org.nd4j.linalg.cpu.nativecpu.ops.NativeOpExecutioner Maven / Gradle / Ivy
/*******************************************************************************
* 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
******************************************************************************/
package org.nd4j.linalg.cpu.nativecpu.ops;
import lombok.Data;
import lombok.Getter;
import lombok.NonNull;
import lombok.extern.slf4j.Slf4j;
import lombok.val;
import org.bytedeco.javacpp.*;
import org.nd4j.compression.impl.AbstractCompressor;
import org.nd4j.linalg.api.buffer.DataBuffer;
import org.nd4j.linalg.api.concurrency.AffinityManager;
import org.nd4j.linalg.api.environment.Nd4jEnvironment;
import org.nd4j.linalg.api.memory.pointers.PagedPointer;
import org.nd4j.linalg.api.ndarray.INDArray;
import org.nd4j.linalg.api.ops.*;
import org.nd4j.linalg.api.ops.aggregates.Aggregate;
import org.nd4j.linalg.api.ops.aggregates.Batch;
import org.nd4j.linalg.api.ops.executioner.DefaultOpExecutioner;
import org.nd4j.linalg.api.ops.executioner.OpStatus;
import org.nd4j.linalg.api.ops.impl.accum.Variance;
import org.nd4j.linalg.api.ops.performance.PerformanceTracker;
import org.nd4j.linalg.api.rng.Random;
import org.nd4j.linalg.api.shape.Shape;
import org.nd4j.linalg.cache.ConstantHandler;
import org.nd4j.linalg.cache.TADManager;
import org.nd4j.linalg.compression.CompressionDescriptor;
import org.nd4j.linalg.compression.CompressionType;
import org.nd4j.linalg.compression.ThresholdCompression;
import org.nd4j.linalg.cpu.nativecpu.CpuTADManager;
import org.nd4j.linalg.exception.ND4JIllegalStateException;
import org.nd4j.linalg.factory.Nd4j;
import org.nd4j.linalg.memory.MemcpyDirection;
import org.nd4j.linalg.primitives.Pair;
import org.nd4j.linalg.util.ArrayUtil;
import org.nd4j.nativeblas.LongPointerWrapper;
import org.nd4j.nativeblas.NativeOps;
import org.nd4j.nativeblas.NativeOpsHolder;
import org.nd4j.nativeblas.Nd4jCpu;
import java.util.*;
/**
*
* Native operation
* executioner in c++
*
* @author Adam Gibson
*/
@Slf4j
public class NativeOpExecutioner extends DefaultOpExecutioner {
private NativeOps loop = NativeOpsHolder.getInstance().getDeviceNativeOps();
private ConstantHandler constantHandler = Nd4j.getConstantHandler();
@Getter
private CpuTADManager tadManager = new CpuTADManager();
//thread locals for custom op inputs and outputs to prevent allocations
//every time exec(CustomOp) is called
private ThreadLocal> inputShapes = new ThreadLocal<>();
private ThreadLocal> inputBuffers = new ThreadLocal<>();
private ThreadLocal> outputShapes = new ThreadLocal<>();
private ThreadLocal> outputBuffers = new ThreadLocal<>();
private ThreadLocal> tArgsPointer = new ThreadLocal<>();
private ThreadLocal> halfArgsPointer = new ThreadLocal<>();
protected Map customOps = null;
protected ThreadLocal extraz = new ThreadLocal<>();
/**
* Instead of allocating new memory chunks for each batch invocation, we reuse them on thread/opNum basis
* Since for NativeOpExecutioner all executions are synchronous
*/
private ThreadLocal> batchPointers = new ThreadLocal<>();
private ThreadLocal> memoryBlocks = new ThreadLocal<>();
public NativeOpExecutioner() {
tadManager.init(loop, constantHandler);
}
@Override
public Op exec(Op op) {
checkForCompression(op);
if (op instanceof ScalarOp) {
ScalarOp s = (ScalarOp) op;
exec(s);
}
else if(op instanceof GradientOp) {
op.exec();
}
else if (op instanceof TransformOp) {
TransformOp t = (TransformOp) op;
exec(t);
} else if (op instanceof Accumulation) {
Accumulation ac = (Accumulation) op;
exec(ac);
} else if (op instanceof IndexAccumulation) {
IndexAccumulation iac = (IndexAccumulation) op;
exec(iac); //Currently using DefaultOpExecutioner
} else if (op instanceof BroadcastOp) {
BroadcastOp broadcastOp = (BroadcastOp) op;
exec(broadcastOp, broadcastOp.getDimension());
}
else if(op instanceof ShapeOp) {
ShapeOp shapeOp = (ShapeOp) op;
exec(shapeOp);
} else if (op instanceof RandomOp) {
RandomOp rngOp = (RandomOp) op;
exec(rngOp, Nd4j.getRandom());
}
return op;
}
@Override
public INDArray exec(IndexAccumulation op, int... dimension) {
if (dimension == null || dimension.length == 0)
dimension = new int[] {Integer.MAX_VALUE};
checkForCompression(op);
validateDataType(Nd4j.dataType(), op);
if (extraz.get() == null)
extraz.set(new PointerPointer(32));
dimension = Shape.normalizeAxis(op.x().rank(), dimension);
for (int i = 0; i < dimension.length; i++) {
if (dimension[i] < 0)
dimension[i] += op.x().rank();
}
//do op along all dimensions
if (dimension.length == op.x().rank())
dimension = new int[] {Integer.MAX_VALUE};
boolean keepDims;
boolean newFormat;
if(op instanceof BaseIndexAccumulation) {
keepDims = ((BaseIndexAccumulation) op).isKeepDims();
newFormat = ((BaseIndexAccumulation) op).isNewFormat();
} else {
keepDims = false;
newFormat = false;
}
long[] retShape = reductionShape(op.x(), dimension, newFormat, keepDims);
if(op.z() == null || op.x() == op.z()) {
val ret = Nd4j.create(retShape);
op.setZ(ret);
} else if(!Arrays.equals(retShape, op.z().shape())){
throw new IllegalStateException("Z array shape does not match expected return type for op " + op
+ ": expected shape " + Arrays.toString(retShape) + ", z.shape()=" + Arrays.toString(op.z().shape()));
}
//do op along all dimensions
if (dimension.length == op.x().rank())
dimension = new int[] {Integer.MAX_VALUE};
Pointer dimensionAddress = constantHandler.getConstantBuffer(dimension).addressPointer();
Pair tadBuffers = tadManager.getTADOnlyShapeInfo(op.x(), dimension);
Pointer hostTadShapeInfo = tadBuffers.getFirst().addressPointer();
DataBuffer offsets = tadBuffers.getSecond();
Pointer hostTadOffsets = offsets == null ? null : offsets.addressPointer();
PointerPointer dummy = extraz.get().put(hostTadShapeInfo, hostTadOffsets);
long st = profilingHookIn(op, tadBuffers.getFirst());
Pointer x = op.x().data().addressPointer();
Pointer z = op.z().data().addressPointer();
if (op.x().data().dataType() == DataBuffer.Type.DOUBLE) {
if (op.z().isScalar()) {
int res = (int) loop.execIndexReduceScalarDouble(dummy, op.opNum(),
(DoublePointer) op.x().data().addressPointer(),
(LongPointer) op.x().shapeInfoDataBuffer().addressPointer(),
(DoublePointer) getPointerForExtraArgs(op));
op.setFinalResult(res);
op.z().putScalar(0, (float) res);
} else {
loop.execIndexReduceDouble(dummy, op.opNum(), (DoublePointer) x,
(LongPointer) op.x().shapeInfoDataBuffer().addressPointer(),
(DoublePointer) getPointerForExtraArgs(op), (DoublePointer) z,
(LongPointer) op.z().shapeInfoDataBuffer().addressPointer(),
(IntPointer) dimensionAddress, dimension.length);
}
} else {
if (op.z().isScalar()) {
int res = (int) loop.execIndexReduceScalarFloat(dummy, op.opNum(),
(FloatPointer) op.x().data().addressPointer(),
(LongPointer) op.x().shapeInfoDataBuffer().addressPointer(),
(FloatPointer) getPointerForExtraArgs(op));
op.setFinalResult(res);
op.z().putScalar(0, (float) res);
} else {
loop.execIndexReduceFloat(dummy, op.opNum(), (FloatPointer) op.x().data().addressPointer(),
(LongPointer) op.x().shapeInfoDataBuffer().addressPointer(),
(FloatPointer) getPointerForExtraArgs(op),
(FloatPointer) op.z().data().addressPointer(),
(LongPointer) op.z().shapeInfoDataBuffer().addressPointer(),
(IntPointer) dimensionAddress, dimension.length);
}
}
profilingHookOut(op, st);
return op.z();
}
@Override
public INDArray exec(Accumulation op, int... dimension) {
dimension = Shape.normalizeAxis(op.x().rank(), dimension);
validateDataType(Nd4j.dataType(), op);
if (extraz.get() == null)
extraz.set(new PointerPointer(32));
long[] maxShape = Shape.getMaxShape(op.x(),op.y());
for (int i = 0; i < dimension.length; i++)
if (dimension[i] >= maxShape.length && dimension[i] != Integer.MAX_VALUE)
throw new ND4JIllegalStateException("Op target dimension " + Arrays.toString(dimension)
+ " contains element that higher then rank of op.X: [" + op.x().rank() + "]");
for (int i = 0; i < dimension.length; i++) {
if (dimension[i] < 0)
dimension[i] += op.x().rank();
}
//do op along all dimensions
if (dimension.length == op.x().rank())
dimension = new int[] {Integer.MAX_VALUE};
boolean keepDims;
boolean newFormat;
if(op instanceof BaseAccumulation) {
keepDims = op.isKeepDims();
newFormat = ((BaseAccumulation) op).isNewFormat();
} else {
keepDims = false;
newFormat = false;
}
long[] retShape = reductionShape(op.x(), dimension, newFormat, keepDims);
if (op.x().isVector() && op.x().length() == ArrayUtil.prod(retShape) && ArrayUtil.prodLong(retShape) > 1 && op.y() == null)
return op.noOp();
/**
* This is the result array.
