mirror/eigen
2
0
Fork 0
mirror of https://gitlab.com/libeigen/eigen.git synced 2024-12-21 07:19:46 +08:00
Code Issues Packages Projects Releases Wiki Activity
bc128f9f3b
eigen/unsupported/test/cxx11_tensor_reduction_sycl.cpp

148 lines
6.2 KiB
C++
Raw Normal View History

Removed the sycl include from Eigen/Core and moved it to Unsupported/Eigen/CXX11/Tensor; added TensorReduction for sycl (full reduction and partial reduction); added TensorReduction test case for sycl (full reduction and partial reduction); fixed the tile size on TensorSyclRun.h based on the device max work group size;
2016-11-05 02:18:19 +08:00
// This file is part of Eigen, a lightweight C++ template library
// for linear algebra.
//
// Copyright (C) 2015
// Mehdi Goli Codeplay Software Ltd.
// Ralph Potter Codeplay Software Ltd.
// Luke Iwanski Codeplay Software Ltd.
// Contact: <eigen@codeplay.com>
//
// This Source Code Form is subject to the terms of the Mozilla
// Public License v. 2.0. If a copy of the MPL was not distributed
// with this file, You can obtain one at http://mozilla.org/MPL/2.0/.
#define EIGEN_TEST_NO_LONGDOUBLE
#define EIGEN_TEST_NO_COMPLEX
#define EIGEN_TEST_FUNC cxx11_tensor_reduction_sycl
Reducing warnings in Sycl backend.
2017-02-01 23:29:53 +08:00
#define EIGEN_DEFAULT_DENSE_INDEX_TYPE int64_t
Removed the sycl include from Eigen/Core and moved it to Unsupported/Eigen/CXX11/Tensor; added TensorReduction for sycl (full reduction and partial reduction); added TensorReduction test case for sycl (full reduction and partial reduction); fixed the tile size on TensorSyclRun.h based on the device max work group size;
2016-11-05 02:18:19 +08:00
#define EIGEN_USE_SYCL
#include "main.h"
#include <unsupported/Eigen/CXX11/Tensor>
Reducing warnings in Sycl backend.
2017-02-01 23:29:53 +08:00
template <typename DataType, int DataLayout, typename IndexType>
Converting all sycl buffers to uninitialised device only buffers; adding memcpyHostToDevice and memcpyDeviceToHost on syclDevice; modifying all examples to obey the new rules; moving sycl queue creating to the device based on Benoit suggestion; removing the sycl specefic condition for returning m_result in TensorReduction.h according to Benoit suggestion.
2016-11-09 01:08:02 +08:00
static void test_full_reductions_sycl(const Eigen::SyclDevice& sycl_device) {
Removed the sycl include from Eigen/Core and moved it to Unsupported/Eigen/CXX11/Tensor; added TensorReduction for sycl (full reduction and partial reduction); added TensorReduction test case for sycl (full reduction and partial reduction); fixed the tile size on TensorSyclRun.h based on the device max work group size;
2016-11-05 02:18:19 +08:00
Reducing warnings in Sycl backend.
2017-02-01 23:29:53 +08:00
const IndexType num_rows = 452;
const IndexType num_cols = 765;
array<IndexType, 2> tensorRange = {{num_rows, num_cols}};
Removed the sycl include from Eigen/Core and moved it to Unsupported/Eigen/CXX11/Tensor; added TensorReduction for sycl (full reduction and partial reduction); added TensorReduction test case for sycl (full reduction and partial reduction); fixed the tile size on TensorSyclRun.h based on the device max work group size;
2016-11-05 02:18:19 +08:00
Reducing warnings in Sycl backend.
2017-02-01 23:29:53 +08:00
Tensor<DataType, 2, DataLayout, IndexType> in(tensorRange);
Tensor<DataType, 0, DataLayout, IndexType> full_redux;
Tensor<DataType, 0, DataLayout, IndexType> full_redux_gpu;
Converting all sycl buffers to uninitialised device only buffers; adding memcpyHostToDevice and memcpyDeviceToHost on syclDevice; modifying all examples to obey the new rules; moving sycl queue creating to the device based on Benoit suggestion; removing the sycl specefic condition for returning m_result in TensorReduction.h according to Benoit suggestion.
