Horton

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

Amit Langote <amitlangote09@gmail.com>

View thread View on postgr.es
From: Amit Langote <amitlangote09@gmail.com>
To: David Rowley <dgrowleyml@gmail.com>
Cc: Alvaro Herrera <alvherre@alvh.no-ip.org>, Andres Freund <andres@anarazel.de>, Andrew Dunstan <andrew@dunslane.net>, Michael Paquier <michael@paquier.xyz>, John Naylor <john.naylor@enterprisedb.com>, "Jonathan S. Katz" <jkatz@postgresql.org>, Tom Lane <tgl@sss.pgh.pa.us>, Peter Geoghegan <pg@bowt.ie>, PostgreSQL-development <pgsql-hackers@postgresql.org>, Nikita Glukhov <n.gluhov@postgrespro.ru>
Date: 2022-07-22T05:49:33Z
Lists: pgsql-hackers 
On Fri, Jul 22, 2022 at 1:13 PM David Rowley <dgrowleyml@gmail.com> wrote:
> On Fri, 22 Jul 2022 at 15:22, Amit Langote <amitlangote09@gmail.com> wrote:
> > BTW, the only way I found to *forcefully* exercise llvm_compile_expr()
> > is to add `set jit_above_cost to 0` at the top of the test file, or
> > are we missing a force_jit_mode, like there is force_parallel_mode?
>
> I don't think we'd need any setting to hide the JIT counters from
> EXPLAIN ANALYZE since those only show with COSTS ON, which we tend not
> to do.

Ah, makes sense.

> I think for testing, you could just zero all the jit*above_cost GUCs.
>
> If you look at the config_extra in [1], you'll see that animal runs
> the tests with modified JIT parameters.
>
> BTW, I was working on code inside llvm_compile_expr() a few days ago
> and I thought I'd gotten the new evaluation steps I was adding correct
> as it worked fine with jit_above_cost=0, but on further testing, it
> crashed with jit_inline_above_cost=0. Might be worth doing both to see
> if everything works as intended.

Thanks for the pointer.

So I didn't see things going bust on re-testing with all
jit_*_above_cost parameters set to 0, so maybe the
llvm_compile_expression() additions are alright.

-- 
Thanks, Amit Langote
EDB: http://www.enterprisedb.com

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