[Webkit-unassigned] [Bug 149338] [ES6] "super" and "this" should be lexically bound inside an arrow function and should live in a JSLexicalEnvironment

[bugzilla-daemon at webkit.org]

https://bugs.webkit.org/show_bug.cgi?id=149338

--- Comment #77 from Saam Barati <sbarati at apple.com> ---
(In reply to comment #76)
> (In reply to comment #73)
> > I renamed the bug to better indicate what the patch is doing.
> > 
> > How is performance on this bug? Can you run the perf tests
> > just to make sure we haven't regressed on things non-arrow function related.
> > We may have regressed slightly on some arrow-function tests, but that's
> > expected. Also, this will get better once you implement some basic static
> > analysis for "this" usage so we don't store "this" in the environment unless
> > we really need to.
> 
> See result by link https://bugs.webkit.org/attachment.cgi?id=265179, Results
> are neutral except following tests:
> arrowfunction-call                                95.9133+-5.5082     !   
> 338.3422+-14.9020       ! definitely 3.5276x slower
> 
> and small performance degradation:
>  function-with-eval                               708.1514+-8.0792     !   
> 748.3060+-10.7312       ! definitely 1.0567x slower
> is-object-or-null-tricky-internal-function   
>                                                      59.8600+-1.4660     !  
> 74.9355+-10.1182       ! definitely 1.2518x slower
> setter                                            50.4615+-2.1227     !    
> 64.9183+-12.1934       ! definitely 1.2865x slower

These are expected.
Eval is slower b/c we must be pessimistic and assume that it could have an arrow function.
The arrowfunction-call is a tiny micro-benchmark. It will get back to normal speed once you implement the patch that statically analyzes an arrow function to see if it really needs the lexical environment created for it.

-- 
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/20151110/4f9891d6/attachment.html>