Re: LuaJIT 2 clone-on-write for tables

[Alexander Gladysh <agladysh@...>]

On Wed, Jan 19, 2011 at 21:26, Mike Pall <mikelu-1101@mike.de> wrote:
> Alexander Gladysh wrote:

>> So, the question: is there a chance for LuaJIT 2 to somehow
>> automagically optimize out tclone() calls for the nodes that are never
>> changed?

> Nope. That's beyond what one can reasonably expect a compiler
> to do. Compilers don't do magic -- it only looks like that.

Ok, no problem, I kind of expected that answer. :-)

Another silly question, while we're on it: is it possible, in theory,
to create some magical "jit.tclone()" builtin, that would make such
clone-on-write optimization possible?

Alexander.