CpuResidualOp.ixx File Reference

CPU implementation of the residual (y = x + F(x)) binary operation. More...

#include <memory>
#include <vector>
#include <string>
#include <iostream>
#include <stdexcept>
#include <math.h>
import Dnn.ITensor;
import Dnn.Tensor;
import Dnn.Components.ResidualConfig;
import Compute.ExecutionContext;
import Compute.CpuDevice;
import Compute.CpuMemoryResource;
import Compute.DeviceType;
import Compute.MemoryResource;
import Dnn.TensorDataType;
import Dnn.ComponentConfig;
import Compute.IExecutionContext;
import Compute.OperationRegistry;
import Compute.OperationType;
import Compute.OperationBase;
import Compute.BinaryOperation;

Classes
class	Mila::Dnn::Compute::CpuResidualOp
	CPU Residual operation (FP32) implementing BinaryOperation interface. More...
class	Mila::Dnn::Compute::CpuResidualOpRegistrar
	Registrar for CPU Residual operation (FP32). More...

Namespaces
namespace	Mila
	Mila main API namespace.
namespace	Mila::Dnn
namespace	Mila::Dnn::Compute

Macros
#define	_USE_MATH_DEFINES

Detailed Description

CPU implementation of the residual (y = x + F(x)) binary operation.

Provides a CPU device implementation of the Residual operation using the device-agnostic BinaryOperation interface. This implementation uses the abstract TensorDataType::FP32 precision and ITensor interfaces for inputs and outputs. Registration uses the canonical operation name "ResidualOp".

Notes:

• The class accepts an optional CPU execution context; if provided it will be validated to ensure it's for a CPU device.
• Forward/backward are implemented in terms of raw host buffers obtained from the ITensor abstraction. OpenMP is used when available.

Macro Definition Documentation

_USE_MATH_DEFINES

#define _USE_MATH_DEFINES