From 64e610361f76b0c68dbae338ad13ce5a1d9fbd33 Mon Sep 17 00:00:00 2001 From: Raymond Knopp Date: Sun, 31 Oct 2021 22:43:20 +0100 Subject: [PATCH] merge from Laurent's TX-optimization-lts (cherry-picked dlsch_encoding component) --- cmake_targets/CMakeLists.txt | 8 +- executables/nr-gnb.c | 10 +- openair1/PHY/CODING/nrLDPC_defs.h | 19 +- .../nrLDPC_encoder/ldpc_encode_parity_check.c | 13 +- .../nrLDPC_encoder/ldpc_encoder_optim.c | 18 +- .../nrLDPC_encoder/ldpc_encoder_optim8seg.c | 17 +- .../ldpc_encoder_optim8segmulti.c | 95 ++--- openair1/PHY/NR_TRANSPORT/nr_dlsch.c | 9 +- openair1/PHY/NR_TRANSPORT/nr_dlsch.h | 1 + openair1/PHY/NR_TRANSPORT/nr_dlsch_coding.c | 387 ++++++++---------- openair1/PHY/NR_TRANSPORT/pucch_rx.c | 2 +- .../NR_UE_TRANSPORT/nr_dlsch_demodulation.c | 12 +- openair1/PHY/defs_gNB.h | 14 - openair1/SCHED_NR_UE/phy_procedures_nr_ue.c | 42 +- openair1/SIMULATION/NR_PHY/dlschsim.c | 10 +- openair1/SIMULATION/NR_PHY/dlsim.c | 47 ++- openair2/LAYER2/NR_MAC_gNB/main.c | 2 +- 17 files changed, 293 insertions(+), 413 deletions(-) diff --git a/cmake_targets/CMakeLists.txt b/cmake_targets/CMakeLists.txt index 9592334125..86b024d361 100644 --- a/cmake_targets/CMakeLists.txt +++ b/cmake_targets/CMakeLists.txt @@ -314,9 +314,9 @@ add_definitions("-DSTDC_HEADERS=1 -DHAVE_SYS_TYPES_H=1 -DHAVE_SYS_STAT_H=1 -DHAV set(commonOpts "-pipe -Wno-packed-bitfield-compat -fPIC -Wall -fno-strict-aliasing -rdynamic") set(CMAKE_C_FLAGS - "${CMAKE_C_FLAGS} ${C_FLAGS_PROCESSOR} ${commonOpts} -std=gnu99 -funroll-loops") + "${CMAKE_C_FLAGS} ${C_FLAGS_PROCESSOR} ${commonOpts} -std=gnu99 -funroll-loops -march=native") set(CMAKE_CXX_FLAGS - "${CMAKE_CXX_FLAGS} ${C_FLAGS_PROCESSOR} ${commonOpts} -std=c++11") + "${CMAKE_CXX_FLAGS} ${C_FLAGS_PROCESSOR} ${commonOpts} -std=c++11 -march=native") # cuda compiler bug (limitation) on complex macro definition if (CUDA_FOUND) @@ -332,11 +332,11 @@ endif () add_definitions("-DASN_DISABLE_OER_SUPPORT") ######################### -set(CMAKE_EXE_LINKER_FLAGS "${CMAKE_EXE_LINKER_FLAGS} -ggdb3 -Wl,-rpath -Wl,${CMAKE_CURRENT_BINARY_DIR}") +set(CMAKE_EXE_LINKER_FLAGS "${CMAKE_EXE_LINKER_FLAGS} -ggdb2 -Wl,-rpath -Wl,${CMAKE_CURRENT_BINARY_DIR}") ######################### # set a flag for changes in the source code # these changes are related to hardcoded path to include .h files -set(debugOpt "-ggdb3 -DMALLOC_CHECK_=3 -fno-delete-null-pointer-checks") +set(debugOpt "-ggdb2 -DMALLOC_CHECK_=3 -fno-delete-null-pointer-checks") set(CMAKE_C_FLAGS_DEBUG "${debugOpt} -O0") set(CMAKE_C_FLAGS_RELWITHDEBINFO "${debugOpt} -O2") set(CMAKE_C_FLAGS_RELEASE "-O3") diff --git a/executables/nr-gnb.c b/executables/nr-gnb.c index 95b983190e..439f4691d3 100644 --- a/executables/nr-gnb.c +++ b/executables/nr-gnb.c @@ -387,15 +387,15 @@ void init_gNB_Tpool(int inst) { LOG_I(PHY,"Number of threads requested in config file: %d, Number of threads available on this machine: %d\n",gNB->pusch_proc_threads,numCPU); int threadCnt = min(numCPU, gNB->pusch_proc_threads); if (threadCnt < 2) LOG_E(PHY,"Number of threads for gNB should be more than 1. Allocated only %d\n",threadCnt); - char ul_pool[80]; - sprintf(ul_pool,"-1"); + char pool[80]; + sprintf(pool,"-1"); int s_offset = 0; for (int icpu=1; icputhreadPool, false); + if (getenv("noThreads")) strcpy(pool, "n"); + initTpool(pool, gNB->threadPool, false); // ULSCH decoder result FIFO gNB->respDecode = (notifiedFIFO_t*) malloc(sizeof(notifiedFIFO_t)); initNotifiedFIFO(gNB->respDecode); diff --git a/openair1/PHY/CODING/nrLDPC_defs.h b/openair1/PHY/CODING/nrLDPC_defs.h index b96067f4b6..c902ecb2b6 100644 --- a/openair1/PHY/CODING/nrLDPC_defs.h +++ b/openair1/PHY/CODING/nrLDPC_defs.h @@ -22,6 +22,7 @@ // encoder interface #ifndef __NRLDPC_DEFS__H__ #define __NRLDPC_DEFS__H__ +#include #include "openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_types.h" /** \brief LDPC encoder @@ -36,13 +37,27 @@ \param 9-12 time_stats_t *tinput,*tprep, *tparity,*toutput */ typedef struct { - int n_segments; // optim8seg + unsigned int n_segments; // optim8seg unsigned int macro_num; // optim8segmulti unsigned char gen_code; //orig time_stats_t *tinput; time_stats_t *tprep; time_stats_t *tparity; time_stats_t *toutput; + int Kr; + uint32_t Kb; + uint32_t *Zc; + void *harq; + /// Encoder BG + uint8_t BG; + /// Interleaver outputs + unsigned char *output; + /// Number of bits in "small" code segments + uint32_t K; + /// Number of "Filler" bits + uint32_t F; + /// LDPC-code outputs + uint8_t *d[MAX_NUM_NR_DLSCH_SEGMENTS]; }encoder_implemparams_t; #define INIT0_LDPCIMPLEMPARAMS {0,0,0,NULL,NULL,NULL,NULL} typedef int(*nrLDPC_encoderfunc_t)(unsigned char **,unsigned char **,int,int,short, short, encoder_implemparams_t*); @@ -56,4 +71,4 @@ typedef int(*nrLDPC_encoderfunc_t)(unsigned char **,unsigned char **,int,int,sho \param p_profiler LDPC profiler statistics */ typedef int32_t(*nrLDPC_decoderfunc_t)(t_nrLDPC_dec_params* , int8_t*, int8_t* , t_nrLDPC_procBuf* , t_nrLDPC_time_stats* ); -#endif \ No newline at end of file +#endif diff --git a/openair1/PHY/CODING/nrLDPC_encoder/ldpc_encode_parity_check.c b/openair1/PHY/CODING/nrLDPC_encoder/ldpc_encode_parity_check.c index 738fd49b7d..2286846695 100644 --- a/openair1/PHY/CODING/nrLDPC_encoder/ldpc_encode_parity_check.c +++ b/openair1/PHY/CODING/nrLDPC_encoder/ldpc_encode_parity_check.c @@ -73,9 +73,18 @@ -static inline void encode_parity_check_part_optim(uint8_t *c,uint8_t *d, short BG,short Zc,short Kb) +static void encode_parity_check_part_optim(uint8_t *cc,uint8_t *d, short BG,short Zc,short Kb, int simd_size, int ncols) { - + unsigned char