Module grid 

Continuous-grid numerical integration — Simpson / trapezoidal rules (Kahan- compensated, SIMD-accelerated chunking) over zero-copy mmap-backed f64 grids, plus the DMA-alignment / state-suspension helpers. Relocated here from modalities::calculus as STEM math; the VM opcode surface that dispatches these stays in modalities::calculus (the modality), the numbers live in the solver.

Structs

ContinuousGrid

Zero-copy continuous data grid view.

Enums

AlignmentError

CalculusError

SimdWidth

Functions

detect_cache_line_size

detect_simd_width

integrate_simpsons_chunked

Simpson’s rule integration (standard, without Kahan compensation).

integrate_simpsons_kahan

Simpson’s rule integration with Kahan summation for precision.

integrate_trapezoidal_chunked

Trapezoidal rule integration (fallback for simpler integrands).

pack_f32_pair

Bit-packs two f32 values into a single 64-bit context field Used for packing step_size and Kahan compensation into the Quin context field

resolve_aligned_byte_offset

Translates a starting float boundary into a strictly 4096-aligned byte offset assuming the underlying grid is an array of contiguous 64-bit floats (8 bytes).

unpack_f32_pair

Unpacks a 64-bit context field back into two f32 values