adpcg Struct Reference

Working struct for the adaptive CG solver. More...

#include <adpcg.h>

Data Fields
void *	A
	LHS data.
void *	r
	Residual.
void *	rnew
	Workspace array.
void *	d
	Workspace array.
void *	pinvr
	Workspace array.
void *	Ad
	Workspace array.
void *	x
	CG solution vector.
void *	aux
	CG auxiliary array.
void *	btmp
	CG temporary RHS array.
cgint	ptype
	Pre-conditioner type.
void *	chol
	Cholesky pre-conditioner.
void *	diag
	Diagonal pre-conditioner.
double	tol
	Relative tolerance of CG.
double	rnrm
	Residual norm.
double	avgsvtime
	Averate solution time of previous CG solves.
double	avgfctime
	Average factorization time of previous CG solves.
double	currenttime
	Buffer of solution time in the current round.
double	latesttime
	Time for the latest solve.
cgint	n
	Dimension of linear system.
cgint	niter
	Number of iterations in most recent solve.
cgint	maxiter
	Maximum number of iterations.
cgint	status
	Solution status.
cgint	reuse
	Reuse Cholesky pre-conditioner.
cgint	nused
	Number of rounds current Cholesky pre-conditioner is already used.
cgint	nmaxiter
	Number of non-successfull solves.
cgint	restart
	Restart frequency.
cgint	nfactors
	Number of factorizes performed so far.
cgint	nrounds
	Number of rounds of solves.
cgint	nsolverd
	Number of solves in a round.
cgint	nsolves
	Number of linear systems solved.
void(*	v_init )(void *v)
	Initialize vector.
cgint(*	v_alloc )(void *v, cgint n)
	Allocate memory for v.
void(*	v_free )(void *v)
	Free the internal memory of vector.
void(*	v_copy )(void *s, void *t)
	Copy s to t.
void(*	v_reset )(void *v)
	Reset vector to 0.
void(*	v_norm )(void *v, double *nrm)
	nrm = norm(v)
void(*	v_axpy )(double a, void *x, void *y)
	y = y + a * x
void(*	v_axpby )(double a, void *x, double b, void *y)
	y = a * x + b * y
void(*	v_zaxpby )(void *z, double a, void *x, double b, void *y)
	z = a * x + b * y
void(*	v_dot )(void *x, void *y, double *xTy)
	xTy = x' * y
cgint(*	A_chol )(void *A)
	Compute Cholesky of A.
cgint(*	A_cond )(void *A)
	Evaluate conditioning of A.
void(*	A_getdiag )(void *A, void *diag)
	Compute diagonal pre-conditioner diag = diag(A)
cgint(*	diagpcd )(void *diag, void *v)
	Diagonal Preconditioning operation v = P \ v.
cgint(*	cholpcd )(void *chol, void *v, void *aux)
	Cholesky Preconditioning operation v = P \ v.
void(*	Av )(void *A, void *v, void *Av)
	Compute Av = A * v.
cgint(*	Ainv )(void *A, void *v, void *aux)
	Solve Ainvv = A \ v.

Detailed Description

Working struct for the adaptive CG solver.

To make the CG solver more general, the operations on

1. vector 2. matrix 3. matrix-vector are presented in abstract function pointers and users can define their own implementations for various purposes

More details: the following operations are expected Vector:

1. v_init(v) Initialize a vector struct
2. v_alloc(v) Allocate internal memory for a vector struct
3. v_free(v) Free the internal memory for a vector struct
4. v_copy(s, t) Copy content from vector s to vector t
5. v_reset(v) Set the content of vector 0
6. v_norm(v, &nrm) Compute norm of v
7. v_axpy(a, x, y) Compute y = y + a * x
8. v_zaxpby(z, a, x, b, y) Compute z = a * x + b * y
9. v_dot(x, y, &dot) Compute x' * y

Matrix:

1. A_chol(A) Perform Cholesky decomposition
2. is_illcond = A_cond(A) Get a boolean variable reflecting the conditioning of A

Matrix-vector:

1. diagpcd(diag, v) Perform diagonal pre-conditioning
2. cholpcf(chol, v) Perform Cholesky pre-conditioning

---

The documentation for this struct was generated from the following file:

• src/adpcg.h