Struct IrDlrTransform 

pub struct IrDlrTransform { /* private fields */ }

Linear map between the coefficients of an IR basis and of a DLR on the same domain.

The DLR basis function of pole ω_i is expanded in the IR basis as u_i = Σ_l T[l, i] U_l with T[l, i] = -s_l V_l(ω_i) (w_i / w^{IR}_i), where w_i and w^{IR}_i are the pole weights of the DLR and of the IR kernel. DLR -> IR applies T; IR -> DLR is its least-squares inverse.

Implementations

impl IrDlrTransform

pub fn ir_size(&self) -> usize

IR basis size.

pub fn dlr_size(&self) -> usize

Number of DLR poles.

pub fn matrix(&self) -> &Matrix<f64>

The ir_size x dlr_size matrix T mapping DLR to IR coefficients.

pub fn ir_to_dlr_nd<T: FitScalar>( &self, backend: Option<&GemmBackendHandle>, gl: &TypedTensor<T>, dim: usize, ) -> Result<TypedTensor<T>, Error>

IR -> DLR along axis dim (least squares).

The mathematical transform does not depend on the axis placement or the memory order of the arrays, but the floating-point result can: a different axis placement selects a different contraction geometry, and the GEMM backend decides how that geometry is blocked. Bitwise agreement across axis placements is not guaranteed, and conditioning or cancellation can amplify the difference (SpM-lab/sparse-ir-rs#326).

Errors

The errors of DiscreteLehmannRepresentation::from_ir_nd.

pub fn dlr_to_ir_nd<T: FitScalar>( &self, backend: Option<&GemmBackendHandle>, g_dlr: &TypedTensor<T>, dim: usize, ) -> Result<TypedTensor<T>, Error>

DLR -> IR along axis dim.

The mathematical transform does not depend on the axis placement or the memory order of the arrays, but the floating-point result can: a different axis placement selects a different contraction geometry, and the GEMM backend decides how that geometry is blocked. Bitwise agreement across axis placements is not guaranteed, and conditioning or cancellation can amplify the difference (SpM-lab/sparse-ir-rs#326).

Errors

The errors of DiscreteLehmannRepresentation::to_ir_nd.