Horton

Re: [PoC] Improve dead tuple storage for lazy vacuum

Masahiko Sawada <sawada.mshk@gmail.com>

View thread View on postgr.es
From: Masahiko Sawada <sawada.mshk@gmail.com>
To: John Naylor <johncnaylorls@gmail.com>
Cc: PostgreSQL-development <pgsql-hackers@postgresql.org>, Nathan Bossart <nathandbossart@gmail.com>, Andres Freund <andres@anarazel.de>, Matthias van de Meent <boekewurm+postgres@gmail.com>, Yura Sokolov <y.sokolov@postgrespro.ru>
Date: 2023-12-11T06:17:37Z
Lists: pgsql-hackers

Attachments

On Fri, Dec 8, 2023 at 9:44 PM Masahiko Sawada <sawada.mshk@gmail.com> wrote:
>
> On Fri, Dec 8, 2023 at 7:46 PM John Naylor <johncnaylorls@gmail.com> wrote:
> >
> > On Fri, Dec 8, 2023 at 3:06 PM Masahiko Sawada <sawada.mshk@gmail.com> wrote:
> > >
> > > BTW Given that the actual value size can be calculated only by the
> > > caller, how does the tree know if the value is embedded or not? It's
> > > probably related to how to store combined pointer/value slots.
> >
> > Right, this is future work. At first, variable-length types will have
> > to be single-value leaves. In fact, the idea for storing up to 3
> > offsets in the bitmap header could be done this way -- it would just
> > be a (small) single-value leaf.
>
> Agreed.
>
> >
> > (Reminder: Currently, fixed-length values are compile-time embeddable
> > if the platform pointer size is big enough.)
> >
> > > If leaf
> > > nodes have a bitmap array that indicates the corresponding slot is an
> > > embedded value or a pointer to a value, it would be easy.
> >
> > That's the most general way to do it. We could do it much more easily
> > with a pointer tag, although for the above idea it may require some
> > endian-aware coding. Both were mentioned in the paper, I recall.
>
> True. Probably we can use the combined pointer/value slots approach
> only if the tree is able to use the pointer tagging. That is, if the
> caller allows the tree to use one bit of the value.
>
> I'm going to update the patch based on the recent discussion (RT_SET()
> and variable-length values) etc., and post the patch set early next
> week.

I've attached the updated patch set. From the previous patch set, I've
merged patches 0007 to 0010. The other changes such as adding RT_GET()
still are unmerged for now, for discussion. Probably we can make them
as follow-up patches as we discussed. 0011 to 0015 patches are new
changes for v44 patch set, which removes RT_SEARCH() and RT_SET() and
support variable-length values.

Regards,

-- 
Masahiko Sawada
Amazon Web Services: https://aws.amazon.com

Commits

  1. radixtree: Fix SIGSEGV at update of embeddable value to non-embeddable.

  2. Get rid of anonymous struct

  3. Teach radix tree to embed values at runtime

  4. Teach TID store to skip bitmap for small numbers of offsets

  5. Use bump context for TID bitmaps stored by vacuum

  6. Fix alignment of stack variable

  7. Use TidStore for dead tuple TIDs storage during lazy vacuum.

  8. Rethink create and attach APIs of shared TidStore.

  9. Fix inconsistent function prototypes with function definitions.

  10. Fix a calculation in TidStoreCreate().

  11. Fix potential integer handling issue in radixtree.h.

  12. Add TIDStore, to store sets of TIDs (ItemPointerData) efficiently.

  13. Fix link error for test_radixtree module on Windows

  14. Blind attempt to fix ODR violations

  15. Fix incorrect format specifier for int64

  16. Fix redefinition of typedefs

  17. Add template for adaptive radix tree

  18. Fix signedness error in 9f225e992 for gcc

  19. Introduce helper SIMD functions for small byte arrays

  20. Optimize vacuuming of relations with no indexes.

  21. Add bound check before bsearch() for performance

  22. Allocate consecutive blocks during parallel seqscans