cavis/libnd4j/include/legacy/cpu/NativeOpExecutioner.cpp

1357 lines
61 KiB
C++
Raw Normal View History

2019-06-06 14:21:15 +02:00
/*******************************************************************************
* 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
******************************************************************************/
#include <vector>
#include <system/pointercast.h>
#include "legacy/NativeOpExecutioner.h"
2019-06-06 14:21:15 +02:00
#include <types/types.h>
#include <helpers/LoopKind.h>
#include <loops/pairwise_bool.h>
#include <loops/broadcasting_bool.h>
#include <loops/scalar_bool.h>
2019-06-06 14:21:15 +02:00
#include <loops/pairwise_int.h>
#include <loops/broadcasting_int.h>
#include <loops/scalar_int.h>
2019-06-06 14:21:15 +02:00
#include <loops/transform_float.h>
#include <loops/transform_bool.h>
#include <loops/transform_any.h>
#include <loops/transform_same.h>
#include <loops/transform_strict.h>
#include <loops/reduce_float.h>
#include <loops/reduce_same.h>
#include <loops/reduce_bool.h>
#include <loops/reduce_long.h>
#include <loops/broadcasting.h>
#include <loops/indexreduce.h>
#include <loops/pairwise_transform.h>
#include <loops/reduce_float.h>
#include <loops/reduce3.h>
#include <loops/summarystatsreduce.h>
#include <loops/transform_same.h>
#include <loops/scalar.h>
#include <loops/random.h>
#include <system/pointercast.h>
2019-06-06 14:21:15 +02:00
#include <exceptions/datatype_exception.h>
#include <array/TadPack.h>
#include <helpers/ConstantTadHelper.h>
2019-06-06 14:21:15 +02:00
[WIP] More of CUDA (#95) * initial commit Signed-off-by: raver119 <raver119@gmail.com> * Implementation of hashcode cuda helper. Working edition. * Fixed parallel test input arangements. * Fixed tests for hashcode op. * Fixed shape calculation for image:crop_and_resize op and test. * NativeOps tests. Initial test suite. * Added tests for indexReduce methods. * Added test on execBroadcast with NDArray as dimensions. * Added test on execBroadcastBool with NDArray as dimensions. * Added tests on execPairwiseTransform and execPairwiseTransofrmBool. * Added tests for execReduce with scalar results. * Added reduce tests for non-empty dims array. * Added tests for reduce3. * Added tests for execScalar. * Added tests for execSummaryStats. * - provide cpu/cuda code for batch_to_space - testing it Signed-off-by: Yurii <yurii@skymind.io> * - remove old test for batch_to_space (had wrong format and numbers were not checked) Signed-off-by: Yurii <yurii@skymind.io> * Fixed complilation errors with test. * Added test for execTransformFloat. * Added test for execTransformSame. * Added test for execTransformBool. * Added test for execTransformStrict. * Added tests for execScalar/execScalarBool with TADs. * Added test for flatten. * - provide cpu/cuda code for space_to_Batch operaion Signed-off-by: Yurii <yurii@skymind.io> * Added test for concat. * comment unnecessary stuff in s_t_b Signed-off-by: Yurii <yurii@skymind.io> * Added test for specialConcat. * Added tests for memcpy/set routines. * Fixed pullRow cuda test. * Added pullRow test. * Added average test. * - correct typo in NDArray::applyPairwiseTransform(nd4j::pairwise::BoolOps op...) Signed-off-by: Yurii <yurii@skymind.io> * - debugging and fixing cuda tests in JavaInteropTests file Signed-off-by: Yurii <yurii@skymind.io> * - correct some tests Signed-off-by: Yurii <yurii@skymind.io> * Added test for shuffle. * Fixed ops declarations. * Restored omp and added shuffle test. * Added convertTypes test. * Added tests for execRandom. Eliminated usage of RandomBuffer with NativeOps. * Added sort tests. * Added tests for execCustomOp. * - further debuging and fixing tests terminated with crash Signed-off-by: Yurii <yurii@skymind.io> * Added tests for calculateOutputShapes. * Addded Benchmarks test. * Commented benchmark tests. * change assertion Signed-off-by: raver119 <raver119@gmail.com> * Added tests for apply_sgd op. Added cpu helper for that op. * Implement cuda helper for aplly_sgd op. Fixed tests for NativeOps. * Added test for assign broadcastable. * Added tests for assign_bp op. * Added tests for axpy op. * - assign/execScalar/execTransformAny signature change - minor test fix Signed-off-by: raver119 <raver119@gmail.com> * Fixed axpy op. * meh Signed-off-by: raver119 <raver119@gmail.com> * - fix tests for nativeOps::concat Signed-off-by: Yurii <yurii@skymind.io> * sequential transform/scalar Signed-off-by: raver119 <raver119@gmail.com> * allow nested parallelism Signed-off-by: raver119 <raver119@gmail.com> * assign_bp leak fix Signed-off-by: raver119 <raver119@gmail.com> * block setRNG fix Signed-off-by: raver119 <raver119@gmail.com> * enable parallelism by default Signed-off-by: raver119 <raver119@gmail.com> * enable nested parallelism by default Signed-off-by: raver119 <raver119@gmail.com> * Added cuda implementation for row_count helper. * Added implementation for tnse gains op helper. * - take into account possible situations when input arrays are empty in reduce_ cuda stuff Signed-off-by: Yurii <yurii@skymind.io> * Implemented tsne/edge_forces op cuda-based helper. Parallelized cpu-based helper for edge_forces. * Added kernel for tsne/symmetrized op heleper. * Implementation of tsne/symmetrized op cuda helper. Working edition. * Eliminated waste printfs. * Added test for broadcastgradientargs op. * host-only fallback for empty reduce float Signed-off-by: raver119 <raver119@gmail.com> * - some tests fixes Signed-off-by: Yurii <yurii@skymind.io> * - correct the rest of reduce_ stuff Signed-off-by: Yurii <yurii@skymind.io> * - further correction of reduce_ stuff Signed-off-by: Yurii <yurii@skymind.io> * Added test for Cbow op. Also added cuda implementation for cbow helpers. * - improve code of stack operation for scalar case Signed-off-by: Yurii <yurii@skymind.io> * - provide cuda kernel for gatherND operation Signed-off-by: Yurii <yurii@skymind.io> * Implementation of cbow helpers with cuda kernels. * minor tests tweaks Signed-off-by: raver119 <raver119@gmail.com> * minor tests tweaks Signed-off-by: raver119 <raver119@gmail.com> * - further correction of cuda stuff Signed-off-by: Yurii <yurii@skymind.io> * Implementatation of cbow op helper with cuda kernels. Working edition. * Skip random testing for cudablas case. * lstmBlockCell context fix Signed-off-by: raver119 <raver119@gmail.com> * Added tests for ELU and ELU_BP ops. * Added tests for eq_scalar, gt_scalar, gte_scalar and lte_scalar ops. * Added tests for neq_scalar. * Added test for noop. * - further work on clipbynorm_bp Signed-off-by: Yurii <yurii@skymind.io> * - get rid of concat op call, use instead direct concat helper call Signed-off-by: Yurii <yurii@skymind.io> * lstmBlockCell context fix Signed-off-by: raver119 <raver119@gmail.com> * Added tests for lrelu and lrelu_bp. * Added tests for selu and selu_bp. * Fixed lrelu derivative helpers. * - some corrections in lstm Signed-off-by: Yurii <yurii@skymind.io> * operator * result shape fix Signed-off-by: raver119 <raver119@gmail.com> * - correct typo in lstmCell Signed-off-by: Yurii <yurii@skymind.io> * few tests fixed Signed-off-by: raver119 <raver119@gmail.com> * CUDA inverse broadcast bool fix Signed-off-by: raver119 <raver119@gmail.com> * disable MMAP test for CUDA Signed-off-by: raver119 <raver119@gmail.com> * BooleanOp syncToDevice Signed-off-by: raver119 <raver119@gmail.com> * meh Signed-off-by: raver119 <raver119@gmail.com> * additional data types for im2col/col2im Signed-off-by: raver119 <raver119@gmail.com> * Added test for firas_sparse op. * one more RandomBuffer test excluded Signed-off-by: raver119 <raver119@gmail.com> * Added tests for flatten op. * Added test for Floor op. * bunch of tests fixed Signed-off-by: raver119 <raver119@gmail.com> * mmulDot tests fixed Signed-off-by: raver119 <raver119@gmail.com> * more tests fixed Signed-off-by: raver119 <raver119@gmail.com> * Implemented floordiv_bp op and tests. * Fixed scalar case with cuda implementation for bds. * - work on cuda kernel for clip_by_norm backprop op is completed Signed-off-by: Yurii <yurii@skymind.io> * Eliminate cbow crach. * more tests fixed Signed-off-by: raver119 <raver119@gmail.com> * more tests fixed Signed-off-by: raver119 <raver119@gmail.com> * Eliminated abortion with batched nlp test. * more tests fixed Signed-off-by: raver119 <raver119@gmail.com> * Fixed shared flag initializing. * disabled bunch of cpu workspaces tests Signed-off-by: raver119 <raver119@gmail.com> * scalar operators fix: missing registerSpecialUse call Signed-off-by: raver119 <raver119@gmail.com> * Fixed logdet for cuda and tests. * - correct clipBynorm_bp Signed-off-by: Yurii <yurii@skymind.io> * Fixed crop_and_resize shape datatype. * - correct some mmul tests Signed-off-by: Yurii <yurii@skymind.io>
2019-08-02 19:01:03 +02:00
#ifdef _OPENMP
#include <omp.h>
#include <helpers/ConstantTadHelper.h>
[WIP] More of CUDA (#95) * initial commit Signed-off-by: raver119 <raver119@gmail.com> * Implementation of hashcode cuda helper. Working edition. * Fixed parallel test input arangements. * Fixed tests for hashcode op. * Fixed shape calculation for image:crop_and_resize op and test. * NativeOps tests. Initial test suite. * Added tests for indexReduce methods. * Added test on execBroadcast with NDArray as dimensions. * Added test on execBroadcastBool with NDArray as dimensions. * Added tests on execPairwiseTransform and execPairwiseTransofrmBool. * Added tests for execReduce with scalar results. * Added reduce tests for non-empty dims array. * Added tests for reduce3. * Added tests for execScalar. * Added tests for execSummaryStats. * - provide cpu/cuda code for batch_to_space - testing it Signed-off-by: Yurii <yurii@skymind.io> * - remove old test for batch_to_space (had wrong format and numbers were not checked) Signed-off-by: Yurii <yurii@skymind.io> * Fixed complilation errors with test. * Added test for execTransformFloat. * Added test for execTransformSame. * Added test for execTransformBool. * Added test for execTransformStrict. * Added tests for execScalar/execScalarBool with TADs. * Added test for flatten. * - provide cpu/cuda code for space_to_Batch operaion Signed-off-by: Yurii <yurii@skymind.io> * Added test for concat. * comment unnecessary stuff in s_t_b Signed-off-by: Yurii <yurii@skymind.io> * Added test for specialConcat. * Added tests for memcpy/set routines. * Fixed pullRow cuda test. * Added pullRow test. * Added average test. * - correct typo in NDArray::applyPairwiseTransform(nd4j::pairwise::BoolOps op...) Signed-off-by: Yurii <yurii@skymind.io> * - debugging and fixing cuda tests in JavaInteropTests file Signed-off-by: Yurii <yurii@skymind.io> * - correct some tests Signed-off-by: Yurii <yurii@skymind.io> * Added test for shuffle. * Fixed ops declarations. * Restored omp and added shuffle test. * Added convertTypes test. * Added tests for execRandom. Eliminated usage of RandomBuffer with NativeOps. * Added sort tests. * Added tests for execCustomOp. * - further debuging and fixing tests terminated with crash Signed-off-by: Yurii <yurii@skymind.io> * Added tests for calculateOutputShapes. * Addded Benchmarks test. * Commented benchmark tests. * change assertion Signed-off-by: raver119 <raver119@gmail.com> * Added tests for apply_sgd op. Added cpu helper for that op. * Implement cuda helper for aplly_sgd op. Fixed tests for NativeOps. * Added test for assign broadcastable. * Added tests for assign_bp op. * Added tests for axpy op. * - assign/execScalar/execTransformAny signature change - minor test fix Signed-off-by: raver119 <raver119@gmail.com> * Fixed axpy op. * meh Signed-off-by: raver119 <raver119@gmail.com> * - fix tests for nativeOps::concat Signed-off-by: Yurii <yurii@skymind.io> * sequential transform/scalar Signed-off-by: raver119 <raver119@gmail.com> * allow nested parallelism Signed-off-by: raver119 <raver119@gmail.com> * assign_bp leak fix Signed-off-by: raver119 <raver119@gmail.com> * block setRNG fix Signed-off-by: raver119 <raver119@gmail.com> * enable parallelism by default Signed-off-by: raver119 <raver119@gmail.com> * enable nested parallelism by default Signed-off-by: raver119 <raver119@gmail.com> * Added cuda implementation for row_count helper. * Added implementation for tnse gains op helper. * - take into account possible situations when input arrays are empty in reduce_ cuda stuff Signed-off-by: Yurii <yurii@skymind.io> * Implemented tsne/edge_forces op cuda-based helper. Parallelized cpu-based helper for edge_forces. * Added kernel for tsne/symmetrized op heleper. * Implementation of tsne/symmetrized op cuda helper. Working edition. * Eliminated waste printfs. * Added test for broadcastgradientargs op. * host-only fallback for empty reduce float Signed-off-by: raver119 <raver119@gmail.com> * - some tests fixes Signed-off-by: Yurii <yurii@skymind.io> * - correct the rest of reduce_ stuff Signed-off-by: Yurii <yurii@skymind.io> * - further correction of reduce_ stuff Signed-off-by: Yurii <yurii@skymind.io> * Added test for Cbow op. Also added cuda implementation for cbow helpers. * - improve code of stack operation for scalar case Signed-off-by: Yurii <yurii@skymind.io> * - provide cuda kernel for gatherND operation Signed-off-by: Yurii <yurii@skymind.io> * Implementation of cbow helpers with cuda kernels. * minor tests tweaks Signed-off-by: raver119 <raver119@gmail.com> * minor tests tweaks Signed-off-by: raver119 <raver119@gmail.com> * - further correction of cuda stuff Signed-off-by: Yurii <yurii@skymind.io> * Implementatation of cbow op helper with cuda kernels. Working edition. * Skip random testing for cudablas case. * lstmBlockCell context fix Signed-off-by: raver119 <raver119@gmail.com> * Added tests for ELU and ELU_BP ops. * Added tests for eq_scalar, gt_scalar, gte_scalar and lte_scalar ops. * Added tests for neq_scalar. * Added test for noop. * - further work on clipbynorm_bp Signed-off-by: Yurii <yurii@skymind.io> * - get rid of concat op call, use instead direct concat helper call Signed-off-by: Yurii <yurii@skymind.io> * lstmBlockCell context fix Signed-off-by: raver119 <raver119@gmail.com> * Added tests for lrelu and lrelu_bp. * Added tests for selu and selu_bp. * Fixed lrelu derivative helpers. * - some corrections in lstm Signed-off-by: Yurii <yurii@skymind.io> * operator * result shape fix Signed-off-by: raver119 <raver119@gmail.com> * - correct typo in lstmCell Signed-off-by: Yurii <yurii@skymind.io> * few tests fixed Signed-off-by: raver119 <raver119@gmail.com> * CUDA inverse broadcast bool fix Signed-off-by: raver119 <raver119@gmail.com> * disable MMAP test for CUDA Signed-off-by: raver119 <raver119@gmail.com> * BooleanOp syncToDevice Signed-off-by: raver119 <raver119@gmail.com> * meh Signed-off-by: raver119 <raver119@gmail.com> * additional data types for im2col/col2im Signed-off-by: raver119 <raver119@gmail.com> * Added test for firas_sparse op. * one more RandomBuffer test excluded Signed-off-by: raver119 <raver119@gmail.com> * Added tests for flatten op. * Added test for Floor op. * bunch of tests fixed Signed-off-by: raver119 <raver119@gmail.com> * mmulDot tests fixed Signed-off-by: raver119 <raver119@gmail.com> * more tests fixed Signed-off-by: raver119 <raver119@gmail.com> * Implemented floordiv_bp op and tests. * Fixed scalar case with cuda implementation for bds. * - work on cuda kernel for clip_by_norm backprop op is completed Signed-off-by: Yurii <yurii@skymind.io> * Eliminate cbow crach. * more tests fixed Signed-off-by: raver119 <raver119@gmail.com> * more tests fixed Signed-off-by: raver119 <raver119@gmail.com> * Eliminated abortion with batched nlp test. * more tests fixed Signed-off-by: raver119 <raver119@gmail.com> * Fixed shared flag initializing. * disabled bunch of cpu workspaces tests Signed-off-by: raver119 <raver119@gmail.com> * scalar operators fix: missing registerSpecialUse call Signed-off-by: raver119 <raver119@gmail.com> * Fixed logdet for cuda and tests. * - correct clipBynorm_bp Signed-off-by: Yurii <yurii@skymind.io> * Fixed crop_and_resize shape datatype. * - correct some mmul tests Signed-off-by: Yurii <yurii@skymind.io>
2019-08-02 19:01:03 +02:00
#endif
2019-06-06 14:21:15 +02:00
////////////////////////////////////////////////////////////////////////
/**
*
* @param opNum
* @param hX
* @param hXShapeInfo
* @param extraParams
* @param hZ
* @param hZShapeInfo
*/
void NativeOpExecutioner::execIndexReduceScalar(sd::LaunchContext *lc, int opNum,
2019-06-06 14:21:15 +02:00
void *hX, Nd4jLong *hXShapeInfo,
void *dX, Nd4jLong *dXShapeInfo,
void *extraParams,
void *hZ, Nd4jLong *hZShapeInfo,
void *dZ, Nd4jLong *dZShapeInfo) {
