From: | Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> |
---|---|
To: | Peter Eisentraut <peter(dot)eisentraut(at)enterprisedb(dot)com> |
Cc: | Krunal Bauskar <krunalbauskar(at)gmail(dot)com>, Michael Paquier <michael(at)paquier(dot)xyz>, pgsql-hackers(at)postgresql(dot)org |
Subject: | Re: Improving spin-lock implementation on ARM. |
Date: | 2020-11-27 23:45:57 |
Message-ID: | 734069.1606520757@sss.pgh.pa.us |
Views: | Raw Message | Whole Thread | Download mbox | Resend email |
Thread: | |
Lists: | pgsql-hackers |
Peter Eisentraut <peter(dot)eisentraut(at)enterprisedb(dot)com> writes:
> I tried this on a M1 MacBook Air. I cannot reproduce these results.
> The unpatched numbers are about in the neighborhood of what you showed,
> but the patched numbers are only about a few percent better, not the
> 1.5x or 2x change that you showed.
After redoing the test, I can't find any outside-the-noise difference
at all between HEAD and the patch. So clearly, I screwed up yesterday.
The most likely theory is that I managed to measure an assert-enabled
build of HEAD.
It might be that this hardware is capable of showing a difference with a
better-tuned pgbench test, but with an untuned pgbench run, we just aren't
sufficiently sensitive to the spinlock properties. (Which I guess is good
news, really.)
One thing that did hold up is that the thermal performance of this box
is pretty ridiculous. After being beat on for a solid hour, the fan
still hasn't turned on to any noticeable level, and the enclosure is
only a little warm to the touch. Try that with Intel hardware ;-)
regards, tom lane
From | Date | Subject | |
---|---|---|---|
Next Message | Andreas Karlsson | 2020-11-28 00:04:09 | What to do about the broken btree_gist for inet/cidr? |
Previous Message | Michael Banck | 2020-11-27 22:49:59 | [Doc Patch] Clarify that CREATEROLE roles can GRANT default roles |