cusp::monitor< ValueType > Class Template Reference

Detailed description

template<typename ValueType>
class cusp::monitor< ValueType >

Implements standard convergence criteria and reporting for iterative solvers.

Template Parameters

ValueType

scalar type used in the solver (e.g. float or cusp::complex<double>).

Overview

The monitor terminates iteration when the residual norm satisfies the condition ||b - A x|| <= absolute_tolerance + relative_tolerance * ||b|| or when the iteration limit is reached. Classes to monitor iterative solver progress, check for convergence, etc. Follows the implementation of Iteration in the ITL:

See Also

http://www.osl.iu.edu/research/itl/doc/Iteration.html

Example

The following code snippet demonstrates how to configure the monitor and use it with an iterative solver.

#include <cusp/csr_matrix.h>
#include <cusp/monitor.h>
#include <cusp/krylov/cg.h>
#include <cusp/gallery/poisson.h>

int main(void)
{
    // create an empty sparse matrix structure (CSR format)
    cusp::csr_matrix<int, float, cusp::device_memory> A;

    // initialize matrix
    cusp::gallery::poisson5pt(A, 10, 10);

    // allocate storage for solution (x) and right hand side (b)
    cusp::array1d<float, cusp::device_memory> x(A.num_rows, 0);
    cusp::array1d<float, cusp::device_memory> b(A.num_rows, 1);

    // set stopping criteria:
    //  iteration_limit    = 100
    //  relative_tolerance = 1e-6
    cusp::monitor<float> monitor(b, 100, 1e-6);

    // solve the linear system A x = b
    cusp::krylov::cg(A, x, b, monitor);

    // report solver results
    if (monitor.converged())
    {
        std::cout << "Solver converged to " << monitor.relative_tolerance() << " relative tolerance";
        std::cout << " after " << monitor.iteration_count() << " iterations" << std::endl;
    }
    else
    {
        std::cout << "Solver reached iteration limit " << monitor.iteration_limit() << " before converging";
        std::cout << " to " << monitor.relative_tolerance() << " relative tolerance " << std::endl;
    }

    return 0;
}

Examples:

ainv.cu, bicgstab.cu, cg.cu, cg_m.cu, cg_raw.cu, cr.cu, custom_amg.cu, diagonal.cu, gmres.cu, monitor.cu, smoothed_aggregation.cu, stencil.cu, and verbose_monitor.cu.

Definition at line 101 of file monitor.h.

#include <monitor.h>

Inheritance diagram for cusp::monitor< ValueType >:

Public Types
typedef cusp::norm_type < ValueType >::type	Real

Public Methods
template<typename VectorType >
	monitor (const VectorType &b, const size_t iteration_limit=500, const Real relative_tolerance=1e-5, const Real absolute_tolerance=0, const bool verbose=false)
	Constructs a monitor for a given right-hand-side b. More...
void	operator++ (void)
	Increments the iteration count.
bool	converged (void) const
	Indicates whether the last tested residual satisfies the convergence tolerance. More...
Real	residual_norm (void) const
	Euclidean norm of last residual. More...
size_t	iteration_count (void) const
	Returns the number of iterations that the monitor has executed. More...
size_t	iteration_limit (void) const
	Returns the maximum number of iterations. More...
Real	relative_tolerance (void) const
	Returns the relative tolerance. More...
Real	absolute_tolerance (void) const
	Returns the absolute tolerance. More...
Real	tolerance (void) const
	Return the tolerance equal to absolute_tolerance() + relative_tolerance() * ||b||. More...
template<typename Vector >
bool	finished (const Vector &r)
	Applies convergence criteria to determine whether iteration is finished. More...
void	set_verbose (bool verbose_=true)
	Sets the verbosity level of the monitor. More...
bool	is_verbose (void)
	Gets the verbosity level of the monitor. More...
template<typename Vector >
void	reset (const Vector &b)
	Resets the monitor using same convergence criteria. More...
void	print (void)
	Prints the number of iterations and convergence history information.
Real	immediate_rate (void)
	Returns the immedidate convergence rate. More...
Real	geometric_rate (void)
	Returns the geometric convergence rate. More...
Real	average_rate (void)
	Returns the average convergence rate. More...

Public Members
cusp::array1d< Real, cusp::host_memory >	residuals