Efficient Hash Probes on Modern Processors

Bucketized versions of Cuckoo hashing can achieve 95– 99% occupancy, without any space overhead for pointers or other structures. However, such methods typically need to consult multiple hash buckets per probe, and have therefore been seen as having worse probe performance than conventional techniques for large tables. We consider workloads typical of database and stream processing, in which keys and payloads are small, and in which a large number of probes are processed in bulk. We show how to improve probe performance by (a) eliminating branch instructions from the probe code, enabling better scheduling and latency-hiding by modern processors, and (b) using SIMD instructions to process multiple keys/payloads in parallel. We show that on modern architectures, probes to a bucketized Cuckoo hash table can be processed much faster than conventional hash table probes, for both small and large memory-resident tables. On a Pentium 4, a probe is two to four times faster, while on the Cell SPE processor a probe is ten times faster.

By: Kenneth A. Ross

Published in: RC24100 in 2006

LIMITED DISTRIBUTION NOTICE:

This Research Report is available. This report has been submitted for publication outside of IBM and will probably be copyrighted if accepted for publication. It has been issued as a Research Report for early dissemination of its contents. In view of the transfer of copyright to the outside publisher, its distribution outside of IBM prior to publication should be limited to peer communications and specific requests. After outside publication, requests should be filled only by reprints or legally obtained copies of the article (e.g., payment of royalties). I have read and understand this notice and am a member of the scientific community outside or inside of IBM seeking a single copy only.

rc24100.pdf

Questions about this service can be mailed to reports@us.ibm.com .