org.nd4j.linalg.api.ops.aggregates.impl

Class AggregateGEMM

java.lang.Object

org.nd4j.linalg.api.ops.aggregates.BaseAggregate

org.nd4j.linalg.api.ops.aggregates.impl.AggregateGEMM

All Implemented Interfaces:

Aggregate

public class AggregateGEMM
extends BaseAggregate

Field Summary

Fields inherited from class org.nd4j.linalg.api.ops.aggregates.BaseAggregate

arguments, finalResult, indexingArguments, intArrayArguments, realArguments, shapes

Constructor Summary

Constructors

Constructor and Description
AggregateGEMM()
AggregateGEMM(int Order, int TransA, int TransB, int M, int N, int K, double alpha, @NonNull INDArray A, int lda, @NonNull INDArray B, int ldb, double beta, @NonNull INDArray C, int ldc)

Method Summary

Modifier and Type	Method and Description
int	getSharedMemorySize() This method returns amount of shared memory required for this specific Aggregate.
int	getThreadsPerInstance() This method returns desired number of threads per Aggregate instance PLEASE NOTE: this method is especially important for CUDA backend.
int	maxArguments() This method returns maximum number of shapes being passed per Aggregate
int	maxIndexArguments() This method returns maximum number of IndexArguments per Aggregate
int	maxIntArrays() This method returns maximum number of IntArrays being passed per Aggregate
int	maxIntArraySize() This method returns maximum length for IntArrays, if any
int	maxRealArguments() This method returns maximum number of real (float/double) per Aggregate
int	maxShapes() This method returns maximum number of shapes being passed per Aggregate
String	name()
int	opNum()

Methods inherited from class org.nd4j.linalg.api.ops.aggregates.BaseAggregate

getArguments, getFinalResult, getIndexingArguments, getIntArrayArguments, getRealArguments, getRequiredBatchMemorySize, getShapes, setFinalResult

Methods inherited from class java.lang.Object

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

Constructor Detail

AggregateGEMM

public AggregateGEMM()

AggregateGEMM

public AggregateGEMM(int Order,
                     int TransA,
                     int TransB,
                     int M,
                     int N,
                     int K,
                     double alpha,
                     @NonNull
                     @NonNull INDArray A,
                     int lda,
                     @NonNull
                     @NonNull INDArray B,
                     int ldb,
                     double beta,
                     @NonNull
                     @NonNull INDArray C,
                     int ldc)

Method Detail

name

public String name()

Returns:

opNum

public int opNum()

Returns:

maxArguments

public int maxArguments()

Description copied from interface: Aggregate

This method returns maximum number of shapes being passed per Aggregate

Returns:

maxShapes

public int maxShapes()

Description copied from interface: Aggregate

This method returns maximum number of shapes being passed per Aggregate

Returns:

maxIntArrays

public int maxIntArrays()

Description copied from interface: Aggregate

This method returns maximum number of IntArrays being passed per Aggregate

Returns:

maxIntArraySize

public int maxIntArraySize()

Description copied from interface: Aggregate

This method returns maximum length for IntArrays, if any

Returns:

maxIndexArguments

public int maxIndexArguments()

Description copied from interface: Aggregate

This method returns maximum number of IndexArguments per Aggregate

Returns:

maxRealArguments

public int maxRealArguments()

Description copied from interface: Aggregate

This method returns maximum number of real (float/double) per Aggregate

Returns:

getSharedMemorySize

public int getSharedMemorySize()

Description copied from interface: Aggregate

This method returns amount of shared memory required for this specific Aggregate. PLEASE NOTE: this method is especially important for CUDA backend. On CPU backend it might be ignored, depending on Aggregate.

Returns:

getThreadsPerInstance

public int getThreadsPerInstance()

Description copied from interface: Aggregate

This method returns desired number of threads per Aggregate instance PLEASE NOTE: this method is especially important for CUDA backend. On CPU backend it might be ignored, depending on Aggregate.

Returns: