Re: Storage management, was Compiler writers will love this language

> That is deceptive, and many of the books on garbage collection
> approach propaganda on the issue. I agree with Lex Spoon here.


Probably because tracing GC has generally been criticized for its slowness,
so there's a lot of understandable defensiveness.


The fact that both ref-count and tracing GC are still in widespread use
after more than 40 years of competition is a strong indication that there's
no clear winner.


> |> Short of resorting to some kind of hybrid with a tracing algorithm,
> |> I don't believe anyone has ever fixed the cyclic graph problem with
> |> reference-counting.
> As it is theoretically impossible to fix, that seems likely :-)


I don't know what counts as a hybrid, but I think the paper below
is not using any kind of "tracing GC" technology:


      http://www.research.ibm.com/people/d/dfb/papers/Bacon03Pure.pdf


Other possible approaches to dealing with cycles can be found in works
on distributed GC.




Stefan

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