[Sidefx-houdini-list] render-memory problem

Mark A. Bolstad mbolstad at arl.army.mil
Fri Aug 4 16:28:03 EDT 2006


Sort of ;-)

<Begin dredging memory for old OS stuff; Proceed with caution>

The maximum addressable space in 32 bits is 4 GB. So, the most a single
application can allocate is 4 GB. However, the OS has a few additional
tricks up it's sleeve, such that it can address a larger space through
the virtual memory management (VMM) system. So in theory, you could have
two 3GB processes running simultaneously, one in RAM, and one swapped
out, or both partially swapped out. And the reason this works is that
most VMM systems use a paging system through the file system (which uses
64-bit pointers), so you only need to keep track of the pages swapped in
and out, and not the physical addresses (although you do need to map
pages to memory addresses and to the file system, but that's another
lecture).

</Begin dredging memory for old OS stuff; Proceed with caution>

Mark

P. Alex Lim wrote:
> Hi Mark,
>
> Just out of curiosity, if you have 4GB of RAM and 2GB of swap memory,
> and once you hit that 4GB RAM limit, does the system still goes into
> swap or is the 4GB "it"?
>
> I wasn't too sure if this limit applies to RAM usage only or if it's the
> 32bits OS in general.
>
>
> Thanks,
> Alex
>
>
>
>
> Mark A. Bolstad wrote:
>   
>> Jed Schwartz wrote:
>>     
>>> On Aug 4, 2006, at 11:46 AM, Mark A. Bolstad wrote:
>>>
>>>       
>>>> Are you running 32 or 64-bits?
>>>>         
>>> 32- Originally (jan 06) I set this box up as a 64, but had way too
>>> many linux sys problems, so we went back to 32-
>>>       
>> Interesting, all of my systems are RHEL4 on x86_64 with absolutely on
>> problems (I have about 30 of them).
>>
>>     
>>>> If 32-bits, see if running 64-bit Houdini on a 64-bit box works.
>>>>
>>>>         
>>> Hmm- I thought there was a 2 gig limit with linux 32- (RH E4), a 3
>>> gig limit with 64- ??-
>>>       
>> Actually, the largest integer you can represent in 32-bits is
>> 2147483648 (2^31), however, if you use unsigned integers than you get
>> full 32 bits or 4294967296 (2^32). Hence, 4 GB is the largest
>> addressable memory with 32-bit systems. With 64-bits the maximum
>> addressable space is 16 exobytes (kilo-, mega-, giga-, tera-, peta-,
>> exo-)
>>
>>     
>>>> 4 GB is the limit for a 32-bit system, and given the size of your
>>>> core file, that's my suspicion. A lot of memory allocators grow the
>>>> amount of memory allocated each call to try to reduce overhead. You
>>>> may have hit a magic threshold where one more allocation send you
>>>> over the 4GB barrier.
>>>>
>>>> Depending on the platform, an out-of-memory error may be in a system
>>>> log.
>>>>         
>>> I'll check the log fie-
>>>
>>> Thanks,
>>>
>>> Jed-
>>>       
>>>> Mark
>>>>
>>>> Jed Schwartz wrote:
>>>>         
>>>>> I am getting a 3 gig core file-
>>>>> Specifically which cache setting?-
>>>>> 10 years at this and still so much to learn!
>>>>>
>>>>> Jed-
>>>>>
>>>>> On Aug 4, 2006, at 10:26 AM, Peter Robbinson wrote:
>>>>>
>>>>>           
>>>>>> Morning,
>>>>>>
>>>>>> Have you possibly missed a cache setting and it is dumping to a
>>>>>> harddrive?
>>>>>>
>>>>>> PeterR
>>>>>>
>>>>>> On Fri, 2006-04-08 at 10:09 -0400, Jed Schwartz wrote:
>>>>>>             
>>>>>>> I'm getting a total freeze up in the middle of a render on frame 741
>>>>>>> on one of several systems in the studio-
>>>>>>> All other systems render the frame without problem- All running the
>>>>>>> latest H-build- 8.1.681
>>>>>>>
>>>>>>> Am I correct to assume that we have some flakey memory on this
>>>>>>> newest
>>>>>>> and fastes system-?
>>>>>>>
>>>>>>> Thanks!!
>>>>>>>
>>>>>>> Jed-
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>> _______________________________________________
>>>>>>> Sidefx-houdini-list mailing list
>>>>>>> Sidefx-houdini-list at sidefx.com
>>>>>>> https://lists.sidefx.com:443/mailman/listinfo/sidefx-houdini-list
>>>>>>>
>>>>>>>               
>>>>>> --"gravity is not a force, it is a boundary layer"
>>>>>> "everything is coincident"
>>>>>> "Love: the state of suspended anticipation"
>>>>>>
>>>>>> _______________________________________________
>>>>>> Sidefx-houdini-list mailing list
>>>>>> Sidefx-houdini-list at sidefx.com
>>>>>> https://lists.sidefx.com:443/mailman/listinfo/sidefx-houdini-list
>>>>>>
>>>>>>             
>>>>> _______________________________________________
>>>>> Sidefx-houdini-list mailing list
>>>>> Sidefx-houdini-list at sidefx.com
>>>>> https://lists.sidefx.com:443/mailman/listinfo/sidefx-houdini-list
>>>>>           
>>>> _______________________________________________
>>>> Sidefx-houdini-list mailing list
>>>> Sidefx-houdini-list at sidefx.com
>>>> https://lists.sidefx.com:443/mailman/listinfo/sidefx-houdini-list
>>>>
>>>>         
>>> _______________________________________________
>>> Sidefx-houdini-list mailing list
>>> Sidefx-houdini-list at sidefx.com
>>> https://lists.sidefx.com:443/mailman/listinfo/sidefx-houdini-list
>>>       
>> _______________________________________________
>> Sidefx-houdini-list mailing list
>> Sidefx-houdini-list at sidefx.com
>> https://lists.sidefx.com:443/mailman/listinfo/sidefx-houdini-list
>>     
>
> _______________________________________________
> Sidefx-houdini-list mailing list
> Sidefx-houdini-list at sidefx.com
> https://lists.sidefx.com:443/mailman/listinfo/sidefx-houdini-list
>   




More information about the Sidefx-houdini-list mailing list