int main(int argc,char **argv)
{
if(argc < 2)
{
fprintf(stderr,"Usage...\n");
exit(1);
}
printf("%lld\n",mydu(argv[1])/2);
exit(0);
}
inline void
time_step_parallel(zfp::array2d& u, const Constants& c)
{
#ifdef _OPENMP
// flush shared cache to ensure cache consistency across threads
u.flush_cache();
// compute du/dt in parallel
zfp::array2d du(c.nx, c.ny, u.rate(), 0, u.cache_size());
#pragma omp parallel
{
// create read-only private view of entire array u
zfp::array2d::private_const_view myu(&u);
// create read-write private view into rectangular subset of du
zfp::array2d::private_view mydu(&du);
mydu.partition(omp_get_thread_num(), omp_get_num_threads());
// process rectangular region owned by this thread
for (uint j = 0; j < mydu.size_y(); j++) {
int y = mydu.global_y(j);
if (1 <= y && y <= c.ny - 2)
for (uint i = 0; i < mydu.size_x(); i++) {
int x = mydu.global_x(i);
if (1 <= x && x <= c.nx - 2) {
double uxx = (myu(x - 1, y) - 2 * myu(x, y) + myu(x + 1, y)) / (c.dx * c.dx);
double uyy = (myu(x, y - 1) - 2 * myu(x, y) + myu(x, y + 1)) / (c.dy * c.dy);
mydu(i, j) = c.dt * c.k * (uxx + uyy);
}
}
}
// compress all private cached blocks to shared storage
mydu.flush_cache();
}
// take forward Euler step in serial
for (uint i = 0; i < u.size(); i++)
u[i] += du[i];
#endif
}
int64_t mydu(const char *path)
{// ../a/b/c/d/e
struct stat statres;
char nextpath[PATHSIZE];
glob_t globres;
int i;
int64_t sum = 0;
if(lstat(path,&statres) < 0)
{
perror("lstat()");
exit(1);
}
if(!S_ISDIR(statres.st_mode))
return statres.st_blocks;
// is a dir
// "../a/b/c/d/e/*" "../a/b/c/d/e/.*"
strncpy(nextpath,path,PATHSIZE);
strncat(nextpath,"/*",PATHSIZE);
glob(nextpath,0,NULL,&globres);
/*if error*/
strncpy(nextpath,path,PATHSIZE);
strncat(nextpath,"/.*",PATHSIZE);
glob(nextpath,GLOB_APPEND,NULL,&globres);
for(i = 0 ; i < globres.gl_pathc; i++)
{
if(path_noloop(globres.gl_pathv[i]))
sum += mydu(globres.gl_pathv[i]);
}
sum += statres.st_blocks;
return sum;
}
