size_t key_index(dict *d, size_t key) {
size_t index = 0;
size_t slot_index = 0;
index += d->checkpoints[key / 1024];
slot_index = (key / 1024) * 16;
key %= 1024;
while (key >= 64) {
index += popcountll(d->slots[slot_index++]);
key -= 64;
}
index += popcountll(((1ULL << key) - 1) & d->slots[slot_index]);
return index;
}
size_t num_keys(dict *d) {
size_t num = 0;
for (size_t i = 0; i < d->num_slots; i++) {
num += popcountll(d->slots[i]);
}
return num;
}
void finalize_dict(dict *d) {
d->checkpoints = malloc(((d->num_slots / 16) + 1) * sizeof(size_t)); // FIXME
size_t checkpoint = 0;
for (size_t i = 0; i < d->num_slots; i++) {
if (i % 16 == 0) {
d->checkpoints[i / 16] = checkpoint;
}
checkpoint += popcountll(d->slots[i]);
}
}
unsigned picture_pool_Reset(picture_pool_t *pool)
{
unsigned ret;
vlc_mutex_lock(&pool->lock);
assert(pool->refs > 0);
ret = pool->picture_count - popcountll(pool->available);
pool->available = (1ULL << pool->picture_count) - 1;
vlc_mutex_unlock(&pool->lock);
return ret;
}
void computeCardinalities(CUDA_WORD * differences, int * cardIndvSets,int cardDiff, int setSize){
int i,j, offset;
#pragma omp parallel private(i,j,offset) if(cardDiff > CHUNKSIZE) num_threads(12)
{
// fprintf(stderr,"TID %d\n",omp_get_thread_num());
#pragma omp for schedule(dynamic,CHUNKSIZE) nowait
for (i=0; i < cardDiff; i++){
offset = i*setSize;
cardIndvSets[i]=0;
for(j=offset; j < offset+setSize; j++) {
cardIndvSets[i]+= popcountll(differences[j]);
}
}
} //parallel OMP
}
