void* thread_function(void* thread_id)
{
int *temp = (int*)thread_id;
int i, j = *temp;
int k = 0, NUM_ITER = 50;
get_data(j);
printf("done getting data\n");
pthread_barrier_wait(&barr);
#ifdef DUMP
if(j == 0)
{
m5_dump_stats(0, 0);
m5_reset_stats(0, 0);
}
#endif
for(k = 0; k < NUM_ITER; k++)
{
int half_point = DIMY/(2*NUM_TH);
int elem_to_process = DIMY/NUM_TH;
for(i = 1; i <= levels; i++)
{
haar2_step1(elem_to_process, j, half_point);
half_point = half_point >> 1;
elem_to_process = elem_to_process >> 1;
}
}
// pthread_barrier_wait(&barr);
#ifdef DUMP
if(j == 15)
{
m5_dump_stats(0, 0);
m5_reset_stats(0, 0);
}
#endif
/*
//printf("Just after barrier %d\n", j);
#ifdef DUMP
if(j == 0)
{
m5_dump_stats(0, 0);
m5_reset_stats(0, 0);
}
#endif
//printf("before writing image ID %d\n", j);
writeImage(j);
*/
}
void
do_dump_stats(int argc, char *argv[])
{
uint64_t ints[2];
parse_int_args(argc, argv, ints, 2);
m5_dump_stats(ints[0], ints[1]);
}
int main(int argc, char *argv[])
{
int rank;
int n_ranks = 16;
int n = 10;
int END = 50;
MPI_Init(&argc, &argv);
MPI_Comm_size(MPI_COMM_WORLD, &n_ranks);
MPI_Comm_rank(MPI_COMM_WORLD, &rank);
MPI_Barrier(MPI_COMM_WORLD);
// printf("crossing bar1 %d\n", rank);
#ifdef DUMP
if(rank == 0)
{ m5_dump_stats(0, 0);
m5_reset_stats(0, 0);
}
#endif
for(n = 0; n < END; n++)
{
if(rank == 0)
{
u8 cipher_text[16] = {0xec,0xae,0xb0,0x2a,0xf2,0x51,0x45,0x25,0xb4,0x19,0x18,0x70,0x10,0x2,0x5,0x12};
rijndaelDecrypt_Master(decpt_rk, cipher_text, rank);
}
else if(rank == n_ranks - 1)
{
u8 decipher_text[16] = {0};
rijndaelDecrypt_Final(decpt_rk, decipher_text, rank);
}
else
rijndaelDecrypt_Middle(decpt_rk, rank);
}
//printf("before bar2 %d\n", rank);
MPI_Barrier(MPI_COMM_WORLD);
//printf("done with thread %d\n", rank);
MPI_Finalize();
return 0;
}
/*
* Runs benchmark.
*/
int main(int argc, char **argv)
{
int i; /* Loop index. */
int *map; /* Map of clusters. */
uint64_t end; /* End time. */
uint64_t start; /* Start time. */
vector_t *data; /* Data points. */
readargs(argc, argv);
printf("CAPBench - KM kernel\n");
printf(" # of threads: %d \n", nthreads);
timer_init();
srandnum(seed);
omp_set_num_threads(nthreads);
/* Benchmark initialization. */
start = timer_get();
data = smalloc(p->npoints*sizeof(vector_t));
for (i = 0; i < p->npoints; i++)
{
data[i] = vector_create(p->dimension);
vector_random(data[i]);
}
end = timer_get();
/* Cluster data. */
printf("Entering in KM ROI - Clustering data...\n");
start = timer_get();
m5_reset_stats(0,0);
map = kmeans(data, p->npoints, p->ncentroids, p->mindistance);
end = timer_get();
m5_dump_stats(0,0);
printf("KM total time: %f\n", timer_diff(start, end)*MICROSEC);
/* House keeping. */
free(map);
for (i = 0; i < p->npoints; i++)
vector_destroy(data[i]);
free(data);
return (0);
}
int main(int argc, char** argv)
{
int i, j;
int nx = NX;
int ny = NX;
/* Variable declaration/allocation. */
double *A = (double *) malloc(ny*nx*sizeof(double));
double *s = (double *) malloc(ny*sizeof(double));
double *q = (double *) malloc(nx*sizeof(double));
double *r = (double *) malloc(ny*sizeof(double));
