RE: [boost] uBLAS formal review

From: Paul A. Bristow (boost_at_[hidden])
Date: 2002-06-25 06:06:22

I strongly support acceptance of uBLAS into Boost.

It seems to provide a good framework for matrix tasks in C++,
with an acceptable efficiency using expression templates.

It compiles with MSVC.net, Standard compiler version 7.0, (and 6.5)
and so should be widely applicable. However, it is not fully 'Standard'
conforming as it fails to compile with MS language extensions disabled.  It
is not yet clear if this is a serious fault.
There are no warnings at level 3. There are very many warnings C4100
unreferenced formal parameter and C4512 'class' : assignment operator could
not be generated which can probably be ignored,
or code changed to avoid them. Strict 'For scope' can be enforced.

I have built and run tests and benchmarks provided in debug and release mode
(but I have not been able to test the compiler optimisation as I only have
the non-optimising 'standard' compiler version, not the 'professional').
I have not examined the actual assembler code produced.
Nor have I investigated the various options provided as Macro values
(the documentation on these is very thin both in source and manual!)

Compiler versions rumoured to be under development might be sigificantly
more efficient.

The programs obviously cause some compiler stress as the heap size must be
increased from the default /Zm500 to /Zm1000.  Compile times are acceptable,
less than a minute or two for the sample code provided.

The basic documentation is fine (though some editorial work would be good).
Far more examples would be useful.

The layout is not quite the Boost C++ style recommendations.

The tests provided are an excellent proof of concept, but are not
exhaustive.
There are no examples of very small and very large size objects, and no
examples
of objects other than float, double, complex. For example, I would find an
interval type
most interesting - this is something that neither FORTRAN nor C can possibly
achieve.

From: "Herve Bronnimann" <hbr_at_[hidden]>

On Thu, Jun 13, 2002 at 11:07:25PM -0500, Dan'l Miller wrote:
>   What is the status of the interval arithmetic library?

It seems like lots of people need this interval library, me too.
I discussed with Jens a few months ago, and got permission to work on
it with the goal of a public review and final integration into boost.
As I am in academia, and not industry, I hired a student and a post-doc
working on it, and they're very good, so I expect there'll be a public
review by the end of the summer.

Best,            Hervé

Apart from some nice coding (whose internals are completely beyond me!),
Joerg Walter and Mathias Koch are providing rapid feedback and corrections.

There are also some tantilising links to the LAPACK algorithms -
surely THE crucial building block (along with decent math functions and
constants)
for C++ to finally become a viable system for real-life calculations
currently done using FORTRAN.  uBLAS promises to be a vital component.

Paul

Dr Paul A Bristow, hetp Chromatography
Prizet Farmhouse
Kendal, Cumbria
LA8 8AB      UK
[phone]
Mobile [phone]
mailto:pbristow_at_[hidden]

> -----Original Message-----
> From: boost-bounces_at_[hidden]
> [mailto:boost-bounces_at_[hidden]]On Behalf Of Ed Brey
> Sent: Friday, June 21, 2002 11:04 PM
> To: boost_at_[hidden]
> Subject: [boost] uBLAS formal review
>