|
|
Ublas : |
Subject: Re: [ublas] Move Semantics
From: Jesse Perla (jesseperla_at_[hidden])
Date: 2009-09-10 17:09:38
On Thu, Sep 10, 2009 at 10:04 AM, Nasos Iliopoulos <nasos_i_at_[hidden]>wrote:
> Wouldn't changing vectors' operator = allow that?
> From:
>
> *BOOST_UBLAS_INLINE*
>
> *vector &operator = (const vector &v) {*
>
> *data () = v.data ();*
>
> *return *this;*
>
> *}*
>
> to
>
> // pass by value for move semantics
>
> BOOST_UBLAS_INLINE
>
> vector &operator = (vector v) {
>
> data().swap(v.data());
> return *this;
> }
>
I believe the above is called shallow copy and changes the current semantics
of ublas (see
http://www.osl.iu.edu/research/mtl/mtl4/doc/shallow_copy_problems.html) .
For example:
matrix<double> A(2,2);
//..... set A....
matrix<double> B = A;
Now they point at the same thing, not a copy/move.
Also, what about rvalues. e.g.:
matrix<double> A = 2 * B;
And last, since this is using expression templates, the "data" vector may
not exist, and will frequently not have the same type.
I don't know that much about this stuff, but my understanding is that rvalue
refereces and move semantics allow a library to overload the rvalue
assignment to do a this kind of pointer swapping. But this stuff is
probably very difficult to do.
-Jesse