[Webkit-unassigned] [Bug 139847] JavaScriptCore deeply nested "call" performance issue

[bugzilla-daemon at webkit.org]

https://bugs.webkit.org/show_bug.cgi?id=139847

--- Comment #7 from Filip Pizlo <fpizlo at apple.com> ---
(In reply to comment #6)
> My guess is we're recursively emitting bytecode in a way that has
> exponential blowup.

Yeah that's totally what's happening.  That's hilarious.

We should just have a back-off on that optimization.  Like, maintain a count of how deep in the "doubling" due to .call, .apply, or other jneq_ptr-based opts.  If more than K deep then don't do the optimization.

Probably the most optimal way to do it would be upside down: don't do the optimization if there are more than K people below you in the tree who want to do it, so that the optimization kicks in for the leaves of that gross call tree.

-- 
You are receiving this mail because:
You are the assignee for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.webkit.org/pipermail/webkit-unassigned/attachments/20170214/02f86d6c/attachment.html>