Horton

Re: Allowing printf("%m") only where it actually works

Tom Lane <tgl@sss.pgh.pa.us>

View thread View on postgr.es
From: Tom Lane <tgl@sss.pgh.pa.us>
To: Thomas Munro <thomas.munro@enterprisedb.com>
Cc: Alvaro Herrera <alvherre@2ndquadrant.com>, PostgreSQL Hackers <pgsql-hackers@lists.postgresql.org>
Date: 2018-08-12T19:35:09Z
Lists: pgsql-hackers

Attachments

I wrote:
> ... So we would be taking a hit on most platforms, but I've not really
> seen sprintf as a major component of very many profiles.  Moreover,
> at least for the elog/ereport use-case, we'd be buying back some
> nontrivial part of that hit by getting rid of expand_fmt_string().
> Also worth noting is that we've never made any effort at all to
> micro-optimize snprintf.c, so maybe there's some gold to be mined
> there to reduce the penalty.

Oh, the plot thickens: apparently, a very large chunk of the time
in that test scenario went into the %g format item, which I think
we can all agree isn't performance-critical for Postgres.  Changing
the test case to test %lld in place of %g, I get (on the same
five platforms as before)

RHEL6:

snprintf time = 357.634 ms total, 0.000357634 ms per iteration
pg_snprintf time = 281.708 ms total, 0.000281708 ms per iteration
ratio = 0.788

macOS:

snprintf time = 155.86 ms total, 0.00015586 ms per iteration
pg_snprintf time = 104.074 ms total, 0.000104074 ms per iteration
ratio = 0.668

FreeBSD:

snprintf time = 268.883 ms total, 0.000268883 ms per iteration
pg_snprintf time = 185.294 ms total, 0.000185294 ms per iteration
ratio = 0.689

OpenBSD:

snprintf time = 276.418 ms total, 0.000276418 ms per iteration
pg_snprintf time = 153.334 ms total, 0.000153334 ms per iteration
ratio = 0.555

NetBSD:

snprintf time = 1174.13 ms total, 0.00117413 ms per iteration
pg_snprintf time = 1508.82 ms total, 0.00150882 ms per iteration
ratio = 1.285

So there's actually a plausible argument to be made that we'd
get a net speed win on most platforms and test cases.

			regards, tom lane

Commits

  1. In pg_log_generic(), be more paranoid about preserving errno.

  2. Make src/common/exec.c's error logging less ugly.

  3. Select appropriate PG_PRINTF_ATTRIBUTE for recent NetBSD.

  4. Fix detection of the result type of strerror_r().

  5. Try another way to detect the result type of strerror_r().

  6. Clean up *printf macros to avoid conflict with format archetypes.

  7. Fix link failures due to snprintf/strerror changes.

  8. Implement %m in src/port/snprintf.c, and teach elog.c to rely on that.

  9. Always use our own versions of *printf().

  10. Incorporate strerror_r() into src/port/snprintf.c, too.

  11. Convert elog.c's useful_strerror() into a globally-used strerror wrapper.

  12. Revert "Distinguish printf-like functions that support %m from those that don't."

  13. Produce compiler errors if errno is referenced inside elog/ereport calls.

  14. Distinguish printf-like functions that support %m from those that don't.

  15. Fix unportable usage of printf("%m").

  16. Be more robust when strerror() doesn't give a useful result.