[Python-3000] Delayed reference counting idea

[Barry Warsaw]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On Sep 22, 2006, at 7:54 AM, Marcin 'Qrczak' Kowalczyk wrote:

> Greg Ewing <greg.ewing at canterbury.ac.nz> writes:
>
>> I know. What I mean to say, I think, is can it be designed so that
>> there cannot be any pauses longer than there would have been if
>> freeing had been performed as early as possible by refcounting.
>
> The question is misleading: refcounting also causes pauses, but at
> different times and with different length distribution. An incremental
> GC generally has pauses which are incomparable to pauses of  
> refcounting,
> i.e. it has longer pauses where refcounting had shorter pauses and
> vice versa.

Python's cyclic gc can also cause long pauses if you end up with a  
ton of objects in say generation 2, because it takes time just to  
traverse them even if they can't yet be collected.

- -Barry

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.5 (Darwin)

iQCVAwUBRRPe2nEjvBPtnXfVAQLspwQAtFhE1UJprBg8Cf/jH6obpaP+4+T2GHsP
ZW2IUcp41jZanfVcrOOEVERyR5saQtsCRoZtaN8XTxOJ1P1ZBCXZId0kGc39MQBW
9J4RDoQ4WTdXQFIaN+15OHIkKDIaLFakX0/smdwjHfAm8QI8D8EbEoetbsu5q0nq
MbfLHc7kk2U=
=Dz8F
-----END PGP SIGNATURE-----