mbuf: make m_pullup() an inline over most expected case
ClosedPublic
Actions

Authored by glebius on Mon, Oct 12, 11:26 PM.

Details

Reviewers

pouria
gallatin
zlei

Group Reviewers

network
transport

Commits

rG671babc79497: mbuf: make m_pullup() an inline over most expected case

Summary

Implement a most typical use of m_pullup() as an inline and leave only the
heavy part as a function. A sweeping change to make benefit of the new
version follows.

Diff Detail

Repository

rG FreeBSD src repository

Lint

Lint Skipped

Unit

Tests Skipped

Build Status

Buildable 78110
Build 74993: arc lint + arc unit

Event Timeline

glebius created this revision.Mon, Oct 12, 11:26 PM

Herald added subscribers: olce, imp. · View Herald TranscriptMon, Oct 12, 11:26 PM

glebius requested review of this revision.Mon, Oct 12, 11:26 PM

Harbormaster completed remote builds in B78110: Diff 189684.Mon, Oct 12, 11:26 PM

glebius added a child revision: D60689: net*: use inlined m_pullup().Mon, Oct 12, 11:26 PM

gallatin accepted this revision.Mon, Oct 12, 11:43 PM

This revision is now accepted and ready to land.Mon, Oct 12, 11:43 PM

I like this idea !

sys/sys/mbuf.h
865	3rd party drivers should be re-compiled, otherwise they may fail to load, as they may still reference to the old symbol `m_pullup`. Any plan to MFC this to stable branches ?
880	This inline function is too small. Does it deserves a `__always_inline` ?

glebius added inline comments.Tue, Oct 13, 5:54 AM

sys/sys/mbuf.h
865	I don't have a strong desire to MFC it, but I will not object if anybody wants to handle it.
880	I prefer to trust the compiler and don't give it such explicit instruction. I bet our version of clang does the right thing and we don't care about others. It is just a very tiny performance optimization.