Re: [CHANGESET]: First attempt at a single precision type.

[Jaroslav Hajek]

On Mon, Apr 28, 2008 at 7:28 PM, John W. Eaton <address@hidden> wrote:
> On 28-Apr-2008, Jaroslav Hajek wrote:
>
>  | still, the code duplication seems to be massive. The different Fortran
>  | function names can be handled by overloaded wrappers. OTOH, it is
>  | certainly desirable to have as much code compiled as possible, and
>  | users aren't supposed to instantiate Matrix for anything other than
>  | float and double.
>
>  I'd really like to avoid the duplication as much as possible, even if
>  it's likely that there will only be two instantiated types.
>
>  Although it seems that we would need to have explicit specializations
>  for functions that call BLAS and LAPACK routines, I think we can get
>  the compiler to do a lot of that work for us if we use functors and
>  some wrappers, though I do agree that using functors to wrap functions
>  that take many arguments (like the BLAS and LAPACK functions do) will
>  add some complication to the code.  Template tricks can avoid the need
>  to define specific functors for each type signature we will encounter,
>  but it means building up the machinery to do it (I think the text C++
>  Templates: The Complete Guide by Vandevoorde and Josuttis outlines all
>  the techniques we would need.
>
>  Whether we use functors or not, it might be good to wrap the BLAS and
>  LAPACK routines anyway, so we can avoid the details of how character
>  arguments are passed.
>
>
>  | Is there any possibility to put a template for a method in a .cc file,
>  | explicitly instantiate there, and just export the explicit instances?
>  |
>  | I'm thinking about something like this:
>  | file RealMatrix.h:
>  |
>  | template <typename T>
>  | class RealMatrix
>  | {
>  |   bool is_symmetric(void) const;
>  | };
>  |
>  | template class RealMatrix<float>;
>  | typedef RealMatrix<float> FloatMatrix;
>  |
>  | template class RealMatrix<double>;
>  | typedef RealMatrix<double> Matrix;
>  |
>  | // external explicit instances ????
>  | template void RealMatrix<float>::is_symmetric();
>  | template void RealMatrix<double>::is_symmetric();
>  |
>  | file RealMatrix.cc:
>  |
>  | # include RealMatrix.h
>  | template<typename T>
>  | void
>  | RealMatrix<T>:: is_symmetric(void)
>  | {
>  |  // code here
>  | }
>  |
>  |
>  | It seems not to be possible with g++. In that case, the best way to
>  | keep maximum code precompiled is probably David's, despite the code
>  | duplication.
>
>  I'm not sure I understand the problem.  We already have some fucntions
>  defined in MArray2 (for example).  If needed, there are forwarding
>  functions in the Matrix class so that the return types of these
>  functions are correct.  In the case of is_symmetric, the return type
>  should always be bool, so I don't think a forwarding function would
>  even be necessary.  I would expect inheritance rules to make this
>  fairly simple.
>

Never mind. I just didn't know whether a particular template instance
can be linked from external .o file. I found out that this is normally
possible though.

>  jwe
>
>



-- 
RNDr. Jaroslav Hajek
computing expert
Aeronautical Research and Test Institute (VZLU)
Prague, Czech Republic
url: www.highegg.matfyz.cz