Page MenuHomeFreeBSD
Differential D49974

queue(3): Wrap all QMD_PANIC() guards with __predict_false()
ClosedPublic
Actions

Authored by olce on Wed, Apr 23, 10:05 AM.
Tags
None
Referenced Files
F115992521: D49974.diff
Thu, May 1, 7:57 AM
Unknown Object (File)
Mon, Apr 28, 4:39 PM
Unknown Object (File)
Thu, Apr 24, 9:41 PM
Unknown Object (File)
Thu, Apr 24, 12:12 PM
Unknown Object (File)
Thu, Apr 24, 3:04 AM
Unknown Object (File)
Wed, Apr 23, 11:35 PM
Unknown Object (File)
Wed, Apr 23, 9:09 PM
Unknown Object (File)
Wed, Apr 23, 8:43 PM
View All 11 Files
Subscribers
imp

Details

Reviewers
markj
emaste
Commits
rG542e14a59bca: queue(3): Wrap QMD_ASSERT()'s guard with __predict_false()
rG613d66b5e17d: queue(3): Wrap QMD_ASSERT()'s guard with __predict_false()
Summary

All these guards are bound to be almost always false (obviously).

Diff Detail

Repository
rG FreeBSD src repository
Lint
Lint Not Applicable
Unit
Tests Not Applicable

Event Timeline

olce created this revision.Wed, Apr 23, 10:05 AM
Herald added a subscriber: imp. · View Herald TranscriptWed, Apr 23, 10:05 AM
olce requested review of this revision.Wed, Apr 23, 10:05 AM
Harbormaster completed remote builds in B63661: Diff 154152.Wed, Apr 23, 10:05 AM
olce added a parent revision: D49973: queue(3): Debug macros: Finer control knobs, userland support.Wed, Apr 23, 10:19 AM
olce added a child revision: D49975: queue(3): Add simple tests for some macros for all list/tailq types.
emaste accepted this revision.Wed, Apr 23, 12:23 PM

I imagine the compiler might infer this already but I don't think it hurts readability in any significant way, so sure.

This revision is now accepted and ready to land.Wed, Apr 23, 12:23 PM
olce added a comment.Wed, Apr 23, 12:43 PM
In D49974#1139786, @emaste wrote:

I imagine the compiler might infer this already but I don't think it hurts readability in any significant way, so sure.

It could (for the regular panic() case, and more generally "noreturn" functions), although there might be cases where the "noreturn" path is actually the most probable, but this is arguably niche and could understandably be ignored by the compiler. Cases where one does not want to actually panic, but just print a diagnostic message and continue, would benefit from this but are likely edge cases. I was not too sure about doing this change, but as you say it doesn't hurt either.

Closed by commit rG613d66b5e17d: queue(3): Wrap QMD_ASSERT()'s guard with __predict_false() (authored by olce). · Explain WhyMon, Apr 28, 12:24 PM
This revision was automatically updated to reflect the committed changes.
olce added a commit: rG613d66b5e17d: queue(3): Wrap QMD_ASSERT()'s guard with __predict_false().
olce added a commit: rG542e14a59bca: queue(3): Wrap QMD_ASSERT()'s guard with __predict_false().Thu, May 1, 7:55 PM

Revision Contents
Changeset List

PathSize
sys/
sys/
2 lines

Diff 154400

View Options

sys/sys/queue.h

Loading...