On Thu, Jan 22, 2009 at 2:34 PM, Pavel Kaderavek <kada@xxxxxxxxxxxxx> wrote:
If I understood, your note is also valid for weights_csa1 and weights_csa2
(weights_csaC.py is different because it is connected to the crosscorelation
term and in the weights_dipY is necessary to add dimension for each variable
to include the possibility to work with an arbitrary number of the
interacting spins ... but I think it is the same issue as
in dipY_jw_mf.py), isn't it?
Yes, I've had a look at that other code now and I think it applies to
all. The difficulty is with the i index in data.ci_dipY[i]. An
elegent solution will need to be found for this one, so that the code
isn't duplicated. Maybe there could be one 'data' object per index
here. I think this would be the simplest and fastest (number
crunching wise) solution. Alternatively the index i could be optional
- i.e. a keyword argument in the form of:
def calc_dipY_S2_jw(data, params, i=None):
and if 'i' is not None, then the index used, otherwise the normal code
used. This, however, introduces an unnecessary if-statement at the
lowest level of number crunching code which will cause a slowdown.
Therefore I think we should avoid this. Do you think it would be
feasible to create an array of these data objects, one for each
neighbouring dipole, and then loop over them and operate one by one?
This will allow all code in maths_fns/, except for 'mf.py' to remain
unchanged (with the benefit of being easier, quicker, and more elegent
to implement).
Regards,
Edward
Re: cst branch development