2016-11-09 01:08:02 +08:00
Removed the sycl include from Eigen/Core and moved it to Unsupported/Eigen/CXX11/Tensor; added TensorReduction for sycl (full reduction and partial reduction); added TensorReduction test case for sycl (full reduction and partial reduction); fixed the tile size on TensorSyclRun.h based on the device max work group size;
2016-11-05 02:18:19 +08:00
in.setRandom();
full_redux = in.sum();
Modifying TensorDeviceSycl.h to always create buffer of type uint8_t and convert them to the actual type at the execution on the device; adding the queue interface class to separate the lifespan of sycl queue and buffers,created for that queue, from Eigen::SyclDevice; modifying sycl tests to support the evaluation of the results for both row major and column major data layout on all different devices that are supported by Sycl{CPU; GPU; and Host}.
2016-11-19 00:20:42 +08:00
DataType* gpu_in_data = static_cast<DataType*>(sycl_device.allocate(in.dimensions().TotalSize()*sizeof(DataType)));
DataType* gpu_out_data =(DataType*)sycl_device.allocate(sizeof(DataType));
Removed the sycl include from Eigen/Core and moved it to Unsupported/Eigen/CXX11/Tensor; added TensorReduction for sycl (full reduction and partial reduction); added TensorReduction test case for sycl (full reduction and partial reduction); fixed the tile size on TensorSyclRun.h based on the device max work group size;
2016-11-05 02:18:19 +08:00
Reducing warnings in Sycl backend.
2017-02-01 23:29:53 +08:00
TensorMap<Tensor<DataType, 2, DataLayout, IndexType> > in_gpu(gpu_in_data, tensorRange);
TensorMap<Tensor<DataType, 0, DataLayout, IndexType> > out_gpu(gpu_out_data);
Removed the sycl include from Eigen/Core and moved it to Unsupported/Eigen/CXX11/Tensor; added TensorReduction for sycl (full reduction and partial reduction); added TensorReduction test case for sycl (full reduction and partial reduction); fixed the tile size on TensorSyclRun.h based on the device max work group size;
2016-11-05 02:18:19 +08:00
Modifying TensorDeviceSycl.h to always create buffer of type uint8_t and convert them to the actual type at the execution on the device; adding the queue interface class to separate the lifespan of sycl queue and buffers,created for that queue, from Eigen::SyclDevice; modifying sycl tests to support the evaluation of the results for both row major and column major data layout on all different devices that are supported by Sycl{CPU; GPU; and Host}.
2016-11-19 00:20:42 +08:00
sycl_device.memcpyHostToDevice(gpu_in_data, in.data(),(in.dimensions().TotalSize())*sizeof(DataType));
Converting all sycl buffers to uninitialised device only buffers; adding memcpyHostToDevice and memcpyDeviceToHost on syclDevice; modifying all examples to obey the new rules; moving sycl queue creating to the device based on Benoit suggestion; removing the sycl specefic condition for returning m_result in TensorReduction.h according to Benoit suggestion.
2016-11-09 01:08:02 +08:00
out_gpu.device(sycl_device) = in_gpu.sum();
Modifying TensorDeviceSycl.h to always create buffer of type uint8_t and convert them to the actual type at the execution on the device; adding the queue interface class to separate the lifespan of sycl queue and buffers,created for that queue, from Eigen::SyclDevice; modifying sycl tests to support the evaluation of the results for both row major and column major data layout on all different devices that are supported by Sycl{CPU; GPU; and Host}.
2016-11-19 00:20:42 +08:00
sycl_device.memcpyDeviceToHost(full_redux_gpu.data(), gpu_out_data, sizeof(DataType));
Converting all sycl buffers to uninitialised device only buffers; adding memcpyHostToDevice and memcpyDeviceToHost on syclDevice; modifying all examples to obey the new rules; moving sycl queue creating to the device based on Benoit suggestion; removing the sycl specefic condition for returning m_result in TensorReduction.h according to Benoit suggestion.
2016-11-09 01:08:02 +08:00
// Check that the CPU and GPU reductions return the same result.
