Re: size of variable

From: Gus Correa <gus_at_nyahnyahspammersnyahnyah>
Date: Wed Jun 08 2011 - 20:35:05 MDT

Hi Debasish

Everybody faces the same problem, computer memory is always
much smaller than disk space.

Read what Dennis Shea said about NCL 6.0.0.
You may want do install it, if you have an older version.
The older versions aren't able to
store in the computer memory variables larger than 2GB,
even if your computer is 64-bit.
With 6.0.0 the sky is the limit, you can have variables
as large as your available memory, if your machine is 64-bit.
(But sill have the 2GB limitation if the computer is 32-bit.)

I hope this helps,
Gus Correa

Debasish wrote:
> Thanks Gus for your response.
> I am agree with you about dividing the large data into small parts
> which was my last option. I guess now this is my only option
>
> Debasish
>
> On 2011-06-08, at 16:09, Gus Correa wrote:
>
>> Debasish wrote:
>>> Dear NCL Users,
>>>
>>> I am trying to declare a variable: x=new((/34,12784,130,193/), float, -999);
>>> I got following error: fatal:New: requested size of variable
>>> (43622076160 bytes) exceeds the current maximum allowed on this system.
>>> I realized that the error comes from large dimension of the variable.
>>> Is there anyway in NCL to handle this type of large dimensional variable.
>>>
>>> Thanks
>>>
>>> Debasish
>>> _______________________________________________
>>> ncl-talk mailing list
>>> List instructions, subscriber options, unsubscribe:
>>> http://mailman.ucar.edu/mailman/listinfo/ncl-talk
>> Hi Debasish
>>
>> The variable seems to require (at least) 43.6GB of memory.
>> Does your computer have this much? (and some more for the OS, NCL, etc)
>> Also, if your computer is 32-bit, you probably will have trouble with
>> variables larger than 2GB, maybe 4GB, anyway.
>>
>> Sometimes it is hard to store in memory all data that
>> we have stored in large disk files.
>>
>> Can't you work with parts of the variable at a time?
>> Say, time slices, or a chunk of a few levels, do whatever data reduction
>> you need to do, then accumulate the reduced results in another (smaller)
>> variable?
>>
>> My two cents,
>> Gus Correa
>> _______________________________________________
>> ncl-talk mailing list
>> List instructions, subscriber options, unsubscribe:
>> http://mailman.ucar.edu/mailman/listinfo/ncl-talk
>

_______________________________________________
ncl-talk mailing list
List instructions, subscriber options, unsubscribe:
http://mailman.ucar.edu/mailman/listinfo/ncl-talk
Received on Wed Jun 8 20:35:20 2011

This archive was generated by hypermail 2.1.8 : Thu Jun 09 2011 - 14:56:48 MDT