Sorry for the slow reply. It's been a lazy summer.
I'm cc-ing this anonymously to blitz-dev, hope this is okay.
The library is free, open-source, and will remain so. I have
no intentions of taking it commercial, ever.
I will be working primarily on other projects but will continue
to support blitz. Response time to support questions will slow
down when I am busy with other things. When I move on (in 3
years), the library will follow me.
> 1. Long term license development.
> ---------------------------------
>
> I recently ran into the big problems when Diffpack choosed to become a
> commercial package. Really sad story. Dead project with that kind of license
> and the lack of support for STL features (biggest advantage of blitz++)
>
> So I would like to know: Since You announced a change in the license policy,
> will this library stay free for private people and educational purposes or are
> You going to restrict the free versions e.g. in vector size?
>
> How long are You going to sit on that project? Will it live on
> when You leave the place You are at now?
>
> I ask this question because right now I am still in the decision phase
> whether to heavily rely on Your package or not.
>
>
> 2. Simple examples for expression templates
> -------------------------------------------
>
> Right know I am working on a solid mechanics project building
> up an implicit Galerkin-solver for tensorial nonlinear PDEs.
>
> I believe that Blitz++ is a good platform to build up this kind of stuff,
> also because its interface to LAPACK, f77 et al. seems to be quite simple.
> I plan to contribute this stuff to the public if it gets out of beta stage.
> (Blitz++ users www-page?)
>
> I read some of Your papers (Expr. Templates, Slides, C++ Templ. as Part.Eval.).
> I understood traits, I did understand code-generation a la metadot, but
> I am sorry, but I did not get the point on how to realize expression templates,
> for (x+1)/x, especially the DExpr<DExprIdentity ...>-stuff
>
> * please: Do You have some code examples (not those kill-the-beginner-
> blitz++-lib.cpp, but maybe more than the code fragments
> in "Expression Templates") perhaps at least the code of
> template<...> class DExpr {..}
>
> I would really enjoy to understand this method, because
> its potential is great for my tasks.
> (e.g. automatic build up of the implicit newton-solver-matrices)
I added examples to the beginning of the expression templates
paper:
http://extreme.indiana.edu/~tveldhui/papers/Expression-Templates/exprtmpl.html
>
> Thanks.
>
>
> 3. Advantage of Expression Templates diminished for run-time instantiation?
> ---------------------------------------------------------------------------
>
> (Should have been contributed to BZSUPPORT ...)
>
> I wondered, whether the advantages of epression templates used in
> BLITZ++ are still big enouth, when vector and matrix sizes are only known
> at run time. Is efficiency still great? Should I look forward
> to static vector sizes (big problem together with adaptivity),
> or should I forget about this problem.
The advantages are still there for dynamically sized vectors/arrays.
Especially if they are large. If they are small, then there is some
overhead from expression templates. But benchmark results show that
blitz is faster than non-e.t. array libraries (e.g. valarray) even
for small vectors.
> 4. Feature request: SymmetricTinyMatrix soon available?
> -------------------------------------------------------
>
> My SymmetricTensor template class uses a TinyVector right now, but
> a lot of headache would be gone with SymmetricTinyMatrix from blitz++.
>
What kind of features would you need from a SymmetricTinyMatrix class?
Cheers,
Todd
--------------------- blitz-dev list --------------------------------
* To subscribe/unsubscribe: use the handy web form at
http://oonumerics.org/blitz/lists.html
This archive was generated by hypermail 2b29 : Wed Feb 20 2002 - 04:30:10 EST