[macstl-dev] [ANN] macstl 0.3 -- correction

[Glen Low]

Ilya:

On 29/06/2005, at 11:59 PM, Ilya Lipovsky wrote:

> Glen,
>
> Thank you for including my initials in the release announcement. I  
> feel somewhat bad, however, that my contribution wasn't in quite  
> the domain of added functionality. So, let me correct it a bit by  
> suggesting and, actually, providing a unary function "conj" (for  
> conjugation):
>
> in functional.h:
>
>
> // added by il at the body, inside namespace impl [begin]
>         template <typename T> struct conjugate: public
>             std::unary_function <T, T>
>         {
>           T operator() (const T& lhs) const
>           {
>             return conj (lhs);
>           }
>         };
> // added by il [end]
>
> in valarray_base.h:
>
>     DEFINE_VALARRAY_UNARY_FUNCTION (conj, conjugate) /* added by il */
>
>
> in vec.h:
>
>      DEFINE_VEC_COMMON_UNARY_FUNCTION(conj,stdext,conjugate) /*  
> added by il */
>
>
> in impl/vec_altivec.h:
>
>   // conjugate - added by il
>
>   template <> struct conjugate <macstl::vec <stdext::complex  
> <float>, 2> >
>   {
>     typedef macstl::vec <stdext::complex <float>, 2> argument_type;
>     typedef macstl::vec <stdext::complex <float>, 2> result_type;
>
>     result_type operator() (const argument_type& lhs) const
>     {
>       using namespace macstl;
>
>       const vec <stdext::complex <float>, 2> conj_filter = vec  
> <stdext::complex <float>, 2>::set<0x0000000080000000ULL,
>     0x0000000080000000ULL>();
>       return  data_cast<result_type>(altivec::vxor (lhs,  
> conj_filter));
>     }
>   };
>
> //----
>
> For NaNs:
> All signaling denormals get converted into signaling denormals.
> All quiet denormals get converted into quiet denormals.
>
> For reals:
> The imaginary parts become negatives of themselves. The real parts  
> stay the same.
>
> The above was tested and verified to produce correct results in  
> MacSTL v.2.2 - haven't tried your newest, .3, gem yet.
>
> I would like to add more functionality in the future.

Excellent!! This would rate as a * initialled item for 0.3.1 -- BTW,  
that's what presents a contributed item in the version history as  
Matt LeBlanc's was in 0.3 [*MLe]. Since it also shows you understand  
the macstl function extension mechanism, I'm particularly happy.

One minor correction, although I haven't put it through the compiler  
yet, is that altivec::vxor should only work on vec <float, 4>, so you  
should do a convert to float before that. (Hmm... somewhat unexpected  
behavior... )

Cheers, Glen Low


---
pixelglow software | simply brilliant stuff
www.pixelglow.com
aim: pixglen

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.pixelglow.com/lists/archive/macstl-dev/attachments/20050630/4558b692/attachment.html