Class OpenCLKernelPipeline

Namespace

DotCompute.Backends.OpenCL.Execution

Assembly

DotCompute.Backends.OpenCL.dll

Production-grade multi-kernel execution pipeline with dependency resolution, automatic optimization, and comprehensive profiling.

public sealed class OpenCLKernelPipeline : IAsyncDisposable

Inheritance

object

OpenCLKernelPipeline

Implements

IAsyncDisposable

Inherited Members

object.Equals(object)

object.Equals(object, object)

object.GetHashCode()

object.GetType()

object.ReferenceEquals(object, object)

object.ToString()

Remarks

The OpenCLKernelPipeline enables complex multi-stage compute workflows with:

• Automatic dependency resolution via topological sorting
• Parallel execution of independent stages
• Intermediate buffer management and reuse
• Event-based synchronization for optimal throughput
• Pipeline-level profiling with stage breakdown
• Cycle detection to prevent invalid pipelines
• Automatic kernel fusion opportunities identification
• Memory reuse optimization across stages

Example use cases: - Image processing pipelines (blur → sharpen → color correct) - Machine learning inference (preprocess → inference → postprocess) - Scientific simulations (initialize → iterate → analyze) - Data processing workflows (load → transform → reduce)

Constructors

OpenCLKernelPipeline(OpenCLKernelExecutionEngine, OpenCLMemoryPoolManager, OpenCLProfiler, ILogger<OpenCLKernelPipeline>)

Initializes a new instance of the OpenCLKernelPipeline class.

public OpenCLKernelPipeline(OpenCLKernelExecutionEngine executor, OpenCLMemoryPoolManager memoryPool, OpenCLProfiler profiler, ILogger<OpenCLKernelPipeline> logger)

Parameters

executor OpenCLKernelExecutionEngine

Kernel execution engine for running stages.

memoryPool OpenCLMemoryPoolManager

Memory pool manager for intermediate buffers.

profiler OpenCLProfiler

Profiler for performance metrics.

logger ILogger<OpenCLKernelPipeline>

Logger for diagnostic information.

Exceptions

ArgumentNullException

Thrown if any parameter is null.

Methods

CreatePipeline(string)

Creates a new pipeline builder for constructing multi-kernel workflows.

public OpenCLKernelPipeline.PipelineBuilder CreatePipeline(string name)

Parameters

name string

Descriptive name for the pipeline.

Returns

OpenCLKernelPipeline.PipelineBuilder

A pipeline builder for configuring stages and dependencies.

Exceptions

ArgumentException

Thrown if name is null or whitespace.

DisposeAsync()

Asynchronously disposes the pipeline manager and releases all resources.

public ValueTask DisposeAsync()

Returns

ValueTask

ExecutePipelineAsync(Pipeline, Dictionary<string, object>, CancellationToken)

Executes a configured pipeline with automatic dependency resolution and optimization.

public Task<OpenCLKernelPipeline.PipelineResult> ExecutePipelineAsync(OpenCLKernelPipeline.Pipeline pipeline, Dictionary<string, object> inputs, CancellationToken cancellationToken = default)

Parameters

pipeline OpenCLKernelPipeline.Pipeline

The pipeline to execute.

inputs Dictionary<string, object>

Input buffers and parameters for the pipeline.

cancellationToken CancellationToken

Cancellation token for async operations.

Returns

Task<OpenCLKernelPipeline.PipelineResult>

Pipeline execution result containing outputs and profiling data.

Exceptions

ArgumentNullException

Thrown if pipeline or inputs is null.

OpenCLPipelineException

Thrown if pipeline validation or execution fails.

GetStatistics()

Gets comprehensive pipeline execution statistics.

public OpenCLKernelPipeline.PipelineStatistics GetStatistics()

Returns

OpenCLKernelPipeline.PipelineStatistics

Statistics about pipeline usage and performance.