Refactoring of magnet integrators

[lfarv]

Here is a full refactoring of the magnet passmethods.
It uses include files which can be combined to generate all the different magnet passmethods. At the moment, this is based on:

6 Propagation methods:

• drift_fast.h
• drift_expanded.h
• drift_exact.h
• drift_exactbend.h
• drift_exactstrbend.h
• drift_E2.h

5 kick methods

• kick_kn,h
• kick_exactkn.h
• kick_k1h_kn.h
• kick_h_k0h_k1h_kn.h
• kick_E2.h

3 geometry descriptions

• straight_multipole.h
• curved_dipole.h
• straight_dipole.h

2 integrators

• 4th order Forest-Ruth
• 6th order Yoshida

And a single integrator template: magnet_template.h

With this, the standard StrMPoleSymplectic4Pass pass method is written as:

#define MAGNET_PASS `StrMPoleSymplectic4Pass`
#define INTEGRATOR_4

#include "drift_fast.h"    /* drift */
#include "kick_kn.h"  /* kick */
#include "straight_multipole.h"

#include "magnet_template.h"

And the corresponding radiative passmethod is written as:

#define MAGNET_PASS StrMPoleSymplectic4RadPass
#define INTEGRATOR_4
#define RADIATION
#define DIFFUSION

#include "drift_expanded.h"
#include "kick_kn.h"
#include "straight_multipole.h"

#include "magnet_template.h"

Obviously you can't mix randomly propagation and kicks. But creating new passmethods is easy, like for instance a 6th order StrMPoleSymplectic6Pass:

#define MAGNET_PASS StrMPoleSymplectic6Pass
#define INTEGRATOR_6

#include "drift_fast.h"    /* drift */
#include "kick_kn.h"  /* kick */
#include "straight_multipole.h"

#include "magnet_template.h"

Or a quantum diffusion version ExactSectorBendQuantPass:

#define MAGNET_PASS ExactSectorBendQuantPass
#define INTEGRATOR_6
#define QUANTUM
#define NO_OMP  /* because of problems with random generator and OpenMP */

#include "drift_exactbend.h"
#include "kick_k1h_kn.h"
#include "curved_dipole.h"

#include "magnet_template.h"

[simoneliuzzo]

Dear @lfarv,

to review I may simply run some basic tracking, DA, etc on this branch compared to the master?
May be you already did this.

thank you
best regards
Simone

[lfarv]

@simoneliuzzo: of course you are welcome to use the branch for all kinds of optics computation: this helps a lot in validating the modifications. I recall the expected behaviour:

• the computations should be strictly identical to the ones in the "Quad_curvature_correction" branch, but differ from the master branch for bending magnet core and end faces. The differences are only in the organisation of the code,
• the speed should be equivalent the the master branch,
• I made successful comparison of single magnet tracking with XSuite, but I did not run all the various cases that you propose, so please try.