maki/openairinterface5g
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Merge remote-tracking branch 'origin/develop' into NR_UE_acknack

Browse Source
This commit is contained in:
francescomani
2021-12-02 09:23:27 +01:00
parent facf3aa6ac 0ac69fb164
commit b765b02129
251 changed files with 21317 additions and 16581 deletions
Download Patch File Download Diff File Expand all files Collapse all files
openair1/PHY/CODING/DOC/LDPCImplementation.md
+158 -5
View File
@@ -1,11 +1,13 @@
#LDPC coder/decoder implementation
# LDPC coder/decoder implementation
The LDPC coder and decoder are implemented in a shared library, dynamically loaded at run-time using the [oai shared library loader](file://../../../../common/utils/DOC/loader.md). The code loading the LDPC library is in [nrLDPC_load.c](file://../nrLDPC_load.c), in function `load_nrLDPClib`, which must be called at init time.
## Selecting the LDPC library at run time
By default the function `int load_nrLDPClib(void)` looks for `libldpc.so`, this default behavior can be changed using the oai loader configuration options in the configuration file or from the command line as shown below:
>loading `libldpc_optim8seg.so` instead of `libldpc.so`
#### Examples of ldpc shared lib selection when running nr softmodem's:
loading `libldpc_optim8seg.so` instead of `libldpc.so`:
```
./nr-softmodem -O libconfig:gnb.band78.tm1.106PRB.usrpx300.conf:dbgl5 --loader.ldpc.shlibversion _optim8seg
@@ -18,9 +20,160 @@ By default the function `int load_nrLDPClib(void)` looks for `libldpc.so`, this
........................
```
Today, this mechanism is not available in the `ldpctest` phy simulator which doesn't initialize the [configuration module](file://../../../../common/config/DOC/config.md). loads `libldpc.so` and `libldpc_orig.so` to compare the performance of the two implementations.
loading `libldpc_cl.so` instead of `libldpc.so`:
###LDPC libraries
Libraries implementing the LDPC algorithms must be named `libldpc<_version>.so`, they must implement two functions: `nrLDPC_decod` and `nrLDPC_encod`. The prototypes for these functions is defined in [nrLDPC_defs.h](file://nrLDPC_defs.h).
`make ldpc_cl`
`cp ../../../openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_decoder_kernels_CL.cl`
`./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
------------------------------------------------
[HW] Platform 0, OpenCL profile FULL_PROFILE
[HW] Platform 0, OpenCL version OpenCL 2.1 LINUX
[HW] Device 0 is available
[HW] Device 0, type 2 = 0x00000002: cpu
[HW] Device 0, number of Compute Units: 8
[HW] Device 0, max Work Items dimension: 3
[HW] Device 0, max Work Items size for dimension: 0 8192
[HW] Device 0, max Work Items size for dimension: 1 8192
[HW] Device 0, max Work Items size for dimension: 2 8192
[New Thread 0x7fffcc258700 (LWP 3945123)]
[New Thread 0x7fffc3e57700 (LWP 3945124)]
[New Thread 0x7fffcbe57700 (LWP 3945125)]
[New Thread 0x7fffcba56700 (LWP 3945126)]
[New Thread 0x7fffcb254700 (LWP 3945128)]
[New Thread 0x7fffcb655700 (LWP 3945127)]
[New Thread 0x7fffcae53700 (LWP 3945129)]
[HW] Platform 1, OpenCL profile FULL_PROFILE
[HW] Platform 1, OpenCL version OpenCL 2.0 beignet 1.3
[New Thread 0x7fffc965a700 (LWP 3945130)]
[Thread 0x7fffc965a700 (LWP 3945130) exited]
[HW] Device 0 is available
[HW] Device 0, type 4 = 0x00000004: gpu
[HW] Device 0, number of Compute Units: 20
[HW] Device 0, max Work Items dimension: 3
[HW] Device 0, max Work Items size for dimension: 0 512
[HW] Device 0, max Work Items size for dimension: 1 512
[HW] Device 0, max Work Items size for dimension: 2 512
-----------------------------------------------------------------
```
`./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] loader.ldpc 1 options set from command line
[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
[HW] Device 0, type 2 = 0x00000002: cpu
[HW] Device 0, number of Compute Units: 8
[HW] Device 0, max Work Items dimension: 3
[HW] Device 0, max Work Items size for dimension: 0 8192
[HW] Device 0, max Work Items size for dimension: 1 8192
[HW] Device 0, max Work Items size for dimension: 2 8192
[New Thread 0x7fffecccc700 (LWP 3945413)]
[New Thread 0x7fffec8cb700 (LWP 3945415)]
[New Thread 0x7fffec4ca700 (LWP 3945414)]
[New Thread 0x7fffdf7fd700 (LWP 3945417)]
[New Thread 0x7fffdfbfe700 (LWP 3945418)]
[New Thread 0x7fffdffff700 (LWP 3945416)]
[New Thread 0x7fffd73fc700 (LWP 3945419)]
[HW] Platform 1, OpenCL profile FULL_PROFILE
[HW] Platform 1, OpenCL version OpenCL 2.0 beignet 1.3
[New Thread 0x7fffde105700 (LWP 3945420)]
[Thread 0x7fffde105700 (LWP 3945420) exited]
[HW] Device 0 is available
[HW] Device 0, type 4 = 0x00000004: gpu
[HW] Device 0, number of Compute Units: 20
[HW] Device 0, max Work Items dimension: 3
[HW] Device 0, max Work Items size for dimension: 0 512
[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.
#### Examples of ldpc shared lib selection when running ldpctest:
Loading libldpc_cuda.so, the cuda implementation of the ldpc decoder:
```$ ./ldpctest -v _cuda
Initializing random number generator, seed 0
block length 8448:
n_trials 1:
SNR0 -2.000000:
[CONFIG] get parameters from cmdline , debug flags: 0x00400000
[CONFIG] log_config: 2/3 parameters successfully set
[CONFIG] log_config: 53/53 parameters successfully set
[CONFIG] log_config: 53/53 parameters successfully set
[CONFIG] log_config: 16/16 parameters successfully set
[CONFIG] log_config: 16/16 parameters successfully set
log init done
[CONFIG] loader: 2/2 parameters successfully set
[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
Initializing random number generator, seed 0
block length 8448:
n_trials 1:
SNR0 -2.000000:
[CONFIG] get parameters from cmdline , debug flags: 0x00400000
[CONFIG] log_config: 2/3 parameters successfully set
[CONFIG] log_config: 53/53 parameters successfully set
[CONFIG] log_config: 53/53 parameters successfully set
[CONFIG] log_config: 16/16 parameters successfully set
[CONFIG] log_config: 16/16 parameters successfully set
log init done
[CONFIG] loader: 2/2 parameters successfully set
[CONFIG] loader.ldpc: 1/2 parameters successfully set
[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
[HW] Device 0, type 2 = 0x00000002: cpu
[HW] Device 0, number of Compute Units: 8
[HW] Device 0, max Work Items dimension: 3
[HW] Device 0, max Work Items size for dimension: 0 8192
[HW] Device 0, max Work Items size for dimension: 1 8192
[HW] Device 0, max Work Items size for dimension: 2 8192
[HW] Platform 1, OpenCL profile FULL_PROFILE
[HW] Platform 1, OpenCL version OpenCL 2.0 beignet 1.3
[HW] Device 0 is available
[HW] Device 0, type 4 = 0x00000004: gpu
[HW] Device 0, number of Compute Units: 20
[HW] Device 0, max Work Items dimension: 3
[HW] Device 0, max Work Items size for dimension: 0 512
[HW] Device 0, max Work Items size for dimension: 1 512
[HW] Device 0, max Work Items size for dimension: 2 512
................................
​```
```
### LDPC libraries
Libraries implementing the LDPC algorithms must be named `libldpc<_version>.so`, they must implement three functions: `nrLDPC_initcall` `nrLDPC_decod` and `nrLDPC_encod`. The prototypes for these functions is defined in [nrLDPC_defs.h](file://nrLDPC_defs.h).
`libldpc_cuda.so`has been tested with the `ldpctest` executable, usage from the softmodem's has to be tested.
`libldpc_cl`is under development.
[oai Wikis home](https://gitlab.eurecom.fr/oai/openairinterface5g/wikis/home)
openair1/PHY/CODING/TESTBENCH/ldpctest.c
+19 -32
View File
@@ -101,8 +101,8 @@ int test_ldpc(short No_iteration,
unsigned int *crc_misses,
time_stats_t *time_optim,
time_stats_t *time_decoder,
n_iter_stats_t *dec_iter,
short run_cuda)
n_iter_stats_t *dec_iter
)
{
//clock initiate
//time_stats_t time,time_optim,tinput,tprep,tparity,toutput, time_decoder;
@@ -393,28 +393,13 @@ int test_ldpc(short No_iteration,
decParams.R=code_rate_vec[R_ind];//13;
decParams.numMaxIter=No_iteration;
decParams.outMode = nrLDPC_outMode_BIT;
decParams.block_length=block_length;
//decParams.outMode =nrLDPC_outMode_LLRINT8;
#ifdef CUDA_FLAG
set_compact_BG(Zc,BG);
init_LLR_DMA_for_CUDA(&decParams, (int8_t*)channel_output_fixed[j], (int8_t*)estimated_output[j], block_length);
#endif
nrLDPC_initcall(&decParams, (int8_t*)channel_output_fixed[j], (int8_t*)estimated_output[j]);
for(j=0;j<n_segments;j++) {
start_meas(time_decoder);
#ifdef CUDA_FLAG
if(run_cuda){
n_iter = nrLDPC_decoder_LYC(&decParams, (int8_t*)channel_output_fixed[j], (int8_t*)estimated_output[j], block_length, time_decoder);
}
else{
// decode the sequence
// decoder supports BG2, Z=128 & 256
//esimated_output=ldpc_decoder(channel_output_fixed, block_length, No_iteration, (double)((float)nom_rate/(float)denom_rate));
///nrLDPC_decoder(&decParams, channel_output_fixed, estimated_output, NULL);
n_iter = nrLDPC_decoder(&decParams, (int8_t*)channel_output_fixed[j], (int8_t*)estimated_output[j], p_nrLDPC_procBuf, p_decoder_profiler);
}
#else
n_iter = nrLDPC_decoder(&decParams, (int8_t*)channel_output_fixed[j], (int8_t*)estimated_output[j], p_nrLDPC_procBuf, p_decoder_profiler);
#endif
stop_meas(time_decoder);
stop_meas(time_decoder);
}
//for (i=(Kb+nrows) * Zc-5;i<(Kb+nrows) * Zc;i++)
@@ -514,17 +499,14 @@ int test_ldpc(short No_iteration,
int main(int argc, char *argv[])
{
#ifdef CUDA_FLAG
warmup_for_GPU();
#endif
unsigned int errors, errors_bit, crc_misses;
double errors_bit_uncoded;
short block_length=8448; // decoder supports length: 1201 -> 1280, 2401 -> 2560
char *ldpc_version=NULL; /* version of the ldpc decoder library to use (XXX suffix to use when loading libldpc_XXX.so */
short No_iteration=5;
int n_segments=1;
//double rate=0.333;
short run_cuda = 0;
int nom_rate=1;
int denom_rate=3;
@@ -544,7 +526,7 @@ int main(int argc, char *argv[])
short BG=0,Zc,Kb=0;
while ((c = getopt (argc, argv, "q:r:s:S:l:G:n:d:i:t:u:h")) != -1)
while ((c = getopt (argc, argv, "q:r:s:S:l:G:n:d:i:t:u:hv:")) != -1)
switch (c)
{
case 'q':
@@ -564,7 +546,7 @@ int main(int argc, char *argv[])
break;
case 'G':
run_cuda = atoi(optarg);
ldpc_version="_cuda";
break;
case 'n':
@@ -590,9 +572,11 @@ int main(int argc, char *argv[])
case 'u':
test_uncoded = atoi(optarg);
break;
case 'v':
ldpc_version=strdup(optarg);
break;
case 'h':
default:
default:
printf("CURRENTLY SUPPORTED CODE RATES: \n");
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");
@@ -609,6 +593,7 @@ int main(int argc, char *argv[])
printf("-t SNR simulation step, Default: 0.1\n");
printf("-i Max decoder iterations, Default: 5\n");
printf("-u Set SNR per coded bit, Default: 0\n");
printf("-v XXX Set ldpc shared library version. libldpc_XXX.so will be used \n");
exit(1);
break;
}
@@ -619,7 +604,10 @@ int main(int argc, char *argv[])
printf("SNR0 %f: \n", SNR0);
load_nrLDPClib();
if (ldpc_version != NULL)
load_nrLDPClib(ldpc_version);
else
load_nrLDPClib(NULL);
load_nrLDPClib_ref("_orig", &encoder_orig);
//for (block_length=8;block_length<=MAX_BLOCK_LENGTH;block_length+=8)
@@ -691,8 +679,7 @@ int main(int argc, char *argv[])
&crc_misses,
time_optim,
time_decoder,
dec_iter,
run_cuda);
dec_iter);
printf("SNR %f, BLER %f (%u/%d)\n", SNR, (float)decoded_errors[i]/(float)n_trials, decoded_errors[i], n_trials);
printf("SNR %f, BER %f (%u/%d)\n", SNR, (float)errors_bit/(float)n_trials/(float)block_length/(double)n_segments, decoded_errors[i], n_trials);
openair1/PHY/CODING/defs_NB_IoT.h
-8
View File
@@ -30,14 +30,6 @@
#include <stdint.h> // for uint8/16/32_t
/* check if this ifndef is required for NB-IoT ?!
