Re: [boost] Re: questions on ublas ?

From: Toon Knapen (toon.knapen_at_[hidden])
Date: 2001-12-07 03:29:09

jhrwalter wrote:

> --- In boost_at_y..., Peter Schmitteckert (boost) <boost_at_s...> wrote:
> 
>>Salut,
>>
>>On Thursday 06 December 2001 10:45, Toon Knapen wrote:
>>
>>
>>>>>For instance if I
>>>>>need a temporary matrix in a loop, but in every loop the total 
>>>>>
> size
> 
>>>>>of the matrix can be different, I can allocate a matrix with 
>>>>>
> some
> 
>>>>>maximal size and resize it in every loop to something smaller 
>>>>>
> in every
> 
>>>>>loop in constant amortised time ?
>>>>>
>>>>Currently not, but it should be possible to change this. May be 
>>>>
> we
> 
>>>>need to differentiate size() and capacity()?
>>>>
>>>Definitly. This would minimise memory allocation !
>>>
>>I also need this feature to avoid unneccessary allocation,
>>but here should be the possiblilty to finally release unneeded 
>>
> memory.
> 
>>What about the following idea:
>>-> resize( n1, n2)  garanties "to hold size1 rows of size2 
>>
> elements".
> 
>>-> reserve(n3, n4) garanties that as long n3>=n1, n4 >=n2 no 
>>
> allocations in 
> 
>>resize(n1,n2) are performed. if n3<n1 or n4 < n2 the matrix is 
>>
> allowed to
> 
>>shrink to the reserved size if requested in a subsequent resize(..) 
>>
> in
> 
>>order to minimize wasted memory.
>>But this would lead to the notion of a minimal_capacity().
>>
> 
> Do we really need reserve()? IMO the allocation behaviour of resize() 
> is a quality of implementation issue, especially since resize() is a 
> method of the proxy concept. If we intend to make it's behaviour 
> transparent, we'd have to introduce capacity() first.

I find this really necessary (there's also a good reason why they have 
capacity and size in std::vector). Everytime you shrink the vector or 
matrix (i.e. resize(x) with x < size() ), you are going to delete[] the 
old data and new[] another array. This is very time consuming.
I would suggest to allign totally with what std::vector does : you can 
resize without reallocation (if the capacity is large enough) and you 
can easily shrink the capacity by copying to a smaller vector.

> OTOH capacity() won't be generic.

please elaborate.

> BTW, do we need to consider std::allocator?
In the long term : yes. But no allocater can be as smart as what you 

can do with capacity() and resize(). So the allocator is not a solution

to the above problem.
An allocator would be nice e.g. to optimise allocation of many small 
matrices.