void
neighbor_get(int *target, int *src, int elements, int me, int npes, int loops)
{
int i, neighbor_pe;
double start_time, elapsed_time;
long total_bytes = loops * elements * sizeof(*src);
if (me==0 && Verbose) {
fprintf(stdout, "%s: %d loops of get(%ld bytes) from neighbor, %d PEs: ",
__FUNCTION__, loops, (elements*sizeof(*src)), npes);
fflush(stdout);
}
shmem_barrier_all();
neighbor_pe = (me + 1) % npes;
start_time = shmemx_wtime();
for(i = 0; i < loops; i++)
shmem_int_get(target, src, elements, neighbor_pe);
elapsed_time = shmemx_wtime() - start_time;
if (me==0 && Verbose) {
printf("%7.3f secs\n", elapsed_time);
printf(" %7.5f usecs / get(), %ld Kbytes @ %7.4f MB/sec\n\n",
(elapsed_time/((double)loops*npes))*1000000.0,
(total_bytes/1024),
((double)total_bytes/(1024.0*1024.0)) / elapsed_time );
}
shmem_barrier_all();
}
int main(){
int me, npes;
start_pes(0);
me = shmem_my_pe();
npes = shmem_n_pes();
for (i = 0; i < _SHMEM_REDUCE_SYNC_SIZE; i += 1)
{
pSync[i] = _SHMEM_SYNC_VALUE;
}
x = 42;
y = 0;
if(me==0){
shmem_barrier_all();
temp = x+y;
}
else {
shmem_barrier_all();
if(me==1){
old = shmem_int_finc (&y, 0);
shmem_int_sum_to_all(&y,&x,1,1,0,npes-1,pWrk,pSync);
x= x+10;
shmem_int_get(&y,&y,1,0);
}
else{
shmem_int_sum_to_all(&y,&x, 1,1,0,npes-1,pWrk,pSync);
x=y*0.23;
}
}
shmem_barrier_all();
if (me == 0) {
printf("value in temp is %d (should be 42)\n", temp);
}
return 0;
}
void
many2one_get(int *target, int *src, int elements, int me, int npes, int loops)
{
int i, pe;
double start_time, elapsed_time;
long total_bytes = loops * elements * sizeof(*src) * (npes - 1);
if (me == 0) {
fprintf(stdout,"%s: %d loops of get(%ld bytes) from %d PEs: ",
__FUNCTION__, loops, (elements*sizeof(*src)), npes-1);
fflush(stdout);
}
shmem_barrier_all();
if (me == 0) {
start_time = shmemx_wtime();
for(i = 0; i < loops; i++) {
for(pe = 1; pe < npes; pe++)
shmem_int_get(target, src, elements, pe);
}
elapsed_time = shmemx_wtime() - start_time;
if (Verbose) {
printf("%7.3f secs\n", elapsed_time);
printf(" %7.5f usecs / get(), %ld Kbytes @ %7.4f MB/sec\n\n",
(elapsed_time/((double)loops*(npes-1)))*1000000.0,
(total_bytes/1024),
((double)total_bytes/(1024.0*1024.0)) / elapsed_time );
}
}
shmem_barrier_all();
}
/****************************************************************************
* Place for Test Item functions
***************************************************************************/
static int test_item1(void)
{
int rc = TC_PASS;
int myPe = shmem_my_pe();
int myPeer = myPe + ( myPe % 2 ? -1 : 1 ) ;
int nPe = shmem_n_pes();
int remainderPe = nPe - (nPe % 2);
static int statArray[ARRAY_SIZE];
int* dynamicArray = shmalloc( ARRAY_SIZE * sizeof(int) );
int iterate;
for (iterate = 0; iterate < ARRAY_SIZE; iterate++)
{
if (myPe != remainderPe)
{
int tryIterate;
int putNum, getNum;
for (tryIterate = 0; tryIterate < TRY_SIZE; tryIterate++)
{
putNum = iterate + myPe;
shmem_int_put(&statArray[iterate], &putNum, 1, myPeer);
shmem_int_put(&dynamicArray[iterate], &putNum, 1, myPeer);
}
shmem_fence();
shmem_int_get(&getNum, &statArray[iterate], 1, myPeer);
if (getNum != putNum)
{
rc = TC_FAIL;
}
shmem_int_get(&getNum, &dynamicArray[iterate], 1, myPeer);
if (getNum != putNum)
{
rc = TC_FAIL;
}
}
shmem_barrier_all();
}
shfree(dynamicArray);
return rc;
}
void update_ghostcells ( int **buffer, int height, int width, int fhlimit,int localheight, int heightoffset )
{
int i, j, k;
shmem_barrier_all() ;
if ( size < 2 ) {
/* Nothing to be done */
return;
}
if ( rank != 0 ) {
if ( shmem_addr_accessible(&buffer[0][0], rank-1)) {
if ( (rank - 1 ) == 0 )
{
for ( i = 0; i < 2; i++ ) shmem_int_get ( &(buffer[i][0]), &(buffer[i+(localheight-2)][0]), width, rank-1);
}
else {
for ( i = 0; i < 2; i++ ) shmem_int_get ( &(buffer[i][0]), &(buffer[i+(localheight)][0]), width, rank-1);
}
}
else {
printf("Not_Accessible_Error_01");
}
if ( shmem_addr_accessible(&buffer[0][0], rank-1)) {
if ( (rank - 1 ) == 0 )
{
for ( i = 0; i < 2; i++ ) shmem_int_put ( &(buffer[i+(localheight)][0]), &(buffer[i+heightoffset][0]), width, rank-1);
}
else {
for ( i = 0; i < 2; i++ ) shmem_int_put ( &(buffer[i+(localheight)+2][0]), &(buffer[i+heightoffset][0]), width, rank-1);
}
}
else {
printf("Not_Accessible_Error_02");
}
}
shmem_barrier_all ();
return;
}
int main (void)
{
static int aaa, bbb;
int num_pes, my_pe, peer;
start_pes(0);
num_pes = _num_pes();
my_pe = _my_pe();
