Module simd 

SIMD (Single Instruction, Multiple Data) optimization utilities.

This module provides SIMD optimization infrastructure:

• Platform detection (AVX2, AVX-512, NEON, etc.)
• Vectorization hints for the compiler
• SIMD-friendly data layouts and alignment
• Automatic vectorization checking
• Performance benchmarking for SIMD operations

Example

use tensorlogic_infer::{SimdCapabilities, SimdOptimizer, AlignedBuffer};

// Check SIMD capabilities
let caps = SimdCapabilities::detect();
println!("AVX2 supported: {}", caps.has_avx2());

// Create aligned buffer for SIMD operations
let buffer = AlignedBuffer::<f32>::new(1024);

// Optimize operations for SIMD
let optimizer = SimdOptimizer::new(caps);
let optimized_graph = optimizer.optimize(&graph)?;

Structs

AlignedBuffer

Aligned buffer for SIMD operations.

SimdCapabilities

SIMD capabilities detection.

SimdOptimizationHints

SIMD optimization hints.

Enums

CpuArchitecture

CPU architecture.

SimdError

SIMD optimization errors.

SimdInstructionSet

SIMD instruction set.

Functions

alignment_offset

Get the alignment offset needed to align a pointer.

is_simd_aligned

Check if a pointer is aligned for SIMD operations.

vectorize_hint

Vectorization hint for the compiler.