Skip to main content

polish

Function polish 

Source
pub fn polish(
    a: &CsrMatrix<f64>,
    b: &[f64],
    x: &mut [DoubleDouble],
    solve: &mut dyn FnMut(&mut [f64]),
    opts: &PolishOptions,
) -> Result<PolishReport, VerifyError>
Expand description

Algorithm P. x enters as the approximate solution (trailing components zero for an f64 solution) and leaves polished; solve overwrites its argument s with ≈ A⁻¹ s in f64.

The residual norm never increases (invariant P1 of Chapter 11.10): a correction that makes it worse is undone. The result is not a certificate; residual_of_dd certifies it.

§Errors

VerifyError::Dimension, also when a is not square.

§Panics

If a violates the invariants of its type.