Re: [boost] Re: Standard Matrix interfaces

From: Gabriel Dos Reis (Gabriel.Dos-Reis_at_[hidden])
Date: 2001-08-20 15:54:50

helmut.zeisel_at_[hidden] writes:

| --- In boost_at_y..., Gabriel Dos Reis <Gabriel.Dos-Reis_at_c...> wrote:
| > Dietmar Kuehl <dietmar_kuehl_at_y...> writes:
| > 
| > | Hi,
| > | 
| > | Helmut Zeisel wrote:
| > | > However, if we wait for a matrix library that covers
| > | > all these issues, we will never have a standard C++ matrix
| interface.
| > | 
| > | You don't want to define a matrix class anyway! You want to define
| > | a generic interface for matrices which is used to access some
| > | representation.
| > 
| > I couldn't agree more.
| > 
| > One should take in separating algorithms from storage
| > layout/management. 
| 
| As I wrote already in a previous answer, it depends on the purpose
| of the "matrix interface".
| If it is used for writing new algorithms from scratch,
| you are right (although I doubt whether efficient
| algorithms can be written without knowledge e.g. whether
| the matrix is stored row by row or column by column).

Most Blas algorithms can be written generically, given sufficient
clean and efficient interface.  Implementations details like storage
layout are handled via traits. 

-- Gaby