Mixing
How to synchronize CUDA threads when they are in the same loop and we need to synchronize them to execute...
0
I have written some code, and now I want to implement this on CUDA GPU but I'm new to synchronization. Below I'm presenting the code and I want to that LOOP1 to be executed by all threads (hence I want to this portion to take advantage of CUDA and the remaining portion (the portion other from the LOOP1) is to be executed by only a single thread.
do{
point_set = master_Q[(*num_mas) - 1].q;
List* temp = point_set;
List* pa = point_set;
if(master_Q[num_mas[0] - 1].max)
max_level = (int) (ceilf(il2 * log(master_Q[num_mas[0] - 1].max)));
*num_mas = (*num_mas) - 1;
while(point_set){
List* insert_ele = temp;
while(temp){
insert_ele = temp;
if((insert_ele->dist[insert_ele->dist_index-1] <= pow(2, max_level-1)) || (top_level == max_level)){
if(point_set == temp){
point_set = temp->next;
pa = temp->next;
}
else{
pa->next = temp->next;
}
temp = NULL;
List* new_point_set = point_set;
float maximum_dist = 0;
if(parent->p_index != insert_ele->point_index){
List* tmp = new_point_set;
float *b = &(data[(insert_ele->point_index)*point_len]);
**LOOP 1:** while(tmp){
float *c = &(data[(tmp->point_index)*point_len]);
float sum = 0.;
for(int j = 0; j < point_len; j+=2){
float d1 = b[j] - c[j];
float d2 = b[j+1] - c[j+1];
d1 *= d1;
d2 *= d2;
sum = sum + d1 + d2;
}
tmp->dist[tmp->dist_index] = sqrt(sum);
if(maximum_dist < tmp->dist[tmp->dist_index])
maximum_dist = tmp->dist[tmp->dist_index];
tmp->dist_index = tmp->dist_index+1;
tmp = tmp->next;
}
max_distance = maximum_dist;
}
while(new_point_set || insert_ele){
List* far, *par, *tmp, *tmp_new;
far = NULL;
tmp = new_point_set;
tmp_new = NULL;
float level_dist = pow(2, max_level-1);
float maxdist = 0, maxp = 0;
while(tmp){
if(tmp->dist[(tmp->dist_index)-1] > level_dist){
if(maxdist < tmp->dist[tmp->dist_index-1])
maxdist = tmp->dist[tmp->dist_index-1];
if(tmp == new_point_set){
new_point_set = tmp->next;
par = tmp->next;
}
else{
par->next = tmp->next;
}
if(far == NULL){
far = tmp;
tmp_new = far;
}
else{
tmp_new->next = tmp;
tmp_new = tmp;
}
if(parent->p_index != insert_ele->point_index)
tmp->dist_index = tmp->dist_index - 1;
tmp = tmp->next;
tmp_new->next = NULL;
}
else{
par = tmp;
if(maxp < tmp->dist[(tmp->dist_index)-1])
maxp = tmp->dist[(tmp->dist_index)-1];
tmp = tmp->next;
}
}
if(0 == maxp){
tmp = new_point_set;
aloc_mem[*tree_index].p_index = insert_ele->point_index;
aloc_mem[*tree_index].no_child = 0;
aloc_mem[*tree_index].level = max_level--;
parent->children_index[parent->no_child++] = *tree_index;
parent = &(aloc_mem[*tree_index]);
tree_index[0] = tree_index[0]+1;
while(tmp){
aloc_mem[*tree_index].p_index = tmp->point_index;
aloc_mem[(*tree_index)].no_child = 0;
aloc_mem[(*tree_index)].level = master_Q[(*cur_count_Q)-1].level;
parent->children_index[parent->no_child] = *tree_index;
parent->no_child = parent->no_child + 1;
(*tree_index)++;
tmp = tmp->next;
}
cur_count_Q[0] = cur_count_Q[0]-1;
new_point_set = NULL;
}
master_Q[*num_mas].q = far;
master_Q[*num_mas].parent = parent;
master_Q[*num_mas].valid = true;
master_Q[*num_mas].max = maxdist;
master_Q[*num_mas].level = max_level;
