Hi, Thomas! > 14 окт. 2018 г., в 9:18, Thomas Munro <thomas.mu...@enterprisedb.com> > написал(а): > > + /* Prefetch the bucket for the next key */ > + uint32 next_hash = hash_uint32(DatumGetInt32(keyval) + 1); > + uint32 next_bucket = next_hash % hashtable->nbuckets; > + __builtin_prefetch(&hashtable->buckets.unshared[next_bucket]);
+1, I also think that we should use __builtin_prefetch these days (years, actually). Exactly after listening Anastassia Ailamaki's (author of referenced paper) talk on VLDB I've proposed to do that for B-tree [0], but did not really pursuit that idea. [0] https://www.postgresql.org/message-id/3B774C9E-01E8-46A7-9642-7830DC1108F1%40yandex-team.ru
