Module simd_agg 

SIMD-accelerated aggregation kernels for timeseries f64 columns.

Runtime CPU detection selects the fastest available path:

• AVX-512 (512-bit, 8 f64/op) — Intel Xeon, AMD Zen 4+
• AVX2+FMA (256-bit, 4 f64/op) — most x86_64 since 2013
• NEON (128-bit, 2 f64/op) — ARM64 (Graviton, Apple Silicon)
• Scalar fallback — auto-vectorized by LLVM

All kernels use Kahan compensated summation for numerical accuracy.

Structs

TsSimdRuntime

SIMD runtime for timeseries f64 aggregation.

Functions

ts_runtime

Get the global timeseries SIMD runtime.