API Reference

All functions accept GPU arrays from any DLPack-compatible library (CuPy, PyTorch, MATLAB gpuArrays). Position and offset arrays are cast to float32 internally; index arrays are cast to int32. See conventions for the unit system and array layout.

Core Operations

Function	Description
das	Delay-and-sum, compounding over the sequence dimension
das_sparse	Delay-and-sum with per-target sequence subsets
greens	Frequency-domain Green's function summation (requires SM 70+)
truncate_rank	Rank truncation filter (truncated SVD)
einsum	Binary tensor contraction
interpolate	Structured 3D grid interpolation
Interpolator	Reusable interpolation plan (Python only)

Tensor Operations

Function	Description
gather	Gather elements along a dimension
scatter	Scatter elements along a dimension
contiguous_copy	Copy a strided tensor into a contiguous buffer

Utilities

Function	Description
cdist	Pairwise Euclidean distance
rect_dist	Distance from points to an axis-aligned rectangle
angle	Angle between vectors
Timer	GPU timing via CUDA events