Re: New LuaJIT benchmark results

[Florian Weimer <fw@...>]

* Geoff Leyland:

> On 10/03/2010, at 10:34 AM, Florian Weimer wrote:
>
>> * Geoff Leyland:
>> 
>>> what's the story with the binary-trees benchmark?  Is it a GC thing,
>>> a recursive thing or is it a bad benchmark?
>> 
>> ...But perhaps using closures instead of tables can
>> result in competitive performance.
>
> I tried.  Either I'm not very good at closures or it's slower.

It currently is.  But I think it's possible to change closure layout
more easily than to change table layout.  After all, you can
statically check that captured variables are not written to after
capture, and remove the upvalues indirection.  But it's a rather
complex transformation.