peer = (my_pe + 1) % num_pes;
printf("Process %d gets message from %d (%d processes in ring)\n", my_pe, peer, num_pes);
shmem_int_get(&aaa, &bbb, 1, peer);
shmem_barrier_all();
printf("Process %d exiting\n", my_pe);
return 0;
}
int recurse(int i, int ilog, int pes, int j, int nelems) {
int next_i,next_ilog,next_pes,k;
if(ilog <= npes) {
if(me % ilog ==0) {
if(me == 4) printf("\n");
// printf("\nDEBUG%d --> %d", me+i, me);
shmem_int_get(&A[nelems], A, nelems, me+i);
/* printf("\nDEBUG(%d)imported list: ",me);
for( k = 0; k < nelems; k++){
printf("%d ,",Aux[k]);
}
printf("\n");
for(k=0; k < nelems; k++){
A[nelems+k] = Aux[k];
} */
merge (0,nelems-1, nelems*2-1);
nelems = nelems *2;
/* printf("\nDEBUG(%d)new A: ",me);
for( k = 0; k < nelems; k++){
printf("A[%d]= %d \n",k,A[k]);
}
printf("\n");*/
next_pes = pes/2;
next_i = 2*i;
next_ilog = 2*ilog;
shmem_barrier(0,j , next_pes, pSync);
j = j+1;
recurse(next_i, next_ilog, next_pes, j, nelems);
}
}
}
int
main(int argc, char* argv[])
{
int me, neighbor;
int ret = 0;
shmem_init();
bbb = me = shmem_my_pe();
neighbor = (me + 1) % shmem_n_pes();
shmem_barrier_all();
shmem_int_get( &aaa, &bbb, 1, neighbor );
shmem_barrier_all();
if (aaa != neighbor ) ret = 1;
shmem_finalize();
return ret;
}
int
main(int argc, char **argv)
{
int i,j;
int nextpe;
int me, npes;
int success1,success2,success3, success4, success5, success6, success7, success8;
short dest1[N];
int dest2[N];
long dest3[N];
long double dest4[N];
long long dest5[N];
double dest6[N];
float dest7[N];
char *dest8;
short dest9;
int dest10;
long dest11;
double dest12;
float dest13;
short *src1;
int *src2;
long *src3;
long double *src4;
long long *src5;
double *src6;
float *src7;
char *src8;
short *src9;
int *src10;
long *src11;
double *src12;
float *src13;
start_pes(0);
me = _my_pe();
npes = _num_pes();
if(npes>1){
success1 =0;
success2 =0;
success3 =0;
success4 =0;
success5 =0;
success6 =0;
success7 =0;
success8 =0;
dest8 = (char *)malloc(N*sizeof(char));
for (i = 0; i < N; i += 1) {
dest1[i] = -9;
dest2[i] = -9;
dest3[i] = -9;
dest4[i] = -9;
dest5[i] = -9;
dest6[i] = -9;
dest7[i] = -9.0;
dest8[i] = -9;
}
dest9 = -9;
dest10 = -9;
dest11 = -9;
dest12 = -9;
dest13 = -9;
src1 = (short *)shmalloc( N * sizeof(*src1) );
src2 = (int *)shmalloc( N * sizeof(*src2) );
src3 = (long *)shmalloc( N * sizeof(*src3) );
src4 = (long double *)shmalloc( N * sizeof(*src4) );
src5 = (long long*)shmalloc( N * sizeof(*src5) );
src6 = (double *)shmalloc( N * sizeof(*src6) );
src7 = (float *)shmalloc( N * sizeof(*src7) );
src8 = (char *)shmalloc( 4 * sizeof(*src8) );
src9 = (short *)shmalloc( sizeof(*src9) );
src10 = (int *)shmalloc( sizeof(*src10) );
src11 = (long *)shmalloc( sizeof(*src11) );
src12 = (double *)shmalloc( sizeof(*src12) );
src13 = (float *)shmalloc( sizeof(*src13) );
for (i = 0; i < N; i += 1) {
src1[i] = (short)me;
src2[i] = me;
src3[i] = (long)me;
src4[i] = (long double)me;
src5[i] = (long long)me;
src6[i] = (double)me;
src7[i] = (float)me;
src8[i] = (char)me;
}
*src9 = (short)me;
*src10 = me;
*src11 = (long)me;
*src12 = (double)me;
*src13 = (float)me;
nextpe = (me + 1) % npes;
/*Testing shmem_short_get, shmem_short_get, shmem_int_get, shmem_long_get, shmem_longdouble_get, shmem_longlong_get, shmem_double_get, shmem_float_get, shmem_getmem*/
shmem_barrier_all();
shmem_short_get(dest1, src1, N, nextpe);
shmem_int_get(dest2, src2, N, nextpe);
shmem_long_get(dest3, src3, N, nextpe);
shmem_longdouble_get(dest4, src4, N, nextpe);
shmem_longlong_get(dest5, src5, N, nextpe);
shmem_double_get(dest6, src6, N, nextpe);
shmem_float_get(dest7, src7, N, nextpe);
shmem_getmem(dest8, src8, N*sizeof(char), nextpe);
shmem_barrier_all();
if(me == 0){
for (i = 0; i < N; i += 1) {
if(dest1[i] != ( 1)){
success1=1;
}
if(dest2[i] != ( 1)){
success2=1;
}
if(dest3[i] != ( 1)){
success3=1;
}
if(dest4[i] != ( 1)){
success4=1;
}
if(dest5[i] != ( 1)){
success5=1;
}
if(dest6[i] != ( 1)){
success6=1;
}
if(dest7[i] != ( 1)){
success7=1;
}
if(dest8[i] != ( 1)){
success8=1;
}
}
if(success1==0)
printf("Test shmem_short_get: Passed\n");
else
printf("Test shmem_short_get: Failed\n");
if(success2==0)
printf("Test shmem_int_get: Passed\n");
else
printf("Test shmem_int_get: Failed\n");
if(success3==0)
printf("Test shmem_long_get: Passed\n");
else
printf("Test shmem_long_get: Failed\n");
if(success4==0)
printf("Test shmem_longdouble_get: Passed\n");
else