* We create it only if we hadn't provided it before
*/
INDArray ret;
if (op.z() == null || op.z() == op.x()) {
if (op.isComplexAccumulation()) {
long xT = op.x().tensorssAlongDimension(dimension);
long yT = op.y().tensorssAlongDimension(dimension);
ret = Nd4j.create(xT, yT);
} else {
if (op.y() != null) {
//2 options here: either pairwise, equal sizes - OR every X TAD vs. entirety of Y
if(op.x().lengthLong() == op.y().lengthLong()) {
//Pairwise
if (op.x().tensorssAlongDimension(dimension) != op.y().tensorssAlongDimension(dimension)) {
throw new ND4JIllegalStateException("Number of TADs along dimension don't match: (x shape = " +
Arrays.toString(op.x().shape()) + ", y shape = " + Arrays.toString(op.y().shape()) +
", dimension = " + Arrays.toString(dimension) + ")");
}
} else {
//Every X TAD vs. entirety of Y
val xTADSize = op.x().lengthLong() / op.x().tensorssAlongDimension(dimension);
if (xTADSize != op.y().length()) {
throw new ND4JIllegalStateException("Size of TADs along dimension don't match for pairwise execution:" +
" (x TAD size = " + xTADSize + ", y size = " + op.y().lengthLong());
}
}
}
ret = Nd4j.create(retShape);
}
op.setZ(ret);
} else {
// compare length
long shapeProduct = (retShape.length == 0 ? 1 : ArrayUtil.prodLong(retShape));
if (!op.isComplexAccumulation() && op.z().lengthLong() != shapeProduct)
throw new ND4JIllegalStateException("Shape of target array for reduction [" + Arrays.toString(op.z().shape()) + "] doesn't match expected [" + Arrays.toString(retShape) + "]");
else if (op.isComplexAccumulation()) {
long xT = op.x().tensorssAlongDimension(dimension);
long yT = op.y().tensorssAlongDimension(dimension);
if (op.z().lengthLong() != xT * yT)
throw new ND4JIllegalStateException("Shape of target array for reduction [" + Arrays.toString(op.z().shape()) + "] doesn't match expected [" + (xT * yT) + "]");
}
if (op.x().data().dataType() == DataBuffer.Type.DOUBLE) {
op.z().assign(op.zeroDouble());
} else {
op.z().assign(op.zeroFloat());
}
ret = op.z();
}
/**
* Returns the {@link Shape#createShapeInformation(int[], int[], int, int, char)}
* and the associated offsets for each {@link INDArray#tensorAlongDimension(int, int...)}
* The first item is the shape information. The second one is the offsets.
*/
Pair tadBuffers = tadManager.getTADOnlyShapeInfo(op.x(), dimension);
Pair yTadBuffers = null;
/**
* Note that we use addresses in libnd4j.
* We use reinterpret cast in c to take the long
* we pass to JNI. This manages overhead.
*/
Pointer hostTadShapeInfo = tadBuffers.getFirst().addressPointer();
DataBuffer offsets = tadBuffers.getSecond();
Pointer hostTadOffsets = offsets == null ? null : offsets.addressPointer();
// we're going to check, if that's TAD vs TAD comparison or TAD vs full array. if later - we're going slightly different route
boolean tvf = false;
if (op.y() != null) {
if (op.x().tensorAlongDimension(0, dimension).lengthLong() == op.y().lengthLong()) {
tvf = true;
}
}
if (op.isComplexAccumulation()) {
yTadBuffers = tadManager.getTADOnlyShapeInfo(op.y(), dimension);
if (op.x().tensorAlongDimension(0, dimension).lengthLong() != op.y().tensorAlongDimension(0, dimension).lengthLong())
throw new ND4JIllegalStateException("Impossible to issue AllDistances operation: TAD lengths mismatch along given dimension: " +
"x TAD length = " + op.x().tensorAlongDimension(0, dimension).lengthLong() + ", y TAD length " +
op.y().tensorAlongDimension(0, dimension).lengthLong());
}
/**
* This is a pointer to a pointer in c.
*/
// FIXME: we need something better then 3rd element being non-null here...
PointerPointer dummy = extraz.get().put(hostTadShapeInfo, hostTadOffsets, tvf ? hostTadOffsets : null);
long st = profilingHookIn(op, tadBuffers.getFirst());
/**
* Note because dimension arrays don't change,
* we use an {@link ConstantHandler} which knows how to reserve memory
* for immutable buffers for the dimensions.
* This gives us a pointer which is passed around in libnd4j.
*/
Pointer dimensionAddress = constantHandler.getConstantBuffer(dimension).addressPointer();
if (op.x().data().dataType() == DataBuffer.Type.DOUBLE) {
if (op instanceof Variance) {
if (ret.isScalar()) {
ret.putScalar(0, loop.execSummaryStatsScalarDouble(dummy, op.opNum(),
(DoublePointer) op.x().data().addressPointer(),
(LongPointer) op.x().shapeInfoDataBuffer().addressPointer(),
(DoublePointer) getPointerForExtraArgs(op), ((Variance) op).isBiasCorrected()));
} else {
Variance var = (Variance) op;
loop.execSummaryStatsDouble(dummy, op.opNum(), (DoublePointer) op.x().data().addressPointer(),
(LongPointer) op.x().shapeInfoDataBuffer().addressPointer(),
(DoublePointer) getPointerForExtraArgs(op),
(DoublePointer) op.z().data().addressPointer(),
(LongPointer) op.z().shapeInfoDataBuffer().addressPointer(),
(IntPointer) dimensionAddress, dimension.length, var.isBiasCorrected());
}
}
//pairwise reduction like similarity of two arrays
else if (op.y() != null && op.getOpType() == Op.Type.REDUCE3) {
if (op.isComplexAccumulation()) {
loop.execReduce3AllDouble(dummy, op.opNum(), (DoublePointer) op.x().data().addressPointer(),
(LongPointer) op.x().shapeInfoDataBuffer().addressPointer(),
(DoublePointer) getPointerForExtraArgs(op),
(DoublePointer) op.y().data().addressPointer(),
(LongPointer) op.y().shapeInfoDataBuffer().addressPointer(),
(DoublePointer) op.z().data().addressPointer(),
(LongPointer) op.z().shapeInfoDataBuffer().addressPointer(),
(IntPointer) dimensionAddress, dimension.length,
(LongPointer) tadBuffers.getFirst().addressPointer(),
new LongPointerWrapper(tadBuffers.getSecond().addressPointer()),
(LongPointer) yTadBuffers.getFirst().addressPointer(),
new LongPointerWrapper(yTadBuffers.getSecond().addressPointer())
);
} else if (ret.isScalar()) {
ret.putScalar(0, loop.execReduce3ScalarDouble(dummy, op.opNum(),
(DoublePointer) op.x().data().addressPointer(),
(LongPointer) op.x().shapeInfoDataBuffer().addressPointer(),
(DoublePointer) getPointerForExtraArgs(op),
(DoublePointer) op.y().data().addressPointer(),
(LongPointer) op.y().shapeInfoDataBuffer().addressPointer()));
} else {
loop.execReduce3Double(dummy, op.opNum(), (DoublePointer) op.x().data().addressPointer(),
(LongPointer) op.x().shapeInfoDataBuffer().addressPointer(),
(DoublePointer) getPointerForExtraArgs(op),
(DoublePointer) op.y().data().addressPointer(),
(LongPointer) op.y().shapeInfoDataBuffer().addressPointer(),
(DoublePointer) op.z().data().addressPointer(),
(LongPointer) op.z().shapeInfoDataBuffer().addressPointer(),
(IntPointer) dimensionAddress, dimension.length);
}
} else {
if (ret.isScalar()) {
ret.putScalar(0, loop.execReduceScalarDouble(dummy, op.opNum(),
(DoublePointer) op.x().data().addressPointer(),
(LongPointer) op.x().shapeInfoDataBuffer().addressPointer(),
(DoublePointer) getPointerForExtraArgs(op)));
} else {
loop.execReduceDouble(dummy, op.opNum(), (DoublePointer) op.x().data().addressPointer(),
(LongPointer) op.x().shapeInfoDataBuffer().addressPointer(),
(DoublePointer) getPointerForExtraArgs(op),
(DoublePointer) op.z().data().addressPointer(),
(LongPointer) op.z().shapeInfoDataBuffer().addressPointer(),
(IntPointer) dimensionAddress, dimension.length);
}
}
} else {
if (op instanceof Variance) {
Variance variance = (Variance) op;
if (ret.isScalar()) {
ret.putScalar(0, loop.execSummaryStatsScalarFloat(dummy, op.opNum(),
(FloatPointer) op.x().data().addressPointer(),
(LongPointer) op.x().shapeInfoDataBuffer().addressPointer(),
(FloatPointer) getPointerForExtraArgs(op), variance.isBiasCorrected()));
} else {
loop.execSummaryStatsFloat(dummy, op.opNum(), (FloatPointer) op.x().data().addressPointer(),
(LongPointer) op.x().shapeInfoDataBuffer().addressPointer(),
(FloatPointer) getPointerForExtraArgs(op),
(FloatPointer) op.z().data().addressPointer(),
(LongPointer) op.z().shapeInfoDataBuffer().addressPointer(),
(IntPointer) dimensionAddress, dimension.length, variance.isBiasCorrected());
}
}
else if (op.y() != null && op.getOpType() == Op.Type.REDUCE3) {
if (op.isComplexAccumulation()) {
loop.execReduce3AllFloat(dummy, op.opNum(), (FloatPointer) op.x().data().addressPointer(),
(LongPointer) op.x().shapeInfoDataBuffer().addressPointer(),
(FloatPointer) getPointerForExtraArgs(op),
(FloatPointer) op.y().data().addressPointer(),
(LongPointer) op.y().shapeInfoDataBuffer().addressPointer(),
(FloatPointer) op.z().data().addressPointer(),
(LongPointer) op.z().shapeInfoDataBuffer().addressPointer(),
(IntPointer) dimensionAddress, dimension.length,
(LongPointer) tadBuffers.getFirst().addressPointer(),
new LongPointerWrapper(tadBuffers.getSecond().addressPointer()),
(LongPointer) yTadBuffers.getFirst().addressPointer(),
new LongPointerWrapper(yTadBuffers.getSecond().addressPointer())
);
} else if (ret.isScalar()) {
ret.putScalar(0, loop.execReduce3ScalarFloat(dummy, op.opNum(),
(FloatPointer) op.x().data().addressPointer(),
(LongPointer) op.x().shapeInfoDataBuffer().addressPointer(),
(FloatPointer) getPointerForExtraArgs(op),
(FloatPointer) op.y().data().addressPointer(),
(LongPointer) op.y().shapeInfoDataBuffer().addressPointer()));
} else {
loop.execReduce3Float(dummy, op.opNum(), (FloatPointer) op.x().data().addressPointer(),
(LongPointer) op.x().shapeInfoDataBuffer().addressPointer(),
(FloatPointer) getPointerForExtraArgs(op),
(FloatPointer) op.y().data().addressPointer(),
(LongPointer) op.y().shapeInfoDataBuffer().addressPointer(),
(FloatPointer) op.z().data().addressPointer(),
(LongPointer) op.z().shapeInfoDataBuffer().addressPointer(),
(IntPointer) dimensionAddress, dimension.length);
}
} else {
if (ret.isScalar()) {
ret.putScalar(0, loop.execReduceScalarFloat(dummy, op.opNum(),
(FloatPointer) op.x().data().addressPointer(),
(LongPointer) op.x().shapeInfoDataBuffer().addressPointer(),
(FloatPointer) getPointerForExtraArgs(op)));
} else {
loop.execReduceFloat(dummy, op.opNum(), (FloatPointer) op.x().data().addressPointer(),
(LongPointer) op.x().shapeInfoDataBuffer().addressPointer(),
(FloatPointer) getPointerForExtraArgs(op),
(FloatPointer) op.z().data().addressPointer(),
(LongPointer) op.z().shapeInfoDataBuffer().addressPointer(),
(IntPointer) dimensionAddress, dimension.length);
}
}
}
return ret;
}
/**
* ScalarOp along dimension
* @param op
* @param dimension
*/
private void invoke(ScalarOp op, int[] dimension) {
dimension = Shape.normalizeAxis(op.x().rank(), dimension);
// do tad magic
/**
* Returns the {@link Shape#createShapeInformation(int[], int[], int, int, char)}
* and the associated offsets for each {@link INDArray#tensorAlongDimension(int, int...)}
* The first item is the shape information. The second one is the offsets.
