Re: Increasing array size in c++

[Edwin Castro]

swagat wrote:
> Hi all!
>
> My debian system gives me segmentation fault on increasing the array
> size beyond certain value.

I'm sure *any* system would give you a segmentation fault in the 
scenario you describe below.

> I have a variable with 5 dimensions A[25][25][25][6][6]. IF I want to
> go for higher value of dimensions, say 30, it gives me segmentation
> fault.

If each dimension only had 2 elements (A[2][2]....[2]), the type of A 
was char (1 byte long) and you had 30 dimensions, then you would end up 
with a memory buffer 1*2^30 bytes long! That's an entire GB of memory!!

Say that each dimension has the same number of elements N (which may not 
be the case), the size of the data type in number of bytes is T, and you 
have D dimensions then you'll use up T*N^D bytes of memory. You'd need 
quite a large amount of RAM to support what you're trying to do!!!


Suppose sizeof(typeof(A)) = 15 and you wanted 25 values per dimension 
and you want 30 dimensions, then you'd need 15*25^30 bytes. That's 
13010426069826053208089433610439300537109375 bytes or 
12116903504194741331124150762743596 GB. I don't care how much virtual 
memory you have, you'll never have enough!


Perhaps you want more values in each of the 5 preexisting dimensions...?
In that case you'd have A[30][30][30][30][30] and if sizeof(typeof(A)) 
is 2 bytes, then you'll need 2*30^5 bytes or 46 MB. If sizeof(typeof(A)) 
is 15 bytes, then you'll need 15*30^5 bytes or 347 MB. You'll still need 
lots of memory.

> How to deal with higher dimensional arrays? How can i increase the
> memory allocated to data types? 

As opposed to memory allocated to something else? I'm not sure what you 
mean here. I still hope my message will help you figure out how much 
memory you'll need. If you really need something that large try to 
parallelize the computation so that it can be computed by multiple 
computers in parallel or sequential chunks by one computer.

> Please help.
>
>
> Regards,
> swagat