On Thu, May 12, 2022 at 10:38:42AM +1200, David Rowley wrote:
> On Wed, 11 May 2022 at 14:38, Justin Pryzby <[email protected]> wrote:
> > I wonder if this is also relevant.
> >
> > 65014000b35 Replace polyphase merge algorithm with a simple balanced k-way
> > merge.
>
> Thanks for highlighting that. It very much is relevant. In fact, it
> seems to account for most of the 25% I mentioned. That particular
> test was sorting 10 million tuples with 4MB of work_mem.
>
> I think that "Improve sorting performance (Heikki Linnakangas)" should
> be moved out from "E.1.3.1.2. Indexes" and put below "E.1.3.1.4.
> General Performance"
Yes, good point, moved.
> The text likely should include the words "disk-based" so that it's
> clear that it's not the same as the other line about "in-memory
> sorts". I'd also be open to just having a single line too. I'd vote
> to put Heikki's name first if we did that.
>
> Maybe:
>
> * Improve performance of sorting tuples (Heikki Linnakangas, Ronan
> Dunklau, David Rowley, Thomas Munro)
>
> This improves the merging performance of individual on-disk sort
> batches, reduces memory consumption for in-memory sorts and reduces
> CPU overheads for certain in-memory sorts.
I kept separate entries:
<!--
Author: Heikki Linnakangas <[email protected]>
2021-10-18 [65014000b] Replace polyphase merge algorithm with a simple
balanced
Author: Heikki Linnakangas <[email protected]>
2021-10-25 [166f94377] Clarify the logic in a few places in the new
balanced me
-->
<listitem>
<para>
Improve performance for sorts that exceed work_mem (Heikki Linnakangas)
</para>
<para>
Specifically, switch to a batch sorting algorithm that uses more
output streams internally.
</para>
</listitem>
<!--
Author: David Rowley <[email protected]>
2021-07-22 [91e9e89dc] Make nodeSort.c use Datum sorts for single
column sorts
Author: David Rowley <[email protected]>
2022-04-04 [40af10b57] Use Generation memory contexts to store tuples
in sorts
Author: John Naylor <[email protected]>
2022-04-02 [697492434] Specialize tuplesort routines for different
kinds of abb
-->
<listitem>
<para>
Improve performance and reduce memory consumption of in-memory
sorts (Ronan Dunklau, David Rowley, Thomas Munro)
</para>
</listitem>
--
Bruce Momjian <[email protected]> https://momjian.us
EDB https://enterprisedb.com
Indecision is a decision. Inaction is an action. Mark Batterson