num_mas[0] = num_mas[0]+1;
if(0 != maxp){
aloc_mem[*tree_index].p_index = insert_ele->point_index;
aloc_mem[*tree_index].no_child = 0;
aloc_mem[*tree_index].level = max_level;
parent->children_index[parent->no_child++] = *tree_index;
parent = &(aloc_mem[*tree_index]);
tree_index[0] = tree_index[0]+1;
if(maxp){
int new_level = ((int) (ceilf(il2 * log(maxp)))) +1;
if (new_level < (max_level-1))
max_level = new_level;
else
max_level--;
}
else
max_level--;
}
if( 0 == maxp )
insert_ele = NULL;
}
}
else{
if(NULL == temp->next){
master_Q[*num_mas].q = point_set;
master_Q[*num_mas].parent = parent;
master_Q[*num_mas].valid = true;
master_Q[*num_mas].level = max_level;
num_mas[0] = num_mas[0]+1;
}
pa = temp;
temp = temp->next;
}
}
if((*num_mas) > 1){
List *temp2 = master_Q[(*num_mas)-1].q;
while(temp2){
List* temp3 = master_Q[(*num_mas)-2].q;
master_Q[(*num_mas)-2].q = temp2;
if((master_Q[(*num_mas)-1].parent)->p_index != (master_Q[(*num_mas)-2].parent)->p_index){
temp2->dist_index = temp2->dist_index - 1;
}
temp2 = temp2->next;
master_Q[(*num_mas)-2].q->next = temp3;
}
num_mas[0] = num_mas[0]-1;
}
point_set = master_Q[(*num_mas)-1].q;
temp = point_set;
pa = point_set;
parent = master_Q[(*num_mas)-1].parent;
max_level = master_Q[(*num_mas)-1].level;
if(master_Q[(*num_mas)-1].max)
if( max_level > ((int) (ceilf(il2 * log(master_Q[(*num_mas)-1].max)))) +1)
max_level = ((int) (ceilf(il2 * log(master_Q[(*num_mas)-1].max)))) +1;
num_mas[0] = num_mas[0]-1;
}
}while(*num_mas > 0);
cuda parallel-processing gpu
edited Jan 2 at 21:10
halfer
14.7k759116
asked May 12 '10 at 7:03
VickeyVickey
468
add a comment |
0
I have written some code, and now I want to implement this on CUDA GPU but I'm new to synchronization. Below I'm presenting the code and I want to that LOOP1 to be executed by all threads (hence I want to this portion to take advantage of CUDA and the remaining portion (the portion other from the LOOP1) is to be executed by only a single thread.
do{
point_set = master_Q[(*num_mas) - 1].q;
List* temp = point_set;
List* pa = point_set;
if(master_Q[num_mas[0] - 1].max)
max_level = (int) (ceilf(il2 * log(master_Q[num_mas[0] - 1].max)));
*num_mas = (*num_mas) - 1;
while(point_set){
List* insert_ele = temp;
while(temp){
insert_ele = temp;
if((insert_ele->dist[insert_ele->dist_index-1] <= pow(2, max_level-1)) || (top_level == max_level)){
if(point_set == temp){
point_set = temp->next;
pa = temp->next;
}
else{
pa->next = temp->next;
}
temp = NULL;
List* new_point_set = point_set;
float maximum_dist = 0;
if(parent->p_index != insert_ele->point_index){
List* tmp = new_point_set;
float *b = &(data[(insert_ele->point_index)*point_len]);
**LOOP 1:** while(tmp){
float *c = &(data[(tmp->point_index)*point_len]);
float sum = 0.;
for(int j = 0; j < point_len; j+=2){
float d1 = b[j] - c[j];
float d2 = b[j+1] - c[j+1];
d1 *= d1;
d2 *= d2;
sum = sum + d1 + d2;
}
tmp->dist[tmp->dist_index] = sqrt(sum);
if(maximum_dist < tmp->dist[tmp->dist_index])
maximum_dist = tmp->dist[tmp->dist_index];
tmp->dist_index = tmp->dist_index+1;
tmp = tmp->next;
}
max_distance = maximum_dist;
}
while(new_point_set || insert_ele){
List* far, *par, *tmp, *tmp_new;
far = NULL;
tmp = new_point_set;