[WIP] More of CUDA (#95) * initial commit Signed-off-by: raver119 <raver119@gmail.com> * Implementation of hashcode cuda helper. Working edition. * Fixed parallel test input arangements. * Fixed tests for hashcode op. * Fixed shape calculation for image:crop_and_resize op and test. * NativeOps tests. Initial test suite. * Added tests for indexReduce methods. * Added test on execBroadcast with NDArray as dimensions. * Added test on execBroadcastBool with NDArray as dimensions. * Added tests on execPairwiseTransform and execPairwiseTransofrmBool. * Added tests for execReduce with scalar results. * Added reduce tests for non-empty dims array. * Added tests for reduce3. * Added tests for execScalar. * Added tests for execSummaryStats. * - provide cpu/cuda code for batch_to_space - testing it Signed-off-by: Yurii <yurii@skymind.io> * - remove old test for batch_to_space (had wrong format and numbers were not checked) Signed-off-by: Yurii <yurii@skymind.io> * Fixed complilation errors with test. * Added test for execTransformFloat. * Added test for execTransformSame. * Added test for execTransformBool. * Added test for execTransformStrict. * Added tests for execScalar/execScalarBool with TADs. * Added test for flatten. * - provide cpu/cuda code for space_to_Batch operaion Signed-off-by: Yurii <yurii@skymind.io> * Added test for concat. * comment unnecessary stuff in s_t_b Signed-off-by: Yurii <yurii@skymind.io> * Added test for specialConcat. * Added tests for memcpy/set routines. * Fixed pullRow cuda test. * Added pullRow test. * Added average test. * - correct typo in NDArray::applyPairwiseTransform(nd4j::pairwise::BoolOps op...) Signed-off-by: Yurii <yurii@skymind.io> * - debugging and fixing cuda tests in JavaInteropTests file Signed-off-by: Yurii <yurii@skymind.io> * - correct some tests Signed-off-by: Yurii <yurii@skymind.io> * Added test for shuffle. * Fixed ops declarations. * Restored omp and added shuffle test. * Added convertTypes test. * Added tests for execRandom. Eliminated usage of RandomBuffer with NativeOps. * Added sort tests. * Added tests for execCustomOp. * - further debuging and fixing tests terminated with crash Signed-off-by: Yurii <yurii@skymind.io> * Added tests for calculateOutputShapes. * Addded Benchmarks test. * Commented benchmark tests. * change assertion Signed-off-by: raver119 <raver119@gmail.com> * Added tests for apply_sgd op. Added cpu helper for that op. * Implement cuda helper for aplly_sgd op. Fixed tests for NativeOps. * Added test for assign broadcastable. * Added tests for assign_bp op. * Added tests for axpy op. * - assign/execScalar/execTransformAny signature change - minor test fix Signed-off-by: raver119 <raver119@gmail.com> * Fixed axpy op. * meh Signed-off-by: raver119 <raver119@gmail.com> * - fix tests for nativeOps::concat Signed-off-by: Yurii <yurii@skymind.io> * sequential transform/scalar Signed-off-by: raver119 <raver119@gmail.com> * allow nested parallelism Signed-off-by: raver119 <raver119@gmail.com> * assign_bp leak fix Signed-off-by: raver119 <raver119@gmail.com> * block setRNG fix Signed-off-by: raver119 <raver119@gmail.com> * enable parallelism by default Signed-off-by: raver119 <raver119@gmail.com> * enable nested parallelism by default Signed-off-by: raver119 <raver119@gmail.com> * Added cuda implementation for row_count helper. * Added implementation for tnse gains op helper. * - take into account possible situations when input arrays are empty in reduce_ cuda stuff Signed-off-by: Yurii <yurii@skymind.io> * Implemented tsne/edge_forces op cuda-based helper. Parallelized cpu-based helper for edge_forces. * Added kernel for tsne/symmetrized op heleper. * Implementation of tsne/symmetrized op cuda helper. Working edition. * Eliminated waste printfs. * Added test for broadcastgradientargs op. * host-only fallback for empty reduce float Signed-off-by: raver119 <raver119@gmail.com> * - some tests fixes Signed-off-by: Yurii <yurii@skymind.io> * - correct the rest of reduce_ stuff Signed-off-by: Yurii <yurii@skymind.io> * - further correction of reduce_ stuff Signed-off-by: Yurii <yurii@skymind.io> * Added test for Cbow op. Also added cuda implementation for cbow helpers. * - improve code of stack operation for scalar case Signed-off-by: Yurii <yurii@skymind.io> * - provide cuda kernel for gatherND operation Signed-off-by: Yurii <yurii@skymind.io> * Implementation of cbow helpers with cuda kernels. * minor tests tweaks Signed-off-by: raver119 <raver119@gmail.com> * minor tests tweaks Signed-off-by: raver119 <raver119@gmail.com> * - further correction of cuda stuff Signed-off-by: Yurii <yurii@skymind.io> * Implementatation of cbow op helper with cuda kernels. Working edition. * Skip random testing for cudablas case. * lstmBlockCell context fix Signed-off-by: raver119 <raver119@gmail.com> * Added tests for ELU and ELU_BP ops. * Added tests for eq_scalar, gt_scalar, gte_scalar and lte_scalar ops. * Added tests for neq_scalar. * Added test for noop. * - further work on clipbynorm_bp Signed-off-by: Yurii <yurii@skymind.io> * - get rid of concat op call, use instead direct concat helper call Signed-off-by: Yurii <yurii@skymind.io> * lstmBlockCell context fix Signed-off-by: raver119 <raver119@gmail.com> * Added tests for lrelu and lrelu_bp. * Added tests for selu and selu_bp. * Fixed lrelu derivative helpers. * - some corrections in lstm Signed-off-by: Yurii <yurii@skymind.io> * operator * result shape fix Signed-off-by: raver119 <raver119@gmail.com> * - correct typo in lstmCell Signed-off-by: Yurii <yurii@skymind.io> * few tests fixed Signed-off-by: raver119 <raver119@gmail.com> * CUDA inverse broadcast bool fix Signed-off-by: raver119 <raver119@gmail.com> * disable MMAP test for CUDA Signed-off-by: raver119 <raver119@gmail.com> * BooleanOp syncToDevice Signed-off-by: raver119 <raver119@gmail.com> * meh Signed-off-by: raver119 <raver119@gmail.com> * additional data types for im2col/col2im Signed-off-by: raver119 <raver119@gmail.com> * Added test for firas_sparse op. * one more RandomBuffer test excluded Signed-off-by: raver119 <raver119@gmail.com> * Added tests for flatten op. * Added test for Floor op. * bunch of tests fixed Signed-off-by: raver119 <raver119@gmail.com> * mmulDot tests fixed Signed-off-by: raver119 <raver119@gmail.com> * more tests fixed Signed-off-by: raver119 <raver119@gmail.com> * Implemented floordiv_bp op and tests. * Fixed scalar case with cuda implementation for bds. * - work on cuda kernel for clip_by_norm backprop op is completed Signed-off-by: Yurii <yurii@skymind.io> * Eliminate cbow crach. * more tests fixed Signed-off-by: raver119 <raver119@gmail.com> * more tests fixed Signed-off-by: raver119 <raver119@gmail.com> * Eliminated abortion with batched nlp test. * more tests fixed Signed-off-by: raver119 <raver119@gmail.com> * Fixed shared flag initializing. * disabled bunch of cpu workspaces tests Signed-off-by: raver119 <raver119@gmail.com> * scalar operators fix: missing registerSpecialUse call Signed-off-by: raver119 <raver119@gmail.com> * Fixed logdet for cuda and tests. * - correct clipBynorm_bp Signed-off-by: Yurii <yurii@skymind.io> * Fixed crop_and_resize shape datatype. * - correct some mmul tests Signed-off-by: Yurii <yurii@skymind.io>
2019-08-02 19:01:03 +02:00
auto xType = sd::ArrayOptions::dataType(hXShapeInfo);
auto zType = sd::ArrayOptions::dataType(hZShapeInfo);
2019-06-06 14:21:15 +02:00
auto hz = reinterpret_cast<Nd4jLong*>(hZ);
BUILD_DOUBLE_SELECTOR(xType, zType, hz[0] = functions::indexreduce::IndexReduce, ::execScalar(opNum,hX,hXShapeInfo,extraParams), LIBND4J_TYPES, INDEXING_TYPES);
2019-06-06 14:21:15 +02:00
}
////////////////////////////////////////////////////////////////////////
/**
*
* @param opNum
* @param hX
* @param hXShapeInfo
* @param extraParams
* @param hZ
* @param hZShapeInfo
* @param dimension
* @param dimensionLength
*/
void NativeOpExecutioner::execIndexReduce(sd::LaunchContext *lc,
2019-06-06 14:21:15 +02:00
int opNum,
void *hX, Nd4jLong *hXShapeInfo,
void *dX, Nd4jLong *dXShapeInfo,
void *extraParams,
void *hZ, Nd4jLong *hZShapeInfo,
void *dZ, Nd4jLong *dZShapeInfo,
int *dimension, int dimensionLength,
Nd4jLong *tadShapeInfo, Nd4jLong *tadOffsets) {
2019-06-06 14:21:15 +02:00
auto xType = sd::ArrayOptions::dataType(hXShapeInfo);
auto zType = sd::ArrayOptions::dataType(hZShapeInfo);
2019-06-06 14:21:15 +02:00
Nd4jLong* hz = reinterpret_cast<Nd4jLong*>(hZ);
BUILD_DOUBLE_SELECTOR(xType, zType, functions::indexreduce::IndexReduce, ::exec(opNum, hX, hXShapeInfo, extraParams, hz, hZShapeInfo, dimension, dimensionLength, tadShapeInfo, tadOffsets), LIBND4J_TYPES, INDEXING_TYPES);
2019-06-06 14:21:15 +02:00
// BUILD_SINGLE_SELECTOR(xType, functions::indexreduce::IndexReduce, ::exec(opNum, hX, hXShapeInfo, dX, dXShapeInfo, extraParams, hZ, hZShapeInfo, dZ, dZShapeInfo, dimension, dimensionLength, tadShapeInfo, tadOffsets), LIBND4J_TYPES);
}
////////////////////////////////////////////////////////////////////////
/**
*
* @param opNum
* @param hX
* @param hXShapeInfo
* @param hY
* @param hYShapeInfo
* @param hZ
* @param hZShapeInfo
* @param dimension
* @param dimensionLength
*/
void NativeOpExecutioner::execBroadcast(sd::LaunchContext *lc,
2019-06-06 14:21:15 +02:00
int opNum,
void *hX, Nd4jLong *hXShapeInfo,
void *dX, Nd4jLong *dXShapeInfo,
void *hY, Nd4jLong *hYShapeInfo,
void *dY, Nd4jLong *dYShapeInfo,
void *hZ, Nd4jLong *hZShapeInfo,
void *dZ, Nd4jLong *dZShapeInfo,
int *dimension, int dimensionLength,
Nd4jLong *tadOnlyShapeInfo, Nd4jLong *tadOffsets,
Nd4jLong *tadOnlyShapeInfoZ,Nd4jLong *tadOffsetsZ) {
[WIP] More of CUDA (#95) * initial commit Signed-off-by: raver119 <raver119@gmail.com> * Implementation of hashcode cuda helper. Working edition. * Fixed parallel test input arangements. * Fixed tests for hashcode op. * Fixed shape calculation for image:crop_and_resize op and test. * NativeOps tests. Initial test suite. * Added tests for indexReduce methods. * Added test on execBroadcast with NDArray as dimensions. * Added test on execBroadcastBool with NDArray as dimensions. * Added tests on execPairwiseTransform and execPairwiseTransofrmBool. * Added tests for execReduce with scalar results. * Added reduce tests for non-empty dims array. * Added tests for reduce3. * Added tests for execScalar. * Added tests for execSummaryStats. * - provide cpu/cuda code for batch_to_space - testing it Signed-off-by: Yurii <yurii@skymind.io> * - remove old test for batch_to_space (had wrong format and numbers were not checked) Signed-off-by: Yurii <yurii@skymind.io> * Fixed complilation errors with test. * Added test for execTransformFloat. * Added test for execTransformSame. * Added test for execTransformBool. * Added test for execTransformStrict. * Added tests for execScalar/execScalarBool with TADs. * Added test for flatten. * - provide cpu/cuda code for space_to_Batch operaion Signed-off-by: Yurii <yurii@skymind.io> * Added test for concat. * comment unnecessary stuff in s_t_b Signed-off-by: Yurii <yurii@skymind.io> * Added test for specialConcat. * Added tests for memcpy/set routines. * Fixed pullRow cuda test. * Added pullRow test. * Added average test. * - correct typo in NDArray::applyPairwiseTransform(nd4j::pairwise::BoolOps op...) Signed-off-by: Yurii <yurii@skymind.io> * - debugging and fixing cuda tests in JavaInteropTests file Signed-off-by: Yurii <yurii@skymind.io> * - correct some tests Signed-off-by: Yurii <yurii@skymind.io> * Added test for shuffle. * Fixed ops declarations. * Restored omp and added shuffle test. * Added convertTypes test. * Added tests for execRandom. Eliminated usage of RandomBuffer with NativeOps. * Added sort tests. * Added tests for execCustomOp. * - further debuging and fixing tests terminated with crash Signed-off-by: Yurii <yurii@skymind.io> * Added tests for calculateOutputShapes. * Addded Benchmarks test. * Commented benchmark tests. * change assertion Signed-off-by: raver119 <raver119@gmail.com> * Added tests for apply_sgd op. Added cpu helper for that op. * Implement cuda helper for aplly_sgd op. Fixed tests for NativeOps. * Added test for assign broadcastable. * Added tests for assign_bp op. * Added tests for axpy op. * - assign/execScalar/execTransformAny signature change - minor test fix Signed-off-by: raver119 <raver119@gmail.com> * Fixed axpy op. * meh Signed-off-by: raver119 <raver119@gmail.com> * - fix tests for nativeOps::concat Signed-off-by: Yurii <yurii@skymind.io> * sequential transform/scalar Signed-off-by: raver119 <raver119@gmail.com> * allow nested parallelism Signed-off-by: raver119 <raver119@gmail.com> * assign_bp leak fix Signed-off-by: raver119 <raver119@gmail.com> * block setRNG fix Signed-off-by: raver119 <raver119@gmail.com> * enable parallelism by default Signed-off-by: raver119 <raver119@gmail.com> * enable nested parallelism by default Signed-off-by: raver119 <raver119@gmail.com> * Added cuda implementation for row_count helper. * Added implementation for tnse gains op helper. * - take into account possible situations when input arrays are empty in reduce_ cuda stuff Signed-off-by: Yurii <yurii@skymind.io> * Implemented tsne/edge_forces op cuda-based helper. Parallelized cpu-based helper for edge_forces. * Added kernel for tsne/symmetrized op heleper. * Implementation of tsne/symmetrized op cuda helper. Working edition. * Eliminated waste printfs. * Added test for broadcastgradientargs op. * host-only fallback for empty reduce float Signed-off-by: raver119 <raver119@gmail.com> * - some tests fixes Signed-off-by: Yurii <yurii@skymind.io> * - correct the rest of reduce_ stuff Signed-off-by: Yurii <yurii@skymind.io> * - further correction of reduce_ stuff Signed-off-by: Yurii <yurii@skymind.io> * Added test for Cbow op. Also added cuda implementation for cbow helpers. * - improve code of stack operation for scalar case Signed-off-by: Yurii <yurii@skymind.io> * - provide cuda kernel for gatherND operation Signed-off-by: Yurii <yurii@skymind.io> * Implementation of cbow helpers with cuda kernels. * minor tests tweaks Signed-off-by: raver119 <raver119@gmail.com> * minor tests tweaks Signed-off-by: raver119 <raver119@gmail.com> * - further correction of cuda stuff Signed-off-by: Yurii <yurii@skymind.io> * Implementatation of cbow op helper with cuda kernels. Working edition. * Skip random testing for cudablas case. * lstmBlockCell context fix Signed-off-by: raver119 <raver119@gmail.com> * Added tests for ELU and ELU_BP ops. * Added tests for eq_scalar, gt_scalar, gte_scalar and lte_scalar ops. * Added tests for neq_scalar. * Added test for noop. * - further work on clipbynorm_bp Signed-off-by: Yurii <yurii@skymind.io> * - get rid of concat op call, use instead direct concat helper call Signed-off-by: Yurii <yurii@skymind.io> * lstmBlockCell context fix Signed-off-by: raver119 <raver119@gmail.com> * Added tests for lrelu and lrelu_bp. * Added tests for selu and selu_bp. * Fixed lrelu derivative helpers. * - some corrections in lstm Signed-off-by: Yurii <yurii@skymind.io> * operator * result shape fix Signed-off-by: raver119 <raver119@gmail.com> * - correct typo in lstmCell Signed-off-by: Yurii <yurii@skymind.io> * few tests fixed Signed-off-by: raver119 <raver119@gmail.com> * CUDA inverse broadcast bool fix Signed-off-by: raver119 <raver119@gmail.com> * disable MMAP test for CUDA Signed-off-by: raver119 <raver119@gmail.com> * BooleanOp syncToDevice Signed-off-by: raver119 <raver119@gmail.com> * meh Signed-off-by: raver119 <raver119@gmail.com> * additional data types for im2col/col2im Signed-off-by: raver119 <raver119@gmail.com> * Added test for firas_sparse op. * one more RandomBuffer test excluded Signed-off-by: raver119 <raver119@gmail.com> * Added tests for flatten op. * Added test for Floor op. * bunch of tests fixed Signed-off-by: raver119 <raver119@gmail.com> * mmulDot tests fixed Signed-off-by: raver119 <raver119@gmail.com> * more tests fixed Signed-off-by: raver119 <raver119@gmail.com> * Implemented floordiv_bp op and tests. * Fixed scalar case with cuda implementation for bds. * - work on cuda kernel for clip_by_norm backprop op is completed Signed-off-by: Yurii <yurii@skymind.io> * Eliminate cbow crach. * more tests fixed Signed-off-by: raver119 <raver119@gmail.com> * more tests fixed Signed-off-by: raver119 <raver119@gmail.com> * Eliminated abortion with batched nlp test. * more tests fixed Signed-off-by: raver119 <raver119@gmail.com> * Fixed shared flag initializing. * disabled bunch of cpu workspaces tests Signed-off-by: raver119 <raver119@gmail.com> * scalar operators fix: missing registerSpecialUse call Signed-off-by: raver119 <raver119@gmail.com> * Fixed logdet for cuda and tests. * - correct clipBynorm_bp Signed-off-by: Yurii <yurii@skymind.io> * Fixed crop_and_resize shape datatype. * - correct some mmul tests Signed-off-by: Yurii <yurii@skymind.io>
2019-08-02 19:01:03 +02:00
auto xType = sd::ArrayOptions::dataType(hXShapeInfo);
auto yType = sd::ArrayOptions::dataType(hYShapeInfo);
auto zType = sd::ArrayOptions::dataType(hZShapeInfo);
2019-06-06 14:21:15 +02:00
if (shape::isEmpty(hXShapeInfo) || shape::isEmpty(hYShapeInfo))
return;
2019-06-06 14:21:15 +02:00
#ifdef __ND4J_EXPERIMENTAL__
BUILD_PAIRWISE_SELECTOR(xType, yType, zType, functions::broadcast::Broadcast, ::exec(opNum, hX, hXShapeInfo, hY, hYShapeInfo, hZ, hZShapeInfo, dimension, dimensionLength, tadOnlyShapeInfo, tadOffsets, tadOnlyShapeInfoZ, tadOffsetsZ), LIBND4J_TYPES, LIBND4J_TYPES);
#else
auto loopKind = sd::LoopKind::deduceKindOfLoopBroadcast(hXShapeInfo, hYShapeInfo, hZShapeInfo);
auto func = PRAGMA_THREADS_FOR {
BUILD_SINGLE_SELECTOR_THRICE(xType, functions::broadcast::Broadcast, ::exec(opNum, hX, hXShapeInfo, hY, hYShapeInfo, hZ, hZShapeInfo, dimension, dimensionLength, tadOnlyShapeInfo, tadOffsets, tadOnlyShapeInfoZ, tadOffsetsZ, loopKind, start, stop), LIBND4J_TYPES);
};
Nd4jLong numTads = 0;
switch (loopKind) {
case sd::LoopKind::BROADCAST_SCALAR_X: {
numTads = shape::length(hXShapeInfo);
}
break;
case sd::LoopKind::BROADCAST_SCALAR_Y: {
numTads = shape::length(hYShapeInfo);
}
break;
case sd::LoopKind::BROADCAST_3D: {
numTads = shape::sizeAt(hZShapeInfo, 0);
}
break;
case sd::LoopKind::BROADCAST_4D: {
numTads = shape::sizeAt(hZShapeInfo, 0) * shape::sizeAt(hZShapeInfo, 1);
}
break;
case sd::LoopKind::BROADCAST_5D: {
numTads = shape::sizeAt(hZShapeInfo, 0) * shape::sizeAt(hZShapeInfo, 1);
}
break;
default: {
auto xLen = shape::length(hXShapeInfo);
auto yLen = shape::length(hYShapeInfo);
numTads = xLen / yLen;
}
}
samediff::Threads::parallel_tad(func, 0, numTads);
2019-06-06 14:21:15 +02:00
#endif
}
void NativeOpExecutioner::execInverseBroadcast(sd::LaunchContext *lc,
2019-06-06 14:21:15 +02:00
int opNum,
void *hX, Nd4jLong *hXShapeInfo,
void *dX, Nd4jLong *dXShapeInfo,
void *hY, Nd4jLong *hYShapeInfo,
void *dY, Nd4jLong *dYShapeInfo,
void *hZ, Nd4jLong *hZShapeInfo,
void *dZ, Nd4jLong *dZShapeInfo,
int *dimension, int dimensionLength,
Nd4jLong *tadOnlyShapeInfo, Nd4jLong *tadOffsets,
Nd4jLong *tadOnlyShapeInfoZ,Nd4jLong *tadOffsetsZ) {
auto xType = sd::ArrayOptions::dataType(hXShapeInfo);
auto yType = sd::ArrayOptions::dataType(hYShapeInfo);
auto zType = sd::ArrayOptions::dataType(hZShapeInfo);
2019-06-06 14:21:15 +02:00
if (shape::isEmpty(hXShapeInfo) || shape::isEmpty(hYShapeInfo))
return;
