adding LDPC encoder and decoder (from mwc18 branch)
This commit is contained in:
Florian Kaltenberger
@@ -1064,12 +1064,18 @@ set(PHY_TURBOSRC
|
||||
${OPENAIR1_DIR}/PHY/CODING/3gpplte_turbo_decoder_sse_16bit.c
|
||||
${OPENAIR1_DIR}/PHY/CODING/3gpplte_turbo_decoder_avx2_16bit.c
|
||||
${OPENAIR1_DIR}/PHY/CODING/3gpplte_turbo_decoder.c
|
||||
)
|
||||
set(PHY_LDPCSRC
|
||||
${OPENAIR1_DIR}/PHY/CODING/nrLDPC_decoder/nrLDPC_decoder.c
|
||||
${OPENAIR1_DIR}/PHY/CODING/nrLDPC_encoder/ldpc_encoder.c
|
||||
${OPENAIR1_DIR}/PHY/CODING/nrLDPC_encoder/ldpc_encoder2.c
|
||||
${OPENAIR1_DIR}/PHY/CODING/nrLDPC_encoder/ldpc_generate_coefficient.c
|
||||
)
|
||||
set(PHY_TURBOIF
|
||||
${OPENAIR1_DIR}/PHY/CODING/coding_load.c
|
||||
)
|
||||
|
||||
add_library(coding MODULE ${PHY_TURBOSRC} )
|
||||
add_library(coding MODULE ${PHY_TURBOSRC} ${PHY_LDPCSRC})
|
||||
set(PHY_SRC
|
||||
# depend on code generation from asn1c
|
||||
${RRC_FULL_DIR}/asn1_constants.h
|
||||
|
||||
@@ -0,0 +1,355 @@
|
||||
/*
|
||||
* 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
|
||||
*/
|
||||
|
||||
#include <stdlib.h>
|
||||
#include <math.h>
|
||||
#include <stdio.h>
|
||||
#include <string.h>
|
||||
|
||||
#include "SIMULATION/TOOLS/defs.h"
|
||||
|
||||
|
||||
|
||||
// 4-bit quantizer
|
||||
char quantize4bit(double D,double x)
|
||||
{
|
||||
double qxd;
|
||||
qxd = floor(x/D);
|
||||
// printf("x=%f,qxd=%f\n",x,qxd);
|
||||
|
||||
if (qxd <= -8)
|
||||
qxd = -8;
|
||||
else if (qxd > 7)
|
||||
qxd = 7;
|
||||
|
||||
return((char)qxd);
|
||||
}
|
||||
|
||||
char quantize(double D,double x,unsigned char B)
|
||||
{
|
||||
double qxd;
|
||||
char maxlev;
|
||||
qxd = floor(x/D);
|
||||
// printf("x=%f,qxd=%f\n",x,qxd);
|
||||
maxlev = 1<<(B-1);
|
||||
|
||||
if (qxd <= -maxlev)
|
||||
qxd = -maxlev;
|
||||
else if (qxd >= maxlev)
|
||||
qxd = maxlev-1;
|
||||
|
||||
return((char)qxd);
|
||||
}
|
||||
|
||||
#define MAX_BLOCK_LENGTH 8448
|
||||
|
||||
int test_ldpc(short No_iteration,
|
||||
int nom_rate,
|
||||
int denom_rate,
|
||||
double SNR,
|
||||
unsigned char qbits,
|
||||
short block_length,
|
||||
unsigned int ntrials,
|
||||
unsigned int *errors,
|
||||
unsigned int *crc_misses)
|
||||
{
|
||||
//clock initiate
|
||||
time_stats_t time,time_optim,tinput,tprep,tparity,toutput;
|
||||
opp_enabled=1;
|
||||
cpu_freq_GHz = get_cpu_freq_GHz();
|
||||
//short test_input[block_length];
|
||||
unsigned char *test_input[MAX_NUM_DLSCH_SEGMENTS]={NULL,NULL,NULL,NULL,NULL,NULL,NULL,NULL};;
|
||||
//short *c; //padded codeword
|
||||
unsigned char *estimated_output;
|
||||
unsigned char *channel_input[MAX_NUM_DLSCH_SEGMENTS];
|
||||
unsigned char *channel_input_optim[MAX_NUM_DLSCH_SEGMENTS];
|
||||
double *channel_output;
|
||||
double *modulated_input;
|
||||
char *channel_output_fixed;
|
||||
unsigned int i,j,trial=0;
|
||||
short BG,Zc,Kb,nrows,ncols;
|
||||
int no_punctured_columns,removed_bit;
|
||||
int i1;
|
||||
//Table of possible lifting sizes
|
||||
short lift_size[51]= {2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,18,20,22,24,26,28,30,32,36,40,44,48,52,56,60,64,72,80,88,96,104,112,120,128,144,160,176,192,208,224,240,256,288,320,352,384};
|
||||
int n_segments=8;
|
||||
|
||||
t_nrLDPC_dec_params decParams;
|
||||
int n_iter;
|
||||
|
||||
*errors=0;
|
||||
*crc_misses=0;
|
||||
|
||||
// generate input block
|
||||
for(j=0;j<MAX_NUM_DLSCH_SEGMENTS;j++) {
|
||||
test_input[j]=(unsigned char *)malloc16(sizeof(unsigned char) * block_length/8);
|
||||
channel_input[j] = (unsigned char *)malloc16(sizeof(unsigned char) * 68*384);
|
||||
channel_input_optim[j] = (unsigned char *)malloc16(sizeof(unsigned char) * 68*384);
|
||||
}
|
||||
modulated_input = (double *)malloc(sizeof(double) * 68*384);
|
||||
channel_output = (double *)malloc(sizeof(double) * 68*384);
|
||||
channel_output_fixed = (char *)malloc16(sizeof(char) * 68*384);
|
||||
estimated_output = (unsigned char*) malloc16(sizeof(unsigned char) * block_length/8);
|
||||
|
||||
reset_meas(&time);
|
||||
reset_meas(&time_optim);
|
||||
reset_meas(&tinput);
|
||||
reset_meas(&tprep);
|
||||
reset_meas(&tparity);
|
||||
reset_meas(&toutput);
|
||||
|
||||
for (j=0;j<MAX_NUM_DLSCH_SEGMENTS;j++) {
|
||||
for (i=0; i<block_length/8; i++) {
|
||||
test_input[j][i]=(unsigned char) rand();
|
||||
//test_input[j][i]=j%256;
|
||||
}
|
||||
}
|
||||
|
||||
//determine number of bits in codeword
|
||||
//if (block_length>3840)
|
||||
{
|
||||
BG=1;
|
||||
Kb = 22;
|
||||
nrows=46; //parity check bits
|
||||
ncols=22; //info bits
|
||||
}
|
||||
/*else if (block_length<=3840)
|
||||
{
|
||||
BG=2;
|
||||
nrows=42; //parity check bits
|
||||
ncols=10; // info bits
|
||||
|
||||
if (block_length>640)
|
||||
Kb = 10;
|
||||
else if (block_length>560)
|
||||
Kb = 9;
|
||||
else if (block_length>192)
|
||||
Kb = 8;
|
||||
else
|
||||
Kb = 6;
|
||||
}*/
|
||||
|
||||
//find minimum value in all sets of lifting size
|
||||
Zc=0;
|
||||
|
||||
for (i1=0; i1 < 51; i1++)
|
||||
{
|
||||
if (lift_size[i1] >= (double) block_length/Kb)
|
||||
{
|
||||
Zc = lift_size[i1];
|
||||
//printf("%d\n",Zc);
|
||||
break;
|
||||
}
|
||||
}
|
||||
|
||||
printf("ldpc_test: block_length %d, BG %d, Zc %d, Kb %d\n",block_length,BG, Zc, Kb);
|
||||
no_punctured_columns=(int)((nrows-2)*Zc+block_length-block_length*(1/((float)nom_rate/(float)denom_rate)))/Zc;
|
||||
// printf("puncture:%d\n",no_punctured_columns);
|
||||
removed_bit=(nrows-no_punctured_columns-2) * Zc+block_length-(int)(block_length/((float)nom_rate/(float)denom_rate));
|
||||
if (ntrials==0)
|
||||
ldpc_encoder_orig(test_input[0],channel_input[0], block_length, nom_rate, denom_rate, 1);
|
||||
|
||||
for (trial=0; trial < ntrials; trial++)
|
||||
{
|
||||
|
||||
//// encoder
|
||||
start_meas(&time);
|
||||
for(j=0;j<n_segments;j++) {
|
||||
ldpc_encoder_orig(test_input[j], channel_input[j],block_length,nom_rate,denom_rate,0);
|
||||
}
|
||||
stop_meas(&time);
|
||||
|
||||
start_meas(&time_optim);
|
||||
ldpc_encoder_optim_8seg(test_input,channel_input_optim,block_length,nom_rate,denom_rate,n_segments,&tinput,&tprep,&tparity,&toutput);
|
||||
/*for(j=0;j<n_segments;j++) {
|
||||
ldpc_encoder_optim(test_input[j],channel_input_optim[j],block_length,nom_rate,denom_rate,&tinput,&tprep,&tparity,&toutput);
|
||||
}*/
|
||||
stop_meas(&time_optim);
|
||||
|
||||
if (ntrials==1)
|
||||
for (j=0;j<n_segments;j++)
|
||||
for (i = 0; i < block_length+(nrows-no_punctured_columns) * Zc - removed_bit; i++)
|
||||
if (channel_input[j][i]!=channel_input_optim[j][i]) {
|
||||
printf("differ in seg %d pos %d (%d,%d)\n",j,i,channel_input[j][i],channel_input_optim[j][i]);
|
||||
return (-1);
|
||||
}
|
||||
|
||||
//print_meas_now(&time, "", stdout);
|
||||
|
||||
// for (i=0;i<6400;i++)
|
||||
//printf("channel_input[%d]=%d\n",i,channel_input[i]);
|
||||
//printf("%d ",channel_input[i]);
|
||||
|
||||
//if ((BG==2) && (Zc==128||Zc==256))
|
||||
if (0)
|
||||
{
|
||||
for (i = 2*Zc; i < (Kb+nrows-no_punctured_columns) * Zc-removed_bit; i++)
|
||||
{
|
||||
#ifdef DEBUG_CODER
|
||||
if ((i&0xf)==0)
|
||||
printf("\ne %d..%d: ",i,i+15);
|
||||
#endif
|
||||
|
||||
if (channel_input[0][i-2*Zc]==0)
|
||||
modulated_input[i]=1/sqrt(2); //QPSK
|
||||
else
|
||||
modulated_input[i]=-1/sqrt(2);
|
||||
|
||||
channel_output[i] = modulated_input[i] + gaussdouble(0.0,1.0) * 1/sqrt(2*SNR);
|
||||
channel_output_fixed[i] = (char) ((channel_output[i]*128)<0?(channel_output[i]*128-0.5):(channel_output[i]*128+0.5)); //fixed point 9-7
|
||||
//printf("llr[%d]=%d\n",i,channel_output_fixed[i]);
|
||||
}
|
||||
|
||||
//for (i=(Kb+nrows) * Zc-5;i<(Kb+nrows) * Zc;i++)
|
||||
//{
|
||||
// printf("channel_input[%d]=%d\n",i,channel_input[i]);
|
||||
//printf("%lf %d\n",channel_output[i], channel_output_fixed[i]);
|
||||
//printf("v[%d]=%lf\n",i,modulated_input[i]);}
|
||||
#ifdef DEBUG_CODER
|
||||
printf("\n");
|
||||
exit(-1);
|
||||
#endif
|
||||
|
||||
decParams.BG=BG;
|
||||
decParams.Z=Zc;
|
||||
decParams.R=13;
|
||||
decParams.numMaxIter=6;
|
||||
decParams.outMode = nrLDPC_outMode_BIT;
|
||||
|
||||
// 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);
|
||||
|
||||
//for (i=(Kb+nrows) * Zc-5;i<(Kb+nrows) * Zc;i++)
|
||||
// printf("esimated_output[%d]=%d\n",i,esimated_output[i]);
|
||||
|
||||
//count errors
|
||||
for (i=0; i<block_length>>3; i++)
|
||||
{
|
||||
if (estimated_output[i] != test_input[0][i])
|
||||
{
|
||||
//printf("error pos %d (%d, %d)\n",i,estimated_output[i],test_input[i]);
|
||||
*errors = (*errors) + 1;
|
||||
break;
|
||||
}
|
||||
}
|
||||
|
||||
}
|
||||
/*else if (trial==0)
|
||||
printf("decoder is not supported\n");*/
|
||||
}
|
||||
|
||||
for(j=0;j<MAX_NUM_DLSCH_SEGMENTS;j++) {
|
||||
free(test_input[j]);
|
||||
free(channel_input[j]);
|
||||
free(channel_input_optim[j]);
|
||||
}
|
||||
free(modulated_input);
|
||||
free(channel_output);
|
||||
free(channel_output_fixed);
|
||||
free(estimated_output);
|
||||
|
||||
print_meas(&time,"ldpc_encoder",NULL,NULL);
|
||||
print_meas(&time_optim,"ldpc_encoder_optim",NULL,NULL);
|
||||
print_meas(&tinput,"ldpc_encoder_optim(input)",NULL,NULL);
|
||||
print_meas(&tprep,"ldpc_encoder_optim(prep)",NULL,NULL);
|
||||
print_meas(&tparity,"ldpc_encoder_optim(parity)",NULL,NULL);
|
||||
print_meas(&toutput,"ldpc_encoder_optim(output)",NULL,NULL);
|
||||
return *errors;
|
||||
}
|
||||
|
||||
int main(int argc, char *argv[])
|
||||
{
|
||||
unsigned int errors,crc_misses;
|
||||
short block_length=576; // decoder supports length: 1201 -> 1280, 2401 -> 2560
|
||||
short No_iteration=25;
|
||||
//double rate=0.333;
|
||||
int nom_rate=1;
|
||||
int denom_rate=3;
|
||||
double SNR0=-2.0,SNR,SNR_lin;
|
||||
unsigned char qbits=4;
|
||||
unsigned int decoded_errors[100]; // initiate the size of matrix equivalent to size of SNR
|
||||
int c,i=0;
|
||||
|
||||
int n_trials = 1;
|
||||
|
||||
randominit(0);
|
||||
|
||||
while ((c = getopt (argc, argv, "q:r:s:l:n:d:")) != -1)
|
||||
switch (c)
|
||||
{
|
||||
case 'q':
|
||||
qbits = atoi(optarg);
|
||||
break;
|
||||
|
||||
case 'r':
|
||||
nom_rate = atoi(optarg);
|
||||
break;
|
||||
|
||||
case 'd':
|
||||
denom_rate = atoi(optarg);
|
||||
break;
|
||||
|
||||
case 'l':
|
||||
block_length = atoi(optarg);
|
||||
break;
|
||||
|
||||
case 'n':
|
||||
n_trials = atoi(optarg);
|
||||
break;
|
||||
|
||||
case 's':
|
||||
SNR0 = atoi(optarg);
|
||||
break;
|
||||
|
||||
default:
|
||||
abort ();
|
||||
}
|
||||
|
||||
printf("the decoder supports BG2, Kb=10, Z=128 & 256\n");
|
||||
printf(" range of blocklength: 1201 -> 1280, 2401 -> 2560\n");
|
||||
printf("block length %d: \n", block_length);
|
||||
printf("rate: %d/%d\n",nom_rate,denom_rate);
|
||||
|
||||
//for (block_length=8;block_length<=MAX_BLOCK_LENGTH;block_length+=8)
|
||||
for (SNR=SNR0;SNR<SNR0+1.0;SNR+=1.0)
|
||||
{
|
||||
SNR_lin = pow(10,SNR/10);
|
||||
decoded_errors[i]=test_ldpc(No_iteration,
|
||||
nom_rate,
|
||||
denom_rate,
|
||||
SNR_lin, // noise standard deviation
|
||||
qbits,
|
||||
block_length, // block length bytes
|
||||
n_trials,
|
||||
&errors,
|
||||
&crc_misses);
|
||||
printf("SNR %f, BLER %f (%d/%d)\n",SNR,(float)decoded_errors[i]/(float)n_trials,decoded_errors[i],n_trials);
|
||||
i=i+1;
|
||||
}
|
||||
|
||||
return(0);
|
||||
}
|
||||
|
||||
|
||||
|
||||
Binary file not shown.
|
After Width: | Height: | Size: 3.6 KiB |
Binary file not shown.