*/
Pair tadBuffers = tadManager.getTADOnlyShapeInfo(op.x(), dimension);
Pointer hostTadShapeInfo = tadBuffers.getFirst().addressPointer();
Pointer hostTadOffsets = tadBuffers.getSecond().addressPointer();
Pointer devTadShapeInfoZ = null;
Pointer devTadOffsetsZ = null;
/**
* Returns the {@link Shape#createShapeInformation(int[], int[], int, int, char)}
* and the associated offsets for each {@link INDArray#tensorAlongDimension(int, int...)}
* The first item is the shape information. The second one is the offsets.
*
* Note that this is the *result* TAD information. An op is always input (x) and output (z)
* for result.
* This is for assigning the result to of the operation along
* the proper dimension.
*/
Pair tadBuffersZ = tadManager.getTADOnlyShapeInfo(op.z(), dimension);
devTadShapeInfoZ = tadBuffersZ.getFirst().addressPointer();
devTadOffsetsZ = tadBuffersZ.getSecond().addressPointer();
if (extraz.get() == null)
extraz.set(new PointerPointer(32));
PointerPointer dummy = extraz.get().put(hostTadShapeInfo, hostTadOffsets, devTadShapeInfoZ, devTadOffsetsZ);
if (op.x().data().dataType() == DataBuffer.Type.FLOAT) {
loop.execScalarFloat(dummy, op.opNum(), (FloatPointer) op.x().data().addressPointer(),
(LongPointer) op.x().shapeInfoDataBuffer().addressPointer(),
(FloatPointer) op.z().data().addressPointer(),
(LongPointer) op.z().shapeInfoDataBuffer().addressPointer(),
(FloatPointer) op.y().data().addressPointer(), (FloatPointer) getPointerForExtraArgs(op),
(IntPointer) Nd4j.getConstantHandler().getConstantBuffer(dimension).addressPointer(), dimension.length);
} else if (op.x().data().dataType() == DataBuffer.Type.DOUBLE) {
loop.execScalarDouble(dummy, op.opNum(), (DoublePointer) op.x().data().addressPointer(),
(LongPointer) op.x().shapeInfoDataBuffer().addressPointer(),
(DoublePointer) op.z().data().addressPointer(),
(LongPointer) op.z().shapeInfoDataBuffer().addressPointer(),
(DoublePointer) op.y().data().addressPointer(), (DoublePointer) getPointerForExtraArgs(op),
(IntPointer) Nd4j.getConstantHandler().getConstantBuffer(dimension).addressPointer(), dimension.length);
}
}
private void exec(ScalarOp op) {
if (executionMode() == ExecutionMode.JAVA) {
super.exec(op);
} else {
long st = profilingHookIn(op);
validateDataType(Nd4j.dataType(), op);
if (op.x().lengthLong() != op.z().lengthLong())
throw new ND4JIllegalStateException("op.X length should be equal to op.Z length: " +
"x.length()=" + op.x().length() + ", z.length()=" + op.z().length() + " - x shape info = ["
+ Arrays.toString(op.x().shapeInfoDataBuffer().asInt()) + "], z shape info = ["
+ Arrays.toString(op.z().shapeInfoDataBuffer().asInt()) + "]");
if (op.getDimension() != null) {
invoke(op, op.getDimension());
return;
}
if (op.x().data().dataType() == DataBuffer.Type.DOUBLE) {
if (op.x().elementWiseStride() >= 1 && !op.isExecSpecial() && op.z().elementWiseStride() >= 1 && !op.isExecSpecial() && op.x().ordering() == op.z().ordering()) {
loop.execScalarDouble(null, op.opNum(), (DoublePointer) op.x().data().addressPointer(),
op.x().elementWiseStride(),
(DoublePointer) op.z().data().addressPointer(),
op.z().elementWiseStride(), op.scalar().doubleValue(),
(DoublePointer) getPointerForExtraArgs(op), op.n());
} else
loop.execScalarDouble(null,
op.opNum(),
(DoublePointer) op.x().data().addressPointer(),
(LongPointer) op.x().shapeInfoDataBuffer().addressPointer(),
(DoublePointer) op.z().data().addressPointer(),
(LongPointer) op.z().shapeInfoDataBuffer().addressPointer(),
op.scalar().doubleValue(),
(DoublePointer) getPointerForExtraArgs(op));
} else {
if (op.x().elementWiseStride() >= 1 && !op.isExecSpecial() && op.z().elementWiseStride() >= 1 && !op.isExecSpecial() && op.x().ordering() == op.z().ordering()) {
loop.execScalarFloat(null, op.opNum(), (FloatPointer) op.x().data().addressPointer(),
op.x().elementWiseStride(), (FloatPointer) op.z().data().addressPointer(),
op.z().elementWiseStride(), op.scalar().floatValue(),
(FloatPointer) getPointerForExtraArgs(op), op.n());
} else
loop.execScalarFloat(null, op.opNum(), (FloatPointer) op.x().data().addressPointer(),
(LongPointer) op.x().shapeInfoDataBuffer().addressPointer(),
(FloatPointer) op.z().data().addressPointer(),
(LongPointer) op.z().shapeInfoDataBuffer().addressPointer(),
op.scalar().floatValue(), (FloatPointer) getPointerForExtraArgs(op));
}
profilingHookOut(op, st);
}
}
private Pointer getPointerForExtraArgs(Op op) {
if (op.extraArgs() != null)
return op.extraArgsDataBuff().addressPointer();
return null;
}
private void exec(TransformOp op) {
long st = 0;
validateDataType(Nd4j.dataType(), op);
if (extraz.get() == null)
extraz.set(new PointerPointer(32));
PointerPointer dummy = extraz.get();
// Pow operations might be special
if (op.opNum() == 7) {
if (op.y() != null && op.y().isScalar()) {
op.setY(Nd4j.valueArrayOf(op.x().shape(), op.y().getDouble(0)));
}
}
/**
* This is the {@link org.nd4j.linalg.api.ops.impl.transforms.IsMax}
* operation.
*
* @see {@link Op#extraArgs()}
* for what an extra argument is in an op.
*
* The extra argument in the op here is the {@link org.nd4j.linalg.api.ops.impl.transforms.IsMax#IsMax(INDArray, int...)}
* dimension to do the ismax along
*/
if (op.opNum() == 41 && op.extraArgs() != null) {
int[] dimension = new int[(int) op.extraArgs()[0]];
for (int i = 0; i < dimension.length; i++) {
dimension[i] = (int) op.extraArgs()[i + 1];
}
/**
* Returns the {@link Shape#createShapeInformation(int[], int[], int, int, char)}
* and the associated offsets for each {@link INDArray#tensorAlongDimension(int, int...)}
* The first item is the shape information. The second one is the offsets.
