neureka.backend.main.operations.other.internal

Class CPUReduce

java.lang.Object

neureka.backend.main.operations.other.internal.CPUReduce

All Implemented Interfaces:

ImplementationFor<CPU>

public class CPUReduce
extends java.lang.Object
implements ImplementationFor<CPU>

An implementation of the min and max algorithm running on the CPU. This algorithm splits the provided input tensor into chucks which are then reduced to local min and max values. This happens iteratively until only a single value is left. Each workload also returns the index of the found min/max value, which is important for backpropagation...

Nested Class Summary

Nested Classes

Modifier and Type	Class and Description
static class	CPUReduce.Type

Constructor Summary

Constructors

Constructor and Description
CPUReduce(CPUReduce.Type type)

Method Summary

Modifier and Type	Method and Description
Tensor<java.lang.Integer>	run(ExecutionCall<CPU> call) This method is the entrypoint for a concrete implementation of the algorithm to which instances of this interface belong and the device on which this is implemented.

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Constructor Detail

CPUReduce

public CPUReduce(CPUReduce.Type type)

Method Detail

run

public Tensor<java.lang.Integer> run(ExecutionCall<CPU> call)

Description copied from interface: ImplementationFor

This method is the entrypoint for a concrete implementation of the algorithm to which instances of this interface belong and the device on which this is implemented. One has to keep in mind that the implementation details with respect to the target device are specific to the methods of the "TargetDevice" type on which this call should run...

Specified by:

run in interface ImplementationFor<CPU>

Parameters:

call - The call which ought to be executed on this implementation.