Better multithreading performance on Lua

[Asko Kauppi <askok@...>]

I intended to suggest a change in Lua, only to realize it already  
suits my needs perfectly!  :)

That is, states created in other threads (Lua Lanes) can be given  
their own allocator functions (or pools), so as to avoid cache line  
conflicts when two states would be using memory areas close to each  
other (< 64 bytes on current x86).

Currently, Lanes uses the default allocator for all states.

Question is, has anyone got performance knowledge about a change like  
this, in using Lua and multiple states running in different threads.  
I will eventually make the change, but it would be nice to know what  
to expect.

Oh, and thanks to the Authors for already having the necessary tools  
Right There.

-asko