//#ifndef NO_OPENAIR1
//#include "PHY/defs_NB_IoT.h"
//#else
//#include "PHY/TOOLS/time_meas.h"
//#endif
*/
#define CRC24_A_NB_IoT 0
#define CRC24_B_NB_IoT 1
#define CRC16_NB_IoT 2
openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_decoder.c
+2 -1
View File
@@ -46,7 +46,8 @@
#endif
static inline uint32_t nrLDPC_decoder_core(int8_t* p_llr, int8_t* p_out, t_nrLDPC_procBuf* p_procBuf, uint32_t numLLR, t_nrLDPC_lut* p_lut, t_nrLDPC_dec_params* p_decParams, t_nrLDPC_time_stats* p_profiler);
void nrLDPC_initcall(t_nrLDPC_dec_params* p_decParams, int8_t* p_llr, int8_t* p_out) {
}
int32_t nrLDPC_decod(t_nrLDPC_dec_params* p_decParams, int8_t* p_llr, int8_t* p_out, t_nrLDPC_procBuf* p_procBuf, t_nrLDPC_time_stats* p_profiler)
{
uint32_t numLLR;
openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_decoder_CL.c
+436
View File
@@ -0,0 +1,436 @@
/*
* 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 nrLDPC_decoder_CL.c
* \brief ldpc decoder, openCL implementaion
* \author Francois TABURET
* \date 2021
* \version 1.0
* \note initial implem - translation of cuda version
*/
/* uses HW component id for log messages ( --log_config.hw_log_level <warning| info|debug|trace>) */
#include <stdio.h>
#include <unistd.h>
#include <cuda_runtime.h>
#include <CL/opencl.h>
#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
#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;
#include "../nrLDPC_decoder_LYC/bgs/BG1_compact_in_C.h"
typedef struct{
cl_uint max_CU;
cl_uint max_WID;
size_t *max_WIS;
} ocldev_t;
typedef struct {
cl_kernel cnp_kernel_1st;
cl_kernel cnp_kernel;
cl_kernel vnp_kernel_normal;
cl_kernel pack_decoded_bit;
} oclkernels_t;
typedef struct{
cl_uint num_devices;
cl_device_id devices[MAX_OCLDEV];
ocldev_t ocldev[MAX_OCLDEV];
cl_context context;
cl_program program;
cl_mem dev_h_compact1;
cl_mem dev_h_compact2;
cl_mem dev_const_llr;
cl_mem dev_llr;
cl_mem dev_dt;
cl_mem dev_tmp;
oclkernels_t kernels[MAX_OCLDEV];
cl_command_queue queue[MAX_OCLDEV];
} oclruntime_t;
typedef struct{
oclruntime_t runtime[MAX_OCLRUNTIME];
} ocl_t;
ocl_t ocl;
void set_compact_BG(int Zc,short BG){
cl_uint rt;
int row,col;
if(BG == 1){
row = 46;
col = 68;
}
else{
row = 42;
col = 52;
}
int compact_row = 30;
int compact_col = 19;
if(BG==2){compact_row = 10, compact_col = 23;}
int memorySize_h_compact1 = row * compact_col * sizeof(h_element);
int memorySize_h_compact2 = compact_row * col * sizeof(h_element);
int lift_index = 0;
short lift_set[][9] = {
{2,4,8,16,32,64,128,256},
{3,6,12,24,48,96,192,384},
{5,10,20,40,80,160,320},
{7,14,28,56,112,224},
{9,18,36,72,144,288},
{11,22,44,88,176,352},
{13,26,52,104,208},
{15,30,60,120,240},
{0}
};
for(int i = 0; lift_set[i][0] != 0; i++){
for(int j = 0; lift_set[i][j] != 0; j++){
if(Zc == lift_set[i][j]){
lift_index = i;
break;
}
}
}
printf("\nZc = %d BG = %d\n",Zc,BG);
ocl.runtime[0].dev_h_compact1 = clCreateBuffer(ocl.runtime[0].context, CL_MEM_READ_ONLY|CL_MEM_HOST_WRITE_ONLY, memorySize_h_compact1, NULL, (cl_int *)&rt);
AssertFatal(rt == CL_SUCCESS, "Error %d creating buffer dev_h_compact1 for platform %i \n" , (int)rt, 0);
ocl.runtime[0].dev_h_compact2 = clCreateBuffer(ocl.runtime[0].context, CL_MEM_READ_ONLY|CL_MEM_HOST_WRITE_ONLY, memorySize_h_compact2, NULL, (cl_int *)&rt);
AssertFatal(rt == CL_SUCCESS, "Error %d creating buffer dev_h_compact2 for platform %i \n" , (int)rt, 0);
h_element *h1;
h_element *h2;
switch(lift_index){
case 0:
h1 = host_h_compact1_I0;
h2 = host_h_compact2_I0;
break;
case 1:
h1 = host_h_compact1_I1;
h2 = host_h_compact2_I1;
break;
case 2:
h1 = host_h_compact1_I2;
h2 = host_h_compact2_I2;
break;
case 3:
h1 = host_h_compact1_I3;
h2 = host_h_compact2_I3;
break;
case 4:
h1 = host_h_compact1_I4;
h2 = host_h_compact2_I4;
break;
case 5:
h1 = host_h_compact1_I5;
h2 = host_h_compact2_I5;
break;
case 6:
h1 = host_h_compact1_I6;
h2 = host_h_compact2_I6;
break;
case 7:
h1 = host_h_compact1_I7;
h2 = host_h_compact2_I7;
break;
default:
AssertFatal(0, "Invalid lift_index value %i\n" , lift_index);
break;
}
rt = clEnqueueWriteBuffer(ocl.runtime[0].queue[0], ocl.runtime[0].dev_h_compact1, CL_TRUE, 0,memorySize_h_compact1, h1, 0, NULL, NULL);
AssertFatal(rt == CL_SUCCESS, "Error %d moving h_compact1 memory to pltf %i dev %i\n" , (int)rt, 0,0);
rt = clEnqueueWriteBuffer(ocl.runtime[0].queue[0], ocl.runtime[0].dev_h_compact2, CL_TRUE, 0,memorySize_h_compact2, h2, 0, NULL, NULL);
AssertFatal(rt == CL_SUCCESS, "Error %d moving h_compact2 memory to pltf %i dev %i\n" , (int)rt, 0,0);
// return 0;
}
void cl_error_callback(const char* errinfo, const void* private_info, size_t cb, void* user_data) {
oclruntime_t *runtime = (oclruntime_t *)user_data;
LOG_E(HW,"OpenCL accelerator error %s\n", errinfo );
}
char *clutil_getstrdev(int intdev) {
static char retstring[255]="";
char *retptr=retstring;
retptr+=sprintf(retptr,"0x%08x: ",(uint32_t)intdev);
if (intdev & CL_DEVICE_TYPE_CPU)
retptr+=sprintf(retptr,"%s","cpu ");
if (intdev & CL_DEVICE_TYPE_GPU)
retptr+=sprintf(retptr,"%s","gpu ");
if (intdev & CL_DEVICE_TYPE_ACCELERATOR)
retptr+=sprintf(retptr,"%s","acc ");
return retstring;
}
void get_CompilErr(cl_program program, int pltf) {
// Determine the size of the log
size_t log_size;
for(int i=0; i<ocl.runtime[pltf].num_devices;i++) {
clGetProgramBuildInfo(program, ocl.runtime[pltf].devices[i], CL_PROGRAM_BUILD_LOG, 0, NULL, &log_size);
// Allocate memory for the log
char *log = (char *) malloc(log_size);
// Get the log
clGetProgramBuildInfo(program, ocl.runtime[pltf].devices[i], CL_PROGRAM_BUILD_LOG, log_size, log, NULL);
// Print the log
printf("%s\n", log);
free(log);
}
}
size_t load_source(char **source_str) {
int MAX_SOURCE_SIZE=(500*132);
FILE *fp;
size_t source_size;
fp = fopen("nrLDPC_decoder_kernels_CL.cl", "r");
AssertFatal(fp,"failed to open cl source: %s\n",strerror(errno));
*source_str = (char*)malloc(MAX_SOURCE_SIZE);
source_size = fread( *source_str, 1, MAX_SOURCE_SIZE, fp);
fclose( fp );
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];
cl_uint num_platforms_found;
int context_ok=0;
cl_uint rt = clGetPlatformIDs( sizeof(platforms)/sizeof(cl_platform_id), platforms, &num_platforms_found );
AssertFatal(rt == CL_SUCCESS, "clGetPlatformIDs error %d\n" , (int)rt);
AssertFatal( num_platforms_found>0 , "clGetPlatformIDs: no cl compatible platform found\n");
for (int i=0 ; i<(int)num_platforms_found ; i++) {
char stringval[255];
rt = clGetPlatformInfo(platforms[i],CL_PLATFORM_PROFILE, sizeof(stringval),stringval,NULL);
AssertFatal(rt == CL_SUCCESS, "clGetPlatformInfo PROFILE error %d\n" , (int)rt);
LOG_I(HW,"Platform %i, OpenCL profile %s\n", i,stringval );
rt = clGetPlatformInfo(platforms[i],CL_PLATFORM_VERSION, sizeof(stringval),stringval,NULL);
AssertFatal(rt == CL_SUCCESS, "clGetPlatformInfo VERSION error %d\n" , (int)rt);
LOG_I(HW,"Platform %i, OpenCL version %s\n", i,stringval );
rt = clGetDeviceIDs(platforms[i],CL_DEVICE_TYPE_ALL, sizeof(ocl.runtime[i].devices)/sizeof(cl_device_id),ocl.runtime[i].devices,&(ocl.runtime[i].num_devices));
AssertFatal(rt == CL_SUCCESS, "clGetDeviceIDs error %d\n" , (int)rt);
int devok=0;
for (int j=0; j<ocl.runtime[i].num_devices; j++) {
cl_bool abool;
rt = clGetDeviceInfo(ocl.runtime[i].devices[j],CL_DEVICE_AVAILABLE, sizeof(abool),&abool,NULL);
AssertFatal(rt == CL_SUCCESS, "clGetDeviceInfo DEVICE_AVAILABLE error %d\n" , (int)rt);
LOG_I(HW,"Device %i is %s available\n", j, (abool==CL_TRUE?"":"not"));
cl_device_type devtype;
rt = clGetDeviceInfo(ocl.runtime[i].devices[j],CL_DEVICE_TYPE, sizeof(cl_device_type),&devtype,NULL);
AssertFatal(rt == CL_SUCCESS, "clGetDeviceInfo DEVICE_TYPE error %d\n" , (int)rt);
LOG_I(HW,"Device %i, type %d = %s\n", j,(int)devtype, clutil_getstrdev(devtype));
rt = clGetDeviceInfo(ocl.runtime[i].devices[j],CL_DEVICE_MAX_COMPUTE_UNITS,sizeof(ocl.runtime[i].ocldev[j].max_CU),&(ocl.runtime[i].ocldev[j].max_CU),NULL);
AssertFatal(rt == CL_SUCCESS, "clGetDeviceInfo MAX_COMPUTE_UNITS error %d\n" , (int)rt);
LOG_I(HW,"Device %i, number of Compute Units: %d\n", j,ocl.runtime[i].ocldev[j].max_CU);
rt = clGetDeviceInfo(ocl.runtime[i].devices[j],CL_DEVICE_MAX_WORK_ITEM_DIMENSIONS,sizeof(ocl.runtime[i].ocldev[j].max_WID),&(ocl.runtime[i].ocldev[j].max_WID),NULL);
AssertFatal(rt == CL_SUCCESS, "clGetDeviceInfo MAX_WORK_ITEM_DIMENSIONS error %d\n" , (int)rt);
LOG_I(HW,"Device %i, max Work Items dimension: %d\n", j,ocl.runtime[i].ocldev[j].max_WID);
ocl.runtime[i].ocldev[j].max_WIS = (size_t *)malloc(ocl.runtime[i].ocldev[j].max_WID * sizeof(size_t));
rt = clGetDeviceInfo(ocl.runtime[i].devices[j],CL_DEVICE_MAX_WORK_ITEM_SIZES,sizeof(ocl.runtime[i].ocldev[j].max_WID)*sizeof(size_t),ocl.runtime[i].ocldev[j].max_WIS,NULL);
AssertFatal(rt == CL_SUCCESS, "clGetDeviceInfo MAX_WORK_ITEM_SIZES error %d\n" , (int)rt);
for(int k=0; k<ocl.runtime[i].ocldev[j].max_WID;k++)
LOG_I(HW,"Device %i, max Work Items size for dimension: %d %u\n", j,k,(uint32_t)ocl.runtime[i].ocldev[j].max_WIS[k]);
devok++;
}
if (devok >0) {
ocl.runtime[i].context = clCreateContext(NULL, ocl.runtime[i].num_devices, ocl.runtime[i].devices, cl_error_callback, &(ocl.runtime[i]), (cl_int *)&rt);
AssertFatal(rt == CL_SUCCESS, "Error %d creating context for platform %i\n" , (int)rt, i);
for(int dev=0; dev<ocl.runtime[i].num_devices; dev++) {
ocl.runtime[i].queue[dev] = clCreateCommandQueueWithProperties(ocl.runtime[i].context,ocl.runtime[i].devices[dev] , 0, (cl_int *)&rt);
AssertFatal(rt == CL_SUCCESS, "Error %d creating command queue for platform %i device %i\n" , (int)rt, i,dev);
}
ocl.runtime[i].dev_const_llr = 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_const_llr for platform %i \n" , (int)rt, i);
ocl.runtime[i].dev_llr = clCreateBuffer(ocl.runtime[i].context, CL_MEM_READ_WRITE|CL_MEM_HOST_WRITE_ONLY, 68*384, NULL, (cl_int *)&rt);
AssertFatal(rt == CL_SUCCESS, "Error %d creating buffer dev_llr for platform %i \n" , (int)rt, i);
ocl.runtime[i].dev_dt = clCreateBuffer(ocl.runtime[i].context, CL_MEM_READ_WRITE|CL_MEM_HOST_NO_ACCESS, 46*68*384, NULL, (cl_int *)&rt);
AssertFatal(rt == CL_SUCCESS, "Error %d creating buffer dev_dt for platform %i \n" , (int)rt, i);
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);
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);
if (rt == CL_BUILD_PROGRAM_FAILURE) {
get_CompilErr(program,i);
}
AssertFatal(rt == CL_SUCCESS, "Error %d buildding program for platform %i \n" , rt, i);
for(int dev=0; dev<ocl.runtime[i].num_devices; dev++) {
ocl.runtime[i].kernels[dev].cnp_kernel_1st = clCreateKernel(program, "ldpc_cnp_kernel_1st_iter", (cl_int *)&rt);
AssertFatal(rt == CL_SUCCESS, "Error %d creating kernel %s platform %i, dev %i\n" , (int)rt,"ldpc_cnp_kernel_1st_iter", i,dev);
ocl.runtime[i].kernels[dev].cnp_kernel = clCreateKernel(program, "ldpc_cnp_kernel", (cl_int *)&rt);
AssertFatal(rt == CL_SUCCESS, "Error %d creating kernel %s platform %i, dev %i\n" , (int)rt,"ldpc_cnp_kernel", i,dev);
ocl.runtime[i].kernels[dev].vnp_kernel_normal = clCreateKernel(program, "ldpc_vnp_kernel_normal", (cl_int *)&rt);
AssertFatal(rt == CL_SUCCESS, "Error %d creating kernel %s platform %i, dev %i\n" , (int)rt,"ldpc_vnp_kernel_normal", i,dev);