@@ -0,0 +1,844 @@
|
||||
%
|
||||
% 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
|
||||
%
|
||||
|
||||
\documentclass{article}
|
||||
|
||||
\usepackage[a4paper, total={6in, 8in}]{geometry}
|
||||
|
||||
\usepackage{amsmath}
|
||||
\usepackage{amsfonts}
|
||||
\usepackage{amssymb}
|
||||
\usepackage{booktabs}
|
||||
\usepackage{url}
|
||||
\usepackage{tcolorbox}
|
||||
|
||||
\usepackage{tikz}
|
||||
\usetikzlibrary{arrows,decorations,shapes,backgrounds,patterns}
|
||||
\usepackage{pgfplots}
|
||||
\pgfplotsset{compat=newest}
|
||||
\definecolor{green}{RGB}{32,127,43}
|
||||
\usetikzlibrary{calc}
|
||||
|
||||
\usepackage{listings}
|
||||
\lstdefinestyle{customc}{
|
||||
belowcaptionskip=1\baselineskip,
|
||||
breaklines=true,
|
||||
frame=L,
|
||||
xleftmargin=\parindent,
|
||||
language=C,
|
||||
showstringspaces=false,
|
||||
basicstyle=\footnotesize\ttfamily,
|
||||
keywordstyle=\bfseries\color{green!40!black},
|
||||
commentstyle=\itshape\color{purple!40!black},
|
||||
identifierstyle=\color{blue},
|
||||
stringstyle=\color{orange},
|
||||
}
|
||||
\lstset{escapechar=@,style=customc}
|
||||
|
||||
\title{NR LDPC Decoder}
|
||||
\author{Sebastian Wagner (TCL)}
|
||||
\date{\today}
|
||||
|
||||
\def\0{\mathbf{0}}
|
||||
\def\b{\mathbf{b}}
|
||||
\def\Bbb{\mathbb{B}}
|
||||
\def\Bcal{\mathcal{B}}
|
||||
\def\c{\mathbf{c}}
|
||||
\def\C{\mathbf{C}}
|
||||
\def\Cbb{\mathbb{C}}
|
||||
\def\Ccal{\mathcal{C}}
|
||||
\def\eqdef{\triangleq}
|
||||
\def\g{\mathbf{g}}
|
||||
\def\G{\mathbf{G}}
|
||||
\def\Gcal{\mathcal{G}}
|
||||
\def\h{\mathbf{h}}
|
||||
\def\H{\mathbf{H}}
|
||||
\def\Hbg{\mathbf{H}_\mathrm{BG}}
|
||||
\def\Hbgo{\mathbf{H}_\mathrm{BG1}}
|
||||
\def\Hbgt{\mathbf{H}_\mathrm{BG2}}
|
||||
\def\I{\mathbf{I}}
|
||||
\def\Kb{{K_b}}
|
||||
\def\m{\mathbf{m}}
|
||||
\def\Mb{{M_b}}
|
||||
\def\Nb{{N_b}}
|
||||
\def\Nbb{\mathbb{N}}
|
||||
\def\n{\mathbf{n}}
|
||||
\def\nr{{n_{\rm r}}}
|
||||
\def\nt{{n_{\rm t}}}
|
||||
\def\s{\mathbf{s}}
|
||||
\def\SNR{\mathsf{SNR}}
|
||||
\def\y{\mathbf{y}}
|
||||
\def\z{\mathbf{z}}
|
||||
\def\Z{\mathbf{Z}}
|
||||
\def\Zc{{Z_c}}
|
||||
|
||||
|
||||
\def\herm{\mathsf{H}}
|
||||
\def\trans{\mathsf{T}}
|
||||
\def\EE{\mathsf{E}}
|
||||
\newcommand{\sgn}{\operatorname{sgn}}
|
||||
|
||||
\begin{document}
|
||||
|
||||
\maketitle
|
||||
|
||||
\begin{tikzpicture}[remember picture,overlay]
|
||||
\node[anchor=north west,inner sep=0pt] at (current page.north west)
|
||||
{\includegraphics[scale=0.5]{logo.png}};
|
||||
\end{tikzpicture}
|
||||
|
||||
|
||||
\begin{center}Currently Supported:\end{center}
|
||||
\tcbox[center]{
|
||||
\begin{tabular}{lll}
|
||||
\toprule
|
||||
\textbf{BG} & \textbf{Lifting Size Z} & \textbf{Code Rate R} \\
|
||||
\midrule
|
||||
1 & all & 1/3, 2/3, 8/9 \\
|
||||
2 & all & 1/5, 1/3, 2/3 \\
|
||||
\bottomrule
|
||||
\end{tabular}
|
||||
}
|
||||
|
||||
\tableofcontents
|
||||
|
||||
\newpage
|
||||
\section{Introduction}
|
||||
\label{sec:introduction}
|
||||
|
||||
Low Density Parity Check (LDPC) codes have been developed by Gallager in 1963 \cite{gallager1962low}. They are linear error correcting codes that are capacity-achieving for large block length and are completely described by their Parity Check Matrix (PCM) $\H^{M\times N}$. The PCM $\H$ defines $M$ constraints on the codeword $\c$ of length $N$ such that
|
||||
\begin{equation}
|
||||
\label{eq:29}
|
||||
\H\c = \0.
|
||||
\end{equation}
|
||||
The number of information bits $B$ that can be encoded with $\H$ is given by $B=N-M$. Hence the code rate $R$ of $\H$ reads
|
||||
\begin{equation}
|
||||
\label{eq:37}
|
||||
R = \frac{B}{N} = 1-\frac{M}{N}.
|
||||
\end{equation}
|
||||
|
||||
|
||||
\subsection{LDPC in NR}
|
||||
\label{sec:ldpc-nr}
|
||||
|
||||
NR uses quasi-cyclic (QC) Protograph LDPC codes, i.e. a smaller graph, called Base Graph (BG), is defined and utilized to construct the larger PCM. This has the advantage that the large PCM does not have to be stored in memory and allows for a more efficient implementation while maintaining good decoding properties.
|
||||
Two BGs $\Hbg\in\Nbb^{\Mb\times \Nb}$ are defined in NR:
|
||||
\begin{enumerate}
|
||||
\item $\Hbgo\in\Nbb^{46\times 68}$
|
||||
\item $\Hbgt\in\Nbb^{42\times 52}$
|
||||
\end{enumerate}
|
||||
where $\Nbb$ is the set of integers. For instance the first 3 rows and 13 columns of BG2 are given by
|
||||
|
||||
\setcounter{MaxMatrixCols}{30}
|
||||
\begin{equation*}
|
||||
\label{eq:33}
|
||||
\Hbgt =
|
||||
\begin{bmatrix}
|
||||
9 & 117 & 204 & 26 & \emptyset & \emptyset & 189 & \emptyset & \emptyset & 205 & 0 & 0 & \emptyset & \emptyset \\
|
||||
127 & \emptyset & \emptyset & 166 & 253 & 125 & 226 & 156 & 224 & 252 & \emptyset & 0 & 0 & \emptyset \\
|
||||
81 & 114 & \emptyset & 44 & 52 & \emptyset & \emptyset & \emptyset & 240 & \emptyset & 1 & \emptyset & 0 & 0
|
||||
\end{bmatrix}.
|
||||
\end{equation*}
|
||||
|
||||
To obtain the PCM $\H$ from the BG $\Hbg$, each element $\Hbg(i,j)$ in the BG is replaced by a lifting matrix of size $\Zc\times \Zc$ according to
|
||||
\begin{equation}
|
||||
\label{eq:35}
|
||||
\Hbg(i,j) =
|
||||
\begin{cases}
|
||||
\0 & \textrm{if}~ \Hbg(i,j)=\emptyset \\
|
||||
\I_{P_{ij}} & \textrm{otherwise}
|
||||
\end{cases}
|
||||
\end{equation}
|
||||
where $\I_{P_{ij}}$ is the identity matrix circularly shifted to the right by $P_{ij} = \Hbg(i,j)\mod \Zc$. Hence, the resulting PCM $\H$ will be of size $\Mb\Zc\times\Nb\Zc$.
|
||||
|
||||
The lifting size $\Zc$ depends on the number of bits to encode. To limit the complexity, a discrete set $\mathcal{Z}$ of possible values of $\Zc$ has been defined in \cite{3gpp2017_38212} and the optimal value $\Zc$ is calculated according to
|
||||
\begin{equation}
|
||||
\label{eq:36}
|
||||
\Zc = \min_{\Z\in\mathcal{Z}}\left[Z\geq\frac{B}{\Nb}\right].
|
||||
\end{equation}
|
||||
|
||||
The base rate of the two BGs is $1/3$ and $1/5$ for BG1 and BG2, respectively. That is, BG1 encodes $K=22\Zc$ bits and BG2 encodes $K=10\Zc$ bits. Note that the first 2 columns of BG 1 and 2 are always punctured, that is after encoding, the first $2\Zc$ bits are discarded and not transmitted.
|
||||
For instance, consider $B=500$ information bits to encode using BG2, \eqref{eq:36} yields $\Zc=64$ hence $K=640$. Since $K>B$, $K-B=140$ filler bits are appended to the information bits. The PCM $\Hbgt$ is of size $2688\times 3328$ and the $640$ bits $\b$ are encoded according to \eqref{eq:29} at a rate $R \approx 0.192$. To achieve the higher base rate of $0.2$, the first $128$ are punctured, i.e. instead of transmitting all $3328$ bits, only $3200$ are transmitted resulting in the desired rate $R=640/3200=0.2$.
|
||||
|
||||
\subsection{LDPC Decoding}
|
||||
\label{sec:ldpc-decoding}
|
||||
|
||||
The decoding of codeword $\c$ can be achieved via the classical message passing algorithm. This algorithm can be illustrated best using the Tanner graph of the PCM. The rows of the PCM are called check nodes (CN) since they represent the parity check equations. The parity check equation of each of these check nodes involves various bits in the codeword. Similarly, every column of the PCM corresponds to a bit and each bit is involved in several parity check equations. In the Tanner graph representation, the bits are called bit nodes (BN). Let's go back to the previous example of BG2 and assume $\Zc=2$, hence the first 3 rows and 13 columns of BG2 $\Hbgt$ read
|
||||
\begin{equation*}
|
||||
\label{eq:36}
|
||||
\Hbgt =
|
||||
\begin{bmatrix}
|
||||
1 & 1 & 0 & 0 & \emptyset & \emptyset & 1 & \emptyset & \emptyset & 1 & 0 & 0 & \emptyset & \emptyset \\
|
||||
1 & \emptyset & \emptyset & 0 & 1 & 1 & 0 & 0 & 0 & 0 & \emptyset & 0 & 0 & \emptyset \\
|
||||
1 & 0 & \emptyset & 0 & 0 & \emptyset & \emptyset & \emptyset & 0 & \emptyset & 1 & \emptyset & 0 & 0
|
||||
\end{bmatrix}.
|
||||
\end{equation*}
|
||||
Replacing the elements according to \eqref{eq:35}, we obtain the first 6 rows and 26 columns of the PCM as
|
||||
\begin{equation*}
|
||||
\label{eq:39}
|
||||
\H =
|
||||
\begin{bmatrix}
|
||||
0 & 1 & 0 & 1 & 1 & 0 & 1 & 0 & 0 & 0 & 0 & 0 & 0 & 1 & 0 & 0 & 0 & 0 & 0 & 1 & 1 & 0 & 1 & 0 & 0 & 0 & 0 & 0\\
|
||||
1 & 0 & 1 & 0 & 0 & 1 & 0 & 1 & 0 & 0 & 0 & 0 & 1 & 0 & 0 & 0 & 0 & 0 & 1 & 0 & 0 & 1 & 0 & 1 & 0 & 0 & 0 & 0\\
|
||||
0 & 1 & 0 & 0 & 0 & 0 & 1 & 0 & 0 & 1 & 0 & 1 & 1 & 0 & 1 & 0 & 1 & 0 & 1 & 0 & 0 & 0 & 1 & 0 & 1 & 0 & 0 & 0\\
|
||||
1 & 0 & 0 & 0 & 0 & 0 & 0 & 1 & 1 & 0 & 1 & 0 & 0 & 1 & 0 & 1 & 0 & 1 & 0 & 1 & 0 & 0 & 0 & 1 & 0 & 1 & 0 & 0\\
|
||||
0 & 1 & 1 & 0 & 0 & 0 & 1 & 0 & 1 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 1 & 0 & 0 & 0 & 0 & 1 & 0 & 0 & 1 & 0 & 1 & 0\\
|
||||
1 & 0 & 0 & 1 & 0 & 0 & 0 & 1 & 0 & 1 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 1 & 0 & 0 & 1 & 0 & 0 & 0 & 0 & 1 & 0 & 1
|
||||
\end{bmatrix}.
|
||||
\end{equation*}
|
||||
|
||||
The Tanner graph of the first 8 BNs is shown in Figure \ref{fig:tannergraph}.
|
||||
|
||||
\begin{figure}[ht]
|
||||
\label{fig:tannergraph}
|
||||
\centering
|
||||
\def\ww{0.3cm}
|
||||
\def\hh{0.3cm}
|
||||
\tikzstyle{cnode}=[fill=white,rectangle,draw=black,thick,inner sep=2pt, minimum height=\hh,minimum width=\ww, rounded corners=1pt,text width=\ww]
|
||||
\tikzstyle{vnode}=[fill=white,circle,draw=black,thick,inner sep=2pt, minimum height=\hh,minimum width=\ww, rounded corners=1pt,text width=\ww]
|
||||
\tikzstyle{connector}=[<->,>=latex',semithick]
|
||||
|
||||
\begin{tikzpicture}
|
||||
\tikzstyle{every node}=[node distance=1.5cm,text centered]
|
||||
% Check nodes
|
||||
\node[cnode, label=above:$v_0$] (v0) {};
|
||||
\node[cnode, label=above:$v_1$, right of=v0] (v1) {};
|
||||
\node[cnode, label=above:$v_2$, right of=v1] (v2) {};
|
||||
% Variable nodes
|
||||
\node[vnode, label=below:$c_3$, below of=v1, node distance=1.5cm] (c3) {};
|
||||
\node[vnode, label=below:$c_2$, left of=c3, node distance=1.5cm] (c2) {};
|
||||
\node[vnode, label=below:$c_1$, left of=c2, node distance=1.5cm] (c1) {};
|
||||
\node[vnode, label=below:$c_0$, left of=c1, node distance=1.5cm] (c0) {};
|
||||
\node[vnode, label=below:$c_4$, right of=c3, node distance=1.5cm] (c4) {};
|
||||
\node[vnode, label=below:$c_5$, right of=c4, node distance=1.5cm] (c5) {};
|
||||
\node[vnode, label=below:$c_6$, right of=c5, node distance=1.5cm] (c6) {};
|
||||
|
||||
% Draw edges
|
||||
\draw (c0) edge[connector] (v1);
|
||||
\draw (c1) edge[connector] (v0);
|
||||
\draw (c1) edge[connector] (v2);
|
||||
\draw (c2) edge[connector] (v1);
|
||||
\draw (c3) edge[connector] (v0);
|
||||
\draw (c4) edge[connector] (v0);
|
||||
\draw (c4) edge[connector] (v2);
|
||||
\draw (c5) edge[connector] (v1);
|
||||
\draw (c6) edge[connector] (v0);
|
||||
\draw (c6) edge[connector] (v2);
|
||||
|
||||
\end{tikzpicture}
|
||||
|
||||
\caption{Tanner graph for first 7 bits nodes and 3 check nodes from \eqref{eq:39}.}
|
||||
\end{figure}
|
||||
|
||||
The message passing algorithm is an iterative algorithm where probabilities of the bits (being either 0 or 1) are exchanged between the BNs and CNs. After sufficient iterations, the probabilities will have either converged to either 0 or 1 and the parity check equations will be satisfied, at this point, the codeword has been decoded correctly.
|
||||
|
||||
\newpage
|
||||
\section{LDPC Decoder Implementation}
|
||||
\label{sec:ldpc-implementation}
|
||||
|
||||
The implementation on a general purpose processor (GPP) has to take advantage of potential instruction extension of the processor architecture. We focus on the Intel x86 instruction set architecture (ISA) and its advanced vector extension (AVX). In particular, we utilize AVX2 with its 256-bit single instruction multiple data (SIMD) format. In order to utilize AVX2 to speed up the processing at the CNs and BNs, the corresponding data has to be ordered/aligned in a specific way. The processing flow of the LDPC decoder is depicted in \ref{fig:ldpc_decoder_flow}.
