void* roundrobin(void* tparam) {
ptrdiff_t tid = (ptrdiff_t)tparam;
int offset = tid*N_ELEMS;
/* fprintf(stderr,"Starting thread %lu with offset %d\n",tid,offset); */
int nextpe = (shmem_my_pe()+1)%shmem_n_pes();
int prevpe = (shmem_my_pe()-1 + shmem_n_pes())%shmem_n_pes();
shmem_long_put(target+offset, source+offset, N_ELEMS, nextpe);
/* fprintf(stderr,"Thread %lu done first put\n",tid); */
pthread_barrier_wait(&fencebar);
if(tid == 0) shmem_barrier_all();
pthread_barrier_wait(&fencebar);
shmem_long_get(source+offset, target+offset, N_ELEMS, prevpe);
/* fprintf(stderr,"Thread %lu done first get\n",tid); */
pthread_barrier_wait(&fencebar);
if(tid == 0) shmem_barrier_all();
pthread_barrier_wait(&fencebar);
shmem_long_get(target+offset, source+offset, N_ELEMS, nextpe);
/* fprintf(stderr,"Thread %lu done second get\n",tid); */
pthread_barrier_wait(&fencebar);
if(tid == 0) shmem_barrier_all();
pthread_barrier_wait(&fencebar);
/* fprintf(stderr,"Done thread %lu\n",tid); */
return 0;
}
/****************************************************************************
* Place for Test Item functions
***************************************************************************/
static int test_item1(void)
{
int rc = TC_PASS;
long* shmem_addr = NULL;
int my_proc = 0;
int root = 0;
int i = 0, j = 0;
long original_value, prev_value, new_value, check_value;
long *lock_value = 0;
my_proc = _my_pe();
shmem_addr = shmalloc(sizeof(*shmem_addr));
lock_value = shmalloc(sizeof(*lock_value));
*lock_value = 0;
shmem_barrier_all();
if (my_proc < (int)sizeof(long))
{
for (i = 0; ((i < COUNT_VALUE) && (rc == TC_PASS)); i++)
{
shmem_long_get(&original_value, shmem_addr, 1, root);
shmem_set_lock(lock_value);
new_value = my_proc;
prev_value = shmem_long_cswap(shmem_addr, original_value, new_value, root);
for (j = 0; j < CHECK_COUNT_VALUE; j++)
{
shmem_long_get(&check_value, shmem_addr, 1, root);
if (check_value != new_value && check_value != prev_value)
{
rc = TC_FAIL;
break;
}
}
shmem_clear_lock(lock_value);
}
}
if (shmem_addr)
{
shfree(shmem_addr);
}
shfree(lock_value);
return rc;
}
int
main (int argc, char **argv)
{
long dest;
long src;
int me, npes;
shmem_init ();
me = shmem_my_pe ();
npes = shmem_n_pes ();
src = 42;
shmem_barrier_all ();
if (me == 0) {
shmem_long_get (&dest, &src, 1, 1);
}
shmem_barrier_all ();
shmem_finalize ();
return 0;
}
int
main(int argc, char* argv[])
{
int i;
shmem_init();
if (shmem_n_pes() == 1) {
printf("%s: Requires number of PEs > 1\n", argv[0]);
shmem_finalize();
return 0;
}
if (shmem_my_pe() == 0) {
memset(target, 0, sizeof(target));
/* put 10 elements into target on PE 1 */
shmem_long_get(target, source, 10, 1);
}
shmem_barrier_all(); /* sync sender and receiver */
if (shmem_my_pe() == 0) {
if (0 != memcmp(source, target, sizeof(long) * 10)) {
fprintf(stderr,"[%d] Src & Target mismatch?\n",shmem_my_pe());
for (i = 0 ; i < 10 ; ++i) {
printf("%ld,%ld ", source[i], target[i]);
}
printf("\n");
shmem_global_exit(1);
}
}
shmem_finalize();
return 0;
}
void
FORTRANIFY (shmem_get) (long *target, const long *source, int *size, int *pe)
{
shmem_long_get (target, source, *size, *pe);
}
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 i,j;
long modj,oldj,oldxmodj,newcount;
int my_pe,n_pes;
size_t max_elements_bytes;
static long *x;
shmem_init();
my_pe = shmem_my_pe();
n_pes = shmem_n_pes();