[WIP] More of CUDA (#95) * initial commit Signed-off-by: raver119 <raver119@gmail.com> * Implementation of hashcode cuda helper. Working edition. * Fixed parallel test input arangements. * Fixed tests for hashcode op. * Fixed shape calculation for image:crop_and_resize op and test. * NativeOps tests. Initial test suite. * Added tests for indexReduce methods. * Added test on execBroadcast with NDArray as dimensions. * Added test on execBroadcastBool with NDArray as dimensions. * Added tests on execPairwiseTransform and execPairwiseTransofrmBool. * Added tests for execReduce with scalar results. * Added reduce tests for non-empty dims array. * Added tests for reduce3. * Added tests for execScalar. * Added tests for execSummaryStats. * - provide cpu/cuda code for batch_to_space - testing it Signed-off-by: Yurii <yurii@skymind.io> * - remove old test for batch_to_space (had wrong format and numbers were not checked) Signed-off-by: Yurii <yurii@skymind.io> * Fixed complilation errors with test. * Added test for execTransformFloat. * Added test for execTransformSame. * Added test for execTransformBool. * Added test for execTransformStrict. * Added tests for execScalar/execScalarBool with TADs. * Added test for flatten. * - provide cpu/cuda code for space_to_Batch operaion Signed-off-by: Yurii <yurii@skymind.io> * Added test for concat. * comment unnecessary stuff in s_t_b Signed-off-by: Yurii <yurii@skymind.io> * Added test for specialConcat. * Added tests for memcpy/set routines. * Fixed pullRow cuda test. * Added pullRow test. * Added average test. * - correct typo in NDArray::applyPairwiseTransform(nd4j::pairwise::BoolOps op...) Signed-off-by: Yurii <yurii@skymind.io> * - debugging and fixing cuda tests in JavaInteropTests file Signed-off-by: Yurii <yurii@skymind.io> * - correct some tests Signed-off-by: Yurii <yurii@skymind.io> * Added test for shuffle. * Fixed ops declarations. * Restored omp and added shuffle test. * Added convertTypes test. * Added tests for execRandom. Eliminated usage of RandomBuffer with NativeOps. * Added sort tests. * Added tests for execCustomOp. * - further debuging and fixing tests terminated with crash Signed-off-by: Yurii <yurii@skymind.io> * Added tests for calculateOutputShapes. * Addded Benchmarks test. * Commented benchmark tests. * change assertion Signed-off-by: raver119 <raver119@gmail.com> * Added tests for apply_sgd op. Added cpu helper for that op. * Implement cuda helper for aplly_sgd op. Fixed tests for NativeOps. * Added test for assign broadcastable. * Added tests for assign_bp op. * Added tests for axpy op. * - assign/execScalar/execTransformAny signature change - minor test fix Signed-off-by: raver119 <raver119@gmail.com> * Fixed axpy op. * meh Signed-off-by: raver119 <raver119@gmail.com> * - fix tests for nativeOps::concat Signed-off-by: Yurii <yurii@skymind.io> * sequential transform/scalar Signed-off-by: raver119 <raver119@gmail.com> * allow nested parallelism Signed-off-by: raver119 <raver119@gmail.com> * assign_bp leak fix Signed-off-by: raver119 <raver119@gmail.com> * block setRNG fix Signed-off-by: raver119 <raver119@gmail.com> * enable parallelism by default Signed-off-by: raver119 <raver119@gmail.com> * enable nested parallelism by default Signed-off-by: raver119 <raver119@gmail.com> * Added cuda implementation for row_count helper. * Added implementation for tnse gains op helper. * - take into account possible situations when input arrays are empty in reduce_ cuda stuff Signed-off-by: Yurii <yurii@skymind.io> * Implemented tsne/edge_forces op cuda-based helper. Parallelized cpu-based helper for edge_forces. * Added kernel for tsne/symmetrized op heleper. * Implementation of tsne/symmetrized op cuda helper. Working edition. * Eliminated waste printfs. * Added test for broadcastgradientargs op. * host-only fallback for empty reduce float Signed-off-by: raver119 <raver119@gmail.com> * - some tests fixes Signed-off-by: Yurii <yurii@skymind.io> * - correct the rest of reduce_ stuff Signed-off-by: Yurii <yurii@skymind.io> * - further correction of reduce_ stuff Signed-off-by: Yurii <yurii@skymind.io> * Added test for Cbow op. Also added cuda implementation for cbow helpers. * - improve code of stack operation for scalar case Signed-off-by: Yurii <yurii@skymind.io> * - provide cuda kernel for gatherND operation Signed-off-by: Yurii <yurii@skymind.io> * Implementation of cbow helpers with cuda kernels. * minor tests tweaks Signed-off-by: raver119 <raver119@gmail.com> * minor tests tweaks Signed-off-by: raver119 <raver119@gmail.com> * - further correction of cuda stuff Signed-off-by: Yurii <yurii@skymind.io> * Implementatation of cbow op helper with cuda kernels. Working edition. * Skip random testing for cudablas case. * lstmBlockCell context fix Signed-off-by: raver119 <raver119@gmail.com> * Added tests for ELU and ELU_BP ops. * Added tests for eq_scalar, gt_scalar, gte_scalar and lte_scalar ops. * Added tests for neq_scalar. * Added test for noop. * - further work on clipbynorm_bp Signed-off-by: Yurii <yurii@skymind.io> * - get rid of concat op call, use instead direct concat helper call Signed-off-by: Yurii <yurii@skymind.io> * lstmBlockCell context fix Signed-off-by: raver119 <raver119@gmail.com> * Added tests for lrelu and lrelu_bp. * Added tests for selu and selu_bp. * Fixed lrelu derivative helpers. * - some corrections in lstm Signed-off-by: Yurii <yurii@skymind.io> * operator * result shape fix Signed-off-by: raver119 <raver119@gmail.com> * - correct typo in lstmCell Signed-off-by: Yurii <yurii@skymind.io> * few tests fixed Signed-off-by: raver119 <raver119@gmail.com> * CUDA inverse broadcast bool fix Signed-off-by: raver119 <raver119@gmail.com> * disable MMAP test for CUDA Signed-off-by: raver119 <raver119@gmail.com> * BooleanOp syncToDevice Signed-off-by: raver119 <raver119@gmail.com> * meh Signed-off-by: raver119 <raver119@gmail.com> * additional data types for im2col/col2im Signed-off-by: raver119 <raver119@gmail.com> * Added test for firas_sparse op. * one more RandomBuffer test excluded Signed-off-by: raver119 <raver119@gmail.com> * Added tests for flatten op. * Added test for Floor op. * bunch of tests fixed Signed-off-by: raver119 <raver119@gmail.com> * mmulDot tests fixed Signed-off-by: raver119 <raver119@gmail.com> * more tests fixed Signed-off-by: raver119 <raver119@gmail.com> * Implemented floordiv_bp op and tests. * Fixed scalar case with cuda implementation for bds. * - work on cuda kernel for clip_by_norm backprop op is completed Signed-off-by: Yurii <yurii@skymind.io> * Eliminate cbow crach. * more tests fixed Signed-off-by: raver119 <raver119@gmail.com> * more tests fixed Signed-off-by: raver119 <raver119@gmail.com> * Eliminated abortion with batched nlp test. * more tests fixed Signed-off-by: raver119 <raver119@gmail.com> * Fixed shared flag initializing. * disabled bunch of cpu workspaces tests Signed-off-by: raver119 <raver119@gmail.com> * scalar operators fix: missing registerSpecialUse call Signed-off-by: raver119 <raver119@gmail.com> * Fixed logdet for cuda and tests. * - correct clipBynorm_bp Signed-off-by: Yurii <yurii@skymind.io> * Fixed crop_and_resize shape datatype. * - correct some mmul tests Signed-off-by: Yurii <yurii@skymind.io>
2019-08-02 19:01:03 +02:00
if (!sd::Environment::getInstance()->isExperimentalBuild())
if ((yType != xType && yType != sd::DataType::BOOL) || xType != zType)
throw sd::datatype_exception::build("NativeOps::execBroadcast both operands must have same data type", xType, yType);
2019-06-06 14:21:15 +02:00
#ifdef __ND4J_EXPERIMENTAL__
BUILD_PAIRWISE_SELECTOR(xType, yType, zType, functions::broadcast::Broadcast, ::execInverse(opNum, hX, hXShapeInfo, hY, hYShapeInfo, hZ, hZShapeInfo, dimension, dimensionLength, tadOnlyShapeInfo, tadOffsets, tadOnlyShapeInfoZ, tadOffsetsZ), LIBND4J_TYPES, LIBND4J_TYPES);
#else
auto func = PRAGMA_THREADS_FOR {
BUILD_SINGLE_SELECTOR_THRICE(xType, functions::broadcast::Broadcast, ::execInverse(opNum, hX, hXShapeInfo, hY, hYShapeInfo, hZ, hZShapeInfo, dimension, dimensionLength, tadOnlyShapeInfo, tadOffsets, tadOnlyShapeInfoZ, tadOffsetsZ, start, stop), LIBND4J_TYPES);
};
auto xLen = shape::length(hXShapeInfo);
auto yLen = shape::length(hYShapeInfo);
auto numTads = yLen / xLen;
samediff::Threads::parallel_tad(func, 0, numTads);
2019-06-06 14:21:15 +02:00
#endif
}
////////////////////////////////////////////////////////////////////////
void NativeOpExecutioner::execBroadcastBool(sd::LaunchContext *lc,
2019-06-06 14:21:15 +02:00
int opNum,
void *hX, Nd4jLong *hXShapeInfo,
void *dX, Nd4jLong *dXShapeInfo,
void *hY, Nd4jLong *hYShapeInfo,
void *dY, Nd4jLong *dYShapeInfo,
void *hZ, Nd4jLong *hZShapeInfo,
void *dZ, Nd4jLong *dZShapeInfo,
void *extraParams,
2019-06-06 14:21:15 +02:00
int *dimension, int dimensionLength,
Nd4jLong *tadOnlyShapeInfo, Nd4jLong *tadOffsets,
Nd4jLong *tadOnlyShapeInfoZ,Nd4jLong *tadOffsetsZ) {
2019-06-06 14:21:15 +02:00
auto xType = sd::ArrayOptions::dataType(hXShapeInfo);
auto yType = sd::ArrayOptions::dataType(hYShapeInfo);
auto zType = sd::ArrayOptions::dataType(hZShapeInfo);
2019-06-06 14:21:15 +02:00
if (shape::isEmpty(hXShapeInfo) || shape::isEmpty(hYShapeInfo))
return;
auto func = PRAGMA_THREADS_FOR {
BUILD_DOUBLE_SELECTOR(xType, zType, functions::broadcast::BroadcastBool, ::exec(opNum, hX, hXShapeInfo, hY, hYShapeInfo, hZ, hZShapeInfo, extraParams, dimension, dimensionLength, tadOnlyShapeInfo, tadOffsets, tadOnlyShapeInfoZ, tadOffsetsZ, start, stop), LIBND4J_TYPES, BOOL_TYPES);
};
auto xLen = shape::length(hXShapeInfo);
auto yLen = shape::length(hYShapeInfo);
auto numTads = xLen / yLen;
samediff::Threads::parallel_tad(func, 0, numTads);
2019-06-06 14:21:15 +02:00
}
void NativeOpExecutioner::execInverseBroadcastBool(sd::LaunchContext *lc,
2019-06-06 14:21:15 +02:00
int opNum,
void *hX, Nd4jLong *hXShapeInfo,
void *dX, Nd4jLong *dXShapeInfo,
void *hY, Nd4jLong *hYShapeInfo,
void *dY, Nd4jLong *dYShapeInfo,
void *hZ, Nd4jLong *hZShapeInfo,
void *dZ, Nd4jLong *dZShapeInfo,
void *extraParams,
2019-06-06 14:21:15 +02:00
int *dimension, int dimensionLength,
Nd4jLong *tadOnlyShapeInfo, Nd4jLong *tadOffsets,
Nd4jLong *tadOnlyShapeInfoZ,Nd4jLong *tadOffsetsZ) {
[WIP] More of CUDA (#95) * initial commit Signed-off-by: raver119 <raver119@gmail.com> * Implementation of hashcode cuda helper. Working edition. * Fixed parallel test input arangements. * Fixed tests for hashcode op. * Fixed shape calculation for image:crop_and_resize op and test. * NativeOps tests. Initial test suite. * Added tests for indexReduce methods. * Added test on execBroadcast with NDArray as dimensions. * Added test on execBroadcastBool with NDArray as dimensions. * Added tests on execPairwiseTransform and execPairwiseTransofrmBool. * Added tests for execReduce with scalar results. * Added reduce tests for non-empty dims array. * Added tests for reduce3. * Added tests for execScalar. * Added tests for execSummaryStats. * - provide cpu/cuda code for batch_to_space - testing it Signed-off-by: Yurii <yurii@skymind.io> * - remove old test for batch_to_space (had wrong format and numbers were not checked) Signed-off-by: Yurii <yurii@skymind.io> * Fixed complilation errors with test. * Added test for execTransformFloat. * Added test for execTransformSame. * Added test for execTransformBool. * Added test for execTransformStrict. * Added tests for execScalar/execScalarBool with TADs. * Added test for flatten. * - provide cpu/cuda code for space_to_Batch operaion Signed-off-by: Yurii <yurii@skymind.io> * Added test for concat. * comment unnecessary stuff in s_t_b Signed-off-by: Yurii <yurii@skymind.io> * Added test for specialConcat. * Added tests for memcpy/set routines. * Fixed pullRow cuda test. * Added pullRow test. * Added average test. * - correct typo in NDArray::applyPairwiseTransform(nd4j::pairwise::BoolOps op...) Signed-off-by: Yurii <yurii@skymind.io> * - debugging and fixing cuda tests in JavaInteropTests file Signed-off-by: Yurii <yurii@skymind.io> * - correct some tests Signed-off-by: Yurii <yurii@skymind.io> * Added test for shuffle. * Fixed ops declarations. * Restored omp and added shuffle test. * Added convertTypes test. * Added tests for execRandom. Eliminated usage of RandomBuffer with NativeOps. * Added sort tests. * Added tests for execCustomOp. * - further debuging and fixing tests terminated with crash Signed-off-by: Yurii <yurii@skymind.io> * Added tests for calculateOutputShapes. * Addded Benchmarks test. * Commented benchmark tests. * change assertion Signed-off-by: raver119 <raver119@gmail.com> * Added tests for apply_sgd op. Added cpu helper for that op. * Implement cuda helper for aplly_sgd op. Fixed tests for NativeOps. * Added test for assign broadcastable. * Added tests for assign_bp op. * Added tests for axpy op. * - assign/execScalar/execTransformAny signature change - minor test fix Signed-off-by: raver119 <raver119@gmail.com> * Fixed axpy op. * meh Signed-off-by: raver119 <raver119@gmail.com> * - fix tests for nativeOps::concat Signed-off-by: Yurii <yurii@skymind.io> * sequential transform/scalar Signed-off-by: raver119 <raver119@gmail.com> * allow nested parallelism Signed-off-by: raver119 <raver119@gmail.com> * assign_bp leak fix Signed-off-by: raver119 <raver119@gmail.com> * block setRNG fix Signed-off-by: raver119 <raver119@gmail.com> * enable parallelism by default Signed-off-by: raver119 <raver119@gmail.com> * enable nested parallelism by default Signed-off-by: raver119 <raver119@gmail.com> * Added cuda implementation for row_count helper. * Added implementation for tnse gains op helper. * - take into account possible situations when input arrays are empty in reduce_ cuda stuff Signed-off-by: Yurii <yurii@skymind.io> * Implemented tsne/edge_forces op cuda-based helper. Parallelized cpu-based helper for edge_forces. * Added kernel for tsne/symmetrized op heleper. * Implementation of tsne/symmetrized op cuda helper. Working edition. * Eliminated waste printfs. * Added test for broadcastgradientargs op. * host-only fallback for empty reduce float Signed-off-by: raver119 <raver119@gmail.com> * - some tests fixes Signed-off-by: Yurii <yurii@skymind.io> * - correct the rest of reduce_ stuff Signed-off-by: Yurii <yurii@skymind.io> * - further correction of reduce_ stuff Signed-off-by: Yurii <yurii@skymind.io> * Added test for Cbow op. Also added cuda implementation for cbow helpers. * - improve code of stack operation for scalar case Signed-off-by: Yurii <yurii@skymind.io> * - provide cuda kernel for gatherND operation Signed-off-by: Yurii <yurii@skymind.io> * Implementation of cbow helpers with cuda kernels. * minor tests tweaks Signed-off-by: raver119 <raver119@gmail.com> * minor tests tweaks Signed-off-by: raver119 <raver119@gmail.com> * - further correction of cuda stuff Signed-off-by: Yurii <yurii@skymind.io> * Implementatation of cbow op helper with cuda kernels. Working edition. * Skip random testing for cudablas case. * lstmBlockCell context fix Signed-off-by: raver119 <raver119@gmail.com> * Added tests for ELU and ELU_BP ops. * Added tests for eq_scalar, gt_scalar, gte_scalar and lte_scalar ops. * Added tests for neq_scalar. * Added test for noop. * - further work on clipbynorm_bp Signed-off-by: Yurii <yurii@skymind.io> * - get rid of concat op call, use instead direct concat helper call Signed-off-by: Yurii <yurii@skymind.io> * lstmBlockCell context fix Signed-off-by: raver119 <raver119@gmail.com> * Added tests for lrelu and lrelu_bp. * Added tests for selu and selu_bp. * Fixed lrelu derivative helpers. * - some corrections in lstm Signed-off-by: Yurii <yurii@skymind.io> * operator * result shape fix Signed-off-by: raver119 <raver119@gmail.com> * - correct typo in lstmCell Signed-off-by: Yurii <yurii@skymind.io> * few tests fixed Signed-off-by: raver119 <raver119@gmail.com> * CUDA inverse broadcast bool fix Signed-off-by: raver119 <raver119@gmail.com> * disable MMAP test for CUDA Signed-off-by: raver119 <raver119@gmail.com> * BooleanOp syncToDevice Signed-off-by: raver119 <raver119@gmail.com> * meh Signed-off-by: raver119 <raver119@gmail.com> * additional data types for im2col/col2im Signed-off-by: raver119 <raver119@gmail.com> * Added test for firas_sparse op. * one more RandomBuffer test excluded Signed-off-by: raver119 <raver119@gmail.com> * Added tests for flatten op. * Added test for Floor op. * bunch of tests fixed Signed-off-by: raver119 <raver119@gmail.com> * mmulDot tests fixed Signed-off-by: raver119 <raver119@gmail.com> * more tests fixed Signed-off-by: raver119 <raver119@gmail.com> * Implemented floordiv_bp op and tests. * Fixed scalar case with cuda implementation for bds. * - work on cuda kernel for clip_by_norm backprop op is completed Signed-off-by: Yurii <yurii@skymind.io> * Eliminate cbow crach. * more tests fixed Signed-off-by: raver119 <raver119@gmail.com> * more tests fixed Signed-off-by: raver119 <raver119@gmail.com> * Eliminated abortion with batched nlp test. * more tests fixed Signed-off-by: raver119 <raver119@gmail.com> * Fixed shared flag initializing. * disabled bunch of cpu workspaces tests Signed-off-by: raver119 <raver119@gmail.com> * scalar operators fix: missing registerSpecialUse call Signed-off-by: raver119 <raver119@gmail.com> * Fixed logdet for cuda and tests. * - correct clipBynorm_bp Signed-off-by: Yurii <yurii@skymind.io> * Fixed crop_and_resize shape datatype. * - correct some mmul tests Signed-off-by: Yurii <yurii@skymind.io>
2019-08-02 19:01:03 +02:00
auto xType = sd::ArrayOptions::dataType(hXShapeInfo);
auto yType = sd::ArrayOptions::dataType(hYShapeInfo);
auto zType = sd::ArrayOptions::dataType(hZShapeInfo);
2019-06-06 14:21:15 +02:00
if (shape::isEmpty(hXShapeInfo) || shape::isEmpty(hYShapeInfo))
return;
if (!sd::Environment::getInstance()->isExperimentalBuild())
if (yType != xType || sd::DataType::BOOL != zType)
throw sd::datatype_exception::build("NativeOps::execInverseBroadcastBool both operands must have same data type", xType, yType);
2019-06-06 14:21:15 +02:00
auto func = PRAGMA_THREADS_FOR {
BUILD_DOUBLE_SELECTOR(xType, zType, functions::broadcast::BroadcastBool, ::execInverse(opNum, hX, hXShapeInfo, hY, hYShapeInfo, hZ, hZShapeInfo, extraParams, dimension, dimensionLength, tadOnlyShapeInfo, tadOffsets, tadOnlyShapeInfoZ, tadOffsetsZ, start, stop), LIBND4J_TYPES, BOOL_TYPES);
};
