/*******************************************************************************
* 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
******************************************************************************/
//
// @author Yurii Shyrma (iuriish@yahoo.com), created on 14.03.2019
//
#ifndef LIBND4J_LOOPS_H
#define LIBND4J_LOOPS_H
#include <functional>
#include <pointercast.h>
#include <shape.h>
#include <LoopKind.h>
#include <OmpLaunchHelper.h>
#include <DataTypeUtils.h>
#include <ops.h>
#include <indexreduce.h>
#include <helpers/ConstantTadHelper.h>
#include <openmp_pragmas.h>
namespace nd4j {
template <typename X, typename Z, typename E>
class ND4J_EXPORT ReductionLoops {
protected:
public:
template <typename OpType>
static FORCEINLINE void loopReduce(X* x, Nd4jLong* xShapeInfo, Z* z, Nd4jLong* zShapeInfo, Nd4jLong* tadShapeInfo, Nd4jLong* tadOffsets, E* extraParams);
};
template <typename X, typename Z>
class ReductionFloatLoops : public ReductionLoops<X,Z,Z> {
public:
static void wrapper(const int opNum, X* x, Nd4jLong* xShapeInfo, Z* z, Nd4jLong* zShapeInfo, Nd4jLong* tadShapeInfo, Nd4jLong* tadOffsets, Z* extraParams);
template <typename OpType>
static void innerloopReduce(X* x, Nd4jLong* xShapeInfo, Z* z, Nd4jLong* zShapeInfo, Nd4jLong* tadShapeInfo, Nd4jLong* tadOffsets, Z* extraParams);
};
template <typename X, typename Z>
class ND4J_EXPORT ReductionBoolLoops : public ReductionLoops<X,Z,X> {
public:
static void wrapper(const int opNum, X* x, Nd4jLong* xShapeInfo, Z* z, Nd4jLong* zShapeInfo, Nd4jLong* tadShapeInfo, Nd4jLong* tadOffsets, X* extraParams);
template <typename OpType>
static void innerloopReduce(X* x, Nd4jLong* xShapeInfo, Z* z, Nd4jLong* zShapeInfo, Nd4jLong* tadShapeInfo, Nd4jLong* tadOffsets, X* extraParams);
};
template <typename X, typename Z>
class ND4J_EXPORT ReductionLongLoops : public ReductionLoops<X,Z,X> {
public:
static void wrapper(const int opNum, X* x, Nd4jLong* xShapeInfo, Z* z, Nd4jLong* zShapeInfo, Nd4jLong* tadShapeInfo, Nd4jLong* tadOffsets, X* extraParams);
template <typename OpType>
static void innerloopReduce(X* x, Nd4jLong* xShapeInfo, Z* z, Nd4jLong* zShapeInfo, Nd4jLong* tadShapeInfo, Nd4jLong* tadOffsets, X* extraParams);
};
template <typename X>
class ND4J_EXPORT ReductionSameLoops : public ReductionLoops<X,X,X> {
public:
static void wrapper(const int opNum, X* x, Nd4jLong* xShapeInfo, X* z, Nd4jLong* zShapeInfo, Nd4jLong* tadShapeInfo, Nd4jLong* tadOffsets, X* extraParams);
template <typename OpType>
static void innerloopReduce(X* x, Nd4jLong* xShapeInfo, X* z, Nd4jLong* zShapeInfo, Nd4jLong* tadShapeInfo, Nd4jLong* tadOffsets, X* extraParams);
};
template <typename X, typename Z>
class ND4J_EXPORT IndexReductionLoops {
private:
public:
static void wrapIndexReduce(const int opNum, void* x, Nd4jLong* xShapeInfo, void* z, Nd4jLong* zShapeInfo, Nd4jLong* tadShapeInfo, Nd4jLong* tadOffsets, void* extraParams);
template <typename OpType>
static void loopIndexReduce(X* x, Nd4jLong* xShapeInfo, Z* z, Nd4jLong* zShapeInfo, Nd4jLong* tadShapeInfo, Nd4jLong* tadOffsets, X* extraParams);
};
template <typename X, typename Z, typename E>
class ND4J_EXPORT TransformLoops {
public:
template<typename OpType, bool doParallel>
static FORCEINLINE void loopTransform(X* x, Nd4jLong* xShapeInfo, Z* z, Nd4jLong* zShapeInfo, E* extraParams);
};
template <typename X, typename Z>
class ND4J_EXPORT Reduction3Loops {
public:
template <typename OpType>
static FORCEINLINE void loopReduce3(X* x, Nd4jLong* xShapeInfo, X* y, Nd4jLong* yShapeInfo, Z* z, Nd4jLong* zShapeInfo, int* dims, int dimsLen, Z* extraParams);
template <typename OpType>
static FORCEINLINE void loopReduce3All(X* x, Nd4jLong* xShapeInfo, X* y, Nd4jLong* yShapeInfo, Z* z, Nd4jLong* zShapeInfo, Nd4jLong* xTadShapeInfo, Nd4jLong* xTadOffsets, Nd4jLong* yTadShapeInfo, Nd4jLong* yTadOffsets, Z* extraParams);
static void wrapper(const int opNum, X* x, Nd4jLong* xShapeInfo, X* y, Nd4jLong* yShapeInfo, Z* z, Nd4jLong* zShapeInfo, int* dims, int dimsLen, Z* extraParams);
static void wrapperAll(const int opNum, X* x, Nd4jLong* xShapeInfo, X* y, Nd4jLong* yShapeInfo, Z* z, Nd4jLong* zShapeInfo, Nd4jLong* xTadShapeInfo, Nd4jLong* xTadOffsets, Nd4jLong* yTadShapeInfo, Nd4jLong* yTadOffsets, Z* extraParams);
template <typename OpType>
static void innerloopReduce3(X* x, Nd4jLong* xShapeInfo, X* y, Nd4jLong* yShapeInfo, Z* z, Nd4jLong* zShapeInfo, int* dims, int dimsLen, Z* extraParams);
template <typename OpType>
static void innerloopReduce3All(X* x, Nd4jLong* xShapeInfo, X* y, Nd4jLong* yShapeInfo, Z* z, Nd4jLong* zShapeInfo, Nd4jLong* xTadShapeInfo, Nd4jLong* xTadOffsets, Nd4jLong* yTadShapeInfo, Nd4jLong* yTadOffsets, Z* extraParams);
};
/*
//////////////////////////////////////////////////////////////////////////////
template<typename X, typename Y, typename Z>
void Loops::loopXYZ(const X* x, const Nd4jLong* xShapeInfo,
const Y* y, const Nd4jLong* yShapeInfo,
Z* z, const Nd4jLong* zShapeInfo,
Z* extraParams,
std::function<Z(X,Y,Z*)> op) {
const LoopKind::Kind kindOfLoop = LoopKind::deduceKindOfLoopXYZ(xShapeInfo, yShapeInfo, zShapeInfo);
const Nd4jLong* xShape = shape::shapeOf(xShapeInfo);
const Nd4jLong* xStride = shape::stride(xShapeInfo);
const Nd4jLong* yStride = shape::stride(yShapeInfo);
const Nd4jLong* zStride = shape::stride(zShapeInfo);
const Nd4jLong len = shape::length(xShapeInfo);
OmpLaunchHelper threadsInfo(len);
switch (kindOfLoop) {
case LoopKind::EWS1: {
PRAGMA_OMP_PARALLEL_THREADS(threadsInfo._numThreads)
{
const auto threadNum = omp_get_thread_num();
const auto threadOffset = threadsInfo.getThreadOffset(threadNum);
const auto lenPerThread = static_cast<uint>(threadsInfo.getItersPerThread(threadNum));
const auto xi = x + threadOffset;
const auto yi = y + threadOffset;
auto zi = z + threadOffset;
PRAGMA_OMP_SIMD
for (uint i = 0; i < lenPerThread; i++)
zi[i] = op(xi[i], yi[i], extraParams);
}
}
break;
case LoopKind::EWSNONZERO: {
const uint xEws = shape::elementWiseStride(xShapeInfo);
const uint yEws = shape::elementWiseStride(yShapeInfo);