VERIFY_IS_APPROX(full_redux_gpu(), full_redux());
Removed the sycl include from Eigen/Core and moved it to Unsupported/Eigen/CXX11/Tensor; added TensorReduction for sycl (full reduction and partial reduction); added TensorReduction test case for sycl (full reduction and partial reduction); fixed the tile size on TensorSyclRun.h based on the device max work group size;
2016-11-05 02:18:19 +08:00
Converting all sycl buffers to uninitialised device only buffers; adding memcpyHostToDevice and memcpyDeviceToHost on syclDevice; modifying all examples to obey the new rules; moving sycl queue creating to the device based on Benoit suggestion; removing the sycl specefic condition for returning m_result in TensorReduction.h according to Benoit suggestion.
2016-11-09 01:08:02 +08:00
sycl_device.deallocate(gpu_in_data);
sycl_device.deallocate(gpu_out_data);
}
Reducing warnings in Sycl backend.
2017-02-01 23:29:53 +08:00
template <typename DataType, int DataLayout, typename IndexType>
Converting all sycl buffers to uninitialised device only buffers; adding memcpyHostToDevice and memcpyDeviceToHost on syclDevice; modifying all examples to obey the new rules; moving sycl queue creating to the device based on Benoit suggestion; removing the sycl specefic condition for returning m_result in TensorReduction.h according to Benoit suggestion.
2016-11-09 01:08:02 +08:00
static void test_first_dim_reductions_sycl(const Eigen::SyclDevice& sycl_device) {
Removed the sycl include from Eigen/Core and moved it to Unsupported/Eigen/CXX11/Tensor; added TensorReduction for sycl (full reduction and partial reduction); added TensorReduction test case for sycl (full reduction and partial reduction); fixed the tile size on TensorSyclRun.h based on the device max work group size;
2016-11-05 02:18:19 +08:00
Reducing warnings in Sycl backend.
2017-02-01 23:29:53 +08:00
IndexType dim_x = 145;
IndexType dim_y = 1;
IndexType dim_z = 67;
Removed the sycl include from Eigen/Core and moved it to Unsupported/Eigen/CXX11/Tensor; added TensorReduction for sycl (full reduction and partial reduction); added TensorReduction test case for sycl (full reduction and partial reduction); fixed the tile size on TensorSyclRun.h based on the device max work group size;
2016-11-05 02:18:19 +08:00
Reducing warnings in Sycl backend.
2017-02-01 23:29:53 +08:00
array<IndexType, 3> tensorRange = {{dim_x, dim_y, dim_z}};
Eigen::array<IndexType, 1> red_axis;
Removed the sycl include from Eigen/Core and moved it to Unsupported/Eigen/CXX11/Tensor; added TensorReduction for sycl (full reduction and partial reduction); added TensorReduction test case for sycl (full reduction and partial reduction); fixed the tile size on TensorSyclRun.h based on the device max work group size;
2016-11-05 02:18:19 +08:00
red_axis[0] = 0;
Reducing warnings in Sycl backend.
2017-02-01 23:29:53 +08:00
array<IndexType, 2> reduced_tensorRange = {{dim_y, dim_z}};
Removed the sycl include from Eigen/Core and moved it to Unsupported/Eigen/CXX11/Tensor; added TensorReduction for sycl (full reduction and partial reduction); added TensorReduction test case for sycl (full reduction and partial reduction); fixed the tile size on TensorSyclRun.h based on the device max work group size;
2016-11-05 02:18:19 +08:00
Reducing warnings in Sycl backend.
2017-02-01 23:29:53 +08:00
Tensor<DataType, 3, DataLayout, IndexType> in(tensorRange);
Tensor<DataType, 2, DataLayout, IndexType> redux(reduced_tensorRange);
Tensor<DataType, 2, DataLayout, IndexType> redux_gpu(reduced_tensorRange);
Converting all sycl buffers to uninitialised device only buffers; adding memcpyHostToDevice and memcpyDeviceToHost on syclDevice; modifying all examples to obey the new rules; moving sycl queue creating to the device based on Benoit suggestion; removing the sycl specefic condition for returning m_result in TensorReduction.h according to Benoit suggestion.
2016-11-09 01:08:02 +08:00
in.setRandom();
Removed the sycl include from Eigen/Core and moved it to Unsupported/Eigen/CXX11/Tensor; added TensorReduction for sycl (full reduction and partial reduction); added TensorReduction test case for sycl (full reduction and partial reduction); fixed the tile size on TensorSyclRun.h based on the device max work group size;
2016-11-05 02:18:19 +08:00
Converting all sycl buffers to uninitialised device only buffers; adding memcpyHostToDevice and memcpyDeviceToHost on syclDevice; modifying all examples to obey the new rules; moving sycl queue creating to the device based on Benoit suggestion; removing the sycl specefic condition for returning m_result in TensorReduction.h according to Benoit suggestion.
