Re: [boost] Re: ublas performance

From: boost (boost_at_[hidden])
Date: 2001-11-23 18:18:13

Salut,

> I guess, this is a rather domain specific vector space decomposition
> (if dim (C) is prime for example, the decomposition seems to be
> useless).
you can call this a special decomposition, neverthetheless it has a
very import field of application : quantum mechanical lattice systems.
Once you want to describe solid state physics at small scale, you
will always have these kind of structure. 
The origin lies in the fact that a lattice site m can be described by 
operators acting on V_m, and the the comple vector space is a product 
of all V_m.
dim(C) won't be prime since it the product of dim(A) and dim(B),
since it is constructed in that way.

> I'll believe that ;-)
In addition there are some another symmetries leading to blocking .... :)

> I guess, your list of matrices conceptually can be viewed as tensor
> or as (domain specific?) special sparse matrix format. But may be,
> that's a matter of taste.
Sure, the structure is not universal, it results form the fact the
the operator is a sum of small oder medium sized number of
product states of A and B.

Best wishes,
Peter

P.S: Please excuse if I there's a large number of typos above,
I just came back from a sampling of wine from toscany :)