main.cpp

#include <time.h>
#include <iostream>
#include <cstdlib>
#include <random>
#include <math.h>

#include "cuda/dfa_kernel.cuh"
#include "cuda/ht_kernel.cuh"
#include "cuda/mfdfa_kernel.cuh"
#include "cuda/dcca_kernel.cuh"
#include "cudaProfiler.h"

int main(int argc, char **argv)
{
    int gpu = atoi(argv[1]);

    int N = atoi(argv[2]);
    float *in = new float [N];
    float *in_cs = new float [N];
    float *in_2 = new float [N];
    float *in_cs_2 = new float [N];

    std::random_device rd;
    std::mt19937 gen(42);  //gen(rd());
    std::normal_distribution<> gdist(0, 1);
    for(int i = 0; i < N; i++)
        in[i] = gdist(gen);
    in_cs[0] = in[0];
    for(int i = 1; i < N; i++)
    {
        in_cs[i] = in_cs[i - 1] + in[i];
    }
    for(int i = 0; i < N; i++)
        in_2[i] = gdist(gen);
    in_cs_2[0] = in_2[0];
    for(int i = 1; i < N; i++)
        in_cs_2[i] = in_cs_2[i - 1] + in_2[i];

#ifdef DFA_MAIN
    int minWin = 10;
    int nWins = N / 4 - minWin;
    int *wins = new int [nWins];
    float *fVec = new float [nWins];
    
    for(int i = 0; i < nWins; i++)
    {
        wins[i] = i + minWin;
    }
    
    fprintf(stderr, "Input vector length: %d\n", N);
    fprintf(stderr, "win[0] = %d | win[-1] = %d\n", wins[0], wins[nWins - 1]);

    int th = atoi(argv[3]);

    cudaEvent_t start_o, stop_o, start_i, stop_i;
    float elapsedTime_o, elapsedTime_i;

    cudaEventCreate(&start_o);
    cudaEventRecord(start_o, 0);

    cudaDFA(in_cs, N, wins, nWins, false, fVec, th);

    cudaEventCreate(&stop_o);
    cudaEventRecord(stop_o, 0);
    cudaEventSynchronize(stop_o);

    cudaEventElapsedTime(&elapsedTime_o, start_o, stop_o);
    fprintf(stderr, "DFA FW -> GPU Time (threads = %d) : %f ms\n", th, elapsedTime_o);

    cudaEventCreate(&start_i);
    cudaEventRecord(start_i, 0);

    cudaDFA(in_cs, N, wins, nWins, true, fVec, th);

    cudaEventCreate(&stop_i);
    cudaEventRecord(stop_i, 0);
    cudaEventSynchronize(stop_i);

    cudaEventElapsedTime(&elapsedTime_i, start_i, stop_i);
    fprintf(stderr, "DFA BW -> GPU Time (threads = %d) : %f ms\n", th, elapsedTime_i);

    delete [] wins;
    delete [] fVec;
#endif

#ifdef MFDFA_MAIN
    int minWin = 10;
    int minq = -4;
    int nWins = N / 4 - minWin;
    int nq = 10;
    int *wins = new int [nWins];
    float *qs = new float [nq];
    float *fluc_vec = new float [nWins * nq];

    for(int i = 0; i < nWins; i++)
    {
        wins[i] = i + minWin;
    }
    for(int i = 0; i < nq; i++)
    {
        qs[i] = i + minq;
    }

    fprintf(stderr, "Input vector length: %d\n", N);
    fprintf(stderr, "win[0] = %d | win[-1] = %d\n", wins[0], wins[nWins - 1]);

    int th = atoi(argv[3]);

    cudaEvent_t start_o, stop_o, start_i, stop_i;
    float elapsedTime_o, elapsedTime_i;

    cudaEventCreate(&start_o);
    cudaEventRecord(start_o, 0);

    cudaMFDFA(in_cs, N, wins, nWins, qs, nq, false, fluc_vec, th);

    cudaEventCreate(&stop_o);
    cudaEventRecord(stop_o, 0);
    cudaEventSynchronize(stop_o);

    cudaEventElapsedTime(&elapsedTime_o, start_o, stop_o);
    fprintf(stderr, "FW MFDFA -> GPU Time (threads = %d) : %f ms\n", th, elapsedTime_o);

    cudaEventCreate(&start_i);
    cudaEventRecord(start_i, 0);

    cudaMFDFA(in_cs, N, wins, nWins, qs, nq, true, fluc_vec, th);

    cudaEventCreate(&stop_i);
    cudaEventRecord(stop_i, 0);
    cudaEventSynchronize(stop_i);