const uint zEws = shape::elementWiseStride(zShapeInfo);
PRAGMA_OMP_PARALLEL_THREADS(threadsInfo._numThreads)
{
const auto threadNum = omp_get_thread_num();
const auto threadOffset = threadsInfo.getThreadOffset(threadNum);
const auto lenPerThread = static_cast<uint>(threadsInfo.getItersPerThread(threadNum));
const auto xi = x + threadOffset * xEws;
const auto yi = y + threadOffset * yEws;
auto zi = z + threadOffset * zEws;
PRAGMA_OMP_SIMD
for (uint i = 0; i < lenPerThread; i++)
zi[i*zEws] = op(xi[i*xEws], yi[i*yEws], extraParams);
}
}
break;
case LoopKind::RANK1: {
PRAGMA_OMP_PARALLEL_FOR
for (uint i0 = 0; i0 < len; ++i0)
z[i0 * zStride[0]] = op(x[i0 * xStride[0]], y[i0 * yStride[0]], extraParams);
}
break;
case LoopKind::RANK2: {
PRAGMA_OMP_PARALLEL_FOR_SIMD
for (uint i0 = 0; i0 < xShape[0]; ++i0)
for (uint i1 = 0; i1 < xShape[1]; ++i1)
z[i0 * zStride[0] + i1 * zStride[1]] = op(x[i0 * xStride[0] + i1 * xStride[1]], y[i0 * yStride[0] + i1 * yStride[1]], extraParams);
}
break;
case LoopKind::RANK3: {
PRAGMA_OMP_PARALLEL_FOR_SIMD_COLLAPSE(2)
for (uint i0 = 0; i0 < xShape[0]; ++i0)
for (uint i1 = 0; i1 < xShape[1]; ++i1)
for (uint i2 = 0; i2 < xShape[2]; ++i2)
z[i0*zStride[0]+i1*zStride[1]+i2*zStride[2]] = op(x[i0*xStride[0]+i1*xStride[1]+i2*xStride[2]], y[i0*yStride[0]+i1*yStride[1]+i2*yStride[2]], extraParams);
}
break;
case LoopKind::RANK4: {
PRAGMA_OMP_PARALLEL_FOR_SIMD_COLLAPSE(3)
for (uint i0 = 0; i0 < xShape[0]; ++i0)
for (uint i1 = 0; i1 < xShape[1]; ++i1)
for (uint i2 = 0; i2 < xShape[2]; ++i2)
for (uint i3 = 0; i3 < xShape[3]; ++i3)
z[i0*zStride[0]+i1*zStride[1]+i2*zStride[2]+i3*zStride[3]] = op(x[i0*xStride[0]+i1*xStride[1]+i2*xStride[2]+i3*xStride[3]], y[i0*yStride[0]+i1*yStride[1]+i2*yStride[2]+i3*yStride[3]], extraParams);
}
break;
case LoopKind::RANK5: {
PRAGMA_OMP_PARALLEL_FOR_SIMD_COLLAPSE(4)
for (uint i0 = 0; i0 < xShape[0]; ++i0)
for (uint i1 = 0; i1 < xShape[1]; ++i1)
for (uint i2 = 0; i2 < xShape[2]; ++i2)
for (uint i3 = 0; i3 < xShape[3]; ++i3)
for (uint i4 = 0; i4 < xShape[4]; ++i4)
z[i0*zStride[0]+i1*zStride[1]+i2*zStride[2]+i3*zStride[3]+i4*zStride[4]] = op(x[i0*xStride[0]+i1*xStride[1]+i2*xStride[2]+i3*xStride[3]+i4*xStride[4]], y[i0*yStride[0]+i1*yStride[1]+i2*yStride[2]+i3*yStride[3]+i4*yStride[4]], extraParams);
}
break;
default: {
uint xShapeInfoCast[MAX_RANK];
uint yShapeInfoCast[MAX_RANK];
uint zShapeInfoCast[MAX_RANK];
bool canCastX = DataTypeUtils::castShapeInfo(xShapeInfo, xShapeInfoCast);
bool canCastY = DataTypeUtils::castShapeInfo(yShapeInfo, yShapeInfoCast);
bool canCastZ = DataTypeUtils::castShapeInfo(zShapeInfo, zShapeInfoCast);
PRAGMA_OMP_PARALLEL_THREADS(threadsInfo._numThreads)
{
auto threadNum = omp_get_thread_num();
auto threadOffset = threadsInfo.getThreadOffset(threadNum);
auto lenPerThread = static_cast<uint>(threadsInfo.getItersPerThread(threadNum));
PRAGMA_OMP_SIMD
for (uint i = 0; i < lenPerThread; i++) {
auto xOffset = shape::indexOffset(i + threadOffset, xShapeInfo, xShapeInfoCast, len, canCastX);
auto yOffset = shape::indexOffset(i + threadOffset, yShapeInfo, yShapeInfoCast, len, canCastY);
auto zOffset = shape::indexOffset(i + threadOffset, zShapeInfo, zShapeInfoCast, len, canCastZ);
z[zOffset] = op(x[xOffset], y[yOffset], extraParams);
}
}
}
}
}
*/
//////////////////////////////////////////////////////////////////////////////
template<typename X, typename Z, typename E>
template <typename OpType>
void nd4j::ReductionLoops<X, Z, E>::loopReduce(X* x, Nd4jLong* xShapeInfo,
Z* z, Nd4jLong* zShapeInfo,
Nd4jLong* tadShapeInfo, Nd4jLong* tadOffsets,
E* extraParams) {
const LoopKind::Kind kindOfLoop = LoopKind::deduceKindOfLoopTadXZ(xShapeInfo, zShapeInfo, tadShapeInfo);
const Nd4jLong zLen = shape::length(zShapeInfo);
const Nd4jLong tadLen = shape::length(tadShapeInfo);
const uint tadEws = shape::elementWiseStride(tadShapeInfo);
const uint zEws = shape::elementWiseStride(zShapeInfo);
const Nd4jLong* tadShape = shape::shapeOf(tadShapeInfo);
const Nd4jLong* tadStride = shape::stride(tadShapeInfo);
int numThreads = OmpLaunchHelper::tadThreads(tadLen, zLen);
switch (kindOfLoop) {
//*********************************************//
// case LoopKind::SMALLARR2DX: {
// shape::printShapeInfoLinear(xShapeInfo);
// shape::printShapeInfoLinear(zShapeInfo);
// const auto xLen = zLen * tadLen;
// for (uint i = 0; i < xLen; ++i) {
// const auto zOffset = shape::subArrayOffset(i, xShapeInfo, zShapeInfo, dimsToExclude, dimsLen);
// const uint tadInd = (i / tadEws) % tadLen;
// auto startVal = tadInd ? z[zOffset] : static_cast<Z>(OpType::startingValue(x));
// z[zOffset] = OpType::update(startVal, OpType::op(x[i], extraParams), extraParams);
// if(tadInd == tadLen - 1)
// z[zOffset] = OpType::postProcess(z[zOffset], tadLen, extraParams);
// printf("%u - %lld\n", i, zOffset);
// }
// }
case LoopKind::SMALLARR2DX: {
const auto uTadLen = static_cast<uint>(tadLen);
const auto uZLenMinusOne = static_cast<uint>(zLen - 1);
const auto xLen = static_cast<uint>(zLen * uTadLen);
const auto sv = static_cast<Z>(OpType::startingValue(x));
for (uint i = 0; i <= uZLenMinusOne; i++)
z[i] = OpType::startingValue(x);
uint zOffset = 0;
for (uint i = 0; i < xLen; ++i) {
z[zOffset] = OpType::update(z[zOffset], OpType::op(x[i], extraParams), extraParams);
zOffset = zOffset == uZLenMinusOne ? 0 : zOffset + 1;
}
for (uint i = 0; i <= uZLenMinusOne; i++)
z[i] = OpType::postProcess(z[i], tadLen, extraParams);
}
break;
//*********************************************//
case LoopKind::EWS1: {
PRAGMA_OMP_PARALLEL_FOR_THREADS(numThreads)
for (uint i = 0; i < zLen; i++) {
auto tad = x + tadOffsets[i];
auto start = OpType::startingValue(tad);
for (uint j = 0; j < tadLen; j++)
start = OpType::update(start, OpType::op(tad[j], extraParams), extraParams);
z[i] = OpType::postProcess(start, tadLen, extraParams);
}
}
break;
//*********************************************//
case LoopKind::EWSNONZERO: {
PRAGMA_OMP_PARALLEL_FOR_THREADS(numThreads)
for (uint i = 0; i < zLen; i++) {
auto tad = x + tadOffsets[i];
