Module math 

SIMD-accelerated mathematical operations

This module provides mathematical functions with an architecture-specific fast path for a subset of operations. Key operations (sqrt, abs, floor, ceil, round) use NEON hardware instructions directly on aarch64. Transcendental functions (exp, log, sin, cos, …) currently call the standard library’s scalar libm implementation (e.g. f32::exp, f32::ln) on every target, including aarch64 — there is no SIMD polynomial approximation implemented for transcendentals today.

Architecture Support

• aarch64: NEON intrinsics for sqrt (vsqrtq_f32), abs (vabsq_f32), and floor/ceil/round (vrndmq_f32/vrndpq_f32/vrndaq_f32). Transcendental functions (exp, ln, log2, log10, sin, cos, …) fall back to scalar libm calls even on this target.
• All other targets (including x86-64): Scalar fallback with auto-vectorization hints for every operation; there is no hand-written SSE2/SSE4.1/AVX2 path in this module today.

Supported Operations

• Power/Root: sqrt, cbrt, pow, exp, exp2
• Logarithms: log, log2, log10
• Trigonometric: sin, cos, tan, asin, acos, atan, atan2
• Hyperbolic: sinh, cosh, tanh
• Special: abs, signum, floor, ceil, round, fract

Performance

Expected speedup over scalar: 3-6x for most operations

Example

use oxigeo_algorithms::simd::math::{sqrt_f32, exp_f32};

let data = vec![1.0, 4.0, 9.0, 16.0];
let mut result = vec![0.0; 4];

sqrt_f32(&data, &mut result)?;
assert_eq!(result, vec![1.0, 2.0, 3.0, 4.0]);

Functions

abs_f32

Compute absolute value element-wise using hardware SIMD

acos_f32

Compute arccosine element-wise

asin_f32

Compute arcsine element-wise

atan2_f32

Compute two-argument arctangent element-wise: atan2(y, x)

atan_f32

Compute arctangent element-wise

ceil_f32

Compute ceiling element-wise using hardware SIMD

cos_f32

Compute cosine element-wise

cosh_f32

Compute hyperbolic cosine element-wise

exp2_f32

Compute 2^x element-wise

exp_f32

Compute exponential (e^x) element-wise

floor_f32

Compute floor element-wise using hardware SIMD

fract_f32

Compute fractional part element-wise: fract(x) = x - floor(x)

ln_f32

Compute natural logarithm (ln) element-wise

log2_f32

Compute base-2 logarithm element-wise

log10_f32

Compute base-10 logarithm element-wise

pow_f32

Compute power (base^exponent) element-wise

round_f32

Compute round (nearest integer) element-wise using hardware SIMD

sin_f32

Compute sine element-wise

sinh_f32

Compute hyperbolic sine element-wise

sqrt_f32

Compute square root element-wise using hardware SIMD

tan_f32

Compute tangent element-wise

tanh_f32

Compute hyperbolic tangent element-wise