c[2*22*Zc*simd_size] __attribute__((aligned(32))); //double size matrix of c + + 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;i1176) || (BG==2 && Zc>64)) { // extend matrix if(impp->tprep != NULL) start_meas(impp->tprep); - for (i1=0; i1 < ncols; i1++) - { - memcpy(&c_extension[2*i1*Zc], &c[i1*Zc], Zc*sizeof(unsigned char)); - memcpy(&c_extension[(2*i1+1)*Zc], &c[i1*Zc], Zc*sizeof(unsigned char)); - } - for (i1=1;i1tprep != NULL) stop_meas(impp->tprep); //parity check part if(impp->tparity != NULL) start_meas(impp->tparity); - encode_parity_check_part_optim(c_extension, d, BG, Zc, Kb); + encode_parity_check_part_optim(c, d, BG, Zc, Kb,simd_size, ncols); if(impp->tparity != NULL) stop_meas(impp->tparity); } else { diff --git a/openair1/PHY/CODING/nrLDPC_encoder/ldpc_encoder_optim8seg.c b/openair1/PHY/CODING/nrLDPC_encoder/ldpc_encoder_optim8seg.c index a974e86e27..460ee80a21 100644 --- a/openair1/PHY/CODING/nrLDPC_encoder/ldpc_encoder_optim8seg.c +++ b/openair1/PHY/CODING/nrLDPC_encoder/ldpc_encoder_optim8seg.c @@ -102,7 +102,6 @@ int nrLDPC_encod(unsigned char **test_input,unsigned char **channel_input,int Zc unsigned char c[22*Zc] __attribute__((aligned(32))); //padded input, unpacked, max size unsigned char d[46*Zc] __attribute__((aligned(32))); //coded parity part output, unpacked, max size - unsigned char c_extension[2*22*Zc*simd_size] __attribute__((aligned(32))); //double size matrix of c // calculate number of punctured bits no_punctured_columns=(int)((nrows-2)*Zc+block_length-block_length*rate)/Zc; @@ -151,24 +150,10 @@ int nrLDPC_encod(unsigned char **test_input,unsigned char **channel_input,int Zc if ((BG==1 && Zc>176) || (BG==2 && Zc>64)) { // extend matrix if(impp->tprep != NULL) start_meas(impp->tprep); - for (i1=0; i1 < ncols; i1++) - { - memcpy(&c_extension[2*i1*Zc], &c[i1*Zc], Zc*sizeof(unsigned char)); - memcpy(&c_extension[(2*i1+1)*Zc], &c[i1*Zc], Zc*sizeof(unsigned char)); - } - for (i1=1;i1tprep != NULL) stop_meas(impp->tprep); //parity check part if(impp->tparity != NULL) start_meas(impp->tparity); - encode_parity_check_part_optim(c_extension, d, BG, Zc, Kb); + encode_parity_check_part_optim(c, d, BG, Zc, Kb, simd_size, ncols); if(impp->tparity != NULL) stop_meas(impp->tparity); } else { diff --git a/openair1/PHY/CODING/nrLDPC_encoder/ldpc_encoder_optim8segmulti.c b/openair1/PHY/CODING/nrLDPC_encoder/ldpc_encoder_optim8segmulti.c index b21f875b32..56a28822f2 100644 --- a/openair1/PHY/CODING/nrLDPC_encoder/ldpc_encoder_optim8segmulti.c +++ b/openair1/PHY/CODING/nrLDPC_encoder/ldpc_encoder_optim8segmulti.c @@ -38,18 +38,19 @@ #include "common/utils/LOG/log.h" #include "PHY/TOOLS/time_meas.h" #include "openair1/PHY/CODING/nrLDPC_defs.h" - -//#define DEBUG_LDPC +#define DEBUG_LDPC #include "ldpc_encode_parity_check.c" #include "ldpc_generate_coefficient.c" -int nrLDPC_encod(unsigned char **test_input,unsigned char **channel_input,int Zc,int Kb,short block_length, short BG, encoder_implemparams_t *impp) +int nrLDPC_encod(unsigned char **input,unsigned char **output,int Zc,int Kb,short block_length, short BG, encoder_implemparams_t *impp) { + //set_log(PHY, 4); + - short nrows=0,ncols=0; - int i,i1,j,rate=3; + int nrows=0,ncols=0; + int rate=3; int no_punctured_columns,removed_bit; //Table of possible lifting sizes char temp; @@ -95,8 +96,8 @@ int nrLDPC_encod(unsigned char **test_input,unsigned char **channel_input,int Zc } #ifdef DEBUG_LDPC - LOG_D(PHY,"ldpc_encoder_optim_8seg: BG %d, Zc %d, Kb %d, block_length %d, segments %d\n",BG,Zc,Kb,block_length,n_segments); - LOG_D(PHY,"ldpc_encoder_optim_8seg: PDU (seg 0) %x %x %x %x\n",test_input[0][0],test_input[0][1],test_input[0][2],test_input[0][3]); + LOG_D(PHY,"ldpc_encoder_optim_8seg: BG %d, Zc %d, Kb %d, block_length %d, segments %d\n",BG,Zc,Kb,block_length,impp->n_segments); + LOG_D(PHY,"ldpc_encoder_optim_8seg: 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); @@ -104,10 +105,8 @@ int nrLDPC_encod(unsigned char **test_input,unsigned char **channel_input,int Zc if ((Zc&31) > 0) simd_size = 16; else simd_size = 32; - unsigned char c[22*Zc] __attribute__((aligned(32))); //padded input, unpacked, max size - unsigned char d[46*Zc] __attribute__((aligned(32))); //coded parity part output, unpacked, max size - - unsigned char c_extension[2*22*Zc*simd_size] __attribute__((aligned(32))); //double size matrix of c + unsigned char cc[22*Zc] __attribute__((aligned(32))); //padded input, unpacked, max size + unsigned char dd[46*Zc] __attribute__((aligned(32))); //coded parity part output, unpacked, max size // calculate number of punctured bits no_punctured_columns=(int)((nrows-2)*Zc+block_length-block_length*rate)/Zc; @@ -115,8 +114,8 @@ int nrLDPC_encod(unsigned char **test_input,unsigned char **channel_input,int Zc //printf("%d\n",no_punctured_columns); //printf("%d\n",removed_bit); // unpack input - memset(c,0,sizeof(unsigned char) * ncols * Zc); - memset(d,0,sizeof(unsigned char) * nrows * Zc); + memset(cc,0,sizeof(unsigned char) * ncols * Zc); + memset(dd,0,sizeof(unsigned char) * nrows * Zc); if(impp->tinput != NULL) start_meas(impp->tinput); #if 0 @@ -124,29 +123,29 @@ int nrLDPC_encod(unsigned char **test_input,unsigned char **channel_input,int Zc //for (j=0; j>(i&7); + temp = (input[j][i/8]&(1<<(i&7)))>>(i&7); //printf("c(%d,%d)=%d\n",j,i,temp); c[i] |= (temp << (j-macro_segment)); } } #else #ifdef __AVX2__ - for (i=0; i>5; i++) { - c256 = _mm256_and_si256(_mm256_cmpeq_epi8(_mm256_andnot_si256(_mm256_shuffle_epi8(_mm256_set1_epi32(((uint32_t*)test_input[macro_segment])[i]), shufmask),andmask),zero256),masks[0]); + for (int i=0; i>5; i++) { + c256 = _mm256_and_si256(_mm256_cmpeq_epi8(_mm256_andnot_si256(_mm256_shuffle_epi8(_mm256_set1_epi32(((uint32_t*)input[macro_segment])[i]), shufmask),andmask),zero256),masks[0]); //for (j=1; j>5)<<5;i>5)<<5;i>(i&7)))>>(7-(i&7)); + temp = (input[j][i/8]&(128>>(i&7)))>>(7-(i&7)); //printf("c(%d,%d)=%d\n",j,i,temp); - c[i] |= (temp << (j-macro_segment)); + cc[i] |= (temp << (j-macro_segment)); } } #else @@ -159,28 +158,14 @@ int nrLDPC_encod(unsigned char **test_input,unsigned char **channel_input,int Zc if ((BG==1 && Zc>176) || (BG==2 && Zc>64)) { // extend matrix if(impp->tprep != NULL) start_meas(impp->tprep); - for (i1=0; i1 < ncols; i1++) - { - memcpy(&c_extension[2*i1*Zc], &c[i1*Zc], Zc*sizeof(unsigned char)); - memcpy(&c_extension[(2*i1+1)*Zc], &c[i1*Zc], Zc*sizeof(unsigned char)); - } - for (i1=1;i1tprep != NULL) stop_meas(impp->tprep); //parity check part if(impp->tparity != NULL) start_meas(impp->tparity); - encode_parity_check_part_optim(c_extension, d, BG, Zc, Kb); + encode_parity_check_part_optim(cc, dd, BG, Zc, Kb, simd_size, ncols); if(impp->tparity != NULL) stop_meas(impp->tparity); } else { - if (encode_parity_check_part_orig(c, d, BG, Zc, Kb, block_length)!=0) { + if (encode_parity_check_part_orig(cc, dd, BG, Zc, Kb, block_length)!=0) { printf("Problem with encoder\n"); return(-1); } @@ -188,8 +173,8 @@ int nrLDPC_encod(unsigned char **test_input,unsigned char **channel_input,int Zc if(impp->toutput != NULL) start_meas(impp->toutput); // information part and puncture columns /* - memcpy(&channel_input[0], &c[2*Zc], (block_length-2*Zc)*sizeof(unsigned char)); - memcpy(&channel_input[block_length-2*Zc], &d[0], ((nrows-no_punctured_columns) * Zc-removed_bit)*sizeof(unsigned char)); + memcpy(&output[0], &c[2*Zc], (block_length-2*Zc)*sizeof(unsigned char)); + memcpy(&output[block_length-2*Zc], &d[0], ((nrows-no_punctured_columns) * Zc-removed_bit)*sizeof(unsigned char)); */ #ifdef __AVX2__ if ((((2*Zc)&31) == 0) && (((block_length-(2*Zc))&31) == 0)) { @@ -197,34 +182,34 @@ int nrLDPC_encod(unsigned char **test_input,unsigned char **channel_input,int Zc //AssertFatal(((block_length-(2*Zc))&31) == 0,"block_length-(2*Zc) needs to be a multiple of 32 for now\n"); uint32_t l1 = (block_length-(2*Zc))>>5; uint32_t l2 = ((nrows-no_punctured_columns) * Zc-removed_bit)>>5; - __m256i *c256p = (__m256i *)&c[2*Zc]; - __m256i *d256p = (__m256i *)&d[0]; + __m256i *c256p = (__m256i *)&cc[2*Zc]; + __m256i *d256p = (__m256i *)&dd[0]; // if (((block_length-(2*Zc))&31)>0) l1++; - for (i=0;i0) l2++; - for (i1=0;i1>(j-macro_segment))&1; - for (i=0;i<((nrows-no_punctured_columns) * Zc-removed_bit);i++) + for (int j=macro_segment; j < macro_segment_end; j++) + output[j][i] = (cc[2*Zc+i]>>(j-macro_segment))&1; + for (int i=0;i<((nrows-no_punctured_columns) * Zc-removed_bit);i++) //for (j=0; j>(j-macro_segment))&1; + for (int j=macro_segment; j < macro_segment_end; j++) + output[j][block_length-2*Zc+i] = (dd[i]>>(j-macro_segment))&1; } #else diff --git a/openair1/PHY/NR_TRANSPORT/nr_dlsch.c b/openair1/PHY/NR_TRANSPORT/nr_dlsch.c index 80e6ce8bc3..abae2f0cc4 100644 --- a/openair1/PHY/NR_TRANSPORT/nr_dlsch.c +++ b/openair1/PHY/NR_TRANSPORT/nr_dlsch.c @@ -180,9 +180,12 @@ uint8_t nr_generate_pdsch(processingData_L1tx_t *msgTx, /// CRC, coding, interleaving and rate matching AssertFatal(harq->pdu!=NULL,"harq->pdu is null\n"); + unsigned char output[rel15->rbSize * NR_SYMBOLS_PER_SLOT * NR_NB_SC_PER_RB * 8 * NR_MAX_NB_LAYERS] __attribute__((aligned(32))); + bzero(output,rel15->rbSize * NR_SYMBOLS_PER_SLOT * NR_NB_SC_PER_RB * 8 * NR_MAX_NB_LAYERS); start_meas(dlsch_encoding_stats); nr_dlsch_encoding(gNB, - harq->pdu, frame, slot, dlsch, frame_parms,tinput,tprep,tparity,toutput, + harq->pdu, frame, slot, dlsch, frame_parms, output, + tinput,tprep,tparity,toutput, dlsch_rate_matching_stats, dlsch_interleaving_stats, dlsch_segmentation_stats); @@ -197,7 +200,7 @@ uint8_t nr_generate_pdsch(processingData_L1tx_t *msgTx, printf("\nEncoded payload:\n"); for (int i=0; i>3; i++) { for (int j=0; j<8; j++) - printf("%d", harq->f[(i<<3)+j]); + printf("%d", output[(i<<3)+j]); printf("\t"); } printf("\n"); @@ -210,7 +213,7 @@ uint8_t nr_generate_pdsch(processingData_L1tx_t *msgTx, for (int q=0; qNrOfCodewords; q++) memset((void*)scrambled_output[q], 0, (encoded_length>>5)*sizeof(uint32_t)); for (int q=0; qNrOfCodewords; q++) - nr_pdsch_codeword_scrambling_optim(harq->f, + nr_pdsch_codeword_scrambling_optim(output, encoded_length, q, rel15->dataScramblingId, diff --git a/openair1/PHY/NR_TRANSPORT/nr_dlsch.h b/openair1/PHY/NR_TRANSPORT/nr_dlsch.h index 4b6ecfe412..64b7ecd025 100644 --- a/openair1/PHY/NR_TRANSPORT/nr_dlsch.h +++ b/openair1/PHY/NR_TRANSPORT/nr_dlsch.h @@ -85,6 +85,7 @@ int nr_dlsch_encoding(PHY_VARS_gNB *gNB, uint8_t slot, NR_gNB_DLSCH_t *dlsch, NR_DL_FRAME_PARMS* frame_parms, + unsigned char * output, time_stats_t *tinput, time_stats_t *tprep, time_stats_t *tparity, diff --git a/openair1/PHY/NR_TRANSPORT/nr_dlsch_coding.c b/openair1/PHY/NR_TRANSPORT/nr_dlsch_coding.c index c84669dba6..5620018b73 100644 --- a/openair1/PHY/NR_TRANSPORT/nr_dlsch_coding.c +++ b/openair1/PHY/NR_TRANSPORT/nr_dlsch_coding.c @@ -50,70 +50,47 @@ //#define DEBUG_DLSCH_FREE 1 -void free_gNB_dlsch(NR_gNB_DLSCH_t **dlschptr, uint16_t N_RB) -{ +void free_gNB_dlsch(NR_gNB_DLSCH_t **dlschptr, uint16_t N_RB) { int r; - NR_gNB_DLSCH_t *dlsch = *dlschptr; - uint16_t a_segments = MAX_NUM_NR_DLSCH_SEGMENTS; //number of segments to be allocated - if (dlsch) { + if (dlsch) { if (N_RB != 273) { a_segments = a_segments*N_RB; a_segments = a_segments/273 +1; } - + #ifdef DEBUG_DLSCH_FREE LOG_D(PHY,"Freeing dlsch %p\n",dlsch); #endif NR_DL_gNB_HARQ_t *harq = &dlsch->harq_process; + if (harq->b) { free16(harq->b, a_segments * 1056); harq->b = NULL; #ifdef DEBUG_DLSCH_FREE LOG_D(PHY, "Freeing harq->b (%p)\n", harq->b); #endif - - if (harq->e) { - free16(harq->e, 14 * N_RB * 12 * 8); - harq->e = NULL; + } + #ifdef DEBUG_DLSCH_FREE - printf("Freeing dlsch process %d e (%p)\n", i, harq->e); + LOG_D(PHY, "Freeing dlsch process %d c (%p)\n", i, harq->c); #endif - } - - if (harq->f) { - free16(harq->f, 14 * N_RB * 12 * 8); - harq->f = NULL; + + for (r = 0; r < a_segments; r++) { #ifdef DEBUG_DLSCH_FREE - printf("Freeing dlsch process %d f (%p)\n", i, harq->f); + LOG_D(PHY, "Freeing dlsch process %d c[%d] (%p)\n", i, r, harq->c[r]); #endif - } - -#ifdef DEBUG_DLSCH_FREE - LOG_D(PHY, "Freeing dlsch process %d c (%p)\n", i, harq->c); -#endif - - for (r = 0; r < a_segments; r++) { -#ifdef DEBUG_DLSCH_FREE - LOG_D(PHY, "Freeing dlsch process %d c[%d] (%p)\n", i, r, harq->c[r]); -#endif - - if (harq->c[r]) { - free16(harq->c[r], 1056); - harq->c[r] = NULL; - } - if (harq->d[r]) { - free16(harq->d[r], 3 * 8448); - harq->d[r] = NULL; - } + + if (harq->c[r]) { + free16(harq->c[r], 1056); + harq->c[r] = NULL; } } + free16(dlsch, sizeof(NR_gNB_DLSCH_t)); + *dlschptr = NULL; } - - free16(dlsch, sizeof(NR_gNB_DLSCH_t)); - *dlschptr = NULL; } NR_gNB_DLSCH_t *new_gNB_dlsch(NR_DL_FRAME_PARMS *frame_parms, @@ -121,8 +98,7 @@ NR_gNB_DLSCH_t *new_gNB_dlsch(NR_DL_FRAME_PARMS *frame_parms, unsigned char Mdlharq, uint32_t Nsoft, uint8_t abstraction_flag, - uint16_t N_RB) -{ + uint16_t N_RB) { unsigned char i,r,aa,layer; int re; uint16_t a_segments = MAX_NUM_NR_DLSCH_SEGMENTS; //number of segments to be allocated @@ -133,10 +109,8 @@ NR_gNB_DLSCH_t *new_gNB_dlsch(NR_DL_FRAME_PARMS *frame_parms, } uint16_t dlsch_bytes = a_segments*1056; // allocated bytes per segment - NR_gNB_DLSCH_t *dlsch = malloc16(sizeof(NR_gNB_DLSCH_t)); AssertFatal(dlsch, "cannot allocate dlsch\n"); - bzero(dlsch,sizeof(NR_gNB_DLSCH_t)); dlsch->Kmimo = Kmimo; dlsch->Mdlharq = Mdlharq; @@ -144,11 +118,12 @@ NR_gNB_DLSCH_t *new_gNB_dlsch(NR_DL_FRAME_PARMS *frame_parms, dlsch->Nsoft = Nsoft; for (layer=0; layerue_spec_bf_weights[layer] = (int32_t**)malloc16(64*sizeof(int32_t*)); + dlsch->ue_spec_bf_weights[layer] = (int32_t **)malloc16(64*sizeof(int32_t *)); for (aa=0; aa<64; aa++) { dlsch->ue_spec_bf_weights[layer][aa] = (int32_t *)malloc16(OFDM_SYMBOL_SIZE_COMPLEX_SAMPLES*sizeof(int32_t)); - for (re=0;reue_spec_bf_weights[layer][aa][re] = 0x00007fff; } } @@ -160,7 +135,8 @@ NR_gNB_DLSCH_t *new_gNB_dlsch(NR_DL_FRAME_PARMS *frame_parms, for (int q=0; qmod_symbs[q] = (int32_t *)malloc16(NR_MAX_PDSCH_ENCODED_LENGTH*sizeof(int32_t)); - dlsch->calib_dl_ch_estimates = (int32_t**)malloc16(64*sizeof(int32_t*)); + dlsch->calib_dl_ch_estimates = (int32_t **)malloc16(64*sizeof(int32_t *)); + for (aa=0; aa<64; aa++) { dlsch->calib_dl_ch_estimates[aa] = (int32_t *)malloc16(OFDM_SYMBOL_SIZE_COMPLEX_SAMPLES*sizeof(int32_t)); } @@ -172,7 +148,6 @@ NR_gNB_DLSCH_t *new_gNB_dlsch(NR_DL_FRAME_PARMS *frame_parms, NR_DL_gNB_HARQ_t *harq = &dlsch->harq_process; bzero(harq, sizeof(NR_DL_gNB_HARQ_t)); - harq->b = malloc16(dlsch_bytes); AssertFatal(harq->b, "cannot allocate memory for harq->b\n"); harq->pdu = malloc16(dlsch_bytes); @@ -188,71 +163,33 @@ NR_gNB_DLSCH_t *new_gNB_dlsch(NR_DL_FRAME_PARMS *frame_parms, // 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); AssertFatal(harq->c[r], "cannot allocate harq->c[%d]\n", r); - harq->d[r] = malloc16(68 * 384); - AssertFatal(harq->d[r], "cannot allocate harq->d[%d]\n", r); // max size for coded output bzero(harq->c[r], 8448); - bzero(harq->d[r], (3 * 8448)); } - harq->e = malloc16(N_RB * NR_SYMBOLS_PER_SLOT * NR_NB_SC_PER_RB * 8 * NR_MAX_NB_LAYERS); - AssertFatal(harq->e, "cannot allocate harq->e\n"); - bzero(harq->e, N_RB * NR_SYMBOLS_PER_SLOT * NR_NB_SC_PER_RB * 8 * NR_MAX_NB_LAYERS); - 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); - return(dlsch); } -void clean_gNB_dlsch(NR_gNB_DLSCH_t *dlsch) -{ - - unsigned char Mdlharq; - unsigned char i,j,r; - +void clean_gNB_dlsch(NR_gNB_DLSCH_t *dlsch) { AssertFatal(dlsch!=NULL,"dlsch is null\n"); - Mdlharq = dlsch->Mdlharq; + unsigned char Mdlharq = dlsch->Mdlharq; dlsch->rnti = 0; dlsch->active = 0; - NR_DL_gNB_HARQ_t *harq=&dlsch->harq_process; - - for (i=0; i<10; i++) { + for (int i=0; i<10; i++) { dlsch->harq_ids[0][i] = Mdlharq; dlsch->harq_ids[1][i] = Mdlharq; } - for (i=0; id[r]) - harq->d[r][j] = NR_NULL; - } } -int nr_dlsch_encoding(PHY_VARS_gNB *gNB, - unsigned char *a, - int frame, - uint8_t slot, - NR_gNB_DLSCH_t *dlsch, - NR_DL_FRAME_PARMS* frame_parms, - time_stats_t *tinput,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) -{ - - unsigned int G; - unsigned int crc=1; - NR_DL_gNB_HARQ_t *harq = &dlsch->harq_process; +void ldpc8blocks( void *p) { + encoder_implemparams_t *impp=(encoder_implemparams_t *) p; + NR_DL_gNB_HARQ_t *harq = (NR_DL_gNB_HARQ_t *)impp->harq; + uint16_t Kr= impp->K; nfapi_nr_dl_tti_pdsch_pdu_rel15_t *rel15 = &harq->pdsch_pdu.pdsch_pdu_rel15; + uint8_t mod_order = rel15->qamModOrder[0]; uint16_t nb_rb = rel15->rbSize; uint8_t nb_symb_sch = rel15->NrOfSymbols; - uint32_t A, Kb, F=0; - uint32_t *Zc = &dlsch->harq_process.Z; - uint8_t mod_order = rel15->qamModOrder[0]; - uint16_t Kr=0,r; - uint32_t r_offset=0; - uint32_t E; - uint8_t Ilbrm = 1; - uint32_t Tbslbrm = 950984; //max tbs + uint16_t length_dmrs = get_num_dmrs(rel15->dlDmrsSymbPos); + uint32_t A = rel15->TBSize[0]<<3; uint8_t nb_re_dmrs; if (rel15->dmrsConfigType==NFAPI_NR_DMRS_TYPE1) @@ -260,24 +197,115 @@ int nr_dlsch_encoding(PHY_VARS_gNB *gNB, else nb_re_dmrs = 4*rel15->numDmrsCdmGrpsNoData; - uint16_t length_dmrs = get_num_dmrs(rel15->dlDmrsSymbPos); - uint16_t R=rel15->targetCodeRate[0]; + unsigned int G = nr_get_G(nb_rb, nb_symb_sch, nb_re_dmrs, length_dmrs,mod_order,rel15->nrOfLayers); + LOG_D(PHY,"dlsch coding A %d Kr %d G %d (nb_rb %d, nb_symb_sch %d, nb_re_dmrs %d, length_dmrs %d, mod_order %d)\n", + A,impp->K,G, nb_rb,nb_symb_sch,nb_re_dmrs,length_dmrs,(int)mod_order); + // nrLDPC_encoder output is in "d" + // let's make this interface happy! + uint8_t tmp[8][68 * 384]__attribute__((aligned(32))); + for (int rr=impp->macro_num*8, i=0; rr < impp->n_segments && rr < (impp->macro_num+1)*8; rr++,i++ ) + impp->d[rr]=tmp[i]; + nrLDPC_encoder(harq->c,impp->d,*impp->Zc, impp->Kb,Kr,impp->BG,impp); + // Compute where to place in output buffer that is concatenation of all segments + uint32_t r_offset=0; + for (int i=0; i < impp->macro_num*8; i++ ) + r_offset+=nr_get_E(G, impp->n_segments, mod_order, rel15->nrOfLayers, i); + for (int rr=impp->macro_num*8; rr < impp->n_segments && rr < (impp->macro_num+1)*8; rr++ ) { + if (impp->F>0) { + // writing into positions d[r][k-2Zc] as in clause 5.3.2 step 2) in 38.212 + memset(&impp->d[rr][Kr-impp->F-2*(*impp->Zc)], impp->F, NR_NULL); + } + +#ifdef DEBUG_DLSCH_CODING + LOG_D(PHY,"rvidx in encoding = %d\n", rel15->rvIndex[0]); +#endif + uint32_t E = nr_get_E(G, impp->n_segments, mod_order, rel15->nrOfLayers, rr); + //#ifdef DEBUG_DLSCH_CODING + LOG_D(NR_PHY,"Rate Matching, Code segment %d/%d (coded bits (G) %u, E %d, Filler bits %d, Filler offset %d mod_order %d, nb_rb %d)...\n", + rr, + impp->n_segments, + G, + E, + impp->F, + Kr-impp->F-2*(*impp->Zc), + mod_order,nb_rb); + // for tbslbrm calculation according to 5.4.2.1 of 38.212 + uint8_t Nl = 4; + + if (rel15->nrOfLayers < Nl) + Nl = rel15->nrOfLayers; + + uint32_t Tbslbrm = nr_compute_tbslbrm(rel15->mcsTable[0],nb_rb,Nl); + uint8_t Ilbrm = 1; + + uint8_t e[E]; + bzero (e, E); + nr_rate_matching_ldpc(Ilbrm, + Tbslbrm, + impp->BG, + *impp->Zc, + impp->d[rr], + e, + impp->n_segments, + impp->F, + Kr-impp->F-2*(*impp->Zc), + rel15->rvIndex[0], + E); +#ifdef DEBUG_DLSCH_CODING + + for (int i =0; i<16; i++) + printf("output ratematching e[%d]= %d r_offset %u\n", i,harq->e[i+r_offset], r_offset); + +#endif + nr_interleaving_ldpc(E, + mod_order, + e, + impp->output+r_offset); +#ifdef DEBUG_DLSCH_CODING + + for (int i =0; i<16; i++) + printf("output interleaving f[%d]= %d r_offset %u\n", i,impp->output[i+r_offset], r_offset); + + if (r==impp->n_segments-1) + write_output("enc_output.m","enc",impp->output,G,1,4); + +#endif + r_offset += E; + } +} + +int nr_dlsch_encoding(PHY_VARS_gNB *gNB, + unsigned char *a, + int frame, + uint8_t slot, + NR_gNB_DLSCH_t *dlsch, + NR_DL_FRAME_PARMS *frame_parms, + unsigned char * output, + time_stats_t *tinput,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) { + encoder_implemparams_t impp; + impp.output=output; + unsigned int crc=1; + NR_DL_gNB_HARQ_t *harq = &dlsch->harq_process; + nfapi_nr_dl_tti_pdsch_pdu_rel15_t *rel15 = &harq->pdsch_pdu.pdsch_pdu_rel15; + impp.Zc = &dlsch->harq_process.Z; float Coderate = 0.0; - uint8_t Nl = 4; - VCD_SIGNAL_DUMPER_DUMP_FUNCTION_BY_NAME(VCD_SIGNAL_DUMPER_FUNCTIONS_gNB_DLSCH_ENCODING, VCD_FUNCTION_IN); + uint32_t A = rel15->TBSize[0]<<3; - A = rel15->TBSize[0]<<3; if ( dlsch->rnti != SI_RNTI ) - trace_NRpdu(DIRECTION_DOWNLINK, a, rel15->TBSize[0], 0, WS_C_RNTI, dlsch->rnti, frame, slot,0, 0); + trace_NRpdu(DIRECTION_DOWNLINK, a, rel15->TBSize[0], 0, WS_C_RNTI, dlsch->rnti, frame, slot,0, 0); NR_gNB_SCH_STATS_t *stats=NULL; int first_free=-1; - for (int i=0;idlsch_stats[i].rnti == 0 && first_free == -1) { first_free = i; stats=&gNB->dlsch_stats[i]; } + if (gNB->dlsch_stats[i].rnti == dlsch->rnti) { stats=&gNB->dlsch_stats[i]; break; @@ -290,179 +318,94 @@ int nr_dlsch_encoding(PHY_VARS_gNB *gNB, stats->current_RI = rel15->nrOfLayers; stats->current_Qm = rel15->qamModOrder[0]; } - G = nr_get_G(nb_rb, nb_symb_sch, nb_re_dmrs, length_dmrs,mod_order,rel15->nrOfLayers); - - LOG_D(PHY,"dlsch coding A %d G %d (nb_rb %d, nb_symb_sch %d, nb_re_dmrs %d, length_dmrs %d, mod_order %d)\n", - A,G, nb_rb,nb_symb_sch,nb_re_dmrs,length_dmrs,(int)mod_order); if (A > 3824) { // Add 24-bit crc (polynomial A) to payload crc = crc24a(a,A)>>8; - a[A>>3] = ((uint8_t*)&crc)[2]; - a[1+(A>>3)] = ((uint8_t*)&crc)[1]; - a[2+(A>>3)] = ((uint8_t*)&crc)[0]; + a[A>>3] = ((uint8_t *)&crc)[2]; + a[1+(A>>3)] = ((uint8_t *)&crc)[1]; + a[2+(A>>3)] = ((uint8_t *)&crc)[0]; //printf("CRC %x (A %d)\n",crc,A); //printf("a0 %d a1 %d a2 %d\n", a[A>>3], a[1+(A>>3)], a[2+(A>>3)]); - harq->B = A+24; // harq->b = a; - AssertFatal((A / 8) + 4 <= MAX_NR_DLSCH_PAYLOAD_BYTES, "A %d is too big (A/8+4 = %d > %d)\n", A, (A / 8) + 4, MAX_NR_DLSCH_PAYLOAD_BYTES); - memcpy(harq->b, a, (A / 8) + 4); // why is this +4 if the CRC is only 3 bytes? - } - else { + } else { // Add 16-bit crc (polynomial A) to payload crc = crc16(a,A)>>16; - a[A>>3] = ((uint8_t*)&crc)[1]; - a[1+(A>>3)] = ((uint8_t*)&crc)[0]; + a[A>>3] = ((uint8_t *)&crc)[1]; + a[1+(A>>3)] = ((uint8_t *)&crc)[0]; //printf("CRC %x (A %d)\n",crc,A); //printf("a0 %d a1 %d \n", a[A>>3], a[1+(A>>3)]); - harq->B = A+16; // harq->b = a; - AssertFatal((A / 8) + 3 <= MAX_NR_DLSCH_PAYLOAD_BYTES, "A %d is too big (A/8+3 = %d > %d)\n", A, (A / 8) + 3, MAX_NR_DLSCH_PAYLOAD_BYTES); - memcpy(harq->b, a, (A / 8) + 3); // using 3 bytes to mimic the case of 24 bit crc } - if (R<1000) - Coderate = (float) R /(float) 1024; + + if (rel15->targetCodeRate[0]<1000) + Coderate = (float)rel15->targetCodeRate[0] /(float) 