tmp_new = NULL;
float level_dist = pow(2, max_level-1);
float maxdist = 0, maxp = 0;
while(tmp){
if(tmp->dist[(tmp->dist_index)-1] > level_dist){
if(maxdist < tmp->dist[tmp->dist_index-1])
maxdist = tmp->dist[tmp->dist_index-1];
if(tmp == new_point_set){
new_point_set = tmp->next;
par = tmp->next;
}
else{
par->next = tmp->next;
}
if(far == NULL){
far = tmp;
tmp_new = far;
}
else{
tmp_new->next = tmp;
tmp_new = tmp;
}
if(parent->p_index != insert_ele->point_index)
tmp->dist_index = tmp->dist_index - 1;
tmp = tmp->next;
tmp_new->next = NULL;
}
else{
par = tmp;
if(maxp < tmp->dist[(tmp->dist_index)-1])
maxp = tmp->dist[(tmp->dist_index)-1];
tmp = tmp->next;
}
}
if(0 == maxp){
tmp = new_point_set;
aloc_mem[*tree_index].p_index = insert_ele->point_index;
aloc_mem[*tree_index].no_child = 0;
aloc_mem[*tree_index].level = max_level--;
parent->children_index[parent->no_child++] = *tree_index;
parent = &(aloc_mem[*tree_index]);
tree_index[0] = tree_index[0]+1;
while(tmp){
aloc_mem[*tree_index].p_index = tmp->point_index;
aloc_mem[(*tree_index)].no_child = 0;
aloc_mem[(*tree_index)].level = master_Q[(*cur_count_Q)-1].level;
parent->children_index[parent->no_child] = *tree_index;
parent->no_child = parent->no_child + 1;
(*tree_index)++;
tmp = tmp->next;
}
cur_count_Q[0] = cur_count_Q[0]-1;
new_point_set = NULL;
}
master_Q[*num_mas].q = far;
master_Q[*num_mas].parent = parent;
master_Q[*num_mas].valid = true;
master_Q[*num_mas].max = maxdist;
master_Q[*num_mas].level = max_level;
num_mas[0] = num_mas[0]+1;
if(0 != maxp){
aloc_mem[*tree_index].p_index = insert_ele->point_index;
aloc_mem[*tree_index].no_child = 0;
aloc_mem[*tree_index].level = max_level;
parent->children_index[parent->no_child++] = *tree_index;
parent = &(aloc_mem[*tree_index]);
tree_index[0] = tree_index[0]+1;
if(maxp){
int new_level = ((int) (ceilf(il2 * log(maxp)))) +1;
if (new_level < (max_level-1))
max_level = new_level;
else
max_level--;
}
else
max_level--;
}
if( 0 == maxp )
insert_ele = NULL;
}
}
else{
if(NULL == temp->next){
master_Q[*num_mas].q = point_set;
master_Q[*num_mas].parent = parent;
master_Q[*num_mas].valid = true;
master_Q[*num_mas].level = max_level;
num_mas[0] = num_mas[0]+1;
}
pa = temp;
temp = temp->next;
}
}
if((*num_mas) > 1){
List *temp2 = master_Q[(*num_mas)-1].q;
while(temp2){
List* temp3 = master_Q[(*num_mas)-2].q;
master_Q[(*num_mas)-2].q = temp2;
if((master_Q[(*num_mas)-1].parent)->p_index != (master_Q[(*num_mas)-2].parent)->p_index){
temp2->dist_index = temp2->dist_index - 1;
}
temp2 = temp2->next;
master_Q[(*num_mas)-2].q->next = temp3;
}
num_mas[0] = num_mas[0]-1;
}
point_set = master_Q[(*num_mas)-1].q;
temp = point_set;
pa = point_set;
parent = master_Q[(*num_mas)-1].parent;
max_level = master_Q[(*num_mas)-1].level;
if(master_Q[(*num_mas)-1].max)
if( max_level > ((int) (ceilf(il2 * log(master_Q[(*num_mas)-1].max)))) +1)
max_level = ((int) (ceilf(il2 * log(master_Q[(*num_mas)-1].max)))) +1;
num_mas[0] = num_mas[0]-1;
}
}while(*num_mas > 0);
cuda parallel-processing gpu
edited Jan 2 at 21:10
halfer
14.7k759116
asked May 12 '10 at 7:03
VickeyVickey
468
3
Your formatting is seriously broken. Make sure your code is all indented by at least four spaces.
