Struct ForwardModel 

pub struct ForwardModel<'a> { /* private fields */ }

Allocation-friendly CPU implementation of the image-plane FPM forward model.

Public spectrum and two-dimensional destination views must have standard C-style row-major layout. The layout is validated once at each public computational boundary and nonstandard views are rejected without copying. Methods returning an Array2 allocate their result; _into methods borrow caller-provided workspace and destinations.

Implementations

impl<'a> ForwardModel<'a>

pub fn new(model: &'a ImagePlaneModel) -> Result<Self>

Validates model and creates a CPU-backed evaluator borrowing it.

pub fn with_backend( model: &'a ImagePlaneModel, backend: Arc<dyn Backend>, ) -> Result<Self>

Validates model and creates an evaluator using the shared execution backend.

pub fn model(&self) -> &ImagePlaneModel

Returns the compiled model borrowed by this evaluator.

pub fn workspace(&self) -> Result<ForwardWorkspace>

Allocates reusable buffers sized for the model’s low-resolution grid.

pub fn extract_patch( &self, object_spectrum: ArrayView2<'_, Complex64>, source: usize, ) -> Result<Array2<Complex64>>

Allocates and returns one source’s low-resolution complex Fourier patch.

pub fn insert_patch_update( &self, object_spectrum: ArrayViewMut2<'_, Complex64>, source: usize, update: &[Complex64], scale: f64, ) -> Result<()>

Adds scale * update through the adjoint crop operator into object_spectrum.

pub fn apply_pupil( &self, patch: &[Complex64], pupil: &Pupil, destination: &mut [Complex64], ) -> Result<()>

Multiplies a row-major complex patch by the same-shaped sampled pupil.

pub fn forward_source_field( &self, object_spectrum: ArrayView2<'_, Complex64>, pupil: &Pupil, source: usize, ) -> Result<Array2<Complex64>>

Coherent low-resolution field for one illumination source.

pub fn forward_field( &self, object_spectrum: ArrayView2<'_, Complex64>, pupil: &Pupil, frame: usize, ) -> Result<Array2<Complex64>>

Coherent field for a non-multiplexed frame. A coded frame with exactly one source is represented by a field scaled by the square-root weight.

pub fn forward_intensity( &self, object_spectrum: ArrayView2<'_, Complex64>, pupil: &Pupil, frame: usize, ) -> Result<Array2<f64>>

Allocates the predicted (row, column) intensity array for one acquisition frame.

Coded frames are incoherent weighted sums of source intensities. Optional frame gain and optical background are applied to the result.

pub fn forward_intensity_into( &self, object_spectrum: ArrayView2<'_, Complex64>, pupil: &Pupil, frame: usize, workspace: &mut ForwardWorkspace, destination: &mut [f64], ) -> Result<()>

Writes one predicted frame to row-major destination, reusing workspace.

pub fn forward_intensity_stack_into( &self, object_spectrum: ArrayView2<'_, Complex64>, pupil: &Pupil, destination: &mut [f64], worker_count: usize, ) -> Result<()>

Predicts every measured frame into a contiguous [frame][row][column] destination using independent worker scratch.

Frame order and floating-point results within each frame are unchanged by worker_count. Values larger than the frame count are capped; zero is rejected. A worker count of one executes directly without spawning.

pub fn forward_intensity_stack( &self, object_spectrum: ArrayView2<'_, Complex64>, pupil: &Pupil, worker_count: usize, ) -> Result<Vec<f64>>

Allocation-owning convenience wrapper for Self::forward_intensity_stack_into.

pub fn residual( &self, object_spectrum: ArrayView2<'_, Complex64>, pupil: &Pupil, frame: usize, measured: &[f64], ) -> Result<Vec<f64>>

Returns row-major predicted - measured intensity residuals for one frame.

pub fn amplitude_projection( &self, field: &[Complex64], measured_intensity: &[f64], frame: usize, epsilon: f64, ) -> Result<Vec<Complex64>>

Replaces coherent-field amplitude with measured amplitude while retaining phase.

Frame gain and background are inverted before taking the square root of measured intensity. epsilon is a positive floor for dark predicted amplitudes.

pub fn frame_loss( &self, object_spectrum: ArrayView2<'_, Complex64>, pupil: &Pupil, frame: usize, measured: &[f64], loss_type: LossType, ) -> Result<f64>

Evaluates loss_type between one predicted and measured intensity frame.