Table of Contents

API Reference

Welcome to the Orleans.GpuBridge.Core API reference documentation.

Namespaces Overview

Orleans.GpuBridge.Abstractions

Core interfaces and contracts for GPU acceleration in Orleans.

  • IGpuBridge - Main bridge interface for GPU operations
  • IGpuKernel - Kernel execution contract
  • IGpuAccelerated - Marker interface for GPU-capable grains
  • GpuBridgeOptions - Configuration options

Orleans.GpuBridge.Abstractions.Clocks

Temporal correctness primitives.

  • IHybridLogicalClock - HLC timestamp generation
  • IVectorClock - Vector clock for causal ordering
  • HybridTimestamp - Immutable HLC timestamp value
  • VectorTimestamp - Immutable vector clock value

Orleans.GpuBridge.Abstractions.Placement

GPU-aware grain placement strategies.

  • IGpuPlacementDirector - Placement decision interface
  • GpuPlacementStrategy - Placement strategy base class
  • QueueDepthAwarePlacement - Ring kernel queue-aware placement

Orleans.GpuBridge.Runtime

Runtime implementation and DI integration.

  • KernelCatalog - Kernel registration and resolution
  • DeviceBroker - GPU device management
  • ServiceCollectionExtensions - AddGpuBridge() extension

Orleans.GpuBridge.Runtime.Clocks

Clock implementation and synchronization.

  • HybridLogicalClock - HLC implementation
  • VectorClock - Vector clock implementation
  • ClockSourceSelector - Clock source selection
  • SoftwarePtpClockSource - Software PTP synchronization

Orleans.GpuBridge.Runtime.Resident

GPU-resident actor support.

  • IGpuResidentManager - GPU memory management
  • GpuResidentHandle - Handle to GPU-resident state
  • RingKernelRuntime - Ring kernel lifecycle management

Orleans.GpuBridge.Grains

Orleans grain implementations.

  • GpuBatchGrain - Batch processing grain
  • GpuResidentGrain - GPU-resident data grain
  • GpuStreamGrain - Stream processing grain
  • HypergraphVertexGrain - Vertex actor implementation
  • HypergraphHyperedgeGrain - Hyperedge actor implementation

Orleans.GpuBridge.BridgeFX

High-level pipeline API.

  • GpuPipeline - Fluent API for batch processing
  • PipelineBuilder - Pipeline configuration builder
  • PipelineResult - Execution result container

Orleans.GpuBridge.Backends.DotCompute

DotCompute GPU backend integration.

  • DotComputeBackendAdapter - Backend adapter implementation
  • DotComputeRingKernelRuntime - Ring kernel runtime for DotCompute
  • KernelAttributeMapper - DotCompute attribute mapping

Quick Reference

Service Registration

services.AddGpuBridge(options =>
{
    options.PreferGpu = true;
    options.EnableFallbackToCpu = true;
    options.GpuFramework = GpuFramework.DotCompute;
});

Kernel Registration

services.AddGpuBridge()
    .AddKernel(k => k
        .Id("matrix-multiply")
        .In<float[]>()
        .Out<float[]>()
        .FromFactory(sp => new MatrixKernel()));

Pipeline Execution

var results = await GpuPipeline<float[], float[]>
    .For(grainFactory, "matrix-multiply")
    .WithBatchSize(1000)
    .ExecuteAsync(data);

Version Information

  • Current Version: 0.1.0
  • Target Framework: .NET 9.0
  • GPU Framework: DotCompute 0.5.1

See individual namespace documentation for detailed API information.

Edit this page