dlsch_demodulation.c +++ b/openair1/PHY/NR_UE_TRANSPORT/nr_dlsch_demodulation.c @@ -460,7 +460,6 @@ int nr_rx_pdsch(PHY_VARS_NR_UE *ue, stop_meas(&ue->generic_stat_bis[proc->thread_id][slot]); if (cpumeas(CPUMEAS_GETSTATE)) LOG_D(PHY, "[AbsSFN %u.%d] Slot%d Symbol %d first_symbol_flag %d: Channel Level %5.2f \n",frame,nr_slot_rx,slot,symbol,first_symbol_flag,ue->generic_stat_bis[proc->thread_id][slot].p_time/(cpuf*1000.0)); - start_meas(&ue->generic_stat_bis[proc->thread_id][slot]); // Now channel compensation if (dlsch0_harq->mimo_modegeneric_stat_bis[proc->thread_id][slot]); if (cpumeas(CPUMEAS_GETSTATE)) LOG_D(PHY, "[AbsSFN %u.%d] Slot%d Symbol %d log2_maxh %d channel_level %d: Channel Comp %5.2f \n", frame, nr_slot_rx, slot, symbol, pdsch_vars[gNB_id]->log2_maxh, proc->channel_level, ue->generic_stat_bis[proc->thread_id][slot].p_time/(cpuf*1000.0)); + start_meas(&ue->generic_stat_bis[proc->thread_id][slot]); if (frame_parms->nb_antennas_rx > 1) { diff --git a/openair1/SCHED_NR_UE/phy_procedures_nr_ue.c b/openair1/SCHED_NR_UE/phy_procedures_nr_ue.c index 30480b89d9..9033c432fe 100644 --- a/openair1/SCHED_NR_UE/phy_procedures_nr_ue.c +++ b/openair1/SCHED_NR_UE/phy_procedures_nr_ue.c @@ -1858,6 +1858,7 @@ int phy_procedures_nrUE_RX(PHY_VARS_NR_UE *ue, LOG_D(PHY, "[SFN %d] Slot1: Pdsch Proc %5.2f\n",nr_slot_rx,ue->pdsch_procedures_stat[proc->thread_id].p_time/(cpuf*1000.0)); LOG_D(PHY, "[SFN %d] Slot0 Slot1: Dlsch Proc %5.2f\n",nr_slot_rx,ue->dlsch_procedures_stat[proc->thread_id].p_time/(cpuf*1000.0)); } + // deactivate dlsch once dlsch proc is done ue->dlsch[proc->thread_id][gNB_id][0]->active = 0; From 712f85760bac7f310d38d606fc587039dfd7e8eb Mon Sep 17 00:00:00 2001 From: frtabu Date: Fri, 10 Dec 2021 18:47:26 +0100 Subject: [PATCH 09/23] fix measurment flag enforced in nr_dlsch_decoding.c --- openair1/PHY/NR_UE_TRANSPORT/nr_dlsch_decoding.c | 1 - 1 file changed, 1 deletion(-) diff --git a/openair1/PHY/NR_UE_TRANSPORT/nr_dlsch_decoding.c b/openair1/PHY/NR_UE_TRANSPORT/nr_dlsch_decoding.c index b8cd59eead..95c451fa1a 100644 --- a/openair1/PHY/NR_UE_TRANSPORT/nr_dlsch_decoding.c +++ b/openair1/PHY/NR_UE_TRANSPORT/nr_dlsch_decoding.c @@ -642,7 +642,6 @@ uint32_t nr_dlsch_decoding(PHY_VARS_NR_UE *phy_vars_ue, if (LOG_DEBUGFLAG(DEBUG_DLSCH_DECOD)) LOG_I(PHY,"Segmentation: C %d, K %d\n",harq_process->C,harq_process->K); - opp_enabled=1; Kr = harq_process->K; // [hna] overwrites this line "Kr = p_decParams->Z*kb" Kr_bytes = Kr>>3; offset = 0; From 87bb0ffab06720bbd042eac91d113a5578bd898b Mon Sep 17 00:00:00 2001 From: frtabu Date: Wed, 24 Nov 2021 11:00:05 +0100 Subject: [PATCH 10/23] Fix time measurment LOGS in phy_procedures_nr_UE.c --- openair1/SCHED_NR_UE/phy_procedures_nr_ue.c | 17 +++++++---------- 1 file changed, 7 insertions(+), 10 deletions(-) diff --git a/openair1/SCHED_NR_UE/phy_procedures_nr_ue.c b/openair1/SCHED_NR_UE/phy_procedures_nr_ue.c index 9033c432fe..f2465127aa 100644 --- a/openair1/SCHED_NR_UE/phy_procedures_nr_ue.c +++ b/openair1/SCHED_NR_UE/phy_procedures_nr_ue.c @@ -1021,16 +1021,13 @@ bool nr_ue_dlsch_procedures(PHY_VARS_NR_UE *ue, LOG_D(PHY, "In %s DL PDU length in bits: %d, in bytes: %d \n", __FUNCTION__, dlsch0->harq_processes[harq_pid]->TBS, dlsch0->harq_processes[harq_pid]->TBS / 8); - - - - stop_meas(&ue->dlsch_decoding_stats[proc->thread_id]); - if (cpumeas(CPUMEAS_GETSTATE)) { - LOG_D(PHY, " --> Unscrambling for CW0 %5.3f\n", - (ue->dlsch_unscrambling_stats.p_time)/(cpuf*1000.0)); - LOG_D(PHY, "AbsSubframe %d.%d --> LDPC Decoding for CW0 %5.3f\n", - frame_rx%1024, nr_slot_rx,(ue->dlsch_decoding_stats[proc->thread_id].p_time)/(cpuf*1000.0)); - } + stop_meas(&ue->dlsch_decoding_stats[proc->thread_id]); + if (cpumeas(CPUMEAS_GETSTATE)) { + LOG_D(PHY, " --> Unscrambling for CW0 %5.3f\n", + (ue->dlsch_unscrambling_stats.p_time)/(cpuf*1000.0)); + LOG_D(PHY, "AbsSubframe %d.