auto start = OpType::startingValue(tad);
for (uint j = 0; j < tadLen; j++)
start = OpType::update(start, OpType::op(tad[j * tadEws], extraParams), extraParams);
z[i * zEws] = OpType::postProcess(start, tadLen, extraParams);
}
}
break;
//*********************************************//
case LoopKind::RANK1: {
PRAGMA_OMP_PARALLEL_FOR_THREADS(numThreads)
for (uint i = 0; i < zLen; i++) {
auto tad = x + tadOffsets[i];
auto start = OpType::startingValue(tad);
for (uint i0 = 0; i0 < tadLen; ++i0)
start = OpType::update(start, OpType::op(tad[i0 * tadStride[0]], extraParams), extraParams);
z[i] = OpType::postProcess(start, tadLen, extraParams);
}
}
break;
//*********************************************//
case LoopKind::RANK2: {
PRAGMA_OMP_PARALLEL_FOR_THREADS(numThreads)
for (uint i = 0; i < zLen; ++i) {
auto tad = x + tadOffsets[i];
auto start = OpType::startingValue(tad);
for (uint i0 = 0; i0 < tadShape[0]; ++i0)
for (uint i1 = 0; i1 < tadShape[1]; ++i1)
start = OpType::update(start, OpType::op(tad[i0*tadStride[0] + i1*tadStride[1]], extraParams), extraParams);
z[i] = OpType::postProcess(start, tadLen, extraParams);
}
}
break;
//*********************************************//
case LoopKind::RANK3: {
PRAGMA_OMP_PARALLEL_FOR_THREADS(numThreads)
for (uint i = 0; i < zLen; ++i) {
auto tad = x + tadOffsets[i];
auto start = OpType::startingValue(tad);
for (uint i0 = 0; i0 < tadShape[0]; ++i0)
for (uint i1 = 0; i1 < tadShape[1]; ++i1)
for (uint i2 = 0; i2 < tadShape[2]; ++i2)
start = OpType::update(start, OpType::op(tad[i0*tadStride[0] + i1*tadStride[1] + i2*tadStride[2]], extraParams), extraParams);
z[i] = OpType::postProcess(start, tadLen, extraParams);
}
}
break;
//*********************************************//
case LoopKind::RANK4: {
PRAGMA_OMP_PARALLEL_FOR_THREADS(numThreads)
for (uint i = 0; i < zLen; ++i) {
auto tad = x + tadOffsets[i];
auto start = OpType::startingValue(tad);
for (uint i0 = 0; i0 < tadShape[0]; ++i0)
for (uint i1 = 0; i1 < tadShape[1]; ++i1)
for (uint i2 = 0; i2 < tadShape[2]; ++i2)
for (uint i3 = 0; i3 < tadShape[3]; ++i3)
start = OpType::update(start, OpType::op(tad[i0*tadStride[0] + i1*tadStride[1] + i2*tadStride[2] + i3*tadStride[3]], extraParams), extraParams);
z[i] = OpType::postProcess(start, tadLen, extraParams);
}
}
break;
//*********************************************//
case LoopKind::RANK5: {
PRAGMA_OMP_PARALLEL_FOR_THREADS(numThreads)
for (uint i = 0; i < zLen; ++i) {
auto tad = x + tadOffsets[i];
auto start = OpType::startingValue(tad);
for (uint i0 = 0; i0 < tadShape[0]; ++i0)
for (uint i1 = 0; i1 < tadShape[1]; ++i1)
for (uint i2 = 0; i2 < tadShape[2]; ++i2)
for (uint i3 = 0; i3 < tadShape[3]; ++i3)
for (uint i4 = 0; i4 < tadShape[4]; ++i4)
start = OpType::update(start, OpType::op(tad[i0*tadStride[0] + i1*tadStride[1] + i2*tadStride[2] + i3*tadStride[3] + i4*tadStride[4] ], extraParams), extraParams);
z[i] = OpType::postProcess(start, tadLen, extraParams);
}
}
break;
//*********************************************//
case LoopKind::X_EWSNONZERO: {
uint castZShapeInfo[MAX_RANK];
const bool canCastZ = nd4j::DataTypeUtils::castShapeInfo<uint>(zShapeInfo, castZShapeInfo);
PRAGMA_OMP_PARALLEL_FOR_THREADS(numThreads)
for (uint i = 0; i < zLen; i++) {
auto tad = x + tadOffsets[i];
auto start = OpType::startingValue(tad);
for (uint j = 0; j < tadLen; j++)
start = OpType::update(start, OpType::op(tad[j * tadEws], extraParams), extraParams);
auto zOffset = shape::indexOffset(i, zShapeInfo, castZShapeInfo, zLen, canCastZ);
z[zOffset] = OpType::postProcess(start, tadLen, extraParams);
}
}
break;
//*********************************************//
case LoopKind::Z_EWSNONZERO: {
uint castTadShapeInfo[MAX_RANK];
const bool canCastTad = nd4j::DataTypeUtils::castShapeInfo<uint>(tadShapeInfo, castTadShapeInfo);
PRAGMA_OMP_PARALLEL_FOR_THREADS(numThreads)
for (uint i = 0; i < zLen; i++) {
auto tad = x + tadOffsets[i];
auto start = OpType::startingValue(tad);
for (uint j = 0; j < tadLen; j++) {
auto tadOffset = shape::indexOffset(j, tadShapeInfo, castTadShapeInfo, tadLen, canCastTad);
start = OpType::update(start, OpType::op(tad[tadOffset], extraParams), extraParams);
}
z[i * zEws] = OpType::postProcess(start, tadLen, extraParams);
}
}
break;
//*********************************************//
// default: {
// uint castTadShapeInfo[MAX_RANK];
// uint castZShapeInfo[MAX_RANK];
// const bool canCastTad = nd4j::DataTypeUtils::castShapeInfo<uint>(tadShapeInfo, castTadShapeInfo);
// const bool canCastZ = nd4j::DataTypeUtils::castShapeInfo<uint>(zShapeInfo, castZShapeInfo);
// PRAGMA_OMP_PARALLEL_FOR_THREADS(numThreads)
// for (uint i = 0; i < zLen; i++) {
// auto tad = x + tadOffsets[i];
// auto start = OpType::startingValue(tad);
// for (uint j = 0; j < tadLen; j++) {
// auto tadOffset = shape::indexOffset(j, tadShapeInfo, castTadShapeInfo, tadLen, canCastTad);
// start = OpType::update(start, OpType::op(tad[tadOffset], extraParams), extraParams);
// }
// auto zOffset = shape::indexOffset(i, zShapeInfo, castZShapeInfo, zLen, canCastZ);
// z[zOffset] = OpType::postProcess(start, tadLen, extraParams);
// }
// }
//*********************************************//
default: {
Nd4jLong* innertadOffsets = new Nd4jLong[tadLen];
shape::calcOffsets(tadShapeInfo, innertadOffsets);
uint castZShapeInfo[MAX_RANK];
const bool canCastZ = nd4j::DataTypeUtils::castShapeInfo<uint>(zShapeInfo, castZShapeInfo);
PRAGMA_OMP_PARALLEL_FOR_THREADS(numThreads)
for (uint i = 0; i < zLen; i++) {
auto tad = x + tadOffsets[i];
auto start = OpType::startingValue(tad);
for (uint j = 0; j < tadLen; j++)
start = OpType::update(start, OpType::op(tad[innertadOffsets[j]], extraParams), extraParams);
auto zOffset = shape::indexOffset(i, zShapeInfo, castZShapeInfo, zLen, canCastZ);
z[zOffset] = OpType::postProcess(start, tadLen, extraParams);
}
delete []innertadOffsets;
}
//*********************************************//
// default: {
// Nd4jLong* innertadOffsets = new Nd4jLong[tadLen];
// shape::calcOffsets(tadShapeInfo, innertadOffsets);
