Re: [WIP] speeding up GIN build with parallel workers

On Fri, 18 Mar 2016 20:40:16 +0300
Dmitry Ivanov <d(dot)ivanov(at)postgrespro(dot)ru> wrote:

> - Currently there's no documentation regarding parallel gin build
> feature and provided GUC variables.

> You could replace direct access to 'rd_id' field with the
> RelationGetRelid macro.

> Parameter 'r' is unused, you could remove it.

> Some of the functions and pieces of code that you've added do not
> comply to the formatting conventions, e. g.

On Fri, 18 Mar 2016 11:18:46 -0400
Robert Haas <robertmhaas(at)gmail(dot)com> wrote:

> Strong +1. The pool of background workers is necessarily quite
> limited and you can't just gobble them up. I'm not saying that it's
> absolutely essential that the leader can also participate, but saying
> that 1 active leader + 1 worker is only 2% faster than 1 passive
> leader + 2 workers is not comparing apples to apples.

On Fri, 18 Mar 2016 09:12:59 +0530
Amit Kapila <amit(dot)kapila16(at)gmail(dot)com> wrote:

> If I understand the above data correctly, then it seems to indicate
> that majority of the work is done in processing the data, so I think
> it should be better if master and worker both can work together.

Thank you for the support and feedback! Fixed all of these in the new
version. Also added new queries to 'gin' regression test, which
immediately helped me catch one silly bug. The new patch is
attached.

Regards,

Constantin S. Pan
Postgres Professional: http://www.postgrespro.com
The Russian Postgres Company