%d --> LDPC Decoding for CW0 %5.3f\n", + frame_rx%1024, nr_slot_rx,(ue->dlsch_decoding_stats[proc->thread_id].p_time)/(cpuf*1000.0)); + } if(is_cw1_active) { // start ldpc decode for CW 1 From 9cd976d6f76cdfd1daaba90b8fc1fd7f440e59cc Mon Sep 17 00:00:00 2001 From: Raymond Knopp Date: Fri, 3 Dec 2021 18:13:52 +0100 Subject: [PATCH 11/23] phy simulators remove measurments printf and check/allow log level config --- openair1/PHY/NR_UE_TRANSPORT/nr_dlsch_demodulation.c | 1 + 1 file changed, 1 insertion(+) diff --git a/openair1/PHY/NR_UE_TRANSPORT/nr_dlsch_demodulation.c b/openair1/PHY/NR_UE_TRANSPORT/nr_dlsch_demodulation.c index 4412115d1e..b4edae972a 100644 --- a/openair1/PHY/NR_UE_TRANSPORT/nr_dlsch_demodulation.c +++ b/openair1/PHY/NR_UE_TRANSPORT/nr_dlsch_demodulation.c @@ -460,6 +460,7 @@ int nr_rx_pdsch(PHY_VARS_NR_UE *ue, stop_meas(&ue->generic_stat_bis[proc->thread_id][slot]); if (cpumeas(CPUMEAS_GETSTATE)) LOG_D(PHY, "[AbsSFN %u.%d] Slot%d Symbol %d first_symbol_flag %d: Channel Level %5.2f \n",frame,nr_slot_rx,slot,symbol,first_symbol_flag,ue->generic_stat_bis[proc->thread_id][slot].p_time/(cpuf*1000.0)); + start_meas(&ue->generic_stat_bis[proc->thread_id][slot]); // Now channel compensation if (dlsch0_harq->mimo_mode Date: Wed, 24 Nov 2021 11:00:05 +0100 Subject: [PATCH 12/23] Fix time measurment LOGS in phy_procedures_nr_UE.c --- openair1/SCHED_NR_UE/phy_procedures_nr_ue.c | 3 +++ 1 file changed, 3 insertions(+) diff --git a/openair1/SCHED_NR_UE/phy_procedures_nr_ue.c b/openair1/SCHED_NR_UE/phy_procedures_nr_ue.c index f2465127aa..fde3434138 100644 --- a/openair1/SCHED_NR_UE/phy_procedures_nr_ue.c +++ b/openair1/SCHED_NR_UE/phy_procedures_nr_ue.c @@ -1391,6 +1391,7 @@ void *UE_thread_slot1_dl_processing(void *arg) { stop_meas(&ue->ue_front_end_per_slot_stat[proc->thread_id][1]); + if (cpumeas(CPUMEAS_GETSTATE)) LOG_D(PHY, "[AbsSFN %d.%d] Slot1: FFT + Channel Estimate + Pdsch Proc Slot0 %5.2f \n",frame_rx,nr_slot_rx,ue->ue_front_end_per_slot_stat[proc->thread_id][1].p_time/(cpuf*1000.0)); @@ -1851,6 +1852,7 @@ int phy_procedures_nrUE_RX(PHY_VARS_NR_UE *ue, dlsch_parallel); stop_meas(&ue->dlsch_procedures_stat[proc->thread_id]); + if (cpumeas(CPUMEAS_GETSTATE)) { LOG_D(PHY, "[SFN %d] Slot1: Pdsch Proc %5.2f\n",nr_slot_rx,ue->pdsch_procedures_stat[proc->thread_id].p_time/(cpuf*1000.0)); LOG_D(PHY, "[SFN %d] Slot0 Slot1: Dlsch Proc %5.2f\n",nr_slot_rx,ue->dlsch_procedures_stat[proc->thread_id].p_time/(cpuf*1000.0)); @@ -1940,6 +1942,7 @@ phy_procedures_emos_UE_RX(ue,slot,gNB_id); VCD_SIGNAL_DUMPER_DUMP_FUNCTION_BY_NAME(VCD_SIGNAL_DUMPER_FUNCTIONS_PHY_PROCEDURES_UE_RX, VCD_FUNCTION_OUT); stop_meas(&ue->phy_proc_rx[proc->thread_id]); + if (cpumeas(CPUMEAS_GETSTATE)) LOG_D(PHY, "------FULL RX PROC [SFN %d]: %5.2f ------\n",nr_slot_rx,ue->phy_proc_rx[proc->thread_id].p_time/(cpuf*1000.0)); From 379d17a5d977c8e7fadccb7e94b7722440b49340 Mon Sep 17 00:00:00 2001 From: Raymond Knopp Date: Fri, 3 Dec 2021 18:13:52 +0100 Subject: [PATCH 13/23] phy simulators remove measurments printf and check/allow log level config --- openair1/SCHED_NR_UE/phy_procedures_nr_ue.c | 15 +++++---------- 1 file changed, 5 insertions(+), 10 deletions(-) diff --git a/openair1/SCHED_NR_UE/phy_procedures_nr_ue.c b/openair1/SCHED_NR_UE/phy_procedures_nr_ue.c index fde3434138..d524d7713f 100644 --- a/openair1/SCHED_NR_UE/phy_procedures_nr_ue.c +++ b/openair1/SCHED_NR_UE/phy_procedures_nr_ue.c @@ -1078,19 +1078,15 @@ bool nr_ue_dlsch_procedures(PHY_VARS_NR_UE *ue, LOG_T(PHY,"CWW sequential dlsch decoding, ret1 = %d\n", ret1); } - stop_meas(&ue->dlsch_decoding_stats[proc->thread_id]); if (cpumeas(CPUMEAS_GETSTATE)) { - LOG_D(PHY, " --> Unscrambling for CW1 %5.3f\n", - (ue->dlsch_unscrambling_stats.p_time)/(cpuf*1000.0)); - LOG_D(PHY, "AbsSubframe %d.%d --> ldpc Decoding for CW1 %5.3f\n", - frame_rx%1024, nr_slot_rx,(ue->dlsch_decoding_stats[proc->thread_id].p_time)/(cpuf*1000.0)); + LOG_D(PHY, " --> Unscrambling for CW1 %5.3f\n", + (ue->dlsch_unscrambling_stats.p_time)/(cpuf*1000.0)); + LOG_D(PHY, "AbsSubframe %d.