[cap-talk] Specs for our 'ideal' language?

[David Barbour]

On Sun, Sep 27, 2009 at 7:22 AM, Sandro Magi <naasking at higherlogics.com> wrote:
> IIRC, the quantified cost of virtual memory is on the same order as the
> quantified cost of garbage collection (~30% of total execution time), so
> I don't think there's a clear winner here. The measured costs of GC may
> also be skewed as virtual memory adversely affects GC far more than
> memory unsafe programs [1].
>
> Sandro
>
> [1] http://lambda-the-ultimate.org/node/2391
>

Indeed. But the two needn't adversely affect one another. Virtual
memory primitives make for good GC primitives, especially in the
presence of multi-core concurrency, as indicated both by the paper you
reference and the papers it references. But it seems most Operating
Systems today interfere with taking advantage.

[11] A. W. Appel and K. Li. Virtual memory primitives for user
programs. ACM SIGPLAN Notices, 26(4):96–107, 1991.