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 FORTRANIFY (shmem_addr_accessible) (void *addr, int *pe)
{
return shmem_addr_accessible (addr, *pe);
}
/* 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 me, num_procs, l, j;
int Verbose = 0;
start_pes(0);
me = _my_pe();
num_procs = _num_pes();
if ( num_procs < 2 ) {
if (me ==0)
printf("PE[0] requires 2 or more PEs?\n");
return 1;
}
for (l = 0 ; l < loops ; ++l) {
if ((src_int = shmalloc(2*num_procs*sizeof(int))) == NULL) {
printf("PE-%d int shmalloc() failed?\n", me);
exit(1);
}
dst_int = &src_int[num_procs];
for(j=0; j < num_procs; j++) {
src_int[j] = 4;
dst_int[j] = 0;
}
if ((src_float = shmalloc(2*num_procs*sizeof(float))) == NULL) {
printf("PE-%d float shmalloc() failed?\n", me);
exit(1);
}
dst_float = &src_float[num_procs];
for(j=0; j < num_procs; j++) {
src_float[j] = 4.0;
dst_float[j] = 0.0;
}
if ((src_double = shmalloc(2*num_procs*sizeof(double))) == NULL) {
printf("PE-%d double shmalloc() failed?\n", me);
exit(1);
}
dst_double = &src_double[num_procs];
for(j=0; j < num_procs; j++) {
src_double[j] = 8.0;
dst_double[j] = 0.0;
}
if ((src_long = shmalloc(2*num_procs*sizeof(long))) == NULL) {
printf("PE-%d long shmalloc() failed?\n", me);
exit(1);
}
dst_long = &src_long[num_procs];
for(j=0; j < num_procs; j++) {
src_long[j] = 8;
dst_long[j] = 0;
}
if ((src_llong = shmalloc(2*num_procs*sizeof(long long))) == NULL) {
printf("PE-%d long shmalloc() failed?\n", me);
exit(1);
}
dst_llong = &src_llong[num_procs];
for(j=0; j < num_procs; j++) {
src_llong[j] = 16;
dst_llong[j] = 0;
}
shmem_barrier_all();
if ( me != 0 ) {
/* is 'src_*' accessible from PE0? should be. */
if (!shmem_addr_accessible(src_int,0)) {
printf("PE-%d local src_int %p not accessible from PE-%d?\n",
me, (void*)src_int, 0);
exit(1);
}
if (!shmem_addr_accessible(src_float,0)) {
printf("PE-%d local src_float %p not accessible from PE-%d?\n",
me, (void*)src_float, 0);
exit(1);
}
if (!shmem_addr_accessible(src_double,0)) {
printf("PE-%d local src_double %p not accessible from PE-%d?\n",
me, (void*)src_double, 0);
exit(1);
}
if (!shmem_addr_accessible(src_long,0)) {
printf("PE-%d local src_long %p not accessible from PE-%d?\n",
me, (void*)src_long, 0);
exit(1);
}
if (!shmem_addr_accessible(src_llong,0)) {
printf("PE-%d local src_llong %p not accessible from PE-%d?\n",
me, (void*)src_llong, 0);
exit(1);
}
}
shmem_barrier_all();
if ( me == 0 ) {
shmem_quiet();
for(j=1; j < num_procs; j++) {
dst_int[j] = shmem_int_swap(src_int+j,0,j);
if (dst_int[j] != 4) {
printf("PE-%d dst_int[%d] %d != 4?\n",me,j,dst_int[j]);
exit(1);
}
}
shmem_barrier_all();
/* verify remote data */
for(j=1; j < num_procs; j++) {
itmp = shmem_int_g(src_int+j,j);
if (itmp != 0) {
printf("PE-0 int PE[%d] rem(%d) != 0?\n",j,itmp);
exit(1);
}
/* swap back */
dst_int[j] = shmem_int_swap(src_int+j,dst_int[j],j);
if (dst_int[j] != 0) {
printf("PE-0 dst_int[%d] %d != 0?\n",j,dst_int[j]);
exit(1);
}
itmp = shmem_int_g(src_int+j,j);
if (itmp != 4) {
printf("PE-0 PE[%d] rem %d != 4?\n",j,itmp);
