Re: Memory leak ?

FROM : Jean Bovet
DATE : Mon Mar 28 23:57:57 2005

Thanks a lot for the explanation. So why the "real memory" column of
the Activity Monitor still count these pages if they are marked as free
? Is there a way to know the number of "used" page ? All we really care
about is how much "used" page our app is using ;-)

Jean

Le 28 mars 05, à 13:14, James Housley a écrit :

> On Mar 28, 2005, at 3:51 PM, Jean Bovet wrote:
>

>> I have been looking at the memory allocation using the ObjectAlloc
>> tool and it seems that all allocated objects are correctly released.
>> As Matt Neuburg wrote, we cannot trust the Activity Monitor numbers
>> for memory leak so my question is the following: why is the "real
>> memory" column under the activity monitor growing up to 70Mb and then
>> stays around 50Mb (were it should be 15Mb) ? Is this memory
>> considered as "free" ? How can I be sure ?
>>

>
> You have to understand that the underlying kernel is Un*x/BSD based. 
> Those pages in memory are marked as free usable to the VM subsystem,
> but the pages are still in memory at this point.  What this means is
> that they are available for the next process that needs them.  The VM
> subsystem is not going to actively swap them out because that would be
> work with no benefit.  The next step is either another process is
> going to allocate them and use them, or they will be swapped out when
> some other process needs to swap some of its pages in or back in.  I
> asked almost the same question about one of my programs on FreeBSD, a
> close relative.
>
> Jim
>
> --
>
> /"\  ASCII Ribbon Campaign  .
> \ / - NO HTML/RTF in e-mail  .
>  X  - NO Word docs in e-mail .
> / \ -----------------------------------------------------------------
> <email_removed>      http://www.FreeBSD.org    The Power to Serve
> <email_removed>  http://www.TheHousleys.net
> ---------------------------------------------------------------------
> Do not meddle in the affairs of dragons, for you are crunchy and taste
>    good with ketchup.