/*
COPYRIGHT (2011-2012) by:
Kevin Marco Erler (author), http://www.kevinerler.de
AIU-FSU Jena (co-owner), http://www.astro.uni-jena.de
SBSZ Jena-Göschwitz (co-owner), http://www.sbsz-jena.de
BSZ-Hermsdorf (co-owner), http://www.bszh.de
Advanced Licensing (dual license: COPYRIGHT and following licenses):
License (international): CC-BY v3.0-unported or later - link: http://creativecommons.org/licenses/by/3.0/deed.en
License (Germany):       CC-BY v3.0-DE       or later - link: http://creativecommons.org/licenses/by/3.0/de/
------------------
Compilation requirements:
Packages (x86-64):
  GCC >v4.2, compat. libstdc++ and GOMP v3.0
Normal-Compile with g++-Compiler (Red Hat GCC 4.4.5-6 x86-64 tested) + OpenMP v3.0 ([lib]GOMP v3.0 x86-64 tested)
  g++ -std=c++0x -m64 -fopenmp -Wall -Wextra -pedantic -pedantic-errors -lgomp -lm -s <source.cpp> -o <dest>
Release-Compile with g++-Compiler (Red Hat GCC 4.4.5-6 x86-64 tested) + OpenMP v3.0 ([lib]GOMP v3.0 x86-64 tested)
  g++ -std=c++0x -m64 -fopenmp -Wall -Wextra -pedantic -pedantic-errors -lgomp -lm -O3 -s <source.cpp> -o <dest>
Debug-Compile with g++-Compiler (Red Hat GCC 4.4.5-6 x86-64 tested) + OpenMP v3.0 ([lib]GOMP v3.0 x86-64 tested)
  g++ -std=c++0x -m64 -fopenmp -Wall -Wextra -pedantic -pedantic-errors -lgomp -lm -g -ggdb3 <source.cpp> -o <dest>
*/

// Includes of C/C++-Librarys for INTs, REAL/FLOATs, STRINGS, Math-Calc and I/O
#include <climits>
#include <cstdint>
#include <cinttypes>
#include <cfloat>
#include <cwchar>
#include <string>  //std:string
#include <string.h>
#include <cstring>
#include <cstdlib>
#include <cstdio>
#include <iostream>
#include <sstream>
#include <cmath>

// Conditional compilation (conditional include) of the OpenMP-Mainlib for OpenMP-Support
#ifdef _OPENMP
#include <omp.h>
#endif

using namespace std;

#define free(x) free(x); *x=NULL
#define PRId128 "s"
#define PRIi128 "s"
#define PRIu128 "s"

const uint64_t UINT64_MIN     = 0;
const __int128_t INT128_MIN   = (__int128_t)((-170141183460469231731.687303715884105728) * pow(10,18));
const __int128_t INT128_MAX   = (__int128_t)(( 170141183460469231731.687303715884105727) * pow(10,18));
const __uint128_t UINT128_MAX = (__uint128_t)((340282366920938463463.374607431768211455) * pow(10,18));
const __uint128_t UINT128_MIN = 0/* * pow(10,18)*/;

std::ostream &operator<<(std::ostream &out, __uint128_t x)
{
  if(x >= 10)
  {
    out << x / 10;
  }
  return out << static_cast<unsigned>(x % 10);
}

std::ostream &operator<<(std::ostream &out, __int128_t x)
{
  if(x < 0)
  {
    out << '-';
    x = -x;
  }
  return out << static_cast<__uint128_t>(x);
}

string INT128ToSTR(__int128_t x)
{
  std::stringstream sstr;
  sstr<<x;
  return sstr.str();
}
#define INT128ToCSTR(x) (INT128ToSTR(x)).c_str()

string UINT128ToSTR(__uint128_t x)
{
  std::stringstream sstr;
  sstr<<x;
  return sstr.str();
}
#define UINT128ToCSTR(x) (UINT128ToSTR(x)).c_str()

const __uint128_t NumValues = 400000000ULL;

int main(int argc, char *argv[])
{
  // Runtime manipulation of OpenMP-state variables
  //omp_set_num_threads(4);
  omp_set_dynamic(0);

  // data declarations and implementations
  double starttime = 0.00, sdelay = 0.00, pdelay = 0.00;
  __uint128_t A_s[NumValues] = {0ULL}, \
              B_s[NumValues] = {0ULL}, \
              C_s[NumValues] = {0ULL}, \
              A_p[NumValues] = {0ULL}, \
              B_p[NumValues] = {0ULL}, \
              C_p[NumValues] = {0ULL};
  bool ResultsAreCorrect = false;

  std::cout << "Vektoraddition (1D)                                         (128-Bit)\n"
            << "=====================================================================\n"
            << "Initialisierung:";

