Re: calloc ( ) failed ... but enough RAM ...
From: xavier (cose_at_cose.fr)
Date: 02/17/05
- Next message: Marc Schodermayr: "learning MFC"
- Previous message: Scott McPhillips [MVP]: "Re: Add Items to MSFlexGrid"
- In reply to: Joseph M. Newcomer: "Re: calloc ( ) failed ... but enough RAM ..."
- Next in thread: Murrgon: "Re: calloc ( ) failed ... but enough RAM ..."
- Messages sorted by: [ date ] [ thread ]
Date: Thu, 17 Feb 2005 14:07:24 +0100 To: "Joseph M. Newcomer" <newcomer@flounder.com>
it is true. memory fragmentation is certainly the problem.
I know how many RAM available has my system by simply looking into the
Windows 2000 Task Manager.
There probably is an error in your compute: Try to call calloc( ) for
400 Moctets , it about instantaneously works fine ... are you sure
talking about "ms" ( milliseconds) for a page fault ?
If you are right, it would mean that allocate 10 Mo ( 800 Mo / 80 ... )
would take at best 7mn/80 = 5 seconds ... ?
The solution is - as said Stephen Kellett - to cut the computation and
writing of my datas ...
Thanks and Regards ,
Xavier
Joseph M. Newcomer a écrit:
> And what,exactly, makes you think you "have enough RAM free"? You just asked for
> 800*1024*1024, or 800 Megabytes of memory. It is unlikely in the extreme that you would
> have a slot in your address space large enough to hold that. You are talking about
> keeping 200 or 300MB free, but you have no basis for making that statement.
>
> You are ignoring issues like address space fragmentation. This is significant. You can
> safely assume that allocating a contiguous chunk of memory this size is unlikely.
>
> Is your swap space large enough to handle an allocation this size?
>
> BY the way, why "calloc"? This has to zero the memory, which means it has to touch each of
> the (1024x1024x800)/4096 pages, or 204,800 pages you are allocating. That will generate
> 204,800 page faults.Assuming 20ms per page fault, this will take 4,096 seconds to execute,
> or a little over an hour (68 minutes). At 10ms per page fault, it will take a half hour.
> Being really optimistic and saying 2ms per page fault, it will take about seven minutes to
> zero this memory out. There's something deeply wrong with what you are trying to do.
>
> There is no way to answer your question about how much space you can allocate; you can't
> answer this until you allocate it, because previous allocation patterns have also
> fragmented memory. So you have both heap fragmentation and address space fragmentation
> coming into play here.
>
> What are you really trying to do?
> joe
>
> On Wed, 16 Feb 2005 12:29:42 +0100, xavier <cose@cose.fr> wrote:
>
>
>>Hello to all
>>
>>I need to allocate a big buffer but calloc failed , even if I do have
>>enough RAM free.
>>I have 1200 Mo free in RAM when calloc( ) is called , but it works only
>> if I calloc 900 Mo in my application.
>>
>>
>>
>>see the code below: it try to allocate the max RAM memory
>>
>>
>>"
>>unsigned short *tabtest = NULL ;
>>int memtaille = 800*1024 ;
>>int mempas = 1024 ;
>>while( (tabtest == NULL) && (memtaille>mempas) )
>>{
>> memtaille -= mempas ;
>> tabtest = (unsigned short *)calloc( memtaille*1024 , sizeof(unsigned
>>short) );
>>}
>>if(tabtest != NULL)
>>{
>> free(tabtest);
>> tabtest = NULL ;
>>}
>>"
>>
>>in a new MFC project , this code can allocate all the free (available)
>>memory in RAM.
>>
>>in my application , depending on where in the code I put this code, it
>>will always keep 200 or 300 Mo free in RAM.
>>
>>why ? I don t think I have memory leaks ( no bugs ... no debug messages
>> when application exits ... )
>>how to allocate all the RAM available ?
>>how to know exactly how much RAM I can allocate with calloc() before
>>calling it ?
>>
>>
>>Thanks.
>>Xavier
>>
>>
>
>
> Joseph M. Newcomer [MVP]
> email: newcomer@flounder.com
> Web: http://www.flounder.com
> MVP Tips: http://www.flounder.com/mvp_tips.htm
- Next message: Marc Schodermayr: "learning MFC"
- Previous message: Scott McPhillips [MVP]: "Re: Add Items to MSFlexGrid"
- In reply to: Joseph M. Newcomer: "Re: calloc ( ) failed ... but enough RAM ..."
- Next in thread: Murrgon: "Re: calloc ( ) failed ... but enough RAM ..."
- Messages sorted by: [ date ] [ thread ]
Relevant Pages
|
