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
210
211
212
213
214
215
216
217
|
#include <torch/csrc/autograd/input_buffer.h>
#include <ATen/BatchedTensorImpl.h>
#include <ATen/SparseCsrTensorUtils.h>
#include <ATen/SparseTensorUtils.h>
#include <ATen/TensorOperators.h>
#include <c10/core/DeviceGuard.h>
#include <c10/core/Event.h>
#include <c10/core/StreamGuard.h>
#include <c10/util/Optional.h>
#include <cstddef>
#include <utility>
#include <vector>
namespace torch {
namespace autograd {
namespace {
// look what you made me do >.<
// Divergent paths for per-Impl stream recording that leak implementation
// details of the impls should not be needed here.
// See https://github.com/pytorch/pytorch/issues/60306
// TODO: clean this up when https://github.com/pytorch/pytorch/issues/60306 is
// improved
void record_stream_any_impl(Variable& var, c10::Stream& stream) {
const auto guard = c10::impl::VirtualGuardImpl(c10::DeviceType::CUDA);
if (C10_UNLIKELY(at::isBatchedTensor(var))) {
auto* impl = at::maybeGetBatchedImpl(var);
if (impl) {
guard.recordDataPtrOnStream(impl->value().storage().data_ptr(), stream);
} else {
TORCH_INTERNAL_ASSERT(false, "Expected batched tensor");
}
} else {
switch (var.layout()) {
case c10::kSparseCsr:
case c10::kSparseCsc:
case c10::kSparseBsr:
case c10::kSparseBsc: {
auto* impl = at::sparse_csr::get_sparse_csr_impl(var);
guard.recordDataPtrOnStream(
impl->values().storage().data_ptr(), stream);
guard.recordDataPtrOnStream(
impl->compressed_indices().storage().data_ptr(), stream);
guard.recordDataPtrOnStream(
impl->plain_indices().storage().data_ptr(), stream);
break;
}
case c10::kSparse: {
auto* impl = at::sparse::get_sparse_impl(var);
guard.recordDataPtrOnStream(
impl->values().storage().data_ptr(), stream);
guard.recordDataPtrOnStream(
impl->indices().storage().data_ptr(), stream);
break;
}
case c10::kStrided:
guard.recordDataPtrOnStream(var.storage().data_ptr(), stream);
break;
default:
TORCH_INTERNAL_ASSERT(
false, "Unknown layout in record_stream_any_impl");
}
}
}
} // anonymous namespace
static void accumulate(
std::vector<Variable>& buffer,
const size_t pos,
Variable&& var) {
TORCH_INTERNAL_ASSERT(pos < buffer.size());
auto& old_var = buffer[pos];
// ATen doesn't route sparse additions correctly...
// do dense + sparse in-place if possible
if (old_var.is_sparse()) {
// It is safe to change the Tensor inplace if the Tensor is only used in
// this buffer (this could be the gradient passed by the user) and that no
// other Tensor is using the same storage.
if (!var.is_sparse() && var.is_contiguous() && var.use_count() == 1 &&
var.storage().use_count() == 1) {
buffer[pos] = var.add_(old_var);
} else {
buffer[pos] = var + old_var;
}
} else {
if (var.is_sparse() && !old_var.is_sparse() && old_var.is_contiguous() &&
old_var.use_count() == 1 && old_var.storage().use_count() == 1) {
buffer[pos] = old_var.add_(var);
} else {
buffer[pos] = old_var + var;
}
}
}
void InputBuffer::add(
size_t pos,
Variable&& var,
const c10::optional<c10::Stream>& opt_producer_stream,
const c10::optional<c10::Stream>& opt_consumer_stream) {
TORCH_INTERNAL_ASSERT(pos < buffer.size());
if (!var.defined()) {
return;
}
// Switches to accumulate device
// The device (and stream) chosen for accumulation is:
// (1) var is not a CUDA variable. Accumulation happens on var's device.