%d --> ldpc Decoding for CW1 %5.3f\n", + frame_rx%1024, nr_slot_rx,(ue->dlsch_decoding_stats[proc->thread_id].p_time)/(cpuf*1000.0)); } - LOG_D(PHY, "harq_pid: %d, TBS expected dlsch1: %d \n", harq_pid, dlsch1->harq_processes[harq_pid]->TBS); + LOG_D(PHY, "harq_pid: %d, TBS expected dlsch1: %d \n", harq_pid, dlsch1->harq_processes[harq_pid]->TBS); } - - LOG_D(PHY," ------ end ldpc decoder for AbsSubframe %d.%d ------ decoded in %d \n", frame_rx, nr_slot_rx, ret); - // send to mac if (ue->if_inst && ue->if_inst->dl_indication) { ue->if_inst->dl_indication(&dl_indication, ul_time_alignment); @@ -1391,7 +1387,6 @@ void *UE_thread_slot1_dl_processing(void *arg) { stop_meas(&ue->ue_front_end_per_slot_stat[proc->thread_id][1]); - if (cpumeas(CPUMEAS_GETSTATE)) LOG_D(PHY, "[AbsSFN %d.%d] Slot1: FFT + Channel Estimate + Pdsch Proc Slot0 %5.2f \n",frame_rx,nr_slot_rx,ue->ue_front_end_per_slot_stat[proc->thread_id][1].p_time/(cpuf*1000.0)); From add1293e9bfa2f28f03dc8e01b44c689390f3d49 Mon Sep 17 00:00:00 2001 From: frtabu Date: Wed, 24 Nov 2021 11:00:05 +0100 Subject: [PATCH 14/23] Fix time measurment LOGS in phy_procedures_nr_UE.c --- openair1/SCHED_NR_UE/phy_procedures_nr_ue.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/openair1/SCHED_NR_UE/phy_procedures_nr_ue.c b/openair1/SCHED_NR_UE/phy_procedures_nr_ue.c index d524d7713f..156aa7fff8 100644 --- a/openair1/SCHED_NR_UE/phy_procedures_nr_ue.c +++ b/openair1/SCHED_NR_UE/phy_procedures_nr_ue.c @@ -1847,7 +1847,6 @@ int phy_procedures_nrUE_RX(PHY_VARS_NR_UE *ue, dlsch_parallel); stop_meas(&ue->dlsch_procedures_stat[proc->thread_id]); - if (cpumeas(CPUMEAS_GETSTATE)) { LOG_D(PHY, "[SFN %d] Slot1: Pdsch Proc %5.2f\n",nr_slot_rx,ue->pdsch_procedures_stat[proc->thread_id].p_time/(cpuf*1000.0)); LOG_D(PHY, "[SFN %d] Slot0 Slot1: Dlsch Proc %5.2f\n",nr_slot_rx,ue->dlsch_procedures_stat[proc->thread_id].p_time/(cpuf*1000.0)); @@ -1941,6 +1940,7 @@ stop_meas(&ue->phy_proc_rx[proc->thread_id]); if (cpumeas(CPUMEAS_GETSTATE)) LOG_D(PHY, "------FULL RX PROC [SFN %d]: %5.2f ------\n",nr_slot_rx,ue->phy_proc_rx[proc->thread_id].p_time/(cpuf*1000.0)); + //#endif //pdsch LOG_D(PHY," ****** end RX-Chain for AbsSubframe %d.%d ****** \n", frame_rx%1024, nr_slot_rx); From 6d2672d7ba4c47a058dbabdec680d9ee12271f46 Mon Sep 17 00:00:00 2001 From: Raymond Knopp Date: Fri, 3 Dec 2021 18:13:52 +0100 Subject: [PATCH 15/23] phy simulators remove measurments printf and check/allow log level config --- openair1/SCHED_NR_UE/phy_procedures_nr_ue.c | 2 -- 1 file changed, 2 deletions(-) diff --git a/openair1/SCHED_NR_UE/phy_procedures_nr_ue.c b/openair1/SCHED_NR_UE/phy_procedures_nr_ue.c index 156aa7fff8..fb34fc6dd0 100644 --- a/openair1/SCHED_NR_UE/phy_procedures_nr_ue.c +++ b/openair1/SCHED_NR_UE/phy_procedures_nr_ue.c @@ -1936,11 +1936,9 @@ phy_procedures_emos_UE_RX(ue,slot,gNB_id); VCD_SIGNAL_DUMPER_DUMP_FUNCTION_BY_NAME(VCD_SIGNAL_DUMPER_FUNCTIONS_PHY_PROCEDURES_UE_RX, VCD_FUNCTION_OUT); stop_meas(&ue->phy_proc_rx[proc->thread_id]); - if (cpumeas(CPUMEAS_GETSTATE)) LOG_D(PHY, "------FULL RX PROC [SFN %d]: %5.2f ------\n",nr_slot_rx,ue->phy_proc_rx[proc->thread_id].p_time/(cpuf*1000.0)); - //#endif //pdsch LOG_D(PHY," ****** end RX-Chain for AbsSubframe %d.%d ****** \n", frame_rx%1024, nr_slot_rx); From aed13183fb8c7d8a57356c0687c19979429d9f6c Mon Sep 17 00:00:00 2001 From: frtabu Date: Wed, 24 Nov 2021 11:00:05 +0100 Subject: [PATCH 16/23] Fix time measurment LOGS in phy_procedures_nr_UE.c --- openair1/SCHED_NR_UE/phy_procedures_nr_ue.c | 3 +++ 1 file changed, 3 insertions(+) diff --git a/openair1/SCHED_NR_UE/phy_procedures_nr_ue.c b/openair1/SCHED_NR_UE/phy_procedures_nr_ue.c index fb34fc6dd0..4ba5cd677d 100644 --- a/openair1/SCHED_NR_UE/phy_procedures_nr_ue.c +++ b/openair1/SCHED_NR_UE/phy_procedures_nr_ue.c @@ -1387,6 +1387,7 @@ void *UE_thread_slot1_dl_processing(void *arg) { stop_meas(&ue->ue_front_end_per_slot_stat[proc->thread_id][1]); + if (cpumeas(CPUMEAS_GETSTATE)) LOG_D(PHY, "[AbsSFN %d.