|
||||
|
||||
\begin{figure}[ht]
|
||||
\label{fig:ldpc_decoder_flow}
|
||||
\centering
|
||||
\def\ww{0.3cm}
|
||||
\def\hh{0.3cm}
|
||||
\tikzstyle{func}=[,draw=none]
|
||||
\tikzstyle{connector}=[->,>=latex',semithick]
|
||||
|
||||
\begin{tikzpicture}
|
||||
\tikzstyle{every node}=[node distance=2.5cm,text centered]
|
||||
% Check nodes
|
||||
% First iteration
|
||||
\node[func] (llr2llrProcBuf) {\texttt{llr2llrProcBuf}};
|
||||
\node[func, above of=llr2llrProcBuf] (llr2CnProcBuf) {\texttt{llr2CnProcBuf}};
|
||||
\node[func, above right of=llr2CnProcBuf] (cnProc1) {\texttt{cnProc}};
|
||||
\node[func, below right of=cnProc1] (cn2bnProcBuf1) {\texttt{cn2bnProcBuf}};
|
||||
\node[func, below of=cn2bnProcBuf1] (bnProcPc1) {\texttt{bnProcPc}};
|
||||
|
||||
% Iterations
|
||||
\node[func, right of=cnProc1, node distance=7cm] (cnProc) {\texttt{cnProc}};
|
||||
\node[func, below right of=cnProc] (cn2bnProcBuf) {\texttt{cn2bnProcBuf}};
|
||||
\node[func, below of=cn2bnProcBuf] (bnProcPc) {\texttt{bnProcPc}};
|
||||
\node[func, below left of=cnProc] (bn2cnProcBuf) {\texttt{bn2cnProcBuf}};
|
||||
\node[func, below of=bn2cnProcBuf] (bnProc) {\texttt{bnProc}};
|
||||
|
||||
% Post processing
|
||||
\node[func, below of=bnProcPc] (llrRes2llrOut) {\texttt{llrRes2llrOut}};
|
||||
\node[func, below of=llrRes2llrOut, node distance=1cm] (llr2bit) {\texttt{llr2bit}};
|
||||
|
||||
% Draw edges
|
||||
\draw (llr2llrProcBuf) edge[connector] (llr2CnProcBuf);
|
||||
\draw (llr2CnProcBuf) edge[connector] (cnProc1);
|
||||
\draw (cnProc1) edge[connector] (cn2bnProcBuf1);
|
||||
\draw (cn2bnProcBuf1) edge[connector] (bnProcPc1);
|
||||
|
||||
\draw (bnProcPc) edge[connector] (bnProc);
|
||||
\draw (bnProc) edge[connector] (bn2cnProcBuf);
|
||||
\draw (bn2cnProcBuf) edge[connector] (cnProc);
|
||||
\draw (cnProc) edge[connector] (cn2bnProcBuf);
|
||||
\draw (cn2bnProcBuf) edge[connector] (bnProcPc);
|
||||
|
||||
\draw (bnProcPc1) edge[connector] (bnProc);
|
||||
|
||||
\draw (bnProcPc) edge[connector] node[left] {iterations done} (llrRes2llrOut);
|
||||
\draw (llrRes2llrOut) edge[connector] (llr2bit);
|
||||
|
||||
% Boxes
|
||||
\node[inner sep=0pt,above right of=cn2bnProcBuf1, node distance = 2.5cm] (ref) {};
|
||||
|
||||
\draw[fill=black,opacity=.2, rounded corners] (llr2llrProcBuf.south west) rectangle ($(ref) + (-.5cm,.5cm)$);
|
||||
\draw[fill=black,opacity=.2, rounded corners] ($(ref) + (.5cm,.5cm)$) rectangle ($(bnProcPc.south east) + (.4cm,0)$);
|
||||
|
||||
\node[func, above of=cnProc1, node distance=.8cm] (iter1) {\textbf{First Iteration}};
|
||||
\node[func, above of=cnProc , node distance=.8cm] (iterX) {\textbf{Subsequent Iterations}};
|
||||
|
||||
\end{tikzpicture}
|
||||
|
||||
\caption{LDPC Decoder processing flow.}
|
||||
\end{figure}
|
||||
|
||||
The functions involved are described in more detail in table \ref{tab:sum_func}.
|
||||
|
||||
\begin{table}[ht]
|
||||
\centering
|
||||
\begin{tabular}{ll}
|
||||
\toprule
|
||||
\textbf{Function} & \textbf{Description} \\
|
||||
\midrule
|
||||
\texttt{llr2llrProcBuf} & Copies input LLRs to LLR processing buffer \\
|
||||
\texttt{llr2CnProcBuf} & Copies input LLRs to CN processing buffer \\
|
||||
\texttt{cnProc} & Performs CN signal processing \\
|
||||
\texttt{cn2bnProcBuf} & Copies the CN results to the BN processing buffer \\
|
||||
\texttt{bnProcPc} & Performs BN processing for parity check and/or hard-decision \\
|
||||
\texttt{bnProc} & Utilizes the results of \texttt{bnProcPc} to compute LLRs for CN processing \\
|
||||
\texttt{bn2cnProcBuf} & Copies the BN results to the CN processing buffer \\
|
||||
\texttt{llrRes2llrOut} & Copies the results of \texttt{bnProcPc} to output LLRs \\
|
||||
\texttt{llr2bit} & Performs hard-decision on the output LLRs \\
|
||||
\bottomrule
|
||||
\end{tabular}
|
||||
\caption{Summary of the LDPC decoder functions.}
|
||||
\label{tab:sum_func}
|
||||
\end{table}
|
||||
|
||||
The input LLRs are assumed to be 8-bit and aligned on 32 bytes. CN processing is carried out in 8-bit whereas BN processing is done in 16 bit. Subsequently, the processing tasks at the CNs and BNs are explained in more detail.
|
||||
|
||||
\newpage
|
||||
\subsection{Check Node Processing}
|
||||
\label{sec:check-node-proc}
|
||||
|
||||
Denote $q_{ij}$ the value from BN $j$ to CN $i$ and let $\Bcal_i$ be the set of connected BNs to the $i$th CN. Then, using the min-sum approximation, CN $i$ has to carry out the following operation for each connected BN.
|
||||
\begin{equation}
|
||||
\label{eq:40}
|
||||
r_{ji} = \prod_{j'\in\Bcal_i\setminus j}\sgn q_{ij'}\min_{j'\in\Bcal_i\setminus j} |q_{ij'}|
|
||||
\end{equation}
|
||||
where $r_{ji}$ is the value returned to BN $j$ from CN $i$. There are $\Mb = \{46,42\}$ CNs in BG 1 and BG 2, respectively. Each of these CNs is connected to only a small number of BNs. The number of connected BNs to CN $i$ is $|\Bcal_i|$. In BG1 and BG2, $|\Bcal_i|=\{3,4,5,6,7,8,9,10,19\}$ and $|\Bcal_i|=\{3,4,5,6,8,10\}$, respectively. The following tables show the number of CNs $M_{|\Bcal_i|}$ that are connected to the same number of BNs.
|
||||
|
||||
\begin{table}[ht]
|
||||
\centering
|
||||
\begin{tabular}{llllllllll}
|
||||
\toprule
|
||||
$|\Bcal_i|$ & 3 & 4 & 5 & 6 & 7 & 8 & 9 & 10 & 19 \\
|
||||
\midrule
|
||||
$M_{|\Bcal_i|}^\mathrm{BG1}$ & 1 & 5 &18 & 8 & 5 & 2 & 2 & 1 & 4 \\
|
||||
$M_{|\Bcal_i|}^\mathrm{BG2}$ & 6 & 20 & 9 & 3 & 0 & 2 & 0 & 2 & 0 \\
|
||||
\bottomrule
|
||||
\end{tabular}
|
||||
\caption{Ceck node groups for BG1 and BG2.}
|
||||
\label{tab:checkNodeGroups}
|
||||
\end{table}
|
||||
|
||||
It can be observed that each CN is at least connected to 3 BNs and there are 9 groups and 5 groups in BG1 and BG2, respectively. Denote the set of CN groups as $\Gcal$ and $M_k$ the number of CNs in group $k\in\Gcal$, e.g. for BG2 $M_4=20\Zc$. Each CN group will be processed separately. The CN processing buffer $p_C^k$ of group $k$ is defined as
|
||||
\begin{equation}
|
||||
\label{eq:44}
|
||||
p_C^k = \{\underbrace{q_{11}q_{21}\dots q_{M_k 1}}_{\text 1. BN},\underbrace{q_{12}q_{22}\dots q_{M_k 2}}_{\text 2. BN},\dots,\underbrace{q_{12}q_{22}\dots q_{M_k k}}_{\text last BN}\}
|
||||
\end{equation}
|
||||
Hence, $|p_C^k| = kM_k$, e.g, $\Zc=128$, $|p_C^4| = 4\cdot 20\cdot 128 = 10240$.
|
||||
|
||||
\begin{lstlisting}[frame=single,caption={Example of CN processing for group 3 from \texttt{cnProc}.},label=code_cnproc] % Start your code-block
|
||||
|
||||
const uint8_t lut_idxCnProcG3[3][2] = {{72,144}, {0,144}, {0,72}};
|
||||
|
||||
// =====================================================================
|
||||
// Process group with 3 BNs
|
||||
|
||||
// Number of groups of 32 CNs for parallel processing
|
||||
M = (lut_numCnInCnGroups[0]*Z)>>5;
|
||||
// Set the offset to each bit within a group in terms of 32 Byte
|
||||
bitOffsetInGroup = (lut_numCnInCnGroups_BG2_R15[0]*NR_LDPC_ZMAX)>>5;
|
||||
|
||||
// Set pointers to start of group 3
|
||||
p_cnProcBuf = (__m256i*) &cnProcBuf [lut_startAddrCnGroups[0]];
|
||||
p_cnProcBufRes = (__m256i*) &cnProcBufRes[lut_startAddrCnGroups[0]];
|
||||
|
||||
// Loop over every BN
|
||||
for (j=0; j<3; j++)
|
||||
{
|
||||
// Set of results pointer to correct BN address
|
||||
p_cnProcBufResBit = p_cnProcBufRes + (j*bitOffsetInGroup);
|
||||
|
||||
// Loop over CNs
|
||||
for (i=0; i<M; i++)
|
||||
{
|
||||
// Abs and sign of 32 CNs (first BN)
|
||||
ymm0 = p_cnProcBuf[lut_idxCnProcG3[j][0] + i];
|
||||
sgn = _mm256_sign_epi8(*p_ones, ymm0);
|
||||
min = _mm256_abs_epi8(ymm0);
|
||||
|
||||
// 32 CNs of second BN
|
||||
ymm0 = p_cnProcBuf[lut_idxCnProcG3[j][1] + i];
|
||||
min = _mm256_min_epu8(min, _mm256_abs_epi8(ymm0));
|
||||
sgn = _mm256_sign_epi8(sgn, ymm0);
|
||||
|
||||
// Store result
|
||||
min = _mm256_min_epu8(min, *p_maxLLR); // 128 in epi8 is -127
|
||||
*p_cnProcBufResBit = _mm256_sign_epi8(min, sgn);
|
||||
p_cnProcBufResBit++;
|
||||
}
|
||||
}
|
||||
|
||||
}
|
||||
\end{lstlisting}
|
||||
|
||||
Once all results of the check node processing $r_{ji}$ have been calculated, they are copied to the bit node processing buffer.
|
||||
|
||||
\subsection{Bit Node Processing}
|
||||
\label{sec:bit-node-processing}
|
||||
|
||||
Denote $r_{ji}$ the value from CN $i$ to BN $j$ and let $\Ccal_j$ be the set of connected CNs to the $j$th BN. Each BN $j$ has to carry out the following operation for every connected CN $i\in\Ccal_j$.
|
||||
\begin{equation}
|
||||
\label{eq:46}
|
||||
q_{ij} = \Lambda_j + \sum_{i'\in\Ccal_j\setminus i}r_{ji'}
|
||||
\end{equation}
|
||||
|
||||
There are $\Nb = \{68,52\}$ BNs in BG 1 and BG 2, respectively. Each of these BNs is connected to only a small number of CNs. The number of connected CNs to BN $j$ is $|\Ccal_j|$. In BG1 and BG2, $|\Ccal_j|=\{1,4,7,8,9,10,11,12,28,30\}$ and $|\Ccal_j|=\{1,5,6,7,8,9,10,12,13,14,16,22,23\}$, respectively. The following tables show the number of BNs $K_{|\Ccal_j|}$ that are connected to the same number of CNs.
|
||||
|
||||
\begin{table}[ht]
|
||||
\centering
|
||||
\begin{tabular}{lllllllllllllllllll}
|
||||
\toprule
|
||||
$|\Ccal_j|$ & 1&4&5&6&7&8&9&10&11&12&13 & 14 & 15 & 16 & 22 & 23 &28&30 \\
|
||||
\midrule
|
||||
$K_{|\Ccal_j|}^\mathrm{BG1}$ & 42 & 1 & 1 & 2 & 4 & 3 & 1 & 4 & 3 & 4 & 1 & 0 & 0 & 0 & 0 & 0 & 1 & 1 \\
|
||||
$K_{|\Ccal_j|}^\mathrm{BG2}$ & 38 & 0 & 2 & 1 & 1 & 1 & 2 & 1 & 0 & 1 & 1 & 1 & 0 & 1 & 1 & 1 & 0 & 0\\
|
||||
\bottomrule
|
||||
\end{tabular}
|
||||
\caption{Bit node groups for BG1 and BG2 for base rates 1/3 and 1/5, respectively.}
|
||||
\label{tab:bitNodeGroups}
|
||||
\end{table}
|
||||
|
||||
The BNs that are connected to a single CN do not need to be considered in the BN processing since \eqref{eq:46} yields $q_{ij} = \Lambda_j$. It can be observed that the grouping is less compact, i.e. there are many groups with only a small number of elements.
|
||||
|
||||
Denote the set of BN groups as $\Bcal$ and $K_k$ the number of BNs in group $k\in\Bcal$, e.g. for BG2 $K_5=2\Zc$. Each BN group will be processed separately. The BN processing buffer $p_B^k$ of group $k$ is defined as
|
||||
\begin{equation}
|
||||
\label{eq:47}
|
||||
p_B^k = \{\underbrace{r_{11}r_{21}\dots r_{K_k 1}}_{\text 1. CN},\underbrace{r_{12}r_{22}\dots r_{K_k 2}}_{\text 2. CN},\dots,\underbrace{r_{12}r_{22}\dots r_{K_k k}}_{\text last CN}\}
|
||||
\end{equation}
|
||||
Hence, $|p_B^k| = kK_k$, e.g, $\Zc=128$, $|p_B^5| = 5\cdot 2\cdot 128 = 1024$.
|
||||
|
||||
Depending on the code rate, some parity bits are not being transmitted. For instance, for BG2 with code rate $R = 1/3$ the last $20\Zc$ bits are discarded. Therefore, the last 20 columns or the last $20\Zc$ parity check equation are not required for decoding. This means that the BN groups shown in table \ref{tab:bitNodeGroups} are depending on the rate.
|
||||
|
||||
\begin{lstlisting}[frame=single,caption={Example of BN processing for group 3 from \texttt{bnProcPc}.},label=code_bnproc] % Start your code-block
|
||||
|
||||
// If elements in group move to next address
|
||||
idxBnGroup++;
|
||||
|
||||
// Number of groups of 32 BNs for parallel processing
|
||||
M = (lut_numBnInBnGroups[2]*Z)>>5;
|
||||
|
||||
// Set the offset to each CN within a group in terms of 16 Byte
|
||||
cnOffsetInGroup = (lut_numBnInBnGroups[2]*NR_LDPC_ZMAX)>>4;
|
||||
|
||||
// Set pointers to start of group 3
|
||||
p_bnProcBuf = (__m128i*) &bnProcBuf [lut_startAddrBnGroups [idxBnGroup]];
|
||||
p_llrProcBuf = (__m128i*) &llrProcBuf [lut_startAddrBnGroupsLlr[idxBnGroup]];
|
||||
p_llrRes = (__m256i*) &llrRes [lut_startAddrBnGroupsLlr[idxBnGroup]];
|
||||
|
||||
// Loop over BNs
|
||||
for (i=0,j=0; i<M; i++,j+=2)
|
||||
{
|
||||
// First 16 LLRs of first CN
|
||||
ymmRes0 = _mm256_cvtepi8_epi16(p_bnProcBuf[j]);
|
||||
ymmRes1 = _mm256_cvtepi8_epi16(p_bnProcBuf[j+1]);
|
||||
|
||||
// Loop over CNs
|
||||
for (k=1; k<3; k++)
|
||||
{
|
||||
ymm0 = _mm256_cvtepi8_epi16(p_bnProcBuf[k*cnOffsetInGroup + j]);
|
||||
ymmRes0 = _mm256_adds_epi16(ymmRes0, ymm0);
|
||||
|
||||
ymm1 = _mm256_cvtepi8_epi16(p_bnProcBuf[k*cnOffsetInGroup + j+1]);
|
||||
ymmRes1 = _mm256_adds_epi16(ymmRes1, ymm1);
|
||||
}
|
||||
|
||||
// Add LLR from receiver input
|
||||
ymm0 = _mm256_cvtepi8_epi16(p_llrProcBuf[j]);
|
||||
ymmRes0 = _mm256_adds_epi16(ymmRes0, ymm0);
|
||||
|
||||
ymm1 = _mm256_cvtepi8_epi16(p_llrProcBuf[j+1]);
|
||||
ymmRes1 = _mm256_adds_epi16(ymmRes1, ymm1);
|
||||
|
||||
// Pack results back to epi8
|
||||
ymm0 = _mm256_packs_epi16(ymmRes0, ymmRes1);
|
||||
// ymm0 = [ymmRes1[255:128] ymmRes0[255:128] ymmRes1[127:0] ymmRes0[127:0]]
|
||||
// p_llrRes = [ymmRes1[255:128] ymmRes1[127:0] ymmRes0[255:128] ymmRes0[127:0]]
|
||||
*p_llrRes = _mm256_permute4x64_epi64(ymm0, 0xD8);
|
||||
|
||||
// Next result
|
||||
p_llrRes++;
|
||||
}
|
||||
}
|
||||
|
||||
\end{lstlisting}
|
||||
|
||||
The sum of the LLRs is carried out in 16 bit for accuracy and is then saturated to 8 bit for CN processing. Saturation after each addition results in significant loss of sensitivity for low code rates.
|
||||
|
||||
\subsection{Mapping to the Processing Buffers}
|
||||
\label{sec:mapp-cn-proc}
|
||||
|
||||
For efficient processing with the AVX instructions, the data is required to be aligned in a certain manner. That is the reason why processing buffers have been introduced. The drawback is that the results of the processing need to copied every time to the processing buffer of the next task. However, the speed up in computation with AVX more than makes up for the time wasted in copying data. The copying is implemented using look-up tables (LUTs) which are described in table \ref{tab:sum_lut}.
