|
|
Ublas : |
From: Jon Pearce (j.pearce_at_[hidden])
Date: 2005-03-30 06:58:09
Hi Michael,
Thank you very much for your fast, detailed reply. no_alias() solves the
problem beautifully
>I think the comment line is only there to explain what we logically is about
>to happen. I don't think it was ever part of the compiled code. It really is
>a comment and not a commented out line! The form using assign_temporary is
>used for efficiency.
>
>
>
(Actually, the commented line does compile ;-) and does eliminate the
temporary, but rather dirty to be modifying the original source!)
>Personally I would remove this completely! That way you would get a compile
>time error when you try and construct such a temporary. Also we could get rid
>of the own_ flag which makes shallow_array_adaptor inefficient.
>
>
>
Absolutely!
>The correct solution would be to use the following for assignment instead
> c.assign ( xx );
>or the equivilent
> noalias( c ) = xx;
>
>Hopefully that works. Have you tried the same with 'array_adaptor'?
>
>
I have now switched to array_adaptor - thank you for pointing this out
I notice that *any* call to internal_resize() in array_adaptor,
unbounded_array and shallow_array_adaptor triggers a reallocation, even
if the new size is smaller than the old
e.g.
BOOST_UBLAS_INLINE
void resize_internal (size_type size, value_type init, bool
preserve = true) {
if (size != size_) {
pointer data = new value_type [size];
...
}
}
Might there be a benefit in replacing this with "if(size < size_)" in a
future version of UBlas? (I think that this is what is done in
std::vector<T>::resize())
Thanks again for your time
Yours gratefully
jon