|
lock<Mutex, TILE_SIZE>(g, table->locks[bkt_idx]); |
I am reading the code for the
reserve interface.
Under what circumstances might multiple threads compete for the same bucket lock?
It seems that there is no such lock contention within the same kernel. So, could there be scenarios where multiple rehash_kernel_for_fast_mode kernels run in parallel?
I can't figure out how it works. Could someone help me?