printf("Test shmem_longdouble_get: Failed\n");
if(success5==0)
printf("Test shmem_longlong_get: Passed\n");
else
printf("Test shmem_longlong_get: Failed\n");
if(success6==0)
printf("Test shmem_double_get: Passed\n");
else
printf("Test shmem_double_get: Failed\n");
if(success7==0)
printf("Test shmem_float_get: Passed\n");
else
printf("Test shmem_float_get: Failed\n");
if(success8==0)
printf("Test shmem_getmem: Passed\n");
else
printf("Test shmem_getmem: Failed\n");
}
shmem_barrier_all();
/*Testing shmem_get32, shmem_get64, shmem_get128 */
if(sizeof(int)==4){
for (i = 0; i < N; i += 1) {
dest2[i] = -9;
dest3[i] = -9;
dest4[i] = -9;
}
success2 = 0;
success3 = 0;
success4 = 0;
shmem_barrier_all();
shmem_get32(dest2, src2, N, nextpe);
shmem_get64(dest3, src3, N, nextpe);
shmem_get128(dest4, src4, N, nextpe);
shmem_barrier_all();
if(me == 0){
for (i = 0; i < N; i += 1) {
if(dest2[i] != ( 1)){
success2=1;
}
if(dest3[i] != ( 1)){
success3=1;
}
if(dest4[i] != ( 1)){
success4=1;
}
}
if(success2==0)
printf("Test shmem_get32: Passed\n");
else
printf("Test shmem_get32: Failed\n");
if(success3==0)
printf("Test shmem_get64: Passed\n");
else
printf("Test shmem_get64: Failed\n");
if(success4==0)
printf("Test shmem_get128: Passed\n");
else
printf("Test shmem_get128: Failed\n");
}
}
else if(sizeof(int)==8){
for (i = 0; i < N; i += 1) {
dest1[i] = -9;
dest2[i] = -9;
dest3[i] = -9;
}
success1 = 0;
success2 = 0;
success3 = 0;
shmem_barrier_all();
shmem_get32(dest1, src1, N, nextpe);
shmem_get64(dest2, src2, N, nextpe);
shmem_get128(dest3, src3, N, nextpe);
shmem_barrier_all();
if(me == 0){
for (i = 0; i < N; i += 1) {
if(dest1[i] != ( 1)){
success1=1;
}
if(dest2[i] != ( 1)){
success2=1;
}
if(dest3[i] != ( 1)){
success3=1;
}
}
if(success1==0)
printf("Test shmem_get32: Passed\n");
else
printf("Test shmem_get32: Failed\n");
if(success2==0)
printf("Test shmem_get64: Passed\n");
else
printf("Test shmem_get64: Failed\n");
if(success3==0)
printf("Test shmem_get128: Passed\n");
else
printf("Test shmem_get128: Failed\n");
}
}
/* Testing shmem_iget32, shmem_iget64, shmem_iget128 */
shmem_barrier_all();
if(sizeof(int)==4){
for (i = 0; i < N; i += 1) {
dest2[i] = -9;
dest3[i] = -9;
dest4[i] = -9;
}
success2 = 0;
success3 = 0;
success4 = 0;
shmem_barrier_all();
shmem_iget32(dest2, src2, 1, 2, N/2, npes-1);
shmem_iget64(dest3, src3, 1, 2, N/2, npes-1);
shmem_iget128(dest4, src4, 1, 2, N/2, npes-1);
shmem_barrier_all();
if(me == 0){
for (i = 0; i < N/2; i += 1) {
if(dest2[i] != (npes-1)){
success2=1;
}
if(dest3[i] != (npes-1)){
success3=1;
}
if(dest4[i] != (npes-1)){
success4=1;
}
}
if(success2==0)
printf("Test shmem_iget32: Passed\n");
else
printf("Test shmem_iget32: Failed\n");
if(success3==0)
printf("Test shmem_iget64: Passed\n");
else
printf("Test shmem_iget64: Failed\n");
if(success4==0)
printf("Test shmem_iget128: Passed\n");
else
printf("Test shmem_iget128: Failed\n");
}
}
else if(sizeof(int)==8){
for (i = 0; i < N; i += 1) {
dest1[i] = -9;
dest2[i] = -9;
dest3[i] = -9;
}
success1 = 0;
success2 = 0;
success3 = 0;
shmem_barrier_all();
shmem_iget32(dest1, src1, 1, 2, N/2, npes-1);
shmem_iget64(dest2, src2, 1, 2, N/2, npes-1);
shmem_iget128(dest3, src3, 1, 2, N/2, npes-1);
shmem_barrier_all();
if(me == 0){
for (i = 0; i < N/2; i += 1) {
if(dest1[i] != (npes-1)){
success1=1;
}
if(dest2[i] != (npes-1)){
success2=1;
}
if(dest3[i] != (npes-1)){
success3=1;
}
}
if(success1==0)
printf("Test shmem_iget32: Passed\n");
else
printf("Test shmem_iget32: Failed\n");
if(success2==0)
printf("Test shmem_iget64: Passed\n");
else
printf("Test shmem_iget64: Failed\n");
if(success3==0)
printf("Test shmem_iget128: Passed\n");
else
printf("Test shmem_iget128: Failed\n");
}
}
/*Testing shmem_short_iget, shmem_int_iget, shmem_long_iget, shmem_double_iget, shmem_float_iget */
for (i = 0; i < N; i += 1) {
dest1[i] = -9;
dest2[i] = -9;
dest3[i] = -9;
dest6[i] = -9;
dest7[i] = -9;
}
success1 = 0;
success2 = 0;
success3 = 0;
success6 = 0;
success7 = 0;
shmem_barrier_all();
shmem_short_iget(dest1, src1, 1, 2, N/2, npes-1);
shmem_int_iget(dest2, src2, 1, 2, N/2, npes-1);
shmem_long_iget(dest3, src3, 1, 2, N/2, npes-1);
shmem_double_iget(dest6, src6, 1, 2, N/2, npes-1);
shmem_float_iget(dest7, src7, 1, 2, N/2, npes-1);
shmem_barrier_all();
if(me == 0){
for (i = 0; i < N/2; i += 1) {
if(dest1[i] != (npes-1)){
success1=1;
}
if(dest2[i] != (npes-1)){
success2=1;
}
