Horton

Re: PG 15 (and to a smaller degree 14) regression due to ExprEvalStep size

David Rowley <dgrowleyml@gmail.com>

View thread View on postgr.es
From: David Rowley <dgrowleyml@gmail.com>
To: Andres Freund <andres@anarazel.de>
Cc: Tom Lane <tgl@sss.pgh.pa.us>, Peter Geoghegan <pg@bowt.ie>, PostgreSQL-development <pgsql-hackers@postgresql.org>, Nikita Glukhov <n.gluhov@postgrespro.ru>, Andrew Dunstan <andrew@dunslane.net>
Date: 2022-06-28T23:40:45Z
Lists: pgsql-hackers

Attachments

On Sat, 18 Jun 2022 at 08:06, Andres Freund <andres@anarazel.de> wrote:
> I also attached my heavily-WIP patches for the ExprEvalStep issues, I
> accidentally had only included a small part of the contents of the json fix.

I've now looked at the 0003 patch.  I like the idea you have about
moving some of the additional fields into ScalarArrayOpExprHashTable.
I think the patch can even go a little further and move the hash_finfo
into there too. This means we don't need to dereference the "op" in
saop_element_hash().

To make this work, I did need to tag the ScalarArrayOpExpr into the
ExprEvalStep.  That's required now since some of the initialization of
the hash function fields is delayed until
ExecEvalHashedScalarArrayOp().  We need to know the
ScalarArrayOpExpr's hashfuncid and inputcollid.

Your v2 patch did shift off some of this initialization work to
ExecEvalHashedScalarArrayOp().  The attached v3 takes that a bit
further.  This saves a bit more work for ScalarArrayOpExprs that are
evaluated 0 times.

Another small thing which I considered doing was to put the
hash_fcinfo_data field as the final field in
ScalarArrayOpExprHashTable so that we could allocate the memory for
the hash_fcinfo_data in the same allocation as the
ScalarArrayOpExprHashTable.  This would reduce the pointer
dereferencing done in saop_element_hash() a bit further.  I just
didn't notice anywhere else where we do that for FunctionCallInfo, so
I resisted doing this.

(There was also a small bug in your patch where you mistakenly cast to
an OpExpr instead of ScalarArrayOpExpr when you were fetching the
inputcollid)

David

Commits

  1. JSON_TABLE: Add support for NESTED paths and columns

  2. Add basic JSON_TABLE() functionality

  3. Add SQL/JSON query functions

  4. Add soft error handling to some expression nodes

  5. Adjust populate_record_field() to handle errors softly

  6. Refactor code used by jsonpath executor to fetch variables

  7. Add more SQL/JSON constructor functions

  8. SQL/JSON: support the IS JSON predicate

  9. SQL/JSON: add standard JSON constructor functions

  10. Add static assertion ensuring sizeof(ExprEvalStep) <= 64 bytes

  11. Remove size increase in ExprEvalStep caused by hashed saops

  12. pgstat: reduce timer overhead by leaving timer running.

  13. expression eval: Fix EEOP_JSON_CONSTRUCTOR and EEOP_JSONEXPR size.

  14. SQL/JSON query functions

  15. Speedup ScalarArrayOpExpr evaluation