1024; else // to scale for mcs 20 and 26 in table 5.1.3.1-2 which are decimal and input 2* in nr_tbs_tools - Coderate = (float) R /(float) 2048; + Coderate = (float)rel15->targetCodeRate[0] /(float) 2048; if ((A <=292) || ((A<=3824) && (Coderate <= 0.6667)) || Coderate <= 0.25) - harq->BG = 2; + impp.BG = 2; else - harq->BG = 1; + impp.BG = 1; start_meas(dlsch_segmentation_stats); - Kb = nr_segmentation(harq->b, harq->c, harq->B, &harq->C, &harq->K, Zc, &harq->F, harq->BG); + impp.Kb = nr_segmentation(harq->b, harq->c, harq->B, &impp.n_segments, &impp.K, impp.Zc, &impp.F, impp.BG); stop_meas(dlsch_segmentation_stats); - F = harq->F; - Kr = harq->K; -#ifdef DEBUG_DLSCH_CODING - uint16_t Kr_bytes; - Kr_bytes = Kr>>3; -#endif - - //printf("segment Z %d k %d Kr %d BG %d C %d\n", *Zc,harq->K,Kr,BG,harq->C); - - for (r=0; rC; r++) { + for (int r=0; rd[r][0]; //channel_input[r] = &harq->d[r][0]; #ifdef DEBUG_DLSCH_CODING - LOG_D(PHY,"Encoder: B %d F %d \n",harq->B, harq->F); - LOG_D(PHY,"start ldpc encoder segment %d/%d\n",r,harq->C); + LOG_D(PHY,"Encoder: B %d F %d \n",harq->B, impp->F); + LOG_D(PHY,"start ldpc encoder segment %d/%d\n",r,impp.n_segments); LOG_D(PHY,"input %d %d %d %d %d \n", harq->c[r][0], harq->c[r][1], harq->c[r][2],harq->c[r][3], harq->c[r][4]); - for (int cnt =0 ; cnt < 22*(*Zc)/8; cnt ++){ + + for (int cnt =0 ; cnt < 22*(*Zc)/8; cnt ++) { LOG_D(PHY,"%d ", harq->c[r][cnt]); } - LOG_D(PHY,"\n"); + LOG_D(PHY,"\n"); #endif //ldpc_encoder_orig((unsigned char*)harq->c[r],harq->d[r],*Zc,Kb,Kr,BG,0); //ldpc_encoder_optim((unsigned char*)harq->c[r],(unsigned char*)&harq->d[r][0],*Zc,Kb,Kr,BG,NULL,NULL,NULL,NULL); } - encoder_implemparams_t impp; - impp.n_segments=harq->C; + impp.tprep = tprep; impp.tinput = tinput; impp.tparity = tparity; impp.toutput = toutput; - for(int j=0;j<(harq->C/8+((harq->C&7)==0 ? 0 : 1));j++) { - impp.macro_num=j; - nrLDPC_encoder(harq->c,harq->d,*Zc,Kb,Kr,harq->BG,&impp); + impp.harq=harq; + notifiedFIFO_t nf; + initNotifiedFIFO(&nf); + int nbJobs=0; + for(int j=0; j<(impp.n_segments/8+((impp.n_segments&7)==0 ? 0 : 1)); j++) { + notifiedFIFO_elt_t *req=newNotifiedFIFO_elt(sizeof(impp), j, &nf, ldpc8blocks); + encoder_implemparams_t* perJobImpp=(encoder_implemparams_t*)NotifiedFifoData(req); + *perJobImpp=impp; + perJobImpp->macro_num=j; + pushTpool(gNB->threadPool,req); + nbJobs++; } - -#ifdef DEBUG_DLSCH_CODING - write_output("enc_input0.m","enc_in0",&harq->c[0][0],Kr_bytes,1,4); - write_output("enc_output0.m","enc0",&harq->d[0][0],(3*8*Kr_bytes)+12,1,4); -#endif - - F = harq->F; - - Kr = harq->K; - for (r=0; rC; r++) { - - if (F>0) { - for (int k=(Kr-F-2*(*Zc)); kd[r][k] = NR_NULL; - //if (k<(Kr-F+8)) - //printf("r %d filler bits [%d] = %d \n", r,k, harq->d[r][k]); - } - } - -#ifdef DEBUG_DLSCH_CODING - LOG_D(PHY,"rvidx in encoding = %d\n", rel15->rvIndex[0]); -#endif - - E = nr_get_E(G, harq->C, mod_order, rel15->nrOfLayers, r); - - //#ifdef DEBUG_DLSCH_CODING - LOG_D(NR_PHY,"Rate Matching, Code segment %d/%d (coded bits (G) %u, E %d, Filler bits %d, Filler offset %d mod_order %d, nb_rb %d)...\n", - r, - harq->C, - G, - E, - F, - Kr-F-2*(*Zc), - mod_order,nb_rb); - - // for tbslbrm calculation according to 5.4.2.1 of 38.212 - if (rel15->nrOfLayers < Nl) - Nl = rel15->nrOfLayers; - - Tbslbrm = nr_compute_tbslbrm(rel15->mcsTable[0],nb_rb,Nl); - - start_meas(dlsch_rate_matching_stats); - nr_rate_matching_ldpc(Ilbrm, - Tbslbrm, - harq->BG, - *Zc, - harq->d[r], - harq->e+r_offset, - harq->C, - F, - Kr-F-2*(*Zc), - rel15->rvIndex[0], - E); - stop_meas(dlsch_rate_matching_stats); -#ifdef DEBUG_DLSCH_CODING - for (int i =0; i<16; i++) - printf("output ratematching e[%d]= %d r_offset %u\n", i,harq->e[i+r_offset], r_offset); -#endif - - start_meas(dlsch_interleaving_stats); - nr_interleaving_ldpc(E, - mod_order, - harq->e+r_offset, - harq->f+r_offset); - stop_meas(dlsch_interleaving_stats); - -#ifdef DEBUG_DLSCH_CODING - for (int i =0; i<16; i++) - printf("output interleaving f[%d]= %d r_offset %u\n", i,harq->f[i+r_offset], r_offset); - - if (r==harq->C-1) - write_output("enc_output.m","enc",harq->f,G,1,4); -#endif - - r_offset += E; + while(nbJobs) { + notifiedFIFO_elt_t *req=pullTpool(&nf, gNB->threadPool); + delNotifiedFIFO_elt(req); + nbJobs--; } - VCD_SIGNAL_DUMPER_DUMP_FUNCTION_BY_NAME(VCD_SIGNAL_DUMPER_FUNCTIONS_gNB_DLSCH_ENCODING, VCD_FUNCTION_OUT); - return 0; } diff --git a/openair1/PHY/NR_TRANSPORT/pucch_rx.c b/openair1/PHY/NR_TRANSPORT/pucch_rx.c index 96842426f5..6e246ba854 100644 --- a/openair1/PHY/NR_TRANSPORT/pucch_rx.c +++ b/openair1/PHY/NR_TRANSPORT/pucch_rx.c @@ -368,7 +368,7 @@ void nr_decode_pucch0(PHY_VARS_gNB *gNB, xrtmag_next = temp; } - av_corr/=nr_sequences/l; + //av_corr/=nr_sequences/l; int xrtmag_dBtimes10 = 10*(int)dB_fixed64(xrtmag/(12*l)); int xrtmag_next_dBtimes10 = 10*(int)dB_fixed64(xrtmag_next/(12*l)); diff --git a/openair1/PHY/NR_UE_TRANSPORT/nr_dlsch_demodulation.c b/openair1/PHY/NR_UE_TRANSPORT/nr_dlsch_demodulation.c index 031c08a72d..618e5680e8 100644 --- a/openair1/PHY/NR_UE_TRANSPORT/nr_dlsch_demodulation.c +++ b/openair1/PHY/NR_UE_TRANSPORT/nr_dlsch_demodulation.c @@ -389,7 +389,7 @@ int nr_rx_pdsch(PHY_VARS_NR_UE *ue, 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_I(PHY, "[AbsSFN %u.