– Marcelo Cantos
May 12 '10 at 7:11
1
if you want just loop1 parallel, maybe post it as a separate listing? other stuff really interfering with comprehension
– Anycorn
May 12 '10 at 7:16
add a comment |
0
0
0
I have written some code, and now I want to implement this on CUDA GPU but I'm new to synchronization. Below I'm presenting the code and I want to that LOOP1 to be executed by all threads (hence I want to this portion to take advantage of CUDA and the remaining portion (the portion other from the LOOP1) is to be executed by only a single thread.
do{
point_set = master_Q[(*num_mas) - 1].q;
List* temp = point_set;
List* pa = point_set;
if(master_Q[num_mas[0] - 1].max)
max_level = (int) (ceilf(il2 * log(master_Q[num_mas[0] - 1].max)));
*num_mas = (*num_mas) - 1;
while(point_set){
List* insert_ele = temp;
while(temp){
insert_ele = temp;
if((insert_ele->dist[insert_ele->dist_index-1] <= pow(2, max_level-1)) || (top_level == max_level)){
if(point_set == temp){
point_set = temp->next;
pa = temp->next;
}
else{
pa->next = temp->next;
}
temp = NULL;
List* new_point_set = point_set;
float maximum_dist = 0;
if(parent->p_index != insert_ele->point_index){
List* tmp = new_point_set;
float *b = &(data[(insert_ele->point_index)*point_len]);
**LOOP 1:** while(tmp){
float *c = &(data[(tmp->point_index)*point_len]);
float sum = 0.;
for(int j = 0; j < point_len; j+=2){
float d1 = b[j] - c[j];
float d2 = b[j+1] - c[j+1];
d1 *= d1;
d2 *= d2;
sum = sum + d1 + d2;
}
tmp->dist[tmp->dist_index] = sqrt(sum);
if(maximum_dist < tmp->dist[tmp->dist_index])
maximum_dist = tmp->dist[tmp->dist_index];
tmp->dist_index = tmp->dist_index+1;
tmp = tmp->next;
}
max_distance = maximum_dist;
}
while(new_point_set || insert_ele){
List* far, *par, *tmp, *tmp_new;
far = NULL;
tmp = new_point_set;
tmp_new = NULL;
float level_dist = pow(2, max_level-1);
float maxdist = 0, maxp = 0;
while(tmp){
if(tmp->dist[(tmp->dist_index)-1] > level_dist){
if(maxdist < tmp->dist[tmp->dist_index-1])
maxdist = tmp->dist[tmp->dist_index-1];
if(tmp == new_point_set){
new_point_set = tmp->next;
par = tmp->next;
}
else{
par->next = tmp->next;
}
if(far == NULL){
far = tmp;
tmp_new = far;
}
else{
tmp_new->next = tmp;
tmp_new = tmp;
}
if(parent->p_index != insert_ele->point_index)
tmp->dist_index = tmp->dist_index - 1;
tmp = tmp->next;
tmp_new->next = NULL;
}
else{
par = tmp;
if(maxp < tmp->dist[(tmp->dist_index)-1])
maxp = tmp->dist[(tmp->dist_index)-1];
tmp = tmp->next;
}
}
if(0 == maxp){
tmp = new_point_set;
aloc_mem[*tree_index].p_index = insert_ele->point_index;
aloc_mem[*tree_index].no_child = 0;
aloc_mem[*tree_index].level = max_level--;
parent->children_index[parent->no_child++] = *tree_index;
parent = &(aloc_mem[*tree_index]);
tree_index[0] = tree_index[0]+1;
while(tmp){
aloc_mem[*tree_index].p_index = tmp->point_index;
aloc_mem[(*tree_index)].no_child = 0;
aloc_mem[(*tree_index)].level = master_Q[(*cur_count_Q)-1].level;
parent->children_index[parent->no_child] = *tree_index;
parent->no_child = parent->no_child + 1;
(*tree_index)++;
tmp = tmp->next;
}
cur_count_Q[0] = cur_count_Q[0]-1;
new_point_set = NULL;
}
master_Q[*num_mas].q = far;
master_Q[*num_mas].parent = parent;
master_Q[*num_mas].valid = true;
master_Q[*num_mas].max = maxdist;
master_Q[*num_mas].level = max_level;
num_mas[0] = num_mas[0]+1;
if(0 != maxp){
