Re: [eigen] Taking advantage of C++11 features in Eigen |
[ Thread Index |
Date Index
| More lists.tuxfamily.org/eigen Archives
]
On 12.02.2014 12:21, Nicola Gigante wrote:
Il giorno 11/feb/2014, alle ore 22:49, Benjamin Schindler <bschindler@xxxxxxxxxxx> ha scritto:
I'd like to add one particular point here: new POD support (Plain old data). It would be grat if a vector of Vector3f could be optimized using the POD optimizations
Not sure how much exactly this would improve things (but I'm not the
C++11 expert). I assume it mostly allows more efficient memmove/memcopy
operations?
In the constructor we currently make some checks of template parameters
and where necessary of alignment. I assume the template parameter checks
could/should be replaced by true static assertions, e.g. similar to the
way BOOST_STATIC_ASSERT is defined. This would also save some NOP calls
from the constructors. I just added a bz-entry for that:
http://eigen.tuxfamily.org/bz/show_bug.cgi?id=742
However, the alignment checks will still be necessary (not for Vector3f,
though).
Of course the first thing to do is to provide Eigen with a set of preprocessor macro to abstract compiler support.
That is definitely a thing which could be improved. Macros.h already has
a starting point for this:
https://bitbucket.org/eigen/eigen/src/90f8b868e6e2/Eigen/src/Core/util/Macros.h?at=default#cl-99
But there are some places left which make their own manual checks. E.g.
MathFunctions.h does:
#if (__cplusplus >= 201103L) && !defined(__CYGWIN__)
instead of something like #ifdef EIGEN_HAS_LOG1P.
Then the first one I’d like to see is certainly move semantics…
As said before, this is already implemented.
Christoph
--
----------------------------------------------
Dipl.-Inf., Dipl.-Math. Christoph Hertzberg
Cartesium 0.049
Universität Bremen
Enrique-Schmidt-Straße 5
28359 Bremen
Tel: +49 (421) 218-64252
----------------------------------------------