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
|
/* Copyright 2016 The TensorFlow Authors. All Rights Reserved.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://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.
==============================================================================*/
#if GOOGLE_CUDA
#define EIGEN_USE_GPU
#include "third_party/eigen3/unsupported/Eigen/CXX11/Tensor"
#include "tensorflow/core/kernels/inplace_ops_functor.h"
#include "tensorflow/core/util/cuda_kernel_helper.h"
namespace tensorflow {
namespace functor {
typedef Eigen::GpuDevice Device;
template <typename T>
__global__ void DoParallelConcatOpKernel(int nthreads, const int64 rows,
const int64 cols, int32 loc,
const T* src, T* dst) {
CUDA_1D_KERNEL_LOOP(idx, nthreads) {
int64 c = idx % cols;
int64 r = (loc % rows + rows) % rows; // Guard index range.
T* p = dst + r * cols + c;
const T* q = src + idx;
*p = ldg(q);
}
}
template <typename T>
Status DoParallelConcatUpdate(const Device& d, const Tensor& value, int32 loc,
Tensor* output) {
const int64 nelem = value.NumElements();
CudaLaunchConfig cfg = GetCudaLaunchConfig(nelem, d);
auto Toutput = output->flat_outer_dims<T>();
const int64 nrows = Toutput.dimension(0);
const int64 ncols = Toutput.dimension(1);
const T* src = value.flat<T>().data();
T* dst = output->flat<T>().data();
DoParallelConcatOpKernel<T>
<<<cfg.block_count, cfg.thread_per_block, 0, d.stream()>>>(
cfg.virtual_thread_count, nrows, ncols, loc, src, dst);
return Status::OK();
}
template <>
Status DoParallelConcat(const Device& d, const Tensor& value, int32 loc,
Tensor* output) {
CHECK_EQ(value.dtype(), output->dtype());
switch (value.dtype()) {
#define CASE(type) \
case DataTypeToEnum<type>::value: \
return DoParallelConcatUpdate<type>(d, value, loc, output); \
break;
CASE(float)
CASE(double)
CASE(Eigen::half)
// Using TF_CALL_GPU_NUMBER_TYPES(CASE) results in the compiler complaining
// that CASE is not defined...hence the above construction
#undef CASE
default:
return errors::InvalidArgument("Unsupported data type: ",
DataTypeString(value.dtype()));
}
return Status::OK();
}
template <typename T, InplaceOpType op>
__global__ void DoInplaceOpKernel(int nthreads, const int64 rows,
const int64 cols, const int64 n, const T* src,
const int32* rowids, T* dst) {
CUDA_1D_KERNEL_LOOP(idx, nthreads) {
int64 r = idx / cols;
int64 c = idx % cols;
r = (rowids[r] % rows + rows) % rows; // Guard index range.
T* p = dst + r * cols + c;
const T* q = src + idx;
switch (op) {
case I_UPDATE:
*p = ldg(q);
break;
case I_ADD:
*p += ldg(q);
break;
case I_SUB:
*p -= ldg(q);
break;
}
}
}
template <typename T>
void DoInplaceOp(const Device& d, InplaceOpType op, const Tensor& i,
const Tensor& v, Tensor* y) {
const int64 nelem = v.NumElements();
CudaLaunchConfig cfg = GetCudaLaunchConfig(nelem, d);
auto Ty = y->flat_outer_dims<T>();
const int64 nrows = Ty.dimension(0);
const int64 ncols = Ty.dimension(1);
const int64 n = i.NumElements();
const T* src = v.flat<T>().data();
// TODO(sjhwang): Check that first dimension fits in int32 range.
const int32* rowids = i.flat<int32>().data();
T* dst = y->flat<T>().data();
switch (op) {
case I_UPDATE:
DoInplaceOpKernel<T, I_UPDATE>
<<<cfg.block_count, cfg.thread_per_block, 0, d.stream()>>>(
cfg.virtual_thread_count, nrows, ncols, n, src, rowids, dst);
break;
case I_ADD:
DoInplaceOpKernel<T, I_ADD>
<<<cfg.block_count, cfg.thread_per_block, 0, d.stream()>>>(
cfg.virtual_thread_count, nrows, ncols, n, src, rowids, dst);
break;
case I_SUB:
DoInplaceOpKernel<T, I_SUB>
<<<cfg.block_count, cfg.thread_per_block, 0, d.stream()>>>(
cfg.virtual_thread_count, nrows, ncols, n, src, rowids, dst);
break;
}
}
template <>
Status DoInplace(const Device& d, InplaceOpType op, const Tensor& i,
const Tensor& v, Tensor* y) {
CHECK_EQ(v.dtype(), y->dtype());
switch (v.dtype()) {
#define CASE(type) \
case DataTypeToEnum<type>::value: \
DoInplaceOp<type>(d, op, i, v, y); \
break;
CASE(float)
CASE(double)
CASE(Eigen::half)
CASE(int64)
#undef CASE
default:
return errors::InvalidArgument("Unsupported data type: ",
DataTypeString(v.dtype()));
}
return Status::OK();
}
template <>
Status DoCopy(const Device& d, const Tensor& x, Tensor* y) {
CHECK_EQ(x.dtype(), y->dtype());
switch (x.dtype()) {
#define CASE(type) \
case DataTypeToEnum<type>::value: \
y->flat<type>().device(d) = x.flat<type>(); \
break;
CASE(float)
CASE(double)
CASE(Eigen::half)
CASE(int64)
#undef CASE
default:
return errors::InvalidArgument("Unsupported dtype: ",
DataTypeString(x.dtype()));
}
return Status::OK();
}
} // end namespace functor
} // namespace tensorflow
#endif // GOOGLE_CUDA
|