// const int zRankMinusOne = shape::rank(zShapeInfo) - 1;
// Nd4jLong* offsetPerDimZ = new Nd4jLong[zRankMinusOne];
// int* idxZ = new int[zRankMinusOne];
// memset(idxZ, 0, sizeof(Nd4jLong) * zRankMinusOne);
// const Nd4jLong* shapeZ = shape::shapeOf(zShapeInfo);
// const Nd4jLong* strideZ = shape::stride(zShapeInfo);
// PRAGMA_OMP_SIMD
// for (int k = 0; k < zRankMinusOne; ++k)
// offsetPerDimZ[k] = (shapeZ[k] - 1) * strideZ[k];
// int dimZ = zRankMinusOne, lZ = 1;
// Nd4jLong initZ = 0, zOffset = 0, e = 1;
// // first iteration
// auto tad = x + tadOffsets[0];
// auto start = OpType::startingValue(tad);
// for (uint j = 0; j < tadLen; j++)
// start = OpType::update(start, OpType::op(tad[innertadOffsets[j]], extraParams), extraParams);
// z[0] = OpType::postProcess(start, OpType::startingValue(x), extraParams);
// // rest iterations
// while (dimZ >= 0) {
// if(shapeZ[dimZ] == 1) { --dimZ; continue; } // ignore dimensions equal to unity
// if(dimZ == zRankMinusOne) { // last dimension
// if(lZ < shapeZ[dimZ]) { zOffset += strideZ[dimZ]; ++lZ;}
// else { lZ = 1; --dimZ; continue; }
// }
// else if(idxZ[dimZ] < shapeZ[dimZ] - 1) { initZ += strideZ[dimZ]; zOffset = initZ; ++idxZ[dimZ]; dimZ = zRankMinusOne; }
// else { initZ -= offsetPerDimZ[dimZ]; idxZ[dimZ--] = 0; continue;}
// start = OpType::startingValue(tad);
// tad = x + tadOffsets[e++];
// for (uint j = 0; j < tadLen; j++)
// start = OpType::update(start, OpType::op(tad[innertadOffsets[j]], extraParams), extraParams);
// z[zOffset] = OpType::postProcess(start, tadLen, extraParams);
// }
// delete []innertadOffsets;
// }
}
}
//////////////////////////////////////////////////////////////////////////////
template <typename X, typename Z, typename E>
template <typename OpType, bool doParallel>
void nd4j::TransformLoops<X,Z,E>::loopTransform(X* x, Nd4jLong* xShapeInfo,
Z* z, Nd4jLong* zShapeInfo,
E* extraParams) {
const LoopKind::Kind kindOfLoop = LoopKind::deduceKindOfLoopXZ(xShapeInfo, zShapeInfo);
const Nd4jLong* xShape = shape::shapeOf(const_cast<Nd4jLong*>(xShapeInfo));
const Nd4jLong* xStride = shape::stride(const_cast<Nd4jLong*>(xShapeInfo));
const Nd4jLong* zStride = shape::stride(const_cast<Nd4jLong*>(zShapeInfo));
const Nd4jLong len = shape::length(xShapeInfo);
OmpLaunchHelper threadsInfo(len, doParallel ? -1 : 1);
switch (kindOfLoop) {
//*********************************************//
case LoopKind::EWS1: {
PRAGMA_OMP_PARALLEL_THREADS(threadsInfo._numThreads)
{
const auto threadNum = omp_get_thread_num();
const auto threadOffset = threadsInfo.getThreadOffset(threadNum);
const auto lenPerThread = static_cast<uint>(threadsInfo.getItersPerThread(threadNum));
const auto xi = x + threadOffset;
const auto zi = z + threadOffset;
PRAGMA_OMP_SIMD
for (uint i = 0; i < lenPerThread; i++)
zi[i] = OpType::op(xi[i], extraParams);
}
}
break;
//*********************************************//
case LoopKind::EWSNONZERO: {
const uint xEws = shape::elementWiseStride(xShapeInfo);
const uint zEws = shape::elementWiseStride(zShapeInfo);
PRAGMA_OMP_PARALLEL_THREADS(threadsInfo._numThreads)
{
const auto threadNum = omp_get_thread_num();
const auto threadOffset = threadsInfo.getThreadOffset(threadNum);
const auto lenPerThread = static_cast<uint>(threadsInfo.getItersPerThread(threadNum));
const auto xi = x + threadOffset * xEws;
auto zi = z + threadOffset * zEws;
PRAGMA_OMP_SIMD
for (uint i = 0; i < lenPerThread; i++)
zi[i*zEws] = OpType::op(xi[i*xEws], extraParams);
}
}
break;
//*********************************************//
case LoopKind::Z_EWSNONZERO: {
const uint zEws = shape::elementWiseStride(zShapeInfo);
uint castXShapeInfo[MAX_RANK];
const bool canCastX = nd4j::DataTypeUtils::castShapeInfo<uint>(xShapeInfo, castXShapeInfo);
PRAGMA_OMP_PARALLEL_THREADS(threadsInfo._numThreads)
{
const auto threadNum = omp_get_thread_num();
const auto threadOffset = threadsInfo.getThreadOffset(threadNum);
const auto lenPerThread = static_cast<uint>(threadsInfo.getItersPerThread(threadNum));
auto zi = z + threadOffset * zEws;
if (zEws > 1) {
PRAGMA_OMP_SIMD
for (uint i = 0; i < lenPerThread; i++) {
const auto xOffset = shape::indexOffset(i + threadOffset, xShapeInfo, castXShapeInfo, len, canCastX);
zi[i * zEws] = OpType::op(x[xOffset], extraParams);
}
} else {
PRAGMA_OMP_SIMD
for (uint i = 0; i < lenPerThread; i++) {
const auto xOffset = shape::indexOffset(i + threadOffset, xShapeInfo, castXShapeInfo, len, canCastX);
zi[i] = OpType::op(x[xOffset], extraParams);
}
}
}
}
break;
//*********************************************//
case LoopKind::RANK1: {
PRAGMA_OMP_PARALLEL_FOR_SIMD_THREADS(threadsInfo._numThreads)
for (uint i0 = 0; i0 < len; ++i0)
z[i0 * zStride[0]] = OpType::op(x[i0 * xStride[0]], extraParams);
}
break;
//*********************************************//
case LoopKind::RANK2: {
auto uXShape0 = static_cast<uint>(xShape[0]);
auto uXShape1 = static_cast<uint>(xShape[1]);
//PRAGMA_OMP_PARALLEL_FOR_SIMD_THREADS(threadsInfo._numThreads)
PRAGMA_OMP_PARALLEL_FOR_SIMD
for (uint i0 = 0; i0 < uXShape0; ++i0) {
auto z0 = i0 * zStride[0];
auto x0 = i0 * xStride[0];
for (uint i1 = 0; i1 < uXShape1; ++i1)
z[z0 + i1 * zStride[1]] = OpType::op(x[x0 + i1 * xStride[1]], extraParams);
}
}
break;
//*********************************************//
case LoopKind::RANK3: {
auto uXShape0 = static_cast<uint>(xShape[0]);
auto uXShape1 = static_cast<uint>(xShape[1]);
auto uXShape2 = static_cast<uint>(xShape[2]);
PRAGMA_OMP_PARALLEL_FOR_SIMD_THREADS_COLLAPSE(threadsInfo._numThreads, 2)
for (uint i0 = 0; i0 < uXShape0; ++i0)
for (uint i1 = 0; i1 < uXShape1; ++i1) {
auto z0 = i0 * zStride[0] + i1 * zStride[1];
auto x0 = i0 * xStride[0] + i1 * xStride[1];
for (uint i2 = 0; i2 < uXShape2; ++i2)
z[z0 + i2 * zStride[2]] = OpType::op(x[x0 + i2 * xStride[2]], extraParams);
}
}
break;
//*********************************************//
case LoopKind::RANK4: {
auto uXShape0 = static_cast<uint>(xShape[0]);
auto uXShape1 = static_cast<uint>(xShape[1]);
auto uXShape2 = static_cast<uint>(xShape[2]);
auto uXShape3 = static_cast<uint>(xShape[3]);