aloc_mem[*tree_index].p_index = insert_ele->point_index;
aloc_mem[*tree_index].no_child = 0;
aloc_mem[*tree_index].level = max_level;
parent->children_index[parent->no_child++] = *tree_index;
parent = &(aloc_mem[*tree_index]);
tree_index[0] = tree_index[0]+1;
if(maxp){
int new_level = ((int) (ceilf(il2 * log(maxp)))) +1;
if (new_level < (max_level-1))
max_level = new_level;
else
max_level--;
}
else
max_level--;
}
if( 0 == maxp )
insert_ele = NULL;
}
}
else{
if(NULL == temp->next){
master_Q[*num_mas].q = point_set;
master_Q[*num_mas].parent = parent;
master_Q[*num_mas].valid = true;
master_Q[*num_mas].level = max_level;
num_mas[0] = num_mas[0]+1;
}
pa = temp;
temp = temp->next;
}
}
if((*num_mas) > 1){
List *temp2 = master_Q[(*num_mas)-1].q;
while(temp2){
List* temp3 = master_Q[(*num_mas)-2].q;
master_Q[(*num_mas)-2].q = temp2;
if((master_Q[(*num_mas)-1].parent)->p_index != (master_Q[(*num_mas)-2].parent)->p_index){
temp2->dist_index = temp2->dist_index - 1;
}
temp2 = temp2->next;
master_Q[(*num_mas)-2].q->next = temp3;
}
num_mas[0] = num_mas[0]-1;
}
point_set = master_Q[(*num_mas)-1].q;
temp = point_set;
pa = point_set;
parent = master_Q[(*num_mas)-1].parent;
max_level = master_Q[(*num_mas)-1].level;
if(master_Q[(*num_mas)-1].max)
if( max_level > ((int) (ceilf(il2 * log(master_Q[(*num_mas)-1].max)))) +1)
max_level = ((int) (ceilf(il2 * log(master_Q[(*num_mas)-1].max)))) +1;
num_mas[0] = num_mas[0]-1;
}
}while(*num_mas > 0);
cuda parallel-processing gpu
edited Jan 2 at 21:10
halfer
14.7k759116
asked May 12 '10 at 7:03
VickeyVickey
468
I have written some code, and now I want to implement this on CUDA GPU but I'm new to synchronization. Below I'm presenting the code and I want to that LOOP1 to be executed by all threads (hence I want to this portion to take advantage of CUDA and the remaining portion (the portion other from the LOOP1) is to be executed by only a single thread.
do{
point_set = master_Q[(*num_mas) - 1].q;
List* temp = point_set;
List* pa = point_set;
if(master_Q[num_mas[0] - 1].max)
max_level = (int) (ceilf(il2 * log(master_Q[num_mas[0] - 1].max)));
*num_mas = (*num_mas) - 1;
while(point_set){
List* insert_ele = temp;
while(temp){
insert_ele = temp;
if((insert_ele->dist[insert_ele->dist_index-1] <= pow(2, max_level-1)) || (top_level == max_level)){
if(point_set == temp){
point_set = temp->next;
pa = temp->next;
}
else{
pa->next = temp->next;
}
temp = NULL;
List* new_point_set = point_set;
float maximum_dist = 0;
if(parent->p_index != insert_ele->point_index){
List* tmp = new_point_set;
float *b = &(data[(insert_ele->point_index)*point_len]);
**LOOP 1:** while(tmp){
float *c = &(data[(tmp->point_index)*point_len]);
float sum = 0.;
for(int j = 0; j < point_len; j+=2){
float d1 = b[j] - c[j];
float d2 = b[j+1] - c[j+1];
d1 *= d1;
d2 *= d2;
sum = sum + d1 + d2;
}
tmp->dist[tmp->dist_index] = sqrt(sum);
if(maximum_dist < tmp->dist[tmp->dist_index])
maximum_dist = tmp->dist[tmp->dist_index];
tmp->dist_index = tmp->dist_index+1;
tmp = tmp->next;
}
max_distance = maximum_dist;
}
while(new_point_set || insert_ele){
List* far, *par, *tmp, *tmp_new;
far = NULL;
tmp = new_point_set;
tmp_new = NULL;
float level_dist = pow(2, max_level-1);
float maxdist = 0, maxp = 0;
while(tmp){
if(tmp->dist[(tmp->dist_index)-1] > level_dist){
if(maxdist < tmp->dist[tmp->dist_index-1])
maxdist = tmp->dist[tmp->dist_index-1];
