forked from ROCm/pytorch
-
Notifications
You must be signed in to change notification settings - Fork 0
/
ForeachPointwiseOp.cu
209 lines (186 loc) · 14.6 KB
/
ForeachPointwiseOp.cu
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
197
198
199
200
201
202
203
204
205
206
207
208
209
#define TORCH_ASSERT_ONLY_METHOD_OPERATORS
#include <ATen/Dispatch.h>
#include <ATen/native/ForeachUtils.h>
#include <ATen/native/cuda/ForeachFunctors.cuh>
#include <ATen/NumericUtils.h>
#ifndef AT_PER_OPERATOR_HEADERS
#include <ATen/NativeFunctions.h>
#else
#include <ATen/ops/_foreach_add_native.h>
#include <ATen/ops/_foreach_addcdiv_native.h>
#include <ATen/ops/_foreach_addcmul_native.h>
#include <ATen/ops/_foreach_div_native.h>
#include <ATen/ops/_foreach_maximum_native.h>
#include <ATen/ops/_foreach_minimum_native.h>
#include <ATen/ops/_foreach_mul_native.h>
#include <ATen/ops/_foreach_sub_native.h>
#include <ATen/ops/empty_like_native.h>
#endif
namespace at { namespace native {
template<template<class> class Op>
std::vector<Tensor> foreach_pointwise_op(TensorList input, TensorList tensors1, TensorList tensors2, const Scalar& scalar) {
std::vector<std::vector<at::Tensor>> tensor_lists;
std::vector<at::Tensor> vec_res;
vec_res.reserve(input.size());
for (const auto& t: input) {
vec_res.emplace_back(at::native::empty_like(t));
}
tensor_lists.emplace_back(input.vec());
tensor_lists.emplace_back(tensors1.vec());
tensor_lists.emplace_back(tensors2.vec());
tensor_lists.emplace_back(std::move(vec_res));
AT_DISPATCH_ALL_TYPES_AND_COMPLEX_AND2(kHalf, kBFloat16, input[0].scalar_type(), "foreach_pointwise_op_cuda", [&]() {
using opmath_t = at::opmath_type<scalar_t>;
multi_tensor_apply<4>(tensor_lists,
PointwiseOpScalarFunctor<scalar_t,
/* depth */ 4,
/* r_args_depth */ 3,
/* res_arg_index */ 3>(),
Op<opmath_t>(),
scalar.to<opmath_t>());
});
return tensor_lists[3];
}
template<template<class> class Op>
void foreach_pointwise_op_(TensorList input, TensorList tensors1, TensorList tensors2, const Scalar& scalar) {
std::vector<std::vector<at::Tensor>> tensor_lists;
tensor_lists.emplace_back(input.vec());
tensor_lists.emplace_back(tensors1.vec());
tensor_lists.emplace_back(tensors2.vec());
AT_DISPATCH_ALL_TYPES_AND_COMPLEX_AND2(kHalf, kBFloat16, input[0].scalar_type(), "foreach_pointwise_op__cuda", [&]() {
using opmath_t = at::opmath_type<scalar_t>;
multi_tensor_apply<3>(tensor_lists,
PointwiseOpScalarFunctor<scalar_t,
/* depth */ 3,
/* r_args_depth */ 3,
/* res_arg_index */ 0>(),
Op<opmath_t>(),
scalar.to<opmath_t>());
});
}
template<template<class> class Op>
void foreach_pointwise_op_(TensorList input, TensorList tensors1, TensorList tensors2, at::ArrayRef<Scalar> scalars) {
std::vector<std::vector<at::Tensor>> tensor_lists;
tensor_lists.reserve(3);
tensor_lists.emplace_back(input.vec());
tensor_lists.emplace_back(tensors1.vec());
tensor_lists.emplace_back(tensors2.vec());
AT_DISPATCH_ALL_TYPES_AND_COMPLEX_AND2(kHalf, kBFloat16, input[0].scalar_type(), "foreach_pointwise_op__cuda", [&]() {
using opmath_t = at::opmath_type<scalar_t>;
multi_tensor_apply<3, opmath_t>(tensor_lists,
scalars,
PointwiseOpScalarListFunctor<scalar_t,
/* depth */ 3,
/* r_args_depth */ 3,
/* res_arg_index */ 0>(),
Op<opmath_t>());
});
}
template<template<class> class Op>
std::vector<Tensor> foreach_pointwise_op(TensorList input, TensorList tensors1, TensorList tensors2, at::ArrayRef<Scalar> scalars) {
std::vector<std::vector<at::Tensor>> tensor_lists;
tensor_lists.reserve(4);
std::vector<at::Tensor> vec_res;
vec_res.reserve(input.size());
for (const auto& t: input) {
vec_res.emplace_back(at::native::empty_like(t));
}
tensor_lists.emplace_back(input.vec());
tensor_lists.emplace_back(tensors1.vec());
tensor_lists.emplace_back(tensors2.vec());
tensor_lists.emplace_back(std::move(vec_res));
AT_DISPATCH_ALL_TYPES_AND_COMPLEX_AND2(kHalf, kBFloat16, input[0].scalar_type(), "foreach_pointwise_op_cuda", [&]() {
using opmath_t = at::opmath_type<scalar_t>;