PRAGMA_OMP_PARALLEL_FOR_SIMD_THREADS_COLLAPSE(threadsInfo._numThreads, 2)
for (uint i0 = 0; i0 < uXShape0; ++i0)
for (uint i1 = 0; i1 < uXShape1; ++i1)
for (uint i2 = 0; i2 < uXShape2; ++i2) {
auto x0 = i0 * xStride[0] + i1 * xStride[1] + i2 * xStride[2];
auto z0 = i0 * zStride[0] + i1 * zStride[1] + i2 * zStride[2];
for (uint i3 = 0; i3 < uXShape3; ++i3)
z[z0 + i3 * zStride[3]] = OpType::op(x[x0 + i3 * xStride[3]], extraParams);
}
}
break;
//*********************************************//
case LoopKind::RANK5: {
auto uXShape0 = static_cast<uint>(xShape[0]);
auto uXShape1 = static_cast<uint>(xShape[1]);
auto uXShape2 = static_cast<uint>(xShape[2]);
auto uXShape3 = static_cast<uint>(xShape[3]);
auto uXShape4 = static_cast<uint>(xShape[4]);
PRAGMA_OMP_PARALLEL_FOR_SIMD_THREADS_COLLAPSE(threadsInfo._numThreads, 3)
for (uint i0 = 0; i0 < uXShape0; ++i0)
for (uint i1 = 0; i1 < uXShape1; ++i1)
for (uint i2 = 0; i2 < uXShape2; ++i2) {
auto z0 = i0 * zStride[0] + i1 * zStride[1] + i2 * zStride[2];
auto x0 = i0 * xStride[0] + i1 * xStride[1] + i2 * xStride[2];
for (uint i3 = 0; i3 < uXShape3; ++i3) {
auto z1 = z0 + i3 * zStride[3];
auto x1 = x0 + i3 * xStride[3];
for (uint i4 = 0; i4 < uXShape4; ++i4)
z[z1 + i4 * zStride[4]] = OpType::op(x[x1 + i4 * xStride[4]], extraParams);
}
}
}
break;
//*********************************************//
default: {
uint xShapeInfoCast[MAX_RANK];
uint zShapeInfoCast[MAX_RANK];
bool canCastX = DataTypeUtils::castShapeInfo(xShapeInfo, xShapeInfoCast);
bool canCastZ = DataTypeUtils::castShapeInfo(zShapeInfo, zShapeInfoCast);
PRAGMA_OMP_PARALLEL_THREADS(threadsInfo._numThreads)
{
auto threadNum = omp_get_thread_num();
auto threadOffset = threadsInfo.getThreadOffset(threadNum);
auto lenPerThread = static_cast<uint>(threadsInfo.getItersPerThread(threadNum));
PRAGMA_OMP_SIMD
for (uint i = 0; i < lenPerThread; i++) {
auto xOffset = shape::indexOffset(i + threadOffset, xShapeInfo, xShapeInfoCast, len, canCastX);
auto zOffset = shape::indexOffset(i + threadOffset, zShapeInfo, zShapeInfoCast, len, canCastZ);
z[zOffset] = OpType::op(x[xOffset], extraParams);
}
}
}
// default: {
// const int xRankMinusOne = shape::rank(xShapeInfo) - 1;
// const int zRankMinusOne = shape::rank(zShapeInfo) - 1;
// printf("%i %i \n", xRankMinusOne, zRankMinusOne);
// uint* xIdx = new uint[xRankMinusOne + 1];
// uint* zIdx = new uint[zRankMinusOne + 1];
// Nd4jLong* xOffsetPerDim = new Nd4jLong[xRankMinusOne];
// Nd4jLong* zOffsetPerDim = new Nd4jLong[zRankMinusOne];
// memset(xIdx, 0, sizeof(uint) * xRankMinusOne);
// memset(zIdx, 0, sizeof(uint) * zRankMinusOne);
// xIdx[xRankMinusOne] = zIdx[zRankMinusOne] = 1;
// const Nd4jLong* xShape = shape::shapeOf(xShapeInfo);
// const Nd4jLong* zShape = shape::shapeOf(zShapeInfo);
// const Nd4jLong* xStride = shape::stride(xShapeInfo);
// const Nd4jLong* zStride = shape::stride(zShapeInfo);
// PRAGMA_OMP_SIMD
// for (int k = 0; k < xRankMinusOne; ++k)
// xOffsetPerDim[k] = (xShape[k] - 1) * xStride[k];
// PRAGMA_OMP_SIMD
// for (int k = 0; k < zRankMinusOne; ++k)
// zOffsetPerDim[k] = (zShape[k] - 1) * zStride[k];
// Nd4jLong xInit = 0, zInit = 0, xOffset = 0, zOffset = 0;
// int jX = xRankMinusOne, jZ = zRankMinusOne;
// // first iteration
// z[0] = OpType::op(x[0], extraParams);
// // rest iterations
// for (uint i = 1; i < len; i++) {
// while(true) {
// if(xShape[jX] == 1) { --jX; continue; }
// if(jX == xRankMinusOne) {
// if(xIdx[jX] < xShape[jX]) { xOffset += xStride[jX]; ++xIdx[jX]; break; }
// else { xIdx[jX] = 1; --jX; continue; }
// }
// else if(xIdx[jX] < xShape[jX] - 1) { xInit += xStride[jX]; xOffset = xInit; ++xIdx[jX]; jX = xRankMinusOne; break; }
// else { xInit -= xOffsetPerDim[jX]; xIdx[jX--] = 0; continue; }
// }
// while(true) {
// if(zShape[jZ] == 1) { --jZ; continue; }
// if(jZ == zRankMinusOne) {
// if(zIdx[jZ] < zShape[jZ]) { zOffset += zStride[jZ]; ++zIdx[jZ]; break; }
// else { zIdx[jZ] = 1; --jZ; continue; }
// }
// else if(zIdx[jZ] < zShape[jZ] - 1) { zInit += zStride[jZ]; zOffset = zInit; ++zIdx[jZ]; jZ = zRankMinusOne; break; }
// else { zInit -= zOffsetPerDim[jZ]; zIdx[jZ--] = 0; continue; }
// }
// z[zOffset] = OpType::op(x[xOffset], extraParams);
// }
// delete []xIdx;
// delete []zIdx;
// delete []xOffsetPerDim;
// delete []zOffsetPerDim;
// }
}
}
//////////////////////////////////////////////////////////////////////////////
template<typename X, typename Z>
template <typename OpType>
void nd4j::Reduction3Loops<X, Z>::loopReduce3(X* x, Nd4jLong* xShapeInfo,
X* y, Nd4jLong* yShapeInfo,
Z* z, Nd4jLong* zShapeInfo,
int* dims, int dimsLen,
Z* extraParameters) {
// both tads have same shape, however strides and ews may differ
Z param0(OpType::startingValue(x)), param1(OpType::startingValue(x)), param2(extraParameters ? extraParameters[0] : OpType::startingValue(x));
Z extraParams[3] = {param0, param1, param2};
const Nd4jLong xLen = shape::length(xShapeInfo);
const Nd4jLong yLen = shape::length(yShapeInfo);
Nd4jLong *xTadShapeInfo = nullptr, *yTadShapeInfo = nullptr, *xTadOffsets = nullptr, *yTadOffsets = nullptr;
TadPack tadPackX, tadPackY;
std::vector<Nd4jLong> zeroOffsets;
if(xLen == yLen) {
tadPackX = nd4j::ConstantTadHelper::getInstance()->tadForDimensions(xShapeInfo, dims, dimsLen);
tadPackY = nd4j::ConstantTadHelper::getInstance()->tadForDimensions(yShapeInfo, dims, dimsLen);
xTadShapeInfo = tadPackX.primaryShapeInfo();
yTadShapeInfo = tadPackY.primaryShapeInfo();
xTadOffsets = tadPackX.primaryOffsets();
yTadOffsets = tadPackY.primaryOffsets();
}
else if(yLen > xLen) {
tadPackY = nd4j::ConstantTadHelper::getInstance()->tadForDimensions(yShapeInfo, dims, dimsLen);
xTadShapeInfo = xShapeInfo;
yTadShapeInfo = tadPackY.primaryShapeInfo();
yTadOffsets = tadPackY.primaryOffsets();
}
else {
tadPackX = nd4j::ConstantTadHelper::getInstance()->tadForDimensions(xShapeInfo, dims, dimsLen);
yTadShapeInfo = yShapeInfo;
xTadShapeInfo = tadPackX.primaryShapeInfo();
xTadOffsets = tadPackX.primaryOffsets();
}
