Struct SimdOps

#[repr(align(32))]
pub struct SimdOps;

SIMD-accelerated operations optimized for HFT market data processing

Provides safe, portable SIMD implementations of common numerical operations used in high-frequency trading applications. All operations use the wide crate for guaranteed safety and cross-platform compatibility.

Design Principles

• Safety first: Zero unsafe code, guaranteed memory safety
• Portability: Works across x86_64, ARM, and other architectures
• Performance: 2-4x speedup through SIMD vectorization
• Simplicity: Clean API that abstracts SIMD complexity

Cache Alignment

32-byte alignment ensures optimal AVX/AVX2 performance and prevents cache line splits that could degrade SIMD operation efficiency.

Implementations

impl SimdOps

pub fn min_f64(values: &[f64]) -> f64

Apply SIMD-accelerated min to an array of f64 values Uses safe portable SIMD operations

pub fn max_f64(values: &[f64]) -> f64

Apply SIMD-accelerated max to an array of f64 values Uses safe portable SIMD operations

pub fn sum_f64(values: &[f64]) -> f64

Apply SIMD-accelerated sum to an array of f64 values Uses safe portable SIMD operations

pub fn avg_f64(values: &[f64]) -> f64

Apply SIMD-accelerated average to an array of f64 values Uses safe portable SIMD operations

pub fn weighted_avg_f64(values: &[f64], weights: &[f64]) -> f64

Apply SIMD-accelerated weighted average to arrays of values and weights Uses safe portable SIMD operations

pub fn convert_to_f64<T: Into<f64> + Copy>(values: &[T]) -> Vec<f64>

Convert an array of decimal prices to f64 for SIMD operations

pub const fn avx2_supported() -> bool

Check if AVX2 is supported at runtime Kept for backward compatibility but no longer needed with portable SIMD

pub fn min_f64_aligned(values: &[f64]) -> f64

Apply SIMD-accelerated min using simd_aligned containers for optimal memory alignment This version ensures data is properly aligned for SIMD operations and uses thread-local buffers

pub fn max_f64_aligned(values: &[f64]) -> f64

Apply SIMD-accelerated max using simd_aligned containers for optimal memory alignment This version uses thread-local buffers to avoid repeated allocations

pub fn sum_f64_aligned(values: &[f64]) -> f64

Apply SIMD-accelerated sum using simd_aligned containers This version uses thread-local buffers to avoid repeated allocations