exit(1);
}
}
for(j=1; j < num_procs; j++) {
dst_float[j] = shmem_float_swap(src_float+j,0.0,j);
if (dst_float[j] != 4.0) {
printf("PE-0 dst_float[%d] %f != 4.0?\n",j,dst_float[j]);
exit(1);
}
/* verify remote data */
ftmp = shmem_float_g(src_float+j,j);
if (ftmp != 0.0) {
printf("PE-0 float rem(%f) != 0.0?\n",ftmp);
exit(1);
}
/* swap back */
dst_float[j] = shmem_float_swap(src_float+j,dst_float[j],j);
if (dst_float[j] != 0.0) {
printf("PE-0 dst_float[%d] %f != 0.0?\n",j,dst_float[j]);
exit(1);
}
ftmp = shmem_float_g(src_float+j,j);
if (ftmp != 4.0) {
printf("PE-%d float rem(%f) != 4.0?\n",me,ftmp);
exit(1);
}
}
for(j=1; j < num_procs; j++) {
dst_double[j] = shmem_double_swap(src_double+j,0.0,j);
if (dst_double[j] != 8.0) {
printf("PE-0 dst_double[%d] %f != 8.0?\n",j,dst_double[j]);
exit(1);
}
/* verify remote data */
dtmp = shmem_double_g(src_double+j,j);
if (dtmp != 0.0) {
printf("PE-0 float rem(%f) != 0.0?\n",dtmp);
exit(1);
}
dst_double[j] = shmem_double_swap(src_double+j,dst_double[j],j);
if (dst_double[j] != 0.0) {
printf("PE-0 dst_double[%d] %f != 0.0?\n",j,dst_double[j]);
exit(1);
}
dtmp = shmem_double_g(src_double+j,j);
if (dtmp != 8.0) {
printf("PE-0 double rem(%f) != 8.0?\n",dtmp);
exit(1);
}
}
for(j=1; j < num_procs; j++) {
dst_long[j] = shmem_long_swap(src_long+j,0,j);
if (dst_long[j] != 8) {
printf("PE-0 dst_long[%d] %ld != 8?\n",j,dst_long[j]);
exit(1);
}
}
shmem_barrier_all();
/* verify remote data */
for(j=1; j < num_procs; j++) {
ltmp = shmem_long_g(src_long+j,j);
if (ltmp != 0) {
printf("PE-0 PE[%d]long rem(%ld) != 0?\n",j,ltmp);
exit(1);
}
/* swap back */
dst_long[j] = shmem_long_swap(src_long+j,dst_long[j],j);
if (dst_long[j] != 0) {
printf("PE-%d dst_long[%d] %ld != 0?\n",me,j,dst_long[j]);
exit(1);
}
ltmp = shmem_long_g(src_long+j,j);
if (ltmp != 8) {
printf("PE-%d long rem(%ld) != 8?\n",me,ltmp);
exit(1);
}
}
for(j=1; j < num_procs; j++) {
dst_llong[j] = shmem_longlong_swap(src_llong+j,0,j);
if (dst_llong[j] != 16) {
printf("PE-%d dst_llong[%d] %lld != 16?\n",me,j,dst_llong[j]);
exit(1);
}
}
shmem_barrier_all();
/* verify remote data */
for(j=1; j < num_procs; j++) {
lltmp = shmem_longlong_g(src_llong+j,j);
if (lltmp != 0) {
printf("PE-%d long long rem(%lld) != 0?\n",me,lltmp);
exit(1);
}
/* swap back */
dst_llong[j] = shmem_longlong_swap(src_llong+j,dst_llong[j],j);
if (dst_llong[j] != 0) {
printf("PE-%d dst_llong[%d] %lld != 0?\n", me,j,dst_llong[j]);
exit(1);
}
lltmp = shmem_longlong_g(src_llong+j,j);
if (lltmp != 16) {
printf("PE-%d longlong rem(%lld) != 16?\n",me,lltmp);
exit(1);
}
}
}
else {
shmem_int_wait_until(&src_int[me],SHMEM_CMP_EQ,0);
shmem_barrier_all();
shmem_long_wait_until(&src_long[me],SHMEM_CMP_EQ,0);
shmem_barrier_all();
shmem_longlong_wait_until(&src_llong[me],SHMEM_CMP_EQ,0);
shmem_barrier_all();
}
shmem_barrier_all();
shfree(src_int);
shfree(src_float);
shfree(src_double);
shfree(src_long);
shfree(src_llong);
}
if (Verbose)
fprintf(stderr,"[%d] exit\n",_my_pe());
return 0;
}
int
main(int argc, char* argv[])