ocl.runtime[i].kernels[dev].pack_decoded_bit = clCreateKernel(program, "pack_decoded_bit", (cl_int *)&rt);
AssertFatal(rt == CL_SUCCESS, "Error %d creating kernel %s platform %i, dev %i\n" , (int)rt,"pack_decoded_bit", i,dev);
}
context_ok++;
}
devok=0;
}
AssertFatal(context_ok>0, "No openCL device available to accelerate ldpc\n");
return 0;
}
void nrLDPC_initcall(t_nrLDPC_dec_params* p_decParams, int8_t* p_llr, int8_t* p_out) {
set_compact_BG(p_decParams->Z,p_decParams->BG);
// init_LLR_DMA(p_decParams, p_llr, p_out);
}
int32_t nrLDPC_decod(t_nrLDPC_dec_params* p_decParams, int8_t* p_llr, int8_t* p_out,t_nrLDPC_procBuf* p_procBuf, t_nrLDPC_time_stats *time_decoder)
{
uint16_t Zc = p_decParams->Z;
uint8_t BG = p_decParams->BG;
// uint8_t numMaxIter = p_decParams->numMaxIter;
int block_length = p_decParams->block_length;
// e_nrLDPC_outMode outMode = p_decParams->outMode;
uint8_t row,col;
if(BG == 1){
row = 46;
col = 68;
}
else{
row = 42;
col = 52;
}
// alloc memory
// unsigned char *hard_decision = (unsigned char*)p_out;
// gpu
int memorySize_llr = col * Zc * sizeof(char) * MC;
// cudaCheck( cudaMemcpyToSymbol(dev_const_llr, p_llr, memorySize_llr_cuda) );
// cudaCheck( cudaMemcpyToSymbol(dev_llr, p_llr, memorySize_llr_cuda) );
int rt = clEnqueueWriteBuffer(ocl.runtime[0].queue[0], ocl.runtime[0].dev_const_llr, CL_TRUE, 0,
memorySize_llr, p_llr, 0, NULL, NULL);
AssertFatal(rt == CL_SUCCESS, "Error %d moving p_llr data to read only memory in pltf %i dev %i\n" , (int)rt, 0,0);
rt = clEnqueueWriteBuffer(ocl.runtime[0].queue[0], ocl.runtime[0].dev_llr, CL_TRUE, 0,
memorySize_llr, p_llr, 0, NULL, NULL);
AssertFatal(rt == CL_SUCCESS, "Error %d moving p_llr data to read-write memory in pltf %i dev %i\n" , (int)rt, 0,0);
// Define CUDA kernel dimension
// int blockSizeX = Zc;
// dim3 dimGridKernel1(row, MC, 1); // dim of the thread blocks
// dim3 dimBlockKernel1(blockSizeX, 1, 1);
// dim3 dimGridKernel2(col, MC, 1);
// dim3 dimBlockKernel2(blockSizeX, 1, 1);
// cudaDeviceSynchronize();
// lauch kernel
size_t global_item_sizek0[2] = {row*Zc,MC}; // Process the entire lists
size_t global_item_sizek1[2] = {col*Zc,MC}; // Process the entire lists
size_t local_item_sizek[2] = {128,1}; // Divide work items into groups of 128
for(int ii = 0; ii < MAX_ITERATION; ii++){
// first kernel
if(ii == 0){
// ldpc_cnp_kernel_1st_iter
// <<<dimGridKernel1, dimBlockKernel1>>>
// ( BG, row, col, Zc);
CLSETKERNELARG(ocl.runtime[0].kernels[0].cnp_kernel_1st, 0, sizeof(cl_mem), (void *)&(ocl.runtime[0].dev_llr));
CLSETKERNELARG(ocl.runtime[0].kernels[0].cnp_kernel_1st, 1, sizeof(cl_mem), (void *)&(ocl.runtime[0].dev_dt));
CLSETKERNELARG(ocl.runtime[0].kernels[0].cnp_kernel_1st, 2, sizeof(cl_mem), (void *)&(ocl.runtime[0].dev_h_compact1));
CLSETKERNELARG(ocl.runtime[0].kernels[0].cnp_kernel_1st, 3, sizeof(int), (void *)&(BG));
CLSETKERNELARG(ocl.runtime[0].kernels[0].cnp_kernel_1st, 4, sizeof(int), (void *)&(row));
CLSETKERNELARG(ocl.runtime[0].kernels[0].cnp_kernel_1st, 5, sizeof(int), (void *)&(col));
CLSETKERNELARG(ocl.runtime[0].kernels[0].cnp_kernel_1st, 6, sizeof(int), (void *)&(Zc));
rt = clEnqueueNDRangeKernel(ocl.runtime[0].queue[0], ocl.runtime[0].kernels[0].cnp_kernel_1st, 2, NULL,
global_item_sizek0, local_item_sizek, 0, NULL, NULL);
AssertFatal(rt == CL_SUCCESS, "Error %d enqueing cnp_kernel_1st \n" , (int)rt);
}else{
// ldpc_cnp_kernel
// <<<dimGridKernel1, dimBlockKernel1>>>
// ( BG, row, col, Zc);
CLSETKERNELARG(ocl.runtime[0].kernels[0].cnp_kernel, 0, sizeof(cl_mem), (void *)&(ocl.runtime[0].dev_llr));
CLSETKERNELARG(ocl.runtime[0].kernels[0].cnp_kernel, 1, sizeof(cl_mem), (void *)&(ocl.runtime[0].dev_dt));
CLSETKERNELARG(ocl.runtime[0].kernels[0].cnp_kernel, 2, sizeof(cl_mem), (void *)&(ocl.runtime[0].dev_h_compact1));
CLSETKERNELARG(ocl.runtime[0].kernels[0].cnp_kernel, 3, sizeof(int), (void *)&(BG));
CLSETKERNELARG(ocl.runtime[0].kernels[0].cnp_kernel, 4, sizeof(int), (void *)&(row));
CLSETKERNELARG(ocl.runtime[0].kernels[0].cnp_kernel, 5, sizeof(int), (void *)&(col));
CLSETKERNELARG(ocl.runtime[0].kernels[0].cnp_kernel, 6, sizeof(int), (void *)&(Zc));
rt = clEnqueueNDRangeKernel(ocl.runtime[0].queue[0], ocl.runtime[0].kernels[0].cnp_kernel, 2, NULL,
global_item_sizek0, local_item_sizek, 0, NULL, NULL);
AssertFatal(rt == CL_SUCCESS, "Error %d enqueing cnp_kernel\n" , (int)rt);
}
// second kernel
// ldpc_vnp_kernel_normal
// <<<dimGridKernel2, dimBlockKernel2>>>
// // (dev_llr, dev_const_llr,BG, row, col, Zc);
// (BG, row, col, Zc);
CLSETKERNELARG(ocl.runtime[0].kernels[0].vnp_kernel_normal, 0, sizeof(cl_mem), (void *)&(ocl.runtime[0].dev_llr));
CLSETKERNELARG(ocl.runtime[0].kernels[0].vnp_kernel_normal, 1, sizeof(cl_mem), (void *)&(ocl.runtime[0].dev_dt));
CLSETKERNELARG(ocl.runtime[0].kernels[0].vnp_kernel_normal, 2, sizeof(cl_mem), (void *)&(ocl.runtime[0].dev_const_llr));
CLSETKERNELARG(ocl.runtime[0].kernels[0].vnp_kernel_normal, 3, sizeof(cl_mem), (void *)&(ocl.runtime[0].dev_h_compact2));
CLSETKERNELARG(ocl.runtime[0].kernels[0].vnp_kernel_normal, 4, sizeof(int), (void *)&(BG));
CLSETKERNELARG(ocl.runtime[0].kernels[0].vnp_kernel_normal, 5, sizeof(int), (void *)&(row));
CLSETKERNELARG(ocl.runtime[0].kernels[0].vnp_kernel_normal, 6, sizeof(int), (void *)&(col));
CLSETKERNELARG(ocl.runtime[0].kernels[0].vnp_kernel_normal, 7, sizeof(int), (void *)&(Zc));
rt = clEnqueueNDRangeKernel(ocl.runtime[0].queue[0], ocl.runtime[0].kernels[0].vnp_kernel_normal, 2, NULL,
global_item_sizek1, local_item_sizek, 0, NULL, NULL);
AssertFatal(rt == CL_SUCCESS, "Error %d enqueing vnp_kernel_normal \n" , (int)rt);
}
// int pack = (block_length/128)+1;
// dim3 pack_block(pack, MC, 1);
// pack_decoded_bit<<<pack_block,128>>>( col, Zc);
CLSETKERNELARG(ocl.runtime[0].kernels[0].pack_decoded_bit, 0, sizeof(cl_mem), (void *)&(ocl.runtime[0].dev_llr));
CLSETKERNELARG(ocl.runtime[0].kernels[0].pack_decoded_bit, 1, sizeof(cl_mem), (void *)&(ocl.runtime[0].dev_tmp));
CLSETKERNELARG(ocl.runtime[0].kernels[0].pack_decoded_bit, 2, sizeof(int), (void *)&(col));
CLSETKERNELARG(ocl.runtime[0].kernels[0].pack_decoded_bit, 3, sizeof(int), (void *)&(Zc));
// Execute the OpenCL kernel on the list
size_t global_item_size[2] = {block_length,MC}; // Process the entire lists
size_t local_item_size[2] = {128,1}; // Divide work items into groups of 128
rt = clEnqueueNDRangeKernel(ocl.runtime[0].queue[0], ocl.runtime[0].kernels[0].pack_decoded_bit, 2, NULL,
global_item_size, local_item_size, 0, NULL, NULL);
AssertFatal(rt == CL_SUCCESS, "Error %d enqueing pack_decoded_bit \n" , (int)rt);
// cudaCheck( cudaMemcpyFromSymbol((void*)hard_decision, (const void*)dev_tmp, (block_length/8)*sizeof(unsigned char)) );
// cudaDeviceSynchronize();
rt = clEnqueueReadBuffer(ocl.runtime[0].queue[0], ocl.runtime[0].dev_tmp, CL_TRUE, 0,
(block_length/8)*sizeof(unsigned char) , p_llr, 0, NULL, NULL);
AssertFatal(rt == CL_SUCCESS, "Error %d moving p_llr data to pltf %i dev %i\n" , (int)rt, 0,0);
return MAX_ITERATION;
}
openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_decoder_kernels_CL.cl
+302
View File
@@ -0,0 +1,302 @@
/*
* 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];
}
}
}
openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_init_mem.h
-2
View File
@@ -34,7 +34,6 @@
#include <stdlib.h>
#include "nrLDPC_types.h"
#ifndef malloc32_clear
/**
\brief Allocates 32 byte aligned memory and initializes to zero
\param size Input size in bytes
@@ -46,7 +45,6 @@ static inline void* malloc32_clear(size_t size)
memset(ptr, 0, size);
return ptr;
}
#endif
/**
\brief Allocates and initializes the internal decoder processing buffers
openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_types.h
+1
View File
@@ -70,6 +70,7 @@ typedef struct nrLDPC_dec_params {
uint16_t Z; /**< Lifting size */
uint8_t R; /**< Decoding rate: Format 15,13,... for code rates 1/5, 1/3,... */
uint8_t numMaxIter; /**< Maximum number of iterations */
int block_length;
e_nrLDPC_outMode outMode; /**< Output format */
} t_nrLDPC_dec_params;
openair1/PHY/CODING/nrLDPC_decoder_LYC/nrLDPC_decoder_LYC.cu
+53 -6
View File
@@ -7,14 +7,14 @@
* \note
* \warning
*/
#include <iostream>
#include <stdio.h>
#include <unistd.h>
#include <cuda_runtime.h>
#include <cuda.h>
#include "PHY/CODING/nrLDPC_decoder/nrLDPC_types.h"
#include "PHY/CODING/nrLDPC_decoder/nrLDPCdecoder_defs.h"
#include "assertions.h"
#include "bgs/BG1_I0"
#include "bgs/BG1_I1"
#include "bgs/BG1_I2"
@@ -462,10 +462,11 @@ void read_BG(int BG, int *h, int row, int col)
}
extern "C"
void init_LLR_DMA_for_CUDA(t_nrLDPC_dec_params* p_decParams, int8_t* p_llr, int8_t* p_out, int block_length){
void init_LLR_DMA(t_nrLDPC_dec_params* p_decParams, int8_t* p_llr, int8_t* p_out){
uint16_t Zc = p_decParams->Z;
uint8_t BG = p_decParams->BG;
int block_length = p_decParams->block_length;
uint8_t row,col;
if(BG == 1){
row = 46;
@@ -483,14 +484,60 @@ void init_LLR_DMA_for_CUDA(t_nrLDPC_dec_params* p_decParams, int8_t* p_llr, int8
}
using namespace std ;
/* from here: entry points in decoder shared lib */
extern "C"
int32_t nrLDPC_decoder_LYC(t_nrLDPC_dec_params* p_decParams, int8_t* p_llr, int8_t* p_out, int block_length, time_stats_t *time_decoder)
int ldpc_autoinit(void) { // called by the library loader
/*int devices = 0;
cudaError_t err = cudaGetDeviceCount(&devices);
AssertFatal(devices>0,"\nNo cuda GPU found\n\n");
const int kb = 1024;
const int mb = kb * kb;
wcout << "NBody.GPU" << endl << "=========" << endl << endl;
wcout << "CUDA version: v" << CUDART_VERSION << endl;
wcout << "CUDA Devices: " << endl << endl;
for(int i = 0; i < devices; ++i)
{
cudaDeviceProp props;
cudaGetDeviceProperties(&props, i);
wcout << i << ": " << props.name << ": " << props.major << "." << props.minor << endl;
wcout << " Global memory: " << props.totalGlobalMem / mb << "mb" << endl;
wcout << " Shared memory: " << props.sharedMemPerBlock / kb << "kb" << endl;
wcout << " Constant memory: " << props.totalConstMem / kb << "kb" << endl;
wcout << " Block registers: " << props.regsPerBlock << endl << endl;
wcout << " Warp size: " << props.warpSize << endl;
wcout << " Threads per block: " << props.maxThreadsPerBlock << endl;
wcout << " Max block dimensions: [ " << props.maxThreadsDim[0] << ", " << props.maxThreadsDim[1] << ", " << props.maxThreadsDim[2] << " ]" << endl;
wcout << " Max grid dimensions: [ " << props.maxGridSize[0] << ", " << props.maxGridSize[1] << ", " << props.maxGridSize[2] << " ]" << endl;
wcout << endl;
}
*/
warmup_for_GPU();
return 0;
}
extern "C"
void nrLDPC_initcall(t_nrLDPC_dec_params* p_decParams, int8_t* p_llr, int8_t* p_out) {
set_compact_BG(p_decParams->Z,p_decParams->BG);
init_LLR_DMA(p_decParams, p_llr, p_out);
}
extern "C"
int32_t nrLDPC_decod(t_nrLDPC_dec_params* p_decParams, int8_t* p_llr, int8_t* p_out,t_nrLDPC_procBuf* p_procBuf, t_nrLDPC_time_stats *time_decoder)
{
uint16_t Zc = p_decParams->Z;
uint8_t BG = p_decParams->BG;
uint8_t numMaxIter = p_decParams->numMaxIter;
int block_length = p_decParams->block_length;
e_nrLDPC_outMode outMode = p_decParams->outMode;
cudaError_t cudaStatus;
uint8_t row,col;
openair1/PHY/CODING/nrLDPC_defs.h
