Re: [Caml-list] Ocaml crash in windows due to running out of memory

[Goswin von Brederlow <goswin-v-b@web.de>]

On Tue, Nov 26, 2013 at 07:53:46AM +0100, Florian Hars wrote:
> Am 25.11.2013 14:55, schrieb Adrien Nader:
> >As far as I know, it doesn't do overcommit. Also, I doubt that would
> >change anything: if you exhaust your address space, you cannot receive a
> >pointer to some place since it would collide with some of your already
> >assigned space.
> 
> That is a non-sequitur, in modern operating systems every process
> gets its own space of 2^64 adresses, no matter how much physical RAM

Or 2^48 on amd64 due to hardware limits. 2^47 under xen. Still plenty
more than physical ram.

> + swap is available. And the common way to get overcommitment (at
> least on unixoids, windows may be different) in't malloc anyway,
> it's fork, where both processes end up with the same address space
> layout, but (logically) corresponding to differen physical memory
> (modulo COW pages).
> 
> - Florian.

Note: malloc() of larger regions usualy uses mmap(), which inside the
kernel sets up a COW region. So basically the same as fork(). Both are
causes for overcommit.

MfG
	Goswin