{
int me, num_pes, l, pe;
int Verbose = 0;
start_pes(0);
me = _my_pe();
num_pes = _num_pes();
for (l = 0 ; l < loops ; ++l) {
if ((src_int = shmalloc(sizeof(int))) == NULL) {
printf("PE-%d int shmalloc() failed?\n", me);
exit(1);
}
*src_int = 4;
dst_int = itmp = 0;
if ((src_long = shmalloc(sizeof(long))) == NULL) {
printf("PE-%d long shmalloc() failed?\n", me);
exit(1);
}
*src_long = 8;
dst_long = ltmp = 0;
if ((src_llong = shmalloc(sizeof(long long))) == NULL) {
printf("PE-%d long long shmalloc() failed?\n", me);
exit(1);
}
*src_llong = 16;
dst_llong = lltmp = 0;
//printf("PE-%d malloc()s done.\n",me);
shmem_barrier_all();
if ( me == 0 ) {
/* integer swap */
itmp = shmem_int_g(src_int,1);
Vprintf("PE-0 Initial Conditions(int) local %d rem(%d)\n",
dst_int,itmp);
dst_int = shmem_int_cswap(src_int,*src_int,0,1);
if (dst_int != 4) {
printf("PE-%d dst_int %d != 4?\n",me,dst_int);
exit(1);
}
/* verify remote data */
itmp = shmem_int_g(src_int,1);
if (itmp != 0) {
printf("PE-%d rem %d != 0?\n",me,itmp);
exit(1);
}
Vprintf("PE-0 1st int_cswap done: local %d rem(%d)\n",dst_int,itmp);
dst_int = shmem_int_cswap(src_int,0,dst_int,1);
if (dst_int != 0) {
printf("PE-%d dst_int %d != 0?\n",me,dst_int);
exit(1);
}
/* verify remote data */
itmp = shmem_int_g(src_int,1);
if (itmp != 4) {
printf("PE-%d rem %d != 4?\n",me,itmp);
exit(1);
}
Vprintf("PE-0 2nd int_swap done: local %d rem(%d)\n",dst_int,itmp);
/* cswap() should not swap as cond(0) != remote(4) */
dst_int = shmem_int_cswap(src_int,0,0,1);
if (dst_int != 4) {
printf("PE-%d int no-swap returned dst_int %d != 4?\n",
me,dst_int);
exit(1);
}
/* verify previous cswap() did not swap */
itmp = shmem_int_g(src_int,1);
if (itmp != 4) {
printf("PE-%d failed cond int_cswap() swapped? rem(%d) != 4?\n",
me,itmp);
exit(1);
}
/* long swap */
ltmp = shmem_long_g(src_long,1);
Vprintf("PE-0 Initial Conditions(long) local %ld rem(%ld)\n",
dst_long,ltmp);
dst_long = shmem_long_cswap(src_long,*src_long,0,1);
if (dst_long != 8) {
printf("PE-%d dst_long %ld != 8?\n",me,dst_long);
exit(1);
}
/* verify remote data */
ltmp = shmem_long_g(src_long,1);
if (ltmp != 0) {
printf("PE-%d long rem(%ld) != 0?\n",me,ltmp);
exit(1);
}
Vprintf("PE-0 1st long_cswap done: local %ld rem(%ld)\n",
dst_long,ltmp);
dst_long = shmem_long_cswap(src_long,0,dst_long,1);
if (dst_long != 0) {
printf("PE-%d dst_long %ld != 0?\n",me,dst_long);
exit(1);
}
/* verify remote data */
ltmp = shmem_long_g(src_long,1);
if (ltmp != 8) {
printf("PE-%d long rem(%ld) != 8?\n",me,ltmp);
exit(1);
}
Vprintf("PE-0 2nd long_swap done: local %ld rem(%ld)\n",
dst_long,ltmp);
/* cswap() should not swap as cond(0) != remote(8) */
dst_long = shmem_long_cswap(src_long,0,0,1);
if (dst_long != 8) {
printf("PE-%d long no-swap returned dst_long %ld != 8?\n",
me,dst_long);
exit(1);
}
/* verify previous cswap() did not swap */
ltmp = shmem_long_g(src_long,1);
if (ltmp != 8) {
printf("PE-%d failed cond long_cswap() swapped? rem(%ld) != 8?\n",
me,ltmp);
exit(1);
}
/* long long swap */