auto xLen = shape::length(hXShapeInfo);
auto yLen = shape::length(hYShapeInfo);
auto numTads = yLen / xLen;
samediff::Threads::parallel_tad(func, 0, numTads);
2019-06-06 14:21:15 +02:00
}
////////////////////////////////////////////////////////////////////////
void NativeOpExecutioner::execBroadcastInt(sd::LaunchContext *lc,
int opNum,
void *hX, Nd4jLong *hXShapeInfo,
void *dX, Nd4jLong *dXShapeInfo,
void *hY, Nd4jLong *hYShapeInfo,
void *dY, Nd4jLong *dYShapeInfo,
void *hZ, Nd4jLong *hZShapeInfo,
void *dZ, Nd4jLong *dZShapeInfo,
int *dimension, int dimensionLength,
Nd4jLong *tadOnlyShapeInfo, Nd4jLong *tadOffsets,
Nd4jLong *tadOnlyShapeInfoZ,Nd4jLong *tadOffsetsZ) {
auto xType = sd::ArrayOptions::dataType(hXShapeInfo);
auto yType = sd::ArrayOptions::dataType(hYShapeInfo);
auto zType = sd::ArrayOptions::dataType(hZShapeInfo);
if (shape::isEmpty(hXShapeInfo) || shape::isEmpty(hYShapeInfo))
return;
if (xType != yType || xType != zType)
throw sd::datatype_exception::build("NativeOpExecutioner::execBroadcastInt", zType, xType, yType);
if (!sd::DataTypeUtils::isZ(zType))
throw sd::datatype_exception::build("NativeOpExecutioner::execBroadcastInt requires integer data type", zType);
auto func = PRAGMA_THREADS_FOR {
BUILD_SINGLE_SELECTOR(xType, functions::broadcast::BroadcastInt, ::exec(opNum, hX, hXShapeInfo, hY, hYShapeInfo, hZ, hZShapeInfo, dimension, dimensionLength, tadOnlyShapeInfo, tadOffsets, tadOnlyShapeInfoZ, tadOffsetsZ, start, stop), INTEGER_TYPES);
};
auto xLen = shape::length(hXShapeInfo);
auto yLen = shape::length(hYShapeInfo);
auto numTads = xLen / yLen;
samediff::Threads::parallel_tad(func, 0, numTads);
}
void NativeOpExecutioner::execInverseBroadcastInt(sd::LaunchContext *lc,
int opNum,
void *hX, Nd4jLong *hXShapeInfo,
void *dX, Nd4jLong *dXShapeInfo,
void *hY, Nd4jLong *hYShapeInfo,
void *dY, Nd4jLong *dYShapeInfo,
void *hZ, Nd4jLong *hZShapeInfo,
void *dZ, Nd4jLong *dZShapeInfo,
int *dimension, int dimensionLength,
Nd4jLong *tadOnlyShapeInfo, Nd4jLong *tadOffsets,
Nd4jLong *tadOnlyShapeInfoZ,Nd4jLong *tadOffsetsZ) {
auto xType = sd::ArrayOptions::dataType(hXShapeInfo);
auto yType = sd::ArrayOptions::dataType(hYShapeInfo);
auto zType = sd::ArrayOptions::dataType(hZShapeInfo);
if (shape::isEmpty(hXShapeInfo) || shape::isEmpty(hYShapeInfo))
return;
if (xType != yType || xType != zType)
throw sd::datatype_exception::build("NativeOpExecutioner::execInverseBroadcastInt", zType, xType, yType);
if (!sd::DataTypeUtils::isZ(zType))
throw sd::datatype_exception::build("NativeOpExecutioner::execInverseBroadcastInt requires integer data type", zType);
auto func = PRAGMA_THREADS_FOR {
BUILD_SINGLE_SELECTOR(xType, functions::broadcast::BroadcastInt,::execInverse(opNum, hX, hXShapeInfo, hY, hYShapeInfo, hZ, hZShapeInfo, dimension, dimensionLength, tadOnlyShapeInfo, tadOffsets, tadOnlyShapeInfoZ, tadOffsetsZ, start, stop), INTEGER_TYPES);
};
auto xLen = shape::length(hXShapeInfo);
auto yLen = shape::length(hYShapeInfo);
auto numTads = yLen / xLen;
samediff::Threads::parallel_tad(func, 0, numTads);
}
2019-06-06 14:21:15 +02:00
////////////////////////////////////////////////////////////////////////
/**
*
* @param opNum
* @param hX
* @param xStride
* @param hY
* @param yStride
* @param hZ
* @param resultStride
* @param extraParams
* @param n
*/
void NativeOpExecutioner::execPairwiseTransform(sd::LaunchContext *lc,
2019-06-06 14:21:15 +02:00
int opNum,
void *hX, Nd4jLong *hXShapeInfo,
void *dX, Nd4jLong *dXShapeInfo,
void *hY, Nd4jLong *hYShapeInfo,
void *dY, Nd4jLong *dYShapeInfo,
void *hZ, Nd4jLong *hZShapeInfo,
void *dZ, Nd4jLong *dZShapeInfo,
void *extraParams) {
auto xType = sd::ArrayOptions::dataType(hXShapeInfo);
auto yType = sd::ArrayOptions::dataType(hYShapeInfo);
auto zType = sd::ArrayOptions::dataType(hZShapeInfo);
2019-06-06 14:21:15 +02:00
if (shape::isEmpty(hXShapeInfo) || shape::isEmpty(hYShapeInfo))
return;
2019-06-06 14:21:15 +02:00
#ifdef __ND4J_EXPERIMENTAL__
BUILD_PAIRWISE_SELECTOR(xType, yType, zType, functions::pairwise_transforms::PairWiseTransform, ::exec(opNum, hX, hXShapeInfo, hY, hYShapeInfo, hZ, hZShapeInfo, extraParams), LIBND4J_TYPES, LIBND4J_TYPES);
#else
auto func = PRAGMA_THREADS_FOR {
BUILD_SINGLE_SELECTOR_THRICE(xType, functions::pairwise_transforms::PairWiseTransform,
::exec(opNum, hX, hXShapeInfo, hY, hYShapeInfo, hZ, hZShapeInfo, extraParams, start, stop),
LIBND4J_TYPES);
};
auto zLen = shape::length(hZShapeInfo);
samediff::Threads::parallel_for(func, 0, zLen, 1, sd::math::nd4j_max<int>(1, sd::math::nd4j_min<int>(zLen / 1024, sd::Environment::getInstance()->maxMasterThreads())));
2019-06-06 14:21:15 +02:00
#endif
}
////////////////////////////////////////////////////////////////////////
void NativeOpExecutioner::execPairwiseBoolTransform(sd::LaunchContext *lc,
2019-06-06 14:21:15 +02:00
int opNum,
void *hX, Nd4jLong *hXShapeInfo,
void *dX, Nd4jLong *dXShapeInfo,
void *hY, Nd4jLong *hYShapeInfo,
void *dY, Nd4jLong *dYShapeInfo,
void *hZ, Nd4jLong *hZShapeInfo,
void *dZ, Nd4jLong *dZShapeInfo,
void *extraParams) {
2019-06-06 14:21:15 +02:00
auto xType = sd::ArrayOptions::dataType(hXShapeInfo);
auto yType = sd::ArrayOptions::dataType(hYShapeInfo);
auto zType = sd::ArrayOptions::dataType(hZShapeInfo);
2019-06-06 14:21:15 +02:00
if (shape::isEmpty(hXShapeInfo) || shape::isEmpty(hYShapeInfo))
return;
if (xType != yType)
throw sd::datatype_exception::build("NativeOpExecutioner::execPairwiseBoolTransform", xType, yType);
if (zType != sd::DataType::BOOL)
throw sd::datatype_exception::build("NativeOpExecutioner::execPairwiseBoolTransform", sd::DataType::BOOL, zType);
auto func = PRAGMA_THREADS_FOR {
BUILD_DOUBLE_SELECTOR(xType, zType, functions::pairwise_transforms::PairWiseBoolTransform, ::exec(opNum, hX, hXShapeInfo, hY, hYShapeInfo, hZ, hZShapeInfo, extraParams, start, stop), LIBND4J_TYPES, BOOL_TYPES);
};
auto zLen = shape::length(hZShapeInfo);
samediff::Threads::parallel_for(func, 0, zLen, 1, sd::math::nd4j_max<int>(1, sd::math::nd4j_min<int>(zLen / 1024, sd::Environment::getInstance()->maxMasterThreads())));
2019-06-06 14:21:15 +02:00
}
////////////////////////////////////////////////////////////////////////
void NativeOpExecutioner::execPairwiseIntTransform(sd::LaunchContext *lc,
int opNum,
void *hX, Nd4jLong *hXShapeInfo,
void *dX, Nd4jLong *dXShapeInfo,
void *hY, Nd4jLong *hYShapeInfo,
void *dY, Nd4jLong *dYShapeInfo,
void *hZ, Nd4jLong *hZShapeInfo,
void *dZ, Nd4jLong *dZShapeInfo,
void *extraParams) {
auto xType = sd::ArrayOptions::dataType(hXShapeInfo);
auto yType = sd::ArrayOptions::dataType(hYShapeInfo);
auto zType = sd::ArrayOptions::dataType(hZShapeInfo);
if (shape::isEmpty(hXShapeInfo) || shape::isEmpty(hYShapeInfo))
return;
if (xType != yType || xType != zType)
throw sd::datatype_exception::build("NativeOpExecutioner::execPairwiseIntTransform", zType, xType, yType);
if (!sd::DataTypeUtils::isZ(zType))
throw sd::datatype_exception::build("NativeOpExecutioner::execSPairwiseInt requires integer data type", zType);
auto func = PRAGMA_THREADS_FOR {
BUILD_SINGLE_SELECTOR(xType, functions::pairwise_transforms::PairWiseIntTransform, ::exec(opNum, hX, hXShapeInfo, hY, hYShapeInfo, hZ, hZShapeInfo, extraParams, start, stop), INTEGER_TYPES);
};
auto zLen = shape::length(hZShapeInfo);
samediff::Threads::parallel_for(func, 0, zLen, 1, sd::math::nd4j_max<int>(1, sd::math::nd4j_min<int>(zLen / 1024, sd::Environment::getInstance()->maxMasterThreads())));
}
2019-06-06 14:21:15 +02:00
////////////////////////////////////////////////////////////////////////
/**
*
* @param opNum
* @param hX
* @param hXShapeInfo
* @param extraParams
* @param hZ
* @param hZShapeInfo
*/
void NativeOpExecutioner::execReduceFloat(sd::LaunchContext *lc,
2019-06-06 14:21:15 +02:00
int opNum,
void *hX, Nd4jLong *hXShapeInfo,
void *dX, Nd4jLong *dXShapeInfo,
void *extraParams,
void *hZ, Nd4jLong *hZShapeInfo,
void *dZ, Nd4jLong *dZShapeInfo,
int *dimension, int dimensionLength,
Nd4jLong *tadShapeInfo, Nd4jLong *tadOffsets) {
[WIP] More of CUDA (#95) * initial commit Signed-off-by: raver119 <raver119@gmail.com> * Implementation of hashcode cuda helper. Working edition. * Fixed parallel test input arangements. * Fixed tests for hashcode op. * Fixed shape calculation for image:crop_and_resize op and test. * NativeOps tests. Initial test suite. * Added tests for indexReduce methods. * Added test on execBroadcast with NDArray as dimensions. * Added test on execBroadcastBool with NDArray as dimensions. * Added tests on execPairwiseTransform and execPairwiseTransofrmBool. * Added tests for execReduce with scalar results. * Added reduce tests for non-empty dims array. * Added tests for reduce3. * Added tests for execScalar. * Added tests for execSummaryStats. * - provide cpu/cuda code for batch_to_space - testing it Signed-off-by: Yurii <yurii@skymind.io> * - remove old test for batch_to_space (had wrong format and numbers were not checked) Signed-off-by: Yurii <yurii@skymind.io> * Fixed complilation errors with test. * Added test for execTransformFloat. * Added test for execTransformSame. * Added test for execTransformBool. * Added test for execTransformStrict. * Added tests for execScalar/execScalarBool with TADs. * Added test for flatten. * - provide cpu/cuda code for space_to_Batch operaion Signed-off-by: Yurii <yurii@skymind.io> * Added test for concat. * comment unnecessary stuff in s_t_b Signed-off-by: Yurii <yurii@skymind.io> * Added test for specialConcat. * Added tests for memcpy/set routines. * Fixed pullRow cuda test. * Added pullRow test. * Added average test. * - correct typo in NDArray::applyPairwiseTransform(nd4j::pairwise::BoolOps op...) Signed-off-by: Yurii <yurii@skymind.io> * - debugging and fixing cuda tests in JavaInteropTests file Signed-off-by: Yurii <yurii@skymind.io> * - correct some tests Signed-off-by: Yurii <yurii@skymind.io> * Added test for shuffle. * Fixed ops declarations. * Restored omp and added shuffle test. * Added convertTypes test. * Added tests for execRandom. Eliminated usage of RandomBuffer with NativeOps. * Added sort tests. * Added tests for execCustomOp. * - further debuging and fixing tests terminated with crash Signed-off-by: Yurii <yurii@skymind.io> * Added tests for calculateOutputShapes. * Addded Benchmarks test. * Commented benchmark tests. * change assertion Signed-off-by: raver119 <raver119@gmail.com> * Added tests for apply_sgd op. Added cpu helper for that op. * Implement cuda helper for aplly_sgd op. Fixed tests for NativeOps. * Added test for assign broadcastable. * Added tests for assign_bp op. * Added tests for axpy op. * - assign/execScalar/execTransformAny signature change - minor test fix Signed-off-by: raver119 <raver119@gmail.com> * Fixed axpy op. * meh Signed-off-by: raver119 <raver119@gmail.com> * - fix tests for nativeOps::concat Signed-off-by: Yurii <yurii@skymind.io> * sequential transform/scalar Signed-off-by: raver119 <raver119@gmail.com> * allow nested parallelism Signed-off-by: raver119 <raver119@gmail.com> * assign_bp leak fix Signed-off-by: raver119 <raver119@gmail.com> * block setRNG fix Signed-off-by: raver119 <raver119@gmail.com> * enable parallelism by default Signed-off-by: raver119 <raver119@gmail.com> * enable nested parallelism by default Signed-off-by: raver119 <raver119@gmail.com> * Added cuda implementation for row_count helper. * Added implementation for tnse gains op helper. * - take into account possible situations when input arrays are empty in reduce_ cuda stuff Signed-off-by: Yurii <yurii@skymind.io> * Implemented tsne/edge_forces op cuda-based helper. Parallelized cpu-based helper for edge_forces. * Added kernel for tsne/symmetrized op heleper. * Implementation of tsne/symmetrized op cuda helper. Working edition. * Eliminated waste printfs. * Added test for broadcastgradientargs op. * host-only fallback for empty reduce float Signed-off-by: raver119 <raver119@gmail.com> * - some tests fixes Signed-off-by: Yurii <yurii@skymind.io> * - correct the rest of reduce_ stuff Signed-off-by: Yurii <yurii@skymind.io> * - further correction of reduce_ stuff Signed-off-by: Yurii <yurii@skymind.io> * Added test for Cbow op. Also added cuda implementation for cbow helpers. * - improve code of stack operation for scalar case Signed-off-by: Yurii <yurii@skymind.io> * - provide cuda kernel for gatherND operation Signed-off-by: Yurii <yurii@skymind.io> * Implementation of cbow helpers with cuda kernels. * minor tests tweaks Signed-off-by: raver119 <raver119@gmail.com> * minor tests tweaks Signed-off-by: raver119 <raver119@gmail.com> * - further correction of cuda stuff Signed-off-by: Yurii <yurii@skymind.io> * Implementatation of cbow op helper with cuda kernels. Working edition. * Skip random testing for cudablas case. * lstmBlockCell context fix Signed-off-by: raver119 <raver119@gmail.com> * Added tests for ELU and ELU_BP ops. * Added tests for eq_scalar, gt_scalar, gte_scalar and lte_scalar ops. * Added tests for neq_scalar. * Added test for noop. * - further work on clipbynorm_bp Signed-off-by: Yurii <yurii@skymind.io> * - get rid of concat op call, use instead direct concat helper call Signed-off-by: Yurii <yurii@skymind.io> * lstmBlockCell context fix Signed-off-by: raver119 <raver119@gmail.com> * Added tests for lrelu and lrelu_bp. * Added tests for selu and selu_bp. * Fixed lrelu derivative helpers. * - some corrections in lstm Signed-off-by: Yurii <yurii@skymind.io> * operator * result shape fix Signed-off-by: raver119 <raver119@gmail.com> * - correct typo in lstmCell Signed-off-by: Yurii <yurii@skymind.io> * few tests fixed Signed-off-by: raver119 <raver119@gmail.com> * CUDA inverse broadcast bool fix Signed-off-by: raver119 <raver119@gmail.com> * disable MMAP test for CUDA Signed-off-by: raver119 <raver119@gmail.com> * BooleanOp syncToDevice Signed-off-by: raver119 <raver119@gmail.com> * meh Signed-off-by: raver119 <raver119@gmail.com> * additional data types for im2col/col2im Signed-off-by: raver119 <raver119@gmail.com> * Added test for firas_sparse op. * one more RandomBuffer test excluded Signed-off-by: raver119 <raver119@gmail.com> * Added tests for flatten op. * Added test for Floor op. * bunch of tests fixed Signed-off-by: raver119 <raver119@gmail.com> * mmulDot tests fixed Signed-off-by: raver119 <raver119@gmail.com> * more tests fixed Signed-off-by: raver119 <raver119@gmail.com> * Implemented floordiv_bp op and tests. * Fixed scalar case with cuda implementation for bds. * - work on cuda kernel for clip_by_norm backprop op is completed Signed-off-by: Yurii <yurii@skymind.io> * Eliminate cbow crach. * more tests fixed Signed-off-by: raver119 <raver119@gmail.com> * more tests fixed Signed-off-by: raver119 <raver119@gmail.com> * Eliminated abortion with batched nlp test. * more tests fixed Signed-off-by: raver119 <raver119@gmail.com> * Fixed shared flag initializing. * disabled bunch of cpu workspaces tests Signed-off-by: raver119 <raver119@gmail.com> * scalar operators fix: missing registerSpecialUse call Signed-off-by: raver119 <raver119@gmail.com> * Fixed logdet for cuda and tests. * - correct clipBynorm_bp Signed-off-by: Yurii <yurii@skymind.io> * Fixed crop_and_resize shape datatype. * - correct some mmul tests Signed-off-by: Yurii <yurii@skymind.io>
2019-08-02 19:01:03 +02:00
auto xType = sd::ArrayOptions::dataType(hXShapeInfo);
auto zType = sd::ArrayOptions::dataType(hZShapeInfo);
2019-06-06 14:21:15 +02:00
// nothing to do here if result is empty
if (shape::isEmpty(hZShapeInfo))
return;
auto func = PRAGMA_THREADS_FOR {
BUILD_DOUBLE_SELECTOR(xType, zType, functions::reduce::ReduceFloatFunction, ::exec(opNum, hX, hXShapeInfo, extraParams, hZ, hZShapeInfo, dimension, dimensionLength, tadShapeInfo, tadOffsets, start, stop), LIBND4J_TYPES, FLOAT_TYPES);
};
const sd::LoopKind::Kind kindOfLoop = sd::LoopKind::deduceKindOfLoopTadXZ(hXShapeInfo, hZShapeInfo, tadShapeInfo);
samediff::Threads::parallel_tad(func, 0, shape::length(hZShapeInfo), 1, kindOfLoop == sd::LoopKind::Kind::SMALLARR2DX ? 1 : sd::Environment::getInstance()->maxMasterThreads());
2019-06-06 14:21:15 +02:00
}
////////////////////////////////////////////////////////////////////////
void NativeOpExecutioner::execReduceSame(sd::LaunchContext *lc,
2019-06-06 14:21:15 +02:00
int opNum,
void *hX, Nd4jLong *hXShapeInfo,
void *dX, Nd4jLong *dXShapeInfo,
void *extraParams,
void *hZ, Nd4jLong *hZShapeInfo,
void *dZ, Nd4jLong *dZShapeInfo,
int *dimension, int dimensionLength,
Nd4jLong *tadShapeInfo, Nd4jLong *tadOffsets) {
2019-06-06 14:21:15 +02:00
auto xType = sd::ArrayOptions::dataType(hXShapeInfo);
auto zType = sd::ArrayOptions::dataType(hZShapeInfo);
2019-06-06 14:21:15 +02:00
// nothing to do here if result is empty
if (shape::isEmpty(hZShapeInfo))
return;
auto func = PRAGMA_THREADS_FOR {
BUILD_SINGLE_SELECTOR(xType, functions::reduce::ReduceSameFunction, ::exec(opNum, hX, hXShapeInfo, extraParams, hZ, hZShapeInfo, dimension, dimensionLength, tadShapeInfo, tadOffsets, start, stop), LIBND4J_TYPES);
};
const sd::LoopKind::Kind kindOfLoop = sd::LoopKind::deduceKindOfLoopTadXZ(hXShapeInfo, hZShapeInfo, tadShapeInfo);
samediff::Threads::parallel_tad(func, 0, shape::length(hZShapeInfo), 1, kindOfLoop == sd::LoopKind::Kind::SMALLARR2DX ? 1 : sd::Environment::getInstance()->maxMasterThreads());
2019-06-06 14:21:15 +02:00
}
////////////////////////////////////////////////////////////////////////
void NativeOpExecutioner::execReduceBool(sd::LaunchContext *lc,
2019-06-06 14:21:15 +02:00
int opNum,
void *hX, Nd4jLong *hXShapeInfo,
void *dX, Nd4jLong *dXShapeInfo,
void *extraParams,
void *hZ, Nd4jLong *hZShapeInfo,
void *dZ, Nd4jLong *dZShapeInfo,
int *dimension, int dimensionLength,
Nd4jLong *tadShapeInfo, Nd4jLong *tadOffsets) {
2019-06-06 14:21:15 +02:00
auto xType = sd::ArrayOptions::dataType(hXShapeInfo);
auto zType = sd::ArrayOptions::dataType(hZShapeInfo);
2019-06-06 14:21:15 +02:00
// nothing to do here if result is empty
if (shape::isEmpty(hZShapeInfo))
return;
auto func = PRAGMA_THREADS_FOR {
BUILD_DOUBLE_SELECTOR(xType, zType, functions::reduce::ReduceBoolFunction, ::exec(opNum, hX, hXShapeInfo, extraParams, hZ, hZShapeInfo, dimension, dimensionLength, tadShapeInfo, tadOffsets, start, stop), LIBND4J_TYPES, BOOL_TYPES);
