SlabHash
SlabHash copied to clipboard
owensgroup
A warp-oriented dynamic hash table for GPUs
Thank you for your time.
I noticed that there were races in the code. For example, in line 45 of src/concurrent_map/warp: uint32_t src_unit_data = (next == SlabHashT::A_INDEX_POINTER) ? *(getPointerFromBucket(src_bucket, laneId)) : *(getPointerFromSlab(next, laneId)); Here a...
Modifications were made to SlabHash's `concurrent_map` as well as SlabAlloc to support dynamic growth of the device-side pool allocator used to extend slab lists. In the SlabHash implementation currently available...
Thanks for the SlabHash implementation. When I am trying to use `gpu_hash_table`, there are no API docs in the source code, which really confuses me. So would you mind to...
Deleting a key replaces the key with an [`EMPTY_PAIR_64`](https://github.com/owensgroup/SlabHash/blob/master/src/concurrent_map/warp/delete.cuh#L71). Then the first deleted key will be overwritten during [unique insertion](https://github.com/owensgroup/SlabHash/blob/master/src/concurrent_map/warp/insert.cuh#L131) without checking all the bucket keys.
A variation of `insertPairUnique` where the last instance of the inserted key always survive.
Hi, I tried to use the ConcurentMap with key as 32bit int and the value as 64 bit int. I test it and there is error saying "misaligned address". I...
After the code has been compiled successfully, I tried to run the simple benchmark **python3 ../bench/bencher.py -m 0 -d 0** but it gives me this error: python3 ../bench/bencher.py -m 0...
