Sometimes we inline a function even though the call has an empty type barrier and we know the caller will be invalidated as soon as the inlined function "returns".
Attachment #8889811 - Flags: review?(bhackett1024)
Attachment #8889811 - Flags: review?(bhackett1024) → review+
Pushed by firstname.lastname@example.org: https://hg.mozilla.org/integration/mozilla-inbound/rev/1455629cb663 Don't Ion-inline functions when the call has an empty type barrier. r=bhackett
I might have regressed AngularJS-TodoMVC-CompletingAllItems-sync by 50% . I'll take a look.
(In reply to Jan de Mooij [:jandem] from comment #3) > I might have regressed AngularJS-TodoMVC-CompletingAllItems-sync by 50% . > I'll take a look. I backed this out for now: https://hg.mozilla.org/integration/mozilla-inbound/rev/91c459dd13e5d05cd5b707ae717801a6b04d6031 This patch actually did the right thing (not inlining a function in a branch that was never taken), but it did affect some other inlining heuristics that I think are broken. I'll get back to this soon.
Status: RESOLVED → REOPENED
Resolution: FIXED → ---
backed out from m-c https://hg.mozilla.org/mozilla-central/rev/91c459dd13e5
You need to log in before you can comment on or make changes to this bug.