batchlas::convergence_error Class Reference

Overview

An iterative kernel did not converge, or a factorisation broke down on the data.

An eigen/SVD sweep budget exhausted, a bidiagonal QR that never deflated, an ILU(k) pivot that was zero with no usable shift: the LAPACK info > 0 case.

Retryable with different parameters (a looser tolerance, a higher sweep cap, a different algorithm, a rescaled input), not with the same ones.

Note

The exception is batch-wide: it says some item failed, not which. Pass the optional per-item info span (syev, syevx, gesvd, steqr, stedc) to learn which.

See also

Error model

Definition at line 170 of file error.hh.

#include <batchlas/error.hh>

Inheritance diagram for batchlas::convergence_error:

Constructors

convergence_error() [1/2]

batchlas::convergence_error::convergence_error ( const std::string & what_arg )

inlineexplicit

Definition at line 172 of file error.hh.

convergence_error() [2/2]

batchlas::convergence_error::convergence_error ( const char * what_arg )

inlineexplicit

Definition at line 173 of file error.hh.