Re: index prefetching

On Mon, Nov 24, 2025 at 3:48 PM Andres Freund <andres@anarazel.de> wrote:
> Huh. I wouldn't have expected -march=native to make a huge difference...

Me neither. On the other hand I find that this area is quite sensitive
to icache misses and branch misprediction penalties. This is partly
due to my holding the patch to a very high standard, in terms of
avoiding regressions (at least for simple point lookup queries and
nestloop join queries).

> 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...

I agree that this precise issue isn't necessarily all that important.

My current focus is on completely separating the I/O prefetching parts
of the patch from the core AM interface changes, while avoiding
regressions shown by various microbenchmarks. My experiments with
-march=native were mostly about that -- not about the heap buffer
locking thing specifically. That was just something I noticed in
passing, and found curious.

--
Peter Geoghegan