// (2) var is a CUDA variable and it, the consumer, and the producer share
// the same device:
// (2a) Uses the consumer's stream as the accumulation stream
// (2b) Syncs the accumulation stream with the producer's stream (if
// different) (2c) Accumulates.
// (3) var is a CUDA variable and it shares a device with the consumer but
// not the producer:
// (3a) Uses the consumer's stream as the accumulation stream
// (3b) Syncs the accumulation stream with the consumer device's default
// stream (3c) Accumulates.
// (4) var is a CUDA variable and it shares a device with the producer but
// not the consumer:
// (4a) Uses the producer device's default stream as the accumulation
// stream (4b) Syncs the accumulation stream with the the producer's
// stream (4c) Accumulates.
// (5) var is a CUDA variable and it does not share a device with the
// consumer or producer.
// Accumulation happens on the var device's default stream.
TORCH_INTERNAL_ASSERT(device_of(var));
c10::optional<c10::Stream> opt_accumulate_stream = c10::nullopt;
if (device_of(var)->is_cuda()) {
const auto on_producer =
opt_producer_stream && device_of(var) == opt_producer_stream->device();
const auto on_consumer =
opt_consumer_stream && device_of(var) == opt_consumer_stream->device();
if (on_producer && on_consumer) {
// (2a)
opt_accumulate_stream = opt_consumer_stream;
if (opt_accumulate_stream != opt_producer_stream) {
// (2b)
auto event = c10::Event{c10::DeviceType::CUDA};
event.record(*opt_producer_stream);
opt_accumulate_stream->wait(event);
record_stream_any_impl(var, *opt_accumulate_stream);
}
} else {
c10::optional<c10::Stream> opt_sync_stream = c10::nullopt;
const auto guard = c10::impl::VirtualGuardImpl{c10::DeviceType::CUDA};
if (on_consumer && !on_producer) {
// (3a)
opt_accumulate_stream = opt_consumer_stream;
opt_sync_stream = guard.getDefaultStream(opt_consumer_stream->device());
} else if (on_producer && !on_consumer) {
// (4a)
opt_accumulate_stream =
guard.getDefaultStream(opt_producer_stream->device());
opt_sync_stream = opt_producer_stream;
} else {
// (5)
opt_accumulate_stream = guard.getDefaultStream(*device_of(var));
}
if (opt_sync_stream && (opt_accumulate_stream != opt_sync_stream)) {
// (3b), (4b)
c10::OptionalDeviceGuard device_guard{opt_sync_stream->device()};
auto event = c10::Event{c10::DeviceType::CUDA};
event.record(*opt_sync_stream);
opt_accumulate_stream->wait(event);
const auto guard = c10::impl::VirtualGuardImpl(c10::DeviceType::CUDA);
record_stream_any_impl(var, *opt_accumulate_stream);
}
}
}
auto& old_var = buffer[pos];
if (!old_var.defined()) {
buffer[pos] = std::move(var);
} else {
if (opt_accumulate_stream) {
c10::OptionalStreamGuard stream_guard{opt_accumulate_stream};
accumulate(buffer, pos, std::move(var));
} else {
// (1) non-CUDA variable
// Accumulation happens on variable's device
c10::OptionalDeviceGuard device_guard{device_of(var)};
accumulate(buffer, pos, std::move(var));
}
}
}
auto InputBuffer::device() const -> at::Device {
// Since we pick the first non-CPU tensor, this won't work with
// mixed device-type operations (e.g., an op that is both CUDA
// and XLA). This is *incredibly* unlikely, so we don't worry
// about it.
for (auto& var : buffer) {
if (var.defined()) {
auto device = var.device();
if (device.type() != at::kCPU) {
return device;
}
}
}
// Only report to the CPU thread if there really were no tensors
// from other devices.
return at::kCPU;
}
auto InputBuffer::variables(InputBuffer&& g) -> std::vector<Variable> {
std::vector<Variable> result = std::move(g.buffer);
return result;
}
} // namespace autograd
} // namespace torch
|