%d] Slot%d Symbol %d type %d: Pilot/Data extraction %5.2f \n", + 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 @@ -411,7 +411,7 @@ int nr_rx_pdsch(PHY_VARS_NR_UE *ue, #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 - LOG_I(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)); + 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) { @@ -472,7 +472,7 @@ int nr_rx_pdsch(PHY_VARS_NR_UE *ue, #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_I(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)); + 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 start_meas(&ue->generic_stat_bis[proc->thread_id][slot]); @@ -524,7 +524,7 @@ int nr_rx_pdsch(PHY_VARS_NR_UE *ue, #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_I(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)); + 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 start_meas(&ue->generic_stat_bis[proc->thread_id][slot]); @@ -583,7 +583,7 @@ int nr_rx_pdsch(PHY_VARS_NR_UE *ue, #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_I(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)); + 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 start_meas(&ue->generic_stat_bis[proc->thread_id][slot]); /* Store the valid DL RE's */ @@ -666,7 +666,7 @@ int nr_rx_pdsch(PHY_VARS_NR_UE *ue, #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_I(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)); + 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 // Please keep it: useful for debugging diff --git a/openair1/PHY/defs_gNB.h b/openair1/PHY/defs_gNB.h index 5c50b5a249..fb4f9a0092 100644 --- a/openair1/PHY/defs_gNB.h +++ b/openair1/PHY/defs_gNB.h @@ -96,20 +96,6 @@ typedef struct { uint32_t subframe; /// MIMO mode for this DLSCH MIMO_mode_t mimo_mode; - /// Concatenated sequences - uint8_t *e; - /// LDPC-code outputs - uint8_t *d[MAX_NUM_NR_DLSCH_SEGMENTS]; - /// Interleaver outputs - uint8_t *f; - /// Number of code segments - uint32_t C; - /// Number of bits in "small" code segments - uint32_t K; - /// Number of "Filler" bits - uint32_t F; - /// Encoder BG - uint8_t BG; /// LDPC lifting size uint32_t Z; } NR_DL_gNB_HARQ_t; diff --git a/openair1/SCHED_NR_UE/phy_procedures_nr_ue.c b/openair1/SCHED_NR_UE/phy_procedures_nr_ue.c index 93839008a6..34fcec07b0 100644 --- a/openair1/SCHED_NR_UE/phy_procedures_nr_ue.c +++ b/openair1/SCHED_NR_UE/phy_procedures_nr_ue.c @@ -843,11 +843,7 @@ 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]); -#if PHYSIM - printf("[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)); -#else 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)); -#endif if(first_symbol_flag) { proc->first_symbol_available = 1; @@ -1041,17 +1037,10 @@ bool nr_ue_dlsch_procedures(PHY_VARS_NR_UE *ue, stop_meas(&ue->dlsch_decoding_stats[proc->thread_id]); -#if PHYSIM - printf(" --> Unscrambling for CW0 %5.3f\n", - (ue->dlsch_unscrambling_stats.p_time)/(cpuf*1000.0)); - printf("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)); -#else - LOG_I(PHY, " --> Unscrambling for CW0 %5.3f\n", + LOG_D(PHY, " --> Unscrambling for CW0 %5.3f\n", (ue->dlsch_unscrambling_stats.p_time)/(cpuf*1000.0)); - LOG_I(PHY, "AbsSubframe %d.%d --> LDPC Decoding for CW0 %5.3f\n", + 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)); -#endif if(is_cw1_active) { // start ldpc decode for CW 1 @@ -1117,17 +1106,10 @@ bool nr_ue_dlsch_procedures(PHY_VARS_NR_UE *ue, stop_meas(&ue->dlsch_decoding_stats[proc->thread_id]); -#if PHYSIM - printf(" --> Unscrambling for CW1 %5.3f\n", - (ue->dlsch_unscrambling_stats.p_time)/(cpuf*1000.0)); - printf("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)); -#else 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)); -#endif LOG_D(PHY,"AbsSubframe %d.%d --> ldpc Decoding for CW1 %5.3f\n", @@ -1438,11 +1420,7 @@ void *UE_thread_slot1_dl_processing(void *arg) { stop_meas(&ue->ue_front_end_per_slot_stat[proc->thread_id][1]); -#if PHYSIM - printf("[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)); -#else 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)); -#endif //wait until pdcch is decoded @@ -1532,11 +1510,7 @@ 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]); -#if PHYSIM - printf("[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)); -#else 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)); -#endif if (pthread_mutex_lock(&proc->mutex_slot1_dl_processing) != 0) { LOG_E( PHY, "[SCHED][UE] error locking mutex for UE RXTX\n" ); @@ -1906,13 +1880,8 @@ int phy_procedures_nrUE_RX(PHY_VARS_NR_UE *ue, dlsch_parallel); stop_meas(&ue->dlsch_procedures_stat[proc->thread_id]); -#if PHYSIM - printf("[SFN %d] Slot1: Pdsch Proc %5.2f\n",nr_slot_rx,ue->pdsch_procedures_stat[proc->thread_id].p_time/(cpuf*1000.0)); - printf("[SFN %d] Slot0 Slot1: Dlsch Proc %5.2f\n",nr_slot_rx,ue->dlsch_procedures_stat[proc->thread_id].p_time/(cpuf*1000.0)); -#else 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)); -#endif @@ -1989,9 +1958,6 @@ if (nr_slot_rx==9) { } stop_meas(&ue->generic_stat); -#if PHYSIM -printf("after tubo until end of Rx %5.2f \n",ue->generic_stat.p_time/(cpuf*1000.0)); -#endif #ifdef EMOS phy_procedures_emos_UE_RX(ue,slot,gNB_id); @@ -2001,11 +1967,7 @@ 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 PHYSIM -printf("------FULL RX PROC [SFN %d]: %5.2f ------\n",nr_slot_rx,ue->phy_proc_rx[proc->thread_id].p_time/(cpuf*1000.0)); -#else 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 //#endif //pdsch diff --git a/openair1/SIMULATION/NR_PHY/dlschsim.c b/openair1/SIMULATION/NR_PHY/dlschsim.c index a0e0183fd1..96bd435674 100644 --- a/openair1/SIMULATION/NR_PHY/dlschsim.c +++ b/openair1/SIMULATION/NR_PHY/dlschsim.c @@ -503,8 +503,10 @@ int main(int argc, char **argv) //printf("crc32: [0]->0x%08x\n",crc24c(test_input, 32)); // generate signal + unsigned char output[rel15->rbSize * NR_SYMBOLS_PER_SLOT * NR_NB_SC_PER_RB * 8 * NR_MAX_NB_LAYERS] __attribute__((aligned(32))); + bzero(output,rel15->rbSize * NR_SYMBOLS_PER_SLOT * NR_NB_SC_PER_RB * 8 * NR_MAX_NB_LAYERS); if (input_fd == NULL) { - nr_dlsch_encoding(gNB, test_input, frame, slot, dlsch, frame_parms,NULL,NULL,NULL,NULL,NULL,NULL,NULL); + nr_dlsch_encoding(gNB, test_input, frame, slot, dlsch, frame_parms,output,NULL,NULL,NULL,NULL,NULL,NULL,NULL); } for (SNR = snr0; SNR < snr1; SNR += snr_step) { @@ -521,11 +523,12 @@ int main(int argc, char **argv) //if (i<16) // printf("encoder output f[%d] = %d\n",i,dlsch->harq_processes[0]->f[i]); + /* if (dlsch->harq_process.f[i] == 0) modulated_input[i] = 1.0; ///sqrt(2); //QPSK else modulated_input[i] = -1.0; ///sqrt(2); - + */ //if (i<16) printf("modulated_input[%d] = %d\n",i,modulated_input[i]); //SNR =10; SNR_lin = pow(10, SNR / 10.0); @@ -548,9 +551,10 @@ int main(int argc, char **argv) channel_output_uncoded[i] = 1; //QPSK demod else channel_output_uncoded[i] = 0; - + /* if (channel_output_uncoded[i] != dlsch->harq_process.f[i]) errors_bit_uncoded = errors_bit_uncoded + 1; + */ } //if (errors_bit_uncoded>10) diff --git a/openair1/SIMULATION/NR_PHY/dlsim.c b/openair1/SIMULATION/NR_PHY/dlsim.c index 2699413f98..7138e41537 100644 --- a/openair1/SIMULATION/NR_PHY/dlsim.c +++ b/openair1/SIMULATION/NR_PHY/dlsim.c @@ -369,6 +369,7 @@ int main(int argc, char **argv) uint8_t n_tx=1,n_rx=1; uint8_t round; uint8_t num_rounds = 4; + char gNBthreads[128]="n"; channel_desc_t *gNB2UE; //uint32_t nsymb,tx_lev,tx_lev1 = 0,tx_lev2 = 0; @@ -430,9 +431,7 @@ int main(int argc, char **argv) FILE *scg_fd=NULL; - - while ((c = getopt (argc, argv, "f:hA:pf:g:in:s:S:t:x:y:z:M:N:F:GR:dPIL:Ea:b:D:e:m:w:T:U:q")) != -1) { - + while ((c = getopt (argc, argv, "f:hA:pf:g:in:s:S:t:x:y:z:M:N:F:GR:dPIL:Ea:b:d:e:m:w:T:U:qX:")) != -1) { switch (c) { case 'f': scg_fd = fopen(optarg,"r"); @@ -635,7 +634,12 @@ int main(int argc, char **argv) dmrs_arg[i] = atoi(argv[optind++]); } break; - + + case 'X': + strncpy(gNBthreads, optarg, sizeof(gNBthreads)); + gNBthreads[sizeof(gNBthreads)-1]=0; + break; + default: case 'h': 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", @@ -671,7 +675,8 @@ int main(int argc, char **argv) printf("-U Change DMRS Config, arguments list DMRS TYPE{0=A,1=B} DMRS AddPos{0:2} DMRS ConfType{1:2}, e.g. -U 3 0 2 1 \n"); printf("-P Print DLSCH performances\n"); printf("-w Write txdata to binary file (one frame)\n"); - printf("-D number of dlsch threads, 0: no dlsch parallelization\n"); + printf("-d number of dlsch threads, 0: no dlsch parallelization\n"); + printf("-X gNB thread pool configuration, n => no threads"); exit (-1); break; } @@ -823,18 +828,18 @@ int main(int argc, char **argv) fs = 122.88e6; bw = 100e6; } - else if (mu == 1 && N_RB_DL == 162) { - fs = 61.44e6; - bw = 60e6; - } - else if (mu == 1 && N_RB_DL == 133) { - fs = 61.44e6; - bw = 50e6; - } else if (mu == 1 && N_RB_DL == 106) { fs = 61.44e6; bw = 40e6; } + else if (mu == 1 && N_RB_DL == 133) { + fs = 61.44e6; + bw = 50e6; + } + else if (mu == 1 && N_RB_DL == 162) { + fs = 61.44e6; + bw = 60e6; + } else if (mu == 3 && N_RB_DL == 66) { fs = 122.88e6; bw = 100e6; @@ -974,8 +979,7 @@ int main(int argc, char **argv) snrRun = 0; gNB->threadPool = (tpool_t*)malloc(sizeof(tpool_t)); - char tp_param[] = "n"; - initTpool(tp_param, gNB->threadPool, true); + initTpool(gNBthreads, gNB->threadPool, true); gNB->resp_L1_tx = (notifiedFIFO_t*) malloc(sizeof(notifiedFIFO_t)); initNotifiedFIFO(gNB->resp_L1_tx); // we create 2 threads for L1 tx processing @@ -1229,8 +1233,8 @@ int main(int argc, char **argv) UE->dlsch[UE_proc.thread_id][0][0]->max_ldpc_iterations+1) n_errors++; - NR_UE_PDSCH **pdsch_vars = UE->pdsch_vars[UE_proc.thread_id]; - int16_t *UE_llr = pdsch_vars[0]->llr[0]; + //NR_UE_PDSCH **pdsch_vars = UE->pdsch_vars[UE_proc.thread_id]; + //int16_t *UE_llr = pdsch_vars[0]->llr[0]; TBS = UE_harq_process->TBS;//rel15->TBSize[0]; uint16_t length_dmrs = get_num_dmrs(rel15->dlDmrsSymbPos); @@ -1245,6 +1249,7 @@ int main(int argc, char **argv) printf("[DLSIM][PTRS] Available bits are: %5u, removed PTRS bits are: %5u \n",available_bits, (ptrsSymbPerSlot * ptrsRePerSymb *rel15->nrOfLayers* 2) ); } + /* for (i = 0; i < available_bits; i++) { if(((gNB_dlsch->harq_process.f[i] == 0) && (UE_llr[i] <= 0)) || @@ -1258,6 +1263,7 @@ int main(int argc, char **argv) } } + */ for (i = 0; i < TBS; i++) { estimated_output_bit[i] = (UE_harq_process->b[i/8] & (1 << (i & 7))) >> (i & 7); @@ -1301,12 +1307,9 @@ int main(int argc, char **argv) printf("\n"); if (print_perf==1) { - printf("\ngNB TX function statistics (per %d us slot, NPRB %d, mcs %d, TBS %d, C %d, Kr %d (Zc %d))\n", + printf("\ngNB TX function statistics (per %d us slot, NPRB %d, mcs %d, TBS %d)\n", 1000>>*scc->ssbSubcarrierSpacing, g_rbSize, g_mcsIndex, - msgDataTx->dlsch[0][0]->harq_process.pdsch_pdu.pdsch_pdu_rel15.TBSize[0]<<3, - msgDataTx->dlsch[0][0]->harq_process.C, - msgDataTx->dlsch[0][0]->harq_process.K, - msgDataTx->dlsch[0][0]->harq_process.K/((msgDataTx->dlsch[0][0]->harq_process.pdsch_pdu.pdsch_pdu_rel15.TBSize[0]<<3)>3824?22:10)); + msgDataTx->dlsch[0][0]->harq_process.pdsch_pdu.pdsch_pdu_rel15.TBSize[0]<<3); printDistribution(gNB->phy_proc_tx_0,table_tx,"PHY proc tx"); printStatIndent2(&gNB->dlsch_encoding_stats,"DLSCH encoding time"); printStatIndent3(&gNB->dlsch_segmentation_stats,"DLSCH segmentation time"); diff --git a/openair2/LAYER2/NR_MAC_gNB/main.c b/openair2/LAYER2/NR_MAC_gNB/main.c index 6e13e4bcc6..41b33d88db 100644 --- a/openair2/LAYER2/NR_MAC_gNB/main.c +++ b/openair2/LAYER2/NR_MAC_gNB/main.c @@ -64,7 +64,7 @@ void *nrmac_stats_thread(void *arg) { fseek(fd,0,SEEK_SET); } fclose(fd); - return (void *)0; + return NULL; } void clear_mac_stats(gNB_MAC_INST *gNB) {