|
||||
|
||||
\begin{table}[ht]
|
||||
\centering
|
||||
\begin{tabular}{ll}
|
||||
\toprule
|
||||
\textbf{LUT} & \textbf{Description} \\
|
||||
\midrule
|
||||
\texttt{lut\_llr2llrProcBuf\_BGX\_ZX\_RX} & Indices for function \texttt{llr2llrProcBuf} \\
|
||||
\texttt{lut\_llr2CnProcBuf\_BGX\_ZX\_RX} & Indices for function \texttt{llr2CnProcBuf} \\
|
||||
\texttt{lut\_cn2bnProcBuf\_BGX\_ZX\_RX} & Indices for functions \texttt{cn2bnProcBuf} and \texttt{bn2cnProcBuf} \\
|
||||
\bottomrule
|
||||
\end{tabular}
|
||||
\caption{Summary of the LUTs.}
|
||||
\label{tab:sum_lut}
|
||||
\end{table}
|
||||
|
||||
These LUTs are depending on the BG, the lifting size and the code rate. Assuming 5 rates for BG2 and 7 rates for BG1, the total number of LUTs is 617.
|
||||
|
||||
\newpage
|
||||
\section{Performance Results}
|
||||
\label{sec:performance-results}
|
||||
|
||||
In this section, the performance in terms of BLER and decoding latency of the current LDPC decoder implementation is verified.
|
||||
|
||||
\subsection{BLER Performance}
|
||||
\label{sec:bler-performance}
|
||||
|
||||
In all simulations, we assume AWGN, QPSK modulation and 8-bit input LLRs. The results are averaged over at least $10\,000$ channel realizations.
|
||||
|
||||
The first set of simulations in Figure \ref{fig:bler-bg2-15} compares the current LDPC decoder implementation to the reference implementation developed by Kien. This reference implementation is called \textit{LDPC Ref} and uses the min-sum algorithm with 2 layers and 16 bit for processing. Out current optimized decoder implementation is referred to as \textit{LDPC Opt}. Moreover, reference results provided by Huawei are also shown.
|
||||
|
||||
\begin{figure}[ht]
|
||||
\centering
|
||||
\begin{tikzpicture}
|
||||
\tikzstyle{every pin}=[fill=white,draw=black]
|
||||
\pgfplotsset{every axis legend/.append style={
|
||||
cells={anchor=west}, at={(1.05,1)}, anchor=north west}}
|
||||
% \pgfplotsset{every axis plot/.append style={smooth}}
|
||||
\pgfplotsset{every axis/.append style={line width=0.5pt}}
|
||||
\pgfplotsset{every axis/.append style={mark options=solid, mark size=2.5pt}}
|
||||
|
||||
\begin{semilogyaxis}[title={}, xlabel={$\SNR$ [dB]}, ylabel={BLER},
|
||||
grid={both}, xmin=-4, xmax=2, xtick={-4,-3.5,...,2}, ymin=0,
|
||||
ymax=1,ytickten={-5,-4,-3,-2,-1,0},legend columns=1]
|
||||
|
||||
% HUAWEI merged BG2 2017-06-15
|
||||
\addplot[black, solid] plot coordinates { (-3.91839,0.01) (-3.5567,0.0001) };
|
||||
|
||||
% Kien's 2-layer 16bit code
|
||||
\addplot[red, solid, mark=o] plot coordinates { (-2.750000,0.915500) (-2.500000,0.576000) (-2.250000,0.165000) (-2.000000,0.017100) (-1.750000,0.000600) (-1.500000,0.000000) (-1.250000,0.000000) (-1.000000,0.000000)};
|
||||
|
||||
% LDPC opt with 16bit BN processing
|
||||
\addplot[blue, solid, mark=square] plot coordinates { (-2.750000,0.998600) (-2.500000,0.953600) (-2.250000,0.718800) (-2.000000,0.299300) (-1.750000,0.053700) (-1.500000,0.005100) (-1.250000,0.000500) (-1.000000,0.000100)};
|
||||
|
||||
% Matlab
|
||||
\addplot[green, solid, mark=triangle] plot coordinates {(-2.750000,0.318200) (-2.500000,0.135900) (-2.250000,0.102000) (-2.000000,0.092300) (-1.750000,0.079200) (-1.500000,0.063100) (-1.250000,0.041400) (-1.000000,0.029600) (-0.750000,0.017500) (-0.500000,0.011800) (-0.250000,0.006100) (0.000000,0.004100) (0.250000,0.002800) (0.500000,0.000800) (0.750000,0.000300) (1.000000,0.000400) };
|
||||
|
||||
|
||||
% \addplot[blue, solid, mark=triangle] plot coordinates {(-2.750000,0.997600) (-2.500000,0.961600) (-2.250000,0.815200) (-2.000000,0.628800) (-1.750000,0.586200) (-1.500000,0.572800) (-1.250000,0.507600) (-1.000000,0.376700) (-0.750000,0.262000) (-0.500000,0.157100) (-0.250000,0.087100) (0.000000,0.045400) (0.250000,0.021000) (0.500000,0.010000) (0.750000,0.004400) (1.000000,0.002600) (1.250000,0.000700) (1.500000,0.000500) (1.750000,0.000000) (2.000000,0.000000) (2.250000,0.000000) (2.500000,0.000000) (2.750000,0.000000) (3.000000,0.000000)};
|
||||
|
||||
% 20 iterations
|
||||
\addplot[red, solid, mark=o] plot coordinates { (-2.750000,0.330300) (-2.500000,0.067800) (-2.250000,0.006000) (-2.000000,0.000100) (-1.750000,0.000000) (-1.500000,0.000000) (-1.250000,0.000000) (-1.000000,0.000000)};
|
||||
|
||||
\addplot[blue, solid, mark=square] plot coordinates {(-2.750000,0.341300) (-2.500000,0.065100) (-2.250000,0.004100) (-2.000000,0.000200) (-1.750000,0.000100) (-1.500000,0.000000) (-1.250000,0.000000) (-1.000000,0.000000)};
|
||||
|
||||
% 5 iterations
|
||||
\addplot[red, solid, mark=o] plot coordinates {(-1.250000,0.781300) (-1.000000,0.421000) (-0.750000,0.140400) (-0.500000,0.028900) (-0.250000,0.003300) (0.000000,0.000300) (0.250000,0.000000) (0.500000,0.000000)};
|
||||
\addplot[blue, solid, mark=square] plot coordinates {(-0.250000,0.705000) (0.000000,0.406200) (0.250000,0.181300) (0.500000,0.061600) (0.750000,0.015900) (1.000000,0.004900) (1.250000,0.000900) (1.500000,0.000200)};
|
||||
\addplot[green, solid, mark=triangle] plot coordinates {(-1.000000,0.778900) (-0.500000,0.226400) (0.000000,0.027400) (0.500000,0.002600) (1.000000,0.000300) };
|
||||
|
||||
% 30 iterations
|
||||
% \addplot[blue, dashed, mark=square] plot coordinates {(-3.000000,0.623800) (-2.750000,0.224100) (-2.500000,0.031600) (-2.250000,0.001100) (-2.000000,0.000000)};
|
||||
|
||||
|
||||
\draw (axis cs:-3.3,0.1) node[fill=white,draw=black] (pint0) {20 iter};
|
||||
\draw (axis cs:-2.3,0.01) node[draw,black,thick,ellipse,minimum height=0.3cm] (ell0) {}; \draw[black,thick] (pint0) -- (ell0);
|
||||
|
||||
\draw (axis cs:0,0.0001) node[fill=white,draw=black] (pint1) {10 iter};
|
||||
\draw (axis cs:-1.6,0.002) node[draw,black,thick,ellipse,minimum width=0.8cm] (ell1) {}; \draw[black,thick] (pint1) -- (ell1);
|
||||
|
||||
\draw (axis cs:1.3,0.2) node[fill=white,draw=black] (pint2) {5 iter};
|
||||
\draw (axis cs:0.3,0.01) node[draw,black,thick,ellipse,minimum width=2cm] (ell2) {}; \draw[black,thick] (pint2) -- (ell2);
|
||||
|
||||
|
||||
\legend{ {Huawei 2017-06-15}\\
|
||||
{LDPC Ref}\\
|
||||
{LDPC Opt}\\
|
||||
{MATLAB 5Glib}\\};
|
||||
|
||||
\end{semilogyaxis}
|
||||
\end{tikzpicture}
|
||||
\caption{BLER vs. SNR, BG2, Rate=1/5, \{5,10,20\} Iterations, B=1280.}
|
||||
\label{fig:bler-bg2-15}
|
||||
\end{figure}
|
||||
|
||||
From Figure \ref{fig:bler-bg2-15} it can be observed that the reference decoder outperforms the current implementation significantly for low to medium number of iterations. The reason is the implementation of 2 layers in the reference decoder, which results in faster convergence for punctured codes and hence requires less iterations to achieve a given BLER target. Note that there is a large performance loss of nearly 6 dB at BLER $10^{-2}$ between the Huawei reference and the current optimized decoder implementation with 5 iterations.
|
||||
|
||||
Moreover, there is a gap of about 1.5 dB between the results provided by Huawei and the current decoder with 20 iterations. The reason is the min-sum approximation algorithm used in both the reference decoder and the current implementation. The gap can be closed by using a tighter approximation like the min-sum with normalization or the lambda-min approach. Moreover, the gap closes for higher code rates which can be observed from Figure \ref{fig:bler-bg2-r23}. The gap is only about 0.6 dB for 50 iterations.
|
||||
|
||||
Concerning the LDPC decoder provided by MATLAB, the performance appears to be rather inconsistent. For 5 iterations, the MATLAB decoder outperforms the optimized decoder most likely due to a tighter approximation used in the check node processing. However, it is inferior to the reference algorithm which suggests that the MATLAB decoder is not optimized for punctured LDPC codes, i.e. no layered processing. For 50 iterations the MATLAB LDPC decoder shows a strange behavior, the slope of the BLER curve is not as expected. This suggests that there might be some internal decoder problems with the NR base graph 2.
|
||||
|
||||
|
||||
\begin{figure}[ht]
|
||||
\centering
|
||||
\begin{tikzpicture}
|
||||
\tikzstyle{every pin}=[fill=white,draw=black]
|
||||
\pgfplotsset{every axis legend/.append style={
|
||||
cells={anchor=west}, at={(1.05,1)}, anchor=north west}}
|
||||
% \pgfplotsset{every axis plot/.append style={smooth}}
|
||||
\pgfplotsset{every axis/.append style={line width=0.5pt}}
|
||||
\pgfplotsset{every axis/.append style={mark options=solid, mark size=2.5pt}}
|
||||
|
||||
\begin{semilogyaxis}[title={}, xlabel={$\SNR$ [dB]}, ylabel={BLER},
|
||||
grid={both}, xmin=3, xmax=5.5, xtick={3,3.5,...,5.5}, ymin=0,
|
||||
ymax=1,ytickten={-5,-4,-3,-2,-1,0},legend columns=1]
|
||||
|
||||
% Kien's 2-layer 16bit code
|
||||
%\addplot[red, solid] plot coordinates { (-2.750000,0.915500) (-2.500000,0.576000) (-2.250000,0.165000) (-2.000000,0.017100) (-1.750000,0.000600) (-1.500000,0.000000) (-1.250000,0.000000) (-1.000000,0.000000)};
|
||||
|
||||
% Huawei
|
||||
\addplot[black, solid] plot coordinates { (3.28392,0.01) (3.73319,0.0001) };
|
||||
|
||||
% LDPC opt with 16bit BN processing
|
||||
\addplot[blue, solid, mark=square] plot coordinates {(4.000000,0.487500) (4.250000,0.163400) (4.500000,0.029800) (4.750000,0.002700) (5.000000,0.000100)};
|
||||
|
||||
%\addplot[blue, dashed, mark=triangle] plot coordinates {(4.000000,0.487500) (4.250000,0.163700) (4.500000,0.030000) (4.750000,0.002900) (5.000000,0.000100)};
|
||||
|
||||
\addplot[blue, dashed, mark=square] plot coordinates {(3.000000,0.911600) (3.250000,0.614100) (3.500000,0.230100) (3.750000,0.036900) (4.000000,0.001100) (4.250000,0.000000) (4.500000,0.000000)};
|
||||
|
||||
|
||||
\legend{ {Huawei 2017-06-15}\\
|
||||
{LDPC Opt 5 iter}\\
|
||||
{LDPC Opt 50 iter}\\};
|
||||
|
||||
\end{semilogyaxis}
|
||||
\end{tikzpicture}
|
||||
\caption{BLER vs. SNR, BG2, Rate=2/3, \{5,50\} Iterations, B=1280.}
|
||||
\label{fig:bler-bg2-r23}
|
||||
\end{figure}
|
||||
|
||||
Figure \ref{fig:bler-bg1-r89} shows the performance of BG1 with largest block size of $B=8448$ and highest code rate $R=8/9$.
|
||||
|
||||
\begin{figure}[ht]
|
||||
\centering
|
||||
\begin{tikzpicture}
|
||||
\tikzstyle{every pin}=[fill=white,draw=black]
|
||||
\pgfplotsset{every axis legend/.append style={
|
||||
cells={anchor=west}, at={(1.05,1)}, anchor=north west}}
|
||||
% \pgfplotsset{every axis plot/.append style={smooth}}
|
||||
\pgfplotsset{every axis/.append style={line width=0.5pt}}
|
||||
\pgfplotsset{every axis/.append style={mark options=solid, mark size=2.5pt}}
|
||||
|
||||
\begin{semilogyaxis}[title={}, xlabel={$\SNR$ [dB]}, ylabel={BLER},
|
||||
grid={both}, xmin=6, xmax=11, xtick={6,6.5,...,11}, ymin=0,
|
||||
ymax=1,ytickten={-5,-4,-3,-2,-1,0},legend columns=1]
|
||||
|
||||
% Huawei
|
||||
\addplot[black, solid] plot coordinates { (6.118717,0.01) (6.291449,0.0001) };
|
||||
|
||||
% LDPC opt 5 iter
|
||||
\addplot[blue, solid, mark=square] plot coordinates {(8.500000,0.350000) (8.750000,0.155100) (9.000000,0.062400) (9.250000,0.023000) (9.500000,0.008700) (9.750000,0.003500) (10.000000,0.000900) (10.250000,0.000300) };
|
||||
|
||||
% LDPC opt 50 iter
|
||||
\addplot[blue, dashed, mark=square] plot coordinates {(6.000000,0.705333) (6.100000,0.353367) (6.200000,0.102100) (6.300000,0.015133) (6.400000,0.000967) (6.500000,0.000000)};
|
||||
|
||||
|
||||
\legend{ {Huawei}\\
|
||||
{LDPC Opt 5 iter}\\
|
||||
{LDPC Opt 50 iter}\\};
|
||||
|
||||
\end{semilogyaxis}
|
||||
\end{tikzpicture}
|
||||
\caption{BLER vs. SNR, BG1, Rate=8/9 \{5,50\} Iterations, B=8448.}
|
||||
\label{fig:bler-bg1-r89}
|
||||
\end{figure}
|
||||
|
||||
From \ref{fig:bler-bg1-r89} it can be observed that the performance gap is only about 0.2 dB if 50 iterations are used. However, for 5 iterations there is still a significant performance loss of about 3.4 dB at BLER $10^{-2}$.
|
||||
|
||||
\newpage
|
||||
\subsection{Decoding Latency}
|
||||
\label{sec:decoding-time}
|
||||
|
||||
This section provides results in terms of decoding latency. That is, the time it takes the decoder to to finish decoding for a given number of iterations. To measure the run time of the decoder we use the OAI tool \texttt{time\_meas.h}. The clock frequency is about 2.9 GHZ, decoder is run on a single core and the results are averaged over $10\,000$ blocks.
|
||||
|
||||
The results in Table \ref{tab:lat-bg2-r15} show the impact of the number of iterations on the decoding latency. It can be observed that the latency roughly doubles if the number of iterations are doubled.
|
||||
|
||||
\begin{table}[ht]
|
||||
\centering
|
||||
\begin{tabular}{lrrr}
|
||||
\toprule
|
||||
\textbf{Function} & \textbf{Time [$\mu s$] (5 it)} & \textbf{Time [$\mu s$] (10 it)} & \textbf{Time [$\mu s$] (20 it)}\\
|
||||
\midrule
|
||||
\texttt{llr2llrProcBuf} & 1.1 & 1.1 & 1.1 \\
|
||||
\texttt{llr2CnProcBuf} & 12.4 & 12.0 & 12.0 \\
|
||||
\texttt{cnProc} & 11.7 & 22.1 & 43.5 \\
|
||||
\texttt{bnProcPc} & 6.6 & 12.1 & 23.8 \\
|
||||
\texttt{bnProc} & 4.2 & 8.1 & 16.2 \\
|
||||
\texttt{cn2bnProcBuf} & 61.3 & 118.3 & 234.9 \\
|
||||
\texttt{bn2cnProcBuf} & 38.1 & 82.5 & 172.3 \\
|
||||
\texttt{llrRes2llrOut} & 3.5 & 3.4 & 3.4 \\
|
||||
\texttt{llr2bit} & 0.2 & 0.1 & 0.1 \\
|
||||
\midrule
|
||||
\textbf{Total} & \textbf{139.4} & \textbf{260.3} & \textbf{508.4} \\
|
||||
\bottomrule
|
||||
\end{tabular}
|
||||
\caption{BG2, Z=128, R=1/5, B=1280, LDPC Opt}
|
||||
\label{tab:lat-bg2-r15}
|
||||
\end{table}
|
||||
|
||||
Table \ref{tab:lat-bg2-i5} shows the impact of the code rate on the latency for a given block size and 5 iterations. It can be observed that the performance gain from code rate 1/3 to 2/3 is about a factor 2.
