#include <boost/thread/thread.hpp>
using namespace boost;
#include <xmmintrin.h>
#define PROC (8)
struct thread1 {
float *x, *y, *ep;
int i0, i1, p;
thread1(float *xx, float *yy, float *ee, int pp, int ii0, int ii1) :
x(xx), y(yy), ep(ee), p(pp), i0(ii0), i1(ii1) {}
void operator()() {
// do computation
float ve[4] = {0, 0, 0, 0};
float e = 0;
for (int i=i0; i<i1; i+=4) {
float half =.5;
_mm_store_ps(&x[i],
_mm_mul_ps(_mm_load1_ps(&half),
_mm_add_ps(_mm_loadu_ps(&y[i+1]),
_mm_loadu_ps(&y[i-1]))));
_mm_store_ps(&ve[0],
_mm_add_ps(_mm_load_ps(&ve[0]),
_mm_mul_ps(_mm_load_ps(&y[i]),
_mm_load_ps(&y[i]))));
}
e += ve[0] + ve[1] + ve[2] + ve[3];
*ep = e;
}
};
int main(int argc, char *argv[]) {
int n = ...;
float *x, *y;
x = new float[n+1];
y = new float[n+1];
... // fill x, y
float e = 0;
float e_vec[PROC];
thread_group grp;
// start threads and wait for termination
for (int i=0; i<PROC; ++i) {
thread1 t(x, y, &e_vec[i], i,
1+((n-1)*i)/PROC, 1+((n-1)*(i+1))/PROC);
grp.create_thread(t);
}
grp.join_all();
for (int i=0; i<PROC; ++i)
e += e_vec[i];
... // output x, e
delete[] x, y;
return 0;
}
|