On Sat, Aug 3, 2013 at 10:58 PM, Bennie Kloosteman <[email protected]>wrote:

> http://dl.acm.org/citation.cfm?doid=1375581.1375586
>
> Immix GC using regions showed some positive results..
>

By my read, there is no concept overlap between an inmix-region, and LIFO
or subtyped regions.

As far as I can tell, Inmix regions (2008) are general use memory-pages
used to make a simple fragmentation-vs-copying tradeoff determined by the
liveliness of the contents of a memory-page (which they call a region).
Other collectors before and since use similar techniques, including C4
(2011 - 4.1 "...C4 chooses to compact pages below a certain liveness
threshold.")

The paper does have a good look at tradeoffs due to the intertwined nature
of collection and allocation schemes.
_______________________________________________
bitc-dev mailing list
[email protected]
http://www.coyotos.org/mailman/listinfo/bitc-dev

Reply via email to