*/
Pair tadBuffers = tadManager.getTADOnlyShapeInfo(op.z(), dimension);
Pointer tad = tadBuffers.getFirst().addressPointer();
DataBuffer offsets = tadBuffers.getSecond();
Pointer off = offsets == null ? null : offsets.addressPointer();
dummy.put(0, tad);
dummy.put(1, off);
st = profilingHookIn(op, tadBuffers.getFirst());
} else
st = profilingHookIn(op);
if (op.x().data().dataType() == DataBuffer.Type.DOUBLE) {
if (op.y() != null) {
int xEWS = op.x().elementWiseStride();
int yEWS = op.y().elementWiseStride();
int zEWS = op.z().elementWiseStride();
boolean xRow = op.x().isRowVector();
boolean yRow = op.y().isRowVector();
boolean zRow = op.z().isRowVector();
if (op.x().length() != op.y().length() || op.x().length() != op.z().length())
throw new ND4JIllegalStateException("X, Y and Z arguments should have the same length for PairwiseTransform " +
op.opName() + ". x: length " + op.x().length() + ", shape " + Arrays.toString(op.x().shape()) +
"; y: " + op.y().length() + ", shape " + Arrays.toString(op.y().shape()) +
"; z: " + op.z().length() + ", shape " + Arrays.toString(op.z().shape()));
if ((xEWS >= 1 && yEWS >= 1
&& xEWS == yEWS && !op.isExecSpecial()
&& op.x().ordering() == op.y().ordering() && op.x().ordering() == op.z().ordering()) || (xEWS >= 1 && yEWS == xEWS && zEWS == xEWS && xRow && yRow && zRow)) {
loop.execPairwiseTransformDouble(dummy, op.opNum(), (DoublePointer) op.x().data().addressPointer(),
xEWS, (DoublePointer) op.y().data().addressPointer(),
yEWS, (DoublePointer) op.z().data().addressPointer(),
zEWS, (DoublePointer) getPointerForExtraArgs(op), op.n());
} else {
loop.execPairwiseTransformDouble(dummy, op.opNum(), (DoublePointer) op.x().data().addressPointer(),
(LongPointer) op.x().shapeInfoDataBuffer().addressPointer(),
(DoublePointer) op.y().data().addressPointer(),
(LongPointer) op.y().shapeInfoDataBuffer().addressPointer(),
(DoublePointer) op.z().data().addressPointer(),
(LongPointer) op.z().shapeInfoDataBuffer().addressPointer(),
(DoublePointer) getPointerForExtraArgs(op));
}
} else {
if (op.x().elementWiseStride() >= 1 && !op.isExecSpecial() && !op.isExecSpecial()
&& op.x().ordering() == op.z().ordering()) {
loop.execTransformDouble(dummy, op.opNum(), (DoublePointer) op.x().data().addressPointer(),
op.x().elementWiseStride(), (DoublePointer) op.z().data().addressPointer(),
op.z().elementWiseStride(), (DoublePointer) getPointerForExtraArgs(op), op.n());
} else {
loop.execTransformDouble(dummy, op.opNum(), (DoublePointer) op.x().data().addressPointer(),
(LongPointer) op.x().shapeInfoDataBuffer().addressPointer(),
(DoublePointer) op.z().data().addressPointer(),
(LongPointer) op.z().shapeInfoDataBuffer().addressPointer(),
(DoublePointer) getPointerForExtraArgs(op));
}
}
} else {
if (op.y() != null) {
int xEWS = op.x().elementWiseStride();
int yEWS = op.y().elementWiseStride();
int zEWS = op.z().elementWiseStride();
boolean xRow = op.x().isRowVector();
boolean yRow = op.y().isRowVector();
boolean zRow = op.z().isRowVector();
if (op.x().length() != op.y().length() || op.x().length() != op.z().length())
throw new ND4JIllegalStateException("X, Y and Z arguments should have the same length for PairwiseTransform. " +
"x: length " + op.x().length() + ", shape " + Arrays.toString(op.x().shape()) +
"; y: " + op.y().length() + ", shape " + Arrays.toString(op.y().shape()) +
"; z: " + op.z().length() + ", shape " + Arrays.toString(op.z().shape()));
if ((xEWS >= 1 && yEWS >= 1
&& xEWS == yEWS && !op.isExecSpecial()
&& op.x().ordering() == op.y().ordering() && op.x().ordering() == op.z().ordering()) || (xEWS >= 1 && yEWS == xEWS && zEWS == xEWS && xRow && yRow && zRow)) {
loop.execPairwiseTransformFloat(dummy, op.opNum(), (FloatPointer) op.x().data().addressPointer(),
xEWS, (FloatPointer) op.y().data().addressPointer(),
yEWS, (FloatPointer) op.z().data().addressPointer(),
zEWS, (FloatPointer) getPointerForExtraArgs(op), op.n());
} else {
loop.execPairwiseTransformFloat(dummy, op.opNum(), (FloatPointer) op.x().data().addressPointer(),
(LongPointer) op.x().shapeInfoDataBuffer().addressPointer(),
(FloatPointer) op.y().data().addressPointer(),
(LongPointer) op.y().shapeInfoDataBuffer().addressPointer(),
(FloatPointer) op.z().data().addressPointer(),
(LongPointer) op.z().shapeInfoDataBuffer().addressPointer(),
(FloatPointer) getPointerForExtraArgs(op));
}
} else {
if (op.x().elementWiseStride() >= 1 && !op.isExecSpecial() && op.x().ordering() == op.z().ordering()) {
loop.execTransformFloat(dummy, op.opNum(), (FloatPointer) op.x().data().addressPointer(),
op.x().elementWiseStride(), (FloatPointer) op.z().data().addressPointer(),
op.z().elementWiseStride(), (FloatPointer) getPointerForExtraArgs(op), op.n());
} else {
loop.execTransformFloat(dummy, op.opNum(), (FloatPointer) op.x().data().addressPointer(),
(LongPointer) op.x().shapeInfoDataBuffer().addressPointer(),
(FloatPointer) op.z().data().addressPointer(),
(LongPointer) op.z().shapeInfoDataBuffer().addressPointer(),
(FloatPointer) getPointerForExtraArgs(op));
}
}
}
profilingHookOut(op, st);
}
@Override
public INDArray exec(BroadcastOp op, int... dimension) {
long st = profilingHookIn(op);
if(dimension == null)
dimension = new int[] {Integer.MAX_VALUE};
dimension = Shape.normalizeAxis(op.x().rank(), dimension);
validateDataType(Nd4j.dataType(), op);
for (int i = 0; i < dimension.length; i++)
if (dimension[i] >= op.x().rank() && dimension[i] != Integer.MAX_VALUE)
throw new ND4JIllegalStateException("Op target dimension " + Arrays.toString(dimension)
+ " contains element that higher then rank of op.X: [" + op.x().rank() + "]");
/**
* Returns the {@link Shape#createShapeInformation(int[], int[], int, int, char)}
* and the associated offsets for each {@link INDArray#tensorAlongDimension(int, int...)}
* The first item is the shape information. The second one is the offsets.
*/
Pair tadBuffers = tadManager.getTADOnlyShapeInfo(op.x(), dimension);
Pointer hostTadShapeInfo = tadBuffers.getFirst().addressPointer();
Pointer hostTadOffsets = tadBuffers.getSecond().addressPointer();
Pointer devTadShapeInfoZ = null;
Pointer devTadOffsetsZ = null;
// if (!Arrays.equals(op.x().shape(),op.z().shape()) || !Arrays.equals(op.x().stride(),op.z().stride()) || op.x().ordering() != op.z().ordering()) {
// that's the place where we're going to have second TAD in place
Pair tadBuffersZ = tadManager.getTADOnlyShapeInfo(op.z(), dimension);
devTadShapeInfoZ = tadBuffersZ.getFirst().addressPointer();
devTadOffsetsZ = tadBuffersZ.getSecond().addressPointer();
/*
log.info("Broascast dimension: {}", Arrays.toString(dimension));
log.info("x shape: {}; x TAD: {}; comp TAD: {}", Arrays.toString(op.x().shapeInfoDataBuffer().asInt()), Arrays.toString(tadBuffers.getFirst().asInt()), Arrays.toString(op.x().tensorAlongDimension(0, dimension).shapeInfoDataBuffer().asInt()));
log.info("z shape: {}; z TAD: {}", Arrays.toString(op.z().shapeInfoDataBuffer().asInt()), Arrays.toString(tadBuffersZ.getFirst().asInt()));
log.info("y shape: {}", Arrays.toString(op.y().shapeInfoDataBuffer().asInt()));
log.info("-------------");
*/
if (extraz.get() == null)
extraz.set(new PointerPointer(32));
PointerPointer dummy = extraz.get().put(hostTadShapeInfo, hostTadOffsets, devTadShapeInfoZ, devTadOffsetsZ);
Pointer dimensionAddress = constantHandler.getConstantBuffer(dimension).addressPointer();
if (op.x().data().dataType() == DataBuffer.Type.DOUBLE) {
loop.execBroadcastDouble(dummy, op.opNum(), (DoublePointer) op.x().data().addressPointer(),
(LongPointer) op.x().shapeInfoDataBuffer().addressPointer(),
(DoublePointer) op.y().data().addressPointer(),
(LongPointer) op.y().shapeInfoDataBuffer().addressPointer(),
(DoublePointer) op.z().data().addressPointer(),
(LongPointer) op.z().shapeInfoDataBuffer().addressPointer(), (IntPointer) dimensionAddress,
dimension.length);
} else {
loop.execBroadcastFloat(dummy, op.opNum(), (FloatPointer) op.x().data().addressPointer(),
(LongPointer) op.x().shapeInfoDataBuffer().addressPointer(),
(FloatPointer) op.y().data().addressPointer(),
(LongPointer) op.y().shapeInfoDataBuffer().addressPointer(),
(FloatPointer) op.z().data().addressPointer(),
(LongPointer) op.z().shapeInfoDataBuffer().addressPointer(), (IntPointer) dimensionAddress,
dimension.length);
}
return op.z();
}
private void exec(IndexAccumulation op) {
if (executionMode() == ExecutionMode.JAVA) {
super.exec(op);
} else {
if(op.z() == op.x() || op.z() == null) {
op.setZ(Nd4j.scalar(0.0));
}
long st = profilingHookIn(op);
validateDataType(Nd4j.dataType(), op);
if (op.x().data().dataType() == DataBuffer.Type.DOUBLE) {
op.setFinalResult((int) loop.execIndexReduceScalarDouble(null, op.opNum(),
(DoublePointer) op.x().data().addressPointer(),
(LongPointer) op.x().shapeInfoDataBuffer().addressPointer(),
(DoublePointer) getPointerForExtraArgs(op)));
} else {
op.setFinalResult((int) loop.execIndexReduceScalarFloat(null, op.opNum(),
(FloatPointer) op.x().data().addressPointer(),
(LongPointer) op.x().shapeInfoDataBuffer().addressPointer(),
(FloatPointer) getPointerForExtraArgs(op)));
}
op.z().assign(op.getFinalResult());
profilingHookOut(op, st);
}
}
private void exec(Accumulation op) {
if (executionMode() == ExecutionMode.JAVA) {
super.exec(op);
}
else if(op.isExecSpecial()) {
op.exec();
}
else {
long st = profilingHookIn(op);
validateDataType(Nd4j.dataType(), op);
if(op.z() == op.x()) {
op.setZ(Nd4j.scalar(0.0));
}
// since we're going to call reduceToScalar, we must ensure equal lengths
if (op.y() != null && op.getOpType() == Op.Type.REDUCE3) {
if (op.x().lengthLong() != op.y().lengthLong())
throw new ND4JIllegalStateException("X and Y operands should have equal lengths. X length: " + op.x().lengthLong() +
", X shape: " + Arrays.toString(op.x().shape()) + "; Y length: " + op.y().lengthLong() +
", Y shape: " + Arrays.toString(op.y().shape()));
}
if (op.x().data().dataType() == DataBuffer.Type.DOUBLE) {
if (op instanceof Variance) {
op.setFinalResult(loop.execSummaryStatsScalarDouble(null, op.opNum(),
(DoublePointer) op.x().data().addressPointer(),
(LongPointer) op.x().shapeInfoDataBuffer().addressPointer(),
(DoublePointer) getPointerForExtraArgs(op), true));
op.z().putScalar(0,op.getFinalResult().doubleValue());
} else if (op.y() != null && op.getOpType() == Op.Type.REDUCE3) {
op.setFinalResult(loop.execReduce3ScalarDouble(null, op.opNum(),
(DoublePointer) op.x().data().addressPointer(),
(LongPointer) op.x().shapeInfoDataBuffer().addressPointer(),
(DoublePointer) getPointerForExtraArgs(op),
(DoublePointer) op.y().data().addressPointer(),
(LongPointer) op.y().shapeInfoDataBuffer().addressPointer()));
op.z().putScalar(0,op.getFinalResult().doubleValue());
} else {
op.setFinalResult(loop.execReduceScalarDouble(null, op.opNum(),
(DoublePointer) op.x().data().addressPointer(),
(LongPointer) op.x().shapeInfoDataBuffer().addressPointer(),
(DoublePointer) getPointerForExtraArgs(op)));
op.z().putScalar(0,op.getFinalResult().doubleValue());
}
} else {
if (op instanceof Variance) {
Variance variance = (Variance) op;
op.setFinalResult(loop.execSummaryStatsScalarFloat(null, op.opNum(),
(FloatPointer) op.x().data().addressPointer(),
(LongPointer) op.x().shapeInfoDataBuffer().addressPointer(),
(FloatPointer) getPointerForExtraArgs(op), variance.isBiasCorrected()));
op.z().putScalar(0,op.getFinalResult().floatValue());
} else if (op.y() != null && op.getOpType() == Op.Type.REDUCE3) {
op.setFinalResult(loop.execReduce3ScalarFloat(null, op.opNum(),
(FloatPointer) op.x().data().addressPointer(),
(LongPointer) op.x().shapeInfoDataBuffer().addressPointer(),
(FloatPointer) getPointerForExtraArgs(op),
(FloatPointer) op.y().data().addressPointer(),
(LongPointer) op.y().shapeInfoDataBuffer().addressPointer()));
op.z().putScalar(0,op.getFinalResult().floatValue());
} else {
op.setFinalResult(loop.execReduceScalarFloat(null, op.opNum(),
(FloatPointer) op.x().data().addressPointer(),
(LongPointer) op.x().shapeInfoDataBuffer().addressPointer(),
(FloatPointer) getPointerForExtraArgs(op)));
op.z().putScalar(0,op.getFinalResult().floatValue());
}
}
profilingHookOut(op, st);
}
}
protected Pointer getPointer(Batch batch) {
if (batchPointers.get() == null)
batchPointers.set(new HashMap());
if (!batchPointers.get().containsKey(batch.opNum())) {
val pointer = new IntPointer(batch.getSample().getRequiredBatchMemorySize() / 4 );
batchPointers.get().put(batch.opNum(), pointer);
return pointer;
}
return batchPointers.get().get(batch.opNum());
}
/**
* This method executes previously built batch
*
* @param batch
*/
@Override
public void exec(Batch batch) {
//profilingHookIn(batch);
IntPointer pointer = (IntPointer) getPointer(batch);
int maxTypes = 5;
int maxIntArrays = batch.getSample().maxIntArrays();
int maxArraySize = batch.getSample().maxIntArraySize();
int indexPos = maxTypes * Batch.getBatchLimit();
int intArraysPos = indexPos + (batch.getSample().maxIndexArguments() * Batch.getBatchLimit());
int realPos = (intArraysPos + (maxIntArrays * maxArraySize * Batch.getBatchLimit()))
/ (Nd4j.dataType() == DataBuffer.Type.DOUBLE ? 2 : 1);
int argsPos = (realPos + ((batch.getSample().maxRealArguments() * Batch.getBatchLimit())))
/ (Nd4j.dataType() == DataBuffer.Type.DOUBLE ? 1 : 2);
int shapesPos = argsPos + (batch.getSample().maxArguments() * Batch.getBatchLimit());
for (int i = 0; i < batch.getNumAggregates(); i++) {
T op = batch.getAggregates().get(i);
// put num arguments
int idx = i * maxTypes;
pointer.put(idx, op.getArguments().size());
pointer.put(idx + 1, op.getShapes().size());
pointer.put(idx + 2, op.getIndexingArguments().size());
pointer.put(idx + 3, op.getRealArguments().size());
pointer.put(idx + 4, op.getIntArrayArguments().size());
// putting indexing arguments
for (int e = 0; e < op.getIndexingArguments().size(); e++) {
idx = indexPos + i * batch.getSample().maxIndexArguments();
pointer.put(idx + e, op.getIndexingArguments().get(e));
}
// putting intArray values
int bsize = maxIntArrays * maxArraySize;
for (int e = 0; e < op.getIntArrayArguments().size(); e++) {
int step = (i * bsize) + (e * maxArraySize);
if (op.getIntArrayArguments().get(e) != null)
for (int x = 0; x < op.getIntArrayArguments().get(e).length; x++) {
idx = intArraysPos + step + x;
pointer.put(idx, op.getIntArrayArguments().get(e)[x]);
}
}
// TODO: variable datatype should be handled here
// putting real arguments
if (Nd4j.dataType() == DataBuffer.Type.FLOAT) {
FloatPointer fPtr = new FloatPointer(pointer);
for (int e = 0; e < op.getRealArguments().size(); e++) {
idx = realPos + i * op.maxRealArguments();
fPtr.put(idx + e, op.getRealArguments().get(e).floatValue());
}
} else if (Nd4j.dataType() == DataBuffer.Type.DOUBLE) {
DoublePointer dPtr = new DoublePointer(pointer);
for (int e = 0; e < op.getRealArguments().size(); e++) {
idx = realPos + (i * op.maxRealArguments());
dPtr.put(idx + e, op.getRealArguments().get(e).doubleValue());
}
}
if (extraz.get() == null)
extraz.set(new PointerPointer(32));
// putting arguments pointers
PointerPointer ptrPtr = new PointerPointer(pointer);//extraz.get().put(pointer);
for (int e = 0; e < op.getArguments().size(); e++) {
idx = argsPos + i * batch.getSample().maxArguments();
if (op.getArguments().get(e) != null) {
ptrPtr.put(idx + e, op.getArguments().get(e).data().addressPointer());
}
}
// putting shape pointers
for (int e = 0; e < op.getShapes().size(); e++) {
idx = shapesPos + i * batch.getSample().maxShapes();
if (op.getShapes().get(e) != null)
ptrPtr.put(idx + e, op.getShapes().get(e).addressPointer());
}
}
if (Nd4j.dataType() == DataBuffer.Type.FLOAT) {
loop.execAggregateBatchFloat(null, batch.getNumAggregates(), batch.opNum(),
batch.getSample().maxArguments(), batch.getSample().maxShapes(),
batch.getSample().maxIntArrays(), batch.getSample().maxIntArraySize(),
batch.getSample().maxIndexArguments(), batch.getSample().maxRealArguments(), pointer);
} else if (Nd4j.dataType() == DataBuffer.Type.DOUBLE) {
loop.execAggregateBatchDouble(null, batch.getNumAggregates(), batch.opNum(),
batch.getSample().maxArguments(), batch.getSample().maxShapes(),
batch.getSample().maxIntArrays(), batch.getSample().maxIntArraySize(),
batch.getSample().maxIndexArguments(), batch.getSample().maxRealArguments(), pointer);
} else {
throw new UnsupportedOperationException("Half precision isn't supported on CPU");
}
}
/**
* This method takes arbitrary
* sized list of {@link Aggregate},
* and packs them into batches
* Note here that this is mainly used for random number generation
* for {@link RandomOp} and things like {@link org.nd4j.linalg.api.rng.distribution.Distribution}
* @param batch the list of {@link Aggregate} to
* execute upon
*/
@Override
public void exec(List batch) {
if (batch.size() == 0)
return;
List> batches = Batch.getBatches(batch);
for (Batch single : batches) {
this.exec(single);
}
}
/**
* This method takes arbitrary
* sized list of {@link Aggregate},
* and packs them into batches
* Note here that this is mainly used for random number generation
* for {@link RandomOp} and things like {@link org.nd4j.linalg.api.rng.distribution.Distribution}
* @param op the list of {@link Aggregate} to
* execute upon
*/
@Override
public void exec(Aggregate op) {
// long st = profilingHookIn(op);
if (memoryBlocks.get() == null)
memoryBlocks.set(new HashMap());
if (memoryBlocks.get().get(op.opNum()) == null)
memoryBlocks.get().put(op.opNum(), new AggregateMemoryBlock(op));
AggregateMemoryBlock block = memoryBlocks.get().get(op.opNum());
int numArguments = op.getArguments().size();
int numIndexArguments = op.getIndexingArguments().size();
int numRealArguments = op.getRealArguments().size();
int numShapes = op.getShapes().size();
int numIntArrays = op.getIntArrayArguments().size();
PointerPointer arguments = block.getArgumentsPointer(); //new PointerPointer(numArguments);
List pointers = new ArrayList<>();
PointerPointer intArrays = block.getArraysPointer(); //new PointerPointer(numIntArrays);
for (int x = 0; x < numArguments; x++) {
arguments.put(x, op.getArguments().get(x) == null ? null
: op.getArguments().get(x).data().addressPointer());
}
PointerPointer shapes = block.getShapesPointer(); //new PointerPointer(numShapes);
for (int x = 0; x < numShapes; x++) {
if (op.getShapes().get(x).dataType() != DataBuffer.Type.INT)
throw new RuntimeException("ShapeBuffers should have INT data opType");
shapes.put(x, op.getShapes().get(x) == null ? null : op.getShapes().get(x).addressPointer());
}
//int[] indexes = new int[numIndexArguments];
IntPointer pointer = block.getIndexingPointer();
for (int x = 0; x < numIndexArguments; x++) {
pointer.put(x, op.getIndexingArguments().get(x));
}
//IntPointer pointer = new IntPointer(indexes);
double[] reals = new double[numRealArguments];
for (int x = 0; x < numRealArguments; x++) {