2016-11-09 01:08:02 +08:00
redux= in.sum(red_axis);
Removed the sycl include from Eigen/Core and moved it to Unsupported/Eigen/CXX11/Tensor; added TensorReduction for sycl (full reduction and partial reduction); added TensorReduction test case for sycl (full reduction and partial reduction); fixed the tile size on TensorSyclRun.h based on the device max work group size;
2016-11-05 02:18:19 +08:00
Modifying TensorDeviceSycl.h to always create buffer of type uint8_t and convert them to the actual type at the execution on the device; adding the queue interface class to separate the lifespan of sycl queue and buffers,created for that queue, from Eigen::SyclDevice; modifying sycl tests to support the evaluation of the results for both row major and column major data layout on all different devices that are supported by Sycl{CPU; GPU; and Host}.
2016-11-19 00:20:42 +08:00
DataType* gpu_in_data = static_cast<DataType*>(sycl_device.allocate(in.dimensions().TotalSize()*sizeof(DataType)));
DataType* gpu_out_data = static_cast<DataType*>(sycl_device.allocate(redux_gpu.dimensions().TotalSize()*sizeof(DataType)));
Removed the sycl include from Eigen/Core and moved it to Unsupported/Eigen/CXX11/Tensor; added TensorReduction for sycl (full reduction and partial reduction); added TensorReduction test case for sycl (full reduction and partial reduction); fixed the tile size on TensorSyclRun.h based on the device max work group size;
2016-11-05 02:18:19 +08:00
Reducing warnings in Sycl backend.
2017-02-01 23:29:53 +08:00
TensorMap<Tensor<DataType, 3, DataLayout, IndexType> > in_gpu(gpu_in_data, tensorRange);
TensorMap<Tensor<DataType, 2, DataLayout, IndexType> > out_gpu(gpu_out_data, reduced_tensorRange);
Removed the sycl include from Eigen/Core and moved it to Unsupported/Eigen/CXX11/Tensor; added TensorReduction for sycl (full reduction and partial reduction); added TensorReduction test case for sycl (full reduction and partial reduction); fixed the tile size on TensorSyclRun.h based on the device max work group size;
2016-11-05 02:18:19 +08:00
Modifying TensorDeviceSycl.h to always create buffer of type uint8_t and convert them to the actual type at the execution on the device; adding the queue interface class to separate the lifespan of sycl queue and buffers,created for that queue, from Eigen::SyclDevice; modifying sycl tests to support the evaluation of the results for both row major and column major data layout on all different devices that are supported by Sycl{CPU; GPU; and Host}.
2016-11-19 00:20:42 +08:00
sycl_device.memcpyHostToDevice(gpu_in_data, in.data(),(in.dimensions().TotalSize())*sizeof(DataType));
Converting all sycl buffers to uninitialised device only buffers; adding memcpyHostToDevice and memcpyDeviceToHost on syclDevice; modifying all examples to obey the new rules; moving sycl queue creating to the device based on Benoit suggestion; removing the sycl specefic condition for returning m_result in TensorReduction.h according to Benoit suggestion.
2016-11-09 01:08:02 +08:00
out_gpu.device(sycl_device) = in_gpu.sum(red_axis);
Modifying TensorDeviceSycl.h to always create buffer of type uint8_t and convert them to the actual type at the execution on the device; adding the queue interface class to separate the lifespan of sycl queue and buffers,created for that queue, from Eigen::SyclDevice; modifying sycl tests to support the evaluation of the results for both row major and column major data layout on all different devices that are supported by Sycl{CPU; GPU; and Host}.
2016-11-19 00:20:42 +08:00
sycl_device.memcpyDeviceToHost(redux_gpu.data(), gpu_out_data, redux_gpu.dimensions().TotalSize()*sizeof(DataType));
Converting all sycl buffers to uninitialised device only buffers; adding memcpyHostToDevice and memcpyDeviceToHost on syclDevice; modifying all examples to obey the new rules; moving sycl queue creating to the device based on Benoit suggestion; removing the sycl specefic condition for returning m_result in TensorReduction.h according to Benoit suggestion.
