This pull request introduces changes to speed up the simulations.
The main (and breaking) change is that now the diffeqsolver family of methods return directly the solution from the DifferentialEquations.jl library, instead of manipulating the result. This surpassingly results in a major speed-up, frequently of 2x or more. However, care should be taken since this change is breaking, mainly since now the order of the time and spin components indices are swapped in the results returned by diffeqsolver. Furthermore, the way to access the interpolator is now also different. The example code in the run folder has been updated to work with the new conventions, but users of previous versions should change their custom scripts.
There are also some other minor changes, such as pre-allocating some variables repeatedly used in the differential equation to minimise the number of allocations when solving the ODE.
This pull request introduces changes to speed up the simulations.
The main (and breaking) change is that now the
diffeqsolverfamily of methods return directly the solution from the DifferentialEquations.jl library, instead of manipulating the result. This surpassingly results in a major speed-up, frequently of 2x or more. However, care should be taken since this change is breaking, mainly since now the order of the time and spin components indices are swapped in the results returned bydiffeqsolver. Furthermore, the way to access the interpolator is now also different. The example code in therunfolder has been updated to work with the new conventions, but users of previous versions should change their custom scripts.There are also some other minor changes, such as pre-allocating some variables repeatedly used in the differential equation to minimise the number of allocations when solving the ODE.