//reals[x] = op.getRealArguments().get(x).doubleValue();
if (Nd4j.dataType() == DataBuffer.Type.FLOAT)
((FloatPointer) block.getRealArgumentsPointer()).put(x, op.getRealArguments().get(x).floatValue());
else
((DoublePointer) block.getRealArgumentsPointer()).put(x, op.getRealArguments().get(x).doubleValue());
}
for (int x = 0; x < numIntArrays; x++) {
IntPointer intPtr = block.getIntArrays().get(x); //new IntPointer(op.getIntArrayArguments().get(x));
intPtr.put(op.getIntArrayArguments().get(x), 0, op.getIntArrayArguments().get(x).length);
intArrays.put(x, intPtr);
pointers.add(intPtr);
}
//INDArray realsBuffer = Nd4j.create(reals);
if (Nd4j.dataType() == DataBuffer.Type.FLOAT) {
loop.execAggregateFloat(null, op.opNum(), arguments, numArguments, shapes, numShapes, pointer,
numIndexArguments, intArrays, numIntArrays, (FloatPointer) block.getRealArgumentsPointer(),
numRealArguments);
} else if (Nd4j.dataType() == DataBuffer.Type.DOUBLE) {
loop.execAggregateDouble(null, op.opNum(), arguments, numArguments, shapes, numShapes, pointer,
numIndexArguments, intArrays, numIntArrays, (DoublePointer) block.getRealArgumentsPointer(),
numRealArguments);
} else {
throw new UnsupportedOperationException("Half precision isn't supported on CPU");
}
}
/**
* This method return set of key/value and
* key/key/value objects,
* describing current environment
*
* @return
*/
@Override
public Properties getEnvironmentInformation() {
Properties properties = super.getEnvironmentInformation();
properties.put(Nd4jEnvironment.BACKEND_KEY, "CPU");
properties.put(Nd4jEnvironment.OMP_THREADS_KEY, loop.ompGetMaxThreads());
properties.put(Nd4jEnvironment.BLAS_THREADS_KEY, Nd4j.factory().blas().getMaxThreads());
properties.put(Nd4jEnvironment.BLAS_VENDOR_KEY, (Nd4j.factory().blas()).getBlasVendor().toString());
properties.put(Nd4jEnvironment.HOST_FREE_MEMORY_KEY, Pointer.maxBytes() - Pointer.totalBytes());
// fill bandwidth information
/*
Note: Environment information is logged as part of ND4J initialization... but PerformanceTracker required
ND4J init to be completed before it can be initialized. Hence we can get a null PerformanceTracker when
OpExecutioner.printEnvironmentInformation() is called as part of ND4J class initialization - even
though PerformanceTracker.getInstance() refers to a static final field (as it may not yet be initialized)
*/
if(PerformanceTracker.getInstance() != null) {
properties.put(Nd4jEnvironment.MEMORY_BANDWIDTH_KEY, PerformanceTracker.getInstance().getCurrentBandwidth());
}
return properties;
}
/**
* This method executes specified RandomOp using default RNG available via Nd4j.getRandom()
*
* @param op
*/
@Override
public INDArray exec(RandomOp op) {
return exec(op, Nd4j.getRandom());
}
/**
* This method executes specific
* RandomOp against specified RNG
*
* @param op
* @param rng
*/
@Override
public INDArray exec(RandomOp op, Random rng) {
if (rng.getStateBuffer() == null)
throw new IllegalStateException(
"You should use one of NativeRandom classes for NativeOperations execution. Op class: " + op.getClass().getName());
long st = profilingHookIn(op);
validateDataType(Nd4j.dataType(), op);
if (op.x() != null && op.y() != null && op.z() != null) {
// triple arg call
if (Nd4j.dataType() == DataBuffer.Type.FLOAT) {
loop.execRandomFloat(null, op.opNum(), rng.getStatePointer(), // rng state ptr
(FloatPointer) op.x().data().addressPointer(),
(LongPointer) op.x().shapeInfoDataBuffer().addressPointer(),
(FloatPointer) op.y().data().addressPointer(),
(LongPointer) op.y().shapeInfoDataBuffer().addressPointer(),
(FloatPointer) op.z().data().addressPointer(),
(LongPointer) op.z().shapeInfoDataBuffer().addressPointer(),
(FloatPointer) op.extraArgsDataBuff().addressPointer());
} else if (Nd4j.dataType() == DataBuffer.Type.DOUBLE) {
loop.execRandomDouble(null, op.opNum(), rng.getStatePointer(), // rng state ptr
(DoublePointer) op.x().data().addressPointer(),
(LongPointer) op.x().shapeInfoDataBuffer().addressPointer(),
(DoublePointer) op.y().data().addressPointer(),
(LongPointer) op.y().shapeInfoDataBuffer().addressPointer(),
(DoublePointer) op.z().data().addressPointer(),
(LongPointer) op.z().shapeInfoDataBuffer().addressPointer(),
(DoublePointer) op.extraArgsDataBuff().addressPointer());
}
} else if (op.x() != null && op.z() != null) {
//double arg call
if (Nd4j.dataType() == DataBuffer.Type.FLOAT) {
loop.execRandomFloat(null, op.opNum(), rng.getStatePointer(), // rng state ptr
(FloatPointer) op.x().data().addressPointer(),
(LongPointer) op.x().shapeInfoDataBuffer().addressPointer(),
(FloatPointer) op.z().data().addressPointer(),
(LongPointer) op.z().shapeInfoDataBuffer().addressPointer(),
(FloatPointer) op.extraArgsDataBuff().addressPointer());
} else if (Nd4j.dataType() == DataBuffer.Type.DOUBLE) {
loop.execRandomDouble(null, op.opNum(), rng.getStatePointer(), // rng state ptr
(DoublePointer) op.x().data().addressPointer(),
(LongPointer) op.x().shapeInfoDataBuffer().addressPointer(),
(DoublePointer) op.z().data().addressPointer(),
(LongPointer) op.z().shapeInfoDataBuffer().addressPointer(),
(DoublePointer) op.extraArgsDataBuff().addressPointer());
}
} else {
// single arg call
if (Nd4j.dataType() == DataBuffer.Type.FLOAT) {
loop.execRandomFloat(null, op.opNum(), rng.getStatePointer(), // rng state ptr
(FloatPointer) op.z().data().addressPointer(),
(LongPointer) op.z().shapeInfoDataBuffer().addressPointer(),
(FloatPointer) op.extraArgsDataBuff().addressPointer());
} else if (Nd4j.dataType() == DataBuffer.Type.DOUBLE) {
loop.execRandomDouble(null, op.opNum(), rng.getStatePointer(), // rng state ptr
(DoublePointer) op.z().data().addressPointer(),
(LongPointer) op.z().shapeInfoDataBuffer().addressPointer(),
(DoublePointer) op.extraArgsDataBuff().addressPointer());
}
}
profilingHookOut(op, st);
return op.z();
}
@Override
public TADManager getTADManager() {
return tadManager;
}
/**
* This class holds memory chunks required for single specific Aggregate op.
* Can be used together with ThreadLocal variables
*/
@Data
private static class AggregateMemoryBlock {
private List intArrays = new ArrayList<>();
private IntPointer indexingPointer;
private Pointer realArgumentsPointer;
private PointerPointer shapesPointer;
private PointerPointer argumentsPointer;
private PointerPointer arraysPointer;
private final int opNum;
private AggregateMemoryBlock(@NonNull Aggregate op) {
opNum = op.opNum();
// creating IntArrays
for (int i = 0; i < op.maxIntArrays(); i++) {
intArrays.add(new IntPointer(op.maxIntArraySize()));
}
// allocating chunk for IndexingArguments
indexingPointer = new IntPointer(op.maxIndexArguments());
// allocating chunk for RealArguments
realArgumentsPointer = Nd4j.dataType() == DataBuffer.Type.DOUBLE ? new DoublePointer(op.maxRealArguments())
: new FloatPointer(op.maxRealArguments());
// allocating chunk for shapesPointer
shapesPointer = new PointerPointer(op.maxShapes());
// allocating chunk for argumentsPointer
argumentsPointer = new PointerPointer(op.maxArguments());
// chunk for intArrays
arraysPointer = new PointerPointer(op.maxIntArrays());
}
@Override
public boolean equals(Object o) {
if (this == o)
return true;
if (o == null || getClass() != o.getClass())
return false;
AggregateMemoryBlock that = (AggregateMemoryBlock) o;
return opNum == that.opNum;
}
@Override
public int hashCode() {
return opNum;
}
}
@Override
public INDArray thresholdEncode(INDArray input, double threshold) {
return thresholdEncode(input, threshold, null);
}
@Override
public INDArray thresholdEncode(INDArray input, double threshold, Integer boundary) {
//val condition = new MatchCondition(input, Conditions.absGreaterThanOrEqual(threshold));
//long t1 = System.currentTimeMillis();
int cntAbs = input.data().dataType() == DataBuffer.Type.FLOAT ? loop.estimateThresholdFloat(null, input.data().addressPointer(), (int) input.length(), (float) threshold)
: input.data().dataType() == DataBuffer.Type.DOUBLE ? loop.estimateThresholdDouble(null, input.data().addressPointer(), (int) input.length(), (float) threshold)
: loop.estimateThresholdHalf(null, input.data().addressPointer(), (int) input.length(), (float) threshold);
//long t2 = System.currentTimeMillis();
if (cntAbs < 2)
return null;
if (boundary != null)
cntAbs = Math.min(cntAbs, boundary);
//log.info("S: {}; T: {}", cntAbs, t2 - t1);
DataBuffer buffer = input.data();
long originalLength = buffer.length() * Nd4j.sizeOfDataType(buffer.dataType());
int compressedLength = cntAbs + 4;
// first 3 elements contain header
DataBuffer encodedBuffer = Nd4j.getMemoryManager().getCurrentWorkspace() == null ? Nd4j.getDataBufferFactory().createInt(4+cntAbs, false) : Nd4j.getDataBufferFactory().createInt(4+cntAbs, false, Nd4j.getMemoryManager().getCurrentWorkspace());
encodedBuffer.put(0, cntAbs);
encodedBuffer.put(1, (int) buffer.length());
encodedBuffer.put(2, Float.floatToIntBits((float) threshold));
// format id
encodedBuffer.put(3, ThresholdCompression.FLEXIBLE_ENCODING);
CompressionDescriptor descriptor = new CompressionDescriptor();
descriptor.setCompressedLength(compressedLength * 4); // sizeOf(INT)
descriptor.setOriginalLength(originalLength);
descriptor.setOriginalElementSize(Nd4j.sizeOfDataType(buffer.dataType()));
descriptor.setNumberOfElements(buffer.length());
descriptor.setCompressionAlgorithm("THRESHOLD");
descriptor.setCompressionType(CompressionType.LOSSLESS);