%d] Slot1: FFT + Channel Estimate + Pdsch Proc Slot0 %5.2f \n",frame_rx,nr_slot_rx,ue->ue_front_end_per_slot_stat[proc->thread_id][1].p_time/(cpuf*1000.0)); @@ -1847,6 +1848,7 @@ int phy_procedures_nrUE_RX(PHY_VARS_NR_UE *ue, dlsch_parallel); stop_meas(&ue->dlsch_procedures_stat[proc->thread_id]); + if (cpumeas(CPUMEAS_GETSTATE)) { LOG_D(PHY, "[SFN %d] Slot1: Pdsch Proc %5.2f\n",nr_slot_rx,ue->pdsch_procedures_stat[proc->thread_id].p_time/(cpuf*1000.0)); LOG_D(PHY, "[SFN %d] Slot0 Slot1: Dlsch Proc %5.2f\n",nr_slot_rx,ue->dlsch_procedures_stat[proc->thread_id].p_time/(cpuf*1000.0)); @@ -1936,6 +1938,7 @@ phy_procedures_emos_UE_RX(ue,slot,gNB_id); VCD_SIGNAL_DUMPER_DUMP_FUNCTION_BY_NAME(VCD_SIGNAL_DUMPER_FUNCTIONS_PHY_PROCEDURES_UE_RX, VCD_FUNCTION_OUT); stop_meas(&ue->phy_proc_rx[proc->thread_id]); + if (cpumeas(CPUMEAS_GETSTATE)) LOG_D(PHY, "------FULL RX PROC [SFN %d]: %5.2f ------\n",nr_slot_rx,ue->phy_proc_rx[proc->thread_id].p_time/(cpuf*1000.0)); From 9159e200028068a842483c5fc3d4018f86cf777e Mon Sep 17 00:00:00 2001 From: Raymond Knopp Date: Fri, 3 Dec 2021 18:13:52 +0100 Subject: [PATCH 17/23] phy simulators remove measurments printf and check/allow log level config --- openair1/SCHED_NR_UE/phy_procedures_nr_ue.c | 1 - 1 file changed, 1 deletion(-) diff --git a/openair1/SCHED_NR_UE/phy_procedures_nr_ue.c b/openair1/SCHED_NR_UE/phy_procedures_nr_ue.c index 4ba5cd677d..d524d7713f 100644 --- a/openair1/SCHED_NR_UE/phy_procedures_nr_ue.c +++ b/openair1/SCHED_NR_UE/phy_procedures_nr_ue.c @@ -1387,7 +1387,6 @@ void *UE_thread_slot1_dl_processing(void *arg) { stop_meas(&ue->ue_front_end_per_slot_stat[proc->thread_id][1]); - if (cpumeas(CPUMEAS_GETSTATE)) LOG_D(PHY, "[AbsSFN %d.%d] Slot1: FFT + Channel Estimate + Pdsch Proc Slot0 %5.2f \n",frame_rx,nr_slot_rx,ue->ue_front_end_per_slot_stat[proc->thread_id][1].p_time/(cpuf*1000.0)); From 6f13d453bfc4ecb4bfb8d1fcd5d660f96d3faf62 Mon Sep 17 00:00:00 2001 From: frtabu Date: Wed, 24 Nov 2021 11:00:05 +0100 Subject: [PATCH 18/23] Fix time measurment LOGS in phy_procedures_nr_UE.c --- openair1/SCHED_NR_UE/phy_procedures_nr_ue.c | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-) diff --git a/openair1/SCHED_NR_UE/phy_procedures_nr_ue.c b/openair1/SCHED_NR_UE/phy_procedures_nr_ue.c index d524d7713f..0aa6f87233 100644 --- a/openair1/SCHED_NR_UE/phy_procedures_nr_ue.c +++ b/openair1/SCHED_NR_UE/phy_procedures_nr_ue.c @@ -1480,6 +1480,7 @@ void *UE_thread_slot1_dl_processing(void *arg) { if (cpumeas(CPUMEAS_GETSTATE)) LOG_D(PHY, "[AbsSFN %d.%d] Slot1: LLR Computation %5.2f \n",frame_rx,nr_slot_rx,ue->pdsch_procedures_per_slot_stat[proc->thread_id][1].p_time/(cpuf*1000.0)); + if (pthread_mutex_lock(&proc->mutex_slot1_dl_processing) != 0) { LOG_E( PHY, "[SCHED][UE] error locking mutex for UE RXTX\n" ); exit_fun("noting to add"); @@ -1847,7 +1848,6 @@ int phy_procedures_nrUE_RX(PHY_VARS_NR_UE *ue, dlsch_parallel); stop_meas(&ue->dlsch_procedures_stat[proc->thread_id]); - if (cpumeas(CPUMEAS_GETSTATE)) { LOG_D(PHY, "[SFN %d] Slot1: Pdsch Proc %5.2f\n",nr_slot_rx,ue->pdsch_procedures_stat[proc->thread_id].p_time/(cpuf*1000.0)); LOG_D(PHY, "[SFN %d] Slot0 Slot1: Dlsch Proc %5.2f\n",nr_slot_rx,ue->dlsch_procedures_stat[proc->thread_id].p_time/(cpuf*1000.0)); @@ -1937,7 +1937,6 @@ phy_procedures_emos_UE_RX(ue,slot,gNB_id); VCD_SIGNAL_DUMPER_DUMP_FUNCTION_BY_NAME(VCD_SIGNAL_DUMPER_FUNCTIONS_PHY_PROCEDURES_UE_RX, VCD_FUNCTION_OUT); stop_meas(&ue->phy_proc_rx[proc->thread_id]); - if (cpumeas(CPUMEAS_GETSTATE)) LOG_D(PHY, "------FULL RX PROC [SFN %d]: %5.2f ------\n",nr_slot_rx,ue->phy_proc_rx[proc->thread_id].p_time/(cpuf*1000.0)); From a631bb2a58b406c69a022a17b0143b3b20d8bca2 Mon Sep 17 00:00:00 2001 From: Raymond Knopp Date: Fri, 3 Dec 2021 18:13:52 +0100 Subject: [PATCH 19/23] phy simulators remove measurments printf and check/allow log level config --- openair1/SCHED_NR_UE/phy_procedures_nr_ue.c | 2 ++ 1 file changed, 2 insertions(+) diff --git a/openair1/SCHED_NR_UE/phy_procedures_nr_ue.c b/openair1/SCHED_NR_UE/phy_procedures_nr_ue.c index 0aa6f87233..af01f4116b 100644 --- a/openair1/SCHED_NR_UE/phy_procedures_nr_ue.c +++ b/openair1/SCHED_NR_UE/phy_procedures_nr_ue.c @@ -1078,6 +1078,7 @@ bool