On Sat, 2005-04-16 at 08:59 -0400, Sandro Magi wrote:
> Eliminating runtime checks before memory operations and the need for
> GC is a potentially significant boost in performance, even if only for
> a subset of programs ("embedded C programs" in the paper).

I agree, but this is a relatively small fraction of what BitC is trying
to address.

The basis of their work is automated region inference. Region inference
isn't going to work for the applications we are really concerned about,
because the inference engine will conclude (correctly) that the region
remains live nearly everywhere that we might call free().

So: this is definitely a useful auxiliary technique for BitC, but it
doesn't address the core issues of memory (de)allocation that we need to
solve.

shap

_______________________________________________
bitc-dev mailing list
[email protected]
http://www.coyotos.org/mailman/listinfo/bitc-dev

Reply via email to