[vsipl++] [patch] Profiling cleanup

[Jules Bergmann]

Don McCoy wrote:
> This patch completes the move of the operation counts for matvec and 
> signal processing functions into impl/ops_info.hpp.  It also cleans up 
> the tags for the FFT's to correspond to recent tutorial changes.

Don,

This looks good.  I have one question below about the names for Fft and 
Fftm, but it otherwise it looks good to check in.

				-- Jules

> Index: src/vsip/impl/ops_info.hpp
> ===================================================================
> --- src/vsip/impl/ops_info.hpp	(revision 147402)
> +++ src/vsip/impl/ops_info.hpp	(working copy)
> @@ -3,7 +3,8 @@
>  /** @file    vsip/impl/ops_info.cpp
>      @author  Jules Bergmann

Don you've taken this file a long way!  Can you add your name to the 
authors?


> +
> +template <int D, typename I, typename O>
> +struct Description
> +{ 
> +  static std::string tag(Domain<D> const &dom, int dir, 
> +    return_mechanism_type rm)
> +  {
> +    std::ostringstream   st;
> +    st << (D == 2 ? "Fftm " : "Fft ")
> +       << (dir == -1 ? "Inv " : "Fwd ")
> +       << Desc_datatype<I>::value() << "-"
> +       << Desc_datatype<O>::value() << " "
> +       << (rm == vsip::by_reference ? "by_ref " : "by_val ")
> +       << dom[0].size();
> +    if (D == 2)
> +       st << "x" << dom[1].size();
> +
> +    return st.str();
> +  } 
> +};

I'm not sure if this logic is right, in particular using the dimension 
of the Domain to distinguish between Fft and Fftm.  It is possible to 
have 1D, 2D, and 3D Fft transforms.  Fftm represents multiple 1D Fft 
transforms on either the rows or columns of a matrix.  A 2D Fft is not 
the same as a Fftm.

-- 
Jules Bergmann
CodeSourcery
jules at codesourcery.com
(650) 331-3385 x705