我有一个while循环嵌套问题，谁能帮我解答一下？
global void kdtree_bbf_kernel(struct kd_node* kd_root,struct kd_node* kd_node_num,struct feature* feat,int n2,int k,int num,int max_nn_chks, struct min_pq min_pq,struct kd_node* expl,struct kd_node* unexpl)
{
int i,j;
int tid=threadIdx.x+blockIdx.xblockDim.x;
if(tid<n2)
{
int t=0;
min_pq[tid].n=0;
min_pq[tid].nallocd=MINPQ_INIT_NALLOCD;
minpq_insert(&min_pq[tid],kd_root,0);
while(min_pq[tid].n>0 && t<max_nn_chks)
{
minpq_extract_min(&min_pq[tid],&expl[tid]);
if(! &expl[tid])
{
printf(“Warning: PQ unexpectedly empty.\n”);
return;
}
explore_to_leaf(&min_pq[tid],&expl[tid],&unexpl[tid],&feat[tid],kd_node_num);
if(! &expl[tid])
{
printf(“Warning: PQ unexpectedly empty.\n”);
return;
}
t++;
}
tid+=blockDim.xgridDim.x;
}
}
device int minpq_insert(struct min_pq* min_pq,struct kd_node* data,int key)
{
int n=min_pq->n;
min_pq->pq_array[n].key=INT_MAX;
min_pq->pq_array[n].data[0]=data[0];
decrease_pq_node_key( min_pq->pq_array, min_pq->n, key );
min_pq->n++;
return 0;
}
device void minpq_extract_min( struct min_pq* min_pq,struct kd_node* expl )
{
if( min_pq->n < 1 )
{
return;
}
expl[0] = min_pq->pq_array[0].data[0];
min_pq->n–;
min_pq->pq_array[0] = min_pq->pq_array[min_pq->n];
restore_minpq_order( min_pq->pq_array, 0, min_pq->n );
}
device void explore_to_leaf(struct min_pq* min_pq,struct kd_node* expl,struct kd_node* unexpl,struct feature* feat,struct kd_node* kd_node_num)
{
double kv;
int ki;
while( expl && ! expl->leaf )
{
ki=expl[0].ki;
kv=expl[0].kv;
if(ki>=feat->d)
{
return;
}
if(feat->descr[ki]<=kv)
{
unexpl[0] = kd_node_num[expl[0].index_right];
expl[0] = kd_node_num[expl[0].index_left];
}
else
{
unexpl[0] = kd_node_num[expl[0].index_left];
expl[0] = kd_node_num[expl[0].index_right];
}
if(minpq_insert( min_pq,unexpl,ABS(kv-feat->descr[ki] ) ) )
{
return;
}
}
}
如果去掉explore_to_leaf函数中的 if(minpq_insert( min_pq,unexpl,ABS(kv-feat->descr[ki] ) ) ) { return; }kernel中的while只循环一次，运行时间为30多ms，如果加上if(minpq_insert( min_pq,unexpl,ABS(kv-feat->descr[ki] ) ) ) { return; }，那么它就和kernel函数中的while循环构成循环嵌套，kernel中的while循环多次，时间就会升到900多毫秒，还没有在CPU上串行快，这是怎么回事啊？应该怎么改啊？