Re: Improving btree performance through specializing by key shape, take 2

On Mon, Sep 25, 2023 at 9:13 AM Matthias van de Meent
<boekewurm+postgres@gmail.com> wrote:
> I think it's fairly complete, and mostly waiting for review.
>
> > I don't have time to do a comprehensive (or even a fairly
> > cursory) analysis of which parts of the patch are helping, and which
> > are marginal or even add no value.
>
> It is a shame that you don't have the time to review this patch.

I didn't say that. Just that I don't have the time (or more like the
inclination) to do most or all of the analysis that might allow us to
arrive at a commitable patch. Most of the work with something like
this is the analysis of the trade-offs, not writing code. There are
all kinds of trade-offs that you could make with something like this,
and the prospect of doing that myself is kind of daunting. Ideally
you'd have made a significant start on that at this point.

> > I have long understood that you gave up on the idea of keeping the
> > bounds across levels of the tree (which does make sense to me), but
> > yesterday the issue became totally muddled by this high key business.
> > That's why I rehashed the earlier discussion, which I had previously
> > understood to be settled.
>
> Understood. I'll see if I can improve the wording to something that is
> more clear about what the optimization entails.

Cool.

--
Peter Geoghegan