const LoopKind::Kind kindOfLoop = LoopKind::deduceKindOfLoopTadXYZ(xTadShapeInfo, yTadShapeInfo, zShapeInfo);
const auto xTadEws = shape::elementWiseStride(xTadShapeInfo);
const auto yTadEws = shape::elementWiseStride(yTadShapeInfo);
const auto zEws = shape::elementWiseStride(zShapeInfo);
const auto zLen = shape::length(zShapeInfo);
const auto tadLen = shape::length(xTadShapeInfo);
const auto tadShape = shape::shapeOf(xTadShapeInfo);
const auto xTadStride = shape::stride(xTadShapeInfo);
const auto yTadStride = shape::stride(xTadShapeInfo);
int numThreads = OmpLaunchHelper::tadThreads(tadLen, zLen);
switch (kindOfLoop) {
//*********************************************//
case LoopKind::EWS1: {
PRAGMA_OMP_PARALLEL_FOR_SIMD_ARGS(num_threads(numThreads) OMP_IF(numThreads > 1) private(extraParams))
for (uint i = 0; i < zLen; ++i) {
extraParams[0] = param0;
extraParams[1] = param1;
extraParams[2] = param2;
const auto xTad = xTadOffsets ? x + xTadOffsets[i] : x;
const auto yTad = yTadOffsets ? y + yTadOffsets[i] : y;
auto start = OpType::startingValue(xTad);
for (uint j = 0; j < tadLen; ++j)
start = OpType::update(start, OpType::op(xTad[j], yTad[j], extraParams), extraParams);
z[i] = OpType::postProcess(start, tadLen, extraParams);
}
}
break;
//*********************************************//
case LoopKind::EWSNONZERO: {
PRAGMA_OMP_PARALLEL_FOR_SIMD_ARGS(num_threads(numThreads) OMP_IF(numThreads > 1) private(extraParams))
for (uint i = 0; i < zLen; ++i) {
extraParams[0] = param0;
extraParams[1] = param1;
extraParams[2] = param2;
const auto xTad = xTadOffsets ? x + xTadOffsets[i] : x;
const auto yTad = yTadOffsets ? y + yTadOffsets[i] : y;
auto start = OpType::startingValue(xTad);
for (uint j = 0; j < tadLen; ++j)
start = OpType::update(start, OpType::op(xTad[j * xTadEws], yTad[j * yTadEws], extraParams), extraParams);
z[i * zEws] = OpType::postProcess(start, tadLen, extraParams);
}
}
break;
//*********************************************//
case LoopKind::RANK1: {
PRAGMA_OMP_PARALLEL_FOR_SIMD_ARGS(num_threads(numThreads) OMP_IF(numThreads > 1) private(extraParams))
for (uint i = 0; i < zLen; i++) {
extraParams[0] = param0;
extraParams[1] = param1;
extraParams[2] = param2;
const auto xTad = xTadOffsets ? x + xTadOffsets[i] : x;
const auto yTad = yTadOffsets ? y + yTadOffsets[i] : y;
auto start = OpType::startingValue(xTad);
for (uint i0 = 0; i0 < tadLen; ++i0) {
const auto xTadOffset = i0 * xTadStride[0];
const auto yTadOffset = i0 * yTadStride[0];
start = OpType::update(start, OpType::op(xTad[xTadOffset], yTad[yTadOffset], extraParams), extraParams);
}
z[i * zEws] = OpType::postProcess(start, tadLen, extraParams);
}
}
break;
//*********************************************//
case LoopKind::RANK2: {
PRAGMA_OMP_PARALLEL_FOR_SIMD_ARGS(num_threads(numThreads) OMP_IF(numThreads > 1) private(extraParams))
for (uint i = 0; i < zLen; i++) {
extraParams[0] = param0;
extraParams[1] = param1;
extraParams[2] = param2;
const auto xTad = xTadOffsets ? x + xTadOffsets[i] : x;
const auto yTad = yTadOffsets ? y + yTadOffsets[i] : y;
auto start = OpType::startingValue(xTad);
for (uint i0 = 0; i0 < tadShape[0]; ++i0) {
for (uint i1 = 0; i1 < tadShape[1]; ++i1) {
const auto xTadOffset = i0 * xTadStride[0] + i1 * xTadStride[1];
const auto yTadOffset = i0 * yTadStride[0] + i1 * yTadStride[1];
start = OpType::update(start, OpType::op(xTad[xTadOffset], yTad[yTadOffset], extraParams), extraParams);
}
}
z[i * zEws] = OpType::postProcess(start, tadLen, extraParams);
}
}
break;
//*********************************************//
case LoopKind::RANK3: {
PRAGMA_OMP_PARALLEL_FOR_SIMD_ARGS(num_threads(numThreads) OMP_IF(numThreads > 1) private(extraParams))
for (uint i = 0; i < zLen; i++) {
extraParams[0] = param0;
extraParams[1] = param1;
extraParams[2] = param2;
const auto xTad = xTadOffsets ? x + xTadOffsets[i] : x;
const auto yTad = yTadOffsets ? y + yTadOffsets[i] : y;
auto start = OpType::startingValue(xTad);
for (uint i0 = 0; i0 < tadShape[0]; ++i0) {
for (uint i1 = 0; i1 < tadShape[1]; ++i1) {
for (uint i2 = 0; i2 < tadShape[2]; ++i2) {
const auto xTadOffset = i0 * xTadStride[0] + i1 * xTadStride[1] + i2 * xTadStride[2];
const auto yTadOffset = i0 * yTadStride[0] + i1 * yTadStride[1] + i2 * yTadStride[2];
start = OpType::update(start, OpType::op(xTad[xTadOffset], yTad[yTadOffset], extraParams), extraParams);
}
}
}
z[i * zEws] = OpType::postProcess(start, tadLen, extraParams);
}
}
break;
//*********************************************//
case LoopKind::RANK4: {
PRAGMA_OMP_PARALLEL_FOR_SIMD_ARGS(num_threads(numThreads) OMP_IF(numThreads > 1) private(extraParams))
for (uint i = 0; i < zLen; i++) {
extraParams[0] = param0;
extraParams[1] = param1;
extraParams[2] = param2;
const auto xTad = xTadOffsets ? x + xTadOffsets[i] : x;
const auto yTad = yTadOffsets ? y + yTadOffsets[i] : y;
auto start = OpType::startingValue(xTad);
for (uint i0 = 0; i0 < tadShape[0]; ++i0) {
for (uint i1 = 0; i1 < tadShape[1]; ++i1) {
for (uint i2 = 0; i2 < tadShape[2]; ++i2) {
for (uint i3 = 0; i3 < tadShape[3]; ++i3) {
const auto xTadOffset = i0 * xTadStride[0] + i1 * xTadStride[1] + i2 * xTadStride[2] + i3 * xTadStride[3];
const auto yTadOffset = i0 * yTadStride[0] + i1 * yTadStride[1] + i2 * yTadStride[2] + i3 * yTadStride[3];
start = OpType::update(start, OpType::op(xTad[xTadOffset], yTad[yTadOffset], extraParams), extraParams);
}
}
}
}
z[i * zEws] = OpType::postProcess(start, tadLen, extraParams);
}
}
break;
//*********************************************//
case LoopKind::RANK5: {
PRAGMA_OMP_PARALLEL_FOR_SIMD_ARGS(num_threads(numThreads) OMP_IF(numThreads > 1) private(extraParams))
for (uint i = 0; i < zLen; i++) {
extraParams[0] = param0;
extraParams[1] = param1;
extraParams[2] = param2;
const auto xTad = xTadOffsets ? x + xTadOffsets[i] : x;
const auto yTad = yTadOffsets ? y + yTadOffsets[i] : y;
auto start = OpType::startingValue(xTad);
for (uint i0 = 0; i0 < tadShape[0]; ++i0) {
for (uint i1 = 0; i1 < tadShape[1]; ++i1) {
for (uint i2 = 0; i2 < tadShape[2]; ++i2) {
for (uint i3 = 0; i3 < tadShape[3]; ++i3) {
