on 16.08.2009 at 23:29 Rutger ter Borg wrote :
Dear Pavel, interesting stuff. Although I think it is promising, what exactly does it have to offer over existing matrix libraries (and uBLAS for that matter)? c++ (read object oriented) handling of entities like
IMHO, for a next generation C++ linear algebra library to be really successful, I think the following properties are desired * split containers from expressions don't get what you mean by that
C = a*A*B + b*C in contrast to uBLAS: C = a*prod(A, B) + b*C //citation from uBLAS docs they are already splitted
* split algorithms from expressions, i.e., enable pluggable/customisable back-ends into the expression templates actually it's rather trivial specialize a template for concrete expression to your taste and you are done
* achieve performance of FORTRAN-based programs (dominant in HPC) that is the hardest part i suppose but i'm sure it is achievable
-- Pavel