Re: Proposal : Parallel Merge Join

From: Tomas Vondra <tomas(dot)vondra(at)2ndquadrant(dot)com>
To: Dilip Kumar <dilipbalaut(at)gmail(dot)com>, Robert Haas <robertmhaas(at)gmail(dot)com>
Cc: pgsql-hackers <pgsql-hackers(at)postgresql(dot)org>
Subject: Re: Proposal : Parallel Merge Join
Date: 2016-12-28 21:45:03
Views: Raw Message | Whole Thread | Download mbox
Lists: pgsql-hackers


On 12/21/2016 04:53 PM, Dilip Kumar wrote:
> On Wed, Dec 21, 2016 at 8:39 PM, Robert Haas <robertmhaas(at)gmail(dot)com> wrote:
>> Committed the refactoring patch with some mild cosmetic adjustments.
> Thanks..
>> As to the second patch:
>> + if (jointype == JOIN_UNIQUE_INNER)
>> + jointype = JOIN_INNER;
>> Isn't this dead code? save_jointype might that value, but jointype won't.
> Yes, it is.
> I have fixed this, and also observed that comment for
> try_partial_mergejoin_path header was having some problem, fixed that
> too.

FWIW, I've done quite a bit of testing on this patch, and also on the
other patches adding parallel index scans and bitmap heap scan. I've
been running TPC-H and TPC-DS on 16GB data sets with each patch, looking
for regressions or crashes.

I haven't found any of that so far, which is good of course. It however
seems the plan changes only for very few queries in those benchmarks
with any of the patches, even after tweaking the costs to make parallel
plans more likely.

I'm going to try with larger scales and also --enable-cassert and post
the results during CF 2017-1.


Tomas Vondra
PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services

In response to


Browse pgsql-hackers by date

  From Date Subject
Next Message Tom Lane 2016-12-28 23:17:26 Re: Improving RLS planning
Previous Message Peter Geoghegan 2016-12-28 20:57:45 Re: Logical tape pause/resume