double *p = (double *) malloc(nx*sizeof(double));
for (i = 0; i < ny; i++)
p[i] = i * M_PI;
for (i = 0; i < nx; i++) {
r[i] = i * M_PI;
for (j = 0; j < ny; j++)
A[i*nx + j] = ((double) i*(j+1))/nx;
}
m5_reset_stats(0,0);
#pragma scop
for (i = 0; i < nx; i++)
s[i] = 0;
for (i = 0; i < nx; i++)
{
q[i] = 0;
for (j = 0; j < ny; j++)
{
s[j] = s[j] + r[i] * A[i*nx + j];
q[i] = q[i] + A[i*nx + j] * p[j];
}
}
#pragma endscop
m5_dump_stats(0,0);
printf("%f, %f, %f, %f, %f\n", A[11], p[11], q[11], r[11], s[11]);
return 0;
}
int main(int argc, char *argv[])
{
int rank;
int n_ranks = 6;
int n = 10;
int i;
MPI_Init(&argc, &argv);
#ifndef DUMP
MPI_Comm_size(MPI_COMM_WORLD, &n_ranks);
#endif
MPI_Comm_rank(MPI_COMM_WORLD, &rank);
//u8 decipher_text[16] = {0};
/*
u8 rk[RK_LEN] = {
0x33, 0xc6, 0x6d, 0x09,
0x3e, 0x68, 0x5c, 0x0f,
0x63, 0x03, 0x13, 0x79,
0x92, 0x7c, 0x2a, 0x09,
};
*/
MPI_Barrier(MPI_COMM_WORLD);
#ifdef DUMP
if(rank == 0)
{ m5_dump_stats(0, 0);
m5_reset_stats(0, 0);
}
#endif
int END = 1000;
for(n = 0; n < END; n++)
{
if(rank == 0)
{
/*
u8 plain_text[16] = {
0x25, 0x5c, 0x31, 0x1d,
0x71, 0xd8, 0x04, 0xb7,
0x43, 0x11, 0x40, 0x86,
0x12, 0xa2, 0xc8, 0x3d
};*/
u8 cipher_text[16] = {
0x7d, 0x9a, 0xd4, 0x86,
0xfd, 0x60, 0x14, 0x4f,
0x57, 0xfd, 0x77, 0x94,
0xa3, 0xf2, 0xa0, 0xfb
};
rijndaelDecrypt_Master(decpt_rk, cipher_text, rank);
/*
putchar('\n');
for (i = 0; i < 16; i++)
{
printf("0x%02x, ", plain_text[i]);
}
putchar('\n');
*/
}
else if(rank == n_ranks - 1)
{
u8 decipher_text[16] = {0};
rijndaelDecrypt_Final(decpt_rk, decipher_text, rank);
//putchar('\n');
/*
for (i = 0; i < 16; i++)
{
printf("0x%02x, ", decipher_text[i]);
}
putchar('\n');
*/
}
else
rijndaelDecrypt_Middle(decpt_rk, rank);
if(n%5 == 0)
{
MPI_Barrier(MPI_COMM_WORLD);
}
}
#ifdef DUMP
if(rank == n_ranks -1)
{ m5_dump_stats(0, 0);
m5_reset_stats(0, 0);
}
#endif
MPI_Finalize();
return 0;
}
int main(int argc, char *argv[])
{
int rank;
int n_ranks, start_rank;
int i,j;
float gamma = 0.25, rho = -0.495266;
float GLOB_SUM = 0, sum = 0;
MPI_Init(&argc, &argv);
MPI_Comm_size(MPI_COMM_WORLD, &n_ranks);
MPI_Comm_rank(MPI_COMM_WORLD, &rank);
printf("before get data in id %d\n", rank);
get_data(rank%4);
start_rank = 6;
n_ranks = 4;
printf("done getting dat rank %d\n", rank);
MPI_Barrier(MPI_COMM_WORLD);
// printf("crossing bar1 %d\n", rank);
for (j = 0; j < INPUT_SIZE; ++j)
{
get_input(rank, start_rank, n_ranks);
sum = compute_svm_sum(rank%4, gamma);
if(rank == start_rank)
{
float tempBuff;
GLOB_SUM = sum;
for (i = start_rank+1; i < start_rank + n_ranks; ++i) {
MPI_Recv(&tempBuff, 1, MPI_FLOAT, i, 0, MPI_COMM_WORLD, MPI_STATUS_IGNORE);
GLOB_SUM = GLOB_SUM + tempBuff;
}
GLOB_SUM -= rho;
}
else {
MPI_Send((float*)&sum, 1, MPI_FLOAT, start_rank, 0, MPI_COMM_WORLD);
}
}
//if(rank != 6)
//printf("before bar2 %d\n", rank);
MPI_Barrier(MPI_COMM_WORLD);
if(rank == 6)
{
#ifdef DUMP
m5_dump_stats(0, 0);
m5_reset_stats(0, 0);
#endif
}
//printf("done with thread %d\n", rank);
if(rank == 6)
printf("global sum = %f\n", GLOB_SUM);
// free_data();
MPI_Finalize();
return 0;
}
