|[6 earlier articles]|
|Re: Static Garbage Collection email@example.com (Matthias-Christian Ott) (2009-05-26)|
|Re: Static Garbage Collection firstname.lastname@example.org (Vincent Belliard) (2009-05-26)|
|Re: Static Garbage Collection DrDiettrich1@aol.com (Hans-Peter Diettrich) (2009-05-29)|
|Re: Static Garbage Collection email@example.com (Armel) (2009-05-29)|
|Re: Static Garbage Collection firstname.lastname@example.org (Marco van de Voort) (2009-05-30)|
|Re: Static Garbage Collection email@example.com (Vincent Belliard) (2009-05-31)|
|Re: Static Garbage Collection firstname.lastname@example.org (Armel) (2009-06-01)|
|Date:||Mon, 1 Jun 2009 23:57:17 +0200|
|Organization:||les newsgroups par Orange|
|References:||09-05-120 09-05-130 09-05-135 09-06-001|
|Posted-Date:||02 Jun 2009 13:25:43 EDT|
> The use of qualified references avoid a lot of job even if it cost a
> little bit more in data (because you must count references separately).
> When an object has at least one reference from a local or global
> variable, in no way it can be destroyed. In that case, when you release
> one reference on the object you don't have extra job to do.
I personnally use the user/maintainance (==qualified) reference
counting principle in all my projects: this is indeed a really
efficient way to handle objects life cycle when the relation between
objects can be foreseen. the thesis referred "Concurrent Cycle
Collection in Reference Counted Systems", June 2001, from David
F. Bacon and V.T. Rajan is simply "another way": sometime you cannot
predict how users will build graphs of those smart pointers, and
having this strategy in those cases can be interesting (in addition to
the faster/more efficient "qualified" references).
Return to the
Search the comp.compilers archives again.