Re: Manipulating complex types as non-contiguous structures in-memory

Andres Freund <andres@anarazel.de>

From: Andres Freund <andres@anarazel.de>
To: Tom Lane <tgl@sss.pgh.pa.us>
Cc: Pavel Stehule <pavel.stehule@gmail.com>, PostgreSQL Hackers <pgsql-hackers@postgresql.org>
Date: 2015-05-11T01:36:02Z
Lists: pgsql-hackers
On 2015-05-10 21:09:14 -0400, Tom Lane wrote:
> Andres Freund <andres@anarazel.de> writes:
> > I'm not sure what exactly to use as a performance benchmark
> > here. For now I chose
> > SELECT * FROM (SELECT ARRAY(SELECT generate_series(1, 10000))) d, generate_series(1, 1000) repeat(i);
> > that'll hit array_out, which uses iterators.
> 
> Hmm, probably those results are swamped by I/O functions though.

I did check with a quick profile, and the iteration itself is a
significant part of the total execution time.

> I'd suggest trying something that exercises array_map(), which
> it looks like means doing an array coercion.  Perhaps like so:

> do $$
> declare a int4[];
> x int;
> begin
>   a := array(select generate_series(1,1000));
>   for i in 1..100000 loop
>     x := array_length(a::int8[], 1);
>   end loop;
> end$$;

with the loop count set to 10000 instead, I get:
before:
after:
tps = 20.940092 (including connections establishing)
after:
tps = 20.568730 (including connections establishing)

Greetings,

Andres Freund


Commits

  1. Use fast path in plpgsql's RETURN/RETURN NEXT in more cases.

  2. Add support for multiple kinds of external toast datums.