if(tmp == new_point_set){
new_point_set = tmp->next;
par = tmp->next;
}
else{
par->next = tmp->next;
}
if(far == NULL){
far = tmp;
tmp_new = far;
}
else{
tmp_new->next = tmp;
tmp_new = tmp;
}
if(parent->p_index != insert_ele->point_index)
tmp->dist_index = tmp->dist_index - 1;
tmp = tmp->next;
tmp_new->next = NULL;
}
else{
par = tmp;
if(maxp < tmp->dist[(tmp->dist_index)-1])
maxp = tmp->dist[(tmp->dist_index)-1];
tmp = tmp->next;
}
}
if(0 == maxp){
tmp = new_point_set;
aloc_mem[*tree_index].p_index = insert_ele->point_index;
aloc_mem[*tree_index].no_child = 0;
aloc_mem[*tree_index].level = max_level--;
parent->children_index[parent->no_child++] = *tree_index;
parent = &(aloc_mem[*tree_index]);
tree_index[0] = tree_index[0]+1;
while(tmp){
aloc_mem[*tree_index].p_index = tmp->point_index;
aloc_mem[(*tree_index)].no_child = 0;
aloc_mem[(*tree_index)].level = master_Q[(*cur_count_Q)-1].level;
parent->children_index[parent->no_child] = *tree_index;
parent->no_child = parent->no_child + 1;
(*tree_index)++;
tmp = tmp->next;
}
cur_count_Q[0] = cur_count_Q[0]-1;
new_point_set = NULL;
}
master_Q[*num_mas].q = far;
master_Q[*num_mas].parent = parent;
master_Q[*num_mas].valid = true;
master_Q[*num_mas].max = maxdist;
master_Q[*num_mas].level = max_level;
num_mas[0] = num_mas[0]+1;
if(0 != maxp){
aloc_mem[*tree_index].p_index = insert_ele->point_index;
aloc_mem[*tree_index].no_child = 0;
aloc_mem[*tree_index].level = max_level;
parent->children_index[parent->no_child++] = *tree_index;
parent = &(aloc_mem[*tree_index]);
tree_index[0] = tree_index[0]+1;
if(maxp){
int new_level = ((int) (ceilf(il2 * log(maxp)))) +1;
if (new_level < (max_level-1))
max_level = new_level;
else
max_level--;
}
else
max_level--;
}
if( 0 == maxp )
insert_ele = NULL;
}
}
else{
if(NULL == temp->next){
master_Q[*num_mas].q = point_set;
master_Q[*num_mas].parent = parent;
master_Q[*num_mas].valid = true;
master_Q[*num_mas].level = max_level;
num_mas[0] = num_mas[0]+1;
}
pa = temp;
temp = temp->next;
}
}
if((*num_mas) > 1){
List *temp2 = master_Q[(*num_mas)-1].q;
while(temp2){
List* temp3 = master_Q[(*num_mas)-2].q;
master_Q[(*num_mas)-2].q = temp2;
if((master_Q[(*num_mas)-1].parent)->p_index != (master_Q[(*num_mas)-2].parent)->p_index){
temp2->dist_index = temp2->dist_index - 1;
}
temp2 = temp2->next;
master_Q[(*num_mas)-2].q->next = temp3;
}
num_mas[0] = num_mas[0]-1;
}
point_set = master_Q[(*num_mas)-1].q;
temp = point_set;
pa = point_set;
parent = master_Q[(*num_mas)-1].parent;
max_level = master_Q[(*num_mas)-1].level;
if(master_Q[(*num_mas)-1].max)
if( max_level > ((int) (ceilf(il2 * log(master_Q[(*num_mas)-1].max)))) +1)
max_level = ((int) (ceilf(il2 * log(master_Q[(*num_mas)-1].max)))) +1;
num_mas[0] = num_mas[0]-1;
}
}while(*num_mas > 0);
cuda parallel-processing gpu
cuda parallel-processing gpu
edited Jan 2 at 21:10
halfer
14.7k759116
asked May 12 '10 at 7:03
VickeyVickey
468
edited Jan 2 at 21:10
halfer
14.7k759116
asked May 12 '10 at 7:03
VickeyVickey
468
edited Jan 2 at 21:10
halfer
14.7k759116
edited Jan 2 at 21:10
halfer
14.7k759116
edited Jan 2 at 21:10
halfer
14.7k759116
14.7k759116
asked May 12 '10 at 7:03
VickeyVickey
468
asked May 12 '10 at 7:03
VickeyVickey
468
asked May 12 '10 at 7:03
VickeyVickey
468
468
3
Your formatting is seriously broken. Make sure your code is all indented by at least four spaces.
