Marc:
What does this mean?
Add support for a SSB which will be used by the generational GC. https://github.com/feeley/gambit/commit/ad379113ae0ed297043dd2d2a43109ba772396b0
Brad
Afficher les réponses par date
On 2013-01-04, at 2:39 PM, Bradley Lucier lucier@math.purdue.edu wrote:
Marc:
What does this mean?
Add support for a SSB which will be used by the generational GC.
Brad
SSB = Sequential Store Buffer. The SSB keeps track of the assignments to heap objects. Assignments might create pointers from an old to a new generation. These "inverse" pointers need to be taken into account when garbage collecting a new generation (without garbage collecting the old generation).
Marc
On Jan 4, 2013, at 2:43 PM, Marc Feeley wrote:
On 2013-01-04, at 2:39 PM, Bradley Lucier lucier@math.purdue.edu wrote:
Marc:
What does this mean?
Add support for a SSB which will be used by the generational GC.
Brad
SSB = Sequential Store Buffer. The SSB keeps track of the assignments to heap objects. Assignments might create pointers from an old to a new generation. These "inverse" pointers need to be taken into account when garbage collecting a new generation (without garbage collecting the old generation).
Marc
This paper
https://bitbucket.org/GregorR/ggggc/src/0769d2c4cb90/doc/write-barrier.pdf
does some slagging on Sequential Store Buffers as part of write barriers. Any comments?
Brad
On 2013-01-04, at 11:48 PM, Bradley Lucier lucier@math.purdue.edu wrote:
This paper
https://bitbucket.org/GregorR/ggggc/src/0769d2c4cb90/doc/write-barrier.pdf
does some slagging on Sequential Store Buffers as part of write barriers. Any comments?
Card-marking is an alternative that I will probably also implement to compare with explicitly recording inverse pointers. I'm not a big fan of card-marking because it imposes some constraints on the object representation (object alignment on card boundaries, or BIBOP) and the scanning can be more expensive due to the conservatism of the cards.
Marc
-
Bradley Lucier -
Marc Feeley