Namespace DotCompute.Backends.CPU.Kernels
Classes
- AdvancedSimdKernels
Advanced SIMD kernel implementations with complete FMA, integer SIMD, enhanced ARM NEON, and modern vectorization techniques.
- AdvancedSimdPatterns
Advanced SIMD patterns for high-performance scientific computing. Implements patterns from the SIMD Playbook for .NET 9 with Native AOT.
- ExecutorPerformanceMetrics
Performance metrics for the kernel executor.
- HardwareSimdKernelExecutor
High-performance SIMD kernel executor with hardware-specific optimizations.
- KernelPerformanceMetrics
Performance metrics for kernel execution.
- OptimizedSimdExecutor
Optimized SIMD kernel executor with advanced performance techniques:
- Instruction-level parallelism with multiple execution units
- Loop unrolling with optimal stride patterns
- Branch prediction optimization
- Cache-friendly memory access patterns
- Prefetch instructions for improved memory bandwidth
- Vectorized operations with fallback paths
- Runtime CPU feature detection and optimization Target: 4-8x performance improvement over scalar code
Enums
- NeonOperation
NEON operation types for enhanced ARM support.
- UnaryOperation
Unary operations supported by the SIMD kernel executor.