I'm wondering when and how to "disable" the GC and do the memory mangement yourself?

"Hoenir" <mrmocool@gmx.de> wrote in message news:f84sio$22an$1@digitalmars.com...
> I'm wondering when and how to "disable" the GC and do the memory mangement yourself?

import std.gc;
std.gc.disable();

Jarrett Billingsley schrieb:
>> I'm wondering when and how to "disable" the GC and do the memory mangement yourself?
> 
> import std.gc;
> std.gc.disable(); 

I don't want to disable it completely. I'm just wondering when it is advantageous to do MM yourself and how to do it.
How to prevent a variable from being deleted by the GC and such stuff.
But thanks for your answer. :)

Hoenir wrote:
> Jarrett Billingsley schrieb:
>>> I'm wondering when and how to "disable" the GC and do the memory mangement yourself?
>>
>> import std.gc;
>> std.gc.disable(); 
> 
> I don't want to disable it completely. I'm just wondering when it is advantageous to do MM yourself and how to do it.

I haven't done this in D so I can't comment.

> How to prevent a variable from being deleted by the GC and such stuff.
> But thanks for your answer. :)

import std.gc;
//...
std.gc.disable();

//Do what ever non gc class allocation you want (you'll have to clean these up specifically yourself).

std.gc.enable();

You may want to do this for deterministic destruction, for optimization reason, or if you allocate from some sort of special memory.

-Joel

Hoenir wrote:
> Jarrett Billingsley schrieb:
>>> I'm wondering when and how to "disable" the GC and do the memory mangement yourself?
>>
>> import std.gc;
>> std.gc.disable(); 
> 
> I don't want to disable it completely. I'm just wondering when it is advantageous to do MM yourself and how to do it.
> How to prevent a variable from being deleted by the GC and such stuff.
> But thanks for your answer. :)

Hmmm, there are a few cases when the GC will not release memory because of pointer misidentification, and there are a couple of cases where the GC won't 'mark' a pointer.

Also, allocating in a tight loop is mostly faster with MM as it stands right now.

For example:

align(1)
struct PackedStruct
{
    short s;
    char *c;
}
//...
PackedStruct* s;
s = new PackedStruct;
s.c = new char[100]; // GC won't mark s.c unless you manually add it as a root

See more here: http://digitalmars.com/d/garbage.html

For non-OOP, basically you do manual MM the same as C w/ a slightly different syntax.

There's more here: http://digitalmars.com/d/memory.html

Short example to get you started:

;---

import std.c.stdlib;

void main()
{
    const ELEMS = 100;
    int* array = cast(int*)malloc(ELEMS * int.sizeof);
//  ...
    free(array);
    array = null;
}

HTH.