Tom Lane wrote:
> I went back and looked at your AllocSetReset patch from last May, and
> liked it better on second consideration.
> I'll fix up the old patch and apply it. I think that's a better answer
> than kluging up the aggregate functions themselves.
Thanks. I also think this fix is better. It speeds up hash join etc. besides
sum(int4) and sum(int2).
Example:
- SQL
select b.bid, count(*)
from accounts a, branches b
where a.bid = b.bid
group by b.bid;
- Query Plan
HashAggregate (cost=23198.06..23198.08 rows=5 width=4)
-> Hash Join (cost=1.06..20698.06 rows=500000 width=4)
Hash Cond: ("outer".bid = "inner".bid)
-> Seq Scan on accounts a (cost=0.00..13197.00 rows=500000
width=4)
-> Hash (cost=1.05..1.05 rows=5 width=4)
-> Seq Scan on branches b (cost=0.00..1.05 rows=5 width=4)
- Result
original: 2.416s
patched: 2.286s
regards,
--- Atsushi Ogawa