Indeed. I did actually start toying with an implementing of this after  
the summit, but I quickly run into the problem of missing  
implementations of important intrinsic method handles - I would have  
needed guardWithTest,  insertArgument, and collectArguments for a  
monomorphic cache, a collectArguments for varargs, and so on. All I  
had to work with was Rémi's backport, which doesn't (yet) implement  
these, so I shelved it for the time being, but the overall direction  
seems to be valid.

On a sidenote, I would also be a happy user of a prebuilt MLVM binary  
(for Mac OS X, no less; alternatively, I might really look into  
installing OpenSolaris into VMWare).

Attila.

On 2009.01.21., at 20:56, John Rose wrote:

> As Attila said at the Summit in September, this new stuff probably
> has strong implications on the design of a MOP.  The main example is
> using method handles to express units of procedural semantics; the
> MOP's job of handing out such units is streamlined if those units can
> be directly invoked (as in invokevirtual or invokedynamic) by the MOP
> client, instead of called through a reflective veil of boxed
> varargs.  Getting this working requires that the client code which
> uses the MOP must be able to emit method handle calls.  If the Java
> language supports these things, then the MOP clients can be written
> using javac not just ASM, which makes a JSR 292-based MOP viable.
>
> -- John


--~--~---------~--~----~------------~-------~--~----~
You received this message because you are subscribed to the Google Groups "JVM 
Languages" group.
To post to this group, send email to [email protected]
To unsubscribe from this group, send email to 
[email protected]
For more options, visit this group at 
http://groups.google.com/group/jvm-languages?hl=en
-~----------~----~----~----~------~----~------~--~---

Reply via email to