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
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
|
// This file is part of Eigen, a lightweight C++ template library
// for linear algebra.
//
// Copyright (C) 2015 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_GENERATOR_H
#define EIGEN_CXX11_TENSOR_TENSOR_GENERATOR_H
namespace Eigen {
/** \class TensorGenerator
* \ingroup CXX11_Tensor_Module
*
* \brief Tensor generator class.
*
*
*/
namespace internal {
template<typename Generator, typename XprType>
struct traits<TensorGeneratorOp<Generator, XprType> > : public traits<XprType>
{
typedef typename XprType::Scalar Scalar;
typedef traits<XprType> XprTraits;
typedef typename XprTraits::StorageKind StorageKind;
typedef typename XprTraits::Index Index;
typedef typename XprType::Nested Nested;
typedef typename remove_reference<Nested>::type _Nested;
static const int NumDimensions = XprTraits::NumDimensions;
static const int Layout = XprTraits::Layout;
};
template<typename Generator, typename XprType>
struct eval<TensorGeneratorOp<Generator, XprType>, Eigen::Dense>
{
typedef const TensorGeneratorOp<Generator, XprType>& type;
};
template<typename Generator, typename XprType>
struct nested<TensorGeneratorOp<Generator, XprType>, 1, typename eval<TensorGeneratorOp<Generator, XprType> >::type>
{
typedef TensorGeneratorOp<Generator, XprType> type;
};
} // end namespace internal
template<typename Generator, typename XprType>
class TensorGeneratorOp : public TensorBase<TensorGeneratorOp<Generator, XprType>, ReadOnlyAccessors>
{
public:
typedef typename Eigen::internal::traits<TensorGeneratorOp>::Scalar Scalar;
typedef typename Eigen::NumTraits<Scalar>::Real RealScalar;
typedef typename XprType::CoeffReturnType CoeffReturnType;
typedef typename Eigen::internal::nested<TensorGeneratorOp>::type Nested;
typedef typename Eigen::internal::traits<TensorGeneratorOp>::StorageKind StorageKind;
typedef typename Eigen::internal::traits<TensorGeneratorOp>::Index Index;
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE TensorGeneratorOp(const XprType& expr, const Generator& generator)
: m_xpr(expr), m_generator(generator) {}
EIGEN_DEVICE_FUNC
const Generator& generator() const { return m_generator; }
EIGEN_DEVICE_FUNC
const typename internal::remove_all<typename XprType::Nested>::type&
expression() const { return m_xpr; }
protected:
typename XprType::Nested m_xpr;
const Generator m_generator;
};
// Eval as rvalue
template<typename Generator, typename ArgType, typename Device>
struct TensorEvaluator<const TensorGeneratorOp<Generator, ArgType>, Device>
{
typedef TensorGeneratorOp<Generator, ArgType> XprType;
typedef typename XprType::Index Index;
typedef typename TensorEvaluator<ArgType, Device>::Dimensions Dimensions;
static const int NumDims = internal::array_size<Dimensions>::value;
typedef typename XprType::Scalar Scalar;
typedef typename XprType::CoeffReturnType CoeffReturnType;
typedef typename PacketType<CoeffReturnType, Device>::type PacketReturnType;
enum {
IsAligned = false,
PacketAccess = (internal::unpacket_traits<PacketReturnType>::size > 1),
BlockAccess = false,
Layout = TensorEvaluator<ArgType, Device>::Layout,
CoordAccess = false, // to be implemented
RawAccess = false
};
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE TensorEvaluator(const XprType& op, const Device& device)
: m_generator(op.generator())
#ifdef EIGEN_USE_SYCL
, m_argImpl(op.expression(), device)
#endif
{
TensorEvaluator<ArgType, Device> argImpl(op.expression(), device);
m_dimensions = argImpl.dimensions();
if (static_cast<int>(Layout) == static_cast<int>(ColMajor)) {
m_strides[0] = 1;
for (int i = 1; i < NumDims; ++i) {
m_strides[i] = m_strides[i - 1] * m_dimensions[i - 1];
}
} else {
m_strides[NumDims - 1] = 1;
for (int i = NumDims - 2; i >= 0; --i) {
m_strides[i] = m_strides[i + 1] * m_dimensions[i + 1];
}
}
}
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const Dimensions& dimensions() const { return m_dimensions; }
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE bool evalSubExprsIfNeeded(Scalar* /*data*/) {
return true;
}
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE void cleanup() {
}
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE CoeffReturnType coeff(Index index) const
{
array<Index, NumDims> coords;
extract_coordinates(index, coords);
return m_generator(coords);
}
template<int LoadMode>
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE PacketReturnType packet(Index index) const
{
const int packetSize = internal::unpacket_traits<PacketReturnType>::size;
EIGEN_STATIC_ASSERT((packetSize > 1), YOU_MADE_A_PROGRAMMING_MISTAKE)
eigen_assert(index+packetSize-1 < dimensions().TotalSize());
EIGEN_ALIGN_MAX typename internal::remove_const<CoeffReturnType>::type values[packetSize];
for (int i = 0; i < packetSize; ++i) {
values[i] = coeff(index+i);
}
PacketReturnType rslt = internal::pload<PacketReturnType>(values);
return rslt;
}
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE TensorOpCost
costPerCoeff(bool) const {
// TODO(rmlarsen): This is just a placeholder. Define interface to make
// generators return their cost.
return TensorOpCost(0, 0, TensorOpCost::AddCost<Scalar>() +
TensorOpCost::MulCost<Scalar>());
}
EIGEN_DEVICE_FUNC Scalar* data() const { return NULL; }
#ifdef EIGEN_USE_SYCL
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const TensorEvaluator<ArgType, Device>& impl() const { return m_argImpl; }
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const Generator& functor() const { return m_generator; }
#endif
protected:
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE
void extract_coordinates(Index index, array<Index, NumDims>& coords) const {
if (static_cast<int>(Layout) == static_cast<int>(ColMajor)) {
for (int i = NumDims - 1; i > 0; --i) {
const Index idx = index / m_strides[i];
index -= idx * m_strides[i];
coords[i] = idx;
}
coords[0] = index;
} else {
for (int i = 0; i < NumDims - 1; ++i) {
const Index idx = index / m_strides[i];
index -= idx * m_strides[i];
coords[i] = idx;
}
coords[NumDims-1] = index;
}
}
Dimensions m_dimensions;
array<Index, NumDims> m_strides;
Generator m_generator;
#ifdef EIGEN_USE_SYCL
TensorEvaluator<ArgType, Device> m_argImpl;
#endif
};
} // end namespace Eigen
#endif // EIGEN_CXX11_TENSOR_TENSOR_GENERATOR_H
|