+2 -1
View File
@@ -45,6 +45,7 @@ typedef struct {
time_stats_t *toutput;
}encoder_implemparams_t;
#define INIT0_LDPCIMPLEMPARAMS {0,0,0,NULL,NULL,NULL,NULL}
typedef void(*nrLDPC_initcallfunc_t)(t_nrLDPC_dec_params* p_decParams, int8_t* p_llr, int8_t* p_out);
typedef int(*nrLDPC_encoderfunc_t)(unsigned char **,unsigned char **,int,int,short, short, encoder_implemparams_t*);
//============================================================================================================================
// decoder interface
@@ -56,4 +57,4 @@ 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
#endif
openair1/PHY/CODING/nrLDPC_extern.h
+5 -2
View File
@@ -23,13 +23,16 @@
#ifdef LDPC_LOADER
nrLDPC_decoderfunc_t nrLDPC_decoder;
nrLDPC_encoderfunc_t nrLDPC_encoder;
nrLDPC_initcallfunc_t nrLDPC_initcall;
#else
/* functions to load the LDPC shared lib, implemented in openair1/PHY/CODING/nrLDPC_load.c */
extern int load_nrLDPClib(void) ;
extern int load_nrLDPClib(char *version) ;
extern int load_nrLDPClib_ref(char *libversion, nrLDPC_encoderfunc_t * nrLDPC_encoder_ptr); // for ldpctest
/* ldpc coder/decoder functions, as loaded by load_nrLDPClib(). */
extern nrLDPC_initcallfunc_t nrLDPC_initcall;
extern nrLDPC_decoderfunc_t nrLDPC_decoder;
extern nrLDPC_encoderfunc_t nrLDPC_encoder;
// inline functions:
#include "openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_init_mem.h"
#endif
#endif
openair1/PHY/CODING/nrLDPC_load.c
+17 -9
View File
@@ -42,22 +42,32 @@
/* function description array, to be used when loading the encoding/decoding shared lib */
static loader_shlibfunc_t shlib_fdesc[2];
static loader_shlibfunc_t shlib_fdesc[3];
char *arg[64]={"ldpctest","-O","cmdlineonly::dbgl0"};
/* 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};
int load_nrLDPClib(void) {
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(3,(char **)arg,CONFIG_ENABLECMDLINEONLY) ;
load_configmodule(0,(char **)NULL,CONFIG_ENABLECMDLINEONLY) ;
logInit();
}
shlib_fdesc[0].fname = "nrLDPC_decod";
shlib_fdesc[1].fname = "nrLDPC_encod";
int ret=load_module_shlib("ldpc",shlib_fdesc,sizeof(shlib_fdesc)/sizeof(loader_shlibfunc_t),NULL);
shlib_fdesc[2].fname = "nrLDPC_initcall";
int ret;
if (version)
ret=load_module_version_shlib(libname,version,shlib_fdesc,sizeof(shlib_fdesc)/sizeof(loader_shlibfunc_t),NULL);
else
ret=load_module_shlib(libname,shlib_fdesc,sizeof(shlib_fdesc)/sizeof(loader_shlibfunc_t),NULL);
AssertFatal( (ret >= 0),"Error loading ldpc decoder");
nrLDPC_decoder = (nrLDPC_decoderfunc_t)shlib_fdesc[0].fptr;
nrLDPC_encoder = (nrLDPC_encoderfunc_t)shlib_fdesc[1].fptr;
nrLDPC_initcall = (nrLDPC_initcallfunc_t)shlib_fdesc[2].fptr;
return 0;
}
@@ -65,10 +75,8 @@ int load_nrLDPClib_ref(char *libversion, nrLDPC_encoderfunc_t * nrLDPC_encoder_p
loader_shlibfunc_t shlib_encoder_fdesc;
shlib_encoder_fdesc.fname = "nrLDPC_encod";
char libpath[64];
sprintf(libpath,"ldpc%s",libversion);
int ret=load_module_shlib(libpath,&shlib_encoder_fdesc,1,NULL);
AssertFatal( (ret >= 0),"Error loading ldpc encoder %s\n",libpath);
int ret=load_module_version_shlib("ldpc",libversion,&shlib_encoder_fdesc,1,NULL);
AssertFatal( (ret >= 0),"Error loading ldpc encoder %s\n",(libversion==NULL)?"":libversion);
*nrLDPC_encoder_ptr = (nrLDPC_encoderfunc_t)shlib_encoder_fdesc.fptr;
return 0;
}
openair1/PHY/CODING/nr_rate_matching.c
+19 -14
View File
@@ -388,7 +388,7 @@ int nr_rate_matching_ldpc(uint8_t Ilbrm,
uint32_t Ncb,ind,k=0,Nref,N;
if (C==0) {
printf("nr_rate_matching: invalid parameters (C %d\n",C);
LOG_E(PHY,"nr_rate_matching: invalid parameters (C %d\n",C);
return -1;
}
@@ -407,16 +407,15 @@ int nr_rate_matching_ldpc(uint8_t Ilbrm,
#ifdef RM_DEBUG
printf("nr_rate_matching_ldpc: E %d, F %d, Foffset %d, k0 %d, Ncb %d, rvidx %d\n", E, F, Foffset,ind, Ncb, rvidx);
#endif
AssertFatal(Foffset <= E,
"Foffset %d > E %d "
"(Ilbrm %d, Tbslbrm %d, Z %d, BG %d, C %d)\n",
Foffset, E,
Ilbrm, Tbslbrm, Z, BG, C);
AssertFatal(Foffset <= Ncb,
"Foffset %d > Ncb %d "
"(Ilbrm %d, Tbslbrm %d, Z %d, BG %d, C %d)\n",
Foffset, Ncb,
Ilbrm, Tbslbrm, Z, BG, C);
if (Foffset > E) {
LOG_E(PHY,"nr_rate_matching: invalid parameters (Foffset %d > E %d)\n",Foffset,E);
return -1;
}
if (Foffset > Ncb) {
LOG_E(PHY,"nr_rate_matching: invalid parameters (Foffset %d > Ncb %d)\n",Foffset,Ncb);
return -1;
}
if (ind >= Foffset && ind < (F+Foffset)) ind = F+Foffset;
@@ -478,7 +477,7 @@ int nr_rate_matching_ldpc_rx(uint8_t Ilbrm,
#endif
if (C==0) {
printf("nr_rate_matching: invalid parameters (C %d\n",C);
LOG_E(PHY,"nr_rate_matching: invalid parameters (C %d\n",C);
return -1;
}
@@ -493,8 +492,14 @@ int nr_rate_matching_ldpc_rx(uint8_t Ilbrm,
}
ind = (index_k0[BG-1][rvidx]*Ncb/N)*Z;
AssertFatal(Foffset <= E,"Foffset %d > E %d\n",Foffset,E);
AssertFatal(Foffset <= Ncb,"Foffset %d > Ncb %d\n",Foffset,Ncb);
if (Foffset > E) {
LOG_E(PHY,"nr_rate_matching: invalid parameters (Foffset %d > E %d)\n",Foffset,E);
return -1;
}
if (Foffset > Ncb) {
LOG_E(PHY,"nr_rate_matching: invalid parameters (Foffset %d > Ncb %d)\n",Foffset,Ncb);
return -1;
}
#ifdef RM_DEBUG
printf("nr_rate_matching_ldpc_rx: Clear %d, E %d, k0 %d, Ncb %d, rvidx %d\n", clear, E, ind, Ncb, rvidx);
openair1/PHY/INIT/nr_init.c
+1 -1
View File
@@ -108,7 +108,7 @@ int phy_init_nr_gNB(PHY_VARS_gNB *gNB,
crcTableInit();
init_scrambling_luts();
init_pucch2_luts();
load_nrLDPClib();
load_nrLDPClib(NULL);
// PBCH DMRS gold sequences generation
nr_init_pbch_dmrs(gNB);
//PDCCH DMRS init
openair1/PHY/INIT/nr_parms.c
+63 -10
View File
@@ -27,6 +27,58 @@
uint32_t nr_subcarrier_spacing[MAX_NUM_SUBCARRIER_SPACING] = {15e3, 30e3, 60e3, 120e3, 240e3};
uint16_t nr_slots_per_subframe[MAX_NUM_SUBCARRIER_SPACING] = {1, 2, 4, 8, 16};
// Table 5.4.3.3-1 38-101
int nr_ssb_table[48][3] = {
{1, 15, nr_ssb_type_A},
{2, 15, nr_ssb_type_A},
{3, 15, nr_ssb_type_A},
{5, 15, nr_ssb_type_A},
{5, 30, nr_ssb_type_B},
{7, 15, nr_ssb_type_A},
{8, 15, nr_ssb_type_A},
{12, 15, nr_ssb_type_A},
{14, 15, nr_ssb_type_A},
{18, 15, nr_ssb_type_A},
{20, 15, nr_ssb_type_A},
{25, 15, nr_ssb_type_A},
{26, 15, nr_ssb_type_A},
{28, 15, nr_ssb_type_A},
{29, 15, nr_ssb_type_A},
{30, 15, nr_ssb_type_A},
{34, 15, nr_ssb_type_A},
{34, 30, nr_ssb_type_C},
{38, 15, nr_ssb_type_A},
{38, 30, nr_ssb_type_C},
{39, 15, nr_ssb_type_A},
{39, 30, nr_ssb_type_C},
{40, 30, nr_ssb_type_C},
{41, 15, nr_ssb_type_A},
{41, 30, nr_ssb_type_C},
{46, 30, nr_ssb_type_C},
{48, 30, nr_ssb_type_C},
{50, 30, nr_ssb_type_C},
{51, 15, nr_ssb_type_A},
{53, 15, nr_ssb_type_A},
{65, 15, nr_ssb_type_A},
{66, 15, nr_ssb_type_A},
{66, 30, nr_ssb_type_B},
{70, 15, nr_ssb_type_A},
{71, 15, nr_ssb_type_A},
{74, 15, nr_ssb_type_A},
{75, 15, nr_ssb_type_A},
{76, 15, nr_ssb_type_A},
{77, 30, nr_ssb_type_C},
{78, 30, nr_ssb_type_C},
{79, 30, nr_ssb_type_C},
{90, 15, nr_ssb_type_A},
{90, 30, nr_ssb_type_C},
{91, 15, nr_ssb_type_A},
{92, 15, nr_ssb_type_A},
{93, 15, nr_ssb_type_A},
{94, 15, nr_ssb_type_A},
{96, 30, nr_ssb_type_C}
};
void set_Lmax(NR_DL_FRAME_PARMS *fp) {
// definition of Lmax according to ts 38.213 section 4.1
@@ -85,27 +137,28 @@ int nr_get_ssb_start_symbol(NR_DL_FRAME_PARMS *fp,uint8_t i_ssb) {
void set_scs_parameters (NR_DL_FRAME_PARMS *fp, int mu, int N_RB_DL)
{
int idx = 0;
switch(mu) {
case NR_MU_0: //15kHz scs
fp->subcarrier_spacing = nr_subcarrier_spacing[NR_MU_0];
fp->slots_per_subframe = nr_slots_per_subframe[NR_MU_0];
fp->ssb_type = nr_ssb_type_A;
while(nr_ssb_table[idx][0]!=fp->nr_band)
idx++;
AssertFatal(nr_ssb_table[idx][1]==15,"SCS %d not applicable to band %d\n",
fp->subcarrier_spacing,fp->nr_band);
break;
case NR_MU_1: //30kHz scs
fp->subcarrier_spacing = nr_subcarrier_spacing[NR_MU_1];
fp->slots_per_subframe = nr_slots_per_subframe[NR_MU_1];
// selection of SS block pattern according to TS 38101-1 Table 5.4.3.3-1 for SCS 30kHz
if (fp->nr_band == 5 || fp->nr_band == 66)
fp->ssb_type = nr_ssb_type_B;
else{
if (fp->nr_band == 41 || fp->nr_band == 38 || ( fp->nr_band > 76 && fp->nr_band < 80) )
fp->ssb_type = nr_ssb_type_C;
else
AssertFatal(1==0,"NR Operating Band n%d not available for SS block SCS with mu=%d\n", fp->nr_band, mu);
while(nr_ssb_table[idx][0]!=fp->nr_band ||
nr_ssb_table[idx][1]!=30) {
AssertFatal(nr_ssb_table[idx][0]<=fp->nr_band,"SCS %d not applicable to band %d\n",
fp->subcarrier_spacing,fp->nr_band);
idx++;
}
fp->ssb_type = nr_ssb_table[idx][2];
break;
case NR_MU_2: //60kHz scs
openair1/PHY/LTE_TRANSPORT/dci.c
-2
View File
@@ -50,8 +50,6 @@
//#define DEBUG_DCI_DECODING 1
//#define DEBUG_PHY
//#undef ALL_AGGREGATION
//extern uint16_t phich_reg[MAX_NUM_PHICH_GROUPS][3];
//extern uint16_t pcfich_reg[4];
openair1/PHY/LTE_TRANSPORT/dci_NB_IoT.h
+1 -4
View File
@@ -32,11 +32,8 @@
#ifndef __DCI_NB_IOT_H__
#define __DCI_NB_IOT_H__
//#ifndef USER_MODE
//#include "PHY/types.h"
//#else
#include <stdint.h>
//#endif
typedef enum
{
openair1/PHY/LTE_TRANSPORT/print_stats.c
-4
View File
@@ -36,10 +36,8 @@
#include "PHY/extern.h"
#include "SCHED/extern.h"
#ifdef OPENAIR2
#include "openair2/LAYER2/MAC/proto.h"
#include "openair2/RRC/L2_INTERFACE/openair_rrc_L2_interface.h"
#endif
extern int mac_get_rrc_status(uint8_t Mod_id,uint8_t eNB_flag,uint8_t index);
@@ -434,10 +432,8 @@ int dump_ue_stats(PHY_VARS_UE *ue, UE_rxtx_proc_t *proc,char *buffer, int length
break;
}
#ifdef OPENAIR2
RRC_status = mac_UE_get_rrc_status(ue->Mod_id, 0);
len += sprintf(&buffer[len],"[UE PROC] RRC status = %d\n",RRC_status);
#endif
len += sprintf(&buffer[len], "[UE PROC] Transmission Mode %d \n",ue->transmission_mode[eNB]);
len += sprintf(&buffer[len], "[UE PROC] PBCH err conseq %d, PBCH error total %d, PBCH FER %d\n",
ue->pbch_vars[eNB]->pdu_errors_conseq,
openair1/PHY/LTE_TRANSPORT/uci_NB_IoT.h
+1 -1
View File
@@ -321,4 +321,4 @@ HLC_subband_cqi_mcs_CBA_20MHz_NB_IoT;
#define MAX_ACK_PAYLOAD_NB_IoT 18
#define MAX_RI_PAYLOAD_NB_IoT 6
#endif
#endif
openair1/PHY/LTE_UE_TRANSPORT/dci_ue.c
-9
View File
@@ -50,9 +50,6 @@
//#define DEBUG_DCI_DECODING 1
//#define DEBUG_PHY
//#undef ALL_AGGREGATION
uint16_t extract_crc(uint8_t *dci,uint8_t dci_len)
{
@@ -2802,7 +2799,6 @@ uint16_t dci_decoding_procedure(PHY_VARS_UE *ue,
if (dci_cnt>old_dci_cnt)
return(dci_cnt);
//#ifdef ALL_AGGREGATION
// Now check UE_SPEC format 1 search spaces at aggregation 8
old_dci_cnt=dci_cnt;
dci_decoding_procedure0(pdcch_vars,0,mode,subframe,
@@ -2836,7 +2832,6 @@ uint16_t dci_decoding_procedure(PHY_VARS_UE *ue,
if (dci_cnt>old_dci_cnt)
return(dci_cnt);
//#endif //ALL_AGGREGATION
} else if (tmode == 3) {
@@ -2946,7 +2941,6 @@ uint16_t dci_decoding_procedure(PHY_VARS_UE *ue,
if (dci_cnt>old_dci_cnt)
return(dci_cnt);
//#ifdef ALL_AGGREGATION
// Now check UE_SPEC format 2_2A search spaces at aggregation 8
LOG_D(PHY," Now check UE_SPEC format 2_2A search spaces at aggregation 8 dci length: %d[bits] %d[bytes]\n",format2A_size_bits,format2A_size_bytes);
old_dci_cnt=dci_cnt;
@@ -3082,7 +3076,6 @@ uint16_t dci_decoding_procedure(PHY_VARS_UE *ue,
if (dci_cnt>old_dci_cnt)
return(dci_cnt);
//#ifdef ALL_AGGREGATION
// Now check UE_SPEC format 2_2A search spaces at aggregation 8
old_dci_cnt=dci_cnt;
dci_decoding_procedure0(pdcch_vars,0,mode,