|
||||
|
||||
\begin{table}[ht]
|
||||
\centering
|
||||
\begin{tabular}{lrrr}
|
||||
\toprule
|
||||
\textbf{Function} & \textbf{Time [$\mu s$] (R=1/5)} & \textbf{Time [$\mu s$] (R=1/3)} & \textbf{Time [$\mu s$] (R=2/3)}\\
|
||||
\midrule
|
||||
\texttt{llr2llrProcBuf} & 3.2 & 2.9 & 2.6 \\
|
||||
\texttt{llr2CnProcBuf} & 36.5 & 25.4 & 14.8 \\
|
||||
\texttt{cnProc} & 33.6 & 25.2 & 13.3 \\
|
||||
\texttt{bnProcPc} & 17.6 & 10.2 & 4.5 \\
|
||||
\texttt{bnProc} & 8.5 & 5.4 & 2.5 \\
|
||||
\texttt{cn2bnProcBuf} & 175.3 & 110.6 & 50.7 \\
|
||||
\texttt{bn2cnProcBuf} & 106.6 & 71.2 & 36.1 \\
|
||||
\texttt{llrRes2llrOut} & 10.2 & 6.3 & 3.3 \\
|
||||
\texttt{llr2bit} & 0.4 & 0.2 & 0.1 \\
|
||||
\midrule
|
||||
\textbf{Total} & \textbf{392.4} & \textbf{258.0} & \textbf{128.2} \\
|
||||
\bottomrule
|
||||
\end{tabular}
|
||||
\caption{BG2, Z=384, B=3840, LDPC Opt, 5 iterations}
|
||||
\label{tab:lat-bg2-i5}
|
||||
\end{table}
|
||||
|
||||
Table \ref{tab:lat-bg1-i5} shows the results for BG1, larges block size and different code rates. The latency difference betwee code rate 1/3 and code rate 2/3 is less than half because upper left corner of the PCM is more dense than the rest of the PCM.
|
||||
|
||||
\begin{table}[ht]
|
||||
\centering
|
||||
\begin{tabular}{lrrr}
|
||||
\toprule
|
||||
\textbf{Function} & \textbf{Time [$\mu s$] (R=1/3)} & \textbf{Time [$\mu s$] (R=2/3)} & \textbf{Time [$\mu s$] (R=8/9)}\\
|
||||
\midrule
|
||||
\texttt{llr2llrProcBuf} & 5.5 & 4.9 & 4.6 \\
|
||||
\texttt{llr2CnProcBuf} & 60.6 & 34.1 & 24.4 \\
|
||||
\texttt{cnProc} & 102.0 & 74.1 & 56.0 \\
|
||||
\texttt{bnProcPc} & 26.0 & 11.0 & 6.4 \\
|
||||
\texttt{bnProc} & 15.7 & 7.4 & 4.5 \\
|
||||
\texttt{cn2bnProcBuf} & 291.0 & 140.8 & 83.1 \\
|
||||
\texttt{bn2cnProcBuf} & 193.6 & 100.5 & 63.0 \\
|
||||
\texttt{llrRes2llrOut} & 13.3 & 6.9 & 5.2 \\
|
||||
\texttt{llr2bit} & 0.4 & 0.2 & 0.2 \\
|
||||
\midrule
|
||||
\textbf{Total} & \textbf{708.9} & \textbf{380.6} & \textbf{248.1}\\
|
||||
\bottomrule
|
||||
\end{tabular}
|
||||
\caption{BG1, Z=384, B=8448, LDPC Opt, 5 iterations}
|
||||
\label{tab:lat-bg1-i5}
|
||||
\end{table}
|
||||
|
||||
From the above results it can be observed that the data transfer between CNs and BNs takes up a significant amount of the run time. However, the performance gain due to AVX instructions in both CN and BN processing is significantly larger than the penalty incurred by the data transfers.
|
||||
|
||||
\section{Parity Check and early stopping Criteria}
|
||||
It is often unnecessary to carry out the maximum number of iterations. After each iteration a parity check \eqref{eq:29} can be computed and if a valid code word is found the decoder can stop. This functionality has been implemented and the additional overhead is reasonable. The PC is carried out in the CN processing buffer and the calculation complexity itself is negligible. However, for the processing it is necessary to move the BN results to the CN buffer which takes time, the overall overhead is at most $10\%$ compared to an algorithm without early stopping criteria with the same number of iterations. The PC has to be activated via the define \texttt{NR\_LDPC\_ENABLE\_PARITY\_CHECK}.
|
||||
|
||||
|
||||
\section{Conclusion}
|
||||
\label{sec:conclusion}
|
||||
|
||||
The results in the previous sections show that the current optimized LDPC implementation full-fills the requirements in terms of decoding latency for low to medium number of iterations at the expanse of a significant loss in BLER performance. To improve BLER performance, it is recommended to implement a layered algorithm and a min-sum algorithm with normalization. Further improvements upon the current implementation are detailed in the next section.
|
||||
|
||||
\newpage
|
||||
\section{Future Work}
|
||||
\label{sec:future-work}
|
||||
|
||||
The improvements upon the current LDPC decoder implementation can be divided into two categories:
|
||||
\begin{enumerate}
|
||||
\item Improved BLER performance
|
||||
\item Reduced decoding latency
|
||||
\end{enumerate}
|
||||
|
||||
\subsection{Improved BLER Performance}
|
||||
\label{sec:impr-bler-perf}
|
||||
|
||||
The BLER performance can be improved by using a tighter approximation than the min-sum approximation. For instance, the min-sum algorithm can be improved by adding a correction factor in the CN processing . The min-sum approximation in \eqref{eq:40} is modified as
|
||||
\begin{equation}
|
||||
\label{eq:50}
|
||||
r_{ji} = \prod_{j'\in\Bcal_i\setminus j}\sgn q_{ij'}\min_{j'\in\Bcal_i\setminus j} |q_{ij'}| + w(q_{ij'})
|
||||
\end{equation}
|
||||
The correction term $w(q_{ij'})$ is defined as
|
||||
\begin{equation}
|
||||
\label{eq:51}
|
||||
w(q_{ij'}) =
|
||||
\begin{cases}
|
||||
c & \textrm{if}~ \\
|
||||
-c & \textrm{if}~ \\
|
||||
0 & \textrm{otherwise}
|
||||
\end{cases}
|
||||
\end{equation}
|
||||
where the constant $c$ is of order $0.5$ typically.
|
||||
|
||||
\subsection{Reduced Decoding Latency}
|
||||
\label{sec:reduc-decod-latency}
|
||||
|
||||
The following improvements will reduce the decoding latency:
|
||||
|
||||
\begin{itemize}
|
||||
\item Adapt to AVX512
|
||||
\item Implement 2/3-layers for faster convergence
|
||||
\end{itemize}
|
||||
|
||||
\paragraph{AVX512:}
|
||||
The computations in the CN and BN processing can be further accelerated by using AVX512 instructions. This improvement will speed-up the CN and BN processing by a approximately a factor of 2.
|
||||
|
||||
\paragraph{Layered processing:}
|
||||
The LDPC code in NR always punctures the first 2 columns of the base graph. Hence, the decoder inserts LLRs with value 0 at their place and needs to retrieve those bits during the decoding process. Instead of computing all the parity equations and then passing the results to the BN processing, it is beneficial to first compute parity equations where at most one punctured BN is connected to that CN. If two punctured BNs are connected than according to \eqref{eq:40}, the result will be again 0. Thus in a first sub-iteration those parity equation are computed and the results are send to BN processing which calculates the results using only those rows in the PCM. In the second sub-iteration the remaining check equation are used.
|
||||
The convergence of this layered approach is much fast since the bit can be retrieved more quickly while the decoding complexity remains the same. Therefore, for a fixed number of iterations the layered algorithm will have a significantly better performance.
|
||||
|
||||
\newpage
|
||||
\bibliographystyle{IEEEtran}
|
||||
\bibliography{./references}
|
||||
|
||||
\end{document}
|
||||
|
||||
%%% Local Variables:
|
||||
%%% mode: latex
|
||||
%%% TeX-master: t
|
||||
%%% End:
|
||||
@@ -0,0 +1,105 @@
|
||||
%
|
||||
% 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
|
||||
%
|
||||
@online{3gpp5gTimeline,
|
||||
author = {3GPP},
|
||||
title = {{3GPP 5G Timeline}},
|
||||
year = 2016,
|
||||
urldate = {2017-06-14},
|
||||
url = {http://www.3gpp.org/images/articleimages/5g_timeline.jpg}
|
||||
}
|
||||
|
||||
@techreport{3gppTR38913,
|
||||
author = "{Technical Specification Group Radio Access Network}",
|
||||
title = "{Study on Scenarios and Requirements for Next Generation Access Technologies}",
|
||||
institution = "{3GPP TR 38.913 V14.2.0}",
|
||||
month = mar,
|
||||
year = 2017,
|
||||
};
|
||||
|
||||
@techreport{iturM2038,
|
||||
author = "{ITU-R}",
|
||||
title = "{IMT Vision -- Framework and overall objectives of the future development of IMT for 2020 and beyond}",
|
||||
institution = "{Radiocommunication Sector of ITU}",
|
||||
month = sep,
|
||||
year = 2015,
|
||||
};
|
||||
|
||||
@techreport{3gpp2014seb,
|
||||
author = "{Samsung, Nokia Networks}",
|
||||
title = "{New SID Proposal: Study on Elevation Beamforming/Full-Dimension (FD) MIMO for LTE}",
|
||||
institution = "3GPP",
|
||||
month = sep,
|
||||
year = 2014,
|
||||
};
|
||||
|
||||
@techreport{3gpp2015fdm,
|
||||
author = "{Technical Specification Group Radio Access Network}",
|
||||
title = "{Study on elevation beamforming / Full-Dimension (FD) Multiple Input Multiple Output (MIMO) for LTE}",
|
||||
institution = "3GPP TR 36.897 V13.0.0",
|
||||
month = jun,
|
||||
year = 2015,
|
||||
};
|
||||
|
||||
|
||||
@techreport{3gpp2008tsg,
|
||||
author = "{Technical Specification Group Radio Access Network;
|
||||
Evolved Universal Terrestrial Radio Access (E-UTRA)}",
|
||||
title = "{Further advancements for E-UTRA physical layer aspects (Release 9)}",
|
||||
institution = "{3GPP TR 36.814 V9.0.0}",
|
||||
month = mar,
|
||||
year = 2010,
|
||||
};
|
||||
|
||||
@techreport{3gpp2011uer,
|
||||
author = "{Technical Specification Group Radio Access Network;
|
||||
Evolved Universal Terrestrial Radio Access (E-UTRA)}",
|
||||
title = "{User Equipment (UE) Radio Transmission and Reception}",
|
||||
institution = "{3GPP TR 36.101 V10.3.0}",
|
||||
month = jun,
|
||||
year = 2011,
|
||||
};
|
||||
|
||||
@techreport{3gpp2009_36211,
|
||||
author = "{3rd Generation Partnership Project}",
|
||||
title = "{Physical Channels and Modulation (Release 8)}",
|
||||
institution = "{3GPP TS 36.211 V8.6.0}",
|
||||
month = mar,
|
||||
year = 2009,
|
||||
};
|
||||
|
||||
@techreport{3gpp2017_38212,
|
||||
author = "{3rd Generation Partnership Project}",
|
||||
title = "{Multiplexing and channel coding (Release 15)}",
|
||||
institution = "{3GPP TS 38.212 V15.0.1}",
|
||||
month = mar,
|
||||
year = 2018,
|
||||
};
|
||||
|
||||
@article{gallager1962low,
|
||||
title={Low-density parity-check codes},
|
||||
author={Gallager, Robert},
|
||||
journal={IRE Transactions on information theory},
|
||||
volume={8},
|
||||
number={1},
|
||||
pages={21--28},
|
||||
year={1962},
|
||||
publisher={IEEE}
|
||||
}
|
||||
File diff suppressed because it is too large
Load Diff
File diff suppressed because it is too large
Load Diff
File diff suppressed because it is too large
Load Diff
@@ -0,0 +1,477 @@
|
||||
/*
|
||||
* Licensed to the OpenAirInterface (OAI) Software Alliance under one or more
|
||||
* contributor license agreements. See the NOTICE file distributed with
|
||||
* this work for additional information regarding copyright ownership.
|
||||
* The OpenAirInterface Software Alliance licenses this file to You under
|
||||
* the OAI Public License, Version 1.1 (the "License"); you may not use this file
|
||||
* except in compliance with the License.
|
||||
* You may obtain a copy of the License at
|
||||
*
|
||||
* http://www.openairinterface.org/?page_id=698
|
||||
*
|
||||
* Unless required by applicable law or agreed to in writing, software
|
||||
* distributed under the License is distributed on an "AS IS" BASIS,
|
||||
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
||||
* See the License for the specific language governing permissions and
|
||||
* limitations under the License.