2016-11-09 01:08:02 +08:00
// Check that the CPU and GPU reductions return the same result.
Reducing warnings in Sycl backend.
2017-02-01 23:29:53 +08:00
for(IndexType j=0; j<reduced_tensorRange[0]; j++ )
for(IndexType k=0; k<reduced_tensorRange[1]; k++ )
Converting all sycl buffers to uninitialised device only buffers; adding memcpyHostToDevice and memcpyDeviceToHost on syclDevice; modifying all examples to obey the new rules; moving sycl queue creating to the device based on Benoit suggestion; removing the sycl specefic condition for returning m_result in TensorReduction.h according to Benoit suggestion.
2016-11-09 01:08:02 +08:00
VERIFY_IS_APPROX(redux_gpu(j,k), redux(j,k));
sycl_device.deallocate(gpu_in_data);
sycl_device.deallocate(gpu_out_data);
}
Reducing warnings in Sycl backend.
2017-02-01 23:29:53 +08:00
template <typename DataType, int DataLayout, typename IndexType>
Converting all sycl buffers to uninitialised device only buffers; adding memcpyHostToDevice and memcpyDeviceToHost on syclDevice; modifying all examples to obey the new rules; moving sycl queue creating to the device based on Benoit suggestion; removing the sycl specefic condition for returning m_result in TensorReduction.h according to Benoit suggestion.
2016-11-09 01:08:02 +08:00
static void test_last_dim_reductions_sycl(const Eigen::SyclDevice &sycl_device) {
Removed the sycl include from Eigen/Core and moved it to Unsupported/Eigen/CXX11/Tensor; added TensorReduction for sycl (full reduction and partial reduction); added TensorReduction test case for sycl (full reduction and partial reduction); fixed the tile size on TensorSyclRun.h based on the device max work group size;
2016-11-05 02:18:19 +08:00
Reducing warnings in Sycl backend.
2017-02-01 23:29:53 +08:00
IndexType dim_x = 567;
IndexType dim_y = 1;
IndexType dim_z = 47;
Removed the sycl include from Eigen/Core and moved it to Unsupported/Eigen/CXX11/Tensor; added TensorReduction for sycl (full reduction and partial reduction); added TensorReduction test case for sycl (full reduction and partial reduction); fixed the tile size on TensorSyclRun.h based on the device max work group size;
2016-11-05 02:18:19 +08:00
Reducing warnings in Sycl backend.
2017-02-01 23:29:53 +08:00
array<IndexType, 3> tensorRange = {{dim_x, dim_y, dim_z}};
Eigen::array<IndexType, 1> red_axis;
Removed the sycl include from Eigen/Core and moved it to Unsupported/Eigen/CXX11/Tensor; added TensorReduction for sycl (full reduction and partial reduction); added TensorReduction test case for sycl (full reduction and partial reduction); fixed the tile size on TensorSyclRun.h based on the device max work group size;
2016-11-05 02:18:19 +08:00
red_axis[0] = 2;
Reducing warnings in Sycl backend.
2017-02-01 23:29:53 +08:00
array<IndexType, 2> reduced_tensorRange = {{dim_x, dim_y}};
Removed the sycl include from Eigen/Core and moved it to Unsupported/Eigen/CXX11/Tensor; added TensorReduction for sycl (full reduction and partial reduction); added TensorReduction test case for sycl (full reduction and partial reduction); fixed the tile size on TensorSyclRun.h based on the device max work group size;
2016-11-05 02:18:19 +08:00
Reducing warnings in Sycl backend.
2017-02-01 23:29:53 +08:00
Tensor<DataType, 3, DataLayout, IndexType> in(tensorRange);
Tensor<DataType, 2, DataLayout, IndexType> redux(reduced_tensorRange);
Tensor<DataType, 2, DataLayout, IndexType> redux_gpu(reduced_tensorRange);
Converting all sycl buffers to uninitialised device only buffers; adding memcpyHostToDevice and memcpyDeviceToHost on syclDevice; modifying all examples to obey the new rules; moving sycl queue creating to the device based on Benoit suggestion; removing the sycl specefic condition for returning m_result in TensorReduction.h according to Benoit suggestion.
