Re: [TCLCORE] Is Tcl bytecode engine is also *in the far future* going to LLVM?

[Simon G. <si...@wh...>]

Alexandre Ferrieux wrote:
>
>
> On Sat, Mar 28, 2009 at 8:06 PM, Tomasz Kosiak <tk...@da...> wrote:
>
>     At this URL
>     http://trentm.com/blog/archives/2009/03/26/unladden-swallow/
>     there is mentioned Google sponsored project to replace CPython’s
>     custom virtual machine with a JIT built on top of LLVM. I have found
>     that news at Planet ActiveState RSS Feed so some people in the
>     community are already aware of this.
>
>     I just wonder what are our long-term plans about Tcl bytecode engine?
>
>
> Interesting. One question to ask, then, is how do the various issues 
> and solution they found (or will find) apply to Tcl.
> Notably, there is the usual question of dynamic typing which makes it 
> uneasy to build large blocks of known-fixed-type values without 
> constant boxing/unboxing (or type verification). Is this different in 
> Python ? Are there optional type annotations ?
>
> Depending on the answers, it would seem that (a) either we're doomed 
> with an underlying type system not really compatible with those 
> superfast optimizations, or (b) we're in roughly the same boat as 
> Python's. If (a) we might target less ambitious goals (like JITting 
> only expr/set blocks with only non-overloaded mathops) or start 
> introducing optional compiler hints. If (b), what about just waiting 
> for this to happen to Python and reaping the surviving good ideas at 
> the end ?
>
> -Alex
>
Does Parrot (http://code.google.com/p/partcl) provide an upgrade path 
for the VM ?

Simon