|
||||
*-------------------------------------------------------------------------------
|
||||
* For more information about the OpenAirInterface (OAI) Software Alliance:
|
||||
* contact@openairinterface.org
|
||||
*/
|
||||
|
||||
/*!\file nrLDPC_decoder.c
|
||||
* \brief Defines the LDPC decoder
|
||||
* \author Sebastian Wagner (TCL Communications) Email: <mailto:sebastian.wagner@tcl.com>
|
||||
* \date 27-03-2018
|
||||
* \version 1.0
|
||||
* \note
|
||||
* \warning
|
||||
*/
|
||||
|
||||
|
||||
#include <stdint.h>
|
||||
#include <immintrin.h>
|
||||
#include "nrLDPC_defs.h"
|
||||
#include "nrLDPC_types.h"
|
||||
#include "nrLDPC_init.h"
|
||||
#include "nrLDPC_mPass.h"
|
||||
#include "nrLDPC_cnProc.h"
|
||||
#include "nrLDPC_bnProc.h"
|
||||
|
||||
#ifdef NR_LDPC_DEBUG_MODE
|
||||
#include "nrLDPC_tools/nrLDPC_debug.h"
|
||||
#endif
|
||||
|
||||
static inline uint32_t nrLDPC_decoder_core(int8_t* p_llr, int8_t* p_out, uint32_t numLLR, t_nrLDPC_lut* p_lut, t_nrLDPC_dec_params* p_decParams, t_nrLDPC_time_stats* p_profiler);
|
||||
|
||||
int32_t nrLDPC_decoder(t_nrLDPC_dec_params* p_decParams, int8_t* p_llr, int8_t* p_out, t_nrLDPC_time_stats* p_profiler)
|
||||
{
|
||||
uint32_t numLLR;
|
||||
uint32_t numIter = 0;
|
||||
t_nrLDPC_lut lut;
|
||||
t_nrLDPC_lut* p_lut = &lut;
|
||||
|
||||
// Initialize decoder core(s) with correct LUTs
|
||||
numLLR = nrLDPC_init(p_decParams, p_lut);
|
||||
|
||||
// Launch LDPC decoder core for one segment
|
||||
numIter = nrLDPC_decoder_core(p_llr, p_out, numLLR, p_lut, p_decParams, p_profiler);
|
||||
|
||||
return numIter;
|
||||
}
|
||||
|
||||
/**
|
||||
\brief Performs LDPC decoding of one code block
|
||||
\param p_llr Input LLRs
|
||||
\param p_out Output vector
|
||||
\param numLLR Number of LLRs
|
||||
\param p_lut Pointer to decoder LUTs
|
||||
\param p_decParams LDPC decoder parameters
|
||||
\param p_profiler LDPC profiler statistics
|
||||
*/
|
||||
static inline uint32_t nrLDPC_decoder_core(int8_t* p_llr, int8_t* p_out, uint32_t numLLR, t_nrLDPC_lut* p_lut, t_nrLDPC_dec_params* p_decParams, t_nrLDPC_time_stats* p_profiler)
|
||||
{
|
||||
uint16_t Z = p_decParams->Z;
|
||||
uint8_t BG = p_decParams->BG;
|
||||
uint8_t numMaxIter = p_decParams->numMaxIter;
|
||||
e_nrLDPC_outMode outMode = p_decParams->outMode;
|
||||
|
||||
// Minimum number of iterations is 1
|
||||
// 0 iterations means hard-decision on input LLRs
|
||||
uint32_t i = 1;
|
||||
// Initialize with parity check fail != 0
|
||||
int32_t pcRes = 1;
|
||||
int8_t* p_llrOut;
|
||||
|
||||
if (outMode == nrLDPC_outMode_LLRINT8)
|
||||
{
|
||||
p_llrOut = p_out;
|
||||
}
|
||||
else
|
||||
{
|
||||
// Use LLR processing buffer as temporary output buffer
|
||||
p_llrOut = (int8_t*) llrProcBuf;
|
||||
}
|
||||
|
||||
|
||||
// Initialization
|
||||
#ifdef NR_LDPC_PROFILER_DETAIL
|
||||
start_meas(&p_profiler->llr2llrProcBuf);
|
||||
#endif
|
||||
nrLDPC_llr2llrProcBuf(p_lut, p_llr, Z, BG);
|
||||
#ifdef NR_LDPC_PROFILER_DETAIL
|
||||
stop_meas(&p_profiler->llr2llrProcBuf);
|
||||
#endif
|
||||
|
||||
#ifdef NR_LDPC_DEBUG_MODE
|
||||
nrLDPC_debug_initBuffer2File(nrLDPC_buffers_LLR_PROC);
|
||||
nrLDPC_debug_writeBuffer2File(nrLDPC_buffers_LLR_PROC);
|
||||
#endif
|
||||
|
||||
#ifdef NR_LDPC_PROFILER_DETAIL
|
||||
start_meas(&p_profiler->llr2CnProcBuf);
|
||||
#endif
|
||||
nrLDPC_llr2CnProcBuf(p_lut, p_llr, numLLR, Z, BG);
|
||||
#ifdef NR_LDPC_PROFILER_DETAIL
|
||||
stop_meas(&p_profiler->llr2CnProcBuf);
|
||||
#endif
|
||||
|
||||
#ifdef NR_LDPC_DEBUG_MODE
|
||||
nrLDPC_debug_initBuffer2File(nrLDPC_buffers_CN_PROC);
|
||||
nrLDPC_debug_writeBuffer2File(nrLDPC_buffers_CN_PROC);
|
||||
#endif
|
||||
|
||||
// First iteration
|
||||
|
||||
// CN processing
|
||||
#ifdef NR_LDPC_PROFILER_DETAIL
|
||||
start_meas(&p_profiler->cnProc);
|
||||
#endif
|
||||
if (BG == 1)
|
||||
{
|
||||
nrLDPC_cnProc_BG1(p_lut, Z);
|
||||
}
|
||||
else
|
||||
{
|
||||
nrLDPC_cnProc_BG2(p_lut, Z);
|
||||
}
|
||||
#ifdef NR_LDPC_PROFILER_DETAIL
|
||||
stop_meas(&p_profiler->cnProc);
|
||||
#endif
|
||||
|
||||
#ifdef NR_LDPC_DEBUG_MODE
|
||||
nrLDPC_debug_initBuffer2File(nrLDPC_buffers_CN_PROC_RES);
|
||||
nrLDPC_debug_writeBuffer2File(nrLDPC_buffers_CN_PROC_RES);
|
||||
#endif
|
||||
|
||||
#ifdef NR_LDPC_PROFILER_DETAIL
|
||||
start_meas(&p_profiler->cn2bnProcBuf);
|
||||
#endif
|
||||
if (BG == 1)
|
||||
{
|
||||
nrLDPC_cn2bnProcBuf_BG1(p_lut, Z);
|
||||
}
|
||||
else
|
||||
{
|
||||
nrLDPC_cn2bnProcBuf(p_lut, Z);
|
||||
}
|
||||
#ifdef NR_LDPC_PROFILER_DETAIL
|
||||
stop_meas(&p_profiler->cn2bnProcBuf);
|
||||
#endif
|
||||
|
||||
#ifdef NR_LDPC_DEBUG_MODE
|
||||
nrLDPC_debug_initBuffer2File(nrLDPC_buffers_BN_PROC);
|
||||
nrLDPC_debug_writeBuffer2File(nrLDPC_buffers_BN_PROC);
|
||||
#endif
|
||||
|
||||
// BN processing
|
||||
#ifdef NR_LDPC_PROFILER_DETAIL
|
||||
start_meas(&p_profiler->bnProcPc);
|
||||
#endif
|
||||
nrLDPC_bnProcPc(p_lut, Z);
|
||||
#ifdef NR_LDPC_PROFILER_DETAIL
|
||||
stop_meas(&p_profiler->bnProcPc);
|
||||
#endif
|
||||
|
||||
#ifdef NR_LDPC_DEBUG_MODE
|
||||
nrLDPC_debug_initBuffer2File(nrLDPC_buffers_LLR_RES);
|
||||
nrLDPC_debug_writeBuffer2File(nrLDPC_buffers_LLR_RES);
|
||||
#endif
|
||||
|
||||
#ifdef NR_LDPC_PROFILER_DETAIL
|
||||
start_meas(&p_profiler->bnProc);
|
||||
#endif
|
||||
nrLDPC_bnProc(p_lut, Z);
|
||||
#ifdef NR_LDPC_PROFILER_DETAIL
|
||||
stop_meas(&p_profiler->bnProc);
|
||||
#endif
|
||||
|
||||
#ifdef NR_LDPC_DEBUG_MODE
|
||||
nrLDPC_debug_initBuffer2File(nrLDPC_buffers_BN_PROC_RES);
|
||||
nrLDPC_debug_writeBuffer2File(nrLDPC_buffers_BN_PROC_RES);
|
||||
#endif
|
||||
|
||||
// BN results to CN processing buffer
|
||||
#ifdef NR_LDPC_PROFILER_DETAIL
|
||||
start_meas(&p_profiler->bn2cnProcBuf);
|
||||
#endif
|
||||
if (BG == 1)
|
||||
{
|
||||
nrLDPC_bn2cnProcBuf_BG1(p_lut, Z);
|
||||
}
|
||||
else
|
||||
{
|
||||
nrLDPC_bn2cnProcBuf(p_lut, Z);
|
||||
}
|
||||
#ifdef NR_LDPC_PROFILER_DETAIL
|
||||
stop_meas(&p_profiler->bn2cnProcBuf);
|
||||
#endif
|
||||
|
||||
#ifdef NR_LDPC_DEBUG_MODE
|
||||
nrLDPC_debug_writeBuffer2File(nrLDPC_buffers_CN_PROC);
|
||||
#endif
|
||||
|
||||
// Parity Check not necessary here since it will fail
|
||||
// because first 2 cols/BNs in BG are punctured and cannot be
|
||||
// estimated after only one iteration
|
||||
|
||||
// First iteration finished
|
||||
|
||||
while ( (i < (numMaxIter-1)) && (pcRes != 0) )
|
||||
{
|
||||
// Increase iteration counter
|
||||
i++;
|
||||
|
||||
// CN processing
|
||||
#ifdef NR_LDPC_PROFILER_DETAIL
|
||||
start_meas(&p_profiler->cnProc);
|
||||
#endif
|
||||
if (BG == 1)
|
||||
{
|
||||
nrLDPC_cnProc_BG1(p_lut, Z);
|
||||
}
|
||||
else
|
||||
{
|
||||
nrLDPC_cnProc_BG2(p_lut, Z);
|
||||
}
|
||||
#ifdef NR_LDPC_PROFILER_DETAIL
|
||||
stop_meas(&p_profiler->cnProc);
|
||||
#endif
|
||||
|
||||
#ifdef NR_LDPC_DEBUG_MODE
|
||||
nrLDPC_debug_writeBuffer2File(nrLDPC_buffers_CN_PROC_RES);
|
||||
#endif
|
||||
|
||||
// Send CN results back to BNs
|
||||
#ifdef NR_LDPC_PROFILER_DETAIL
|
||||
start_meas(&p_profiler->cn2bnProcBuf);
|
||||
#endif
|
||||
if (BG == 1)
|
||||
{
|
||||
nrLDPC_cn2bnProcBuf_BG1(p_lut, Z);
|
||||
}
|
||||
else
|
||||
{
|
||||
nrLDPC_cn2bnProcBuf(p_lut, Z);
|
||||
}
|
||||
#ifdef NR_LDPC_PROFILER_DETAIL
|
||||
stop_meas(&p_profiler->cn2bnProcBuf);
|
||||
#endif
|
||||
|
||||
#ifdef NR_LDPC_DEBUG_MODE
|
||||
nrLDPC_debug_writeBuffer2File(nrLDPC_buffers_BN_PROC);
|
||||
#endif
|
||||
|
||||
// BN Processing
|
||||
#ifdef NR_LDPC_PROFILER_DETAIL
|
||||
start_meas(&p_profiler->bnProcPc);
|
||||
#endif
|
||||
nrLDPC_bnProcPc(p_lut, Z);
|
||||
#ifdef NR_LDPC_PROFILER_DETAIL
|
||||
stop_meas(&p_profiler->bnProcPc);
|
||||
#endif
|
||||
|
||||
#ifdef NR_LDPC_DEBUG_MODE
|
||||
nrLDPC_debug_writeBuffer2File(nrLDPC_buffers_LLR_RES);
|
||||
#endif
|
||||
|
||||
#ifdef NR_LDPC_PROFILER_DETAIL
|
||||
start_meas(&p_profiler->bnProc);
|
||||
#endif
|
||||
nrLDPC_bnProc(p_lut, Z);
|
||||
#ifdef NR_LDPC_PROFILER_DETAIL
|
||||
stop_meas(&p_profiler->bnProc);
|
||||
#endif
|
||||
|
||||
#ifdef NR_LDPC_DEBUG_MODE
|
||||
nrLDPC_debug_writeBuffer2File(nrLDPC_buffers_BN_PROC_RES);
|
||||
#endif
|
||||
|
||||
// BN results to CN processing buffer
|
||||
#ifdef NR_LDPC_PROFILER_DETAIL
|
||||
start_meas(&p_profiler->bn2cnProcBuf);
|
||||
#endif
|
||||
if (BG == 1)
|
||||
{
|
||||
nrLDPC_bn2cnProcBuf_BG1(p_lut, Z);
|
||||
}
|
||||
else
|
||||
{
|
||||
nrLDPC_bn2cnProcBuf(p_lut, Z);
|
||||
}
|
||||
#ifdef NR_LDPC_PROFILER_DETAIL
|
||||
stop_meas(&p_profiler->bn2cnProcBuf);
|
||||
#endif
|
||||
|
||||
#ifdef NR_LDPC_DEBUG_MODE
|
||||
nrLDPC_debug_writeBuffer2File(nrLDPC_buffers_CN_PROC);
|
||||
#endif
|
||||
|
||||
// Parity Check
|
||||
#ifdef NR_LDPC_ENABLE_PARITY_CHECK
|
||||
#ifdef NR_LDPC_PROFILER_DETAIL
|
||||
start_meas(&p_profiler->cnProcPc);
|
||||
#endif
|
||||
if (BG == 1)
|
||||
{
|
||||
pcRes = nrLDPC_cnProcPc_BG1(p_lut, Z);
|
||||
}
|
||||
else
|
||||
{
|
||||
pcRes = nrLDPC_cnProcPc_BG2(p_lut, Z);
|
||||
}
|
||||
#ifdef NR_LDPC_PROFILER_DETAIL
|
||||
stop_meas(&p_profiler->cnProcPc);
|
||||
#endif
|
||||
#endif
|
||||
|
||||
}
|
||||
|
||||
// Last iteration
|
||||
if ( (i < numMaxIter) && (pcRes != 0) )
|
||||
{
|
||||
// Increase iteration counter
|
||||
i++;
|
||||
|
||||
// CN processing
|
||||
#ifdef NR_LDPC_PROFILER_DETAIL
|
||||
start_meas(&p_profiler->cnProc);
|
||||
#endif
|
||||
if (BG == 1)
|
||||
{
|
||||
nrLDPC_cnProc_BG1(p_lut, Z);
|
||||
}
|
||||
else
|
||||
{
|
||||
nrLDPC_cnProc_BG2(p_lut, Z);
|
||||
}
|
||||
#ifdef NR_LDPC_PROFILER_DETAIL
|
||||
stop_meas(&p_profiler->cnProc);
|
||||
#endif
|
||||
|
||||
#ifdef NR_LDPC_DEBUG_MODE
|
||||
nrLDPC_debug_writeBuffer2File(nrLDPC_buffers_CN_PROC_RES);
|
||||
#endif
|
||||
|
||||
// Send CN results back to BNs
|
||||
#ifdef NR_LDPC_PROFILER_DETAIL
|
||||
start_meas(&p_profiler->cn2bnProcBuf);
|
||||
#endif
|
||||
if (BG == 1)
|
||||
{
|
||||
nrLDPC_cn2bnProcBuf_BG1(p_lut, Z);
|
||||
}
|
||||
else
|
||||
{
|
||||
nrLDPC_cn2bnProcBuf(p_lut, Z);
|
||||
}
|
||||
#ifdef NR_LDPC_PROFILER_DETAIL
|
||||
stop_meas(&p_profiler->cn2bnProcBuf);
|
||||
#endif
|
||||
|
||||
#ifdef NR_LDPC_DEBUG_MODE
|
||||
nrLDPC_debug_writeBuffer2File(nrLDPC_buffers_BN_PROC);
|
||||
#endif
|
||||
|
||||
// BN Processing
|
||||
#ifdef NR_LDPC_PROFILER_DETAIL
|
||||
start_meas(&p_profiler->bnProcPc);
|
||||
#endif
|
||||
nrLDPC_bnProcPc(p_lut, Z);
|
||||
#ifdef NR_LDPC_PROFILER_DETAIL
|
||||
stop_meas(&p_profiler->bnProcPc);
|
||||
#endif
|
||||
|
||||
#ifdef NR_LDPC_DEBUG_MODE
|
||||
nrLDPC_debug_writeBuffer2File(nrLDPC_buffers_LLR_RES);
|
||||
#endif
|
||||
|
||||
// If parity check not enabled, no need to send the BN proc results
|
||||
// back to CNs
|
||||
#ifdef NR_LDPC_ENABLE_PARITY_CHECK
|
||||
#ifdef NR_LDPC_PROFILER_DETAIL
|
||||
start_meas(&p_profiler->bnProc);
|
||||
#endif
|
||||
nrLDPC_bnProc(p_lut, Z);
|
||||
#ifdef NR_LDPC_PROFILER_DETAIL
|
||||
stop_meas(&p_profiler->bnProc);
|
||||
#endif
|
||||
|
||||
#ifdef NR_LDPC_DEBUG_MODE
|
||||
nrLDPC_debug_writeBuffer2File(nrLDPC_buffers_BN_PROC_RES);
|
||||
#endif
|
||||
|
||||
// BN results to CN processing buffer
|
||||
#ifdef NR_LDPC_PROFILER_DETAIL
|
||||
start_meas(&p_profiler->bn2cnProcBuf);
|
||||
#endif
|
||||
if (BG == 1)
|
||||
{
|
||||
nrLDPC_bn2cnProcBuf_BG1(p_lut, Z);
|
||||
}
|
||||
else
|
||||
{
|
||||
nrLDPC_bn2cnProcBuf(p_lut, Z);
|
||||
}
|
||||
#ifdef NR_LDPC_PROFILER_DETAIL
|
||||
stop_meas(&p_profiler->bn2cnProcBuf);
|
||||
#endif
|
||||
|
||||
#ifdef NR_LDPC_DEBUG_MODE
|
||||
nrLDPC_debug_writeBuffer2File(nrLDPC_buffers_CN_PROC);
|
||||
#endif
|
||||
|
||||
// Parity Check
|
||||
#ifdef NR_LDPC_PROFILER_DETAIL
|
||||
start_meas(&p_profiler->cnProcPc);
|
||||
#endif
|
||||
if (BG == 1)
|
||||
{
|
||||
pcRes = nrLDPC_cnProcPc_BG1(p_lut, Z);
|
||||
}
|
||||
else
|
||||
{
|
||||
pcRes = nrLDPC_cnProcPc_BG2(p_lut, Z);
|
||||
}
|
||||
#ifdef NR_LDPC_PROFILER_DETAIL
|
||||
stop_meas(&p_profiler->cnProcPc);
|
||||
#endif
|
||||
#endif
|
||||
}
|
||||
|
||||
|
||||
// If maximum number of iterations reached an PC still fails increase number of iterations
|
||||
// Thus, i > numMaxIter indicates that PC has failed
|
||||
if (pcRes != 0)
|
||||
{
|
||||
i++;
|
||||
}
|
||||
|
||||
// Assign results from processing buffer to output
|
||||
#ifdef NR_LDPC_PROFILER_DETAIL
|
||||
start_meas(&p_profiler->llrRes2llrOut);
|
||||
#endif
|
||||
nrLDPC_llrRes2llrOut(p_lut, p_llrOut, numLLR);
|
||||
#ifdef NR_LDPC_PROFILER_DETAIL
|
||||
stop_meas(&p_profiler->llrRes2llrOut);
|
||||
#endif
|
||||
|
||||
// Hard-decision
|
||||
#ifdef NR_LDPC_PROFILER_DETAIL
|
||||
start_meas(&p_profiler->llr2bit);
|
||||
#endif
|
||||
if (outMode == nrLDPC_outMode_BIT)
|
||||
{
|
||||
nrLDPC_llr2bitPacked(p_out, p_llrOut, numLLR);
|
||||
}
|
||||
else if (outMode == nrLDPC_outMode_BITINT8)
|
||||
{
|
||||
nrLDPC_llr2bit(p_out, p_llrOut, numLLR);
|
||||
}
|
||||
|
||||
#ifdef NR_LDPC_PROFILER_DETAIL
|
||||
stop_meas(&p_profiler->llr2bit);
|
||||
#endif
|
||||
|
||||
return i;
|
||||
}
|
||||
@@ -0,0 +1,45 @@
|
||||
/*
|
||||
* Licensed to the OpenAirInterface (OAI) Software Alliance under one or more
|
||||
* contributor license agreements. See the NOTICE file distributed with
|
||||
* this work for additional information regarding copyright ownership.
|
||||
* The OpenAirInterface Software Alliance licenses this file to You under
|
||||
* the OAI Public License, Version 1.1 (the "License"); you may not use this file
|
||||
* except in compliance with the License.
|
||||
* You may obtain a copy of the License at
|
||||
*
|
||||
* http://www.openairinterface.org/?page_id=698
|
||||
*
|
||||
* Unless required by applicable law or agreed to in writing, software
|
||||
* distributed under the License is distributed on an "AS IS" BASIS,
|
||||
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
||||
* See the License for the specific language governing permissions and
|
||||
* limitations under the License.
