Crate sim_lib_interference_compute 

Normalized tile-local f32 lowering over the canonical Tensor executor.

The host keeps physical geometry and absolute phase in f64. Each prepared tile exposes only bounded offsets and residual phase to ordinary open Tensor operations, so no executor needs interference-specific behavior.

The lowering preserves the homogeneous, isotropic, three-dimensional scalar free-field scope. It does not add polarization, impedance, interfaces, obstacles, or diffraction. Amplitude-squared metrics are normalized observables, not physical intensity, power, or energy.

Structs

DenseExecutionEvidence

Execution and phase-range evidence retained with a dense f32 field.

DenseF32Field

A complete finite row-major phasor field evaluated as canonical f32 Tensors.

DifferentialError

Structural or non-finite input rejected before differential reporting.

DifferentialMaximum

Maximum observed error and its fixed tolerance at one cell.

DifferentialReport

Complete comparison report shared by portable and accelerated providers.

DifferentialTolerances

Published fixed tolerances for dense f32 interference conformance.

HardwareEvidenceMetrics

Provider-independent measurements used to classify one hardware run.

HardwareEvidenceReport

Sanitized differential, lifecycle, and determinism evidence for one profile.

InterferenceComputeLib

Loadable library that registers the Tensor study provider.

LoweredTile

One completed tile with separate canonical real and imaginary f32 Tensors.

LoweringError

A deterministic lowering diagnostic.

LoweringPlan

A completely admitted lowering bound to one canonical Tensor executor.

PhaseBudget

Accuracy limits applied to every source on every tile.

PlaneTile

One rectangular row-major tile of a physical sampling plane.

PlaneTileConstants

Tile-center constants for a forward plane source.

PointTileConstants

Tile-center constants for a spherical point source.

ScalarTolerance

Absolute and relative tolerances for one scalar quantity.

SourcePhaseEstimate

Predicted phase-domain accuracy for one source on one tile.

TensorStudyConfig

Admission, crossover, accuracy, and evidence settings for Tensor studies.

TensorStudySnapshot

Stable solver counters and last route.

TensorStudySolver

Study solver that preselects CPU or executes one resident Tensor plan.

TilePlan

Complete deterministic tiling chosen before any Tensor submission.

TileProfile

Provider and allocation limits used to partition and admit a plane.

Enums

ConformanceMetric

A quantity checked by the portable f32 differential contract.

CpuFallbackReason

A reason the Tensor provider was declined before any submission.

HardwareMeasurementResult

Whether a physical-provider observation can satisfy the hardware gate.

PreflightCheck

The validation boundary that refused a lowering request.

ProviderRoute

Last routing outcome observed by a TensorStudySolver.

SourceTileConstants

Prepared constants and phase estimate for one source on one tile.

Constants

HARDWARE_DETERMINISM_REPEATS

Number of same-profile executions required by the hardware contract.

REQUIRED_TENSOR_OPERATION_NAMES

Stable names of the only open Tensor operations emitted by this lowering.

Statics

RECIPES

Checked modeled and optional physical-provider evidence recipes.

Functions

compare_dense_to_reference

Compares one dense f32 result with the deterministic f64 reference field.

compare_materialized_to_reference

Compares one materialized provider result with the deterministic f64 oracle.

interference_compute_lib_symbol

Stable library symbol for the accelerated interference provider.

solve_dense_f32_cpu

Executes the normalized lowering through the canonical dense f32 CPU executor.