Expand description
Enhanced GPU acceleration with real compute shaders and kernel optimization
This module provides production-ready GPU acceleration using compute shaders, advanced memory management, and optimized kernels for metrics computation. Supports CUDA, OpenCL, and WebGPU backends with automatic fallback.
Structs§
- Auto
Tuning Config - Auto-tuning configuration
- Backend
Info - GPU backend information
- Bandwidth
Measurement - Bandwidth measurement
- Cache
Statistics - Cache statistics
- Compute
Kernel - Compute kernel representation
- Compute
Stream - Compute stream for async execution
- Cuda
Backend - Cuda
Context - CUDA context information
- Cuda
Device Properties - CUDA device properties
- Cuda
Kernel Info - CUDA kernel information
- Cuda
Memory Info - CUDA memory allocation info
- Defragmentation
Settings - Defragmentation configuration
- Dependency
Tracker - Dependency tracker for stream synchronization
- Enhanced
GpuEngine - Enhanced GPU compute engine with multiple backend support
- GpuMemory
Handle - GPU memory handle for buffer management
- GpuMemory
Pool - GPU memory pool for efficient allocation
- GpuProfiler
- GPU performance profiler
- Kernel
Cache - Kernel cache for optimized reuse
- Kernel
Features - Kernel features for ML optimization
- Kernel
Optimization Params - Kernel optimization parameters
- Kernel
Optimizer - Automatic kernel optimizer
- Kernel
Parameter - Kernel parameter definition
- Kernel
Params - Kernel execution parameters
- Load
Balancing Config - Load balancing configuration
- Memory
Usage Stats - Memory usage statistics
- Open
ClBackend - Open
ClContext - OpenCL context information
- Open
ClDevice Properties - OpenCL device properties
- Open
ClKernel Info - OpenCL kernel information
- Open
ClMemory Info - OpenCL memory allocation info
- Optimization
Result - Optimization result
- Search
Space - Parameter search space for auto-tuning
- Stream
Manager - Stream manager for concurrent kernel execution
- Stream
Scheduler - Stream scheduler for optimal resource utilization
- Transfer
Measurement - Memory transfer measurement
- Utilization
Measurement - GPU utilization measurement
- WebGpu
Adapter - WebGPU adapter information
- WebGpu
Backend - WebGpu
Buffer Info - WebGPU buffer information
- WebGpu
Device - WebGPU device
- WebGpu
Limits - WebGPU limits
- WebGpu
Pipeline Info - WebGPU compute pipeline information
Enums§
- Allocation
Strategy - Memory allocation strategies
- Eviction
Policy - Cache eviction policies
- Kernel
Parameter Type - Types of kernel parameters
- Memory
Access Pattern - Memory access patterns
- Parallelism
Type - Types of parallelism
- Scheduling
Strategy - Stream scheduling strategies
- Stream
Priority - Stream priority levels
- Stream
Status - Stream execution status
- Transfer
Direction - Transfer direction
- Tuning
Strategy - Auto-tuning strategies
Traits§
- GpuBackend
- Trait for GPU compute backends (CUDA, OpenCL, WebGPU)
- Optimization
Model - Machine learning model for kernel optimization