@@ -3214,7 +3207,6 @@ uint16_t dci_decoding_procedure(PHY_VARS_UE *ue,
if (dci_cnt>old_dci_cnt)
return(dci_cnt);
//#ifdef ALL_AGGREGATION
// Now check UE_SPEC format 1E_2A_M10PRB search spaces at aggregation 8
old_dci_cnt=dci_cnt;
@@ -3249,7 +3241,6 @@ uint16_t dci_decoding_procedure(PHY_VARS_UE *ue,
if (dci_cnt>old_dci_cnt)
return(dci_cnt);
//#endif //ALL_AGGREGATION
}
openair1/PHY/LTE_UE_TRANSPORT/dlsch_demodulation.c
-1
View File
@@ -51,7 +51,6 @@
#include <cblas.h>
#include "linear_preprocessing_rec.h"
#define NOCYGWIN_STATIC
//#define DEBUG_MMSE
openair1/PHY/LTE_UE_TRANSPORT/ulsch_modulation.c
-52
View File
@@ -42,7 +42,6 @@
//#define DEBUG_ULSCH_MODULATION
#ifndef OFDMA_ULSCH
void dft_lte(int32_t *z,int32_t *d, int32_t Msc_PUSCH, uint8_t Nsymb)
{
@@ -371,7 +370,6 @@ void dft_lte(int32_t *z,int32_t *d, int32_t Msc_PUSCH, uint8_t Nsymb)
// printf("\n");
}
#endif
void ulsch_modulation(int32_t **txdataF,
short amp,
uint32_t frame,
@@ -690,57 +688,10 @@ void ulsch_modulation(int32_t **txdataF,
// Transform Precoding
#ifdef OFDMA_ULSCH
for (i=0; i<ulsch_Msymb; i++) {
ulsch->z[i] = ulsch->d[i];
}
#else
dft_lte(ulsch->z,ulsch->d,Msc_PUSCH,ulsch->Nsymb_pusch);
#endif
DevAssert(txdataF);
#ifdef OFDMA_ULSCH
re_offset0 = frame_parms->first_carrier_offset + (ulsch->harq_processes[harq_pid]->first_rb*12);
if (re_offset0>frame_parms->ofdm_symbol_size) {
re_offset0 -= frame_parms->ofdm_symbol_size;
// re_offset0++;
}
// printf("re_offset0 %d\n",re_offset0);
for (j=0,l=0; l<(nsymb-ulsch->srs_active); l++) {
re_offset = re_offset0;
symbol_offset = (int)frame_parms->ofdm_symbol_size*(l+(subframe*nsymb));
#ifdef DEBUG_ULSCH_MODULATION
printf("symbol %d (subframe %d): symbol_offset %d\n",l,subframe,symbol_offset);
#endif
txptr = &txdataF[0][symbol_offset];
if (((frame_parms->Ncp == 0) && ((l==3) || (l==10)))||
((frame_parms->Ncp == 1) && ((l==2) || (l==8)))) {
}
// Skip reference symbols
else {
// printf("copying %d REs\n",Msc_PUSCH);
for (i=0; i<Msc_PUSCH; i++,j++) {
#ifdef DEBUG_ULSCH_MODULATION
printf("re_offset %d (%p): %d,%d\n", re_offset,&ulsch->z[j],((int16_t*)&ulsch->z[j])[0],((int16_t*)&ulsch->z[j])[1]);
#endif
txptr[re_offset++] = ulsch->z[j];
if (re_offset==frame_parms->ofdm_symbol_size)
re_offset = 0;
}
}
}
# else // OFDMA_ULSCH = 0
re_offset0 = frame_parms->first_carrier_offset + (ulsch->harq_processes[harq_pid]->first_rb*12);
if (re_offset0>frame_parms->ofdm_symbol_size) {
@@ -777,8 +728,5 @@ void ulsch_modulation(int32_t **txdataF,
}
}
#endif
VCD_SIGNAL_DUMPER_DUMP_FUNCTION_BY_NAME(VCD_SIGNAL_DUMPER_FUNCTIONS_UE_ULSCH_MODULATION, VCD_FUNCTION_OUT);
}
openair1/PHY/NR_TRANSPORT/nr_dci.c
+1
View File
@@ -35,6 +35,7 @@
#include "nr_dlsch.h"
#include "nr_sch_dmrs.h"
#include "PHY/MODULATION/nr_modulation.h"
#include "common/utils/nr/nr_common.h"
//#define DEBUG_PDCCH_DMRS
//#define DEBUG_DCI
openair1/PHY/NR_TRANSPORT/nr_dci.h
-2
View File
@@ -59,6 +59,4 @@ void nr_fill_ul_dci(PHY_VARS_gNB *gNB,
void nr_fill_cce_list(PHY_VARS_gNB *gNB, uint8_t m,nfapi_nr_dl_tti_pdcch_pdu_rel15_t *);
void get_coreset_rballoc(uint8_t *FreqDomainResource,int *n_rb,int *rb_offset);
#endif //__PHY_NR_TRANSPORT_DCI__H
openair1/PHY/NR_TRANSPORT/nr_dci_tools.c
+1
View File
@@ -31,6 +31,7 @@
*/
#include "nr_dci.h"
#include "common/utils/nr/nr_common.h"
//#define DEBUG_FILL_DCI
openair1/PHY/NR_TRANSPORT/nr_dci_tools_common.c
-62
View File
@@ -1,62 +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 1.1 (the "License"); you may not use this file
* except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.openairinterface.org/?page_id=698
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*-------------------------------------------------------------------------------
* For more information about the OpenAirInterface (OAI) Software Alliance:
* contact@openairinterface.org
*/
/*! \file PHY/NR_TRANSPORT/nr_dci_tools_common.c
* \brief
* \author
* \date 2018
* \version 0.1
* \company Eurecom
* \email:
* \note
* \warning
*/
#include "nr_dci.h"
//#define DEBUG_FILL_DCI
#include "nr_dlsch.h"
void get_coreset_rballoc(uint8_t *FreqDomainResource,int *n_rb,int *rb_offset) {
uint8_t count=0, start=0, start_set=0;
uint64_t bitmap = (((uint64_t)FreqDomainResource[0])<<37)|
(((uint64_t)FreqDomainResource[1])<<29)|
(((uint64_t)FreqDomainResource[2])<<21)|
(((uint64_t)FreqDomainResource[3])<<13)|
(((uint64_t)FreqDomainResource[4])<<5)|
(((uint64_t)FreqDomainResource[5])>>3);
for (int i=0; i<45; i++)
if ((bitmap>>(44-i))&1) {
count++;
if (!start_set) {
start = i;
start_set = 1;
}
}
*rb_offset = 6*start;
*n_rb = 6*count;
}
openair1/PHY/NR_TRANSPORT/nr_dlsch.c
+9 -11
View File
@@ -111,9 +111,9 @@ void nr_pdsch_codeword_scrambling_optim(uint8_t *in,
}
uint8_t nr_generate_pdsch(processingData_L1tx_t *msgTx,
int frame,
int slot) {
void nr_generate_pdsch(processingData_L1tx_t *msgTx,
int frame,
int slot) {
PHY_VARS_gNB *gNB = msgTx->gNB;
NR_gNB_DLSCH_t *dlsch;
@@ -181,11 +181,12 @@ uint8_t nr_generate_pdsch(processingData_L1tx_t *msgTx,
/// CRC, coding, interleaving and rate matching
AssertFatal(harq->pdu!=NULL,"harq->pdu is null\n");
start_meas(dlsch_encoding_stats);
nr_dlsch_encoding(gNB,
harq->pdu, frame, slot, dlsch, frame_parms,tinput,tprep,tparity,toutput,
dlsch_rate_matching_stats,
dlsch_interleaving_stats,
dlsch_segmentation_stats);
if (nr_dlsch_encoding(gNB,
harq->pdu, frame, slot, dlsch, frame_parms,tinput,tprep,tparity,toutput,
dlsch_rate_matching_stats,
dlsch_interleaving_stats,
dlsch_segmentation_stats) == -1)
return;
stop_meas(dlsch_encoding_stats);
#ifdef DEBUG_DLSCH
printf("PDSCH encoding:\nPayload:\n");
@@ -522,9 +523,6 @@ uint8_t nr_generate_pdsch(processingData_L1tx_t *msgTx,
LOG_D(PHY,"beam index for PDSCH allocation already taken\n");
}
}// dlsch loop
return 0;
}
void dump_pdsch_stats(FILE *fd,PHY_VARS_gNB *gNB) {
openair1/PHY/NR_TRANSPORT/nr_dlsch.h
+3 -3
View File
@@ -67,9 +67,9 @@ void nr_fill_dlsch(processingData_L1tx_t *msgTx,
nfapi_nr_dl_tti_pdsch_pdu *pdsch_pdu,
unsigned char *sdu);
uint8_t nr_generate_pdsch(processingData_L1tx_t *msgTx,
int frame,
int slot);
void nr_generate_pdsch(processingData_L1tx_t *msgTx,
int frame,
int slot);
void free_gNB_dlsch(NR_gNB_DLSCH_t **dlschptr, uint16_t N_RB);
void clean_gNB_dlsch(NR_gNB_DLSCH_t *dlsch);
openair1/PHY/NR_TRANSPORT/nr_dlsch_coding.c
+13 -11
View File
@@ -427,17 +427,19 @@ int nr_dlsch_encoding(PHY_VARS_gNB *gNB,
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);
if (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) == -1)
return -1;
stop_meas(dlsch_rate_matching_stats);
#ifdef DEBUG_DLSCH_CODING
for (int i =0; i<16; i++)
openair1/PHY/NR_TRANSPORT/pucch_rx.c
+44 -44
View File
@@ -206,7 +206,9 @@ void nr_decode_pucch0(PHY_VARS_gNB *gNB,
nr_sequences=8>>(1-pucch_pdu->sr_flag);
}
LOG_D(PHY,"pucch0: nr_symbols %d, start_symbol %d, prb_start %d, second_hop_prb %d, group_hop_flag %d, sequence_hop_flag %d, O_ACK %d, O_SR %d, mcs %d initial_cyclic_shift %d\n",pucch_pdu->nr_of_symbols,pucch_pdu->start_symbol_index,pucch_pdu->prb_start,pucch_pdu->second_hop_prb,pucch_pdu->group_hop_flag,pucch_pdu->sequence_hop_flag,pucch_pdu->bit_len_harq,pucch_pdu->sr_flag,mcs[0],pucch_pdu->initial_cyclic_shift);
LOG_D(PHY,"pucch0: nr_symbols %d, start_symbol %d, prb_start %d, second_hop_prb %d, group_hop_flag %d, sequence_hop_flag %d, O_ACK %d, O_SR %d, mcs %d initial_cyclic_shift %d\n",
pucch_pdu->nr_of_symbols,pucch_pdu->start_symbol_index,pucch_pdu->prb_start,pucch_pdu->second_hop_prb,pucch_pdu->group_hop_flag,pucch_pdu->sequence_hop_flag,pucch_pdu->bit_len_harq,
pucch_pdu->sr_flag,mcs[0],pucch_pdu->initial_cyclic_shift);
int cs_ind = get_pucch0_cs_lut_index(gNB,pucch_pdu);
/*
@@ -238,7 +240,7 @@ void nr_decode_pucch0(PHY_VARS_gNB *gNB,
// x_n contains the sequence r_u_v_alpha_delta(n)
int n,i,l;
int n,i;
int prb_offset[2] = {pucch_pdu->bwp_start+pucch_pdu->prb_start, pucch_pdu->bwp_start+pucch_pdu->prb_start};
nr_group_sequence_hopping(pucch_GroupHopping,pucch_pdu->hopping_id,0,slot,&u[0],&v[0]); // calculating u and v value first hop
@@ -266,14 +268,14 @@ void nr_decode_pucch0(PHY_VARS_gNB *gNB,
int64_t xrtmag=0,xrtmag_next=0;
uint8_t maxpos=0;
uint8_t index=0;
for (l=0; l<pucch_pdu->nr_of_symbols; l++) {
for (int l=0; l<pucch_pdu->nr_of_symbols; l++) {
for (int aarx=0;aarx<frame_parms->nb_antennas_rx;aarx++) {
memset((void*)xr[aarx][l],0,24*sizeof(int16_t));
}
}
int n2;
for (l=0; l<pucch_pdu->nr_of_symbols; l++) {
for (int l=0; l<pucch_pdu->nr_of_symbols; l++) {
l2 = l+pucch_pdu->start_symbol_index;
re_offset[l] = (12*prb_offset[l]) + frame_parms->first_carrier_offset;
if (re_offset[l]>= frame_parms->ofdm_symbol_size)
@@ -300,11 +302,10 @@ void nr_decode_pucch0(PHY_VARS_gNB *gNB,
//int32_t no_corr = 0;
int seq_index;
int64_t temp;
int64_t av_corr=0;
for(i=0;i<nr_sequences;i++){
for (l=0;l<pucch_pdu->nr_of_symbols;l++) {
for (int l=0;l<pucch_pdu->nr_of_symbols;l++) {
seq_index = (pucch_pdu->initial_cyclic_shift+
mcs[i]+
gNB->pucch0_lut.lut[cs_ind][slot][l+pucch_pdu->start_symbol_index])%12;
@@ -322,26 +323,30 @@ void nr_decode_pucch0(PHY_VARS_gNB *gNB,
}
}
}
LOG_D(PHY,"PUCCH IDFT[%d/%d] = (%d,%d)=>%f\n",mcs[i],seq_index,corr_re[0][0],corr_im[0][0],10*log10((double)corr_re[0][0]*corr_re[0][0] + (double)corr_im[0][0]*corr_im[0][0]));
if (l>1) LOG_D(PHY,"PUCCH 2nd symbol IDFT[%d/%d] = (%d,%d)=>%f\n",mcs[i],seq_index,corr_re[0][1],corr_im[0][1],10*log10((double)corr_re[0][1]*corr_re[0][1] + (double)corr_im[0][1]*corr_im[0][1]));
if (pucch_pdu->freq_hop_flag == 0 && l==1) {// non-coherent correlation
temp=0;
for (int aa=0;aa<frame_parms->nb_antennas_rx;aa++)
temp+=(int64_t)corr_re[aa][0]*corr_re[aa][0] + (int64_t)corr_im[aa][0]*corr_im[aa][0];
}
else if (pucch_pdu->freq_hop_flag == 0 && l==2) {
int64_t corr_re2=0;
int64_t corr_im2=0;
temp=0;
for (int aa=0;aa<frame_parms->nb_antennas_rx;aa++) {
corr_re2 = (int64_t)corr_re[aa][0]+corr_re[aa][1];
corr_im2 = (int64_t)corr_im[aa][0]+corr_im[aa][1];
// coherent combining of 2 symbols and then complex modulus for single-frequency case
temp+=corr_re2*corr_re2 + corr_im2*corr_im2;
}
}
else if (pucch_pdu->freq_hop_flag == 1) {
LOG_D(PHY,"PUCCH IDFT[%d/%d] = (%d,%d)=>%f\n",
mcs[i],seq_index,corr_re[0][0],corr_im[0][0],
10*log10((double)corr_re[0][0]*corr_re[0][0] + (double)corr_im[0][0]*corr_im[0][0]));
if (pucch_pdu->nr_of_symbols==2)
LOG_D(PHY,"PUCCH 2nd symbol IDFT[%d/%d] = (%d,%d)=>%f\n",
mcs[i],seq_index,corr_re[0][1],corr_im[0][1],
10*log10((double)corr_re[0][1]*corr_re[0][1] + (double)corr_im[0][1]*corr_im[0][1]));
if (pucch_pdu->freq_hop_flag == 0) {
if (pucch_pdu->nr_of_symbols==1) {// non-coherent correlation
temp=0;
for (int aa=0;aa<frame_parms->nb_antennas_rx;aa++)
temp+=(int64_t)corr_re[aa][0]*corr_re[aa][0] + (int64_t)corr_im[aa][0]*corr_im[aa][0];
} else {
int64_t corr_re2=0;
int64_t corr_im2=0;
temp=0;
for (int aa=0;aa<frame_parms->nb_antennas_rx;aa++) {
corr_re2 = (int64_t)corr_re[aa][0]+corr_re[aa][1];
corr_im2 = (int64_t)corr_im[aa][0]+corr_im[aa][1];
// coherent combining of 2 symbols and then complex modulus for single-frequency case
temp+=corr_re2*corr_re2 + corr_im2*corr_im2;
}
}
} else if (pucch_pdu->freq_hop_flag == 1) {