nr_ue_dlsch_procedures(PHY_VARS_NR_UE *ue, LOG_T(PHY,"CWW sequential dlsch decoding, ret1 = %d\n", ret1); } + stop_meas(&ue->dlsch_decoding_stats[proc->thread_id]); if (cpumeas(CPUMEAS_GETSTATE)) { LOG_D(PHY, " --> Unscrambling for CW1 %5.3f\n", @@ -1853,6 +1854,7 @@ int phy_procedures_nrUE_RX(PHY_VARS_NR_UE *ue, LOG_D(PHY, "[SFN %d] Slot0 Slot1: Dlsch Proc %5.2f\n",nr_slot_rx,ue->dlsch_procedures_stat[proc->thread_id].p_time/(cpuf*1000.0)); } + // deactivate dlsch once dlsch proc is done ue->dlsch[proc->thread_id][gNB_id][0]->active = 0; From 88f67ded26c26a49426819535a23ae3863efc9e4 Mon Sep 17 00:00:00 2001 From: frtabu Date: Wed, 24 Nov 2021 11:00:05 +0100 Subject: [PATCH 20/23] Fix time measurment LOGS in phy_procedures_nr_UE.c --- openair1/SCHED_NR_UE/phy_procedures_nr_ue.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/openair1/SCHED_NR_UE/phy_procedures_nr_ue.c b/openair1/SCHED_NR_UE/phy_procedures_nr_ue.c index af01f4116b..1054a40807 100644 --- a/openair1/SCHED_NR_UE/phy_procedures_nr_ue.c +++ b/openair1/SCHED_NR_UE/phy_procedures_nr_ue.c @@ -1481,7 +1481,6 @@ void *UE_thread_slot1_dl_processing(void *arg) { if (cpumeas(CPUMEAS_GETSTATE)) LOG_D(PHY, "[AbsSFN %d.%d] Slot1: LLR Computation %5.2f \n",frame_rx,nr_slot_rx,ue->pdsch_procedures_per_slot_stat[proc->thread_id][1].p_time/(cpuf*1000.0)); - if (pthread_mutex_lock(&proc->mutex_slot1_dl_processing) != 0) { LOG_E( PHY, "[SCHED][UE] error locking mutex for UE RXTX\n" ); exit_fun("noting to add"); @@ -1849,12 +1848,12 @@ int phy_procedures_nrUE_RX(PHY_VARS_NR_UE *ue, dlsch_parallel); stop_meas(&ue->dlsch_procedures_stat[proc->thread_id]); + if (cpumeas(CPUMEAS_GETSTATE)) { LOG_D(PHY, "[SFN %d] Slot1: Pdsch Proc %5.2f\n",nr_slot_rx,ue->pdsch_procedures_stat[proc->thread_id].p_time/(cpuf*1000.0)); LOG_D(PHY, "[SFN %d] Slot0 Slot1: Dlsch Proc %5.2f\n",nr_slot_rx,ue->dlsch_procedures_stat[proc->thread_id].p_time/(cpuf*1000.0)); } - // deactivate dlsch once dlsch proc is done ue->dlsch[proc->thread_id][gNB_id][0]->active = 0; @@ -1942,6 +1941,7 @@ stop_meas(&ue->phy_proc_rx[proc->thread_id]); if (cpumeas(CPUMEAS_GETSTATE)) LOG_D(PHY, "------FULL RX PROC [SFN %d]: %5.2f ------\n",nr_slot_rx,ue->phy_proc_rx[proc->thread_id].p_time/(cpuf*1000.0)); + //#endif //pdsch LOG_D(PHY," ****** end RX-Chain for AbsSubframe %d.%d ****** \n", frame_rx%1024, nr_slot_rx); From 4f5ea3bb6fba92aff8a69cae4b8e497fe994fe7e Mon Sep 17 00:00:00 2001 From: Raymond Knopp Date: Fri, 3 Dec 2021 18:13:52 +0100 Subject: [PATCH 21/23] phy simulators remove measurments printf and check/allow log level config --- openair1/SCHED_NR_UE/phy_procedures_nr_ue.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/openair1/SCHED_NR_UE/phy_procedures_nr_ue.c b/openair1/SCHED_NR_UE/phy_procedures_nr_ue.c index 1054a40807..9b4f667dbe 100644 --- a/openair1/SCHED_NR_UE/phy_procedures_nr_ue.c +++ b/openair1/SCHED_NR_UE/phy_procedures_nr_ue.c @@ -1085,7 +1085,8 @@ bool nr_ue_dlsch_procedures(PHY_VARS_NR_UE *ue, (ue->dlsch_unscrambling_stats.p_time)/(cpuf*1000.0)); LOG_D(PHY, "AbsSubframe %d.%d --> ldpc Decoding for CW1 %5.3f\n", frame_rx%1024, nr_slot_rx,(ue->dlsch_decoding_stats[proc->thread_id].p_time)/(cpuf*1000.0)); - } + } + } LOG_D(PHY, "harq_pid: %d, TBS expected dlsch1: %d \n", harq_pid, dlsch1->harq_processes[harq_pid]->TBS); } // send to mac @@ -1848,7 +1849,6 @@ int phy_procedures_nrUE_RX(PHY_VARS_NR_UE *ue, dlsch_parallel); stop_meas(&ue->dlsch_procedures_stat[proc->thread_id]); - if (cpumeas(CPUMEAS_GETSTATE)) { LOG_D(PHY, "[SFN %d] Slot1: Pdsch Proc %5.2f\n",nr_slot_rx,ue->pdsch_procedures_stat[proc->thread_id].p_time/(cpuf*1000.0)); LOG_D(PHY, "[SFN %d] Slot0 Slot1: Dlsch Proc %5.2f\n",nr_slot_rx,ue->dlsch_procedures_stat[proc->thread_id].p_time/(cpuf*1000.0)); From 1dc47e5db11d3504e3141134f623da2c15fe6d08 Mon Sep 17 00:00:00 2001 From: frtabu Date: Thu, 9 Dec 2021 14:41:06 +0100 Subject: [PATCH 22/23] Switch loglvl to ARNING by default as for other simulators --- openair1/SCHED_NR_UE/phy_procedures_nr_ue.c | 1 - 1 file changed, 1 deletion(-) diff --git a/openair1/SCHED_NR_UE/phy_procedures_nr_ue.c b/openair1/SCHED_NR_UE/phy_procedures_nr_ue.c index 9b4f667dbe..81e2fa0ad5 100644 --- a/openair1/SCHED_NR_UE/phy_procedures_nr_ue.c +++ b/openair1/SCHED_NR_UE/phy_procedures_nr_ue.c @@ -846,7 +846,6 @@ int nr_ue_pdsch_procedures(PHY_VARS_NR_UE *ue, UE_nr_rxtx_proc_t *proc, int gNB_ stop_meas(&ue->dlsch_llr_stats_parallelization[proc->thread_id][slot]); if (cpumeas(CPUMEAS_GETSTATE)) LOG_D(PHY, "[AbsSFN %d.