};
const sd::LoopKind::Kind kindOfLoop = sd::LoopKind::deduceKindOfLoopTadXZ(hXShapeInfo, hZShapeInfo, tadShapeInfo);
samediff::Threads::parallel_tad(func, 0, shape::length(hZShapeInfo), 1, kindOfLoop == sd::LoopKind::Kind::SMALLARR2DX ? 1 : sd::Environment::getInstance()->maxMasterThreads());
2019-06-06 14:21:15 +02:00
}
////////////////////////////////////////////////////////////////////////
void NativeOpExecutioner::execReduceLong(sd::LaunchContext *lc,
2019-06-06 14:21:15 +02:00
int opNum,
void *hX, Nd4jLong *hXShapeInfo,
void *dX, Nd4jLong *dXShapeInfo,
void *extraParams,
void *hZ, Nd4jLong *hZShapeInfo,
void *dZ, Nd4jLong *dZShapeInfo,
int *dimension, int dimensionLength,
Nd4jLong *tadShapeInfo, Nd4jLong *tadOffsets) {
2019-06-06 14:21:15 +02:00
auto xType = sd::ArrayOptions::dataType(hXShapeInfo);
auto zType = sd::ArrayOptions::dataType(hZShapeInfo);
2019-06-06 14:21:15 +02:00
// nothing to do here if result is empty
if (shape::isEmpty(hZShapeInfo))
return;
auto func = PRAGMA_THREADS_FOR {
BUILD_DOUBLE_SELECTOR(xType, zType, functions::reduce::ReduceLongFunction, ::exec(opNum, hX, hXShapeInfo, extraParams, hZ, hZShapeInfo, dimension, dimensionLength, tadShapeInfo, tadOffsets, start, stop), LIBND4J_TYPES, LONG_TYPES);
};
const sd::LoopKind::Kind kindOfLoop = sd::LoopKind::deduceKindOfLoopTadXZ(hXShapeInfo, hZShapeInfo, tadShapeInfo);
samediff::Threads::parallel_tad(func, 0, shape::length(hZShapeInfo), 1, kindOfLoop == sd::LoopKind::Kind::SMALLARR2DX ? 1 : sd::Environment::getInstance()->maxMasterThreads());
2019-06-06 14:21:15 +02:00
}
////////////////////////////////////////////////////////////////////////
/**
*
* @param opNum
* @param hX
* @param hXShapeInfo
* @param extraParams
* @return
*/
void NativeOpExecutioner::execReduceFloatScalar(sd::LaunchContext *lc,
2019-06-06 14:21:15 +02:00
int opNum,
void *hX, Nd4jLong *hXShapeInfo,
void *dX, Nd4jLong *dXShapeInfo,
void *extraParams,
void *hZ, Nd4jLong *hZShapeInfo,
void *dZ, Nd4jLong *dZShapeInfo) {
2019-06-06 14:21:15 +02:00
auto xType = sd::ArrayOptions::dataType(hXShapeInfo);
auto zType = sd::ArrayOptions::dataType(hZShapeInfo);
2019-06-06 14:21:15 +02:00
BUILD_DOUBLE_SELECTOR(xType, zType, functions::reduce::ReduceFloatFunction, ::execScalar(opNum, hX, hXShapeInfo, extraParams, hZ, hZShapeInfo), LIBND4J_TYPES, FLOAT_TYPES);
}
////////////////////////////////////////////////////////////////////////
void NativeOpExecutioner::execReduceSameScalar(sd::LaunchContext *lc,
2019-06-06 14:21:15 +02:00
int opNum,
void *hX, Nd4jLong *hXShapeInfo,
void *dX, Nd4jLong *dXShapeInfo,
void *extraParams,
void *hZ, Nd4jLong *hZShapeInfo,
void *dZ, Nd4jLong *dZShapeInfo) {
2019-06-06 14:21:15 +02:00
auto xType = sd::ArrayOptions::dataType(hXShapeInfo);
2019-06-06 14:21:15 +02:00
BUILD_SINGLE_SELECTOR(xType, functions::reduce::ReduceSameFunction, ::execScalar(opNum, hX, hXShapeInfo, extraParams, hZ, hZShapeInfo), LIBND4J_TYPES);
}
////////////////////////////////////////////////////////////////////////
void NativeOpExecutioner::execReduceBoolScalar(sd::LaunchContext *lc,
2019-06-06 14:21:15 +02:00
int opNum,
void *hX, Nd4jLong *hXShapeInfo,
void *dX, Nd4jLong *dXShapeInfo,
void *extraParams,
void *hZ, Nd4jLong *hZShapeInfo,
void *dZ, Nd4jLong *dZShapeInfo) {
[WIP] More of CUDA (#95) * initial commit Signed-off-by: raver119 <raver119@gmail.com> * Implementation of hashcode cuda helper. Working edition. * Fixed parallel test input arangements. * Fixed tests for hashcode op. * Fixed shape calculation for image:crop_and_resize op and test. * NativeOps tests. Initial test suite. * Added tests for indexReduce methods. * Added test on execBroadcast with NDArray as dimensions. * Added test on execBroadcastBool with NDArray as dimensions. * Added tests on execPairwiseTransform and execPairwiseTransofrmBool. * Added tests for execReduce with scalar results. * Added reduce tests for non-empty dims array. * Added tests for reduce3. * Added tests for execScalar. * Added tests for execSummaryStats. * - provide cpu/cuda code for batch_to_space - testing it Signed-off-by: Yurii <yurii@skymind.io> * - remove old test for batch_to_space (had wrong format and numbers were not checked) Signed-off-by: Yurii <yurii@skymind.io> * Fixed complilation errors with test. * Added test for execTransformFloat. * Added test for execTransformSame. * Added test for execTransformBool. * Added test for execTransformStrict. * Added tests for execScalar/execScalarBool with TADs. * Added test for flatten. * - provide cpu/cuda code for space_to_Batch operaion Signed-off-by: Yurii <yurii@skymind.io> * Added test for concat. * comment unnecessary stuff in s_t_b Signed-off-by: Yurii <yurii@skymind.io> * Added test for specialConcat. * Added tests for memcpy/set routines. * Fixed pullRow cuda test. * Added pullRow test. * Added average test. * - correct typo in NDArray::applyPairwiseTransform(nd4j::pairwise::BoolOps op...) Signed-off-by: Yurii <yurii@skymind.io> * - debugging and fixing cuda tests in JavaInteropTests file Signed-off-by: Yurii <yurii@skymind.io> * - correct some tests Signed-off-by: Yurii <yurii@skymind.io> * Added test for shuffle. * Fixed ops declarations. * Restored omp and added shuffle test. * Added convertTypes test. * Added tests for execRandom. Eliminated usage of RandomBuffer with NativeOps. * Added sort tests. * Added tests for execCustomOp. * - further debuging and fixing tests terminated with crash Signed-off-by: Yurii <yurii@skymind.io> * Added tests for calculateOutputShapes. * Addded Benchmarks test. * Commented benchmark tests. * change assertion Signed-off-by: raver119 <raver119@gmail.com> * Added tests for apply_sgd op. Added cpu helper for that op. * Implement cuda helper for aplly_sgd op. Fixed tests for NativeOps. * Added test for assign broadcastable. * Added tests for assign_bp op. * Added tests for axpy op. * - assign/execScalar/execTransformAny signature change - minor test fix Signed-off-by: raver119 <raver119@gmail.com> * Fixed axpy op. * meh Signed-off-by: raver119 <raver119@gmail.com> * - fix tests for nativeOps::concat Signed-off-by: Yurii <yurii@skymind.io> * sequential transform/scalar Signed-off-by: raver119 <raver119@gmail.com> * allow nested parallelism Signed-off-by: raver119 <raver119@gmail.com> * assign_bp leak fix Signed-off-by: raver119 <raver119@gmail.com> * block setRNG fix Signed-off-by: raver119 <raver119@gmail.com> * enable parallelism by default Signed-off-by: raver119 <raver119@gmail.com> * enable nested parallelism by default Signed-off-by: raver119 <raver119@gmail.com> * Added cuda implementation for row_count helper. * Added implementation for tnse gains op helper. * - take into account possible situations when input arrays are empty in reduce_ cuda stuff Signed-off-by: Yurii <yurii@skymind.io> * Implemented tsne/edge_forces op cuda-based helper. Parallelized cpu-based helper for edge_forces. * Added kernel for tsne/symmetrized op heleper. * Implementation of tsne/symmetrized op cuda helper. Working edition. * Eliminated waste printfs. * Added test for broadcastgradientargs op. * host-only fallback for empty reduce float Signed-off-by: raver119 <raver119@gmail.com> * - some tests fixes Signed-off-by: Yurii <yurii@skymind.io> * - correct the rest of reduce_ stuff Signed-off-by: Yurii <yurii@skymind.io> * - further correction of reduce_ stuff Signed-off-by: Yurii <yurii@skymind.io> * Added test for Cbow op. Also added cuda implementation for cbow helpers. * - improve code of stack operation for scalar case Signed-off-by: Yurii <yurii@skymind.io> * - provide cuda kernel for gatherND operation Signed-off-by: Yurii <yurii@skymind.io> * Implementation of cbow helpers with cuda kernels. * minor tests tweaks Signed-off-by: raver119 <raver119@gmail.com> * minor tests tweaks Signed-off-by: raver119 <raver119@gmail.com> * - further correction of cuda stuff Signed-off-by: Yurii <yurii@skymind.io> * Implementatation of cbow op helper with cuda kernels. Working edition. * Skip random testing for cudablas case. * lstmBlockCell context fix Signed-off-by: raver119 <raver119@gmail.com> * Added tests for ELU and ELU_BP ops. * Added tests for eq_scalar, gt_scalar, gte_scalar and lte_scalar ops. * Added tests for neq_scalar. * Added test for noop. * - further work on clipbynorm_bp Signed-off-by: Yurii <yurii@skymind.io> * - get rid of concat op call, use instead direct concat helper call Signed-off-by: Yurii <yurii@skymind.io> * lstmBlockCell context fix Signed-off-by: raver119 <raver119@gmail.com> * Added tests for lrelu and lrelu_bp. * Added tests for selu and selu_bp. * Fixed lrelu derivative helpers. * - some corrections in lstm Signed-off-by: Yurii <yurii@skymind.io> * operator * result shape fix Signed-off-by: raver119 <raver119@gmail.com> * - correct typo in lstmCell Signed-off-by: Yurii <yurii@skymind.io> * few tests fixed Signed-off-by: raver119 <raver119@gmail.com> * CUDA inverse broadcast bool fix Signed-off-by: raver119 <raver119@gmail.com> * disable MMAP test for CUDA Signed-off-by: raver119 <raver119@gmail.com> * BooleanOp syncToDevice Signed-off-by: raver119 <raver119@gmail.com> * meh Signed-off-by: raver119 <raver119@gmail.com> * additional data types for im2col/col2im Signed-off-by: raver119 <raver119@gmail.com> * Added test for firas_sparse op. * one more RandomBuffer test excluded Signed-off-by: raver119 <raver119@gmail.com> * Added tests for flatten op. * Added test for Floor op. * bunch of tests fixed Signed-off-by: raver119 <raver119@gmail.com> * mmulDot tests fixed Signed-off-by: raver119 <raver119@gmail.com> * more tests fixed Signed-off-by: raver119 <raver119@gmail.com> * Implemented floordiv_bp op and tests. * Fixed scalar case with cuda implementation for bds. * - work on cuda kernel for clip_by_norm backprop op is completed Signed-off-by: Yurii <yurii@skymind.io> * Eliminate cbow crach. * more tests fixed Signed-off-by: raver119 <raver119@gmail.com> * more tests fixed Signed-off-by: raver119 <raver119@gmail.com> * Eliminated abortion with batched nlp test. * more tests fixed Signed-off-by: raver119 <raver119@gmail.com> * Fixed shared flag initializing. * disabled bunch of cpu workspaces tests Signed-off-by: raver119 <raver119@gmail.com> * scalar operators fix: missing registerSpecialUse call Signed-off-by: raver119 <raver119@gmail.com> * Fixed logdet for cuda and tests. * - correct clipBynorm_bp Signed-off-by: Yurii <yurii@skymind.io> * Fixed crop_and_resize shape datatype. * - correct some mmul tests Signed-off-by: Yurii <yurii@skymind.io>
2019-08-02 19:01:03 +02:00
auto xType = sd::ArrayOptions::dataType(hXShapeInfo);
auto zType = sd::ArrayOptions::dataType(hZShapeInfo);
2019-06-06 14:21:15 +02:00
BUILD_DOUBLE_SELECTOR(xType, zType, functions::reduce::ReduceBoolFunction, ::execScalar(opNum, hX, hXShapeInfo, extraParams, hZ, hZShapeInfo), LIBND4J_TYPES, BOOL_TYPES);
}
////////////////////////////////////////////////////////////////////////
void NativeOpExecutioner::execReduceLongScalar(sd::LaunchContext *lc,
2019-06-06 14:21:15 +02:00
int opNum,
void *hX, Nd4jLong *hXShapeInfo,
void *dX, Nd4jLong *dXShapeInfo,
void *extraParams,
void *hZ, Nd4jLong *hZShapeInfo,
void *dZ, Nd4jLong *dZShapeInfo) {
2019-06-06 14:21:15 +02:00
auto xType = sd::ArrayOptions::dataType(hXShapeInfo);
auto zType = sd::ArrayOptions::dataType(hZShapeInfo);
2019-06-06 14:21:15 +02:00
BUILD_DOUBLE_SELECTOR(xType, zType, functions::reduce::ReduceLongFunction, ::execScalar(opNum, hX, hXShapeInfo, extraParams, hZ, hZShapeInfo), LIBND4J_TYPES, LONG_TYPES);
}
////////////////////////////////////////////////////////////////////////
/**
*
* @param opNum
* @param hX
* @param hXShapeInfo
* @param extraParamsVals
* @param hY
* @param hYShapeInfo
* @param hZ
* @param hZShapeInfo
* @param dimension
* @param dimensionLength
*/
void NativeOpExecutioner::execReduce3Scalar(sd::LaunchContext *lc,
2019-06-06 14:21:15 +02:00
int opNum,
void *hX, Nd4jLong *hXShapeInfo,
void *dX, Nd4jLong *dXShapeInfo,
void *extraParamsVals,
void *hY, Nd4jLong *hYShapeInfo,
void *dY, Nd4jLong *dYShapeInfo,
void *hZ, Nd4jLong *hZShapeInfo,
void *dZ, Nd4jLong *dZShapeInfo) {
auto xType = sd::ArrayOptions::dataType(hXShapeInfo);
auto zType = sd::ArrayOptions::dataType(hZShapeInfo);
2019-06-06 14:21:15 +02:00
BUILD_DOUBLE_SELECTOR(xType, zType, functions::reduce3::Reduce3, ::execScalar(opNum, hX, hXShapeInfo, extraParamsVals, hY, hYShapeInfo, hZ, hZShapeInfo), LIBND4J_TYPES, FLOAT_TYPES);
}
////////////////////////////////////////////////////////////////////////
/**
*
* @param opNum
* @param hX
* @param hXShapeInfo
* @param extraParamsVals
* @param hY
* @param hYShapeInfo
* @param hZ
* @param hZShapeInfo
*/
void NativeOpExecutioner::execReduce3(sd::LaunchContext *lc,
2019-06-06 14:21:15 +02:00
int opNum,
void *hX, Nd4jLong *hXShapeInfo,
void *dX, Nd4jLong *dXShapeInfo,
void *extraParamsVals,
void *hY, Nd4jLong *hYShapeInfo,
void *dY, Nd4jLong *dYShapeInfo,
void *hZ, Nd4jLong *hZShapeInfo,
void *dZ, Nd4jLong *dZShapeInfo) {
2019-06-06 14:21:15 +02:00
auto xType = sd::ArrayOptions::dataType(hXShapeInfo);
auto zType = sd::ArrayOptions::dataType(hZShapeInfo);
2019-06-06 14:21:15 +02:00
//BUILD_DOUBLE_SELECTOR(xType, zType, functions::reduce3::Reduce3, ::exec(opNum, hX, hXShapeInfo, extraParamsVals, hY, hYShapeInfo, hZ, hZShapeInfo, nullptr, 0), LIBND4J_TYPES, FLOAT_TYPES);
NativeOpExecutioner::execReduce3Scalar(lc, opNum, hX, hXShapeInfo, dX, dXShapeInfo, extraParamsVals, hY, hYShapeInfo, dY, dYShapeInfo, hZ, hZShapeInfo, dZ, dZShapeInfo);
2019-06-06 14:21:15 +02:00
}
////////////////////////////////////////////////////////////////////////
void NativeOpExecutioner::execReduce3(sd::LaunchContext *lc,
2019-06-06 14:21:15 +02:00
int opNum,
void *hX, Nd4jLong *hXShapeInfo,
void *dX, Nd4jLong *dXShapeInfo,
void *extraParamsVals,
void *hY, Nd4jLong *hYShapeInfo,
void *dY, Nd4jLong *dYShapeInfo,
void *hZ, Nd4jLong *hZShapeInfo,
void *dZ, Nd4jLong *dZShapeInfo,
int *dimension, int dimensionLength,
Nd4jLong *xTadOnlyShapeInfo, Nd4jLong *xTadOffsets,
Nd4jLong *yTadOnlyShapeInfo, Nd4jLong *yTadOffsets) {
2019-06-06 14:21:15 +02:00
auto xType = sd::ArrayOptions::dataType(hXShapeInfo);
auto zType = sd::ArrayOptions::dataType(hZShapeInfo);
2019-06-06 14:21:15 +02:00
const auto xLen = shape::length(hXShapeInfo);
const auto yLen = shape::length(hYShapeInfo);
sd::TadPack tadPack;
if(xLen == yLen) {
tadPack = sd::ConstantTadHelper::getInstance()->tadForDimensions(hXShapeInfo, dimension, dimensionLength);
}
else if(yLen > xLen) {
tadPack = sd::ConstantTadHelper::getInstance()->tadForDimensions(hYShapeInfo, dimension, dimensionLength);
}
else {
tadPack = sd::ConstantTadHelper::getInstance()->tadForDimensions(hXShapeInfo, dimension, dimensionLength);
}
auto func = PRAGMA_THREADS_FOR {
BUILD_DOUBLE_SELECTOR(xType, zType, functions::reduce3::Reduce3, ::exec(opNum, hX, hXShapeInfo, extraParamsVals, hY, hYShapeInfo, hZ, hZShapeInfo, dimension, dimensionLength, start, stop), LIBND4J_TYPES, FLOAT_TYPES);
};
samediff::Threads::parallel_tad(func, 0, tadPack.numberOfTads());
2019-06-06 14:21:15 +02:00
}
////////////////////////////////////////////////////////////////////////
void NativeOpExecutioner::execReduce3All(sd::LaunchContext *lc,
2019-06-06 14:21:15 +02:00
int opNum,
void *hX, Nd4jLong *hXShapeInfo,
void *dX, Nd4jLong *dXShapeInfo,
void *extraParamsVals,
void *hY, Nd4jLong *hYShapeInfo,
void *dY, Nd4jLong *dYShapeInfo,
void *hZ, Nd4jLong *hZShapeInfo,
void *dZ, Nd4jLong *dZShapeInfo,
int *dimension, int dimensionLength,
Nd4jLong *xTadShapeInfo, Nd4jLong *xOffsets,
Nd4jLong *yTadShapeInfo, Nd4jLong *yOffsets) {
2019-06-06 14:21:15 +02:00
auto xType = sd::ArrayOptions::dataType(hXShapeInfo);
auto zType = sd::ArrayOptions::dataType(hZShapeInfo);
2019-06-06 14:21:15 +02:00
auto tadPack = sd::ConstantTadHelper::getInstance()->tadForDimensions(hXShapeInfo, dimension, dimensionLength);
// TODO: make it 2d
auto func = PRAGMA_THREADS_FOR {
BUILD_DOUBLE_SELECTOR(xType, zType, functions::reduce3::Reduce3, ::execAll(opNum, hX, hXShapeInfo, extraParamsVals, hY, hYShapeInfo, hZ, hZShapeInfo, dimension, dimensionLength, xTadShapeInfo, xOffsets, yTadShapeInfo, yOffsets, start, stop), LIBND4J_TYPES, FLOAT_TYPES);
};
samediff::Threads::parallel_tad(func, 0, tadPack.numberOfTads());
2019-06-06 14:21:15 +02:00
}
////////////////////////////////////////////////////////////////////////
void NativeOpExecutioner::execReduce3TAD(sd::LaunchContext *lc,
2019-06-06 14:21:15 +02:00
int opNum,
void *hX, Nd4jLong *hXShapeInfo,
void *dX, Nd4jLong *dXShapeInfo,
void *extraParamsVals,
void *hY, Nd4jLong *hYShapeInfo,
void *dY, Nd4jLong *dYShapeInfo,
void *hZ, Nd4jLong *hZShapeInfo,
void *dZ, Nd4jLong *dZShapeInfo,
int *dimension, int dimensionLength,
Nd4jLong *tadShapeInfo, Nd4jLong *tadOffsets,
Nd4jLong *yTadShapeInfo, Nd4jLong *yTadOffsets) {