//CompressedDataBuffer cbuff = new CompressedDataBuffer(pointer, descriptor);
Nd4j.getNDArrayFactory().convertDataEx(AbstractCompressor.getBufferTypeEx(buffer), buffer.addressPointer(), DataBuffer.TypeEx.THRESHOLD, encodedBuffer.addressPointer(), buffer.length());
Nd4j.getAffinityManager().tagLocation(buffer, AffinityManager.Location.HOST);
return Nd4j.createArrayFromShapeBuffer(encodedBuffer, input.shapeInfoDataBuffer());
}
@Override
public INDArray thresholdDecode(INDArray encoded, INDArray target) {
DataBuffer buffer = encoded.data();
if (buffer.dataType() != DataBuffer.Type.INT)
throw new ND4JIllegalStateException("thresholdEncoded array should have dataType of INT");
long compressedLength = buffer.getInt(0);
long originalLength = buffer.getInt(1);
float threshold = buffer.getInt(2);
if (target.lengthLong() != originalLength)
throw new ND4JIllegalStateException("originalLength ["+ originalLength+"] stored in encoded array doesn't match target length ["+ target.lengthLong()+"]");
DataBuffer.TypeEx typeDst = AbstractCompressor.getBufferTypeEx(target.data());
loop.convertTypes(null, DataBuffer.TypeEx.THRESHOLD.ordinal(), buffer.addressPointer(), target.length(), typeDst.ordinal(), target.data().addressPointer());
return target;
}
@Override
public long bitmapEncode(INDArray indArray, INDArray target, double threshold) {
long length = indArray.lengthLong();
long tLen = target.data().length();
if (tLen != (length / 16 + 5))
throw new ND4JIllegalStateException("Length of target array should be " + (length / 16 + 5));
if (target.data().dataType() != DataBuffer.Type.INT)
throw new ND4JIllegalStateException("Target array should have INT dataType");
DataBuffer buffer = target.data();
buffer.put(0, (int) length);
buffer.put(1, (int) length);
buffer.put(2, Float.floatToIntBits((float) threshold));
// format id
buffer.put(3, ThresholdCompression.BITMAP_ENCODING);
long affected = 0;
if (indArray.data().dataType() == DataBuffer.Type.FLOAT) {
affected = loop.encodeBitmapFloat(null, (FloatPointer) indArray.data().addressPointer(), length, (IntPointer) buffer.addressPointer(), (float) threshold);
} else if (indArray.data().dataType() == DataBuffer.Type.DOUBLE) {
affected = loop.encodeBitmapDouble(null, (DoublePointer) indArray.data().addressPointer(), length, (IntPointer) buffer.addressPointer(), (float) threshold);
} else
throw new UnsupportedOperationException("HALF precision isn't supported on CPU yet");
return affected;
}
@Override
public INDArray bitmapDecode(INDArray encoded, INDArray target) {
if (target.data().dataType() == DataBuffer.Type.FLOAT) {
loop.decodeBitmapFloat(null, encoded.data().addressPointer(), target.length(), (FloatPointer) target.data().addressPointer());
}
return target;
}
@Override
public synchronized Map getCustomOperations() {
if (customOps == null) {
String list = loop.getAllCustomOps();
if (list == null || list.isEmpty()) {
log.warn("No customs ops available!");
customOps = Collections.emptyMap();
return customOps;
}
val map = new HashMap();
String[] split = list.split(";");
for (String op : split) {
if (op == null || op.isEmpty())
continue;
String[] another = op.split(":");
CustomOpDescriptor descriptor = CustomOpDescriptor.builder()
.hash(Long.valueOf(another[1]))
.numInputs(Integer.valueOf(another[2]))
.numOutputs(Integer.valueOf(another[3]))
.allowsInplace(Integer.valueOf(another[4]) == 1)
.numTArgs(Integer.valueOf(another[5]))
.numIArgs(Integer.valueOf(another[6]))
.build();
map.put(another[0], descriptor);
}
customOps = Collections.unmodifiableMap(map);
}
return customOps;
}
private PointerPointer getPointerPointerFrom(ThreadLocal> map,int numArguments) {
if(map.get() == null) {
Map store = new HashMap<>();
store.put(numArguments,new PointerPointer(numArguments));
map.set(store);
return map.get().get(numArguments);
}
else if (map.get().get(numArguments) == null) {
PointerPointer pointerPointer = new PointerPointer(numArguments);
map.get().put(numArguments,pointerPointer);
return pointerPointer;
}
return map.get().get(numArguments);
}
private ShortPointer getShortPointerFrom(ThreadLocal> map,int numArguments) {
if(map.get() == null) {
Map store = new HashMap<>();
store.put(numArguments,new ShortPointer(numArguments));
map.set(store);
return map.get().get(numArguments);
}
else if (map.get().get(numArguments) == null) {
ShortPointer pointerPointer = new ShortPointer(numArguments);
map.get().put(numArguments,pointerPointer);
return pointerPointer;
}
return map.get().get(numArguments);
}
private DoublePointer getDoublePointerFrom(ThreadLocal> map,int numArguments) {
if(map.get() == null) {
Map store = new HashMap<>();
store.put(numArguments,new DoublePointer(numArguments));
map.set(store);
return map.get().get(numArguments);
}
else if (map.get().get(numArguments) == null) {
DoublePointer pointerPointer = new DoublePointer(numArguments);
map.get().put(numArguments,pointerPointer);
return pointerPointer;
}
return map.get().get(numArguments);
}
private PointerPointer getInputShapes(int numArguments) {
return getPointerPointerFrom(inputShapes,numArguments);
}
private PointerPointer getInputBuffers(int numArguments) {
return getPointerPointerFrom(inputBuffers,numArguments);
}
private PointerPointer getOutputShapes(int numArguments) {
return getPointerPointerFrom(outputShapes,numArguments);
}
private PointerPointer getOutputBuffers(int numArguments) {
return getPointerPointerFrom(outputBuffers,numArguments);
}
/**
* This method executes given CustomOp
*
* PLEASE NOTE: You're responsible for input/output validation
* @param op
*/
public void exec(@NonNull CustomOp op) {
long st = profilingHookIn(op);
if (op.numOutputArguments() == 0 && !op.isInplaceCall()) {
try {
val list = this.calculateOutputShape(op);
if (list.isEmpty())
throw new ND4JIllegalStateException("Op name " + op.opName() + " failed to execute. You can't execute non-inplace CustomOp without outputs being specified");
for (val shape: list)
op.addOutputArgument(Nd4j.create(shape));
} catch (Exception e) {
throw new ND4JIllegalStateException("Op name " + op.opName() + " failed to execute. You can't execute non-inplace CustomOp without outputs being specified");
}
}
val name = op.opName().toLowerCase();
val hash = op.opHash();
val inputShapes = getInputShapes(op.numInputArguments());
val inputBuffers = getInputBuffers(op.numInputArguments());
int cnt= 0;
val inputArgs = op.inputArguments();
for (val in: inputArgs) {
if(in == null)
throw new NullPointerException("Input argument is null for op " + op.getClass().getName());
if (!in.isEmpty())
inputBuffers.put(cnt, in.data().addressPointer());
inputShapes.put(cnt++, in.shapeInfoDataBuffer().addressPointer());
}
val outputArgs = op.outputArguments();
for(int i = 0; i < outputArgs.length; i++) {
if(outputArgs[i] == null)
throw new ND4JIllegalStateException("Op output arguments must not be null! Op " + op.getClass().getName());
}
val outputShapes = getOutputShapes(op.numOutputArguments());
val outputBuffers = getOutputBuffers(op.numOutputArguments());
cnt= 0;
for (val out: outputArgs) {
outputBuffers.put(cnt, out.data().addressPointer());
outputShapes.put(cnt++, out.shapeInfoDataBuffer().addressPointer());
}
val iArgs = op.numIArguments() > 0 ? new LongPointer(op.numIArguments()) : null;
cnt = 0;
val iArgs1 = op.iArgs();
for (val i: iArgs1)
iArgs.put(cnt++, i);
if (Nd4j.dataType() == DataBuffer.Type.FLOAT) {
val tArgs = op.numTArguments() > 0 ? new FloatPointer(op.numTArguments()) : null;
val tArgs1 = op.tArgs();
cnt = 0;
for (val t: tArgs1)
tArgs.put(cnt++, (float) t);
OpStatus status = OpStatus.byNumber(loop.execCustomOpFloat(null, hash, inputBuffers, inputShapes, op.numInputArguments(), outputBuffers, outputShapes, op.numOutputArguments(), tArgs, op.numTArguments(), iArgs, op.numIArguments(), op.isInplaceCall()));
if (status != OpStatus.ND4J_STATUS_OK)
throw new ND4JIllegalStateException("Op execution failed: " + status + " - op " + op.getClass().getName());
} else if (Nd4j.dataType() == DataBuffer.Type.DOUBLE) {
val tArgs = op.numTArguments() > 0 ? getDoublePointerFrom(tArgsPointer,op.numTArguments()) : null;
val tArgs1 = op.tArgs();
cnt = 0;
for (val t: tArgs1)
tArgs.put(cnt++, t);
val t = op.numInputArguments();
OpStatus status = OpStatus.ND4J_STATUS_OK;
try {
status = OpStatus.byNumber(loop.execCustomOpDouble(
null,
hash,
inputBuffers,
inputShapes,
op.numInputArguments(),
outputBuffers,
outputShapes,
op.numOutputArguments(),
tArgs, op.numTArguments(),
iArgs, op.numIArguments(),
op.isInplaceCall()));
}catch(Exception e) {
log.error("Failed to execute op " + op.opName() + ". Attempted to execute with " +
String.valueOf(op.numInputArguments()) + " inputs, " +
String.valueOf(op.numOutputArguments()) + " outputs, "+
String.valueOf(op.numTArguments()) + " targs and " +
String.valueOf(op.numIArguments()) + " iargs. " +
"Please see above message (printed out from c++) for a possible cause of error.");
throw e;
}
} else if (Nd4j.dataType() == DataBuffer.Type.HALF) {
val tArgs = op.numTArguments() > 0 ? getShortPointerFrom(halfArgsPointer,op.numTArguments()) : null;
cnt = 0;
val tArgs1 = op.tArgs();
for (val t: tArgs1)
tArgs.put(cnt++, ArrayUtil.toHalf(t));
OpStatus status = OpStatus.byNumber(loop.execCustomOpHalf(null, hash, inputBuffers, inputShapes, op.numInputArguments(),
outputBuffers, outputShapes, op.numOutputArguments(), tArgs, op.numTArguments(), iArgs, op.numIArguments(), op.isInplaceCall()));