if(dest3[i] != (npes-1)){
success3=1;
}
if(dest6[i] != (npes-1)){
success6=1;
}
if(dest7[i] != (npes-1)){
success7=1;
}
}
if(success1==0)
printf("Test shmem_short_iget: Passed\n");
else
printf("Test shmem_short_iget: Failed\n");
if(success2==0)
printf("Test shmem_int_iget: Passed\n");
else
printf("Test shmem_int_iget: Failed\n");
if(success3==0)
printf("Test shmem_long_iget: Passed\n");
else
printf("Test shmem_long_iget: Failed\n");
if(success6==0)
printf("Test shmem_double_iget: Passed\n");
else
printf("Test shmem_double_iget: Failed\n");
if(success7==0)
printf("Test shmem_float_iget: Passed\n");
else
printf("Test shmem_float_iget: Failed\n");
}
/* Testing shmem_double_g, shmem_float_g, shmem_int_g, shmem_long_g, shmem_short_g */
shmem_barrier_all();
dest9 = shmem_short_g(src9, nextpe);
dest10 = shmem_int_g(src10, nextpe);
dest11 = shmem_long_g(src11, nextpe);
dest12 = shmem_double_g(src12, nextpe);
dest13 = shmem_float_g(src13, nextpe);
shmem_barrier_all();
if(me == 0){
if(dest9 == 1)
printf("Test shmem_short_g: Passed\n");
else
printf("Test shmem_short_g: Failed\n");
if(dest10 == 1)
printf("Test shmem_int_g: Passed\n");
else
printf("Test shmem_int_g: Failed\n");
if(dest11 == 1)
printf("Test shmem_long_g: Passed\n");
else
printf("Test shmem_long_g: Failed\n");
if(dest12 == 1)
printf("Test shmem_double_g: Passed\n");
else
printf("Test shmem_double_g: Failed\n");
if(dest13 == 1)
printf("Test shmem_float_g: Passed\n");
else
printf("Test shmem_float_g: Failed\n");
}
shmem_barrier_all();
shfree(src1);
shfree(src2);
shfree(src3);
shfree(src4);
shfree(src5);
shfree(src6);
shfree(src7);
shfree(src8);
}
else{
printf("Number of PEs must be > 1 to test shmem get, test skipped\n");
}
return 0;
}
int
main (int argc, char *argv[])
{
int *sray, *rray;
int *sdisp, *scounts, *rdisp, *rcounts;
int ssize, rsize, i, k, j;
float z;
init_it (&argc, &argv);
scounts = (int *) shmem_malloc (sizeof (int) * numnodes);
rcounts = (int *) shmem_malloc (sizeof (int) * numnodes);
sdisp = (int *) shmem_malloc (sizeof (int) * numnodes);
rdisp = (int *) shmem_malloc (sizeof (int) * numnodes);
/*
! seed the random number generator with a ! different number on each
processor */
seed_random (myid);
/* find out how much data to send */
for (i = 0; i < numnodes; i++) {
random_number (&z);
scounts[i] = (int) (5.0 * z) + 1;
}
printf ("myid= %d scounts=%d %d %d %d\n", myid, scounts[0], scounts[1],
scounts[2], scounts[3]);
/* tell the other processors how much data is coming */
// mpi_err = MPI_Alltoall(scounts,1,MPI_INT, rcounts,1,MPI_INT,
// MPI_COMM_WORLD);
shmem_barrier_all ();
int other, j1;
for (j1 = 0; j1 < numnodes; j1++) {
shmem_int_put (&rcounts[myid], &scounts[j1], 1, j1);
}
shmem_barrier_all ();
printf ("-----myid= %d rcounts=", myid);
for (i = 0; i < numnodes; i++) {
printf ("%d ", rcounts[i]);
}
printf ("\n");
/* write(*,*)"myid= ",myid," rcounts= ",rcounts */
/* calculate displacements and the size of the arrays */
sdisp[0] = 0;
for (i = 1; i < numnodes; i++) {
sdisp[i] = scounts[i - 1] + sdisp[i - 1];
}
rdisp[0] = 0;
for (i = 1; i < numnodes; i++) {
rdisp[i] = rcounts[i - 1] + rdisp[i - 1];
}
ssize = 0;
rsize = 0;
for (i = 0; i < numnodes; i++) {
ssize = ssize + scounts[i];
rsize = rsize + rcounts[i];
}
/* allocate send and rec arrays */
sray = (int *) shmem_malloc (sizeof (int) * 20);
rray = (int *) shmem_malloc (sizeof (int) * 20);
for (i = 0; i < ssize; i++) {
sray[i] = myid;
}
/* send/rec different amounts of data to/from each processor */
// mpi_err = MPI_Alltoallv(sray,scounts,sdisp,MPI_INT,
// rray,rcounts,rdisp,MPI_INT, MPI_COMM_WORLD);
shmem_barrier_all ();
for (j1 = 0; j1 < numnodes; j1++) {
int k1 = sdisp[j1];
static int k2;
shmem_int_get (&k2, &rdisp[myid], 1, j1);
shmem_int_put (rray + k2, sray + k1, scounts[j1], j1);
}
shmem_barrier_all ();
printf ("myid= %d rray=", myid);
for (i = 0; i < rsize; i++) {
printf ("%d ", rray[i]);
}
printf ("\n");
// mpi_err = MPI_Finalize();
shmem_finalize ();
return 0;
}
/* main function */
int main(int argc, char **argv)
{
int localheight=0;
int realheight=0;
int height, width,i, j, k, heightoffset;
int **labels=NULL;
FILE *fpa;
/* initializing shmem values */
start_pes (0);
size = _num_pes ();
rank = _my_pe ();
/* determining inputs */
if (argc != 5 ) {
printf("usage: %s <inputfile> <outputfile> <height> <width>\n", argv[0]);
exit(0);
}
height = atoi ( argv[3]);
width = atoi ( argv[4]);
