Hi all,
I'll be running an open discussion session at PGConf.dev 2026 (Tuesday
schedule [1]) titled "Path to batched and vectorized execution
primitives in core Postgres executor".
I'm especially hoping to hear from folks who have worked on executor
internals, the TAM API, expression evaluation, or custom
scan/execution paths in Postgres --- including those who have resorted
to hooks or extensions to get batched or vectorized behavior and
wished core carried more of that weight --- as well as anyone who has
built or worked on batched/vectorized execution in other database
systems and can speak to which design choices succeeded and which
failed.
Some of you may have followed the "Batching in executor" thread [2]
that I started last year. This session is a chance to hash out the
broader design questions in person.
If you're at the conference and have experience or opinions in this
area, please come. If you would like to attend, it would be nice to
hear from you in advance by replying to this email or off-list.
--
Thanks, Amit Langote
[1] https://2026.pgconf.dev/schedule/tuesday
[2] https://www.postgresql.org/message-id/CA%2BHiwqGkway9dcpTVmRs0Q%3DZsQQ8Nx0MRjy9jfDmR8QFLnd-Mg%40mail.gmail.com