|
||||
*-------------------------------------------------------------------------------
|
||||
* For more information about the OpenAirInterface (OAI) Software Alliance:
|
||||
* contact@openairinterface.org
|
||||
*/
|
||||
|
||||
/*!\file nrLDPC_decoder.h
|
||||
* \brief Defines the LDPC decoder core prototypes
|
||||
* \author Sebastian Wagner (TCL Communications) Email: <mailto:sebastian.wagner@tcl.com>
|
||||
* \date 27-03-2018
|
||||
* \version 1.0
|
||||
* \note
|
||||
* \warning
|
||||
*/
|
||||
|
||||
#ifndef __NR_LDPC_DECODER__H__
|
||||
#define __NR_LDPC_DECODER__H__
|
||||
|
||||
#include "nrLDPC_types.h"
|
||||
|
||||
/**
|
||||
\brief LDPC decoder
|
||||
\param p_decParams LDPC decoder parameters
|
||||
\param p_llr Input LLRs
|
||||
\param p_llrOut Output vector
|
||||
\param p_profiler LDPC profiler statistics
|
||||
*/
|
||||
int32_t nrLDPC_decoder(t_nrLDPC_dec_params* p_decParams, int8_t* p_llr, int8_t* p_llrOut, t_nrLDPC_time_stats* p_profiler);
|
||||
|
||||
#endif
|
||||
@@ -0,0 +1,217 @@
|
||||
/*
|
||||
* Licensed to the OpenAirInterface (OAI) Software Alliance under one or more
|
||||
* contributor license agreements. See the NOTICE file distributed with
|
||||
* this work for additional information regarding copyright ownership.
|
||||
* The OpenAirInterface Software Alliance licenses this file to You under
|
||||
* the OAI Public License, Version 1.1 (the "License"); you may not use this file
|
||||
* except in compliance with the License.
|
||||
* You may obtain a copy of the License at
|
||||
*
|
||||
* http://www.openairinterface.org/?page_id=698
|
||||
*
|
||||
* Unless required by applicable law or agreed to in writing, software
|
||||
* distributed under the License is distributed on an "AS IS" BASIS,
|
||||
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
||||
* See the License for the specific language governing permissions and
|
||||
* limitations under the License.
|
||||
*-------------------------------------------------------------------------------
|
||||
* For more information about the OpenAirInterface (OAI) Software Alliance:
|
||||
* contact@openairinterface.org
|
||||
*/
|
||||
|
||||
/*!\file nrLDPC_defs.h
|
||||
* \brief Defines all constants and buffers for the LDPC decoder
|
||||
* \author Sebastian Wagner (TCL Communications) Email: <mailto:sebastian.wagner@tcl.com>
|
||||
* \date 27-03-2018
|
||||
* \version 1.0
|
||||
* \note
|
||||
* \warning
|
||||
*/
|
||||
|
||||
#ifndef __NR_LDPC_DEFS__H__
|
||||
#define __NR_LDPC_DEFS__H__
|
||||
|
||||
// ==============================================================================
|
||||
// DEFINES
|
||||
|
||||
/** Maximum lifting size */
|
||||
#define NR_LDPC_ZMAX 384
|
||||
|
||||
/** Number of columns in BG1 */
|
||||
#define NR_LDPC_NCOL_BG1 68
|
||||
/** Number of rows in BG1 */
|
||||
#define NR_LDPC_NROW_BG1 46
|
||||
/** Number of edges/entries in BG1 */
|
||||
#define NR_LDPC_NUM_EDGE_BG1 316
|
||||
/** Number of check node (CN) groups in BG1
|
||||
A CN group is defined by its number of connected bit nodes. */
|
||||
#define NR_LDPC_NUM_CN_GROUPS_BG1 9
|
||||
/** First column in BG1 that is connected to only a single CN */
|
||||
#define NR_LDPC_START_COL_PARITY_BG1 26
|
||||
|
||||
/** Number of columns in BG1 for rate 1/3 = 22/(68-2) */
|
||||
#define NR_LDPC_NCOL_BG1_R13 NR_LDPC_NCOL_BG1
|
||||
/** Number of columns in BG1 for rate 2/3 = 22/(35-2) */
|
||||
#define NR_LDPC_NCOL_BG1_R23 35
|
||||
/** Number of columns in BG1 for rate 8/9 ~ 22/(27-2) */
|
||||
#define NR_LDPC_NCOL_BG1_R89 27
|
||||
|
||||
/** Number of bit node (BN) groups in BG1 for rate 1/3
|
||||
A BN group is defined by its number of connected CNs. */
|
||||
#define NR_LDPC_NUM_BN_GROUPS_BG1_R13 30
|
||||
/** Number of bit node (BN) groups in BG1 for rate 2/3 */
|
||||
#define NR_LDPC_NUM_BN_GROUPS_BG1_R23 8
|
||||
/** Number of bit node (BN) groups in BG1 for rate 8/9 */
|
||||
#define NR_LDPC_NUM_BN_GROUPS_BG1_R89 5
|
||||
|
||||
/** Number of columns in BG2 */
|
||||
#define NR_LDPC_NCOL_BG2 52
|
||||
/** Number of rows in BG2 */
|
||||
#define NR_LDPC_NROW_BG2 42
|
||||
/** Number of edges/entries in BG2 */
|
||||
#define NR_LDPC_NUM_EDGE_BG2 197
|
||||
/** Number of check node (CN) groups in BG2
|
||||
A CN group is defined by its number of connected bit nodes. */
|
||||
#define NR_LDPC_NUM_CN_GROUPS_BG2 6
|
||||
/** First column in BG2 that is connected to only a single CN */
|
||||
#define NR_LDPC_START_COL_PARITY_BG2 14
|
||||
|
||||
/** Number of columns in BG2 for rate 1/5 = 10/(52-2) */
|
||||
#define NR_LDPC_NCOL_BG2_R15 NR_LDPC_NCOL_BG2
|
||||
/** Number of columns in BG2 for rate 1/3 = 10/(32-2) */
|
||||
#define NR_LDPC_NCOL_BG2_R13 32
|
||||
/** Number of columns in BG2 for rate 2/3 = 10/(17-2) */
|
||||
#define NR_LDPC_NCOL_BG2_R23 17
|
||||
|
||||
/** Number of bit node (BN) groups in BG2 for rate 1/5
|
||||
A BN group is defined by its number of connected CNs. */
|
||||
#define NR_LDPC_NUM_BN_GROUPS_BG2_R15 23
|
||||
/** Number of bit node (BN) groups in BG2 for rate 1/3 */
|
||||
#define NR_LDPC_NUM_BN_GROUPS_BG2_R13 10
|
||||
/** Number of bit node (BN) groups in BG2 for rate 2/3 */
|
||||
#define NR_LDPC_NUM_BN_GROUPS_BG2_R23 6
|
||||
|
||||
/** Worst case size of the CN processing buffer */
|
||||
#define NR_LDPC_SIZE_CN_PROC_BUF NR_LDPC_NUM_EDGE_BG1*NR_LDPC_ZMAX
|
||||
/** Worst case size of the BN processing buffer */
|
||||
#define NR_LDPC_SIZE_BN_PROC_BUF NR_LDPC_NUM_EDGE_BG1*NR_LDPC_ZMAX
|
||||
|
||||
/** Maximum number of possible input LLR = NR_LDPC_NCOL_BG1*NR_LDPC_ZMAX */
|
||||
#define NR_LDPC_MAX_NUM_LLR 26112
|
||||
|
||||
// ==============================================================================
|
||||
// GLOBAL VARIABLES
|
||||
|
||||
// Aligned on 32 bytes = 256 bits for AVX2
|
||||
/** CN processing buffer */
|
||||
static int8_t cnProcBuf [NR_LDPC_SIZE_CN_PROC_BUF] __attribute__ ((aligned(32)));
|
||||
/** Buffer for CN processing results */
|
||||
static int8_t cnProcBufRes[NR_LDPC_SIZE_CN_PROC_BUF] __attribute__ ((aligned(32)));
|
||||
|
||||
/** BN processing buffer */
|
||||
static int8_t bnProcBuf [NR_LDPC_SIZE_BN_PROC_BUF] __attribute__ ((aligned(32)));
|
||||
/** Buffer for BN processing results */
|
||||
static int8_t bnProcBufRes[NR_LDPC_SIZE_BN_PROC_BUF] __attribute__ ((aligned(32)));
|
||||
|
||||
/** Buffer for LLR results */
|
||||
static int8_t llrRes [NR_LDPC_MAX_NUM_LLR] __attribute__ ((aligned(32)));
|
||||
/** LLR processing buffer */
|
||||
static int8_t llrProcBuf[NR_LDPC_MAX_NUM_LLR] __attribute__ ((aligned(32)));
|
||||
|
||||
/** Start addresses for the cnProcBuf for each CN group in BG1*/
|
||||
static const uint32_t lut_startAddrCnGroups_BG1[NR_LDPC_NUM_CN_GROUPS_BG1] = {0, 1152, 8832, 43392, 61824, 75264, 81408, 88320, 92160};
|
||||
/** Start addresses for the cnProcBuf for each CN group in BG2*/
|
||||
static const uint32_t lut_startAddrCnGroups_BG2[NR_LDPC_NUM_CN_GROUPS_BG2] = {0, 6912, 37632, 54912, 61824, 67968};
|
||||
|
||||
/** Number of BNs of CN group for BG1.
|
||||
E.g. 10 means that there is a CN group where every CN is connected to 10 BNs */
|
||||
static const uint8_t lut_numBnInCnGroups_BG1_R13[NR_LDPC_NUM_CN_GROUPS_BG1] = {3, 4, 5, 6, 7, 8, 9, 10, 19};
|
||||
/** Number of rows/CNs in every CN group for rate = 1/3 BG1, e.g. 5 rows of CNs connected to 4 BNs */
|
||||
static const uint8_t lut_numCnInCnGroups_BG1_R13[NR_LDPC_NUM_CN_GROUPS_BG1] = {1, 5, 18, 8, 5, 2, 2, 1, 4};
|
||||
/** Number of rows/CNs in every CN group for rate = 2/3 BG1, e.g. 3 rows of CNs connected to 7 BNs */
|
||||
static const uint8_t lut_numCnInCnGroups_BG1_R23[NR_LDPC_NUM_CN_GROUPS_BG1] = {1, 0, 0, 0, 3, 2, 2, 1, 4};
|
||||
/** Number of rows/CNs in every CN group for rate = 8/9 BG1, e.g. 4 rows of CNs connected to 19 BNs */
|
||||
static const uint8_t lut_numCnInCnGroups_BG1_R89[NR_LDPC_NUM_CN_GROUPS_BG1] = {1, 0, 0, 0, 0, 0, 0, 0, 4};
|
||||
|
||||
/** Number of connected BNs for every column in BG1 rate = 1/3, e.g. in first column all BNs are connected to 30 CNs */
|
||||
static const uint8_t lut_numEdgesPerBn_BG1_R13[NR_LDPC_START_COL_PARITY_BG1] = {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};
|
||||
/** Number of connected BNs for every column in BG1 rate = 2/3, e.g. in first column all BNs are connected to 12 CNs */
|
||||
static const uint8_t lut_numEdgesPerBn_BG1_R23[NR_LDPC_START_COL_PARITY_BG1] = {12, 11, 4, 5, 5, 3, 4, 5, 5, 3, 6, 6, 6, 6, 5, 3, 6, 5, 6, 4, 5, 6, 6, 3, 3, 2};
|
||||
/** Number of connected BNs for every column in BG1 rate = 8/9, e.g. in first column all BNs are connected to 5 CNs */
|
||||
static const uint8_t lut_numEdgesPerBn_BG1_R89[NR_LDPC_START_COL_PARITY_BG1] = {5, 4, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 2, 2, 2};
|
||||
|
||||
/** Number of BNs of CN group for BG2.
|
||||
E.g. 3 means that there is a CN group where every CN is connected to 3 BNs */
|
||||
static const uint8_t lut_numBnInCnGroups_BG2_R15[NR_LDPC_NUM_CN_GROUPS_BG2] = {3, 4, 5, 6, 8, 10};
|
||||
/** Number of rows/CNs in every CN group for rate = 1/5 BG2, e.g. 6 rows of CNs connected to 3 BNs */
|
||||
static const uint8_t lut_numCnInCnGroups_BG2_R15[NR_LDPC_NUM_CN_GROUPS_BG2] = {6, 20, 9, 3, 2, 2};
|
||||
/** Number of rows/CNs in every CN group for rate = 1/3 BG2, e.g. 8 rows of CNs connected to 4 BNs */
|
||||
static const uint8_t lut_numCnInCnGroups_BG2_R13[NR_LDPC_NUM_CN_GROUPS_BG2] = {0, 8, 7, 3, 2, 2};
|
||||
/** Number of rows/CNs in every CN group for rate = 2/3 BG2, e.g. 1 row of CNs connected to 4 BNs */
|
||||
static const uint8_t lut_numCnInCnGroups_BG2_R23[NR_LDPC_NUM_CN_GROUPS_BG2] = {0, 1, 0, 2, 2, 2};
|
||||
|
||||
/** Number of connected BNs for every column in BG2 rate = 1/5, e.g. in first column all BNs are connected to 22 CNs */
|
||||
static const uint8_t lut_numEdgesPerBn_BG2_R15[NR_LDPC_START_COL_PARITY_BG2] = {22, 23, 10, 5, 5, 14, 7, 13, 6, 8, 9, 16, 9, 12};
|
||||
/** Number of connected BNs for every column in BG2 rate = 1/3, e.g. in first column all BNs are connected to 14 CNs */
|
||||
static const uint8_t lut_numEdgesPerBn_BG2_R13[NR_LDPC_START_COL_PARITY_BG2] = {14, 16, 2, 4, 4, 6, 6, 8, 6, 6, 6, 13, 5, 7};
|
||||
/** Number of connected BNs for every column in BG2 rate = 2/3, e.g. in first column all BNs are connected to 6 CNs */
|
||||
static const uint8_t lut_numEdgesPerBn_BG2_R23[NR_LDPC_START_COL_PARITY_BG2] = { 6, 5, 2, 3, 3, 4, 3, 4, 3, 4, 3, 5, 2, 2};
|
||||
|
||||
// Number of groups for bit node processing
|
||||
/** Number of connected CNs for every column/BN in BG1 for rate = 1/3, Worst case is BG1 with up to 30 CNs connected to one BN
|
||||
E.g. 42 parity BNs connected to single CN */
|
||||
// BG1: 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30
|
||||
static const uint8_t lut_numBnInBnGroups_BG1_R13[NR_LDPC_NUM_BN_GROUPS_BG1_R13] = {42, 0, 0, 1, 1, 2, 4, 3, 1, 4, 3, 4, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 1};
|
||||
/** Number of connected CNs for every column/BN in BG1 for rate = 2/3 */
|
||||
static const uint8_t lut_numBnInBnGroups_BG1_R23[NR_LDPC_NUM_BN_GROUPS_BG1_R13] = { 9, 1, 5, 3, 7, 8, 0, 0, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0};
|
||||
/** Number of connected CNs for every column/BN in BG1 for rate = 8/9 */
|
||||
static const uint8_t lut_numBnInBnGroups_BG1_R89[NR_LDPC_NUM_BN_GROUPS_BG1_R13] = { 1, 3,21, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0};
|
||||
|
||||
/** Number of connected CNs for every column/BN in BG2 for rate = 1/5, Worst case is BG1 with up to 30 CNs connected to one BN
|
||||
E.g. 38 parity BNs connected to single CN */
|
||||
// BG2: 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30
|
||||
static const uint8_t lut_numBnInBnGroups_BG2_R15[NR_LDPC_NUM_BN_GROUPS_BG1_R13] = {38, 0, 0, 0, 2, 1, 1, 1, 2, 1, 0, 1, 1, 1, 0, 1, 0, 0, 0, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 0};
|
||||
/** Number of connected CNs for every column/BN in BG2 for rate = 1/3 */
|
||||
static const uint8_t lut_numBnInBnGroups_BG2_R13[NR_LDPC_NUM_BN_GROUPS_BG1_R13] = {18, 1, 0, 2, 1, 5, 1, 1, 0, 0, 0, 0, 1, 1, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0};
|
||||
/** Number of connected CNs for every column/BN in BG2 for rate = 2/3 */
|
||||
static const uint8_t lut_numBnInBnGroups_BG2_R23[NR_LDPC_NUM_BN_GROUPS_BG1_R13] = { 3, 3, 5, 3, 2, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0};
|
||||
|
||||
// Start addresses for the bnProcBuf for each BN group
|
||||
// BG1
|
||||
/** Start address for every BN group within the BN processing buffer for BG1 rate = 1/3 */
|
||||
static const uint32_t lut_startAddrBnGroups_BG1_R13[NR_LDPC_NUM_BN_GROUPS_BG1_R13] = {0, 16128, 17664, 19584, 24192, 34944, 44160, 47616, 62976, 75648, 94080, 99072, 109824};
|
||||
/** Start address for every BN group within the BN processing buffer for BG1 rate = 2/3 */
|
||||
static const uint32_t lut_startAddrBnGroups_BG1_R23[NR_LDPC_NUM_BN_GROUPS_BG1_R23] = {0, 3456, 4224, 9984, 14592, 28032, 46464, 50688};
|
||||
/** Start address for every BN group within the BN processing buffer for BG1 rate = 8/9 */
|
||||
static const uint32_t lut_startAddrBnGroups_BG1_R89[NR_LDPC_NUM_BN_GROUPS_BG1_R89] = {0, 384, 2688, 26880, 28416};
|
||||
|
||||
/** Start address for every BN group within the LLR processing buffer for BG1 rate = 1/3 */
|
||||
static const uint16_t lut_startAddrBnGroupsLlr_BG1_R13[NR_LDPC_NUM_BN_GROUPS_BG1_R13] = {0, 16128, 16512, 16896, 17664, 19200, 20352, 20736, 22272, 23424, 24960, 25344, 25728};
|
||||
/** Start address for every BN group within the LLR processing buffer for BG1 rate = 2/3 */
|
||||
static const uint16_t lut_startAddrBnGroupsLlr_BG1_R23[NR_LDPC_NUM_BN_GROUPS_BG1_R23] = {0, 3456, 3840, 5760, 6912, 9600, 12672, 13056};
|
||||
/** Start address for every BN group within the LLR processing buffer for BG1 rate = 8/9 */
|
||||
static const uint16_t lut_startAddrBnGroupsLlr_BG1_R89[NR_LDPC_NUM_BN_GROUPS_BG1_R89] = {0, 384, 1536, 9600, 9984};
|
||||
|
||||
// BG2
|
||||
/** Start address for every BN group within the BN processing buffer for BG2 rate = 1/5 */
|
||||
static const uint32_t lut_startAddrBnGroups_BG2_R15[NR_LDPC_NUM_BN_GROUPS_BG2_R15] = {0, 14592, 18432, 20736, 23424, 26496, 33408, 37248, 41856, 46848, 52224, 58368, 66816};
|
||||
/** Start address for every BN group within the BN processing buffer for BG2 rate = 1/3 */
|
||||
