Re: FastMM and garbage collection...
- From: "Thorsten Engler [NexusDB]" <thorsten.englerNO@xxxxxxxxxxxxxxx>
- Date: Tue, 12 Sep 2006 17:55:59 +1000
Atmapuri wrote:
And on the subject: Do you think it is feasible to try garbage colectingExcept that a) you have no control over all the threads in the
TObject's by extending fastMM?
It seems to me that the memory manager has all the information required
to implement something like that...
application and can't suspend them in a "safe" position while doing a
collection cycle and b) you don't have any reliable way to identify
pointers referencing the objects. Such references could be pretty much
everywhere on the heap, in the stack of any thread, in cpu registers,
and even in the "saved" thread context of threads that are currently not
running (which is stored somewhere in the kernel address space IIRC).
So how would your garbage collector determine which objects are no
longer referenced? A garbage collector usually requires that you are in
total control of all threads and have complete type information about
any place where references to garbage collected allocations might be
stored.
Cheers,
Thorsten
.
- Follow-Ups:
- Re: FastMM and garbage collection...
- From: Atmapuri
- Re: FastMM and garbage collection...
- References:
- FastMM and garbage collection...
- From: Atmapuri
- Re: FastMM and garbage collection...
- From: Eric Grange
- Re: FastMM and garbage collection...
- From: Atmapuri
- Re: FastMM and garbage collection...
- From: Pierre le Riche
- Re: FastMM and garbage collection...
- From: Atmapuri
- FastMM and garbage collection...
- Prev by Date: Re: FastMM and garbage collection...
- Next by Date: Re: FastMM and garbage collection...
- Previous by thread: Re: FastMM and garbage collection...
- Next by thread: Re: FastMM and garbage collection...
- Index(es):
Relevant Pages
|