multi_tensor_apply<4, opmath_t>(tensor_lists,
scalars,
PointwiseOpScalarListFunctor<scalar_t,
/* depth */ 4,
/* r_args_depth */ 3,
/* res_arg_index */ 3>(),
Op<opmath_t>());
});
return tensor_lists[3];
}
#define FOREACH_POINTWISE_OP_SCALAR(NAME, OP) \
std::vector<Tensor> foreach_tensor_##NAME##_scalar_cuda(TensorList input, TensorList tensors1, TensorList tensors2, const Scalar& scalar) { \
check_foreach_api_restrictions(input, tensors1, tensors2); \
\
if (!can_use_fast_route({input, tensors1, tensors2}, scalar) || has_integral_tensor(input, /* includeBool */ true)) { \
return at::native::foreach_tensor_##NAME##_scalar_slow(input, tensors1, tensors2, scalar); \
} \
\
return foreach_pointwise_op<OP>(input, tensors1, tensors2, scalar); \
} \
\
void foreach_tensor_##NAME##_scalar_cuda_(TensorList input, TensorList tensors1, TensorList tensors2, const Scalar& scalar) { \
check_foreach_api_restrictions(input, tensors1, tensors2); \
\
if (!can_use_fast_route({input, tensors1, tensors2}, scalar) || has_integral_tensor(input, /* includeBool */ true)) { \
return at::native::foreach_tensor_##NAME##_scalar_slow_(input, tensors1, tensors2, scalar); \
} \
\
foreach_pointwise_op_<OP>(input, tensors1, tensors2, scalar); \
}
#define FOREACH_POINTWISE_OP_SCALARLIST(NAME, OP) \
std::vector<Tensor> foreach_tensor_##NAME##_scalarlist_cuda(TensorList input, TensorList tensors1, TensorList tensors2, at::ArrayRef<Scalar> scalars) { \
check_foreach_api_restrictions(input, tensors1, tensors2, scalars); \
\
if (!can_use_fast_route({input, tensors1, tensors2}, scalars) || has_integral_tensor(input, /* includeBool */ true)) { \
return at::native::foreach_tensor_##NAME##_scalarlist_slow(input, tensors1, tensors2, scalars); \
} \
\
return foreach_pointwise_op<OP>(input, tensors1, tensors2, scalars); \
} \
\
void foreach_tensor_##NAME##_scalarlist_cuda_(TensorList input, TensorList tensors1, TensorList tensors2, at::ArrayRef<Scalar> scalars) { \
check_foreach_api_restrictions(input, tensors1, tensors2, scalars); \
\
if (!can_use_fast_route({input, tensors1, tensors2}, scalars) || has_integral_tensor(input, /* includeBool */ true)) { \
return at::native::foreach_tensor_##NAME##_scalarlist_slow_(input, tensors1, tensors2, scalars); \
} \
\
foreach_pointwise_op_<OP>(input, tensors1, tensors2, scalars); \
}
FOREACH_POINTWISE_OP_SCALAR(addcmul, std::multiplies);
FOREACH_POINTWISE_OP_SCALAR(addcdiv, std::divides);
FOREACH_POINTWISE_OP_SCALARLIST(addcmul, std::multiplies);
FOREACH_POINTWISE_OP_SCALARLIST(addcdiv, std::divides);
// Why bool tensors are pushed to slowpath?
// Because `AT_DISPATCH_ALL_TYPES_AND` is used below.
// TODO(mkozuki): Check whether it's possible to handle bool tensors in fastpath.
#define FOREACH_MAXIMUM_MINIMUM_OP(NAME, OP) \
std::vector<Tensor> foreach_tensor_##NAME##_cuda(TensorList tensors1, TensorList tensors2) { \
check_foreach_api_restrictions(tensors1, tensors2); \
if (!can_use_fast_route({tensors1, tensors2}) || has_bool_tensor(tensors1)) { \
return at::native::foreach_tensor_##NAME##_slow(tensors1, tensors2); \
} \
\
std::vector<std::vector<at::Tensor>> tensor_lists; \
std::vector<at::Tensor> vec_res; \
vec_res.reserve(tensors1.size()); \
for (const auto& t: tensors1) { \
vec_res.emplace_back(at::native::empty_like(t)); \
} \
\
tensor_lists.emplace_back(tensors1.vec()); \
tensor_lists.emplace_back(tensors2.vec()); \
tensor_lists.emplace_back(std::move(vec_res)); \
\
AT_DISPATCH_ALL_TYPES_AND2(kHalf, kBFloat16, tensors1[0].scalar_type(), "foreach_maximum_minimum_op_cuda", [&]() { \
using opmath_t = at::opmath_type<scalar_t>; \
auto op = [] GPU_LAMBDA (opmath_t a, opmath_t b) -> opmath_t { \
opmath_t c = a OP b ? a : b; \
if (_isnan(a)) { \
c = a; \
} \
return c;}; \
multi_tensor_apply<3>(tensor_lists, \
PointwiseOpListFunctor<scalar_t, 3>(), \
op); \
}); \
\
return tensor_lists[2]; \
} \
FOREACH_MAXIMUM_MINIMUM_OP(maximum, >)
FOREACH_MAXIMUM_MINIMUM_OP(minimum, <)
}} // namespace at::native