KernelRamExamples.ExampleRamKernelAdd Class Reference

Example implementation of a RAM‑based compute kernel that adds two integers. More...

Inheritance diagram for KernelRamExamples.ExampleRamKernelAdd:

Collaboration diagram for KernelRamExamples.ExampleRamKernelAdd:

Public Member Functions
	ExampleRamKernelAdd ()
	Initializes a new instance of the example kernel.
Public Member Functions inherited from SystemEx.Device.NativeRAMKernel< TD >
	NativeRAMKernel ()
	Initializes a new RAM kernel instance with an empty buffer set and a default RamSharedBackend backend. The native module is not loaded until Create is called.
bool	Create (string dllPath)
	Loads the native module located at dllPath and performs backend‑specific initialization via OnCreate. Must be called before buffers are added or the kernel is executed.
int	AddBuffer (DeviceBuffer buffer, string name, int flags, object? configs)
	Adds a shared buffer to the kernel and converts the provided DeviceBuffer into a DeviceSharedBuffer<RamSharedBackend> using the RAM backend. Derived classes may override OnAddBuffer to perform validation or backend‑specific configuration.
DeviceSharedBuffer< RamSharedBackend >?	GetBuffer (string name)
	Retrieves a shared buffer by its symbolic name.
bool	RemoveBuffer (string name)
	Removes a buffer by name and buffer type.
bool	RemoveBuffer (int index)
	Removes a buffer by index and buffer type.
virtual bool	BeginRun (string strFunction)
	Begins a kernel execution session by loading the native function specified by strFunction and locking all shared buffers. Derived classes may override OnBegin to perform additional backend‑specific setup.
virtual void	EndRun ()
	Ends the kernel execution session, unlocks all shared buffers, and performs backend cleanup via OnEnd.
bool	IsRunning ()
	Indicates whether the kernel is currently running.
unsafe bool	Run (object? options)
	Executes the kernel asynchronously using the delegate loaded during BeginRun. The actual kernel logic is implemented in OnRun.

Protected Member Functions
override bool	OnCreate ()
	Creates and initializes the buffers required for this kernel.
override bool	OnRun (MyKernelFunction function)
	Executes the native addition kernel.
Protected Member Functions inherited from SystemEx.Device.NativeRAMKernel< TD >
virtual bool	OnBegin (string strFunction)
	Called when a kernel execution session begins. Loads the native function delegate using NativeHost and returns true if the function is successfully resolved.
virtual void	OnEnd ()
	Called when a kernel execution session ends. Derived classes may override this method to perform backend‑specific cleanup.
virtual bool	OnAddBuffer (string name, DeviceSharedBuffer< RamSharedBackend > buffer)
	Called when a buffer is added to the kernel. Derived classes may override this method to validate buffer names or apply backend‑specific configuration.
virtual bool	OnRun (TD function)
	Executes the native kernel function. Derived classes must override this method to implement backend‑specific compute logic using the resolved delegate function .

Additional Inherited Members
Properties inherited from SystemEx.Device.NativeRAMKernel< TD >
RamSharedBackend	Backend [get]
	Gets the backend instance used for RAM‑based shared buffer operations.
Module?	Module [get, protected set]
	Gets the native module loaded for this kernel. The module represents the underlying shared library (DLL, SO, DYLIB) from which the kernel function delegate is resolved. This property is assigned during Create and remains available for the entire lifetime of the kernel instance.
TD?	Function [get, protected set]
	Gets the resolved native kernel function delegate. The delegate is created during BeginRun using the function name provided by the caller and represents the entry point of the native compute routine. If the function cannot be resolved, this property is null.
bool	Running [get]
	Indicates whether the kernel is currently running.

Detailed Description

Example implementation of a RAM‑based compute kernel that adds two integers.

This class demonstrates how to use NativeRAMKernel<T>:

• Buffers are created and registered in OnCreate.
• The native function is executed in OnRun.
• Unmanaged pointers are passed directly to the native DLL.

The example is intentionally simple: it adds two integer values stored in device buffers A and B, and writes the result into C.

Constructor & Destructor Documentation

ExampleRamKernelAdd()

KernelRamExamples.ExampleRamKernelAdd.ExampleRamKernelAdd

(

)

Initializes a new instance of the example kernel.

The native DLL is loaded later via IKernel<T>.Create.

: base() {  }

Member Function Documentation

OnCreate()

override bool KernelRamExamples.ExampleRamKernelAdd.OnCreate ( )

protectedvirtual

Creates and initializes the buffers required for this kernel.

In this example, only buffer B is created here. Buffers A and C are created in Main().

Steps:

1. Allocate a device buffer for integer B.
2. Write the initial value (54) into the buffer.
3. Register the buffer with the kernel.

Returns

true if buffer creation succeeds.

Reimplemented from SystemEx.Device.NativeRAMKernel< TD >.

                                            {
            DeviceBuffer B = new DeviceBuffer(sizeof(int), CacheType.ToDevice);
 
            B.Write(0, 54U, Endian.System);
 
            AddBuffer(B, "BufferB", 0, null);
 
            return base.OnCreate();
        }

Here is the call graph for this function:

OnRun()

override bool KernelRamExamples.ExampleRamKernelAdd.OnRun ( MyKernelFunction function )

protected

Executes the native addition kernel.

The method:

• Retrieves buffers A, B, and C
• Extracts their unmanaged memory handles
• Invokes the native function through the delegate
• Writes the result into buffer C

Parameters

function

The resolved native kernel delegate. If null, the kernel cannot be executed.

Returns

true if execution succeeds.

                                                                    {
 
            
            DeviceSharedBuffer < RamSharedBackend >? A = base.GetBuffer("BufferA")!;
            DeviceSharedBuffer < RamSharedBackend >? B = base.GetBuffer("BufferB")!;
            DeviceSharedBuffer < RamSharedBackend >? C = base.GetBuffer("BufferC")!;
 
            var unmanagedA = (UnmanagedObject)A.HardwareBuffer!;
            var unmanagedB = (UnmanagedObject)B.HardwareBuffer!;
            var unmanagedC = (UnmanagedObject)C.HardwareBuffer!;
 
            unmanagedC.Size = function(unmanagedA.Point, unmanagedA.Size, unmanagedB.Point, unmanagedB.Size, unmanagedC.Point);
 
            return true;
        }

---

The documentation for this class was generated from the following file:

• ExampleRamKernelAdd.cs