JMATRIX - to-do list
--------------------
[x] - priority, 0-lowest, 2-highest
(+) - new function
($) - improving existing function
(!) - bug
(%) - cosmetic / clean up
Improvements to the method
--------------------------
[2] ($) The method stops working for calculations in gaussian basis and big
energies (mor or less for e>25). This IS NOT the program's fault,
this is the method's fault. Try to improve the method to accept
big energies in gaussian basis.
[2] ($) "The lambda problem" - the method works for scaling parameter lambda
equal to unity. For other values, results are not correct. This
is another method's fault, not a program, and needs to be resolved.
Using the scaling parameter may greatly improve the convergence.
[1] (+) Try to speed up the convergence in general, by modyfiing the method
Improvements to the code
------------------------
[2] (+) Introduce the Yukawa potential
[2] ($) Rewrite documentation as it is rather outdated at the moment
[1] ($) Try to speed up the program. Since the computational complexity
is proportional to n^4, this is the main resctriction on the
size of calculations.
[0] ($) Introduce integration with constants points and store results
[0] ($) Minimize memory usage. If n aproaches 5000 in the relativistic
calculations, program uses a few matrices of dimensions
10000 x 10000 x 8 bytes = 800 000 bytes = 0.8Gbytes...
[1] ($) Use only halves of the matrices, as they are all symmetric
[1] ($) look at the v and v_saved, do they really need to be stored
in memory?
[0] ($) Extend and improve the Makefile
[0] (+) Calculate difference between relativistic and non-relativistic
numerical deltas, and compare with difference between analytical ones
[0] (%) Unify matrices dimensions (now some indexes start from 0, some from 1)
[0] (+) Introduce numerical potential (interpolation?)
[0] (+) Introduce quarduple precision