[coyotos-dev] Explicit Persistence Considered Harmful

Jonathan S. Shapiro shap at eros-os.com
Fri Aug 22 10:48:46 CDT 2008


On Fri, 2008-08-22 at 09:06 -0400, Sandro Magi wrote:
> Interesting! All non-concurrent GC algorithms are a hybrid mix of 
> reference counting and tracing [1]...

Umm. I can name 5 non-current collectors off the top of my head that do
not use reference counting at all. Three of those are in widespread use
today...

What the paper alleges is that **high performance** collectors are
hybrids. I'm not even convinced that that is true, but it's more
plausible.

> , so this system definitely counts as 
> GC. The first cycle collector for reference counting was described in 
> the literature in 2001 [2] to the best of my knowledge, but this use 
> might predate it by a good number of years.
> 
> Sandro
> 
> [1] http://researchweb.watson.ibm.com/people/d/dfb/papers/Bacon04Unified.pdf
> [2] http://researchweb.watson.ibm.com/people/d/dfb/papers/Bacon04Unified.pdf
> _______________________________________________
> coyotos-dev mailing list
> coyotos-dev at smtp.coyotos.org
> http://www.coyotos.org/mailman/listinfo/coyotos-dev



More information about the coyotos-dev mailing list