chundoong-lab-ta/SHPC2022/hw3_answer/prefix_sum/prefix_sum.c

#include <immintrin.h>
#include <math.h>

void prefix_sum_sequential(double *out, const double *in, int N) {
  out[0] = in[0];
  for (int i = 1; i < N; ++i) {
    out[i] = in[i] + out[i - 1];
  }
}

void prefix_sum_parallel(double *out, const double *in, int N) {

#pragma omp parallel for
  for (int i = 0; i < N; ++i) {
    out[i] = in[i];
  }

  int st = 1;
  while (st < N) {
#pragma omp parallel for
    for (int i = 0; i < N; ++i) {
      if ((i + 1) % (2 * st) == 0) {
        out[i] = out[i - st] + out[i];
      }
    }
    st *= 2;
  }

  // printf("st-1: %d\n", st-1);
  out[st - 1] = 0;
  st /= 2;

  while (st > 0) {
#pragma omp parallel for
    for (int i = 0; i < N; ++i) {
      if ((i + 1) % (2 * st) == 0) {
        double tmp = out[i];
        out[i] = out[i - st] + out[i];
        out[i - st] = tmp;
      }
    }
    st /= 2;
  }

#pragma omp parallel for
  for (int i = 0; i < N; ++i) {
    out[i] = out[i] + in[i];
  }
}