Re: Reduce pinning in btree indexes

Kyotaro HORIGUCHI <horiguchi.kyotaro@lab.ntt.co.jp> wrote:

> Hello, I measured the performance of this patch considering
> markpos/restorepos. The loss seems to be up to about 10%
> unfortunately.

Thanks for the test case!

I took another look at this optimization, and found that it didn't
really depend on the pin (as I had first thought), so I put it back
(keeping the rest of the patch unchanged).  I saw a 1.4% increase
in run time with the patch (compared to master) for the mark-only
test, and a 1.6% increase in run time for the mark/restore test.
I'll look into what might be causing that.  I have seen bigger
differences just due to changing where executable code crossed
cache boundaries, but this is big enough to worry about; it needs
to be checked.

New patch attached.

--
Kevin Grittner
EDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company