– Marcelo Cantos
May 12 '10 at 7:11
1
if you want just loop1 parallel, maybe post it as a separate listing? other stuff really interfering with comprehension
– Anycorn
May 12 '10 at 7:16
add a comment |
3
Your formatting is seriously broken. Make sure your code is all indented by at least four spaces.
– Marcelo Cantos
May 12 '10 at 7:11
1
if you want just loop1 parallel, maybe post it as a separate listing? other stuff really interfering with comprehension
– Anycorn
May 12 '10 at 7:16
3
3
Your formatting is seriously broken. Make sure your code is all indented by at least four spaces.
– Marcelo Cantos
May 12 '10 at 7:11
Your formatting is seriously broken. Make sure your code is all indented by at least four spaces.
– Marcelo Cantos
May 12 '10 at 7:11
1
1
if you want just loop1 parallel, maybe post it as a separate listing? other stuff really interfering with comprehension
– Anycorn
May 12 '10 at 7:16
if you want just loop1 parallel, maybe post it as a separate listing? other stuff really interfering with comprehension
– Anycorn
May 12 '10 at 7:16
add a comment |
1 Answer
1
active
oldest
votes
1
If you want a single kernel to execute parts of its code on only a single thread within a block, use an if statement to execute a portion of code for only a single threadIdx and then barrier synchronize. Perhaps you should take a stab at writing a kernel and posting that for people to look at.
answered May 12 '10 at 17:16
D3C34C34DD3C34C34D
385210
add a comment |
StackExchange.ifUsing("editor", function () {
StackExchange.using("externalEditor", function () {
StackExchange.using("snippets", function () {
StackExchange.snippets.init();
});
});
}, "code-snippets");
StackExchange.ready(function() {
var channelOptions = {
tags: "".split(" "),
id: "1"
};
initTagRenderer("".split(" "), "".split(" "), channelOptions);
StackExchange.using("externalEditor", function() {
// Have to fire editor after snippets, if snippets enabled
if (StackExchange.settings.snippets.snippetsEnabled) {
StackExchange.using("snippets", function() {
createEditor();
});
}
else {
createEditor();
}
});
function createEditor() {
StackExchange.prepareEditor({
heartbeatType: 'answer',
autoActivateHeartbeat: false,
convertImagesToLinks: true,
noModals: true,
showLowRepImageUploadWarning: true,
reputationToPostImages: 10,
bindNavPrevention: true,
postfix: "",
imageUploader: {
brandingHtml: "Powered by u003ca class="icon-imgur-white" href="https://imgur.com/"u003eu003c/au003e",
contentPolicyHtml: "User contributions licensed under u003ca href="https://creativecommons.org/licenses/by-sa/3.0/"u003ecc by-sa 3.0 with attribution requiredu003c/au003e u003ca href="https://stackoverflow.com/legal/content-policy"u003e(content policy)u003c/au003e",
allowUrls: true
},
onDemand: true,
discardSelector: ".discard-answer"
,immediatelyShowMarkdownHelp:true
});
}
});
draft saved
draft discarded
Sign up or log in
StackExchange.ready(function () {
StackExchange.helpers.onClickDraftSave('#login-link');
});
Sign up using Google
Sign up using Facebook
Sign up using Email and Password
Post as a guest
Required, but never shown
StackExchange.ready(
function () {
StackExchange.openid.initPostLogin('.new-post-login', 'https%3a%2f%2fstackoverflow.com%2fquestions%2f2816617%2fhow-to-synchronize-cuda-threads-when-they-are-in-the-same-loop-and-we-need-to-sy%23new-answer', 'question_page');
}
);
Post as a guest
Required, but never shown
1 Answer
1
active
oldest
votes
1 Answer
1
active
oldest
votes
active
oldest
votes
active
oldest
votes
1
If you want a single kernel to execute parts of its code on only a single thread within a block, use an if statement to execute a portion of code for only a single threadIdx and then barrier synchronize. Perhaps you should take a stab at writing a kernel and posting that for people to look at.