2016-11-09 01:08:02 +08:00
in.setRandom();
redux= in.sum(red_axis);
Modifying TensorDeviceSycl.h to always create buffer of type uint8_t and convert them to the actual type at the execution on the device; adding the queue interface class to separate the lifespan of sycl queue and buffers,created for that queue, from Eigen::SyclDevice; modifying sycl tests to support the evaluation of the results for both row major and column major data layout on all different devices that are supported by Sycl{CPU; GPU; and Host}.
2016-11-19 00:20:42 +08:00
DataType* gpu_in_data = static_cast<DataType*>(sycl_device.allocate(in.dimensions().TotalSize()*sizeof(DataType)));
DataType* gpu_out_data = static_cast<DataType*>(sycl_device.allocate(redux_gpu.dimensions().TotalSize()*sizeof(DataType)));
Converting all sycl buffers to uninitialised device only buffers; adding memcpyHostToDevice and memcpyDeviceToHost on syclDevice; modifying all examples to obey the new rules; moving sycl queue creating to the device based on Benoit suggestion; removing the sycl specefic condition for returning m_result in TensorReduction.h according to Benoit suggestion.
2016-11-09 01:08:02 +08:00
Reducing warnings in Sycl backend.
2017-02-01 23:29:53 +08:00
TensorMap<Tensor<DataType, 3, DataLayout, IndexType> > in_gpu(gpu_in_data, tensorRange);
TensorMap<Tensor<DataType, 2, DataLayout, IndexType> > out_gpu(gpu_out_data, reduced_tensorRange);
Converting all sycl buffers to uninitialised device only buffers; adding memcpyHostToDevice and memcpyDeviceToHost on syclDevice; modifying all examples to obey the new rules; moving sycl queue creating to the device based on Benoit suggestion; removing the sycl specefic condition for returning m_result in TensorReduction.h according to Benoit suggestion.
2016-11-09 01:08:02 +08:00
Modifying TensorDeviceSycl.h to always create buffer of type uint8_t and convert them to the actual type at the execution on the device; adding the queue interface class to separate the lifespan of sycl queue and buffers,created for that queue, from Eigen::SyclDevice; modifying sycl tests to support the evaluation of the results for both row major and column major data layout on all different devices that are supported by Sycl{CPU; GPU; and Host}.
2016-11-19 00:20:42 +08:00
sycl_device.memcpyHostToDevice(gpu_in_data, in.data(),(in.dimensions().TotalSize())*sizeof(DataType));
Converting all sycl buffers to uninitialised device only buffers; adding memcpyHostToDevice and memcpyDeviceToHost on syclDevice; modifying all examples to obey the new rules; moving sycl queue creating to the device based on Benoit suggestion; removing the sycl specefic condition for returning m_result in TensorReduction.h according to Benoit suggestion.
2016-11-09 01:08:02 +08:00
out_gpu.device(sycl_device) = in_gpu.sum(red_axis);
Modifying TensorDeviceSycl.h to always create buffer of type uint8_t and convert them to the actual type at the execution on the device; adding the queue interface class to separate the lifespan of sycl queue and buffers,created for that queue, from Eigen::SyclDevice; modifying sycl tests to support the evaluation of the results for both row major and column major data layout on all different devices that are supported by Sycl{CPU; GPU; and Host}.
2016-11-19 00:20:42 +08:00
sycl_device.memcpyDeviceToHost(redux_gpu.data(), gpu_out_data, redux_gpu.dimensions().TotalSize()*sizeof(DataType));
Removed the sycl include from Eigen/Core and moved it to Unsupported/Eigen/CXX11/Tensor; added TensorReduction for sycl (full reduction and partial reduction); added TensorReduction test case for sycl (full reduction and partial reduction); fixed the tile size on TensorSyclRun.h based on the device max work group size;
2016-11-05 02:18:19 +08:00
// Check that the CPU and GPU reductions return the same result.
Reducing warnings in Sycl backend.
