Namespace DotCompute.Backends.CUDA.RingKernels
Classes
- CudaCompiledRingKernel
Represents a compiled Ring Kernel with PTX module and function pointer.
- CudaDevicePointerBuffer
Simple read-only wrapper for CUDA device memory pointers used by message queues.
- CudaMessageQueue<T>
CUDA-based lock-free message queue for inter-kernel communication.
- CudaRingKernelCompiler
Compiles C# ring kernel definitions to CUDA C persistent kernels.
- CudaRingKernelRuntime
CUDA runtime for managing persistent ring kernels.
- GpuRingBufferBridge<T>
Bidirectional bridge between host IMessageQueue<T> and GPU ring buffer.
- GpuRingBuffer<T>
Manages GPU-resident ring buffer memory for message passing.
- RingKernelConfig
Configuration for a ring kernel compilation.
- RingKernelHandlerTranslator
Translates unified ring kernel methods (RingKernelContext + TRequest) to CUDA device functions.
- RingKernelInfrastructure
Provides CUDA infrastructure headers for Ring Kernel compilation.
- RoutingTableBuilder
Builds and manages GPU-resident routing tables for inter-kernel messaging.
- TopicRegistryBuilder
Builds and manages GPU-resident topic registries for pub/sub messaging.
Structs
- KernelHealthStatus
Health monitoring data for a Ring Kernel (GPU-resident).
- KernelRoutingTable
GPU-resident routing table for kernel-to-kernel message passing.
- MultiKernelBarrier
Multi-kernel barrier for synchronizing persistent Ring Kernels across GPU.
- RingKernelControlBlock
GPU-resident control block for managing persistent ring kernel state.
- TaskDescriptor
Task descriptor for dynamic work queue.
- TaskQueue
Lock-free task queue with work-stealing support for dynamic load balancing.
- TopicRegistry
GPU-resident topic registry for pub/sub messaging.
- TopicSubscription
GPU-resident topic subscription entry for pub/sub messaging.
Interfaces
- IGpuRingBuffer
Non-generic interface for GPU ring buffers, providing access to head/tail pointers.
- IGpuRingBufferBridge
Non-generic interface for GPU ring buffer bridges, enabling polymorphic access when the message type is only known at runtime.
Enums
- CudaRingKernelCompiler.CompilationStage
Compilation stages for Ring Kernel PTX generation.
- GpuBridgeDirection
Direction of data transfer for a GPU ring buffer bridge.
- KernelState
Kernel health state enumeration.