start_pos array should have type uint64_t

superliuxz / DBSCAN

Density-Based Spatial Clustering of Applications with Noise

GNU General Public License v2.0

9 stars 2 forks source link

start_pos array should have type uint64_t #17

Open superliuxz opened 4 years ago

superliuxz commented 4 years ago

start_pos stores the indices of neighbours; and neighbours could have |V|^2 size. Even though |V| is bounded by uint32, |V|^2 is not guaranteed. Hence it's possible to overflow start_pos at 32 bit width.

superliuxz commented 4 years ago

need to change:

data type for totl_num_nbs_
type for start_pos,
and everhwhere these two values are used; for tot_num_nbs, there is a CUDA memcpy D2H operation, need to change the size of that from 32 to 64 as well.
- in calc_start_pos, last_vtx_start_pos should be 64 bit