Expand description
SIMD-accelerated operations for neural networks
This module provides vectorized implementations of common neural network operations using unified SIMD operations from scirs2-core for significant performance improvements and cross-platform compatibility. All functions are feature-gated with “simd” feature.
IMPORTANT: All SIMD operations are delegated to scirs2-core’s unified SIMD abstraction layer in compliance with the project-wide SIMD policy. Direct use of SIMD intrinsics is FORBIDDEN.
Structs§
- SIMD
Operations - SIMD operations for neural network computations (fallback implementation)