if (status != OpStatus.ND4J_STATUS_OK)
throw new ND4JIllegalStateException("Op execution failed: " + status + " - op " + op.getClass().getName());
}
profilingHookOut(op, st);
}
protected long[] getShapeFromPointer(LongPointer ptr) {
val rank = (int) ptr.get(0);
long[] array = new long[rank];
for (int i = 0; i < rank; i++) {
array[i] = ptr.get(i+1);
}
return array;
}
@Override
public List calculateOutputShape(@NonNull CustomOp op) {
val lc = op.opName().toLowerCase();
val hash = op.opHash();
val result = new ArrayList();
if(op.numInputArguments() < 1 && op.getDescriptor().getNumInputs() != -2) {
if(log.isTraceEnabled()){
log.trace("Could not calculate output shape for op {}: number of input args was 0",
op.getClass().getName());
}
return Collections.emptyList();
}
val inputBuffers = new PointerPointer<>(op.numInputArguments());
val inputShapes = new PointerPointer<>(op.numInputArguments());
val inputArgs = op.inputArguments();
int cnt= 0;
for (val in: inputArgs) {
if (!in.isEmpty())
inputBuffers.put(cnt, in.data().addressPointer());
inputShapes.put(cnt++, in.shapeInfoDataBuffer().addressPointer());
}
val iArgs = op.numIArguments() > 0 ? new LongPointer(op.numIArguments()) : null;
cnt = 0;
val iArgs1 = op.iArgs();
for (val i: iArgs1)
iArgs.put(cnt++, i);
if (Nd4j.dataType() == DataBuffer.Type.FLOAT) {
val tArgs = op.numTArguments() > 0 ? new FloatPointer(op.numTArguments()) : null;
val tArgs1 = op.tArgs();
cnt = 0;
for (val t: tArgs1)
tArgs.put(cnt++, (float) t);
val ptrptr= (Nd4jCpu.ShapeList) loop.calculateOutputShapesFloat(null,
hash, inputBuffers, inputShapes, op.numInputArguments(),
tArgs, op.numTArguments(), iArgs, op.numIArguments());
if (ptrptr == null)
throw new RuntimeException();
for (int e = 0; e < ptrptr.size(); e++ )
result.add(getShapeFromPointer(new PagedPointer(ptrptr.at(e)).asLongPointer()));
loop.deleteShapeList(ptrptr);
} else if (Nd4j.dataType() == DataBuffer.Type.DOUBLE) {
val tArgs = op.numTArguments() > 0 ? new DoublePointer(op.numTArguments()) : null;
cnt = 0;
val tArgs1 = op.tArgs();
for (val t: tArgs1)
tArgs.put(cnt++, t);
val ptrptr= (Nd4jCpu.ShapeList) loop.calculateOutputShapesDouble(null,
hash, inputBuffers, inputShapes, op.numInputArguments(), tArgs,
op.numTArguments(), iArgs, op.numIArguments());
if (ptrptr == null)
throw new RuntimeException();
for (int e = 0; e < ptrptr.size(); e++ )
result.add(getShapeFromPointer(new PagedPointer(ptrptr.at(e)).asLongPointer()));
loop.deleteShapeList(ptrptr);
} else if (Nd4j.dataType() == DataBuffer.Type.HALF) {
val tArgs = op.numTArguments() > 0 ? new ShortPointer(op.numTArguments()) : null;
cnt = 0;
val tArgs1 = op.tArgs();
for (val t: tArgs1)
tArgs.put(cnt++, ArrayUtil.toHalf(t));
val ptrptr= (Nd4jCpu.ShapeList) loop.calculateOutputShapesHalf(null, hash, inputBuffers, inputShapes, op.numInputArguments(), tArgs, op.numTArguments(), iArgs, op.numIArguments());
if (ptrptr == null)
throw new RuntimeException();
val numOutputs = getCustomOperations().get(lc).getNumOutputs();
for (int e = 0; e < ptrptr.size(); e++ )
result.add(getShapeFromPointer(new PagedPointer(ptrptr.at(e)).asLongPointer()));
loop.deleteShapeList(ptrptr);
}
if(log.isTraceEnabled()){
String[] arr = new String[result.size()];
for( int i=0; i executeGraph(long id, @NonNull Map map, @NonNull Map reverseMap) {
val ptrBuffers = new PointerPointer(map.size());
val ptrShapes = new PointerPointer(map.size());
val ptrIndices = new IntPointer(map.size());
int cnt = 0;
val keySet = new ArrayList(map.keySet());
for (val key: keySet) {
val array = map.get(key);
ptrBuffers.put(cnt, array.data().addressPointer());
ptrShapes.put(cnt, array.shapeInfoDataBuffer().addressPointer());
ptrIndices.put(cnt, reverseMap.get(key));
cnt++;
}
val newMap = new LinkedHashMap();
if (Nd4j.dataType() == DataBuffer.Type.FLOAT) {
val result = (Nd4jCpu.FloatVariablesSet) loop.executeStoredGraphFloat(null, id, ptrBuffers, ptrShapes, ptrIndices, map.size());
val status = OpStatus.byNumber(result.status());
if (status != OpStatus.ND4J_STATUS_OK)
throw new ND4JIllegalStateException("Op execution failed: " + status);
for (int e = 0; e < result.size(); e++) {
val var = result.at(e);
val nodeId = var.id();
val index = var.index();
val nodeName = var.getName().getString();
val shapeInfo = var.getNDArray().shapeInfo();
val buffer = var.getNDArray().buffer();
val rank = (int) shapeInfo.get(0);
val jshape = new long[rank * 2 + 4];
for (int i = 0; i < jshape.length; i++) {
jshape[i] = shapeInfo.get(i);
}
val shapeOf = Shape.shapeOf(jshape);
val stridesOf = Shape.stridesOf(jshape);
val order = Shape.order(jshape);
val array = Nd4j.create(shapeOf, stridesOf, 0, order);
val perfD = PerformanceTracker.getInstance().helperStartTransaction();
Pointer.memcpy(array.data().addressPointer(), buffer, Shape.lengthOf(shapeOf) * Nd4j.sizeOfDataType());
PerformanceTracker.getInstance().helperRegisterTransaction(0, perfD, Shape.lengthOf(shapeOf) * Nd4j.sizeOfDataType(), MemcpyDirection.HOST_TO_HOST);
//newMap.put(keySet.get(e), array);
// if (map.containsKey(nodeName))
newMap.put(nodeName, array);
}
loop.deleteVariablesSetFloat(result);
} else if (Nd4j.dataType() == DataBuffer.Type.DOUBLE) {
val result = (Nd4jCpu.DoubleVariablesSet) loop.executeStoredGraphDouble(null, id, ptrBuffers, ptrShapes, ptrIndices, map.size());
val status = OpStatus.byNumber(result.status());
if (status != OpStatus.ND4J_STATUS_OK)
throw new ND4JIllegalStateException("Op execution failed: " + status);
for (int e = 0; e < result.size(); e++) {
val var = result.at(e);
val nodeId = var.id();
val index = var.index();
val shapeInfo = var.getNDArray().shapeInfo();
val buffer = var.getNDArray().buffer();
val rank = (int) shapeInfo.get(0);
val jshape = new long[rank * 2 + 4];
for (int i = 0; i < jshape.length; i++) {
jshape[i] = shapeInfo.get(i);
}
val shapeOf = Shape.shapeOf(jshape);
val stridesOf = Shape.stridesOf(jshape);
val order = Shape.order(jshape);
val array = Nd4j.create(shapeOf, stridesOf, 0, order);
val perfX = PerformanceTracker.getInstance().helperStartTransaction();
Pointer.memcpy(array.data().addressPointer(), buffer, Shape.lengthOf(shapeOf) * Nd4j.sizeOfDataType());
PerformanceTracker.getInstance().helperRegisterTransaction(0, perfX, Shape.lengthOf(shapeOf) * Nd4j.sizeOfDataType(), MemcpyDirection.HOST_TO_HOST);
//newMap.put(keySet.get(nodeId), array);
val nodeName = var.getName().getString();
newMap.put(nodeName, array);
}
loop.deleteVariablesSetDouble(result);
} else if (Nd4j.dataType() == DataBuffer.Type.HALF) {
val result = (Nd4jCpu.DoubleVariablesSet) loop.executeStoredGraphHalf(null, id, ptrBuffers, ptrShapes, ptrIndices, map.size());
val status = OpStatus.byNumber(result.status());
if (status != OpStatus.ND4J_STATUS_OK)
throw new ND4JIllegalStateException("Op execution failed: " + status);
for (int e = 0; e < result.size(); e++) {
val var = result.at(e);
val nodeId = var.id();
val index = var.index();
val shapeInfo = var.getNDArray().shapeInfo();
val buffer = var.getNDArray().buffer();
val rank = (int) shapeInfo.get(0);
val jshape = new long[rank * 2 + 4];
for (int i = 0; i < jshape.length; i++) {
jshape[i] = shapeInfo.get(i);
}
val shapeOf = Shape.shapeOf(jshape);
val stridesOf = Shape.stridesOf(jshape);
val order = Shape.order(jshape);
val array = Nd4j.create(shapeOf, stridesOf, 0, order);
val perfD = PerformanceTracker.getInstance().helperStartTransaction();
Pointer.memcpy(array.data().addressPointer(), buffer, Shape.lengthOf(shapeOf) * Nd4j.sizeOfDataType());
PerformanceTracker.getInstance().helperRegisterTransaction(0, perfD, Shape.lengthOf(shapeOf) * Nd4j.sizeOfDataType(), MemcpyDirection.HOST_TO_HOST);
//newMap.put(keySet.get(nodeId), array);
val nodeName = var.getName().getString();
newMap.put(nodeName, array);
}
loop.deleteVariablesSetHalf(result);
}
return newMap;
}
@Override
public void forgetGraph(long id) {
loop.unregisterGraph(null, id);
}
/**
* This method allows to set desired number of elements per thread, for performance optimization purposes.
* I.e. if array contains 2048 elements, and threshold is set to 1024, 2 threads will be used for given op execution.
*
* Default value: 1024
*
* @param threshold
*/
@Override
public void setElementsThreshold(int threshold) {
loop.setElementThreshold(threshold);
}
/**
* This method allows to set desired number of sub-arrays per thread, for performance optimization purposes.
* I.e. if matrix has shape of 64 x 128, and threshold is set to 8, each thread will be processing 8 sub-arrays (sure, if you have 8 core cpu).
* If your cpu has, say, 4, cores, only 4 threads will be spawned, and each will process 16 sub-arrays
*
* Default value: 8
*
* @param threshold
*/
@Override
public void setTadThreshold(int threshold) {
loop.setTADThreshold(threshold);
}
private static long[] reductionShape(INDArray x, int[] dimension, boolean newFormat, boolean keepDims){
boolean wholeArray = Shape.wholeArrayDimension(dimension);
long[] retShape;
if(!newFormat) {
retShape = wholeArray ? new long[] {1, 1} : ArrayUtil.removeIndex(x.shape(), dimension);
//ensure vector is proper shape (if old format)
if (retShape.length == 1) {
if (dimension[0] == 0)
retShape = new long[]{1, retShape[0]};
else
retShape = new long[]{retShape[0], 1};
} else if (retShape.length == 0) {
retShape = new long[]{1, 1};
}
} else {
if(keepDims){
retShape = x.shape().clone();
if(wholeArray){
for( int i=0; i