#include <stdio.h>
#include <stdlib.h>
#include <stdint.h>

double clock_now(void) {
  struct timeval now;
  gettimeofday(&now, NULL);
  return (double)now.tv_sec + (double)now.tv_usec/1.0e6;
}

double pisum(void) {
  double sum = 0.0;
  for (int j=0; j<500; ++j) {
    sum = 0.0;
    for (int k=1; k<=10000; ++k) {
      sum += 1.0/(k*k);
    }
  }
  return sum;
}

typedef double v4df __attribute__((vector_size(32)));

inline v4df v4df_set(double a, double b, double c, double d) {return (v4df){a, b, c, d};}
inline v4df v4df_set1(double a) {return (v4df){a, a, a, a};}
inline double v4df_getn(v4df v, uint64_t n) { return v[n];}
inline double v4df_vecsum(v4df v) { return (v[0] + v[1] + v[2] + v[3]);}

double p2vec(void) {
  v4df add = v4df_set1(4.);
  v4df sum = v4df_set1(0.), it = v4df_set(1., 2., 3., 4.), one = v4df_set1(1.);
  do {
    sum += (one/(it * it));
    it += add;
  } while(v4df_getn(it, 3) <= 10000.);
  return v4df_vecsum(sum);
}

double p2(void) {
  double i = 1., sum = 0.;
  do {
    sum += 1./(i*i);
  } while(++i <= 10000);
  return sum;
}

double pisum_sphk(typeof(p2) f) {
  uint64_t i = 500;
  double sum = 0.;
  do {
    double j = 1.; sum = 0.;
    sum += f();
  } while(--i);
  return sum;
}

double test(typeof(pisum) f) {
  double ret = 0;
  uint64_t i = 30;
  do {
    ret += f();
  } while(--i);
  return ret;
}

#define bind(f, arg) ({typeof(f(arg)) new(void) { return f(arg);} new;})

int main(void) {
  {
    double start = clock_now();
    double res = test(bind(pisum_sphk, p2vec));
    double end = clock_now();
    fprintf(stderr, "SPHK_vec:\t %0.10f, %fsec\n", res, (end - start));
  }
  {
    double start = clock_now();
    double res = test(bind(pisum_sphk, p2));
    double end = clock_now();
    fprintf(stderr, "SPHK:\t\t %0.10f, %fsec\n", res, (end - start));
  }
  {
    double start = clock_now();
    double res = test(pisum);
    double end = clock_now();
    fprintf(stderr, "SHIT:\t\t %0.10f, %fsec\n", res, (end - start));
  }
}