for (uint i4 = 0; i4 < tadShape[4]; ++i4) {
const auto xTadOffset = i0 * xTadStride[0] + i1 * xTadStride[1] + i2 * xTadStride[2] + i3 * xTadStride[3] + i4 * xTadStride[4];
const auto yTadOffset = i0 * yTadStride[0] + i1 * yTadStride[1] + i2 * yTadStride[2] + i3 * yTadStride[3] + i4 * yTadStride[4];
start = OpType::update(start, OpType::op(xTad[xTadOffset], yTad[yTadOffset], extraParams), extraParams);
}
}
}
}
}
z[i * zEws] = OpType::postProcess(start, tadLen, extraParams);
}
}
break;
//*********************************************//
default: {
uint castXTadShapeInfo[MAX_RANK];
const bool canCastXTad = nd4j::DataTypeUtils::castShapeInfo<uint>(xTadShapeInfo, castXTadShapeInfo);
if(shape::haveSameShapeAndStrides(xTadShapeInfo, yTadShapeInfo)) {
PRAGMA_OMP_PARALLEL_FOR_SIMD_ARGS(num_threads(numThreads) OMP_IF(numThreads > 1) private(extraParams))
for (uint i = 0; i < zLen; ++i) {
extraParams[0] = param0;
extraParams[1] = param1;
extraParams[2] = param2;
const auto xTad = xTadOffsets ? x + xTadOffsets[i] : x;
const auto yTad = yTadOffsets ? y + yTadOffsets[i] : y;
auto start = OpType::startingValue(xTad);
for (uint j = 0; j < tadLen; ++j) {
const auto tadOffset = shape::indexOffset(j, xTadShapeInfo, castXTadShapeInfo, tadLen, canCastXTad);
start = OpType::update(start, OpType::op(xTad[tadOffset], yTad[tadOffset], extraParams), extraParams);
}
z[i * zEws] = OpType::postProcess(start, tadLen, extraParams);
}
}
else {
uint castYTadShapeInfo[MAX_RANK];
const bool canCastYTad = nd4j::DataTypeUtils::castShapeInfo<uint>(yTadShapeInfo, castYTadShapeInfo);
PRAGMA_OMP_PARALLEL_FOR_SIMD_ARGS(num_threads(numThreads) OMP_IF(numThreads > 1) private(extraParams))
for (uint i = 0; i < zLen; ++i) {
extraParams[0] = param0;
extraParams[1] = param1;
extraParams[2] = param2;
const auto xTad = xTadOffsets ? x + xTadOffsets[i] : x;
const auto yTad = yTadOffsets ? y + yTadOffsets[i] : y;
auto start = OpType::startingValue(xTad);
for (uint j = 0; j < tadLen; ++j) {
const auto xTadOffset = shape::indexOffset(j, xTadShapeInfo, castXTadShapeInfo, tadLen, canCastXTad);
const auto yTadOffset = shape::indexOffset(j, yTadShapeInfo, castYTadShapeInfo, tadLen, canCastYTad);
start = OpType::update(start, OpType::op(xTad[xTadOffset], yTad[yTadOffset], extraParams), extraParams);
}
z[i * zEws] = OpType::postProcess(start, tadLen, extraParams);
}
}
}
}
}
//////////////////////////////////////////////////////////////////////////////
template<typename X, typename Z>
template <typename OpType>
void nd4j::Reduction3Loops<X, Z>::loopReduce3All(X* x, Nd4jLong* xShapeInfo,
X* y, Nd4jLong* yShapeInfo,
Z* z, Nd4jLong* zShapeInfo,
Nd4jLong* xTadShapeInfo, Nd4jLong* xTadOffsets,
Nd4jLong* yTadShapeInfo, Nd4jLong* yTadOffsets,
Z* extraParameters) {
// both tads have same shape, however strides and ews may differ
Z param0(OpType::startingValue(x)), param1(OpType::startingValue(x)), param2(extraParameters ? extraParameters[0] : OpType::startingValue(x));
Z extraParams[3] = {param0, param1, param2};
const LoopKind::Kind kindOfLoop = LoopKind::deduceKindOfLoopTadXYZ(xTadShapeInfo, yTadShapeInfo, zShapeInfo);
const auto xTadEws = shape::elementWiseStride(xTadShapeInfo);
const auto yTadEws = shape::elementWiseStride(yTadShapeInfo);
const auto zEws = shape::elementWiseStride(zShapeInfo);
const auto zLen = shape::length(zShapeInfo);
const auto tadLen = shape::length(xTadShapeInfo);
const auto numXTads = shape::length(xShapeInfo) / tadLen;
const auto numYTads = shape::length(yShapeInfo) / tadLen;
const auto tadShape = shape::shapeOf(xTadShapeInfo);
const auto xTadStride = shape::stride(xTadShapeInfo);
const auto yTadStride = shape::stride(yTadShapeInfo);
const auto startVal = OpType::startingValue(x);
int numThreads = OmpLaunchHelper::tadThreads(tadLen, numXTads*numYTads);
switch (kindOfLoop) {
//*********************************************//
case LoopKind::EWS1: {
PRAGMA_OMP_PARALLEL_FOR_SIMD_ARGS(collapse(2) num_threads(numThreads) OMP_IF(numThreads > 1) private(extraParams))
for (uint ix = 0; ix < numXTads; ++ix) {
for (uint iy = 0; iy < numYTads; ++iy) {
extraParams[0] = param0;
extraParams[1] = param1;
extraParams[2] = param2;
const auto xTad = x + xTadOffsets[ix];
const auto yTad = y + yTadOffsets[iy];
const auto zInd = ix * numYTads + iy;
auto start = startVal;
for (uint j = 0; j < tadLen; ++j)
start = OpType::update(start, OpType::op(xTad[j], yTad[j], extraParams), extraParams);
z[zInd] = OpType::postProcess(start, tadLen, extraParams);
}
}
}
break;
//*********************************************//
case LoopKind::EWSNONZERO: {
PRAGMA_OMP_PARALLEL_FOR_SIMD_ARGS(collapse(2) num_threads(numThreads) OMP_IF(numThreads > 1) private(extraParams))
for (uint ix = 0; ix < numXTads; ++ix) {
for (uint iy = 0; iy < numYTads; ++iy) {
extraParams[0] = param0;
extraParams[1] = param1;
extraParams[2] = param2;
const auto xTad = x + xTadOffsets[ix];
const auto yTad = y + yTadOffsets[iy];
const auto zInd = ix * numYTads + iy;
auto start = startVal;
for (uint j = 0; j < tadLen; ++j)
start = OpType::update(start, OpType::op(xTad[j * xTadEws], yTad[j * yTadEws], extraParams), extraParams);
z[zInd * zEws] = OpType::postProcess(start, tadLen, extraParams);
}
}
}
break;
//*********************************************//
case LoopKind::RANK1: {
PRAGMA_OMP_PARALLEL_FOR_SIMD_ARGS(collapse(2) num_threads(numThreads) OMP_IF(numThreads > 1) private(extraParams))
for (uint ix = 0; ix < numXTads; ++ix) {
for (uint iy = 0; iy < numYTads; ++iy) {
extraParams[0] = param0;
extraParams[1] = param1;
extraParams[2] = param2;
const auto xTad = x + xTadOffsets[ix];
const auto yTad = y + yTadOffsets[iy];
const auto zInd = ix * numYTads + iy;
auto start = startVal;
for (uint i0 = 0; i0 < tadLen; ++i0) {
const auto xTadOffset = i0 * xTadStride[0];
const auto yTadOffset = i0 * yTadStride[0];
start = OpType::update(start, OpType::op(xTad[xTadOffset], yTad[yTadOffset], extraParams), extraParams);
}
z[zInd * zEws] = OpType::postProcess(start, tadLen, extraParams);
}
}
}
break;
//*********************************************//