static const uint32_t lut_startAddrBnGroups_BG2_R13[NR_LDPC_NUM_BN_GROUPS_BG2_R13] = {0, 6912, 7680, 10752, 12672, 24192, 26880, 29952, 34944, 40320};
|
||||
/** Start address for every BN group within the BN processing buffer for BG2 rate = 2/3 */
|
||||
static const uint32_t lut_startAddrBnGroups_BG2_R23[NR_LDPC_NUM_BN_GROUPS_BG2_R23] = {0, 1152, 3456, 9216, 13824, 17664};
|
||||
|
||||
/** Start address for every BN group within the LLR processing buffer for BG2 rate = 1/5 */
|
||||
static const uint16_t lut_startAddrBnGroupsLlr_BG2_R15[NR_LDPC_NUM_BN_GROUPS_BG2_R15] = {0, 14592, 15360, 15744, 16128, 16512, 17280, 17664, 18048, 18432, 18816, 19200, 19584};
|
||||
/** Start address for every BN group within the LLR processing buffer for BG2 rate = 1/3 */
|
||||
static const uint16_t lut_startAddrBnGroupsLlr_BG2_R13[NR_LDPC_NUM_BN_GROUPS_BG2_R13] = {0, 6912, 7296, 8064, 8448, 10368, 10752, 11136, 11520, 11904};
|
||||
/** Start address for every BN group within the LLR processing buffer for BG2 rate = 2/3 */
|
||||
static const uint16_t lut_startAddrBnGroupsLlr_BG2_R23[NR_LDPC_NUM_BN_GROUPS_BG2_R23] = {0, 1152, 2304, 4224, 5376, 6144};
|
||||
|
||||
/** Vector of 32 '1' in int8 for application with AVX2 */
|
||||
static const int8_t ones256_epi8[32] __attribute__ ((aligned(32))) = {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};
|
||||
/** Vector of 32 '0' in int8 for application with AVX2 */
|
||||
static const int8_t zeros256_epi8[32] __attribute__ ((aligned(32))) = {0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0};
|
||||
/** Vector of 32 '127' in int8 for application with AVX2 */
|
||||
static const int8_t maxLLR256_epi8[32] __attribute__ ((aligned(32))) = {127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127};
|
||||
|
||||
#endif
|
||||
File diff suppressed because it is too large
Load Diff
File diff suppressed because one or more lines are too long
File diff suppressed because one or more lines are too long
File diff suppressed because one or more lines are too long
File diff suppressed because one or more lines are too long
File diff suppressed because one or more lines are too long
File diff suppressed because one or more lines are too long
File diff suppressed because one or more lines are too long
File diff suppressed because one or more lines are too long
File diff suppressed because one or more lines are too long
File diff suppressed because one or more lines are too long
File diff suppressed because one or more lines are too long
File diff suppressed because one or more lines are too long
File diff suppressed because one or more lines are too long
File diff suppressed because one or more lines are too long
File diff suppressed because one or more lines are too long
File diff suppressed because one or more lines are too long
File diff suppressed because one or more lines are too long
File diff suppressed because one or more lines are too long
File diff suppressed because one or more lines are too long
File diff suppressed because one or more lines are too long
File diff suppressed because one or more lines are too long
File diff suppressed because one or more lines are too long
File diff suppressed because one or more lines are too long
File diff suppressed because one or more lines are too long
File diff suppressed because one or more lines are too long
File diff suppressed because one or more lines are too long
File diff suppressed because one or more lines are too long
File diff suppressed because one or more lines are too long
File diff suppressed because one or more lines are too long
File diff suppressed because one or more lines are too long
File diff suppressed because one or more lines are too long
File diff suppressed because one or more lines are too long
File diff suppressed because one or more lines are too long
File diff suppressed because one or more lines are too long
File diff suppressed because one or more lines are too long
File diff suppressed because one or more lines are too long
File diff suppressed because one or more lines are too long
File diff suppressed because one or more lines are too long
File diff suppressed because one or more lines are too long
File diff suppressed because one or more lines are too long
File diff suppressed because one or more lines are too long
File diff suppressed because one or more lines are too long
File diff suppressed because one or more lines are too long
File diff suppressed because one or more lines are too long
File diff suppressed because one or more lines are too long
File diff suppressed because one or more lines are too long
File diff suppressed because one or more lines are too long
File diff suppressed because one or more lines are too long
File diff suppressed because one or more lines are too long
File diff suppressed because one or more lines are too long
File diff suppressed because one or more lines are too long
File diff suppressed because one or more lines are too long
File diff suppressed because one or more lines are too long
File diff suppressed because one or more lines are too long
File diff suppressed because one or more lines are too long
File diff suppressed because one or more lines are too long
File diff suppressed because one or more lines are too long
File diff suppressed because one or more lines are too long
File diff suppressed because one or more lines are too long
File diff suppressed because one or more lines are too long
File diff suppressed because one or more lines are too long
File diff suppressed because one or more lines are too long
File diff suppressed because one or more lines are too long
File diff suppressed because one or more lines are too long
File diff suppressed because one or more lines are too long
File diff suppressed because one or more lines are too long
File diff suppressed because one or more lines are too long
File diff suppressed because one or more lines are too long
File diff suppressed because one or more lines are too long
File diff suppressed because one or more lines are too long
File diff suppressed because one or more lines are too long
File diff suppressed because one or more lines are too long
File diff suppressed because one or more lines are too long
File diff suppressed because one or more lines are too long
File diff suppressed because one or more lines are too long
File diff suppressed because one or more lines are too long
File diff suppressed because one or more lines are too long
File diff suppressed because one or more lines are too long
@@ -0,0 +1 @@
|
||||
static const uint32_t lut_cn2bnProcBuf_BG1_Z2_R13[632] = {111361, 111360, 100224, 100225, 0, 1, 105985, 105984, 107904, 107905, 120193, 120192, 120577, 120576, 109441, 109440, 40704, 40705, 74498, 74499, 43011, 43010, 47233, 47232, 43009, 43008, 41858, 41859, 19200, 19201, 61447, 61446, 23425, 23424, 92547, 92546, 46, 47, 66, 67, 72, 73, 76, 77, 82, 83, 116736, 116737, 105216, 105217, 105601, 105600, 117505, 117504, 117888, 117889, 106368, 106369, 118272, 118273, 106753, 106752, 118657, 118656, 107136, 107137, 119041, 119040, 107521, 107520, 119425, 119424, 119809, 119808, 108289, 108288, 108673, 108672, 109057, 109056, 120961, 120960, 87940, 87941, 30337, 30336, 39552, 39553, 31873, 31872, 46849, 46848, 58370, 58371, 89475, 89474, 87937, 87936, 89476, 89477, 33409, 33408, 89473, 89472, 41857, 41856, 61443, 61442, 30339, 30338, 73345, 73344, 74497, 74496, 61445, 61444, 92545, 92544, 72194, 72195, 87938, 87939, 86400, 86401, 46465, 46464, 31879, 31878, 97536, 97537, 73346, 73347, 89478, 89479, 59907, 59906, 61441, 61440, 31876, 31877, 91013, 91012, 33413, 33412, 91011, 91010, 91009, 91008, 31875, 31874, 98689, 98688, 33411, 33410, 58374, 58375, 97152, 97153, 56834, 56835, 30341, 30340, 73349, 73348, 21890, 21891, 21889, 21888, 22659, 22658, 22657, 22656, 74501, 74500, 59911, 59910, 91015, 91014, 97920, 97921, 92549, 92548, 33414, 33415, 98305, 98304, 23427, 23426, 92551, 92550, 36, 37, 38, 39, 42, 43, 44, 45, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64, 65, 68, 69, 70, 71, 74, 75, 78, 79, 80, 81, 114433, 114432, 103680, 103681, 115585, 115584, 104064, 104065, 115968, 115969, 116353, 116352, 104833, 104832, 117120, 117121, 68736, 68737, 69889, 69888, 55298, 55299, 86405, 86404, 104449, 104448, 71040, 71041, 17281, 17280, 72192, 72193, 83329, 83328, 56833, 56832, 58373, 58372, 71042, 71043, 84864, 84865, 28803, 28802, 59909, 59908, 46080, 46081, 40709, 40708, 41861, 41860, 56839, 56838, 96769, 96768, 40707, 40706, 58369, 58368, 43013, 43012, 59904, 59905, 18816, 18817, 84870, 84871, 71044, 71045, 30343, 30342, 86402, 86403, 86406, 86407, 72197, 72196, 87942, 87943, 18, 19, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 40, 41, 112512, 112513, 102144, 102145, 114048, 114049, 114816, 114817, 115200, 115201, 100992, 100993, 28801, 28800, 102912, 102913, 84869, 84868, 103296, 103297, 45313, 45312, 45697, 45696, 83331, 83330, 28804, 28805, 84867, 84866, 80257, 80256, 81792, 81793, 55296, 55297, 56837, 56836, 69891, 69890, 38403, 38402, 39555, 39554, 68738, 68739, 55302, 55303, 96385, 96384, 52226, 52227, 53762, 53763, 96001, 96000, 69893, 69892, 21123, 21122, 6, 7, 12, 13, 16, 17, 20, 21, 22, 23, 111745, 111744, 113665, 113664, 100608, 100609, 102529, 102528, 66432, 66433, 83333, 83332, 80261, 80260, 55300, 55301, 52228, 52229, 68740, 68741, 66437, 66436, 83334, 83335, 80263, 80262, 18432, 18433, 2, 3, 14, 15, 112129, 112128, 113281, 113280, 38400, 38401, 101760, 101761, 80258, 80259, 81795, 81794, 52225, 52224, 53761, 53760, 66434, 66435, 67586, 67587, 52231, 52230, 53766, 53767, 95233, 95232, 95617, 95616, 38405, 38404, 39556, 39557, 4, 5, 10, 11, 112896, 112897, 101376, 101377, 67585, 67584, 81797, 81796, 53764, 53765, 28806, 28807, 67589, 67588, 81799, 81798, 21120, 21121, 8, 9, 109824, 109825, 110208, 110209, 110592, 110593, 110977, 110976, 99073, 99072, 25729, 25728, 99457, 99456, 99841, 99840, 24192, 24193, 64129, 64128, 27265, 27264, 65280, 65281, 62977, 62976, 44160, 44161, 44545, 44544, 44928, 44929, 16128, 16129, 16513, 16512, 16897, 16896, 37248, 37249, 34944, 34945, 75648, 75649, 36097, 36096, 78721, 78720, 24195, 24194, 34946, 34947, 77185, 77184, 37251, 37250, 75651, 75650, 25731, 25730, 36099, 36098, 78722, 78723, 47616, 47617, 49152, 49153, 27267, 27266, 50688, 50689, 75653, 75652, 77188, 77189, 77187, 77186, 78724, 78725, 62979, 62978, 47619, 47618, 64130, 64131, 65283, 65282, 24197, 24196, 25732, 25733, 49155, 49154, 50691, 50690, 47621, 47620, 49156, 49157, 27269, 27268, 50692, 50693, 94080, 94081, 47623, 47622, 49159, 49158, 50694, 50695, 24199, 24198, 25735, 25734, 94465, 94464, 94848, 94849, 34949, 34948, 64132, 64133, 27271, 27270, 37252, 37253, 62981, 62980, 77190, 77191, 36101, 36100, 65284, 65285, 75655, 75654, 18048, 18049, 20352, 20353, 78727, 78726, 17664, 17665, 19584, 19585, 19586, 19587, 20354, 20355};
|
||||
@@ -0,0 +1 @@
|
||||
static const uint32_t lut_cn2bnProcBuf_BG1_Z2_R23[288] = {52225, 52224, 47616, 47617, 0, 1, 53376, 53377, 49536, 49537, 54912, 54913, 48384, 48385, 13441, 13440, 50304, 50305, 22659, 22658, 25347, 25346, 43393, 43392, 22661, 22660, 25348, 25349, 43394, 43395, 22663, 22662, 25351, 25350, 43398, 43399, 22664, 22665, 25352, 25353, 43403, 43402, 6, 7, 12, 13, 16, 17, 52609, 52608, 54529, 54528, 48000, 48001, 49921, 49920, 22656, 22657, 43397, 43396, 37253, 37252, 43400, 43401, 37256, 37257, 43404, 43405, 37261, 37260, 43406, 43407, 37263, 37262, 8070, 8071, 2, 3, 14, 15, 52993, 52992, 54145, 54144, 13442, 13443, 49152, 49153, 37248, 37249, 40321, 40320, 37251, 37250, 40323, 40322, 37254, 37255, 40326, 40327, 22667, 22666, 25354, 25355, 37259, 37258, 40331, 40330, 22669, 22668, 25356, 25357, 4, 5, 10, 11, 53760, 53761, 48768, 48769, 25345, 25344, 40325, 40324, 40328, 40329, 13444, 13445, 40333, 40332, 40335, 40334, 8072, 8073, 8, 9, 50688, 50689, 51072, 51073, 51456, 51457, 51841, 51840, 46465, 46464, 11137, 11136, 46849, 46848, 47233, 47232, 9984, 9985, 17281, 17280, 12289, 12288, 19968, 19969, 14593, 14592, 14594, 14595, 17283, 17282, 19970, 19971, 4224, 4225, 6145, 6144, 8065, 8064, 12290, 12291, 9986, 9987, 14596, 14597, 11139, 11138, 19973, 19972, 4227, 4226, 14598, 14599, 17285, 17284, 19975, 19974, 28033, 28032, 6147, 6146, 17287, 17286, 34176, 34177, 28034, 28035, 31106, 31107, 8067, 8066, 34178, 34179, 28037, 28036, 31108, 31109, 31105, 31104, 34180, 34181, 28039, 28038, 14601, 14600, 31110, 31111, 34183, 34182, 4229, 4228, 6148, 6149, 17289, 17288, 19977, 19976, 28041, 28040, 31112, 31113, 8069, 8068, 34184, 34185, 28042, 28043, 14603, 14602, 17291, 17290, 19978, 19979, 9989, 9988, 11141, 11140, 31115, 31114, 34186, 34187, 14605, 14604, 31116, 31117, 12293, 12292, 19980, 19981, 28045, 28044, 31118, 31119, 17293, 17292, 34188, 34189, 28047, 28046, 6150, 6151, 6152, 6153, 34191, 34190, 4230, 4231, 4232, 4233, 3456, 3457, 3840, 3841};
|
||||
@@ -0,0 +1 @@
|
||||
static const uint32_t lut_cn2bnProcBuf_BG1_Z2_R89[158] = {29953, 29952, 28032, 28033, 0, 1, 28416, 28417, 28800, 28801, 29184, 29185, 29569, 29568, 26881, 26880, 10753, 10752, 27265, 27264, 27649, 27648, 2688, 2689, 10755, 10754, 18817, 18816, 18818, 18819, 2691, 2690, 2692, 2693, 10757, 10756, 18820, 18821, 2694, 2695, 10759, 10758, 18823, 18822, 18824, 18825, 2696, 2697, 2698, 2699, 10761, 10760, 18827, 18826, 2703, 2702, 2700, 2701, 10763, 10762, 18829, 18828, 2705, 2704, 10767, 10766, 10765, 10764, 18832, 18833, 2706, 2707, 10770, 10771, 18831, 18830, 18834, 18835, 2709, 2708, 10772, 10773, 10769, 10768, 18836, 18837, 2711, 2710, 2713, 2712, 10774, 10775, 18839, 18838, 2715, 2714, 10778, 10779, 10777, 10776, 18841, 18840, 2717, 2716, 10780, 10781, 18843, 18842, 18844, 18845, 2720, 2721, 2719, 2718, 10783, 10782, 18846, 18847, 2723, 2722, 10787, 10786, 10785, 10784, 18848, 18849, 2725, 2724, 10790, 10791, 18851, 18850, 18852, 18853, 2727, 2726, 10792, 10793, 10789, 10788, 18854, 18855, 2729, 2728, 1536, 1537, 1538, 1539, 18857, 18856, 384, 385, 386, 387, 388, 389, 1540, 1541};
|
||||
File diff suppressed because one or more lines are too long
File diff suppressed because one or more lines are too long
File diff suppressed because one or more lines are too long
File diff suppressed because one or more lines are too long
File diff suppressed because one or more lines are too long
File diff suppressed because one or more lines are too long
Some files were not shown because too many files have changed in this diff Show More
Reference in New Issue
Block a user