Re: Compiler writers will love this language

nmm1@cus.cam.ac.uk (Nick Maclaren)
15 Jul 2003 23:33:35 -0400

          From comp.compilers

Related articles
[21 earlier articles]
Re: Compiler writers will love this language ericmuttta@email.com (2003-07-02)
Re: Compiler writers will love this language mwotton@cse.unsw.edu.au (2003-07-03)
Re: Compiler writers will love this language nmm1@cus.cam.ac.uk (2003-07-04)
Re: Compiler writers will love this language nmm1@cus.cam.ac.uk (2003-07-04)
Re: Compiler writers will love this language joachim.durchholz@web.de (Joachim Durchholz) (2003-07-13)
Re: Compiler writers will love this language ericmuttta@email.com (2003-07-15)
Re: Compiler writers will love this language nmm1@cus.cam.ac.uk (2003-07-15)
Re: Storage management, was Compiler writers will love this language dmr@bell-labs.com (Dennis Ritchie) (2003-07-17)
| List of all articles for this month |
From: nmm1@cus.cam.ac.uk (Nick Maclaren)
Newsgroups: comp.compilers
Date: 15 Jul 2003 23:33:35 -0400
Organization: University of Cambridge, England
References: 03-05-211 03-06-106 03-07-012 03-07-089
Keywords: storage
Posted-Date: 15 Jul 2003 23:33:34 EDT

Eric <ericmuttta@email.com> wrote:
>> [Pure reference-count garbage collectors have problems with circular
>> structures that point to themselves.]
>
>I notice here, that you refer to "pure" reference-count GC, implying
>that some GC's are "hybrid" and combine reference-counting and some
>other scheme (supposeddly mark-and-sweep). How would such a hybrid GC
>work? any papers on the subject?


Gug. There certainly were some, but I last looked decades ago. One
approach was to use reference counters for objects that contain no
potential for backpointers, and fancier methods for those that do.
For example, arrays of plain data - even sparse arrays - are safe, and
many applications have ten of those for every structure that isn't
safe.


>> Note that a doubly-linked list falls into this category...


I can't remember whether anyone has said so, but it doesn't if the
list type is known to the language. It does only if the type is
built up from plain pointers.


>[I think I've seen descriptions of GC that use reference counts in places
>where they're known to be safe, other kinds elsewhere. -John]


Yes, I remember that, too.




Regards,
Nick Maclaren.


Post a followup to this message

Return to the comp.compilers page.
Search the comp.compilers archives again.