case LoopKind::RANK2: {
PRAGMA_OMP_PARALLEL_FOR_SIMD_ARGS(collapse(2) num_threads(numThreads) OMP_IF(numThreads > 1) private(extraParams))
for (uint ix = 0; ix < numXTads; ++ix) {
for (uint iy = 0; iy < numYTads; ++iy) {
extraParams[0] = param0;
extraParams[1] = param1;
extraParams[2] = param2;
const auto xTad = x + xTadOffsets[ix];
const auto yTad = y + yTadOffsets[iy];
const auto zInd = ix * numYTads + iy;
auto start = startVal;
for (uint i0 = 0; i0 < tadShape[0]; ++i0) {
for (uint i1 = 0; i1 < tadShape[1]; ++i1) {
const auto xTadOffset = i0 * xTadStride[0] + i1 * xTadStride[1];
const auto yTadOffset = i0 * yTadStride[0] + i1 * yTadStride[1];
start = OpType::update(start, OpType::op(xTad[xTadOffset], yTad[yTadOffset], extraParams), extraParams);
}
}
z[zInd * zEws] = OpType::postProcess(start, tadLen, extraParams);
}
}
}
break;
//*********************************************//
case LoopKind::RANK3: {
PRAGMA_OMP_PARALLEL_FOR_SIMD_ARGS(collapse(2) num_threads(numThreads) OMP_IF(numThreads > 1) private(extraParams))
for (uint ix = 0; ix < numXTads; ++ix) {
for (uint iy = 0; iy < numYTads; ++iy) {
extraParams[0] = param0;
extraParams[1] = param1;
extraParams[2] = param2;
const auto xTad = x + xTadOffsets[ix];
const auto yTad = y + yTadOffsets[iy];
const auto zInd = ix * numYTads + iy;
auto start = startVal;
for (uint i0 = 0; i0 < tadShape[0]; ++i0) {
for (uint i1 = 0; i1 < tadShape[1]; ++i1) {
for (uint i2 = 0; i2 < tadShape[2]; ++i2) {
const auto xTadOffset = i0 * xTadStride[0] + i1 * xTadStride[1] + i2 * xTadStride[2];
const auto yTadOffset = i0 * yTadStride[0] + i1 * yTadStride[1] + i2 * yTadStride[2];
start = OpType::update(start, OpType::op(xTad[xTadOffset], yTad[yTadOffset], extraParams), extraParams);
}
}
}
z[zInd * zEws] = OpType::postProcess(start, tadLen, extraParams);
}
}
}
break;
//*********************************************//
case LoopKind::RANK4: {
PRAGMA_OMP_PARALLEL_FOR_SIMD_ARGS(collapse(2) num_threads(numThreads) OMP_IF(numThreads > 1) private(extraParams))
for (uint ix = 0; ix < numXTads; ++ix) {
for (uint iy = 0; iy < numYTads; ++iy) {
extraParams[0] = param0;
extraParams[1] = param1;
extraParams[2] = param2;
const auto xTad = x + xTadOffsets[ix];
const auto yTad = y + yTadOffsets[iy];
const auto zInd = ix * numYTads + iy;
auto start = startVal;
for (uint i0 = 0; i0 < tadShape[0]; ++i0) {
for (uint i1 = 0; i1 < tadShape[1]; ++i1) {
for (uint i2 = 0; i2 < tadShape[2]; ++i2) {
for (uint i3 = 0; i3 < tadShape[3]; ++i3) {
const auto xTadOffset = i0 * xTadStride[0] + i1 * xTadStride[1] + i2 * xTadStride[2] + i3 * xTadStride[3];
const auto yTadOffset = i0 * yTadStride[0] + i1 * yTadStride[1] + i2 * yTadStride[2] + i3 * yTadStride[3];
start = OpType::update(start, OpType::op(xTad[xTadOffset], yTad[yTadOffset], extraParams), extraParams);
}
}
}
}
z[zInd * zEws] = OpType::postProcess(start, tadLen, extraParams);
}
}
}
break;
//*********************************************//
case LoopKind::RANK5: {
PRAGMA_OMP_PARALLEL_FOR_SIMD_ARGS(collapse(2) num_threads(numThreads) OMP_IF(numThreads > 1) private(extraParams))
for (uint ix = 0; ix < numXTads; ++ix) {
for (uint iy = 0; iy < numYTads; ++iy) {
extraParams[0] = param0;
extraParams[1] = param1;
extraParams[2] = param2;
const auto xTad = x + xTadOffsets[ix];
const auto yTad = y + yTadOffsets[iy];
const auto zInd = ix * numYTads + iy;
auto start = startVal;
for (uint i0 = 0; i0 < tadShape[0]; ++i0) {
for (uint i1 = 0; i1 < tadShape[1]; ++i1) {
for (uint i2 = 0; i2 < tadShape[2]; ++i2) {
for (uint i3 = 0; i3 < tadShape[3]; ++i3) {
for (uint i4 = 0; i4 < tadShape[4]; ++i4) {
const auto xTadOffset = i0 * xTadStride[0] + i1 * xTadStride[1] + i2 * xTadStride[2] + i3 * xTadStride[3] + i4 * xTadStride[4];
const auto yTadOffset = i0 * yTadStride[0] + i1 * yTadStride[1] + i2 * yTadStride[2] + i3 * yTadStride[3] + i4 * yTadStride[4];
start = OpType::update(start, OpType::op(xTad[xTadOffset], yTad[yTadOffset], extraParams), extraParams);
}
}
}
}
}
z[zInd * zEws] = OpType::postProcess(start, tadLen, extraParams);
}
}
}
break;
//*********************************************//
default: {
uint castXTadShapeInfo[MAX_RANK];
const bool canCastXTad = nd4j::DataTypeUtils::castShapeInfo<uint>(xTadShapeInfo, castXTadShapeInfo);
if(shape::haveSameShapeAndStrides(xTadShapeInfo, yTadShapeInfo)) {
PRAGMA_OMP_PARALLEL_FOR_SIMD_ARGS(collapse(2) num_threads(numThreads) OMP_IF(numThreads > 1) private(extraParams))
for (uint ix = 0; ix < numXTads; ++ix) {
for (uint iy = 0; iy < numYTads; ++iy) {
extraParams[0] = param0;
extraParams[1] = param1;
extraParams[2] = param2;
const auto xTad = x + xTadOffsets[ix];
const auto yTad = y + yTadOffsets[iy];
const auto zInd = ix * numYTads + iy;
auto start = startVal;
for (uint j = 0; j < tadLen; ++j) {
const auto tadOffset = shape::indexOffset(j, xTadShapeInfo, castXTadShapeInfo, tadLen, canCastXTad);
start = OpType::update(start, OpType::op(xTad[tadOffset], yTad[tadOffset], extraParams), extraParams);
}
z[zInd * zEws] = OpType::postProcess(start, tadLen, extraParams);
}
}
}
else {
uint castYTadShapeInfo[MAX_RANK];
const bool canCastYTad = nd4j::DataTypeUtils::castShapeInfo<uint>(yTadShapeInfo, castYTadShapeInfo);
PRAGMA_OMP_PARALLEL_FOR_SIMD_ARGS(collapse(2) num_threads(numThreads) OMP_IF(numThreads > 1) private(extraParams))
for (uint ix = 0; ix < numXTads; ++ix) {
for (uint iy = 0; iy < numYTads; ++iy) {
extraParams[0] = param0;
extraParams[1] = param1;
extraParams[2] = param2;
const auto xTad = x + xTadOffsets[ix];
const auto yTad = y + yTadOffsets[iy];
const auto zInd = ix * numYTads + iy;
auto start = startVal;
for (uint j = 0; j < tadLen; ++j) {
const auto xTadOffset = shape::indexOffset(j, xTadShapeInfo, castXTadShapeInfo, tadLen, canCastXTad);
const auto yTadOffset = shape::indexOffset(j, yTadShapeInfo, castYTadShapeInfo, tadLen, canCastYTad);
start = OpType::update(start, OpType::op(xTad[xTadOffset], yTad[yTadOffset], extraParams), extraParams);
}
z[zInd * zEws] = OpType::postProcess(start, tadLen, extraParams);
}
}
}
}
}
}
}
#endif //LIBND4J_LOOPS_H