Namespace DotCompute.Backends.Metal.MPS

Classes

MPSPerformanceMetrics

Performance metrics comparing MPS execution vs custom kernel execution.

MetalMPSDetector

Intelligent detector that analyzes kernel definitions to determine if they can be accelerated using Metal Performance Shaders instead of custom MSL kernels. Provides 3-10x speedup for standard operations like matrix multiply, convolution, etc.

MetalMPSKernel

Compiled kernel wrapper that represents MPS-accelerated operations. This is a lightweight placeholder that indicates a kernel can be executed via MPS. Actual execution is handled by the Metal accelerator's orchestration layer.

MetalMPSOrchestrator

Intelligent orchestrator that automatically selects between MPS and custom kernels based on operation type, data size, and device capabilities.

MetalPerformanceShadersBackend

Metal Performance Shaders backend for optimized ML and linear algebra operations. Provides 3x+ speedup for matrix operations compared to custom kernels.

PerformanceMetrics

Performance metrics for MPS vs CPU operations.

Structs

MPSCapabilities

Capabilities supported by Metal Performance Shaders on the current device.

Enums

MPSOperationType

Extended set of MPS operation types supporting standard compute patterns.