[Webkit-unassigned] [Bug 158083] LLInt should support other types of prototype GetById caching.

[bugzilla-daemon at webkit.org]

https://bugs.webkit.org/show_bug.cgi?id=158083

--- Comment #120 from Caio Lima <ticaiolima at gmail.com> ---
(In reply to comment #118)
> Comment on attachment 291039 [details]
> Patch
> 
> I remember Keith did a patch similar to this in the LLInt for prototype
> values and we didn't see a measurable speedup from it. 
> 
> Should we be doing optimizations like this in the LLInt? Is it worth the
> complexity and maintenance 

I think I can't agree more. It is bringing a complex with practically no performance gains. I remember last time I ran Speedometer in x86_64 and I didn't saw gains. The point is that we can get better improvements if JIT=false, but it doesn't represent real world usage.

> What's the best benchmark result we have from this? Do we have any results
> that are not tailored micro benchmarks?

I don't.

What do you think if we just cache getters and forget about customGetter inline? The patch will be much more simpler and we also can take advantage on "super" access in a further case. I can take a look on that and see if it is worth or not.

-- 
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/20161011/3b3ea14e/attachment.html>