Re: RE: Delaying/avoiding BTreeTupleGetNAtts() call within _bt_compare()

On Mon, Nov 2, 2020 at 9:46 AM Anastasia Lubennikova
<a(dot)lubennikova(at)postgrespro(dot)ru> wrote:
> This thread was inactive for a while. The latest review suggests that it is Ready For Committer.
> I also took a quick look at the patch and agree that it looks sensible. Maybe add a comment before the _bt_compare_inl() to explain the need for this code change.

Actually I am probably going to withdraw this patch soon. The idea is
a plausible way of improving things. But at the same time I cannot
really demonstrate any benefit on hardware that I have access to.

This patch was something I worked on based on a private complaint from
Andres (who is CC'd here now) during an informal conversation at pgDay
SF. If Andres is now in a position to test the patch and can show a
benefit on certain hardware, I may well pick it back up. But as things
stand the evidence in support of the patch is pretty weak. I'm not
going to commit a patch like this unless and until it's crystal clear
what the benefits are.

if Andres cannot spend any time on this in the foreseeable future then
I'll withdraw the patch. I intend to formally withdraw the patch on
November 9th, provided no new information comes to light.

Thanks
--
Peter Geoghegan