On Thu, Mar 14, 2024 at 5:26 PM Tomas Vondra
<tomas.vondra@enterprisedb.com> wrote:
>
> On 3/14/24 19:16, Melanie Plageman wrote:
> > On Thu, Mar 14, 2024 at 03:32:04PM +0200, Heikki Linnakangas wrote:
> >> ...
> >>
> >> Ok, committed that for now. Thanks for looking!
> >
> > Attached v6 is rebased over your new commit. It also has the "fix" in
> > 0010 which moves BitmapAdjustPrefetchIterator() back above
> > table_scan_bitmap_next_block(). I've also updated the Streaming Read API
> > commit (0013) to Thomas' v7 version from [1]. This has the update that
> > we theorize should address some of the regressions in the bitmapheapscan
> > streaming read user in 0014.
> >
>
> Should I rerun the benchmarks with these new patches, to see if it
> really helps with the regressions?
That would be awesome!
I will soon send out a summary of what we investigated off-list about
0010 (though we didn't end up concluding anything). My "fix" (leaving
BitmapAdjustPrefetchIterator() above table_scan_bitmap_next_block())
eliminates the regression in 0010 on the one example that I repro'd
upthread, but it would be good to know if it eliminates the
regressions across some other tests.
I think it would be worthwhile to run the subset of tests which seemed
to fare the worst on 0010 against the patches 0001-0010-- cyclic
uncached on your xeon machine with 4 parallel workers, IIRC -- even
the 1 million scale would do the trick, I think.
And then separately run the subset of tests which seemed to do the
worst on 0014. There were several groups of issues across the
different tests, but I think that the uniform pages data test would be
relevant to use. It showed the regressions with eic 0.
As for the other regressions showing with 0014, I think we would want
to see at least one with fully-in-shared-buffers and one with fully
uncached. Some of the fixes were around pinning fewer buffers when the
blocks were already in shared buffers.
- Melanie