1978 called, wants its garbage collectors back

Posted Sep 4, 2008 13:04 UTC (Thu) by i3839 (guest, #31386)
In reply to: 1978 called, wants its garbage collectors back by rwmj
Parent article: The Kernel Hacker's Bookshelf: UNIX Internals

I thought Python's GC also used reference counting. What better alternatives are there? Scanning memory for anything that looks like a pointer doesn't seem to be one. Perhaps built a reference graph of all allocated memory, and remove all disjunct pieces or something? But that seems to have more overhead than reference counting. I don't know much about GC...

1978 called, wants its garbage collectors back

Posted Sep 4, 2008 13:22 UTC (Thu) by knobunc (guest, #4678) [Link]

The "What the Heck Is?" series of posts by Dan Sugalski was good:
http://www.sidhe.org/~dan/blog/archives/cat_what_the_heck...

In particular he has one about GC:
http://www.sidhe.org/~dan/blog/archives/000200.html

He was posting good overviews of CS topics so that people could work on Parrot (the proposed Perl VM).

But a more thorough review of GC technology on LWN would be appreciated!

-ben

1978 called, wants its garbage collectors back

Posted Sep 4, 2008 14:06 UTC (Thu) by rwmj (subscriber, #5474) [Link] (1 responses)

I don't know much about GC...

So it seems. Have a look at Wikipedia then for the serious end of garbage collection take a look at Richard Jones [no relation]'s garbage collection resources page and in particular his book.

Rich.

1978 called, wants its garbage collectors back

Posted Sep 5, 2008 13:12 UTC (Fri) by i3839 (guest, #31386) [Link]

Umm, Wikipedia just mentions two basic types of GC, a reachability based one and a reference counting based one, and I mentioned both methods without knowing a thing about GC. Same for the "What the heck is: Garbage Collection" article mentioned in the comment above, and the book doesn't seem to mention anything fundamentally different either, just goes into details (same ideas, different algorithms, though interesting ones. Especially the copying one is smart, as it solves the fragmentation problem). Isn't there another paradigm of doing GC out there? Or is the little I know really all that's to know about GC on a high level?

I was totally wrong on the overhead/cost though. ;-)

1978 called, wants its garbage collectors back

Posted Oct 11, 2008 9:24 UTC (Sat) by anomalizer (✭ supporter ✭, #53112) [Link]

If you are interested in GC options in general, try reading up about the various GC developments that have been happening in the Java world. Things have moved a long way from reference counting algoithms. The caveat of course is that what works for Java is not going to work for the kernel.