// full non-coherent combining of 2 symbols for frequency-hopping case
temp=0;
for (int aa=0;aa<frame_parms->nb_antennas_rx;aa++)
@@ -349,7 +354,6 @@ void nr_decode_pucch0(PHY_VARS_gNB *gNB,
}
else AssertFatal(1==0,"shouldn't happen\n");
av_corr+=temp;
if (temp>xrtmag) {
xrtmag_next = xrtmag;
xrtmag=temp;
@@ -359,19 +363,19 @@ void nr_decode_pucch0(PHY_VARS_gNB *gNB,
int64_t temp2=0,temp3=0;;
for (int aa=0;aa<frame_parms->nb_antennas_rx;aa++) {
temp2 += ((int64_t)corr_re[aa][0]*corr_re[aa][0] + (int64_t)corr_im[aa][0]*corr_im[aa][0]);
if (l==2) temp3 += ((int64_t)corr_re[aa][1]*corr_re[aa][1] + (int64_t)corr_im[aa][1]*corr_im[aa][1]);
if (pucch_pdu->nr_of_symbols==2)
temp3 += ((int64_t)corr_re[aa][1]*corr_re[aa][1] + (int64_t)corr_im[aa][1]*corr_im[aa][1]);
}
uci_stats->current_pucch0_stat0= dB_fixed64(temp2);
if (l==2) uci_stats->current_pucch0_stat1= dB_fixed64(temp3);
if ( pucch_pdu->nr_of_symbols==2)
uci_stats->current_pucch0_stat1= dB_fixed64(temp3);
}
else if (temp>xrtmag_next)
xrtmag_next = temp;
}
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));
int xrtmag_dBtimes10 = 10*(int)dB_fixed64(xrtmag/(12*pucch_pdu->nr_of_symbols));
int xrtmag_next_dBtimes10 = 10*(int)dB_fixed64(xrtmag_next/(12*pucch_pdu->nr_of_symbols));
#ifdef DEBUG_NR_PUCCH_RX
printf("PUCCH 0 : maxpos %d\n",maxpos);
#endif
@@ -425,11 +429,9 @@ void nr_decode_pucch0(PHY_VARS_gNB *gNB,
uci_pdu->harq->num_harq = 1;
uci_pdu->harq->harq_confidence_level = no_conf ? 1 : 0;
uci_pdu->harq->harq_list = (nfapi_nr_harq_t*)malloc(1);
uci_pdu->harq->harq_list[0].harq_value = index&0x01;
LOG_D(PHY, "[DLSCH/PDSCH/PUCCH] %d.%d HARQ value %d with confidence level %d (0 is good, 1 is bad) xrt_mag %d xrt_mag_next %d n0 %d (%d,%d) pucch0_thres %d, cqi %d, SNRtimes10 %d, energy %f, sync_pos %d\n",
frame,slot,uci_pdu->harq->harq_list[0].harq_value,uci_pdu->harq->harq_confidence_level,
xrtmag_dBtimes10,xrtmag_next_dBtimes10,max_n0,uci_stats->pucch0_n00,uci_stats->pucch0_n01,
uci_stats->pucch0_thres,cqi,SNRtimes10,10*log10((double)sigenergy),gNB->ulsch_stats[0].sync_pos);
uci_pdu->harq->harq_list[0].harq_value = !(index&0x01);
LOG_D(PHY, "[DLSCH/PDSCH/PUCCH] %d.%d HARQ value %d (0 pass, 1 fail) with confidence level %d (0 is good, 1 is bad) xrt_mag %d xrt_mag_next %d n0 %d (%d,%d) pucch0_thres %d, cqi %d, SNRtimes10 %d, energy %f, sync_pos %d\n",
frame,slot,uci_pdu->harq->harq_list[0].harq_value,uci_pdu->harq->harq_confidence_level,xrtmag_dBtimes10,xrtmag_next_dBtimes10,max_n0,uci_stats->pucch0_n00,uci_stats->pucch0_n01,uci_stats->pucch0_thres,cqi,SNRtimes10,10*log10((double)sigenergy),gNB->ulsch_stats[0].sync_pos);
if (pucch_pdu->sr_flag == 1) {
uci_pdu->sr = calloc(1,sizeof(*uci_pdu->sr));
uci_pdu->sr->sr_indication = (index>1) ? 1 : 0;
@@ -443,12 +445,10 @@ void nr_decode_pucch0(PHY_VARS_gNB *gNB,
uci_pdu->harq->num_harq = 2;
uci_pdu->harq->harq_confidence_level = (no_conf) ? 1 : 0;
uci_pdu->harq->harq_list = (nfapi_nr_harq_t*)malloc(2);
uci_pdu->harq->harq_list[1].harq_value = index&0x01;
uci_pdu->harq->harq_list[0].harq_value = (index>>1)&0x01;
LOG_D(PHY, "[DLSCH/PDSCH/PUCCH] %d.%d HARQ values %d and %d with confidence level %d (0 is good, 1 is bad), xrt_mag %d xrt_mag_next %d n0 %d (%d,%d) pucch0_thres %d, cqi %d, SNRtimes10 %d,sync_pos %d\n",
frame,slot,uci_pdu->harq->harq_list[1].harq_value,uci_pdu->harq->harq_list[0].harq_value,
uci_pdu->harq->harq_confidence_level,xrtmag_dBtimes10,xrtmag_next_dBtimes10,max_n0,
uci_stats->pucch0_n00,uci_stats->pucch0_n01,uci_stats->pucch0_thres,cqi,SNRtimes10,gNB->ulsch_stats[0].sync_pos);
uci_pdu->harq->harq_list[1].harq_value = !(index&0x01);
uci_pdu->harq->harq_list[0].harq_value = !((index>>1)&0x01);
LOG_D(PHY, "[DLSCH/PDSCH/PUCCH] %d.%d HARQ values %d (0 pass, 1 fail) and %d with confidence level %d (0 is good, 1 is bad), xrt_mag %d xrt_mag_next %d n0 %d (%d,%d) pucch0_thres %d, cqi %d, SNRtimes10 %d,sync_pos %d\n",
frame,slot,uci_pdu->harq->harq_list[1].harq_value,uci_pdu->harq->harq_list[0].harq_value,uci_pdu->harq->harq_confidence_level,xrtmag_dBtimes10,xrtmag_next_dBtimes10,max_n0,uci_stats->pucch0_n00,uci_stats->pucch0_n01,uci_stats->pucch0_thres,cqi,SNRtimes10,gNB->ulsch_stats[0].sync_pos);
if (pucch_pdu->sr_flag == 1) {
uci_pdu->sr = calloc(1,sizeof(*uci_pdu->sr));
uci_pdu->sr->sr_indication = (index>3) ? 1 : 0;
openair1/PHY/NR_UE_ESTIMATION/nr_adjust_synch_ue.c
+10 -13
View File
@@ -42,7 +42,6 @@ void nr_adjust_synch_ue(NR_DL_FRAME_PARMS *frame_parms,
short coef)
{
static int max_pos_fil = 0;
static int count_max_pos_ok = 0;
static int first_time = 1;
int max_val = 0, max_pos = 0;
@@ -55,13 +54,14 @@ void nr_adjust_synch_ue(NR_DL_FRAME_PARMS *frame_parms,
LOG_D(PHY,"AbsSubframe %d: rx_offset (before) = %d\n",subframe,ue->rx_offset);
// we only use channel estimates from tx antenna 0 here
for (int i = 0; i < frame_parms->nb_prefix_samples; i++) {
// search for maximum position within the cyclic prefix
for (int i = -frame_parms->nb_prefix_samples/2; i < frame_parms->nb_prefix_samples/2; i++) {
int temp = 0;
int j = (i < 0) ? (i + frame_parms->ofdm_symbol_size) : i;
for (int aa = 0; aa < frame_parms->nb_antennas_rx; aa++) {
int Re = ((int16_t*)ue->pbch_vars[gNB_id]->dl_ch_estimates_time[aa])[(i<<1)];
int Im = ((int16_t*)ue->pbch_vars[gNB_id]->dl_ch_estimates_time[aa])[1+(i<<1)];
int Re = ((int16_t*)ue->pbch_vars[gNB_id]->dl_ch_estimates_time[aa])[(j<<1)];
int Im = ((int16_t*)ue->pbch_vars[gNB_id]->dl_ch_estimates_time[aa])[1+(j<<1)];
temp += (Re*Re/2) + (Im*Im/2);
}
@@ -71,19 +71,16 @@ void nr_adjust_synch_ue(NR_DL_FRAME_PARMS *frame_parms,
}
}
if (max_pos > frame_parms->ofdm_symbol_size/2)
max_pos = max_pos - frame_parms->ofdm_symbol_size;
// filter position to reduce jitter
if (clear == 1)
max_pos_fil = max_pos;
ue->max_pos_fil = max_pos;
else
max_pos_fil = ((max_pos_fil * coef) + (max_pos * ncoef)) >> 15;
ue->max_pos_fil = ((ue->max_pos_fil * coef) + (max_pos * ncoef)) >> 15;
// do not filter to have proactive timing adjustment
//max_pos_fil = max_pos;
//ue->max_pos_fil = max_pos;
int diff = max_pos_fil - sync_pos;
int diff = ue->max_pos_fil - sync_pos;
if (frame_parms->freq_range==nr_FR2)
sync_offset = 2;
@@ -131,7 +128,7 @@ void nr_adjust_synch_ue(NR_DL_FRAME_PARMS *frame_parms,
ue->rx_offset,
clear,
max_pos,
max_pos_fil,
ue->max_pos_fil,
max_val,
sync_pos);
#endif //DEBUG_PHY
openair1/PHY/NR_UE_ESTIMATION/nr_dl_channel_estimation.c
+2 -2
View File
@@ -485,8 +485,8 @@ int nr_pdcch_channel_estimation(PHY_VARS_NR_UE *ue,
#ifdef DEBUG_PDCCH
printf("PDCCH Channel Estimation : ThreadId %d, gNB_id %d ch_offset %d, OFDM size %d, Ncp=%d, Ns=%d, k=%d symbol %d\n",proc->thread_id, gNB_id,ch_offset,ue->frame_parms.ofdm_symbol_size,
ue->frame_parms.Ncp,Ns,k, symbol);
printf("PDCCH Channel Estimation : ThreadId %d, gNB_id %d ch_offset %d, OFDM size %d, Ncp=%d, Ns=%d, symbol %d\n",
proc->thread_id, gNB_id,ch_offset,ue->frame_parms.ofdm_symbol_size,ue->frame_parms.Ncp,Ns,symbol);
#endif
fl = filt16a_l1;
openair1/PHY/NR_UE_TRANSPORT/dci_nr.c
+21 -920
View File
File diff suppressed because it is too large Load Diff
openair1/PHY/NR_UE_TRANSPORT/dci_nr.h
-6
View File
@@ -33,13 +33,7 @@
#ifndef __PHY_NR_UE_TRANSPORT_DCI_NR__H__
#define __PHY_NR_UE_TRANSPORT_DCI_NR__H__
#ifndef USER_MODE
#include "PHY/types.h"
#else
#include <stdint.h>
#endif
#define MAX_DCI_SIZE_BITS 45
openair1/PHY/NR_UE_TRANSPORT/nr_dlsch_decoding.c
+309 -1027
View File
File diff suppressed because it is too large Load Diff
openair1/PHY/NR_UE_TRANSPORT/nr_dlsch_demodulation.c
-6
View File
@@ -43,12 +43,6 @@
#include "PHY/NR_REFSIG/dmrs_nr.h"
#include "common/utils/nr/nr_common.h"
#ifndef USER_MODE
#define NOCYGWIN_STATIC static
#else
#define NOCYGWIN_STATIC
#endif
/* dynamic shift for LLR computation for TM3/4
* set as command line argument, see lte-softmodem.c
* default value: 0
openair1/PHY/NR_UE_TRANSPORT/nr_initial_sync.c
+3 -2
View File
@@ -38,6 +38,7 @@
#include "PHY/NR_UE_ESTIMATION/nr_estimation.h"
#include "SCHED_NR_UE/defs.h"
#include "common/utils/LOG/vcd_signal_dumper.h"
#include "common/utils/nr/nr_common.h"
#include "common_lib.h"
#include <math.h>
@@ -484,9 +485,9 @@ int nr_initial_sync(UE_nr_rxtx_proc_t *proc,
for(int n_ss = 0; n_ss<pdcch_vars->nb_search_space; n_ss++) {
uint8_t nb_symb_pdcch = pdcch_vars->pdcch_config[n_ss].coreset.duration;
int start_symb = pdcch_vars->pdcch_config[n_ss].coreset.StartSymbolIndex;
get_coreset_rballoc(pdcch_vars->pdcch_config[n_ss].coreset.frequency_domain_resource,&coreset_nb_rb,&coreset_start_rb);
for (uint16_t l=0; l<nb_symb_pdcch; l++) {
for (uint16_t l=start_symb; l<start_symb+nb_symb_pdcch; l++) {
nr_slot_fep_init_sync(ue,
proc,
l, // the UE PHY has no notion of the symbols to be monitored in the search space
openair1/PHY/NR_UE_TRANSPORT/nr_pbch.c
-2
View File
@@ -40,9 +40,7 @@
//#define DEBUG_PBCH
//#define DEBUG_PBCH_ENCODING
#ifdef OPENAIR2
//#include "PHY_INTERFACE/defs.h"
#endif
#define PBCH_A 24
openair1/PHY/NR_UE_TRANSPORT/nr_transport_proto_ue.h
-18
View File
@@ -1087,22 +1087,6 @@ uint8_t nr_ue_pusch_common_procedures(PHY_VARS_NR_UE *UE,
NR_DL_FRAME_PARMS *frame_parms,
uint8_t Nl);
uint32_t nr_dlsch_decoding_mthread(PHY_VARS_NR_UE *phy_vars_ue,
UE_nr_rxtx_proc_t *proc,
int eNB_id,
short *dlsch_llr,
NR_DL_FRAME_PARMS *frame_parms,
NR_UE_DLSCH_t *dlsch,
NR_DL_UE_HARQ_t *harq_process,
uint32_t frame,
uint16_t nb_symb_sch,
uint8_t nr_slot_rx,
uint8_t harq_pid,
uint8_t is_crnti,
uint8_t llr8_flag);
void *nr_dlsch_decoding_2thread0(void *arg);
void *nr_dlsch_decoding_2thread1(void *arg);
@@ -1750,8 +1734,6 @@ int nr_rx_pdsch(PHY_VARS_NR_UE *ue,
int32_t generate_nr_prach(PHY_VARS_NR_UE *ue, uint8_t gNB_id, uint8_t subframe);
void dump_nrdlsch(PHY_VARS_NR_UE *ue,uint8_t gNB_id,uint8_t nr_slot_rx,unsigned int *coded_bits_per_codeword,int round, unsigned char harq_pid);
void *dlsch_thread(void *arg);
/**@}*/
#endif
openair1/PHY/NR_UE_TRANSPORT/nr_ulsch_coding.c
+14 -12
View File
@@ -274,7 +274,7 @@ int nr_ulsch_encoding(PHY_VARS_NR_UE *ue,
#ifdef DEBUG_ULSCH_CODING
printf("encoding thinks this is a new packet \n");
#endif
harq_process->first_tx = 0;
harq_process->first_tx = 0;
///////////////////////// a---->| add CRC |---->b /////////////////////////
///////////
/*
@@ -445,17 +445,19 @@ int nr_ulsch_encoding(PHY_VARS_NR_UE *ue,
VCD_SIGNAL_DUMPER_DUMP_FUNCTION_BY_NAME(VCD_SIGNAL_DUMPER_FUNCTIONS_NR_RATE_MATCHING_LDPC, VCD_FUNCTION_IN);
start_meas(&ue->ulsch_rate_matching_stats);
nr_rate_matching_ldpc(Ilbrm,
Tbslbrm,
harq_process->BG,
*pz,
harq_process->d[r],
harq_process->e+r_offset,
harq_process->C,
F,
Kr-F-2*(*pz),
harq_process->pusch_pdu.pusch_data.rv_index,
E);
if (nr_rate_matching_ldpc(Ilbrm,
Tbslbrm,
harq_process->BG,
*pz,
harq_process->d[r],
harq_process->e+r_offset,
harq_process->C,
F,
Kr-F-2*(*pz),
harq_process->pusch_pdu.pusch_data.rv_index,
E) == -1)
return -1;
stop_meas(&ue->ulsch_rate_matching_stats);
VCD_SIGNAL_DUMPER_DUMP_FUNCTION_BY_NAME(VCD_SIGNAL_DUMPER_FUNCTIONS_NR_RATE_MATCHING_LDPC, VCD_FUNCTION_OUT);
openair1/PHY/NR_UE_TRANSPORT/nr_ulsch_ue.c
+2 -1
View File
@@ -162,7 +162,8 @@ void nr_ue_ulsch_procedures(PHY_VARS_NR_UE *UE,
nb_dmrs_re_per_rb, number_dmrs_symbols, mod_order, Nl);