/* creating symmetric memory */
labels = (int **) shmalloc (height*sizeof(int *));
for ( i = 0; i < height; i++ )
{
labels[i] = (int*) shmalloc (width*sizeof(int*));
}
allocate_2D_int_matrix ( &labels, height, width );
/* reading from the input file */
if ( rank == 0 )
{
fpa = fopen ( argv[1], "rb");
fread ( &(labels[0][0]), sizeof(int), height*width, fpa);
fclose ( fpa );
}
/* datapartitioning between the processes */
localheight = height/size;
realheight = localheight + 4 ;
heightoffset = 2;
if ( rank == 0 ) {
realheight -= 2 ;
heightoffset = 0;
}
if ( rank == (size -1) ) {
realheight -= 2;
}
shmem_barrier_all ();
/* distributing input datum to all the symmetric variables */
for ( i=1; i<size; i++ ) {
if ( rank == i )
if ( shmem_addr_accessible(&labels[0][0], 0)) {
if ( rank == ( size -1 ) ) {
k = 0;
}
else {
k = 2;
}
for ( j = 0; j < localheight+k; j++ ) {
shmem_int_get(&labels[j+heightoffset][0], &labels[j+(localheight*rank)][0], width, 0);
}
}
else {
printf("Not_accessible_Error_00");
}
}
/* gray erode calculation */
shmem_barrier_all();
gettimeofday(&start, NULL);
gray_erode(labels, height, width, FILTERHEIGHT,FILTERWIDTH, ITERATIONS, SEGMENTS, localheight, heightoffset );
gettimeofday(&end, NULL);
/* calculating and displaying time taken */
if( rank == 0 )
{
t = (double) ((end.tv_sec * 1000000 + end.tv_usec) - (start.tv_sec * 1000000 + start.tv_usec)) / 1000000.0;
printf("Total Time%f \n",t);
}
/* transferring datum back to root */
shmem_barrier_all ();
for ( k = 0; k < size; k++) {
if ( rank == k ) {
for ( i = 0; i < localheight; i++ ) {
shmem_int_put(&labels[i+(localheight*rank)][0], &labels[i+heightoffset][0], width, 0);
}
}
}
shmem_barrier_all ();
/* creating output file */
if(rank == 0) {
fpa = fopen ( argv[2], "wb");
fwrite( &(labels[0][0]), sizeof(int), height*width, fpa );
fclose( fpa );
}
free_2D_int_matrix ( &labels);
return 0;
}
int
main (int argc, char *argv[])
{
int *sray, *rray;
int *sdisp, *scounts, *rdisp, *rcounts, *rcounts_full;
int ssize, rsize, i, k, j;
float z;
init_it (&argc, &argv);
scounts = (int *) shmem_malloc (sizeof (int) * numnodes);
rcounts = (int *) shmem_malloc (sizeof (int) * numnodes);
rcounts_full = (int *) shmem_malloc (sizeof (int) * numnodes * numnodes);
sdisp = (int *) shmem_malloc (sizeof (int) * numnodes);
rdisp = (int *) shmem_malloc (sizeof (int) * numnodes);
/*
! seed the random number generator with a ! different number on each
processor */
seed_random (myid);
/* find out how much data to send */
for (i = 0; i < numnodes; i++) {
random_number (&z);
scounts[i] = (int) (5.0 * z) + 1;
}
printf ("myid= %d scounts=%d %d %d %d\n", myid, scounts[0], scounts[1],
scounts[2], scounts[3]);
printf ("\n");
/* tell the other processors how much data is coming */
// mpi_err = MPI_Alltoall(scounts,1,MPI_INT, rcounts,1,MPI_INT,
// MPI_COMM_WORLD);
static long psync[_SHMEM_COLLECT_SYNC_SIZE];
for (i = 0; i < _SHMEM_COLLECT_SYNC_SIZE; i++)
psync[i] = _SHMEM_SYNC_VALUE;
shmem_barrier_all ();
int other, j1;
shmem_fcollect32 (rcounts_full, scounts, 4, 0, 0, numnodes, psync);
for (i = 0; i < numnodes; i++) {
rcounts[i] = rcounts_full[i * numnodes + myid];
}
printf ("-----myid= %d rcounts=", myid);
for (i = 0; i < numnodes; i++)
printf ("%d ", rcounts[i]);
printf ("\n");
/* write(*,*)"myid= ",myid," rcounts= ",rcounts */
/* calculate displacements and the size of the arrays */
sdisp[0] = 0;
for (i = 1; i < numnodes; i++) {
sdisp[i] = scounts[i - 1] + sdisp[i - 1];
}
rdisp[0] = 0;
for (i = 1; i < numnodes; i++) {
rdisp[i] = rcounts[i - 1] + rdisp[i - 1];
}
ssize = 0;
rsize = 0;
for (i = 0; i < numnodes; i++) {
ssize = ssize + scounts[i];
rsize = rsize + rcounts[i];
}
/* allocate send and rec arrays */
sray = (int *) shmem_malloc (sizeof (int) * 20);
rray = (int *) shmem_malloc (sizeof (int) * 20);
for (i = 0; i < ssize; i++) {
sray[i] = myid;
}
/* send/rec different amounts of data to/from each processor */
// mpi_err = MPI_Alltoallv(sray,scounts,sdisp,MPI_INT,
// rray,rcounts,rdisp,MPI_INT, MPI_COMM_WORLD);
shmem_barrier_all ();
for (j1 = 0; j1 < numnodes; j1++) {
int k1 = sdisp[j1];
static int k2;
shmem_int_get (&k2, &rdisp[myid], 1, j1);
shmem_int_put (rray + k2, sray + k1, scounts[j1], j1);
}
shmem_barrier_all ();
// not possible, coz even though the rcounts[myid] will be different on
// each PE, the elements collected
// by PE0 from other PE's will be constant.