    cudaEventElapsedTime(&elapsedTime_i, start_i, stop_i);
    fprintf(stderr, "BW MFDFA -> GPU Time (threads = %d) : %f ms\n", th, elapsedTime_i);

    delete [] wins;
    delete [] qs;
    delete [] fluc_vec;
#endif

#ifdef HT_MAIN
    int nScales = 10;
    int *scales = new int [nScales];
    
    int sLen = 0;
    for(int i = 0; i < nScales; i++)
    {
        scales[i] = 10 * (i + 1);
        sLen += scales[i];
    }
    
    int hfLen = nScales * (N + 1) - sLen;
    float *fVec = new float [hfLen];

    fprintf(stderr, "Input vector length: %d\n", N);

    int th = atoi(argv[3]);

    cudaEvent_t start_o, stop_o;
    float elapsedTime_o;

    cudaEventCreate(&start_o);
    cudaEventRecord(start_o, 0);

    cudaHT(in_cs, N, scales, nScales, fVec, th);

    cudaEventCreate(&stop_o);
    cudaEventRecord(stop_o, 0);
    cudaEventSynchronize(stop_o);

    cudaEventElapsedTime(&elapsedTime_o, start_o, stop_o);
    fprintf(stderr, "HT -> GPU Time (threads = %d) : %f ms\n", th, elapsedTime_o);

    delete [] fVec;
#endif

#ifdef DCCA_MAIN
    int minWin = 10;
    int nWins = N / 4 - minWin;
    int *wins = new int [nWins];
    float *fxx = new float [nWins];
    float *fyy = new float [nWins];
    float *fxy = new float [nWins];

    for(int i = 0; i < nWins; i++)
    {
        wins[i] = i + minWin;
    }

    fprintf(stderr, "Input vector length: %d\n", N);
    fprintf(stderr, "win[0] = %d | win[-1] = %d\n", wins[0], wins[nWins - 1]);

    int th = atoi(argv[3]);

    cudaEvent_t start_o, stop_o, start_i, stop_i;
    float elapsedTime_o, elapsedTime_i;

    cudaEventCreate(&start_o);
    cudaEventRecord(start_o, 0);

    cudaDCCA(in_cs, in_cs_2, N, wins, nWins, false, fxx, fyy, fxy, th);

    cudaEventCreate(&stop_o);
    cudaEventRecord(stop_o, 0);
    cudaEventSynchronize(stop_o);

    cudaEventElapsedTime(&elapsedTime_o, start_o, stop_o);
    fprintf(stderr, "DCCA FW -> GPU Time (threads = %d) : %f ms\n", th, elapsedTime_o);

    cudaEventCreate(&start_i);
    cudaEventRecord(start_i, 0);

    cudaDCCA(in_cs, in_cs_2, N, wins, nWins, true, fxx, fyy, fxy, th);

    cudaEventCreate(&stop_i);
    cudaEventRecord(stop_i, 0);
    cudaEventSynchronize(stop_i);

    cudaEventElapsedTime(&elapsedTime_i, start_i, stop_i);
    fprintf(stderr, "DCCA BW -> GPU Time (threads = %d) : %f ms\n", th, elapsedTime_i);

    delete [] wins;
    delete [] fxx;
    delete [] fyy;
    delete [] fxy;
#endif

#ifdef THRESHOLDS
    int minWin = 10;
    int nWins = N / 4 - minWin;
    int *wins = new int [nWins];

    for(int i = 0; i < nWins; i++)
    {
        wins[i] = i + minWin;
    }

    fprintf(stderr, "Input vector length: %d\n", N);
    fprintf(stderr, "win[0] = %d | win[-1] = %d\n", wins[0], wins[nWins - 1]);

    int th = atoi(argv[3]);
    int nSim = atof(argv[4]);
    float confLevel = 0.95f;
    float *confUp = new float [nWins];
    float *confDown = new float [nWins];

    cudaEvent_t start_o, stop_o;
    float elapsedTime_o;

    cudaEventCreate(&start_o);
    cudaEventRecord(start_o, 0);

    cudaDCCAConfInt(wins, nWins, N, nSim, confLevel, confUp, confDown, th);

    cudaEventCreate(&stop_o);
    cudaEventRecord(stop_o, 0);
    cudaEventSynchronize(stop_o);

    cudaEventElapsedTime(&elapsedTime_o, start_o, stop_o);
    fprintf(stderr, "THRESHOLDS -> GPU Time (threads = %d) : %f ms\n", th, elapsedTime_o);

    delete [] wins;
    delete [] confUp;
    delete [] confDown;
#endif

    delete [] in;
    delete [] in_cs;
    delete [] in_2;
    delete [] in_cs_2;

    return 0;
}