Re: index prefetching

Hi,

On 2025-11-23 19:03:44 -0500, Peter Geoghegan wrote:
> On Fri, Nov 21, 2025 at 6:31 PM Andres Freund <andres@anarazel.de> wrote:
> > On 2025-11-21 18:14:56 -0500, Peter Geoghegan wrote:
> > > On Fri, Nov 21, 2025 at 5:38 PM Andres Freund <andres@anarazel.de> wrote:
> > > > Another benfit is that it helps even more when there multiple queries running
> > > > concurrently - the high rate of lock/unlock on the buffer rather badly hurts
> > > > scalability.
> > >
> > > I haven't noticed that effect myself. In fact, it seemed to be the
> > > other way around; it looked like it helped most with very low client
> > > count workloads.
> >
> > It's possible that that effect is more visible on larger machines - I did test
> > that on a 2x 24cores/48 threads machine. I do see a smaller effect on a
> > 2x10c/20t machine.
> 
> Update: I find that when I build Postgres with -march=native, I see
> performance characteristics that are much more in line with what you
> saw when you ran your own experiments (experiments with minimizing the
> number of heap buffer locks acquired during index scans).

Huh. I wouldn't have expected -march=native to make a huge difference...

> Are you in the habit of using -march=native? I'm not.

I occasionally use it, but not regularly - I do however use -O3, as I found
that to actually improve performance sufficiently in plenty cases. And it's
something that's much more generally applicable than -march=native?.


I don't think the precise gains here, particularly basedon on quick
prototypes, make that much of a difference. There's so much more optimization
potential other than the amortization of locking costs...

Greetings,

Andres Freund