- Computation optimisations
- lot of removed prints
Next work: - Sequential algo implemetation
- temperature simulations (check)
- Work on parallelization
To do:
- Remove all checks/prints used to debug
- Check all the flag set in the atom_vec_spin creator (very important for the reverse comm)
- Code DMI/ME interactions
- Start to work on parallel implementation of the integration
- For Paramag. simulations, the option "atom_modify" has to be set
ex: atom_modify sort 1000 4.0 (Freq,Dist).
- Actual time is now printed (c_mag[0] in compute_spin)
- Value of Gilbert's damping corrected
- Now even results for SD/Lammps comp. in purely paramg. or aniso. situations
- Pack and unpack reverse needed corrections (f only was set, not fm)
- Spin temperature is now computed (data c_mag[7] in spin_compute)
To do:
- Fcc with p p p bc is still not working
- If Zeeman/Aniso force not defined, error => to be removed
- Add DMI and ME (see if new file or add in the exchange file)
example: atom_modify sort 1000 4.0 (Freq. of sorting, Cutoff distance)
In order to print the actual time with the total mag., the vector associated to the mag. compute was modified. It is now: [time, Mx, My, Mz, |M|, En_mag]
Optimization of the spin_compute routine: energy and mag. have been gathered in a same loop.
Changes to come:
-Exchange interaction computation to check (loop on neighbors),
-Temperature/random fluctuations to correct (effects too strong),
-Physical results to check,
-Add final interactions (DMI, ME, Dipolar),
-Compute spin temperature (Nurdin and Ma formslisms),
-Work on MPI parallelization,
-Ewald sums to implement (see with Stan's pakage),
-See for prefered magnetic axis (Mitchell's idea),
this is done supporting the combination of type == 2 with style == 0,
i.e. a local scalar, which is not available in C++ (but there one can
access the compute style data member directly. for the python interface,
the pointer is automatically dereferenced and returned as a c_int.
