Module compute 

Chronix Compute Engine

CPU compute engine for Chronix analytics operations — dot products, matrix solving, exponential smoothing, differencing, autocorrelation, and batch statistics. Built on multi-tier SIMD acceleration and rayon parallelism.

Architecture

┌──────────────────────────────────┐
│        ComputeEngine trait       │
├──────────────────────────────────┤
│  batch_dot_product()             │
│  batch_matrix_solve()            │
│  batch_exponential_smooth()      │
│  batch_difference()              │
│  batch_autocorrelation()         │
└──────────────┬───────────────────┘
               │
         ┌─────┴────┐
         │ CpuEngine │
         │ rayon+SIMD│
         └───────────┘

SIMD Operations

Multi-tier SIMD for common statistical operations:

• simd_sum, simd_mean, simd_variance, simd_min_max, simd_dot_product, simd_std_dev
• x86_64: runtime-detected AVX-512F (8×f64) → AVX2+FMA (4×f64) → SSE2 (2×f64, baseline)
• aarch64: NEON (2×f64, baseline)
• Scalar 4-wide ILP fallback on other architectures
• simd_tier() returns the active tier for diagnostics

Buffer Pool

BufferPool<T> provides reusable Vec<T> to minimize allocation overhead in hot compute paths.

Structs

BufferPool

A pool of reusable Vec<T> buffers that avoids repeated allocation on hot compute paths: get() hands out a cleared buffer, put() returns it.

ComputeConfig

Configuration for the compute engine.

CpuEngine

CPU compute engine using SIMD and rayon parallelism.

Enums

ComputeError

Errors from compute operations.

Traits

ComputeEngine

Compute engine trait — defines batch numerical operations used by forecast, anomaly detection, and multivariate engines.

Functions

simd_dot_product

Computes the dot product of two equal-length slices using SIMD.

simd_mean

Computes the arithmetic mean of a slice.

simd_mean_variance

Two-pass mean + sample variance using SIMD kernels.

simd_min_max

Computes the minimum and maximum of a slice simultaneously using SIMD.

simd_population_variance

Computes the population variance of the slice using SIMD (divides by n).

simd_range_filter_i64

SIMD-accelerated timestamp range filter.

simd_std_dev

Computes the sample standard deviation.

simd_sum

Computes the sum of a slice with multi-tier SIMD acceleration and Kahan (compensated) summation for numerical stability.

simd_tier

Returns a human-readable string describing the active SIMD tier.

simd_variance

Computes the population variance given a precomputed mean.