// shmem_collect32(rray_full,sray,rcounts[myid],0,0,numnodes,psync);
printf ("myid= %d rray=", myid);
for (i = 0; i < rsize; i++)
printf ("%d ", rray[i]);
printf ("\n");
// mpi_err = MPI_Finalize();
shmem_finalize ();
return 0;
}
int
main(int argc, char **argv)
{
int loops=DFLT_LOOPS;
char *pgm;
int *Target;
int *Source;
int i, me, npes;
int target_pe;
long bytes;
double time_taken=0.0, start_time;
shmem_init();
me = shmem_my_pe();
npes = shmem_n_pes();
if ((pgm=strrchr(argv[0],'/')))
pgm++;
else
pgm = argv[0];
while ((i = getopt (argc, argv, "hve:l:st")) != EOF) {
switch (i)
{
case 'v':
Verbose++;
break;
case 'e':
if ((elements = atoi_scaled(optarg)) <= 0) {
fprintf(stderr,"ERR: Bad elements count %d\n",elements);
shmem_finalize();
return 1;
}
break;
case 'l':
if ((loops = atoi_scaled(optarg)) <= 0) {
fprintf(stderr,"ERR: Bad loop count %d\n",loops);
shmem_finalize();
return 1;
}
break;
case 's':
Sync++;
break;
case 't':
Track++;
break;
case 'h':
if (me == 0)
usage(pgm);
return 0;
default:
if (me == 0) {
fprintf(stderr,"%s: unknown switch '-%c'?\n",pgm,i);
usage(pgm);
}
shmem_finalize();
return 1;
}
}
target_pe = (me+1) % npes;
total_time = (double *) shmem_malloc( npes * sizeof(double) );
if (!total_time) {
fprintf(stderr,"ERR: bad total_time shmem_malloc(%ld)\n",
(elements * sizeof(double)));
shmem_global_exit(1);
}
Source = (int *) shmem_malloc( elements * sizeof(*Source) );
if (!Source) {
fprintf(stderr,"ERR: bad Source shmem_malloc(%ld)\n",
(elements * sizeof(*Target)));
shmem_free(total_time);
shmem_global_exit(1);
}
Target = (int *) shmem_malloc( elements * sizeof(*Target) );
if (!Target) {
fprintf(stderr,"ERR: bad Target shmem_malloc(%ld)\n",
(elements * sizeof(*Target)));
shmem_free(Source);
shmem_free(total_time);
shmem_global_exit(1);
}
for (i = 0; i < elements; i++) {
Target[i] = -90;
Source[i] = i + 1;
}
bytes = loops * sizeof(int) * elements;
if (Verbose && me==0)
fprintf(stderr, "%s: INFO - %d loops, get %d (int) elements from PE+1\n",
pgm, loops, elements);
shmem_barrier_all();
for(i=0; i < loops; i++) {
start_time = shmemx_wtime();
shmem_int_get( Target, Source, elements, target_pe );
time_taken += shmemx_wtime() - start_time;
if (me==0) {
if ( Track && i > 0 && ((i % 200) == 0))
fprintf(stderr,".%d",i);
}
if (Sync)
shmem_barrier_all();
}
// collect time per node elapsed time.
shmem_double_put( &total_time[me], &time_taken, 1, 0 );
shmem_barrier_all();
for (i = 0; i < elements; i++) {
if (Target[i] != i + 1) {
printf("%d: Error Target[%d] = %d, expected %d\n",
me, i, Target[i], i + 1);
shmem_global_exit(1);
}
}
if ( Track && me == 0 )
fprintf(stderr,"\n");
if (Verbose && me == 0) {
double rate,secs;
// average time
for(i=0,secs=0.0; i < npes; i++)
secs += total_time[i];
secs /= (double)npes;
rate = ((double)bytes/(1024.0*1024.0)) / secs;
printf("%s: ave %5.3f MB/sec (bytes %ld secs %5.3f)\n",
pgm, rate, bytes, secs);
}
shmem_free(total_time);
shmem_free(Target);
shmem_free(Source);
shmem_finalize();
return 0;
}
int
main (int argc, char **argv)
{
int i;
int nextpe;
int me, npes;
int success1, success2, success3, success4, success5, success6, success7,
success8;
short dest1[N];
int dest2[N];
long dest3[N];
long double dest4[N];
long long dest5[N];
double dest6[N];
float dest7[N];
char *dest8;
short dest9;
int dest10;
long dest11;
double dest12;
float dest13;
int fail_count = 0;
shmem_init ();
me = shmem_my_pe ();
npes = shmem_n_pes ();
if (npes > 1) {
success1 = 0;
success2 = 0;
success3 = 0;
success4 = 0;
success5 = 0;
success6 = 0;
success7 = 0;
success8 = 0;
dest8 = (char *) malloc (N * sizeof (char));
for (i = 0; i < N; i += 1) {
dest1[i] = -9;
dest2[i] = -9;
dest3[i] = -9;
dest4[i] = -9;
dest5[i] = -9;
dest6[i] = -9;
dest7[i] = -9.0;
dest8[i] = -9;
}
dest9 = -9;
dest10 = -9;
dest11 = -9;
dest12 = -9;
dest13 = -9;
for (i = 0; i < N; i += 1) {
src1[i] = (short) me;
src2[i] = me;
src3[i] = (long) me;
src4[i] = (long double) me;
src5[i] = (long long) me;
src6[i] = (double) me;
src7[i] = (float) me;
src8[i] = (char) me;
}
src9 = (short) me;
src10 = me;
src11 = (long) me;
src12 = (double) me;
src13 = (float) me;
nextpe = (me + 1) % npes;
/* Testing shmem_short_get, shmem_short_get, shmem_int_get,
shmem_long_get, shmem_longdouble_get, shmem_longlong_get,
shmem_double_get, shmem_float_get, shmem_getmem */
shmem_barrier_all ();
shmem_short_get (dest1, src1, N, nextpe);
shmem_int_get (dest2, src2, N, nextpe);
shmem_long_get (dest3, src3, N, nextpe);
shmem_longdouble_get (dest4, src4, N, nextpe);
shmem_longlong_get (dest5, src5, N, nextpe);
shmem_double_get (dest6, src6, N, nextpe);
shmem_float_get (dest7, src7, N, nextpe);
shmem_getmem (dest8, src8, N * sizeof (char), nextpe);
shmem_barrier_all ();
if (me == 0) {
for (i = 0; i < N; i += 1) {
if (dest1[i] != (1)) {
success1 = 1;
}
if (dest2[i] != (1)) {
success2 = 1;
}
if (dest3[i] != (1)) {
success3 = 1;
}
if (dest4[i] != (1)) {
success4 = 1;
}
if (dest5[i] != (1)) {
success5 = 1;
}
if (dest6[i] != (1)) {
success6 = 1;
}
if (dest7[i] != (1)) {
success7 = 1;
}
if (dest8[i] != (1)) {
success8 = 1;
}
}
if (success1 == 0)
printf ("Test shmem_short_get: Passed\n");
else {
printf ("Test shmem_short_get: Failed\n");
fail_count++;
}
if (success2 == 0)
printf ("Test shmem_int_get: Passed\n");
else {
printf ("Test shmem_int_get: Failed\n");
fail_count++;
}
if (success3 == 0)
printf ("Test shmem_long_get: Passed\n");
else {
printf ("Test shmem_long_get: Failed\n");
fail_count++;
}
if (success4 == 0)
printf ("Test shmem_longdouble_get: Passed\n");
else {
printf ("Test shmem_longdouble_get: Failed\n");
fail_count++;
}
if (success5 == 0)
printf ("Test shmem_longlong_get: Passed\n");
else {
printf ("Test shmem_longlong_get: Failed\n");
fail_count++;
}
if (success6 == 0)
printf ("Test shmem_double_get: Passed\n");
else {
printf ("Test shmem_double_get: Failed\n");
fail_count++;
}
if (success7 == 0)
printf ("Test shmem_float_get: Passed\n");
else {
printf ("Test shmem_float_get: Failed\n");
fail_count++;
}
if (success8 == 0)
printf ("Test shmem_getmem: Passed\n");
else {
printf ("Test shmem_getmem: Failed\n");
fail_count++;
}
}
shmem_barrier_all ();
/* Testing shmem_get32, shmem_get64, shmem_get128 */
if (sizeof (int) == 4) {
for (i = 0; i < N; i += 1) {
dest2[i] = -9;
dest3[i] = -9;
dest4[i] = -9;
}
success2 = 0;
success3 = 0;
success4 = 0;
shmem_barrier_all ();
shmem_get32 (dest2, src2, N, nextpe);
shmem_get64 (dest3, src3, N, nextpe);
shmem_get128 (dest4, src4, N, nextpe);
shmem_barrier_all ();
if (me == 0) {
for (i = 0; i < N; i += 1) {
if (dest2[i] != (1)) {
success2 = 1;
}
if (dest3[i] != (1)) {
success3 = 1;
}
if (dest4[i] != (1)) {
success4 = 1;
}
}
if (success2 == 0)
printf ("Test shmem_get32: Passed\n");
else {
printf ("Test shmem_get32: Failed\n");
fail_count++;
}
if (success3 == 0)
printf ("Test shmem_get64: Passed\n");
else {
printf ("Test shmem_get64: Failed\n");
fail_count++;
}
if (success4 == 0)
printf ("Test shmem_get128: Passed\n");
else {
printf ("Test shmem_get128: Failed\n");
fail_count++;
}
}
}
else if (sizeof (int) == 8) {
for (i = 0; i < N; i += 1) {
dest1[i] = -9;
dest2[i] = -9;
dest3[i] = -9;
}
success1 = 0;
success2 = 0;
success3 = 0;
shmem_barrier_all ();
shmem_get32 (dest1, src1, N, nextpe);
shmem_get64 (dest2, src2, N, nextpe);
shmem_get128 (dest3, src3, N, nextpe);
shmem_barrier_all ();
if (me == 0) {
for (i = 0; i < N; i += 1) {
if (dest1[i] != (1)) {
success1 = 1;
}
if (dest2[i] != (1)) {
success2 = 1;
}
if (dest3[i] != (1)) {
success3 = 1;
}
}
if (success1 == 0)
printf ("Test shmem_get32: Passed\n");
else {
printf ("Test shmem_get32: Failed\n");
fail_count++;
}
if (success2 == 0)
printf ("Test shmem_get64: Passed\n");
else {
printf ("Test shmem_get64: Failed\n");
fail_count++;
}
if (success3 == 0)
printf ("Test shmem_get128: Passed\n");
else {
printf ("Test shmem_get128: Failed\n");
fail_count++;
}
}
}
/* Testing shmem_iget32, shmem_iget64, shmem_iget128 */
shmem_barrier_all ();
if (sizeof (int) == 4) {
for (i = 0; i < N; i += 1) {
dest2[i] = -9;
dest3[i] = -9;
dest4[i] = -9;
}
success2 = 0;
success3 = 0;
success4 = 0;
shmem_barrier_all ();
shmem_iget32 (dest2, src2, 1, 2, N / 2, npes - 1);
shmem_iget64 (dest3, src3, 1, 2, N / 2, npes - 1);
shmem_iget128 (dest4, src4, 1, 2, N / 2, npes - 1);
shmem_barrier_all ();
if (me == 0) {
for (i = 0; i < N / 2; i += 1) {
if (dest2[i] != (npes - 1)) {
success2 = 1;
}
if (dest3[i] != (npes - 1)) {
success3 = 1;
}
if (dest4[i] != (npes - 1)) {
success4 = 1;
}
}
if (success2 == 0)
printf ("Test shmem_iget32: Passed\n");
else {
printf ("Test shmem_iget32: Failed\n");
fail_count++;
}
if (success3 == 0)
printf ("Test shmem_iget64: Passed\n");
else {
printf ("Test shmem_iget64: Failed\n");
fail_count++;
}
if (success4 == 0)
printf ("Test shmem_iget128: Passed\n");
else {
printf ("Test shmem_iget128: Failed\n");
fail_count++;
}
}
}
else if (sizeof (int) == 8) {
for (i = 0; i < N; i += 1) {
dest1[i] = -9;
dest2[i] = -9;
dest3[i] = -9;
}
success1 = 0;
success2 = 0;
success3 = 0;
shmem_barrier_all ();
shmem_iget32 (dest1, src1, 1, 2, N / 2, npes - 1);
shmem_iget64 (dest2, src2, 1, 2, N / 2, npes - 1);
shmem_iget128 (dest3, src3, 1, 2, N / 2, npes - 1);
shmem_barrier_all ();
if (me == 0) {
for (i = 0; i < N / 2; i += 1) {
if (dest1[i] != (npes - 1)) {
success1 = 1;
}
if (dest2[i] != (npes - 1)) {
success2 = 1;
}
if (dest3[i] != (npes - 1)) {
success3 = 1;
}
}
if (success1 == 0)
printf ("Test shmem_iget32: Passed\n");
else {
printf ("Test shmem_iget32: Failed\n");
fail_count++;
}
if (success2 == 0)
printf ("Test shmem_iget64: Passed\n");
else {
printf ("Test shmem_iget64: Failed\n");
fail_count++;
}
if (success3 == 0)