%d] LLR Computation Symbol %d %5.2f \n",frame_rx,nr_slot_rx,m,ue->dlsch_llr_stats_parallelization[proc->thread_id][slot].p_time/(cpuf*1000.0)); - if(first_symbol_flag) { proc->first_symbol_available = 1; } From a7b6c4bcb0f81e4f46235be4958e14b69c1b73e9 Mon Sep 17 00:00:00 2001 From: frtabu Date: Tue, 4 Jan 2022 18:38:51 +0100 Subject: [PATCH 23/23] fix a warning. fix openCL regression after rebase --- cmake_targets/CMakeLists.txt | 4 +- .../nrLDPC_decoder_kernels_CL.c | 292 ++++++++++++++++++ openair1/PHY/LTE_UE_TRANSPORT/pcfich_ue.c | 3 +- openair1/SCHED_NR_UE/phy_procedures_nr_ue.c | 1 - 4 files changed, 295 insertions(+), 5 deletions(-) create mode 100644 openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_decoder_kernels_CL.c diff --git a/cmake_targets/CMakeLists.txt b/cmake_targets/CMakeLists.txt index af3baa2584..3f27379d1d 100644 --- a/cmake_targets/CMakeLists.txt +++ b/cmake_targets/CMakeLists.txt @@ -1532,8 +1532,8 @@ set(PHY_LDPC_CL_SRC ${OPENAIR1_DIR}/PHY/CODING/nrLDPC_encoder/ldpc_encoder_optim8segmulti.c ) add_custom_target( nrLDPC_decoder_kernels_CL - COMMAND gcc ${OPENAIR1_DIR}/PHY/CODING/nrLDPC_decoder/nrLDPC_decoder_kernels_CL.c -dD -DNRLDPC_KERNEL_SOURCE -E -o ${CMAKE_CURRENT_BINARY_DIR}/nrLDPC_decoder_kernels_CL.clc - SOURCES ${OPENAIR1_DIR}/PHY/CODING/nrLDPC_decoder/nrLDPC_decoder_kernels_CL.c + COMMAND gcc ${OPENAIR1_DIR}/PHY/CODING/nrLDPC_decoder/nrLDPC_decoder_CL.c -dD -DNRLDPC_KERNEL_SOURCE -E -o ${CMAKE_CURRENT_BINARY_DIR}/nrLDPC_decoder_kernels_CL.clc + SOURCES ${OPENAIR1_DIR}/PHY/CODING/nrLDPC_decoder/nrLDPC_decoder_CL.c ) set(PHY_NR_CODINGIF diff --git a/openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_decoder_kernels_CL.c b/openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_decoder_kernels_CL.c new file mode 100644 index 0000000000..42b25e41c0 --- /dev/null +++ b/openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_decoder_kernels_CL.c @@ -0,0 +1,292 @@ +/* + * 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_decoder_kernels_CL.cl +* \brief kernel functions for ldpc decoder accelerated via openCL +* \author Francois TABURET +* \date 2021 +* \version 1.0 +* \company Nokia BellLabs France +* \email: francois.taburet@nokia-bell-labs.com +* \note initial implem - translation of cuda version +* \warning +*/ + + +//__global char dev_dt [46*68*384]; +//__local char *dev_t; +//__global char dev_llr[68*384]; +//__global unsigned char dev_tmp[68*384]; + + +#define INT32_MAX 2147483647 + +//__constant h_element dev_h_compact1[46*19] = {}; // used in kernel 1 +//__constant h_element dev_h_compact2[68*30] = {}; // used in kernel 2 + +// __device__ __constantant__ h_element dev_h_compact1[46*19]; // used in kernel 1 +// __device__ __constantant__ h_element dev_h_compact2[68*30]; // used in kernel 2 + +// row and col element count +__constant char h_ele_row_bg1_count[46] = { + 19, 19, 19, 19, 3, 8, 9, 7, 10, 9, + 7, 8, 7, 6, 7, 7, 6, 6, 6, 6, + 6, 6, 5, 5, 6, 5, 5, 4, 5, 5, + 5, 5, 5, 5, 5, 5, 5, 4, 5, 5, + 4, 5, 4, 5, 5, 4}; +__constant char h_ele_col_bg1_count[68] = { + 30, 28, 7, 11, 9, 4, 8, 12, 8, 7, + 12, 10, 12, 11, 10, 7, 10, 10, 13, 7, + 8, 11, 12, 5, 6, 6, 1, 1, 1, 1, + 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, + 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, + 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, + 1, 1, 1, 1, 1, 1, 1, 1}; +__constant char h_ele_row_bg2_count[42] = { + 8, 10, 8, 10, 4, 6, 6, 6, 4, 5, + 5, 5, 4, 5, 5, 4, 5, 5, 4, 4, + 4, 4, 3, 4, 4, 3, 5, 3, 4, 3, + 5, 3, 4, 4, 4, 4, 4, 3, 4, 4, + 4, 4}; +__constant char