  //--------------------------Begin: Initialization of data------------------------------------------

  for(__uint128_t i=0ULL;i<NumValues;++i)
  {
    A_s[i] = A_p[i] = B_s[i] = B_p[i] = (i+1ULL);
  }

  //--------------------------End: Initialization of data--------------------------------------------

  std::cout << "                                                 done\n"
            << "SERIELLE AUSFÜHRUNG:";

  //--------------------------Begin: CPU-serial execution of algorithm-------------------------------

  starttime = omp_get_wtime();
  //CPU-serial algorithm:
  for(__uint128_t j=0ULL;j<NumValues;++j)
  {
    C_s[j] = A_s[j] + B_s[j];
  }
  sdelay = omp_get_wtime()-starttime;
  std::cout << "                                             done\n"; //serial

  //--------------------------End: CPU-serial execution of algorithm---------------------------------

  //--------------------------Begin: CPU-parallel OpenMP-execution of algorithm----------------------

  std::cout << "PARALLELE AUSFÜHRUNG mit ";

  // create parallel region:
  #pragma omp parallel default(none) shared(std::cout, starttime, pdelay, A_p, B_p, C_p)
  {
    #pragma omp master
    {
      std::cout << omp_get_num_threads() << " Threads:";
      starttime = omp_get_wtime();
    }

    //OpenMP-CPU-parallel algorithm
    #pragma omp flush
    #pragma omp for schedule(static)
    for(__uint128_t k=0ULL;k<NumValues;++k)
    {
      C_p[k] = A_p[k] + B_p[k];
    }

    #pragma omp master
    {
      pdelay = omp_get_wtime()-starttime;
      if(omp_get_num_threads() >= 10)
      {
        std::cout << "                             done\n";  //parallel
      }
      else
      {
        std::cout << "                              done\n"; //parallel
      }
    }
  }

  //--------------------------End: CPU-parallel OpenMP-execution of algorithm------------------------

  //--------------------------Analysis of results----------------------------------------------------

  std::cout << "Überprüfe Ergebnisse:";
  for(__uint128_t l=0ULL;l<NumValues;++l)
  {
    if(C_p[l]==C_s[l])
    {
      ResultsAreCorrect = true;
    }
    else
    {
      ResultsAreCorrect = false;
      break;
    }
  }
  std::cout << "                                            done\n";

  std::cout << "\nAuswertung:\n"
            << "*********************************************************************\n"
            << "Anzahl 1D-Eingangsvektoren  A: " << NumValues << '\n'
            << "Anzahl 1D-Eingangsvektoren  B: " << NumValues << '\n'
            << "Anzahl 1D-Ergebnis-Vektoren C: " << NumValues << '\n'
            << "Seriell & parallel richtig gerechnet?:                            " << ((ResultsAreCorrect==true)?"yes\n":" no\n")
            << "Dauer - SERIELL:     " << sdelay << " sec\n"
            << "Dauer - PARALLEL:    " << pdelay << " sec\n"
            << "__________________\n"
            << "Beispiele:\n"
            << "==> 1.(1D)-Vektoraddition:\n"
            << "C1 = A1(" << A_p[0] << ") + B1(" << B_p[0] << ")\n"
            << "C1 = " << C_p[0] << "\n"
            << "==> " << NumValues << ".(1D)-Vektoraddition:\n"
            << "C" << NumValues << " = A" << NumValues << "(" << A_p[NumValues-1] << ") + B" << NumValues << "(" << B_p[NumValues-1] << ")\n"
            << "C" << NumValues << " = " << C_p[NumValues-1] << "\n"
            << "__________________"
            << "\n128-Bit-Werte:\n"
            << "INT128_MIN:                  "  << INT128_MIN << '\n'
            << "INT128_MAX:                   " << INT128_MAX << '\n'
            << "UINT128_MIN:                  " << UINT128_MIN << '\n'
            << "UINT128_MAX:                  " << UINT128_MAX << '\n';
  /*
  // Detailed output
  std::cout << "__________________\n"
            << "Ergebnisliste:\n";
  for(__uint128_t m=0ULL;m<NumValues;++m)
  {
    std::cout << "Seriell:    A" << (m+1) << " = " << A_s[m] << "    B" << (m+1) << " = " << B_s[m] << "    C" << (m+1) << " = " << C_s[m] << '\n';
    std::cout << "Parallel:   A" << (m+1) << " = " << A_p[m] << "    B" << (m+1) << " = " << B_p[m] << "    C" << (m+1) << " = " << C_p[m] << '\n';
  }
  */
  getchar();
  return 0;
}