[WIP] More of CUDA (#95) * initial commit Signed-off-by: raver119 <raver119@gmail.com> * Implementation of hashcode cuda helper. Working edition. * Fixed parallel test input arangements. * Fixed tests for hashcode op. * Fixed shape calculation for image:crop_and_resize op and test. * NativeOps tests. Initial test suite. * Added tests for indexReduce methods. * Added test on execBroadcast with NDArray as dimensions. * Added test on execBroadcastBool with NDArray as dimensions. * Added tests on execPairwiseTransform and execPairwiseTransofrmBool. * Added tests for execReduce with scalar results. * Added reduce tests for non-empty dims array. * Added tests for reduce3. * Added tests for execScalar. * Added tests for execSummaryStats. * - provide cpu/cuda code for batch_to_space - testing it Signed-off-by: Yurii <yurii@skymind.io> * - remove old test for batch_to_space (had wrong format and numbers were not checked) Signed-off-by: Yurii <yurii@skymind.io> * Fixed complilation errors with test. * Added test for execTransformFloat. * Added test for execTransformSame. * Added test for execTransformBool. * Added test for execTransformStrict. * Added tests for execScalar/execScalarBool with TADs. * Added test for flatten. * - provide cpu/cuda code for space_to_Batch operaion Signed-off-by: Yurii <yurii@skymind.io> * Added test for concat. * comment unnecessary stuff in s_t_b Signed-off-by: Yurii <yurii@skymind.io> * Added test for specialConcat. * Added tests for memcpy/set routines. * Fixed pullRow cuda test. * Added pullRow test. * Added average test. * - correct typo in NDArray::applyPairwiseTransform(nd4j::pairwise::BoolOps op...) Signed-off-by: Yurii <yurii@skymind.io> * - debugging and fixing cuda tests in JavaInteropTests file Signed-off-by: Yurii <yurii@skymind.io> * - correct some tests Signed-off-by: Yurii <yurii@skymind.io> * Added test for shuffle. * Fixed ops declarations. * Restored omp and added shuffle test. * Added convertTypes test. * Added tests for execRandom. Eliminated usage of RandomBuffer with NativeOps. * Added sort tests. * Added tests for execCustomOp. * - further debuging and fixing tests terminated with crash Signed-off-by: Yurii <yurii@skymind.io> * Added tests for calculateOutputShapes. * Addded Benchmarks test. * Commented benchmark tests. * change assertion Signed-off-by: raver119 <raver119@gmail.com> * Added tests for apply_sgd op. Added cpu helper for that op. * Implement cuda helper for aplly_sgd op. Fixed tests for NativeOps. * Added test for assign broadcastable. * Added tests for assign_bp op. * Added tests for axpy op. * - assign/execScalar/execTransformAny signature change - minor test fix Signed-off-by: raver119 <raver119@gmail.com> * Fixed axpy op. * meh Signed-off-by: raver119 <raver119@gmail.com> * - fix tests for nativeOps::concat Signed-off-by: Yurii <yurii@skymind.io> * sequential transform/scalar Signed-off-by: raver119 <raver119@gmail.com> * allow nested parallelism Signed-off-by: raver119 <raver119@gmail.com> * assign_bp leak fix Signed-off-by: raver119 <raver119@gmail.com> * block setRNG fix Signed-off-by: raver119 <raver119@gmail.com> * enable parallelism by default Signed-off-by: raver119 <raver119@gmail.com> * enable nested parallelism by default Signed-off-by: raver119 <raver119@gmail.com> * Added cuda implementation for row_count helper. * Added implementation for tnse gains op helper. * - take into account possible situations when input arrays are empty in reduce_ cuda stuff Signed-off-by: Yurii <yurii@skymind.io> * Implemented tsne/edge_forces op cuda-based helper. Parallelized cpu-based helper for edge_forces. * Added kernel for tsne/symmetrized op heleper. * Implementation of tsne/symmetrized op cuda helper. Working edition. * Eliminated waste printfs. * Added test for broadcastgradientargs op. * host-only fallback for empty reduce float Signed-off-by: raver119 <raver119@gmail.com> * - some tests fixes Signed-off-by: Yurii <yurii@skymind.io> * - correct the rest of reduce_ stuff Signed-off-by: Yurii <yurii@skymind.io> * - further correction of reduce_ stuff Signed-off-by: Yurii <yurii@skymind.io> * Added test for Cbow op. Also added cuda implementation for cbow helpers. * - improve code of stack operation for scalar case Signed-off-by: Yurii <yurii@skymind.io> * - provide cuda kernel for gatherND operation Signed-off-by: Yurii <yurii@skymind.io> * Implementation of cbow helpers with cuda kernels. * minor tests tweaks Signed-off-by: raver119 <raver119@gmail.com> * minor tests tweaks Signed-off-by: raver119 <raver119@gmail.com> * - further correction of cuda stuff Signed-off-by: Yurii <yurii@skymind.io> * Implementatation of cbow op helper with cuda kernels. Working edition. * Skip random testing for cudablas case. * lstmBlockCell context fix Signed-off-by: raver119 <raver119@gmail.com> * Added tests for ELU and ELU_BP ops. * Added tests for eq_scalar, gt_scalar, gte_scalar and lte_scalar ops. * Added tests for neq_scalar. * Added test for noop. * - further work on clipbynorm_bp Signed-off-by: Yurii <yurii@skymind.io> * - get rid of concat op call, use instead direct concat helper call Signed-off-by: Yurii <yurii@skymind.io> * lstmBlockCell context fix Signed-off-by: raver119 <raver119@gmail.com> * Added tests for lrelu and lrelu_bp. * Added tests for selu and selu_bp. * Fixed lrelu derivative helpers. * - some corrections in lstm Signed-off-by: Yurii <yurii@skymind.io> * operator * result shape fix Signed-off-by: raver119 <raver119@gmail.com> * - correct typo in lstmCell Signed-off-by: Yurii <yurii@skymind.io> * few tests fixed Signed-off-by: raver119 <raver119@gmail.com> * CUDA inverse broadcast bool fix Signed-off-by: raver119 <raver119@gmail.com> * disable MMAP test for CUDA Signed-off-by: raver119 <raver119@gmail.com> * BooleanOp syncToDevice Signed-off-by: raver119 <raver119@gmail.com> * meh Signed-off-by: raver119 <raver119@gmail.com> * additional data types for im2col/col2im Signed-off-by: raver119 <raver119@gmail.com> * Added test for firas_sparse op. * one more RandomBuffer test excluded Signed-off-by: raver119 <raver119@gmail.com> * Added tests for flatten op. * Added test for Floor op. * bunch of tests fixed Signed-off-by: raver119 <raver119@gmail.com> * mmulDot tests fixed Signed-off-by: raver119 <raver119@gmail.com> * more tests fixed Signed-off-by: raver119 <raver119@gmail.com> * Implemented floordiv_bp op and tests. * Fixed scalar case with cuda implementation for bds. * - work on cuda kernel for clip_by_norm backprop op is completed Signed-off-by: Yurii <yurii@skymind.io> * Eliminate cbow crach. * more tests fixed Signed-off-by: raver119 <raver119@gmail.com> * more tests fixed Signed-off-by: raver119 <raver119@gmail.com> * Eliminated abortion with batched nlp test. * more tests fixed Signed-off-by: raver119 <raver119@gmail.com> * Fixed shared flag initializing. * disabled bunch of cpu workspaces tests Signed-off-by: raver119 <raver119@gmail.com> * scalar operators fix: missing registerSpecialUse call Signed-off-by: raver119 <raver119@gmail.com> * Fixed logdet for cuda and tests. * - correct clipBynorm_bp Signed-off-by: Yurii <yurii@skymind.io> * Fixed crop_and_resize shape datatype. * - correct some mmul tests Signed-off-by: Yurii <yurii@skymind.io>
2019-08-02 19:01:03 +02:00
auto xType = sd::ArrayOptions::dataType(hXShapeInfo);
auto zType = sd::ArrayOptions::dataType(hZShapeInfo);
2019-06-06 14:21:15 +02:00
const auto xLen = shape::length(hXShapeInfo);
const auto yLen = shape::length(hYShapeInfo);
sd::TadPack tadPack;
if(xLen == yLen) {
tadPack = sd::ConstantTadHelper::getInstance()->tadForDimensions(hXShapeInfo, dimension, dimensionLength);
}
else if(yLen > xLen) {
tadPack = sd::ConstantTadHelper::getInstance()->tadForDimensions(hYShapeInfo, dimension, dimensionLength);
}
else {
tadPack = sd::ConstantTadHelper::getInstance()->tadForDimensions(hXShapeInfo, dimension, dimensionLength);
}
auto func = PRAGMA_THREADS_FOR {
BUILD_DOUBLE_SELECTOR(xType, zType, functions::reduce3::Reduce3, ::exec(opNum, hX, hXShapeInfo, extraParamsVals, hY, hYShapeInfo, hZ, hZShapeInfo, dimension, dimensionLength, tadShapeInfo, tadOffsets, start, stop), LIBND4J_TYPES, FLOAT_TYPES);
};
samediff::Threads::parallel_tad(func, 0, tadPack.numberOfTads());
2019-06-06 14:21:15 +02:00
}
////////////////////////////////////////////////////////////////////////
/**
*
* @param opNum
* @param hX
* @param xStride
* @param hZ
* @param resultStride
* @param scalar
* @param extraParams
* @param n
*/
void NativeOpExecutioner::execScalar(sd::LaunchContext *lc,
2019-06-06 14:21:15 +02:00
int opNum,
void *hX, Nd4jLong *hXShapeInfo,
void *dX, Nd4jLong *dXShapeInfo,
void *hZ, Nd4jLong *hZShapeInfo,
void *dZ, Nd4jLong *dZShapeInfo,
void *hScalar, Nd4jLong *hScalarShapeInfo,
void *dScalar, Nd4jLong *dScalarShapeInfo,
[WIP] More of CUDA (#95) * initial commit Signed-off-by: raver119 <raver119@gmail.com> * Implementation of hashcode cuda helper. Working edition. * Fixed parallel test input arangements. * Fixed tests for hashcode op. * Fixed shape calculation for image:crop_and_resize op and test. * NativeOps tests. Initial test suite. * Added tests for indexReduce methods. * Added test on execBroadcast with NDArray as dimensions. * Added test on execBroadcastBool with NDArray as dimensions. * Added tests on execPairwiseTransform and execPairwiseTransofrmBool. * Added tests for execReduce with scalar results. * Added reduce tests for non-empty dims array. * Added tests for reduce3. * Added tests for execScalar. * Added tests for execSummaryStats. * - provide cpu/cuda code for batch_to_space - testing it Signed-off-by: Yurii <yurii@skymind.io> * - remove old test for batch_to_space (had wrong format and numbers were not checked) Signed-off-by: Yurii <yurii@skymind.io> * Fixed complilation errors with test. * Added test for execTransformFloat. * Added test for execTransformSame. * Added test for execTransformBool. * Added test for execTransformStrict. * Added tests for execScalar/execScalarBool with TADs. * Added test for flatten. * - provide cpu/cuda code for space_to_Batch operaion Signed-off-by: Yurii <yurii@skymind.io> * Added test for concat. * comment unnecessary stuff in s_t_b Signed-off-by: Yurii <yurii@skymind.io> * Added test for specialConcat. * Added tests for memcpy/set routines. * Fixed pullRow cuda test. * Added pullRow test. * Added average test. * - correct typo in NDArray::applyPairwiseTransform(nd4j::pairwise::BoolOps op...) Signed-off-by: Yurii <yurii@skymind.io> * - debugging and fixing cuda tests in JavaInteropTests file Signed-off-by: Yurii <yurii@skymind.io> * - correct some tests Signed-off-by: Yurii <yurii@skymind.io> * Added test for shuffle. * Fixed ops declarations. * Restored omp and added shuffle test. * Added convertTypes test. * Added tests for execRandom. Eliminated usage of RandomBuffer with NativeOps. * Added sort tests. * Added tests for execCustomOp. * - further debuging and fixing tests terminated with crash Signed-off-by: Yurii <yurii@skymind.io> * Added tests for calculateOutputShapes. * Addded Benchmarks test. * Commented benchmark tests. * change assertion Signed-off-by: raver119 <raver119@gmail.com> * Added tests for apply_sgd op. Added cpu helper for that op. * Implement cuda helper for aplly_sgd op. Fixed tests for NativeOps. * Added test for assign broadcastable. * Added tests for assign_bp op. * Added tests for axpy op. * - assign/execScalar/execTransformAny signature change - minor test fix Signed-off-by: raver119 <raver119@gmail.com> * Fixed axpy op. * meh Signed-off-by: raver119 <raver119@gmail.com> * - fix tests for nativeOps::concat Signed-off-by: Yurii <yurii@skymind.io> * sequential transform/scalar Signed-off-by: raver119 <raver119@gmail.com> * allow nested parallelism Signed-off-by: raver119 <raver119@gmail.com> * assign_bp leak fix Signed-off-by: raver119 <raver119@gmail.com> * block setRNG fix Signed-off-by: raver119 <raver119@gmail.com> * enable parallelism by default Signed-off-by: raver119 <raver119@gmail.com> * enable nested parallelism by default Signed-off-by: raver119 <raver119@gmail.com> * Added cuda implementation for row_count helper. * Added implementation for tnse gains op helper. * - take into account possible situations when input arrays are empty in reduce_ cuda stuff Signed-off-by: Yurii <yurii@skymind.io> * Implemented tsne/edge_forces op cuda-based helper. Parallelized cpu-based helper for edge_forces. * Added kernel for tsne/symmetrized op heleper. * Implementation of tsne/symmetrized op cuda helper. Working edition. * Eliminated waste printfs. * Added test for broadcastgradientargs op. * host-only fallback for empty reduce float Signed-off-by: raver119 <raver119@gmail.com> * - some tests fixes Signed-off-by: Yurii <yurii@skymind.io> * - correct the rest of reduce_ stuff Signed-off-by: Yurii <yurii@skymind.io> * - further correction of reduce_ stuff Signed-off-by: Yurii <yurii@skymind.io> * Added test for Cbow op. Also added cuda implementation for cbow helpers. * - improve code of stack operation for scalar case Signed-off-by: Yurii <yurii@skymind.io> * - provide cuda kernel for gatherND operation Signed-off-by: Yurii <yurii@skymind.io> * Implementation of cbow helpers with cuda kernels. * minor tests tweaks Signed-off-by: raver119 <raver119@gmail.com> * minor tests tweaks Signed-off-by: raver119 <raver119@gmail.com> * - further correction of cuda stuff Signed-off-by: Yurii <yurii@skymind.io> * Implementatation of cbow op helper with cuda kernels. Working edition. * Skip random testing for cudablas case. * lstmBlockCell context fix Signed-off-by: raver119 <raver119@gmail.com> * Added tests for ELU and ELU_BP ops. * Added tests for eq_scalar, gt_scalar, gte_scalar and lte_scalar ops. * Added tests for neq_scalar. * Added test for noop. * - further work on clipbynorm_bp Signed-off-by: Yurii <yurii@skymind.io> * - get rid of concat op call, use instead direct concat helper call Signed-off-by: Yurii <yurii@skymind.io> * lstmBlockCell context fix Signed-off-by: raver119 <raver119@gmail.com> * Added tests for lrelu and lrelu_bp. * Added tests for selu and selu_bp. * Fixed lrelu derivative helpers. * - some corrections in lstm Signed-off-by: Yurii <yurii@skymind.io> * operator * result shape fix Signed-off-by: raver119 <raver119@gmail.com> * - correct typo in lstmCell Signed-off-by: Yurii <yurii@skymind.io> * few tests fixed Signed-off-by: raver119 <raver119@gmail.com> * CUDA inverse broadcast bool fix Signed-off-by: raver119 <raver119@gmail.com> * disable MMAP test for CUDA Signed-off-by: raver119 <raver119@gmail.com> * BooleanOp syncToDevice Signed-off-by: raver119 <raver119@gmail.com> * meh Signed-off-by: raver119 <raver119@gmail.com> * additional data types for im2col/col2im Signed-off-by: raver119 <raver119@gmail.com> * Added test for firas_sparse op. * one more RandomBuffer test excluded Signed-off-by: raver119 <raver119@gmail.com> * Added tests for flatten op. * Added test for Floor op. * bunch of tests fixed Signed-off-by: raver119 <raver119@gmail.com> * mmulDot tests fixed Signed-off-by: raver119 <raver119@gmail.com> * more tests fixed Signed-off-by: raver119 <raver119@gmail.com> * Implemented floordiv_bp op and tests. * Fixed scalar case with cuda implementation for bds. * - work on cuda kernel for clip_by_norm backprop op is completed Signed-off-by: Yurii <yurii@skymind.io> * Eliminate cbow crach. * more tests fixed Signed-off-by: raver119 <raver119@gmail.com> * more tests fixed Signed-off-by: raver119 <raver119@gmail.com> * Eliminated abortion with batched nlp test. * more tests fixed Signed-off-by: raver119 <raver119@gmail.com> * Fixed shared flag initializing. * disabled bunch of cpu workspaces tests Signed-off-by: raver119 <raver119@gmail.com> * scalar operators fix: missing registerSpecialUse call Signed-off-by: raver119 <raver119@gmail.com> * Fixed logdet for cuda and tests. * - correct clipBynorm_bp Signed-off-by: Yurii <yurii@skymind.io> * Fixed crop_and_resize shape datatype. * - correct some mmul tests Signed-off-by: Yurii <yurii@skymind.io>
2019-08-02 19:01:03 +02:00
void *extraParams, bool allowParallelism) {
auto xType = sd::ArrayOptions::dataType(hXShapeInfo);
auto yType = sd::ArrayOptions::dataType(hScalarShapeInfo);
auto zType = sd::ArrayOptions::dataType(hZShapeInfo);
2019-06-06 14:21:15 +02:00
if (shape::isEmpty(hXShapeInfo) || shape::isEmpty(hScalarShapeInfo))
return;
2019-06-06 14:21:15 +02:00
#ifdef __ND4J_EXPERIMENTAL__
BUILD_PAIRWISE_SELECTOR(xType, yType, zType, functions::scalar::ScalarTransform, ::transform(opNum, hX, hXShapeInfo, hZ, hZShapeInfo, hScalar, extraParams), LIBND4J_TYPES, LIBND4J_TYPES);
#else
if (xType != yType || xType != zType)
throw sd::datatype_exception::build("NativeOpExecutioner::execScalar", zType, xType, yType);
2019-06-06 14:21:15 +02:00
auto func = PRAGMA_THREADS_FOR {
BUILD_SINGLE_SELECTOR_THRICE(xType, functions::scalar::ScalarTransform,::transform(opNum, hX, hXShapeInfo, hZ, hZShapeInfo, hScalar, extraParams, start, stop), LIBND4J_TYPES);
};
auto zLen = shape::length(hZShapeInfo);
samediff::Threads::parallel_for(func, 0, zLen, 1, !allowParallelism ? 1 : sd::math::nd4j_max<int>(1, sd::math::nd4j_min<int>(zLen / 1024, sd::Environment::getInstance()->maxMasterThreads())));
2019-06-06 14:21:15 +02:00
#endif
}
////////////////////////////////////////////////////////////////////////
void NativeOpExecutioner::execScalar(sd::LaunchContext *lc,
2019-06-06 14:21:15 +02:00
int opNum,
void *hX, Nd4jLong *hXShapeInfo,
void *dX, Nd4jLong *dXShapeInfo,
void *extraParams,
void *hZ, Nd4jLong *hZShapeInfo,
void *dZ, Nd4jLong *dZShapeInfo,
void *hScalars, Nd4jLong *hScalarShapeInfo,
void *dScalars, Nd4jLong *dScalarShapeInfo,
int *dimension, int dimensionLength,
Nd4jLong *tadShapeInfo, Nd4jLong *tadOffsets,
Nd4jLong *tadShapeInfoZ, Nd4jLong *tadOffsetsZ) {
auto xType = sd::ArrayOptions::dataType(hXShapeInfo);
auto yType = sd::ArrayOptions::dataType(hScalarShapeInfo);
auto zType = sd::ArrayOptions::dataType(hZShapeInfo);
2019-06-06 14:21:15 +02:00
if (shape::isEmpty(hXShapeInfo) || shape::isEmpty(hScalarShapeInfo))
return;
2019-06-06 14:21:15 +02:00
#ifdef __ND4J_EXPERIMENTAL__
BUILD_PAIRWISE_SELECTOR(xType, yType, zType, functions::scalar::ScalarTransform, ::transform(opNum, hX, hXShapeInfo, extraParams, hZ, hZShapeInfo, hScalars, dimension, dimensionLength, tadShapeInfo, tadOffsets, tadShapeInfoZ, tadOffsetsZ), LIBND4J_TYPES, LIBND4J_TYPES);
#else
if (xType != yType || xType != zType)
throw sd::datatype_exception::build("NativeOpExecutioner::execScalar", zType, xType, yType);
2019-06-06 14:21:15 +02:00
auto func = PRAGMA_THREADS_FOR {
BUILD_SINGLE_SELECTOR_THRICE(xType, functions::scalar::ScalarTransform, ::transform(opNum, hX, hXShapeInfo, extraParams, hZ, hZShapeInfo, hScalars, dimension, dimensionLength, tadShapeInfo, tadOffsets, tadShapeInfoZ, tadOffsetsZ, start, stop), LIBND4J_TYPES);
};
auto yLen = shape::length(hScalarShapeInfo);
samediff::Threads::parallel_tad(func, 0, yLen, 1, sd::math::nd4j_min<int>(yLen, sd::Environment::getInstance()->maxMasterThreads()));
2019-06-06 14:21:15 +02:00
#endif
}
////////////////////////////////////////////////////////////////////////
void NativeOpExecutioner::execScalarBool(sd::LaunchContext *lc,
2019-06-06 14:21:15 +02:00
int opNum,
void *hX, Nd4jLong *hXShapeInfo,