nr_ulsch_encoding(UE, ulsch_ue, frame_parms, harq_pid, G);
if (nr_ulsch_encoding(UE, ulsch_ue, frame_parms, harq_pid, G) == -1)
return;
///////////
////////////////////////////////////////////////////////////////////
openair1/PHY/NR_UE_TRANSPORT/pucch_nr.c
+2 -2
View File
@@ -90,8 +90,8 @@ void nr_generate_pucch0(PHY_VARS_NR_UE *ue,
// the value of u,v (delta always 0 for PUCCH) has to be calculated according to TS 38.211 Subclause 6.3.2.2.1
uint8_t u[2]={0,0},v[2]={0,0};
LOG_D(PHY,"pucch0: slot %d nr_symbols %d, start_symbol %d, prb_start %d, second_hop_prb %d, group_hop_flag %d, sequence_hop_flag %d, mcs %d\n",nr_slot_tx,pucch_pdu->nr_of_symbols,pucch_pdu->start_symbol_index,pucch_pdu->prb_start,pucch_pdu->second_hop_prb,pucch_pdu->group_hop_flag,pucch_pdu->sequence_hop_flag,pucch_pdu->mcs);
LOG_D(PHY,"pucch0: slot %d nr_symbols %d, start_symbol %d, prb_start %d, second_hop_prb %d, group_hop_flag %d, sequence_hop_flag %d, mcs %d\n",
nr_slot_tx,pucch_pdu->nr_of_symbols,pucch_pdu->start_symbol_index,pucch_pdu->prb_start,pucch_pdu->second_hop_prb,pucch_pdu->group_hop_flag,pucch_pdu->sequence_hop_flag,pucch_pdu->mcs);
#ifdef DEBUG_NR_PUCCH_TX
printf("\t [nr_generate_pucch0] sequence generation: variable initialization for test\n");
openair1/PHY/TOOLS/lte_enb_scope.c
-16
View File
@@ -69,18 +69,12 @@ void reset_stats(FL_OBJECT *button, long arg) {
static void *scope_thread_eNB(void *arg) {
# ifdef ENABLE_XFORMS_WRITE_STATS
FILE *eNB_stats;
# endif
struct sched_param sched_param;
int UE_id, CC_id;
int ue_cnt=0;
sched_param.sched_priority = sched_get_priority_min(SCHED_FIFO)+1;
sched_setscheduler(0, SCHED_FIFO,&sched_param);
printf("Scope thread has priority %d\n",sched_param.sched_priority);
# ifdef ENABLE_XFORMS_WRITE_STATS
eNB_stats = fopen("eNB_stats.txt", "w");
#endif
while (!oai_exit) {
ue_cnt=0;
@@ -99,16 +93,6 @@ static void *scope_thread_eNB(void *arg) {
usleep(100*1000);
}
// printf("%s",stats_buffer);
# ifdef ENABLE_XFORMS_WRITE_STATS
if (eNB_stats) {
rewind (eNB_stats);
fwrite (stats_buffer, 1, len, eNB_stats);
fclose (eNB_stats);
}
# endif
pthread_exit((void *)arg);
}
openair1/PHY/TOOLS/lte_ue_scope.c
-16
View File
@@ -72,16 +72,10 @@ void reset_stats(FL_OBJECT *button, long arg) {
static void *scope_thread_UE(void *arg) {
char stats_buffer[16384];
# ifdef ENABLE_XFORMS_WRITE_STATS
FILE *UE_stats, *eNB_stats;
# endif
struct sched_param sched_param;
sched_param.sched_priority = sched_get_priority_min(SCHED_FIFO)+1;
sched_setscheduler(0, SCHED_FIFO,&sched_param);
printf("Scope thread has priority %d\n",sched_param.sched_priority);
# ifdef ENABLE_XFORMS_WRITE_STATS
UE_stats = fopen("UE_stats.txt", "w");
#endif
while (!oai_exit) {
// dump_ue_stats (PHY_vars_UE_g[0][0], &PHY_vars_UE_g[0][0]->proc.proc_rxtx[0],stats_buffer, 0, mode,rx_input_level_dBm);
@@ -94,16 +88,6 @@ static void *scope_thread_UE(void *arg) {
0,7);
// printf("%s",stats_buffer);
}
# ifdef ENABLE_XFORMS_WRITE_STATS
if (UE_stats) {
rewind (UE_stats);
fwrite (stats_buffer, 1, len, UE_stats);
fclose (UE_stats);
}
# endif
pthread_exit((void *)arg);
}
openair1/PHY/TOOLS/nr_phy_scope.c
-3
View File
@@ -517,9 +517,6 @@ void phy_scope_gNB(OAI_phy_scope_t *form,
}
static void *scope_thread_gNB(void *arg) {
scopeData_t *p=(scopeData_t *) arg;
//# ifdef ENABLE_XFORMS_WRITE_STATS
// FILE *gNB_stats = fopen("gNB_stats.txt", "w");
//#endif
size_t stksize=0;
pthread_attr_t atr;
pthread_attr_init(&atr);
openair1/PHY/TOOLS/time_meas.c
+43
View File
@@ -131,6 +131,49 @@ void print_meas(time_stats_t *ts,
}
}
int print_meas_log(time_stats_t *ts,
const char *name,
time_stats_t *total_exec_time,
time_stats_t *sf_exec_time,
char *output)
{
int stroff = 0;
static int first_time = 0;
static double cpu_freq_GHz = 0.0;
if (cpu_freq_GHz == 0.0)
cpu_freq_GHz = get_cpu_freq_GHz();
if (first_time == 0) {
first_time=1;
if ((total_exec_time == NULL) || (sf_exec_time== NULL))
stroff += sprintf(output, "%25s %25s %25s %25s %25s %6f\n","Name","Total","Per Trials", "Num Trials","CPU_F_GHz", cpu_freq_GHz);
else
stroff += sprintf(output+stroff, "%25s %25s %25s %20s %15s %6f\n","Name","Total","Average/Frame","Trials", "CPU_F_GHz", cpu_freq_GHz);
}
if (ts->trials>0) {
//printf("%20s: total: %10.3f ms, average: %10.3f us (%10d trials)\n", name, ts->diff/cpu_freq_GHz/1000000.0, ts->diff/ts->trials/cpu_freq_GHz/1000.0, ts->trials);
if ((total_exec_time == NULL) || (sf_exec_time== NULL)) {
stroff += sprintf(output+stroff, "%25s: %15.3f us; %15d; %15.3f us;\n",
name,
(ts->diff/ts->trials/cpu_freq_GHz/1000.0),
ts->trials,
ts->max/cpu_freq_GHz/1000.0);
} else {
stroff += sprintf(output+stroff, "%25s: %15.3f ms (%5.2f%%); %15.3f us (%5.2f%%); %15d;\n",
name,
(ts->diff/cpu_freq_GHz/1000000.0),
((ts->diff/cpu_freq_GHz/1000000.0)/(total_exec_time->diff/cpu_freq_GHz/1000000.0))*100, // percentage
(ts->diff/ts->trials/cpu_freq_GHz/1000.0),
((ts->diff/ts->trials/cpu_freq_GHz/1000.0)/(sf_exec_time->diff/sf_exec_time->trials/cpu_freq_GHz/1000.0))*100, // percentage
ts->trials);
}
}
return stroff;
}
double get_time_meas_us(time_stats_t *ts)
{
static double cpu_freq_GHz = 0.0;
openair1/PHY/TOOLS/time_meas.h
+1
View File
@@ -88,6 +88,7 @@ static inline void stop_meas(time_stats_t *ts) __attribute__((always_inline));
void print_meas_now(time_stats_t *ts, const char *name, FILE *file_name);
void print_meas(time_stats_t *ts, const char *name, time_stats_t *total_exec_time, time_stats_t *sf_exec_time);
int print_meas_log(time_stats_t *ts, const char *name, time_stats_t *total_exec_time, time_stats_t *sf_exec_time, char *output);
double get_time_meas_us(time_stats_t *ts);
double get_cpu_freq_GHz(void);
openair1/PHY/defs_L1_NB_IoT.h
-17
View File
@@ -48,7 +48,6 @@
#ifdef MEX
#define msg mexPrintf
#else
#ifdef OPENAIR2
#if ENABLE_RAL
#include "collection/hashtable/hashtable.h"
#include "COMMON/ral_messages_types.h"
@@ -56,9 +55,6 @@
#endif
#include "common/utils/LOG/log.h"
#define msg(aRGS...) LOG_D(PHY, ##aRGS)
#else
#define msg printf
#endif
#endif
//use msg in the real-time thread context
#define msg_nrt printf
@@ -196,19 +192,6 @@ typedef struct {
NB_IoT_UE_DLSCH_t *dlsch_rn_MCH[10];
} PHY_VARS_RN_NB_IoT;
/*
#ifdef OCP_FRAMEWORK
#include <enums.h>
#else
//typedef enum {normal_txrx=0,rx_calib_ue=1,rx_calib_ue_med=2,rx_calib_ue_byp=3,debug_prach=4,no_L2_connect=5,calib_prach_tx=6,rx_dump_frame=7,loop_through_memory=8} runmode_t;
*/
// enum transmission_access_mode {
// NO_ACCESS=0,
// POSTPONED_ACCESS,
// CANCELED_ACCESS,
// UNKNOWN_ACCESS,
// SCHEDULED_ACCESS,
// CBA_ACCESS};
typedef enum {
eNodeB_3GPP_NB_IoT=0, // classical eNodeB function
openair1/PHY/defs_UE.h
-4
View File
@@ -87,7 +87,6 @@
#define LOG_W(x, ...) mexPrintf(__VA_ARGS__)
#define LOG_M(x, ...) mexPrintf(__VA_ARGS__)
#else
#ifdef OPENAIR2
#if ENABLE_RAL
#include "collection/hashtable/hashtable.h"
#include "COMMON/ral_messages_types.h"
@@ -95,9 +94,6 @@
#endif
#include "common/utils/LOG/log.h"
#define msg(aRGS...) LOG_D(PHY, ##aRGS)
#else
#define msg printf
#endif
#endif
openair1/PHY/defs_nr_UE.h
+26 -7
View File
@@ -51,16 +51,12 @@
#ifdef MEX
#define msg mexPrintf
#else
#ifdef OPENAIR2
#if ENABLE_RAL
#include "common/utils/hashtable/hashtable.h"
#include "COMMON/ral_messages_types.h"
#include "UTIL/queue.h"
#endif
#define msg(aRGS...) LOG_D(PHY, ##aRGS)
#else
#define msg printf
#endif
#endif
//use msg in the real-time thread context
#define msg_nrt printf
@@ -638,7 +634,7 @@ typedef struct {
uint8_t dciFormat;
uint8_t agregationLevel;
int nb_search_space;
fapi_nr_dl_config_dci_dl_pdu_rel15_t pdcch_config[FAPI_NR_MAX_SS_PER_CORESET];
fapi_nr_dl_config_dci_dl_pdu_rel15_t pdcch_config[FAPI_NR_MAX_SS];
// frame and slot for sib1 in initial sync
uint16_t sfn;
uint16_t slot;
@@ -921,9 +917,10 @@ typedef struct {
uint8_t init_sync_frame;
/// temporary offset during cell search prior to MIB decoding
int ssb_offset;
uint16_t symbol_offset; // offset in terms of symbols for detected ssb in sync
int rx_offset; /// Timing offset
uint16_t symbol_offset; /// offset in terms of symbols for detected ssb in sync
int rx_offset; /// Timing offset
int rx_offset_diff; /// Timing adjustment for ofdm symbol0 on HW USRP
int max_pos_fil; /// Timing offset IIR filter
int time_sync_cell;
/// Timing Advance updates variables
@@ -1081,5 +1078,27 @@ typedef struct nr_rxtx_thread_data_s {
notifiedFIFO_t txFifo;
} nr_rxtx_thread_data_t;
typedef struct LDPCDecode_ue_s {
PHY_VARS_NR_UE *phy_vars_ue;
NR_DL_UE_HARQ_t *harq_process;
t_nrLDPC_dec_params decoderParms;
NR_UE_DLSCH_t *dlsch;
short* dlsch_llr;
int dlsch_id;
int harq_pid;
int rv_index;
int A;
int E;
int Kc;
int Qm;
int Kr_bytes;
int nbSegments;
int segment_r;
int r_offset;
int offset;
int Tbslbrm;
int decodeIterations;
} ldpcDecode_ue_t;
#include "SIMULATION/ETH_TRANSPORT/defs.h"
#endif
openair1/PHY/phy_extern.h
-6
View File
@@ -55,12 +55,6 @@ extern char mode_string[4][20];
extern unsigned char NB_RU;
#ifndef OPENAIR2
extern unsigned char NB_eNB_INST;
extern uint16_t NB_UE_INST;
extern unsigned char NB_RN_INST;
#endif
extern int flag_LA;
extern double sinr_bler_map[MCS_COUNT][2][MCS_TABLE_LENGTH_MAX];
extern double sinr_bler_map_up[MCS_COUNT][2][16];
openair1/PHY/phy_extern_nr_ue.h
-6
View File
@@ -63,12 +63,6 @@ extern char mode_string[4][20];
extern unsigned char NB_RU;
#ifndef OPENAIR2
extern unsigned char NB_eNB_INST;
extern uint16_t NB_UE_INST;
extern unsigned char NB_RN_INST;
#endif
extern int flag_LA;
extern double sinr_bler_map[MCS_COUNT][2][MCS_TABLE_LENGTH_MAX];
extern double sinr_bler_map_up[MCS_COUNT][2][16];
openair1/PHY/phy_extern_ue.h
-6
View File
@@ -53,12 +53,6 @@ extern int flagMag;
extern char mode_string[4][20];
#ifndef OPENAIR2
extern unsigned char NB_eNB_INST;
extern uint16_t NB_UE_INST;
extern unsigned char NB_RN_INST;
#endif
extern int flag_LA;
extern double sinr_bler_map[MCS_COUNT][2][MCS_TABLE_LENGTH_MAX];
extern double sinr_bler_map_up[MCS_COUNT][2][16];
openair1/PHY/phy_vars.h
-7
View File
@@ -55,13 +55,6 @@ const short conjugate2[8]__attribute__((aligned(16))) = {1,-1,1,-1,1,-1,1,-1};
unsigned char NB_RU=0;
#ifndef OPENAIR2
//unsigned char NB_eNB_INST=0;
//uint16_t NB_UE_INST=0;
//unsigned char NB_RN_INST=0;
//unsigned char NB_INST=0;
#endif
int number_of_cards;
openair1/PHY/phy_vars_nr_ue.h
-7
View File
@@ -57,13 +57,6 @@ const short conjugate2[8]__attribute__((aligned(16))) = {1,-1,1,-1,1,-1,1,-1};
unsigned char NB_RU=0;
#ifndef OPENAIR2
unsigned char NB_eNB_INST=0;
uint16_t NB_UE_INST=0;
unsigned char NB_RN_INST=0;
unsigned char NB_INST=0;
#endif
int number_of_cards;
openair1/PHY/phy_vars_ue.h
-8
View File
@@ -47,14 +47,6 @@ const short conjugate2[8]__attribute__((aligned(16))) = {1,-1,1,-1,1,-1,1,-1};
#include "SIMULATION/ETH_TRANSPORT/vars.h"
#ifndef OPENAIR2
unsigned char NB_eNB_INST=0;
uint16_t NB_UE_INST=0;
unsigned char NB_RN_INST=0;
unsigned char NB_INST=0;
#endif
int number_of_cards;
openair1/PHY/types_NB_IoT.h
-5
View File
@@ -22,11 +22,6 @@
#ifndef __openair_TYPES_NB_IOT_H__
#define __openair_TYPES_NB_IOT_H__
#ifdef USER_MODE
#include <stdint.h>
#else
#include <linux/types.h>
#endif
#endif /*__openair_TYPES_NB_IOT_H__ */
openair1/SCHED_NR/phy_procedures_nr_gNB.c
+6 -6
