MAOS
Multithreaded Adaptive Optics Simulator
Loading...
Searching...
No Matches
recon_tomo.c File Reference

Setup tomography parameters. More...

Macros

#define TIMING   0
#define TIC_tm
#define tic_tm
#define toc_tm(A)

Functions

void setup_recon_tomo_reg (recon_t *recon, const parms_t *parms)
void setup_recon_tomo_matrix (recon_t *recon, const parms_t *parms)
void setup_recon_tomo (recon_t *recon, const parms_t *parms, const powfs_t *powfs)
void Tomo_nea_gpt (Tomo_T *data, int nthread, int gpt)
void TomoR (dcell **xout, const void *A, const dcell *gin, const real alpha)
void TomoRt (dcell **gout, const void *A, const dcell *xin, const real alpha)
void TomoL (dcell **xout, const void *A, const dcell *xin, const real alpha)

Detailed Description

Setup tomography parameters.

Function Documentation

◆ setup_recon_tomo_reg()

void setup_recon_tomo_reg ( recon_t * recon,
const parms_t * parms )

Regularization terms for tomography. Can be called multiple times for Cn2 update.

Parameters
[in,out]reconThe reconstruction struct. recon->L2, invpsd, or fractal, and ZZT are set; recon->wt is clipped and normalized.
[in]parmsThe parameter struct.

◆ setup_recon_tomo_matrix()

void setup_recon_tomo_matrix ( recon_t * recon,
const parms_t * parms )

assemble tomography matrix. In CG mode, this function is not executed if tomo.assemble=0, Instead, the algorithm is contained in recon.c. When you modify anything, make sure you also do it there.

For integrated tomograhy:

\(\hat{x}=(G_{lgs}^{T}C_{lgs}^{-1}G_{lgs}+C_{x}^{-1}+G_{ngs}^{T}C_{ngs}^{-1} G_{ngs})^{-1}(G_{lgs}^{T}C_{lgs}^{-1}s_{lgs}+G_{ngs}^{T}C_{ngs}^{-1}s_{ngs}){\equiv}RL^{-1}RR s.\)

For split tomography, the terms regarding NGS are dropped:

\(\hat{x}_{lgs}=(G_{lgs}^{T}C_{lgs}^{-1}G_{lgs}+C_{x}^{-1})^{-1} G_{lgs}^{T}C_{lgs}^{-1}s_{lgs}\equiv R_L^{-1} R_R s.\)

The left hand side of linear equation (inside the inverse) is stored in RL. The right hand side of the linear equation (outside of the inverse) is stored in RR. The terms regarding the NGS are handled using low rank terms. The gradients from LGS and NGS and concatenated to \(s\).

In the LGS part, there is a global tip/tilt removal operator because LGS is insensitive to global tip/tilts.

For details see www.opticsinfobase.org/abstract.cfm?URI=josaa-19-9-1803

Parameters
[in,out]reconThe reconstruction struct. recon->RR and recon->RL are set.
[in]parmsThe parameter struct.

◆ setup_recon_tomo()

void setup_recon_tomo ( recon_t * recon,
const parms_t * parms,
const powfs_t * powfs )

Sets up the tomogrpahy turbulence reconstruction structs including wavefront reconstructor and DM fitting operator

Calls setup_recon_tomo_matrix() and setup_recon_fit_matrix() to setup the tomography and DM fitting matrix.

AHST is handled in setup_ngsmod().

MVST is handled in setup_recon_mvst().

MOAO is handled in setup_recon_moao().

Parameters
[in,out]reconThe reconstruction struct.
[in]parmsThe parameter struct.
[in]powfsThe powfs struct array.
Here is the call graph for this function:

◆ Tomo_nea_gpt()

void Tomo_nea_gpt ( Tomo_T * data,
int nthread,
int gpt )

Apply NEA weighting with optional GP': gg = GP' * NEAI * gg, or gg = NEAI * gg.

Parameters
[in,out]dataThe shared data. data->gg is replaced.
[in]nthreadNumber of threads. Currently ignored and 1 is used.
[in]gptWhether to apply GP'.

◆ TomoR()

void TomoR ( dcell ** xout,
const void * A,
const dcell * gin,
const real alpha )

Apply tomography right hand operator without using assembled matrix. Fast and saves memory. The operation is the same as the Tomo_nea_gpt and Tomo_iprop in TomoL, so merge the implemenations.

xout=HXW'GP'*NEAI(1-TTF*PTTF)*gin.

Parameters
[in,out]xoutThe output on xloc. Allocated if NULL; alpha*RR*gin is accumulated into it.
[in]AThe recon_t struct.
[in]ginThe gradients.
[in]alphaScaling factor.

◆ TomoRt()

void TomoRt ( dcell ** gout,
const void * A,
const dcell * xin,
const real alpha )

Transpose operation of TomoR. From xin to gout: gout=NEAI*(1-TTF*PTTF)*GP*HXW*xin.

Parameters
[in,out]goutThe output gradients. Allocated if NULL; alpha*RR'*xin is accumulated into it.
[in]AThe recon_t struct.
[in]xinThe input on xloc.
[in]alphaScaling factor.

◆ TomoL()

void TomoL ( dcell ** xout,
const void * A,
const dcell * xin,
const real alpha )

Apply tomography left hand side operator without using assembled matrix. Fast and saves memory. Only useful in CG. Accumulates to xout;

Some timing information: Generated in T410s with Intel Core i5 520 M @ 2.4 GHz (1 T is 1 thread, 2 T is 2 thread) and Poweredge with dual Xeon X5355 2.66Ghz (1 P is 1 thread, 8 P is 8 thread) Time 1 T 2 T 1 P 8 P HXW: 7.8 5.4 7.7 5.0 GP: 4.3 2.6 3.6 1.4 remove_mode: 0.3 0.3 0.4 0.6 neai: 0.5 0.3 0.5 0.4 GP': 3.2 2.1 3.4 1.0 HXW': 5.5 4.7 7.4 4.9 cxx: 3.2 2.5 3.5 2.7

Parameters
[in,out]xoutThe output on xloc. Allocated if NULL; alpha*RL*xin is accumulated into it.
[in]AThe recon_t struct.
[in]xinThe input on xloc.
[in]alphaScaling factor.