void *dX, Nd4jLong *dXShapeInfo,
void *hZ, Nd4jLong *hZShapeInfo,
void *dZ, Nd4jLong *dZShapeInfo,
void *hScalar, Nd4jLong *hSscalarShapeInfo,
void *dScalar, Nd4jLong *dSscalarShapeInfo,
[WIP] More of CUDA (#95) * initial commit Signed-off-by: raver119 <raver119@gmail.com> * Implementation of hashcode cuda helper. Working edition. * Fixed parallel test input arangements. * Fixed tests for hashcode op. * Fixed shape calculation for image:crop_and_resize op and test. * NativeOps tests. Initial test suite. * Added tests for indexReduce methods. * Added test on execBroadcast with NDArray as dimensions. * Added test on execBroadcastBool with NDArray as dimensions. * Added tests on execPairwiseTransform and execPairwiseTransofrmBool. * Added tests for execReduce with scalar results. * Added reduce tests for non-empty dims array. * Added tests for reduce3. * Added tests for execScalar. * Added tests for execSummaryStats. * - provide cpu/cuda code for batch_to_space - testing it Signed-off-by: Yurii <yurii@skymind.io> * - remove old test for batch_to_space (had wrong format and numbers were not checked) Signed-off-by: Yurii <yurii@skymind.io> * Fixed complilation errors with test. * Added test for execTransformFloat. * Added test for execTransformSame. * Added test for execTransformBool. * Added test for execTransformStrict. * Added tests for execScalar/execScalarBool with TADs. * Added test for flatten. * - provide cpu/cuda code for space_to_Batch operaion Signed-off-by: Yurii <yurii@skymind.io> * Added test for concat. * comment unnecessary stuff in s_t_b Signed-off-by: Yurii <yurii@skymind.io> * Added test for specialConcat. * Added tests for memcpy/set routines. * Fixed pullRow cuda test. * Added pullRow test. * Added average test. * - correct typo in NDArray::applyPairwiseTransform(nd4j::pairwise::BoolOps op...) Signed-off-by: Yurii <yurii@skymind.io> * - debugging and fixing cuda tests in JavaInteropTests file Signed-off-by: Yurii <yurii@skymind.io> * - correct some tests Signed-off-by: Yurii <yurii@skymind.io> * Added test for shuffle. * Fixed ops declarations. * Restored omp and added shuffle test. * Added convertTypes test. * Added tests for execRandom. Eliminated usage of RandomBuffer with NativeOps. * Added sort tests. * Added tests for execCustomOp. * - further debuging and fixing tests terminated with crash Signed-off-by: Yurii <yurii@skymind.io> * Added tests for calculateOutputShapes. * Addded Benchmarks test. * Commented benchmark tests. * change assertion Signed-off-by: raver119 <raver119@gmail.com> * Added tests for apply_sgd op. Added cpu helper for that op. * Implement cuda helper for aplly_sgd op. Fixed tests for NativeOps. * Added test for assign broadcastable. * Added tests for assign_bp op. * Added tests for axpy op. * - assign/execScalar/execTransformAny signature change - minor test fix Signed-off-by: raver119 <raver119@gmail.com> * Fixed axpy op. * meh Signed-off-by: raver119 <raver119@gmail.com> * - fix tests for nativeOps::concat Signed-off-by: Yurii <yurii@skymind.io> * sequential transform/scalar Signed-off-by: raver119 <raver119@gmail.com> * allow nested parallelism Signed-off-by: raver119 <raver119@gmail.com> * assign_bp leak fix Signed-off-by: raver119 <raver119@gmail.com> * block setRNG fix Signed-off-by: raver119 <raver119@gmail.com> * enable parallelism by default Signed-off-by: raver119 <raver119@gmail.com> * enable nested parallelism by default Signed-off-by: raver119 <raver119@gmail.com> * Added cuda implementation for row_count helper. * Added implementation for tnse gains op helper. * - take into account possible situations when input arrays are empty in reduce_ cuda stuff Signed-off-by: Yurii <yurii@skymind.io> * Implemented tsne/edge_forces op cuda-based helper. Parallelized cpu-based helper for edge_forces. * Added kernel for tsne/symmetrized op heleper. * Implementation of tsne/symmetrized op cuda helper. Working edition. * Eliminated waste printfs. * Added test for broadcastgradientargs op. * host-only fallback for empty reduce float Signed-off-by: raver119 <raver119@gmail.com> * - some tests fixes Signed-off-by: Yurii <yurii@skymind.io> * - correct the rest of reduce_ stuff Signed-off-by: Yurii <yurii@skymind.io> * - further correction of reduce_ stuff Signed-off-by: Yurii <yurii@skymind.io> * Added test for Cbow op. Also added cuda implementation for cbow helpers. * - improve code of stack operation for scalar case Signed-off-by: Yurii <yurii@skymind.io> * - provide cuda kernel for gatherND operation Signed-off-by: Yurii <yurii@skymind.io> * Implementation of cbow helpers with cuda kernels. * minor tests tweaks Signed-off-by: raver119 <raver119@gmail.com> * minor tests tweaks Signed-off-by: raver119 <raver119@gmail.com> * - further correction of cuda stuff Signed-off-by: Yurii <yurii@skymind.io> * Implementatation of cbow op helper with cuda kernels. Working edition. * Skip random testing for cudablas case. * lstmBlockCell context fix Signed-off-by: raver119 <raver119@gmail.com> * Added tests for ELU and ELU_BP ops. * Added tests for eq_scalar, gt_scalar, gte_scalar and lte_scalar ops. * Added tests for neq_scalar. * Added test for noop. * - further work on clipbynorm_bp Signed-off-by: Yurii <yurii@skymind.io> * - get rid of concat op call, use instead direct concat helper call Signed-off-by: Yurii <yurii@skymind.io> * lstmBlockCell context fix Signed-off-by: raver119 <raver119@gmail.com> * Added tests for lrelu and lrelu_bp. * Added tests for selu and selu_bp. * Fixed lrelu derivative helpers. * - some corrections in lstm Signed-off-by: Yurii <yurii@skymind.io> * operator * result shape fix Signed-off-by: raver119 <raver119@gmail.com> * - correct typo in lstmCell Signed-off-by: Yurii <yurii@skymind.io> * few tests fixed Signed-off-by: raver119 <raver119@gmail.com> * CUDA inverse broadcast bool fix Signed-off-by: raver119 <raver119@gmail.com> * disable MMAP test for CUDA Signed-off-by: raver119 <raver119@gmail.com> * BooleanOp syncToDevice Signed-off-by: raver119 <raver119@gmail.com> * meh Signed-off-by: raver119 <raver119@gmail.com> * additional data types for im2col/col2im Signed-off-by: raver119 <raver119@gmail.com> * Added test for firas_sparse op. * one more RandomBuffer test excluded Signed-off-by: raver119 <raver119@gmail.com> * Added tests for flatten op. * Added test for Floor op. * bunch of tests fixed Signed-off-by: raver119 <raver119@gmail.com> * mmulDot tests fixed Signed-off-by: raver119 <raver119@gmail.com> * more tests fixed Signed-off-by: raver119 <raver119@gmail.com> * Implemented floordiv_bp op and tests. * Fixed scalar case with cuda implementation for bds. * - work on cuda kernel for clip_by_norm backprop op is completed Signed-off-by: Yurii <yurii@skymind.io> * Eliminate cbow crach. * more tests fixed Signed-off-by: raver119 <raver119@gmail.com> * more tests fixed Signed-off-by: raver119 <raver119@gmail.com> * Eliminated abortion with batched nlp test. * more tests fixed Signed-off-by: raver119 <raver119@gmail.com> * Fixed shared flag initializing. * disabled bunch of cpu workspaces tests Signed-off-by: raver119 <raver119@gmail.com> * scalar operators fix: missing registerSpecialUse call Signed-off-by: raver119 <raver119@gmail.com> * Fixed logdet for cuda and tests. * - correct clipBynorm_bp Signed-off-by: Yurii <yurii@skymind.io> * Fixed crop_and_resize shape datatype. * - correct some mmul tests Signed-off-by: Yurii <yurii@skymind.io>
2019-08-02 19:01:03 +02:00
void *extraParams, bool allowParallelism) {
auto xType = sd::ArrayOptions::dataType(hXShapeInfo);
auto yType = sd::ArrayOptions::dataType(hSscalarShapeInfo);
auto zType = sd::ArrayOptions::dataType(hZShapeInfo);
2019-06-06 14:21:15 +02:00
if (shape::isEmpty(hXShapeInfo) || shape::isEmpty(hSscalarShapeInfo))
return;
2019-06-06 14:21:15 +02:00
if (xType != yType)
throw sd::datatype_exception::build("NativeOpExecutioner::execScalarBool", xType, yType);
2019-06-06 14:21:15 +02:00
if (zType != sd::DataType::BOOL)
throw sd::datatype_exception::build("NativeOpExecutioner::execScalarBool", sd::DataType::BOOL, zType);
2019-06-06 14:21:15 +02:00
auto func = PRAGMA_THREADS_FOR {
BUILD_DOUBLE_SELECTOR(xType, zType, functions::scalar::ScalarBoolTransform, ::transform(opNum, hX, hXShapeInfo, hZ, hZShapeInfo, hScalar, extraParams, start, stop), LIBND4J_TYPES, BOOL_TYPES);
};
auto zLen = shape::length(hZShapeInfo);
samediff::Threads::parallel_for(func, 0, zLen, 1, !allowParallelism ? 1 : sd::math::nd4j_max<int>(1, sd::math::nd4j_min<int>(zLen / 1024, sd::Environment::getInstance()->maxMasterThreads())));
2019-06-06 14:21:15 +02:00
}
////////////////////////////////////////////////////////////////////////
void NativeOpExecutioner::execScalarBool(sd::LaunchContext *lc,
2019-06-06 14:21:15 +02:00
int opNum,
void *hX, Nd4jLong *hXShapeInfo,
void *dX, Nd4jLong *dXShapeInfo,
void *extraParams,
void *hZ, Nd4jLong *hZShapeInfo,
void *dZ, Nd4jLong *dZShapeInfo,
void *hScalars, Nd4jLong *hScalarShapeInfo,
void *dScalars, Nd4jLong *dScalarShapeInfo,
int *dimension, int dimensionLength,
Nd4jLong *tadShapeInfo, Nd4jLong *tadOffsets,
Nd4jLong *tadShapeInfoZ, Nd4jLong *tadOffsetsZ) {
auto xType = sd::ArrayOptions::dataType(hXShapeInfo);
auto yType = sd::ArrayOptions::dataType(hScalarShapeInfo);
auto zType = sd::ArrayOptions::dataType(hZShapeInfo);
2019-06-06 14:21:15 +02:00
if (shape::isEmpty(hXShapeInfo) || shape::isEmpty(hScalarShapeInfo))
return;
2019-06-06 14:21:15 +02:00
if (xType != yType)
throw sd::datatype_exception::build("NativeOpExecutioner::execScalarBool", xType, yType);
2019-06-06 14:21:15 +02:00
if (zType != sd::DataType::BOOL)
throw sd::datatype_exception::build("NativeOpExecutioner::execScalarBool", sd::DataType::BOOL, zType);
2019-06-06 14:21:15 +02:00
auto func = PRAGMA_THREADS_FOR {
BUILD_DOUBLE_SELECTOR(xType, zType, functions::scalar::ScalarBoolTransform, ::transform(opNum, hX, hXShapeInfo, extraParams, hZ, hZShapeInfo, hScalars, dimension, dimensionLength, tadShapeInfo, tadOffsets, tadShapeInfoZ, tadOffsetsZ, start, stop), LIBND4J_TYPES, BOOL_TYPES);
};
auto yLen = shape::length(hScalarShapeInfo);
samediff::Threads::parallel_tad(func, 0, yLen, 1, sd::math::nd4j_min<int>(yLen, sd::Environment::getInstance()->maxMasterThreads()));
2019-06-06 14:21:15 +02:00
}
////////////////////////////////////////////////////////////////////////
void NativeOpExecutioner::execScalarInt(sd::LaunchContext *lc,
int opNum,
void *hX, Nd4jLong *hXShapeInfo,
void *dX, Nd4jLong *dXShapeInfo,
void *hZ, Nd4jLong *hZShapeInfo,
void *dZ, Nd4jLong *dZShapeInfo,
void *hScalar, Nd4jLong *hSscalarShapeInfo,
void *dScalar, Nd4jLong *dSscalarShapeInfo,
void *extraParams, bool allowParallelism) {
auto xType = sd::ArrayOptions::dataType(hXShapeInfo);
auto yType = sd::ArrayOptions::dataType(hSscalarShapeInfo);
auto zType = sd::ArrayOptions::dataType(hZShapeInfo);
if (shape::isEmpty(hXShapeInfo) || shape::isEmpty(hSscalarShapeInfo))
return;
if (xType != yType || xType != zType)
throw sd::datatype_exception::build("NativeOpExecutioner::execScalarInt", xType, yType);
if (!sd::DataTypeUtils::isZ(zType))
throw sd::datatype_exception::build("NativeOpExecutioner::execScalarInt", sd::DataType::INT32, zType);
auto func = PRAGMA_THREADS_FOR {
BUILD_SINGLE_SELECTOR(xType, functions::scalar::ScalarIntTransform, ::transform(opNum, hX, hXShapeInfo, hZ, hZShapeInfo, hScalar, extraParams, start, stop), INTEGER_TYPES);
};
auto zLen = shape::length(hZShapeInfo);
samediff::Threads::parallel_for(func, 0, zLen, 1, !allowParallelism ? 1 : sd::math::nd4j_max<int>(1, sd::math::nd4j_min<int>(zLen / 1024, sd::Environment::getInstance()->maxMasterThreads())));
}
////////////////////////////////////////////////////////////////////////
void NativeOpExecutioner::execScalarInt(sd::LaunchContext *lc,
int opNum,
void *hX, Nd4jLong *hXShapeInfo,
void *dX, Nd4jLong *dXShapeInfo,
void *extraParams,
void *hZ, Nd4jLong *hZShapeInfo,
void *dZ, Nd4jLong *dZShapeInfo,
void *hScalars, Nd4jLong *hScalarShapeInfo,
void *dScalars, Nd4jLong *dScalarShapeInfo,
int *dimension, int dimensionLength,
Nd4jLong *tadShapeInfo, Nd4jLong *tadOffsets,
Nd4jLong *tadShapeInfoZ, Nd4jLong *tadOffsetsZ) {
auto xType = sd::ArrayOptions::dataType(hXShapeInfo);
auto yType = sd::ArrayOptions::dataType(hScalarShapeInfo);
auto zType = sd::ArrayOptions::dataType(hZShapeInfo);
if (shape::isEmpty(hXShapeInfo) || shape::isEmpty(hScalarShapeInfo))
return;
if (xType != yType || xType != zType)
throw sd::datatype_exception::build("NativeOpExecutioner::execScalarInt", xType, yType);
if (!sd::DataTypeUtils::isZ(zType))
throw sd::datatype_exception::build("NativeOpExecutioner::execScalarInt requires integer data type", zType);
auto func = PRAGMA_THREADS_FOR {
BUILD_SINGLE_SELECTOR(xType, functions::scalar::ScalarIntTransform, ::transform(opNum, hX, hXShapeInfo, extraParams, hZ, hZShapeInfo, hScalars, dimension, dimensionLength, tadShapeInfo, tadOffsets, tadShapeInfoZ, tadOffsetsZ, start, stop), INTEGER_TYPES);
};
auto yLen = shape::length(hScalarShapeInfo);
samediff::Threads::parallel_tad(func, 0, yLen, 1, sd::math::nd4j_min<int>(yLen, sd::Environment::getInstance()->maxMasterThreads()));
}
2019-06-06 14:21:15 +02:00
////////////////////////////////////////////////////////////////////////
/**
*
* @param opNum
* @param hX
* @param hXShapeInfo
* @param extraParams
* @param hZ
* @param hZShapeInfo
*/
void NativeOpExecutioner::execSummaryStats(sd::LaunchContext *lc,
2019-06-06 14:21:15 +02:00
int opNum,
void *hX, Nd4jLong *hXShapeInfo,
void *dX, Nd4jLong *dXShapeInfo,
void *extraParams,
void *hZ, Nd4jLong *hZShapeInfo,
void *dZ, Nd4jLong *dZShapeInfo,
bool biasCorrected) {
2019-06-06 14:21:15 +02:00
auto xType = sd::ArrayOptions::dataType(hXShapeInfo);
auto zType = sd::ArrayOptions::dataType(hZShapeInfo);
2019-06-06 14:21:15 +02:00
BUILD_DOUBLE_SELECTOR(xType, zType, functions::summarystats::SummaryStatsReduce, ::exec(opNum, biasCorrected, hX, hXShapeInfo, extraParams, hZ, hZShapeInfo, nullptr, 1), LIBND4J_TYPES, FLOAT_TYPES);
}
////////////////////////////////////////////////////////////////////////
/**
*
* @param opNum
* @param hX
* @param hXShapeInfo
* @param extraParams
* @param hZ
* @param hZShapeInfo
*/
void NativeOpExecutioner::execSummaryStatsScalar(sd::LaunchContext *lc,
2019-06-06 14:21:15 +02:00
int opNum,
void *hX, Nd4jLong *hXShapeInfo,
void *dX, Nd4jLong *dXShapeInfo,
void *extraParams,
void *hZ, Nd4jLong *hZShapeInfo,
void *dZ, Nd4jLong *dZShapeInfo,
bool biasCorrected) {
2019-06-06 14:21:15 +02:00
auto xType = sd::ArrayOptions::dataType(hXShapeInfo);
auto zType = sd::ArrayOptions::dataType(hZShapeInfo);
2019-06-06 14:21:15 +02:00
BUILD_DOUBLE_SELECTOR(xType, zType, functions::summarystats::SummaryStatsReduce, ::execScalar(opNum, biasCorrected, hX, hXShapeInfo, extraParams, hZ, hZShapeInfo), LIBND4J_TYPES, FLOAT_TYPES);
}
////////////////////////////////////////////////////////////////////////
/**
*
* @param opNum
* @param hX
* @param hXShapeInfo
* @param extraParams
* @param hZ
* @param hZShapeInfo
* @param dimension
* @param dimensionLength
*/
void NativeOpExecutioner::execSummaryStats(sd::LaunchContext *lc,
2019-06-06 14:21:15 +02:00
int opNum,
void *hX, Nd4jLong *hXShapeInfo,
void *dX, Nd4jLong *dXShapeInfo,
void *extraParams,
void *hZ, Nd4jLong *hZShapeInfo,
void *dZ, Nd4jLong *dZShapeInfo,
int *dimension, int dimensionLength,
Nd4jLong *tadShapeInfo, Nd4jLong *tadOffsets,
bool biasCorrected) {
auto xType = sd::ArrayOptions::dataType(hXShapeInfo);
auto zType = sd::ArrayOptions::dataType(hZShapeInfo);
2019-06-06 14:21:15 +02:00
BUILD_DOUBLE_SELECTOR(xType, zType, functions::summarystats::SummaryStatsReduce, ::exec(opNum, biasCorrected, hX, hXShapeInfo, extraParams, hZ, hZShapeInfo, dimension, dimensionLength), LIBND4J_TYPES, FLOAT_TYPES);
}
////////////////////////////////////////////////////////////////////////
/**
*
* @param opNum
* @param hX
* @param xStride
* @param hZ
* @param resultStride
* @param extraParams
* @param n
*/
void NativeOpExecutioner::execTransformFloat(sd::LaunchContext *lc,
2019-06-06 14:21:15 +02:00
int opNum,
void *hX, Nd4jLong *hXShapeInfo,
void *dX, Nd4jLong *dXShapeInfo,
void *hZ, Nd4jLong *hZShapeInfo,
void *dZ, Nd4jLong *dZShapeInfo,
void *extraParams,
Nd4jLong *tadShapeInfo, Nd4jLong *tadOffsets) {
auto xType = sd::ArrayOptions::dataType(hXShapeInfo);
auto zType = sd::ArrayOptions::dataType(hZShapeInfo);
2019-06-06 14:21:15 +02:00
if (shape::isEmpty(hXShapeInfo))
return;
auto func = PRAGMA_THREADS_DO {
BUILD_DOUBLE_SELECTOR(xType, zType, functions::transform::TransformFloat, ::exec(opNum, hX, hXShapeInfo, hZ, hZShapeInfo, extraParams, thread_id, numThreads), LIBND4J_TYPES, FLOAT_TYPES);
};
samediff::Threads::parallel_do(func, sd::math::nd4j_max<int>(1, sd::math::nd4j_min<int>(shape::length(hZShapeInfo) / 1024, sd::Environment::getInstance()->maxMasterThreads())));
2019-06-06 14:21:15 +02:00
}
////////////////////////////////////////////////////////////////////////
void NativeOpExecutioner::execTransformBool(sd::LaunchContext *lc,
2019-06-06 14:21:15 +02:00
int opNum,
void *hX, Nd4jLong *hXShapeInfo,
void *dX, Nd4jLong *dXShapeInfo,
void *hZ, Nd4jLong *hZShapeInfo,
void *dZ, Nd4jLong *dZShapeInfo,
void *extraParams,
Nd4jLong *tadShapeInfo, Nd4jLong *tadOffsets) {
auto xType = sd::ArrayOptions::dataType(hXShapeInfo);
auto zType = sd::ArrayOptions::dataType(hZShapeInfo);
2019-06-06 14:21:15 +02:00
if (shape::isEmpty(hXShapeInfo))
return;
auto func = PRAGMA_THREADS_DO {
BUILD_DOUBLE_SELECTOR(xType, zType, functions::transform::TransformBool, ::exec(opNum, hX, hXShapeInfo, hZ, hZShapeInfo, extraParams, thread_id, numThreads), LIBND4J_TYPES, BOOL_TYPES);
};
samediff::Threads::parallel_do(func, sd::math::nd4j_max<int>(1, sd::math::nd4j_min<int>(shape::length(hZShapeInfo) / 1024, sd::Environment::getInstance()->maxMasterThreads())));
2019-06-06 14:21:15 +02:00
}
////////////////////////////////////////////////////////////////////////
void NativeOpExecutioner::execTransformAny(sd::LaunchContext *lc,
2019-06-06 14:21:15 +02:00
int opNum,
void *hX, Nd4jLong *hXShapeInfo,
void *dX, Nd4jLong *dXShapeInfo,
void *hZ, Nd4jLong *hZShapeInfo,
