On 05/05/2011 21:46, Diego Novillo wrote:
On Thu, May 5, 2011 at 15:16, Diego Novillo<dnovi...@google.com> wrote:
I've always found MELT interesting as an exercise in impossibility and
[ ... ]
There was some confusion about my meaning here. It was meant as a
compliment. Sorry for the confusion.
Diego.
Yeah, Basile's work is very impressive.
From what I can say (Basile might answer better), the idea is to offer
more and more abstraction in order to manipulate as easily as possible
the GCC structures. In particular, MELT emphasizes on offering rich
pattern matching features (on tree, gimple...). On a (very?) long term
view, we could imagine a language centered on pattern matching request,
helping the user to extract informations (such has "return every basic
block that contains a call to foo()").