Garbage Collection in C

[Chris Thomasson]

# FWIW, here is an experimental protoyope of an atomically thread-safe
# reference counted pointer:

How well does it collect cyclic graphs?

Currently, you have to be careful about cycles... An object A that
explicitly references object B, which in turn explicitly references object A
== leak... So, you have to resort to using tricky reference counting
techniques, which are compatible with my current algorithm, to get around
the cycle problem.

However, there is a way to add weak pointers to my algorithm which would
render it immune to the cycle problem... Not sure if I should do it or
not...

Humm...

 

[Paul Connolly]

Pure functional languages don't necessarily have that property - such
a language could perfectly well have a syntax for circular structures.

And also pure functional languages might have optimisations such as in cases
of tail recursion where a containing structure might be constructed before
the elements of that structure - so the garbage collection becomes more
complicated.

 

[af]

Hello,

Paul Connolly wrote:
[...]

Actually I wrote a _little_ bit of assembler this morning, which turned into
this afternoon, as it often does with assembler

BTW, is anyone aware of any assembler/hardware optimizations that can
speed up garbage collection?

And does anyone from the respectful community use high performance
garbage collectors from Harmony project (http://harmony.apache.org)?

With best regards,
Alexei Fedotov