1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
|
// This file is part of Eigen, a lightweight C++ template library
// for linear algebra.
//
// Copyright (C) 2014 Benoit Steiner <benoit.steiner.goog@gmail.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/.
#ifndef EIGEN_CXX11_TENSOR_TENSOR_DEVICE_H
#define EIGEN_CXX11_TENSOR_TENSOR_DEVICE_H
namespace Eigen {
/** \class TensorDevice
* \ingroup CXX11_Tensor_Module
*
* \brief Pseudo expression providing an operator = that will evaluate its argument
* on the specified computing 'device' (GPU, thread pool, ...)
*
* Example:
* C.device(EIGEN_GPU) = A + B;
*
* Todo: thread pools.
* Todo: operator +=, -=, *= and so on.
*/
template <typename ExpressionType, typename DeviceType> class TensorDevice {
public:
TensorDevice(const DeviceType& device, ExpressionType& expression) : m_device(device), m_expression(expression) {}
template<typename OtherDerived>
EIGEN_STRONG_INLINE TensorDevice& operator=(const OtherDerived& other) {
typedef TensorAssignOp<ExpressionType, const OtherDerived> Assign;
Assign assign(m_expression, other);
static const bool Vectorize = TensorEvaluator<const Assign, DeviceType>::PacketAccess;
internal::TensorExecutor<const Assign, DeviceType, Vectorize>::run(assign, m_device);
return *this;
}
template<typename OtherDerived>
EIGEN_STRONG_INLINE TensorDevice& operator+=(const OtherDerived& other) {
typedef typename OtherDerived::Scalar Scalar;
typedef TensorCwiseBinaryOp<internal::scalar_sum_op<Scalar>, const ExpressionType, const OtherDerived> Sum;
Sum sum(m_expression, other);
typedef TensorAssignOp<ExpressionType, const Sum> Assign;
Assign assign(m_expression, sum);
static const bool Vectorize = TensorEvaluator<const Assign, DeviceType>::PacketAccess;
internal::TensorExecutor<const Assign, DeviceType, Vectorize>::run(assign, m_device);
return *this;
}
protected:
const DeviceType& m_device;
ExpressionType& m_expression;
};
#ifdef EIGEN_USE_THREADS
template <typename ExpressionType> class TensorDevice<ExpressionType, ThreadPoolDevice> {
public:
TensorDevice(const ThreadPoolDevice& device, ExpressionType& expression) : m_device(device), m_expression(expression) {}
template<typename OtherDerived>
EIGEN_STRONG_INLINE TensorDevice& operator=(const OtherDerived& other) {
typedef TensorAssignOp<ExpressionType, const OtherDerived> Assign;
Assign assign(m_expression, other);
static const bool Vectorize = TensorEvaluator<const Assign, ThreadPoolDevice>::PacketAccess;
internal::TensorExecutor<const Assign, ThreadPoolDevice, Vectorize>::run(assign, m_device);
return *this;
}
template<typename OtherDerived>
EIGEN_STRONG_INLINE TensorDevice& operator+=(const OtherDerived& other) {
typedef typename OtherDerived::Scalar Scalar;
typedef TensorCwiseBinaryOp<internal::scalar_sum_op<Scalar>, const ExpressionType, const OtherDerived> Sum;
Sum sum(m_expression, other);
typedef TensorAssignOp<ExpressionType, const Sum> Assign;
Assign assign(m_expression, sum);
static const bool Vectorize = TensorEvaluator<const Assign, ThreadPoolDevice>::PacketAccess;
internal::TensorExecutor<const Assign, ThreadPoolDevice, Vectorize>::run(assign, m_device);
return *this;
}
protected:
const ThreadPoolDevice& m_device;
ExpressionType& m_expression;
};
#endif
#if defined(EIGEN_USE_GPU) && defined(__CUDACC__)
template <typename ExpressionType> class TensorDevice<ExpressionType, GpuDevice>
{
public:
TensorDevice(const GpuDevice& device, ExpressionType& expression) : m_device(device), m_expression(expression) {}
template<typename OtherDerived>
EIGEN_STRONG_INLINE TensorDevice& operator=(const OtherDerived& other) {
typedef TensorAssignOp<ExpressionType, const OtherDerived> Assign;
Assign assign(m_expression, other);
internal::TensorExecutor<const Assign, GpuDevice, false>::run(assign, m_device);
return *this;
}
template<typename OtherDerived>
EIGEN_STRONG_INLINE TensorDevice& operator+=(const OtherDerived& other) {
typedef typename OtherDerived::Scalar Scalar;
typedef TensorCwiseBinaryOp<internal::scalar_sum_op<Scalar>, const ExpressionType, const OtherDerived> Sum;
Sum sum(m_expression, other);
typedef TensorAssignOp<ExpressionType, const Sum> Assign;
Assign assign(m_expression, sum);
internal::TensorExecutor<const Assign, GpuDevice, false>::run(assign, m_device);
return *this;
}
protected:
const GpuDevice& m_device;
ExpressionType m_expression;
};
#endif
} // end namespace Eigen
#endif // EIGEN_CXX11_TENSOR_TENSOR_DEVICE_H
|