Re: [eigen] fix double-promotion in 3.3 Core/SpecialFunctions.h

[Christoph Hertzberg <chtz@xxxxxxxxxxxxxxxxxxxxxxxx>]

On 2016-05-05 15:39, Marc Glisse wrote:
> On Thu, 5 May 2016, Christoph Hertzberg wrote:
> > I (wrongfully) expected pow(x, 2) to be evaluated to x*x, at least if
> > the exponent is known at compile time and small. That could work, if
> > pow(float, int) got evaluated to some __builtin_pow function and the
> > compiler decides depending on the size of the exponent how to evaluate
> > it.
>
> That is the case. Gcc has __builtin_pow (takes 2 double) and expands it
> when the second argument is a small integer (though in this case it
> produces (float)((double)x*(double)x)). More relevant builtins are
> __builtin_powf (takes 2 floats) and __builtin_powif (takes a float and
> an int).

Ok, my memory is not as bad as I thought. I recalled that something like 
that existed once. And as Ilja pointed out, at least gcc used to do that 
prior to C++11 (and apparently I missed the C++03 part when I looked it up).

> However, writing pow(x,2) should really be avoided...

Yes, it is kind of annoying though to implement trivial square, cube, 
etc, functions for the few cases you need them (or to add a 
boost-dependency, unless you already do).


Christoph



--
 Dipl. Inf., Dipl. Math. Christoph Hertzberg

 Universität Bremen
 FB 3 - Mathematik und Informatik
 AG Robotik
 Robert-Hooke-Straße 1
 28359 Bremen, Germany

 Zentrale: +49 421 178 45-6611

 Besuchsadresse der Nebengeschäftsstelle:
 Robert-Hooke-Straße 5
 28359 Bremen, Germany

 Tel.:    +49 421 178 45-4021
 Empfang: +49 421 178 45-6600
 Fax:     +49 421 178 45-4150
 E-Mail:  chtz@xxxxxxxxxxxxxxxxxxxxxxxx

 Weitere Informationen: http://www.informatik.uni-bremen.de/robotik