lltmp = shmem_longlong_g(src_llong,1);
Vprintf("PE-0 Initial Conditions(long long) local %lld rem(%lld)\n",
dst_llong,lltmp);
dst_llong = shmem_longlong_cswap(src_llong,*src_llong,0,1);
if (dst_llong != 16) {
printf("PE-%d dst_llong %lld != 16?\n",me,dst_llong);
exit(1);
}
/* verify remote data */
lltmp = shmem_longlong_g(src_llong,1);
if (lltmp != 0) {
printf("PE-%d longlong rem(%lld) != 0?\n",me,lltmp);
exit(1);
}
Vprintf("PE-0 1st longlong_cswap done: local %lld rem(%lld)\n",
dst_llong, lltmp);
dst_llong = shmem_longlong_cswap(src_llong,0,dst_llong,1);
if (dst_llong != 0) {
printf("PE-%d dst_llong %lld != 0?\n",me,dst_llong);
exit(1);
}
/* verify remote data */
lltmp = shmem_longlong_g(src_llong,1);
if (lltmp != 16) {
printf("PE-%d long long rem(%lld) != 16?\n",me,lltmp);
exit(1);
}
Vprintf("PE-0 2nd longlong_swap done: local %lld rem(%lld)\n",
dst_llong,lltmp);
/* cswap() should not swap as cond(0) != remote(8) */
dst_llong = shmem_longlong_cswap(src_llong,0,0,1);
if (dst_llong != 16) {
printf("PE-%d longlong no-swap returned dst_llong %lld != 16?\n",
me,dst_llong);
exit(1);
}
/* verify previous cswap() did not swap */
lltmp = shmem_longlong_g(src_llong,1);
if (lltmp != 16) {
printf("PE-0 failed cond longlong_cswap() swapped? rem(%lld) != 16?\n",
lltmp);
exit(1);
}
}
else {
if (!shmem_addr_accessible(src_int,0)) {
printf("PE-%d local src_int %p not accessible from PE-%d?\n",
me, (void*)src_int, 0);
exit(1);
}
if (!shmem_addr_accessible(src_long,0)) {
printf("PE-%d local src_long %p not accessible from PE-%d?\n",
me, (void*)src_long, 0);
exit(1);
}
if (!shmem_addr_accessible(src_llong,0)) {
printf("PE-%d local src_llong %p not accessible from PE-%d?\n",
me, (void*)src_llong, 0);
exit(1);
}
}
shmem_barrier_all();
/* shmem_*fadd() exercise */
if (me == 0) {
itmp = 0;
ltmp = 0;
lltmp = 0;
*src_int = 0;
*src_long = 0;
*src_llong = 0;
}
shmem_barrier_all();
(void)shmem_int_fadd( &itmp, me+1, 0 );
(void)shmem_long_fadd( <mp, me+1, 0 );
(void)shmem_longlong_fadd( &lltmp, me+1, 0 );
shmem_barrier_all();
if (me == 0) {
int tot;
for(pe=0,tot=0; pe < num_pes; pe++)
tot += pe+1;
if ( itmp != tot )
printf("fadd() total %d != expected %d?\n",itmp,tot);
if ( ltmp != (long)tot )
printf("fadd() total %ld != expected %d?\n",ltmp,tot);
if ( lltmp != (long long)tot )
printf("fadd() total %lld != expected %d?\n",lltmp,tot);
}
shmem_barrier_all();
(void)shmem_int_finc(src_int,0);
(void)shmem_long_finc(src_long,0);
(void)shmem_longlong_finc(src_llong,0);
shmem_barrier_all();
if (me == 0) {
int tot = num_pes;
if ( *src_int != tot )
printf("finc() total %d != expected %d?\n",*src_int,tot);
if ( *src_long != (long)tot )
printf("finc() total %ld != expected %d?\n",*src_long,tot);
if ( *src_llong != (long long)tot )
printf("finc() total %lld != expected %d?\n",*src_llong,tot);
}
shmem_barrier_all();
shfree(src_int);
shfree(src_long);
shfree(src_llong);
}
if (Verbose)
fprintf(stderr,"[%d] exit\n",_my_pe());
return 0;
}
static int
check_it (void *addr)
{
return shmem_addr_accessible (addr, 1);
}