void *dZ, Nd4jLong *dZShapeInfo,
void *extraParams,
[WIP] More of CUDA (#95) * initial commit Signed-off-by: raver119 <raver119@gmail.com> * Implementation of hashcode cuda helper. Working edition. * Fixed parallel test input arangements. * Fixed tests for hashcode op. * Fixed shape calculation for image:crop_and_resize op and test. * NativeOps tests. Initial test suite. * Added tests for indexReduce methods. * Added test on execBroadcast with NDArray as dimensions. * Added test on execBroadcastBool with NDArray as dimensions. * Added tests on execPairwiseTransform and execPairwiseTransofrmBool. * Added tests for execReduce with scalar results. * Added reduce tests for non-empty dims array. * Added tests for reduce3. * Added tests for execScalar. * Added tests for execSummaryStats. * - provide cpu/cuda code for batch_to_space - testing it Signed-off-by: Yurii <yurii@skymind.io> * - remove old test for batch_to_space (had wrong format and numbers were not checked) Signed-off-by: Yurii <yurii@skymind.io> * Fixed complilation errors with test. * Added test for execTransformFloat. * Added test for execTransformSame. * Added test for execTransformBool. * Added test for execTransformStrict. * Added tests for execScalar/execScalarBool with TADs. * Added test for flatten. * - provide cpu/cuda code for space_to_Batch operaion Signed-off-by: Yurii <yurii@skymind.io> * Added test for concat. * comment unnecessary stuff in s_t_b Signed-off-by: Yurii <yurii@skymind.io> * Added test for specialConcat. * Added tests for memcpy/set routines. * Fixed pullRow cuda test. * Added pullRow test. * Added average test. * - correct typo in NDArray::applyPairwiseTransform(nd4j::pairwise::BoolOps op...) Signed-off-by: Yurii <yurii@skymind.io> * - debugging and fixing cuda tests in JavaInteropTests file Signed-off-by: Yurii <yurii@skymind.io> * - correct some tests Signed-off-by: Yurii <yurii@skymind.io> * Added test for shuffle. * Fixed ops declarations. * Restored omp and added shuffle test. * Added convertTypes test. * Added tests for execRandom. Eliminated usage of RandomBuffer with NativeOps. * Added sort tests. * Added tests for execCustomOp. * - further debuging and fixing tests terminated with crash Signed-off-by: Yurii <yurii@skymind.io> * Added tests for calculateOutputShapes. * Addded Benchmarks test. * Commented benchmark tests. * change assertion Signed-off-by: raver119 <raver119@gmail.com> * Added tests for apply_sgd op. Added cpu helper for that op. * Implement cuda helper for aplly_sgd op. Fixed tests for NativeOps. * Added test for assign broadcastable. * Added tests for assign_bp op. * Added tests for axpy op. * - assign/execScalar/execTransformAny signature change - minor test fix Signed-off-by: raver119 <raver119@gmail.com> * Fixed axpy op. * meh Signed-off-by: raver119 <raver119@gmail.com> * - fix tests for nativeOps::concat Signed-off-by: Yurii <yurii@skymind.io> * sequential transform/scalar Signed-off-by: raver119 <raver119@gmail.com> * allow nested parallelism Signed-off-by: raver119 <raver119@gmail.com> * assign_bp leak fix Signed-off-by: raver119 <raver119@gmail.com> * block setRNG fix Signed-off-by: raver119 <raver119@gmail.com> * enable parallelism by default Signed-off-by: raver119 <raver119@gmail.com> * enable nested parallelism by default Signed-off-by: raver119 <raver119@gmail.com> * Added cuda implementation for row_count helper. * Added implementation for tnse gains op helper. * - take into account possible situations when input arrays are empty in reduce_ cuda stuff Signed-off-by: Yurii <yurii@skymind.io> * Implemented tsne/edge_forces op cuda-based helper. Parallelized cpu-based helper for edge_forces. * Added kernel for tsne/symmetrized op heleper. * Implementation of tsne/symmetrized op cuda helper. Working edition. * Eliminated waste printfs. * Added test for broadcastgradientargs op. * host-only fallback for empty reduce float Signed-off-by: raver119 <raver119@gmail.com> * - some tests fixes Signed-off-by: Yurii <yurii@skymind.io> * - correct the rest of reduce_ stuff Signed-off-by: Yurii <yurii@skymind.io> * - further correction of reduce_ stuff Signed-off-by: Yurii <yurii@skymind.io> * Added test for Cbow op. Also added cuda implementation for cbow helpers. * - improve code of stack operation for scalar case Signed-off-by: Yurii <yurii@skymind.io> * - provide cuda kernel for gatherND operation Signed-off-by: Yurii <yurii@skymind.io> * Implementation of cbow helpers with cuda kernels. * minor tests tweaks Signed-off-by: raver119 <raver119@gmail.com> * minor tests tweaks Signed-off-by: raver119 <raver119@gmail.com> * - further correction of cuda stuff Signed-off-by: Yurii <yurii@skymind.io> * Implementatation of cbow op helper with cuda kernels. Working edition. * Skip random testing for cudablas case. * lstmBlockCell context fix Signed-off-by: raver119 <raver119@gmail.com> * Added tests for ELU and ELU_BP ops. * Added tests for eq_scalar, gt_scalar, gte_scalar and lte_scalar ops. * Added tests for neq_scalar. * Added test for noop. * - further work on clipbynorm_bp Signed-off-by: Yurii <yurii@skymind.io> * - get rid of concat op call, use instead direct concat helper call Signed-off-by: Yurii <yurii@skymind.io> * lstmBlockCell context fix Signed-off-by: raver119 <raver119@gmail.com> * Added tests for lrelu and lrelu_bp. * Added tests for selu and selu_bp. * Fixed lrelu derivative helpers. * - some corrections in lstm Signed-off-by: Yurii <yurii@skymind.io> * operator * result shape fix Signed-off-by: raver119 <raver119@gmail.com> * - correct typo in lstmCell Signed-off-by: Yurii <yurii@skymind.io> * few tests fixed Signed-off-by: raver119 <raver119@gmail.com> * CUDA inverse broadcast bool fix Signed-off-by: raver119 <raver119@gmail.com> * disable MMAP test for CUDA Signed-off-by: raver119 <raver119@gmail.com> * BooleanOp syncToDevice Signed-off-by: raver119 <raver119@gmail.com> * meh Signed-off-by: raver119 <raver119@gmail.com> * additional data types for im2col/col2im Signed-off-by: raver119 <raver119@gmail.com> * Added test for firas_sparse op. * one more RandomBuffer test excluded Signed-off-by: raver119 <raver119@gmail.com> * Added tests for flatten op. * Added test for Floor op. * bunch of tests fixed Signed-off-by: raver119 <raver119@gmail.com> * mmulDot tests fixed Signed-off-by: raver119 <raver119@gmail.com> * more tests fixed Signed-off-by: raver119 <raver119@gmail.com> * Implemented floordiv_bp op and tests. * Fixed scalar case with cuda implementation for bds. * - work on cuda kernel for clip_by_norm backprop op is completed Signed-off-by: Yurii <yurii@skymind.io> * Eliminate cbow crach. * more tests fixed Signed-off-by: raver119 <raver119@gmail.com> * more tests fixed Signed-off-by: raver119 <raver119@gmail.com> * Eliminated abortion with batched nlp test. * more tests fixed Signed-off-by: raver119 <raver119@gmail.com> * Fixed shared flag initializing. * disabled bunch of cpu workspaces tests Signed-off-by: raver119 <raver119@gmail.com> * scalar operators fix: missing registerSpecialUse call Signed-off-by: raver119 <raver119@gmail.com> * Fixed logdet for cuda and tests. * - correct clipBynorm_bp Signed-off-by: Yurii <yurii@skymind.io> * Fixed crop_and_resize shape datatype. * - correct some mmul tests Signed-off-by: Yurii <yurii@skymind.io>
2019-08-02 19:01:03 +02:00
Nd4jLong *tadShapeInfo, Nd4jLong *tadOffsets, bool allowParallelism) {
auto xType = sd::ArrayOptions::dataType(hXShapeInfo);
auto zType = sd::ArrayOptions::dataType(hZShapeInfo);
2019-06-06 14:21:15 +02:00
if (shape::isEmpty(hXShapeInfo))
return;
profiling of stack and unstack ops (#261) * - profiling of stack and unstack ops Signed-off-by: Yurii <iuriish@yahoo.com> * - fix bug in cpu concat op Signed-off-by: Yurii <iuriish@yahoo.com> * - correction of cuda stack and unstack Signed-off-by: Yurii <iuriish@yahoo.com> * - change shape.h method which operates with unity dimensions strides Signed-off-by: Yurii <iuriish@yahoo.com> * - rearrange stack tests Signed-off-by: Yurii <iuriish@yahoo.com> * - correct evaluation of smallest stride for moving through contiguous axis Signed-off-by: Yurii <iuriish@yahoo.com> * - forgot to update signature of function strideOverContigAxis in cuda concat and split ops Signed-off-by: Yurii <iuriish@yahoo.com> * - remove ShapeUtils::shapeAsString method applied before input arrays validations Signed-off-by: Yurii <iuriish@yahoo.com> * - further removing of ShapeUtils::shapeAsString Signed-off-by: Yurii <iuriish@yahoo.com> * - take sub-array shapeIndo/offset calculation out of NDArray class - add possibility of contiguous memory copy in execTransformAny op if opNum == assign Signed-off-by: Yurii <iuriish@yahoo.com> * - correct test_empty_scatter_2 in EmptyTests.cpp Signed-off-by: Yurii <iuriish@yahoo.com> * - profiling of slice op Signed-off-by: Yurii <iuriish@yahoo.com> * - get rid of contiguous memcpy for some cases in concat and split ops Signed-off-by: Yurii <iuriish@yahoo.com> * - forgot to declare oid nd4j::SpecialMethods<T>::splitCpuGeneric Signed-off-by: Yurii <iuriish@yahoo.com> * - correct typo in calculation of threads in cuda split op Signed-off-by: Yurii <iuriish@yahoo.com> * - forgot to correct another set of threads variables in split cuda ops Signed-off-by: Yurii <iuriish@yahoo.com> * - further conflicts resolving Signed-off-by: Yurii <iuriish@yahoo.com> Co-authored-by: raver119 <raver119@gmail.com>
2020-03-03 05:32:37 +01:00
if (opNum == sd::transform::Assign && shape::order(hXShapeInfo) == shape::order(hZShapeInfo) && shape::order(hXShapeInfo) == 'c' && xType == zType && shape::elementWiseStride(hXShapeInfo) == 1 && shape::elementWiseStride(hZShapeInfo) == 1) {
profiling of stack and unstack ops (#261) * - profiling of stack and unstack ops Signed-off-by: Yurii <iuriish@yahoo.com> * - fix bug in cpu concat op Signed-off-by: Yurii <iuriish@yahoo.com> * - correction of cuda stack and unstack Signed-off-by: Yurii <iuriish@yahoo.com> * - change shape.h method which operates with unity dimensions strides Signed-off-by: Yurii <iuriish@yahoo.com> * - rearrange stack tests Signed-off-by: Yurii <iuriish@yahoo.com> * - correct evaluation of smallest stride for moving through contiguous axis Signed-off-by: Yurii <iuriish@yahoo.com> * - forgot to update signature of function strideOverContigAxis in cuda concat and split ops Signed-off-by: Yurii <iuriish@yahoo.com> * - remove ShapeUtils::shapeAsString method applied before input arrays validations Signed-off-by: Yurii <iuriish@yahoo.com> * - further removing of ShapeUtils::shapeAsString Signed-off-by: Yurii <iuriish@yahoo.com> * - take sub-array shapeIndo/offset calculation out of NDArray class - add possibility of contiguous memory copy in execTransformAny op if opNum == assign Signed-off-by: Yurii <iuriish@yahoo.com> * - correct test_empty_scatter_2 in EmptyTests.cpp Signed-off-by: Yurii <iuriish@yahoo.com> * - profiling of slice op Signed-off-by: Yurii <iuriish@yahoo.com> * - get rid of contiguous memcpy for some cases in concat and split ops Signed-off-by: Yurii <iuriish@yahoo.com> * - forgot to declare oid nd4j::SpecialMethods<T>::splitCpuGeneric Signed-off-by: Yurii <iuriish@yahoo.com> * - correct typo in calculation of threads in cuda split op Signed-off-by: Yurii <iuriish@yahoo.com> * - forgot to correct another set of threads variables in split cuda ops Signed-off-by: Yurii <iuriish@yahoo.com> * - further conflicts resolving Signed-off-by: Yurii <iuriish@yahoo.com> Co-authored-by: raver119 <raver119@gmail.com>
2020-03-03 05:32:37 +01:00
memcpy(hZ, hX, shape::length(hXShapeInfo) * sd::DataTypeUtils::sizeOfElement(xType));
}
else {
auto func = PRAGMA_THREADS_DO {
BUILD_DOUBLE_SELECTOR(xType, zType, functions::transform::TransformAny, ::exec(opNum, hX, hXShapeInfo, hZ, hZShapeInfo, extraParams, thread_id, numThreads), LIBND4J_TYPES, LIBND4J_TYPES);
};
samediff::Threads::parallel_do(func, sd::math::nd4j_max<int>(1, sd::math::nd4j_min<int>(shape::length(hZShapeInfo) / 1024, sd::Environment::getInstance()->maxMasterThreads())));
profiling of stack and unstack ops (#261) * - profiling of stack and unstack ops Signed-off-by: Yurii <iuriish@yahoo.com> * - fix bug in cpu concat op Signed-off-by: Yurii <iuriish@yahoo.com> * - correction of cuda stack and unstack Signed-off-by: Yurii <iuriish@yahoo.com> * - change shape.h method which operates with unity dimensions strides Signed-off-by: Yurii <iuriish@yahoo.com> * - rearrange stack tests Signed-off-by: Yurii <iuriish@yahoo.com> * - correct evaluation of smallest stride for moving through contiguous axis Signed-off-by: Yurii <iuriish@yahoo.com> * - forgot to update signature of function strideOverContigAxis in cuda concat and split ops Signed-off-by: Yurii <iuriish@yahoo.com> * - remove ShapeUtils::shapeAsString method applied before input arrays validations Signed-off-by: Yurii <iuriish@yahoo.com> * - further removing of ShapeUtils::shapeAsString Signed-off-by: Yurii <iuriish@yahoo.com> * - take sub-array shapeIndo/offset calculation out of NDArray class - add possibility of contiguous memory copy in execTransformAny op if opNum == assign Signed-off-by: Yurii <iuriish@yahoo.com> * - correct test_empty_scatter_2 in EmptyTests.cpp Signed-off-by: Yurii <iuriish@yahoo.com> * - profiling of slice op Signed-off-by: Yurii <iuriish@yahoo.com> * - get rid of contiguous memcpy for some cases in concat and split ops Signed-off-by: Yurii <iuriish@yahoo.com> * - forgot to declare oid nd4j::SpecialMethods<T>::splitCpuGeneric Signed-off-by: Yurii <iuriish@yahoo.com> * - correct typo in calculation of threads in cuda split op Signed-off-by: Yurii <iuriish@yahoo.com> * - forgot to correct another set of threads variables in split cuda ops Signed-off-by: Yurii <iuriish@yahoo.com> * - further conflicts resolving Signed-off-by: Yurii <iuriish@yahoo.com> Co-authored-by: raver119 <raver119@gmail.com>
2020-03-03 05:32:37 +01:00
}
2019-06-06 14:21:15 +02:00
}
////////////////////////////////////////////////////////////////////////
void NativeOpExecutioner::execTransformSame(sd::LaunchContext *lc,
2019-06-06 14:21:15 +02:00
int opNum,
void *hX, Nd4jLong *hXShapeInfo,
void *dX, Nd4jLong *dXShapeInfo,
void *hZ, Nd4jLong *hZShapeInfo,
void *dZ, Nd4jLong *dZShapeInfo,
void *extraParams,
Nd4jLong *tadShapeInfo, Nd4jLong *tadOffsets) {
auto xType = sd::ArrayOptions::dataType(hXShapeInfo);
auto zType = sd::ArrayOptions::dataType(hZShapeInfo);
2019-06-06 14:21:15 +02:00
if (shape::isEmpty(hXShapeInfo))
return;
auto func = PRAGMA_THREADS_DO {
BUILD_SINGLE_SELECTOR(xType, functions::transform::TransformSame, ::exec(opNum, hX, hXShapeInfo, hZ, hZShapeInfo, extraParams, thread_id, numThreads), LIBND4J_TYPES);
};
samediff::Threads::parallel_do(func, sd::math::nd4j_max<int>(1, sd::math::nd4j_min<int>(shape::length(hZShapeInfo) / 1024, sd::Environment::getInstance()->maxMasterThreads())));
2019-06-06 14:21:15 +02:00
}
////////////////////////////////////////////////////////////////////////
void NativeOpExecutioner::execTransformStrict(sd::LaunchContext *lc,
2019-06-06 14:21:15 +02:00
int opNum,
void *hX, Nd4jLong *hXShapeInfo,
void *dX, Nd4jLong *dXShapeInfo,
void *hZ, Nd4jLong *hZShapeInfo,
void *dZ, Nd4jLong *dZShapeInfo,
void *extraParams,
Nd4jLong *tadShapeInfo, Nd4jLong *tadOffsets) {
auto xType = sd::ArrayOptions::dataType(hXShapeInfo);
auto zType = sd::ArrayOptions::dataType(hZShapeInfo);
2019-06-06 14:21:15 +02:00
if (shape::isEmpty(hXShapeInfo))
return;
auto func = PRAGMA_THREADS_DO {
BUILD_SINGLE_SELECTOR(xType, functions::transform::TransformStrict, ::exec(opNum, hX, hXShapeInfo, hZ, hZShapeInfo, extraParams, thread_id, numThreads), FLOAT_TYPES);
};
samediff::Threads::parallel_do(func, sd::math::nd4j_max<int>(1, sd::math::nd4j_min<int>(shape::length(hZShapeInfo) / 1024, sd::Environment::getInstance()->maxMasterThreads())));
2019-06-06 14:21:15 +02:00
}
////////////////////////////////////////////////////////////////////////
void NativeOpExecutioner::execRandom(sd::LaunchContext *lc,
2019-06-06 14:21:15 +02:00
int opNum,
Nd4jPointer state,
void *hZ, Nd4jLong *hZShapeInfo,
void *dZ, Nd4jLong *dZShapeInfo,
void *extraArguments) {
2019-06-06 14:21:15 +02:00
auto zType = sd::ArrayOptions::dataType(hZShapeInfo);
2019-06-06 14:21:15 +02:00
BUILD_SINGLE_SELECTOR(zType, functions::random::RandomFunction, ::execTransform(opNum, state, hZ, hZShapeInfo, extraArguments), FLOAT_TYPES);
auto rng = reinterpret_cast<sd::graph::RandomGenerator*>(state);
rng->rewindH(shape::length(hZShapeInfo));
2019-06-06 14:21:15 +02:00
}
////////////////////////////////////////////////////////////////////////
void NativeOpExecutioner::execRandom(sd::LaunchContext *lc,
2019-06-06 14:21:15 +02:00
int opNum,
Nd4jPointer state,
void *hX, Nd4jLong *hXShapeInfo,
void *dX, Nd4jLong *dXShapeInfo,
void *hZ, Nd4jLong *hZShapeInfo,
void *dZ, Nd4jLong *dZShapeInfo,
void *extraArguments) {
2019-06-06 14:21:15 +02:00
auto zType = sd::ArrayOptions::dataType(hZShapeInfo);
2019-06-06 14:21:15 +02:00
BUILD_SINGLE_SELECTOR(zType, functions::random::RandomFunction, ::execTransform(opNum, state, hX, hXShapeInfo, hZ, hZShapeInfo, extraArguments), FLOAT_TYPES);
auto rng = reinterpret_cast<sd::graph::RandomGenerator*>(state);
rng->rewindH(shape::length(hZShapeInfo));
2019-06-06 14:21:15 +02:00
}
////////////////////////////////////////////////////////////////////////
void NativeOpExecutioner::execRandom(sd::LaunchContext *lc,
2019-06-06 14:21:15 +02:00
int opNum,
Nd4jPointer state,
void *hX, Nd4jLong *hXShapeInfo,
void *dX, Nd4jLong *dXShapeInfo,
void *hY, Nd4jLong *hYShapeInfo,
void *dY, Nd4jLong *dYShapeInfo,
void *hZ, Nd4jLong *hZShapeInfo,
void *dZ, Nd4jLong *dZShapeInfo,
void *extraArguments) {
2019-06-06 14:21:15 +02:00
auto xType = sd::ArrayOptions::dataType(hZShapeInfo);
2019-06-06 14:21:15 +02:00
BUILD_SINGLE_SELECTOR(xType, functions::random::RandomFunction, ::execTransform(opNum, state, hX, hXShapeInfo, hY, hYShapeInfo, hZ, hZShapeInfo, extraArguments), FLOAT_TYPES);
auto rng = reinterpret_cast<sd::graph::RandomGenerator*>(state);
rng->rewindH(shape::length(hZShapeInfo));
2019-06-06 14:21:15 +02:00
}