View File
@@ -403,8 +403,6 @@ void nr_fill_indication(PHY_VARS_gNB *gNB, int frame, int slot_rx, int ULSCH_id,
pthread_mutex_lock(&gNB->UL_INFO_mutex);
int timing_advance_update, cqi;
int sync_pos;
NR_gNB_ULSCH_t *ulsch = gNB->ulsch[ULSCH_id][0];
NR_UL_gNB_HARQ_t *harq_process = ulsch->harq_processes[harq_pid];
NR_gNB_SCH_STATS_t *stats=get_ulsch_stats(gNB,ulsch);
@@ -412,19 +410,20 @@ void nr_fill_indication(PHY_VARS_gNB *gNB, int frame, int slot_rx, int ULSCH_id,
nfapi_nr_pusch_pdu_t *pusch_pdu = &harq_process->ulsch_pdu;
// pdu->data = gNB->ulsch[ULSCH_id+1][0]->harq_processes[harq_pid]->b;
sync_pos = nr_est_timing_advance_pusch(gNB, ULSCH_id); // estimate timing advance for MAC
int sync_pos = nr_est_timing_advance_pusch(gNB, ULSCH_id); // estimate timing advance for MAC
// scale the 16 factor in N_TA calculation in 38.213 section 4.2 according to the used FFT size
uint16_t bw_scaling = 16 * gNB->frame_parms.ofdm_symbol_size / 2048;
int sync_pos_rounded;
// do some integer rounding to improve TA accuracy
int sync_pos_rounded;
if (sync_pos > 0)
sync_pos_rounded = sync_pos + (bw_scaling / 2) - 1;
else
sync_pos_rounded = sync_pos - (bw_scaling / 2) - 1;
sync_pos_rounded = sync_pos - (bw_scaling / 2) + 1;
if (stats) stats->sync_pos = sync_pos;
timing_advance_update = sync_pos_rounded / bw_scaling;
int timing_advance_update = sync_pos_rounded / bw_scaling;
// put timing advance command in 0..63 range
timing_advance_update += 31;
@@ -449,6 +448,7 @@ void nr_fill_indication(PHY_VARS_gNB *gNB, int frame, int slot_rx, int ULSCH_id,
LOG_D(PHY, "Estimated SNR for PUSCH is = %f dB (ulsch_power %f, noise %f)\n", SNRtimes10/10.0,dB_fixed_x10(gNB->pusch_vars[ULSCH_id]->ulsch_power_tot)/10.0,dB_fixed_x10(gNB->pusch_vars[ULSCH_id]->ulsch_noise_power_tot)/10.0);
int cqi;
if (SNRtimes10 < -640) cqi=0;
else if (SNRtimes10 > 635) cqi=255;
else cqi=(640+SNRtimes10)/5;
openair1/SCHED_NR_UE/fapi_nr_ue_l1.c
+19 -9
View File
@@ -154,6 +154,7 @@ int8_t nr_ue_scheduled_response(nr_scheduled_response_t *scheduled_response){
fapi_nr_ul_config_request_t *ul_config = scheduled_response->ul_config;
pthread_mutex_lock(&ul_config->mutex_ul_config);
for (i = 0; i < ul_config->number_pdus; ++i){
AssertFatal(ul_config->ul_config_list[i].pdu_type <= FAPI_NR_UL_CONFIG_TYPES,"pdu_type %d out of bounds\n",ul_config->ul_config_list[i].pdu_type);
@@ -175,6 +176,7 @@ int8_t nr_ue_scheduled_response(nr_scheduled_response_t *scheduled_response){
pusch_config_pdu = &ul_config->ul_config_list[i].pusch_config_pdu;
current_harq_pid = pusch_config_pdu->pusch_data.harq_process_id;
NR_UL_UE_HARQ_t *harq_process_ul_ue = ulsch0->harq_processes[current_harq_pid];
harq_process_ul_ue->status = 0;
if (harq_process_ul_ue){
@@ -184,14 +186,16 @@ int8_t nr_ue_scheduled_response(nr_scheduled_response_t *scheduled_response){
ulsch0->f_pusch = pusch_config_pdu->absolute_delta_PUSCH;
if (scheduled_response->tx_request){
fapi_nr_tx_request_body_t *tx_req_body = &scheduled_response->tx_request->tx_request_body[i];
LOG_D(PHY,"%d.%d Copying %d bytes to harq_process_ul_ue->a (harq_pid %d)\n",scheduled_response->frame,slot,tx_req_body->pdu_length,current_harq_pid);
memcpy(harq_process_ul_ue->a, tx_req_body->pdu, tx_req_body->pdu_length);
harq_process_ul_ue->status = ACTIVE;
scheduled_response->tx_request->number_of_pdus = 0;
if (scheduled_response->tx_request) {
for (int j=0; j<scheduled_response->tx_request->number_of_pdus; j++) {
fapi_nr_tx_request_body_t *tx_req_body = &scheduled_response->tx_request->tx_request_body[j];
if (tx_req_body->pdu_index == i) {
LOG_D(PHY,"%d.%d Copying %d bytes to harq_process_ul_ue->a (harq_pid %d)\n",scheduled_response->frame,slot,tx_req_body->pdu_length,current_harq_pid);
memcpy(harq_process_ul_ue->a, tx_req_body->pdu, tx_req_body->pdu_length);
harq_process_ul_ue->status = ACTIVE;
break;
}
}
}
} else {
@@ -229,7 +233,13 @@ int8_t nr_ue_scheduled_response(nr_scheduled_response_t *scheduled_response){
break;
}
}
memset(ul_config, 0, sizeof(fapi_nr_ul_config_request_t));
if (scheduled_response->tx_request)
scheduled_response->tx_request->number_of_pdus = 0;
ul_config->sfn = 0;
ul_config->slot = 0;
ul_config->number_pdus = 0;
memset(ul_config->ul_config_list, 0, sizeof(ul_config->ul_config_list));
pthread_mutex_unlock(&ul_config->mutex_ul_config);
}
}
return 0;
openair1/SCHED_NR_UE/phy_procedures_nr_ue.c
+31 -56
View File
@@ -768,7 +768,8 @@ int nr_ue_pdsch_procedures(PHY_VARS_NR_UE *ue, UE_nr_rxtx_proc_t *proc, int gNB_
uint16_t s1 = dlsch0_harq->nb_symbols;
bool is_SI = dlsch0->rnti_type == _SI_RNTI_;
LOG_D(PHY,"[UE %d] PDSCH type %d active in nr_slot_rx %d, harq_pid %d (%d), rb_start %d, nb_rb %d, symbol_start %d, nb_symbols %d, DMRS mask %x\n",ue->Mod_id,pdsch,nr_slot_rx,harq_pid,dlsch0_harq->status,pdsch_start_rb,pdsch_nb_rb,s0,s1,dlsch0_harq->dlDmrsSymbPos);
LOG_D(PHY,"[UE %d] PDSCH type %d active in nr_slot_rx %d, harq_pid %d (%d), rb_start %d, nb_rb %d, symbol_start %d, nb_symbols %d, DMRS mask %x\n",
ue->Mod_id,pdsch,nr_slot_rx,harq_pid,dlsch0_harq->status,pdsch_start_rb,pdsch_nb_rb,s0,s1,dlsch0_harq->dlDmrsSymbPos);
for (m = s0; m < (s0 +s1); m++) {
if (dlsch0_harq->dlDmrsSymbPos & (1 << m)) {
@@ -980,37 +981,23 @@ bool nr_ue_dlsch_procedures(PHY_VARS_NR_UE *ue,
start_meas(&ue->dlsch_decoding_stats[proc->thread_id]);
ret = nr_dlsch_decoding(ue,
proc,
gNB_id,
pdsch_vars->llr[0],
&ue->frame_parms,
dlsch0,
dlsch0->harq_processes[harq_pid],
frame_rx,
nb_symb_sch,
nr_slot_rx,
harq_pid,
pdsch==PDSCH?1:0,
dlsch0->harq_processes[harq_pid]->TBS>256?1:0);
if( dlsch_parallel) {
ret = nr_dlsch_decoding_mthread(ue,
proc,
gNB_id,
pdsch_vars->llr[0],
&ue->frame_parms,
dlsch0,
dlsch0->harq_processes[harq_pid],
frame_rx,
nb_symb_sch,
nr_slot_rx,
harq_pid,
pdsch==PDSCH?1:0,
dlsch0->harq_processes[harq_pid]->TBS>256?1:0);
LOG_T(PHY,"dlsch decoding is parallelized, ret = %d\n", ret);
}
else {
ret = nr_dlsch_decoding(ue,
proc,
gNB_id,
pdsch_vars->llr[0],
&ue->frame_parms,
dlsch0,
dlsch0->harq_processes[harq_pid],
frame_rx,
nb_symb_sch,
nr_slot_rx,
harq_pid,
pdsch==PDSCH?1:0,
dlsch0->harq_processes[harq_pid]->TBS>256?1:0);
LOG_T(PHY,"Sequential dlsch decoding , ret = %d\n", ret);
}
@@ -1082,36 +1069,24 @@ bool nr_ue_dlsch_procedures(PHY_VARS_NR_UE *ue,
start_meas(&ue->dlsch_decoding_stats[proc->thread_id]);
ret1 = nr_dlsch_decoding(ue,
proc,
gNB_id,
pdsch_vars->llr[1],
&ue->frame_parms,
dlsch1,
dlsch1->harq_processes[harq_pid],
frame_rx,
nb_symb_sch,
nr_slot_rx,
harq_pid,
pdsch==PDSCH?1:0,//proc->decoder_switch,
dlsch1->harq_processes[harq_pid]->TBS>256?1:0);
if(dlsch_parallel) {
ret1 = nr_dlsch_decoding_mthread(ue,
proc,
gNB_id,
pdsch_vars->llr[1],
&ue->frame_parms,
dlsch1,
dlsch1->harq_processes[harq_pid],
frame_rx,
nb_symb_sch,
nr_slot_rx,
harq_pid,
pdsch==PDSCH?1:0,
dlsch1->harq_processes[harq_pid]->TBS>256?1:0);
LOG_T(PHY,"CW dlsch decoding is parallelized, ret1 = %d\n", ret1);
}
else {
ret1 = nr_dlsch_decoding(ue,
proc,
gNB_id,
pdsch_vars->llr[1],
&ue->frame_parms,
dlsch1,
dlsch1->harq_processes[harq_pid],
frame_rx,
nb_symb_sch,
nr_slot_rx,
harq_pid,
pdsch==PDSCH?1:0,//proc->decoder_switch,
dlsch1->harq_processes[harq_pid]->TBS>256?1:0);
LOG_T(PHY,"CWW sequential dlsch decoding, ret1 = %d\n", ret1);
}
@@ -1733,8 +1708,8 @@ int phy_procedures_nrUE_RX(PHY_VARS_NR_UE *ue,
for (uint16_t l=0; l<nb_symb_pdcch; l++) {
// note: this only works if RBs for PDCCH are contigous!
LOG_D(PHY, "pdcch_channel_estimation: first_carrier_offset %d, BWPStart %d, coreset_start_rb %d\n",
fp->first_carrier_offset, pdcch_vars->pdcch_config[n_ss].BWPStart, coreset_start_rb);
LOG_D(PHY, "pdcch_channel_estimation: first_carrier_offset %d, BWPStart %d, coreset_start_rb %d, coreset_nb_rb %d\n",
fp->first_carrier_offset, pdcch_vars->pdcch_config[n_ss].BWPStart, coreset_start_rb, coreset_nb_rb);
if (coreset_nb_rb > 0)
nr_pdcch_channel_estimation(ue,
openair1/SIMULATION/LTE_PHY/gpib_send.c
+4 -4
View File
@@ -19,10 +19,10 @@
* contact@openairinterface.org
*/
#include<stdio.h>
#include<string.h>
#include<gpib/ib.h>
#include"gpib_send.h"
#include <stdio.h>
#include <string.h>
#include <gpib/ib.h>
#include "gpib_send.h"
void gpib_send(unsigned int gpib_board, unsigned int gpib_device, char *command_string )
{
unsigned short addlist[2] = {gpib_device, NOADDR};
openair1/SIMULATION/NR_PHY/dlschsim.c
+6 -3
View File
@@ -109,6 +109,7 @@ int main(int argc, char **argv)
int ret;
//int run_initial_sync=0;
int loglvl = OAILOG_WARNING;
uint8_t dlsch_threads = 0;
float target_error_rate = 0.01;
uint64_t SSB_positions=0x01;
uint16_t nb_symb_sch = 12;
@@ -138,9 +139,9 @@ int main(int argc, char **argv)
break;*/
/*case 'd':
frame_type = 1;
break;*/
case 'd':
dlsch_threads = atoi(optarg);
break;
case 'g':
switch ((char) *optarg) {
@@ -320,6 +321,7 @@ 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 Output filename (.txt format) for Pe/SNR results\n");
printf("-F Input filename (.txt format) for RX conformance testing\n");
printf("-d number of dlsch threads, 0: no dlsch parallelization\n");
exit(-1);
break;
}
@@ -331,6 +333,7 @@ int main(int argc, char **argv)
if (snr1set == 0)
snr1 = snr0 + 10;
init_dlsch_tpool(dlsch_threads);
if (ouput_vcd)
vcd_signal_dumper_init("/tmp/openair_dump_nr_dlschsim.vcd");
openair1/SIMULATION/NR_PHY/dlsim.c
+2 -6
View File
@@ -688,7 +688,7 @@ int main(int argc, char **argv)
if (snr1set==0)
snr1 = snr0+10;
init_dlsch_tpool(dlsch_threads);
RC.gNB = (PHY_VARS_gNB**) malloc(sizeof(PHY_VARS_gNB *));
@@ -980,11 +980,7 @@ int main(int argc, char **argv)
reset_meas(&msgDataTx->phy_proc_tx);
gNB->phy_proc_tx_0 = &msgDataTx->phy_proc_tx;
pushTpool(gNB->threadPool,msgL1Tx);
if (dlsch_threads ) {
init_dlsch_tpool(dlsch_threads);
pthread_t dlsch0_threads;
threadCreate(&dlsch0_threads, dlsch_thread, (void *)UE, "DLthread", -1, OAI_PRIORITY_RT_MAX-1);
}
for (SNR = snr0; SNR < snr1; SNR += .2) {
varArray_t *table_tx=initVarArray(1000,sizeof(double));
openair1/SIMULATION/NR_PHY/pucchsim.c
+10 -5
View File
@@ -662,14 +662,19 @@ int main(int argc, char **argv)
if (uci_pdu.sr->sr_indication == 0 || uci_pdu.sr->sr_confidence_level == 1)
sr_errors+=1;
}
// harq value 0 -> pass
// confidence value 0 -> good confidence
const int harq_value0 = uci_pdu.harq->harq_list[0].harq_value;
const int harq_value1 = uci_pdu.harq->harq_list[1].harq_value;
const int confidence_lvl = uci_pdu.harq->harq_confidence_level;
if(nr_bit>0){
if (nr_bit==1 && do_DTX == 0)
ack_nack_errors+=(actual_payload^uci_pdu.harq->harq_list[0].harq_value);
ack_nack_errors+=(actual_payload^(!harq_value0));
else if (do_DTX == 0)
ack_nack_errors+=(((actual_payload&1)^uci_pdu.harq->harq_list[0].harq_value)+((actual_payload>>1)^uci_pdu.harq->harq_list[1].harq_value));
else if ((uci_pdu.harq->harq_confidence_level == 0 && uci_pdu.harq->harq_list[0].harq_value == 1) ||
(uci_pdu.harq->harq_confidence_level == 0 && nr_bit == 2 && uci_pdu.harq->harq_list[1].harq_value==1))
ack_nack_errors++;
ack_nack_errors+=(((actual_payload&1)^(!harq_value0))+((actual_payload>>1)^(!harq_value1)));
else if ((!confidence_lvl && !harq_value0) ||
(!confidence_lvl && nr_bit == 2 && !harq_value1))
ack_nack_errors++;
free(uci_pdu.harq->harq_list);
}
}