cavis/libnd4j/include/loops/reduce_same.h

204 lines
7.7 KiB
C++

/*******************************************************************************
* Copyright (c) 2015-2018 Skymind, Inc.
*
* This program and the accompanying materials are made available under the
* terms of the Apache License, Version 2.0 which is available at
* https://www.apache.org/licenses/LICENSE-2.0.
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS, WITHOUT
* WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the
* License for the specific language governing permissions and limitations
* under the License.
*
* SPDX-License-Identifier: Apache-2.0
******************************************************************************/
#ifndef REDUCE_SAME_H
#define REDUCE_SAME_H
#include <dll.h>
//#include <string>
#include <stdio.h>
#include <helpers/shape.h>
#ifdef _OPENMP
#include <omp.h>
#endif
#include <templatemath.h>
#include <nd4jmalloc.h>
#include <pairwise_util.h>
#include <ops/ops.h>
#include <ops/special_accumulation_ops.h>
#include <op_boilerplate.h>
#pragma once
#ifdef __CUDACC__
#include <cuda.h>
#include <cuda_runtime.h>
#endif
#ifndef _OPENMP
#define omp_get_thread_num() 0
#define omp_get_max_threads() 1
#endif
#include "legacy_ops.h"
//an op for the kernel
namespace functions {
namespace reduce {
/**
* A reduce function
* reduces a vector down to
* a subset of itself
* via aggregating member
* elements.
*/
template<typename X>
class ReduceSameFunction {
public:
#ifdef __CUDACC__
template<typename OpType>
static __device__ void aggregatePartials(void *sPartials, Nd4jLong tid, Nd4jLong numItems, void *extraParams);
template<typename OpType>
static __device__ void execScalarCuda( void *vx, Nd4jLong *xShapeInfo, void *extraParams, void *vz, Nd4jLong *zShapeInfo, void *reductionBuffer, Nd4jLong *tadOnlyShapeInfo);
static __device__ void execScalarCudaLegacy(int opNum, void *vx, Nd4jLong *xShapeInfo, void *extraParams, void *vz, Nd4jLong *zShapeInfo, void *reductionBuffer, Nd4jLong *tadOnlyShapeInfo);
template<typename OpType>
static __device__ void transformCudaXD( void *vx, Nd4jLong *xShapeInfo, void *extraParams, void *vz, Nd4jLong *zShapeInfo, int *dimension, int dimensionLength, void *reductionBuffer, Nd4jLong *tadOnlyShapeInfo, Nd4jLong *tadOffsets);
template<typename OpType>
static __host__ void intermediateScalar(dim3 launchDims, cudaStream_t *stream, void *vx, Nd4jLong *xShapeInfo, Nd4jLong *hXShapeInfo, void *extraParams, void *vz, Nd4jLong *zShapeInfo, Nd4jLong *hZShapeInfo, int *dimension, int dimensionLength, void *reductionBuffer, Nd4jLong *tadOnlyShapeInfo);
template<typename OpType>
static __host__ void intermediateXD(dim3 launchDims, cudaStream_t *stream, void *vx, Nd4jLong *xShapeInfo, Nd4jLong *hXShapeInfo, void *extraParams, void *vz, Nd4jLong *zShapeInfo, Nd4jLong *hZShapeInfo, int *dimension, int dimensionLength, void *reductionPointer, Nd4jLong *tadShapeInfo, Nd4jLong *tadOffsets);
static __host__ void execReduceScalar(dim3 launchDims, cudaStream_t *stream, int opNum, void *vx, Nd4jLong *xShapeInfo, Nd4jLong* hXShapeInfo, void *extraParams, void *vz, Nd4jLong *zShapeInfo, Nd4jLong* hZShapeInfo, int *dimension, int dimensionLength, void *reductionBuffer, Nd4jLong *tadOnlyShapeInfo);
static __host__ void execReduceXD(dim3 launchDims, cudaStream_t *stream, int opNum, int rank, void *vx, Nd4jLong *xShapeInfo, Nd4jLong* hXShapeInfo, void *extraParams, void *vz, Nd4jLong *zShapeInfo, Nd4jLong* hZShapeInfo, int *dimension, int dimensionLength, void *reductionPointer, Nd4jLong *tadShapeInfo, Nd4jLong *tadOffsets);
#endif
/**
* Reduce down to 1 number
* @param x the input
* @param xShapeInfo the shape information
* for the input
* @param extraParams the extra params
* @return
*/
template<typename OpType>
static _CUDA_H X execScalar(void *x,
Nd4jLong *xShapeInfo,
void *extraParams);
template<typename OpType>
static _CUDA_H void execScalar(void *x,
Nd4jLong *xShapeInfo,
void *extraParams,
void *z,
Nd4jLong *zShapeInfo);
static X execScalar(int opNum,
void *x,
Nd4jLong *xShapeInfo,
void *extraParams);
static void execScalar(int opNum,
void *x,
Nd4jLong *xShapeInfo,
void *extraParams,
void *z,
Nd4jLong *zShapeInfo);
static void exec(int opNum,
void *x,
Nd4jLong *xShapeInfo,
void *extraParams,
void *result,
Nd4jLong *resultShapeInfoBuffer,
int *dimension,
int dimensionLength,
Nd4jLong *tadShapeInfo,
Nd4jLong *tadOffset);
/**
* Execute on the cpu
* @param x the input data
* @param xShapeInfo the shape information for x
* @param extraParams the extra parameters
* @param result the result buffer
* @param resultShapeInfoBuffer the shape information
* @param dimension the dimension to perform
* the reduce along long
* @param dimensionLength the length of the dimension buffer
*/
template<typename OpType>
static void _CUDA_H exec(void *x,
Nd4jLong *xShapeInfo,
void *extraParams,
void *result,
Nd4jLong *resultShapeInfoBuffer,
int *dimension,
int dimensionLength,
Nd4jLong *tadShapeInfo,
Nd4jLong *tadOffset);
/**
* CPU implementation
* @param x the input data
* @param xShapeInfo the shape information for
* the input data
* @param extraParams the extra parameters for the problem
* @param result the result buffer
* @param resultShapeInfo the shape information
*/
template<typename OpType>
static void _CUDA_H exec(void *x,
Nd4jLong *xShapeInfo,
void *extraParams,
void *result,
Nd4jLong *resultShapeInfo);
/**
* Reduce down to 1 number
* @param x the input
* @param xShapeInfo the shape information
* for the input
* @param extraParams the extra params
* @return
*/
template<typename OpType>
static X _CUDA_H execScalar(void *x,
Nd4jLong xElementWiseStride,
Nd4jLong length,
void *extraParams);
};
#ifdef __CUDACC__
/**
*
* @param extraParams
* @param sPartials
* @param sMemSize
*/
template<typename T>
__device__ void initializeShared(T *extraParams, T **sPartials, int sMemSize);
#endif
}
}
#endif