printf ("Test shmem_iget128: Passed\n");
else {
printf ("Test shmem_iget128: Failed\n");
fail_count++;
}
}
}
/* Testing shmem_short_iget, shmem_int_iget, shmem_long_iget,
shmem_double_iget, shmem_float_iget */
for (i = 0; i < N; i += 1) {
dest1[i] = -9;
dest2[i] = -9;
dest3[i] = -9;
dest6[i] = -9;
dest7[i] = -9;
}
success1 = 0;
success2 = 0;
success3 = 0;
success6 = 0;
success7 = 0;
shmem_barrier_all ();
shmem_short_iget (dest1, src1, 1, 2, N / 2, npes - 1);
shmem_int_iget (dest2, src2, 1, 2, N / 2, npes - 1);
shmem_long_iget (dest3, src3, 1, 2, N / 2, npes - 1);
shmem_double_iget (dest6, src6, 1, 2, N / 2, npes - 1);
shmem_float_iget (dest7, src7, 1, 2, N / 2, npes - 1);
shmem_barrier_all ();
if (me == 0) {
for (i = 0; i < N / 2; i += 1) {
if (dest1[i] != (npes - 1)) {
success1 = 1;
}
if (dest2[i] != (npes - 1)) {
success2 = 1;
}
if (dest3[i] != (npes - 1)) {
success3 = 1;
}
if (dest6[i] != (npes - 1)) {
success6 = 1;
}
if (dest7[i] != (npes - 1)) {
success7 = 1;
}
}
if (success1 == 0)
printf ("Test shmem_short_iget: Passed\n");
else {
printf ("Test shmem_short_iget: Failed\n");
fail_count++;
}
if (success2 == 0)
printf ("Test shmem_int_iget: Passed\n");
else {
printf ("Test shmem_int_iget: Failed\n");
fail_count++;
}
if (success3 == 0)
printf ("Test shmem_long_iget: Passed\n");
else {
printf ("Test shmem_long_iget: Failed\n");
fail_count++;
}
if (success6 == 0)
printf ("Test shmem_double_iget: Passed\n");
else {
printf ("Test shmem_double_iget: Failed\n");
fail_count++;
}
if (success7 == 0)
printf ("Test shmem_float_iget: Passed\n");
else {
printf ("Test shmem_float_iget: Failed\n");
fail_count++;
}
}
/* Testing shmem_double_g, shmem_float_g, shmem_int_g, shmem_long_g,
shmem_short_g */
shmem_barrier_all ();
dest9 = shmem_short_g (&src9, nextpe);
dest10 = shmem_int_g (&src10, nextpe);
dest11 = shmem_long_g (&src11, nextpe);
dest12 = shmem_double_g (&src12, nextpe);
dest13 = shmem_float_g (&src13, nextpe);
shmem_barrier_all ();
if (me == 0) {
if (dest9 == 1)
printf ("Test shmem_short_g: Passed\n");
else {
printf ("Test shmem_short_g: Failed\n");
fail_count++;
}
if (dest10 == 1)
printf ("Test shmem_int_g: Passed\n");
else {
printf ("Test shmem_int_g: Failed\n");
fail_count++;
}
if (dest11 == 1)
printf ("Test shmem_long_g: Passed\n");
else {
printf ("Test shmem_long_g: Failed\n");
fail_count++;
}
if (dest12 == 1)
printf ("Test shmem_double_g: Passed\n");
else {
printf ("Test shmem_double_g: Failed\n");
fail_count++;
}
if (dest13 == 1)
printf ("Test shmem_float_g: Passed\n");
else {
printf ("Test shmem_float_g: Failed\n");
fail_count++;
}
}
shmem_barrier_all ();
if (me == 0) {
if (fail_count == 0)
printf("All Tests Passed\n");
else
printf("%d Tests Failed\n", fail_count);
}
}
else {
printf ("Number of PEs must be > 1 to test shmem get, test skipped\n");
}
shmem_finalize ();
return 0;
}
/* Performance test for shmem_XX_get (latency and bandwidth) */
#include <stdio.h>
#include <stdlib.h>
#include <time.h>
#include <sys/time.h>
#include <shmem.h>
long double time_taken;
long pSync[_SHMEM_REDUCE_SYNC_SIZE];
long double pWrk[_SHMEM_REDUCE_MIN_WRKDATA_SIZE];
//#define N_ELEMENTS 25600/*Data size chosen to be able to capture time required*/
int
main(void)
{
int i,j,k;
int *target;
int *source;
int me, npes;
int nxtpe;
struct timeval start, end;
long double start_time,end_time;
int N_ELEMENTS = (4194304*2)/sizeof(int);
start_pes(0);
me = _my_pe();
npes = _num_pes();
for (i = 0; i < SHMEM_BCAST_SYNC_SIZE; i += 1)
{
pSync[i] = _SHMEM_SYNC_VALUE;
}
nxtpe = (me+1)%npes;
source = (int *) shmalloc( N_ELEMENTS * sizeof(*source) );
target = (int *) shmalloc( N_ELEMENTS * sizeof(*target) );
if(me == 0)
printf("Get Performance test results:\nSize (Bytes)\t\tTime (Microseconds)\t\tBandwidth (Bytes/Second)\n");
for (i = 0; i < N_ELEMENTS; i += 1) {
source[i] = i + 1;
target[i] = -90;
}
shmem_barrier_all();
/*For int put we take average of all the times realized by a pair of PEs, thus
* reducing effects of physical location of PEs*/
for (i=1;i<=N_ELEMENTS;i=i*2)
{
time_taken = 0;
for(j=0;j<10000;j++){
gettimeofday(&start, NULL);
start_time = (start.tv_sec * 1000000.0) + start.tv_usec;
shmem_int_get(target, source, i,nxtpe);
gettimeofday(&end, NULL);
end_time = (end.tv_sec * 1000000.0) + end.tv_usec;
time_taken = time_taken + (end_time - start_time);
}
shmem_longdouble_sum_to_all(&time_taken, &time_taken,1, 0, 0, npes, pWrk, pSync);
if(me == 0){
time_taken = time_taken/(npes*10000); /*Average time across all PEs for one put*/
if (i*sizeof(i) < 1048576)
printf("%ld \t\t\t\t %ld\t\t\t\t %ld\n",i*sizeof(i),time_taken,(i*sizeof(i))/(time_taken*1000000.0));
else
printf("%ld \t\t\t %ld\t\t\t\t %ld\n",i*sizeof(i),time_taken,(i*sizeof(i))/(time_taken*1000000.0));
}
}
shmem_barrier_all();
shfree(target);
shfree(source);
return 0;
}
