param: Refactor rounddown
AbandonedPublic
Actions

Authored by aokblast on Feb 27 2026, 7:17 AM.

Tags

None

Referenced Files

	Unknown Object (File)
	Wed, Mar 18, 6:36 AM

	Unknown Object (File)
	Tue, Mar 17, 8:47 PM

	Unknown Object (File)
	Tue, Mar 17, 7:23 AM

	Unknown Object (File)
	Sun, Mar 15, 6:12 AM

	Unknown Object (File)
	Fri, Mar 13, 7:50 AM

	Unknown Object (File)
	Mon, Mar 2, 10:38 AM

	Unknown Object (File)
	Sun, Mar 1, 11:53 AM

	Unknown Object (File)
	Sun, Mar 1, 11:50 AM

View All 11 Files

Subscribers

imp

Details

Reviewers

imp
markj
jhb

Summary

Obviously, a minus + reminder is faster than divide and multiply in most of the platform.

Diff Detail

Repository

rG FreeBSD src repository

Lint

Lint Passed

Unit

No Test Coverage

Build Status

Buildable 71063
Build 67946: arc lint + arc unit

Event Timeline

aokblast created this revision.Feb 27 2026, 7:17 AM

Herald added a subscriber: imp. · View Herald TranscriptFeb 27 2026, 7:17 AM

aokblast requested review of this revision.Feb 27 2026, 7:17 AM

Harbormaster completed remote builds in B71063: Diff 172848.Feb 27 2026, 7:17 AM

aokblast edited the summary of this revision. (Show Details)Feb 27 2026, 7:17 AM

aokblast added reviewers: imp, markj, jhb.

Does it actually make a difference in generated code? I would expect a compiler to be able to do this transformation.

I did some experiment and you are right, the compiler optimize it above O1.
Also, my assumption is wrong since reminder in aarch64 actually convert into udiv + msub, which is not faster than what we previous have.
Thanks!

Revision Contents
Changeset List

Path

Size

sys/

_param.h

2 lines

Diff 172848

View Options

param: Refactor rounddownAbandonedPublicActions