h_ele_col_bg2_count[52] = { + 22, 23, 10, 5, 5, 14, 7, 13, 6, 8, + 9, 16, 9, 12, 1, 1, 1, 1, 1, 1, + 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, + 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, + 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, + 1, 1}; + + + +// Kernel 1 +__kernel void ldpc_cnp_kernel_1st_iter( __global char * dev_llr, __global char * dev_dt, __local h_element *dev_h_compact1, int BG, int row, int col, int Zc) +{ +// int iMCW = blockIdx.y; // codeword id +// int iBlkRow = blockIdx.x; // block row in h_base +// int iSubRow = threadIdx.x; // row index in sub_block of h_base +// if(blockIdx.x == 0 && threadIdx.x == 1) printf("cnp %d\n", threadIdx.x); + int iMCW = get_group_id(1); // codeword id + int iBlkRow = get_group_id(0); // block row in h_base + int iBlkCol; // block col in h_base + int iSubRow = get_local_id(0);; // row index in sub_block of h_base + int iCol; // overall col index in h_base + int offsetR; + int shift_t; + +// For 2-min algorithm. + int Q_sign = 0; + int sq; + int Q, Q_abs; + int R_temp; + + int sign = 1; + int rmin1 = INT32_MAX; + int rmin2 = INT32_MAX; + char idx_min = 0; + + h_element h_element_t; + int s = (BG==1)? h_ele_row_bg1_count[iBlkRow]:h_ele_row_bg2_count[iBlkRow]; + offsetR = (iMCW * row*col*Zc) + iBlkRow * Zc + iSubRow; // row*col*Zc = size of dev_dt +// if(blockIdx.x == 0 && threadIdx.x == 1) printf("s: %d, offset %d\n", s, offsetR); +// The 1st recursion + for(int i = 0; i < s; i++) // loop through all the ZxZ sub-blocks in a row + { + h_element_t = dev_h_compact1[i*row+iBlkRow]; // compact_col == row + + iBlkCol = h_element_t.y; + shift_t = h_element_t.value; + + shift_t = (iSubRow + shift_t) % Zc; + iCol = (iMCW * col*Zc) + iBlkCol * Zc + shift_t; // col*Zc = size of llr + Q = dev_llr[iCol]; + Q_abs = (Q>0)? Q : -Q; + sq = Q < 0; +// if(blockIdx.x == 0 && threadIdx.x == 1) printf("i %d, icol %d, Q: %d\n", i, iCol, Q); + // quick version + sign = sign * (1 - sq * 2); + Q_sign |= sq << i; + + if (Q_abs < rmin1){ + rmin2 = rmin1; + rmin1 = Q_abs; + idx_min = i; + } else if (Q_abs < rmin2){ + rmin2 = Q_abs; + } + } + +// if(blockIdx.x == 0 && threadIdx.x == 1)printf("min1 %d, min2 %d, min1_idx %d\n", rmin1, rmin2, idx_min); + +// The 2nd recursion + for(int i = 0; i < s; i++){ + // v0: Best performance so far. 0.75f is the value of alpha. + sq = 1 - 2 * ((Q_sign >> i) & 0x01); + R_temp = 0.75f * sign * sq * (i != idx_min ? rmin1 : rmin2); + // write results to global memory + h_element_t = dev_h_compact1[i*row+iBlkRow]; + int addr_temp = offsetR + h_element_t.y * row * Zc; + dev_dt[addr_temp] = R_temp; +// if(blockIdx.x == 0 && threadIdx.x == 1)printf("R_temp %d, temp_addr %d\n", R_temp, addr_temp); + } +} + +// Kernel_1 +__kernel void ldpc_cnp_kernel( __global char * dev_llr, __global char * dev_dt, __local h_element *dev_h_compact1, int BG, int row, int col, int Zc) +{ +// if(blockIdx.x == 0 && threadIdx.x == 1) printf("cnp\n"); +// int iMCW = blockIdx.y; +// int iBlkRow = blockIdx.x; // block row in h_base // block col in h_base +// int iSubRow = threadIdx.x; // row index in sub_block of h_base + int iMCW = get_group_id(1); + int iBlkRow = get_group_id(0); // block row in h_base + int iBlkCol; // block col in h_base + int iSubRow = get_local_id(0);; // row index in sub_block of h_base + int iCol; // overall col index in h_base + int offsetR; + int shift_t; + +// For 2-min algorithm. + int Q_sign = 0; + int sq; + int Q, Q_abs; + int R_temp; + + int sign = 1; + int rmin1 = INT32_MAX; + int rmin2 = INT32_MAX; + char idx_min = 0; + + h_element h_element_t; + int s = (BG==1)? h_ele_row_bg1_count[iBlkRow]: h_ele_row_bg2_count[iBlkRow]; + offsetR = (iMCW *row*col*Zc) + iBlkRow * Zc + iSubRow; // row * col * Zc = size of dev_dt +// if(blockIdx.x == 0 && threadIdx.x == 1) printf("s: %d, offset %d\n", s, offsetR); +// The 1st recursion + for(int i = 0; i < s; i++) // loop through all the ZxZ sub-blocks in a row + { + h_element_t = dev_h_compact1[i*row+iBlkRow]; + + iBlkCol = h_element_t.y; + shift_t = h_element_t.value; + shift_t = (iSubRow + shift_t) % Zc; + iCol = iBlkCol * Zc + shift_t; + + R_temp = dev_dt[offsetR + iBlkCol * row * Zc]; + + Q = dev_llr[iMCW * (col*Zc) + iCol] - R_temp; + Q_abs = (Q>0)? Q : -Q; +// if(blockIdx.x == 0 && threadIdx.x == 1) printf("i %d, icol %d, Q: %d\n", i, iCol, Q); + sq = Q < 0; + sign = sign * (1 - sq * 2); + Q_sign |= sq << i; + + if (Q_abs < rmin1){ + rmin2 = rmin1; + rmin1 = Q_abs; + idx_min = i; + } else if (Q_abs < rmin2){ + rmin2 = Q_abs; + } + + } + +// if(blockIdx.x == 0 && threadIdx.x == 1)printf("min1 %d, min2 %d, min1_idx %d\n", rmin1, rmin2, idx_min); + +// The 2nd recursion + for(int i = 0; i < s; i ++){ + sq = 1 - 2 * ((Q_sign >> i) & 0x01); + R_temp = 0.75f * sign * sq * (i != idx_min ? rmin1 : rmin2); + + + // write results to global memory + h_element_t = dev_h_compact1[i*row+iBlkRow]; + int addr_temp = h_element_t.y * row * Zc + offsetR; + dev_dt[addr_temp] = R_temp; +// if(blockIdx.x == 0 && threadIdx.x == 1)printf("R_temp %d, temp_addr %d\n", R_temp, addr_temp); + } +} + +// Kernel 2: VNP processing +__kernel void +ldpc_vnp_kernel_normal(__global char * dev_llr, __global char * dev_dt, __global char * dev_const_llr, __local h_element *dev_h_compact2, int BG, int row, int col, int Zc) +{ +// int iMCW = blockIdx.y; +// int iBlkCol = blockIdx.x; +// int iSubCol = threadIdx.x; + int iMCW = get_group_id(1); + int iBlkCol = get_group_id(0); + int iBlkRow; + int iSubCol = get_local_id(0); + int iRow; + int iCol; + + int shift_t, sf; + int APP; + + h_element h_element_t; + + // update all the llr values + iCol = iBlkCol * Zc + iSubCol; + APP = dev_const_llr[iMCW *col*Zc + iCol]; + int offsetDt = iMCW *row*col*Zc + iBlkCol * row * Zc; + int s = (BG==1)? h_ele_col_bg1_count[iBlkCol]:h_ele_col_bg2_count[iBlkCol]; + + for(int i = 0; i < s; i++) + { + h_element_t = dev_h_compact2[i*col+iBlkCol]; + + shift_t = h_element_t.value%Zc; + iBlkRow = h_element_t.x; + + sf = iSubCol - shift_t; + sf = (sf + Zc) % Zc; + + iRow = iBlkRow * Zc + sf; + APP = APP + dev_dt[offsetDt + iRow]; + } + if(APP > SCHAR_MAX) APP = SCHAR_MAX; + if(APP < SCHAR_MIN) APP = SCHAR_MIN; + // write back to device global memory + dev_llr[iMCW *col*Zc + iCol] = APP; +} + + +__kernel void pack_decoded_bit(__global unsigned char * dev_llr, __global unsigned char * dev_tmp, int col, int Zc) +{ +// int iMCW = blockIdx.y; +// int btid = threadIdx.x; + unsigned char tmp[128]; + int iMCW = get_group_id(1); + int btid = get_local_id(0); + int tid = iMCW * col*Zc + get_group_id(0)*128 + btid; + tmp[btid] = 0; + + if(dev_llr[tid] < 0){ + tmp[btid] = 1 << (7-(btid&7)); + } +// __syncthreads(); + + if(btid < 16){ + dev_tmp[iMCW * col*Zc + get_group_id(0)*16+btid] = 0; + for(int i = 0; i < 8; i++){ + dev_tmp[iMCW * col*Zc + get_group_id(0)*16+btid] += tmp[btid*8+i]; + } + } +} diff --git a/openair1/PHY/LTE_UE_TRANSPORT/pcfich_ue.c b/openair1/PHY/LTE_UE_TRANSPORT/pcfich_ue.c index 82bb0a2659..83a6b3234a 100644 --- a/openair1/PHY/LTE_UE_TRANSPORT/pcfich_ue.c +++ b/openair1/PHY/LTE_UE_TRANSPORT/pcfich_ue.c @@ -46,10 +46,9 @@ void pcfich_unscrambling(LTE_DL_FRAME_PARMS *frame_parms, uint32_t i; uint8_t reset; - uint32_t x1, x2, s=0; + uint32_t x1=0, x2, s=0; // x1 is set in lte_gold_generic reset = 1; - // x1 is set in lte_gold_generic x2 = ((((2*frame_parms->Nid_cell)+1)*(1+subframe))<<9) + frame_parms->Nid_cell; //this is c_init in 36.211 Sec 6.7.1 for (i=0; i<32; i++) { diff --git a/openair1/SCHED_NR_UE/phy_procedures_nr_ue.c b/openair1/SCHED_NR_UE/phy_procedures_nr_ue.c index 81e2fa0ad5..f6b36ddce7 100644 --- a/openair1/SCHED_NR_UE/phy_procedures_nr_ue.c +++ b/openair1/SCHED_NR_UE/phy_procedures_nr_ue.c @@ -1085,7 +1085,6 @@ bool nr_ue_dlsch_procedures(PHY_VARS_NR_UE *ue, LOG_D(PHY, "AbsSubframe %d.%d --> ldpc Decoding for CW1 %5.3f\n", frame_rx%1024, nr_slot_rx,(ue->dlsch_decoding_stats[proc->thread_id].p_time)/(cpuf*1000.0)); } - } LOG_D(PHY, "harq_pid: %d, TBS expected dlsch1: %d \n", harq_pid, dlsch1->harq_processes[harq_pid]->TBS); } // send to mac