#ifdef HAVE_SET_CACHE_INV
shmem_set_cache_inv();
#endif
/* fail if trying to use only one processor */
if ( n_pes <= 1 ){
fprintf(stderr, "FAIL - test requires at least two PEs\n");
exit(1);
}
if(my_pe == 0)
fprintf(stderr, "shmem_lock_set_clear(%s) n_pes=%d\n", argv[0],n_pes);
/* shmalloc x on all pes (only use the one on PE 0) */
max_elements_bytes = (size_t) (sizeof(long) * n_pes);
x = shmem_malloc( max_elements_bytes );
for(i=0; i<n_pes; i++)
x[i] = 0;
count = 0;
shmem_barrier_all();
for(i=0; i<ITER; i++) {
if (my_pe != 0) {
/* emulate oldj = shmem_long_finc(&count, 0); */
shmem_set_lock(&lock);
shmem_long_get(&oldj,&count,1,0); /* get oldj from PE 0's count */
newcount = oldj+1;
shmem_long_put(&count,&newcount,1,0); /* update count on PE 0 */
shmem_quiet; /* insure that write completes */
shmem_clear_lock(&lock);
/* end of emulation */
modj = (oldj % (n_pes-1)); /* PE 0 is just the counter/checker */
/* increment value in x[modj] */
oldxmodj = shmem_long_finc(&x[modj], 0);
/* printf("PE=%d,oldj=%ld,modj=%ld,oldxmodj=%ld\n",my_pe,oldj,modj,oldxmodj); */
}
}
shmem_barrier_all();
if (my_pe == 0) { /* check x[j] array on PE 0 */
for(j=1 ; j<n_pes; j++) {
if (x[j-1] != (long) ITER)
fprintf(stderr, "FAIL PE %d of %d: x[%d] = %ld expected = %ld\n",
my_pe, n_pes, j-1, x[j-1], (long) ITER);
}
}
shmem_barrier_all();
#ifdef NEEDS_FINALIZE
shmem_finalize();
#endif
return 0;
}
void
shmemi_broadcast32_tree (void *target, const void *source,
size_t nlong,
int PE_root, int PE_start,
int logPE_stride, int PE_size, long *pSync)
{
int child_l, child_r, parent;
const int step = 1 << logPE_stride;
int my_pe = GET_STATE (mype);
int *target_ptr, *source_ptr;
int no_children;
long is_ready, lchild_ready, rchild_ready;
is_ready = 1;
lchild_ready = -1;
rchild_ready = -1;
shmem_long_wait_until (&pSync[0], SHMEM_CMP_EQ, SHMEM_SYNC_VALUE);
shmem_long_wait_until (&pSync[1], SHMEM_CMP_EQ, SHMEM_SYNC_VALUE);
pSync[0] = 0;
pSync[1] = 0;
target_ptr = (int *) target;
source_ptr = (int *) source;
set_2tree (PE_start, step, PE_size, &parent, &child_l, &child_r, my_pe);
no_children = 0;
build_tree (PE_start, step, PE_root, PE_size,
&parent, &child_l, &child_r, my_pe);
shmemi_trace (SHMEM_LOG_BROADCAST,
"before broadcast, R_child = %d L_child = %d",
child_r, child_l);
/* The actual broadcast */
if (PE_size > 1) {
if (my_pe == (PE_start + step * PE_root)) {
pSync[0] = SHMEM_SYNC_VALUE;
if (child_l != -1) {
shmem_long_get (&lchild_ready, (const long *) &pSync[0],
1, child_l);
while (lchild_ready != 0)
shmem_long_get (&lchild_ready, &pSync[0], 1, child_l);
shmem_int_put (target_ptr, source_ptr, nlong, child_l);
shmem_fence ();
shmem_long_put (&pSync[0], &is_ready, 1, child_l);
no_children = 1;
}
if (child_r != -1) {
shmem_long_get (&rchild_ready, &pSync[0], 1, child_r);
while (rchild_ready != 0)
shmem_long_get (&rchild_ready, &pSync[0], 1, child_r);
shmem_int_put (target_ptr, source_ptr, nlong, child_r);
shmem_fence ();
shmem_long_put (&pSync[0], &is_ready, 1, child_r);
no_children = 2;
}
shmem_long_wait_until (&pSync[1], SHMEM_CMP_EQ,
(long) no_children);
pSync[1] = SHMEM_SYNC_VALUE;
}
else {
shmem_long_wait_until (&pSync[0], SHMEM_CMP_EQ, is_ready);
pSync[0] = SHMEM_SYNC_VALUE;
shmemi_trace (SHMEM_LOG_BROADCAST, "inside else");
memcpy (source_ptr, target_ptr, nlong * sizeof (int));
if (child_l != -1) {
shmem_long_get (&lchild_ready, &pSync[0], 1, child_l);