answered May 12 '10 at 17:16
D3C34C34DD3C34C34D
385210
add a comment |
1
If you want a single kernel to execute parts of its code on only a single thread within a block, use an if statement to execute a portion of code for only a single threadIdx and then barrier synchronize. Perhaps you should take a stab at writing a kernel and posting that for people to look at.
answered May 12 '10 at 17:16
D3C34C34DD3C34C34D
385210
add a comment |
1
1
1
If you want a single kernel to execute parts of its code on only a single thread within a block, use an if statement to execute a portion of code for only a single threadIdx and then barrier synchronize. Perhaps you should take a stab at writing a kernel and posting that for people to look at.
answered May 12 '10 at 17:16
D3C34C34DD3C34C34D
385210
If you want a single kernel to execute parts of its code on only a single thread within a block, use an if statement to execute a portion of code for only a single threadIdx and then barrier synchronize. Perhaps you should take a stab at writing a kernel and posting that for people to look at.
answered May 12 '10 at 17:16
D3C34C34DD3C34C34D
385210
answered May 12 '10 at 17:16
D3C34C34DD3C34C34D
385210
answered May 12 '10 at 17:16
D3C34C34DD3C34C34D
385210
answered May 12 '10 at 17:16
D3C34C34DD3C34C34D
385210
385210
add a comment |
add a comment |
draft saved
draft discarded
Thanks for contributing an answer to Stack Overflow!
- Please be sure to answer the question. Provide details and share your research!
But avoid …
- Asking for help, clarification, or responding to other answers.
- Making statements based on opinion; back them up with references or personal experience.
To learn more, see our tips on writing great answers.
draft saved
draft discarded
Sign up or log in
StackExchange.ready(function () {
StackExchange.helpers.onClickDraftSave('#login-link');
});
Sign up using Google
Sign up using Facebook
Sign up using Email and Password
Post as a guest
Required, but never shown
StackExchange.ready(
function () {
StackExchange.openid.initPostLogin('.new-post-login', 'https%3a%2f%2fstackoverflow.com%2fquestions%2f2816617%2fhow-to-synchronize-cuda-threads-when-they-are-in-the-same-loop-and-we-need-to-sy%23new-answer', 'question_page');
}
);
Post as a guest
Required, but never shown
Sign up or log in
StackExchange.ready(function () {
StackExchange.helpers.onClickDraftSave('#login-link');
});
Sign up using Google
Sign up using Facebook
Sign up using Email and Password
Post as a guest
Required, but never shown
Sign up or log in
StackExchange.ready(function () {
StackExchange.helpers.onClickDraftSave('#login-link');
});
Sign up using Google
Sign up using Facebook
Sign up using Email and Password
Post as a guest
Required, but never shown
Sign up or log in
StackExchange.ready(function () {
StackExchange.helpers.onClickDraftSave('#login-link');
});
Sign up using Google
Sign up using Facebook
Sign up using Email and Password
Sign up using Google
Sign up using Facebook
Sign up using Email and Password
Post as a guest
Required, but never shown
Required, but never shown
Required, but never shown
Required, but never shown
Required, but never shown
Required, but never shown
Required, but never shown
Required, but never shown
Required, but never shown
3
Your formatting is seriously broken. Make sure your code is all indented by at least four spaces.
– Marcelo Cantos
May 12 '10 at 7:11
1
if you want just loop1 parallel, maybe post it as a separate listing? other stuff really interfering with comprehension
– Anycorn
May 12 '10 at 7:16