On 03.01.2016 04:14, Esteban Gínez wrote:
Hey guys.
I spent some time today reading about Graal and Truffle, thought there
is still lots for me to learn there. I am somewhat curious if people
have though about writing groovy on top of graal and truffle
Thought yes... for several years actually. But time to do that? Sadly no.
One concern is that given our license we can only use Truffle and most
not use Graal directly. I assume this is not a real problem, for a
prototype anyway.
The question is where to interface with Truffle/Graal. We could safe the
bytecode generation step and instead use the AST (or a shadow of it) to
execute the program. But then again we would have the costly class
resolving step in there all the time. So we would first have to write
out some kind of intermediate format, which is "interpreted" by Graal.
Of course right now, bytecode itself looks like a possible choice for
this. But it is imho questionable if an approach like that will really
allow for the kind of optimization, that makes Graal interesting at all.
So I cannot find the gradual path I would like to have instead of a big
implementation, which is required to get a simple program running. And
then of course you would have to rewrite the meta class system to marry
it with truffle to really get something out of this.
Not saying it is no worth it... just saying this can turn out consuming
a lot of time
bye blackdrag