2017-02-01 23:29:53 +08:00
for(IndexType j=0; j<reduced_tensorRange[0]; j++ )
for(IndexType k=0; k<reduced_tensorRange[1]; k++ )
Removed the sycl include from Eigen/Core and moved it to Unsupported/Eigen/CXX11/Tensor; added TensorReduction for sycl (full reduction and partial reduction); added TensorReduction test case for sycl (full reduction and partial reduction); fixed the tile size on TensorSyclRun.h based on the device max work group size;
2016-11-05 02:18:19 +08:00
VERIFY_IS_APPROX(redux_gpu(j,k), redux(j,k));
Converting all sycl buffers to uninitialised device only buffers; adding memcpyHostToDevice and memcpyDeviceToHost on syclDevice; modifying all examples to obey the new rules; moving sycl queue creating to the device based on Benoit suggestion; removing the sycl specefic condition for returning m_result in TensorReduction.h according to Benoit suggestion.
2016-11-09 01:08:02 +08:00
sycl_device.deallocate(gpu_in_data);
sycl_device.deallocate(gpu_out_data);
Removed the sycl include from Eigen/Core and moved it to Unsupported/Eigen/CXX11/Tensor; added TensorReduction for sycl (full reduction and partial reduction); added TensorReduction test case for sycl (full reduction and partial reduction); fixed the tile size on TensorSyclRun.h based on the device max work group size;
2016-11-05 02:18:19 +08:00
}
Only runs the cxx11_tensor_reduction_sycl on devices that are available.
2016-11-19 08:31:14 +08:00
template<typename DataType> void sycl_reduction_test_per_device(const cl::sycl::device& d){
std::cout << "Running on " << d.template get_info<cl::sycl::info::device::name>() << std::endl;
QueueInterface queueInterface(d);
Modifying TensorDeviceSycl.h to always create buffer of type uint8_t and convert them to the actual type at the execution on the device; adding the queue interface class to separate the lifespan of sycl queue and buffers,created for that queue, from Eigen::SyclDevice; modifying sycl tests to support the evaluation of the results for both row major and column major data layout on all different devices that are supported by Sycl{CPU; GPU; and Host}.
2016-11-19 00:20:42 +08:00
auto sycl_device = Eigen::SyclDevice(&queueInterface);
Only runs the cxx11_tensor_reduction_sycl on devices that are available.
2016-11-19 08:31:14 +08:00
Reducing warnings in Sycl backend.
2017-02-01 23:29:53 +08:00
test_full_reductions_sycl<DataType, RowMajor, int64_t>(sycl_device);
test_first_dim_reductions_sycl<DataType, RowMajor, int64_t>(sycl_device);
test_last_dim_reductions_sycl<DataType, RowMajor, int64_t>(sycl_device);
test_full_reductions_sycl<DataType, ColMajor, int64_t>(sycl_device);
test_first_dim_reductions_sycl<DataType, ColMajor, int64_t>(sycl_device);
test_last_dim_reductions_sycl<DataType, ColMajor, int64_t>(sycl_device);
Modifying TensorDeviceSycl.h to always create buffer of type uint8_t and convert them to the actual type at the execution on the device; adding the queue interface class to separate the lifespan of sycl queue and buffers,created for that queue, from Eigen::SyclDevice; modifying sycl tests to support the evaluation of the results for both row major and column major data layout on all different devices that are supported by Sycl{CPU; GPU; and Host}.
2016-11-19 00:20:42 +08:00
}
Removed the sycl include from Eigen/Core and moved it to Unsupported/Eigen/CXX11/Tensor; added TensorReduction for sycl (full reduction and partial reduction); added TensorReduction test case for sycl (full reduction and partial reduction); fixed the tile size on TensorSyclRun.h based on the device max work group size;
2016-11-05 02:18:19 +08:00
void test_cxx11_tensor_reduction_sycl() {
Fixing LLVM error on TensorMorphingSycl.h on GPU; fixing int64_t crash for tensor_broadcast_sycl on GPU; adding get_sycl_supported_devices() on syclDevice.h.
2016-11-26 00:19:07 +08:00
for (const auto& device :Eigen::get_sycl_supported_devices()) {
CALL_SUBTEST(sycl_reduction_test_per_device<float>(device));
Only runs the cxx11_tensor_reduction_sycl on devices that are available.
2016-11-19 08:31:14 +08:00
}
Removed the sycl include from Eigen/Core and moved it to Unsupported/Eigen/CXX11/Tensor; added TensorReduction for sycl (full reduction and partial reduction); added TensorReduction test case for sycl (full reduction and partial reduction); fixed the tile size on TensorSyclRun.h based on the device max work group size;
2016-11-05 02:18:19 +08:00
}
Reference in New Issue Copy Permalink