while (lchild_ready != 0)
shmem_long_get (&lchild_ready, &pSync[0], 1, child_l);
shmem_int_put (target_ptr, source_ptr, nlong, child_l);
shmem_fence ();
shmem_long_put (&pSync[0], &is_ready, 1, child_l);
no_children = 1;
}
if (child_r != -1) {
shmem_long_get (&rchild_ready, &pSync[0], 1, child_r);
while (rchild_ready != 0)
shmem_long_get (&rchild_ready, &pSync[0], 1, child_r);
shmem_int_put (target_ptr, source_ptr, nlong, child_r);
shmem_fence ();
shmem_long_put (&pSync[0], &is_ready, 1, child_r);
no_children = 2;
}
pSync[0] = SHMEM_SYNC_VALUE;
if (no_children == 0) {
pSync[1] = SHMEM_SYNC_VALUE;
/* TO DO: Is check for parents pSync required? */
shmem_long_inc (&pSync[1], parent);
}
else {
shmem_long_wait_until (&pSync[1], SHMEM_CMP_EQ,
(long) no_children);
pSync[1] = SHMEM_SYNC_VALUE;
/* printf("PE %d incrementing child count on PE
%d\n",my_pe,parent); */
shmem_long_inc (&pSync[1], parent);
}
}
shmemi_trace (SHMEM_LOG_BROADCAST, "at the end of bcast32");
/* shmem_barrier(PE_start, logPE_stride, PE_size, pSync); */
}
}
void
shmemi_barrier_tree (int PE_start, int logPE_stride, int PE_size, long *pSync)
{
int child_l, child_r, parent;
const int step = 1 << logPE_stride;
int my_pe = GET_STATE (mype);
int no_children;
long is_ready, lchild_ready, rchild_ready;
is_ready = 1;
lchild_ready = -1;
rchild_ready = -1;
shmem_long_wait_until (&pSync[0], SHMEM_CMP_EQ, SHMEM_SYNC_VALUE);
shmem_long_wait_until (&pSync[1], SHMEM_CMP_EQ, SHMEM_SYNC_VALUE);
/* printf("Tree barrier\n"); */
set_2tree (PE_start, step, PE_size, &parent, &child_l, &child_r, my_pe);
no_children = 0;
shmemi_trace (SHMEM_LOG_BARRIER,
"before barrier, R_child = %d L_child = %d",
child_r, child_l);
/* The actual barrier */
if (PE_size > 1) {
pSync[0] = 0;
pSync[1] = 0;
if (my_pe == PE_start) {
pSync[0] = SHMEM_SYNC_VALUE;
if (child_l != -1) {
shmem_long_get (&lchild_ready, (const long *) &pSync[0],
1, child_l);
while (lchild_ready != 0)
shmem_long_get (&lchild_ready, &pSync[0], 1, child_l);
shmem_long_put (&pSync[0], &is_ready, 1, child_l);
no_children = 1;
}
if (child_r != -1) {
shmem_long_get (&rchild_ready, &pSync[0], 1, child_r);
while (rchild_ready != 0)
shmem_long_get (&rchild_ready, &pSync[0], 1, child_r);
shmem_long_put (&pSync[0], &is_ready, 1, child_r);
no_children = 2;
}
shmem_long_wait_until (&pSync[1], SHMEM_CMP_EQ,
(long) no_children);
pSync[1] = SHMEM_SYNC_VALUE;
}
else {
shmem_long_wait_until (&pSync[0], SHMEM_CMP_EQ, is_ready);
shmemi_trace (SHMEM_LOG_BARRIER, "inside else");
if (child_l != -1) {
shmem_long_get (&lchild_ready, &pSync[0], 1, child_l);
while (lchild_ready != 0)
shmem_long_get (&lchild_ready, &pSync[0], 1, child_l);
shmem_long_put (&pSync[0], &is_ready, 1, child_l);
no_children = 1;
}
if (child_r != -1) {
shmem_long_get (&rchild_ready, &pSync[0], 1, child_r);
while (rchild_ready != 0)
shmem_long_get (&rchild_ready, &pSync[0], 1, child_r);
shmem_long_put (&pSync[0], &is_ready, 1, child_r);
no_children = 2;
}
pSync[0] = SHMEM_SYNC_VALUE;
if (no_children == 0) {
pSync[1] = SHMEM_SYNC_VALUE;
shmem_long_inc (&pSync[1], parent);
}
else {
shmem_long_wait_until (&pSync[1], SHMEM_CMP_EQ,
(long) no_children);
pSync[1] = SHMEM_SYNC_VALUE;
shmem_long_inc (&pSync[1], parent);
}
}
shmemi_trace (SHMEM_LOG_BARRIER, "at the end of barrier");
}
}
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;
}
