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
|
#include <pybind11/pybind11.h>
#include <torch/csrc/cuda/Stream.h>
#include <torch/csrc/cuda/Module.h>
#include <torch/csrc/Device.h>
#include <torch/csrc/THP.h>
#include <c10/cuda/CUDAGuard.h>
#include <structmember.h>
#include <cuda_runtime_api.h>
PyObject *THCPStreamClass = nullptr;
static PyObject * THCPStream_pynew(
PyTypeObject *type, PyObject *args, PyObject *kwargs) {
HANDLE_TH_ERRORS
int current_device;
THCudaCheck(cudaGetDevice(¤t_device));
int priority = 0;
uint64_t cdata = 0;
static char *kwlist[] = {"priority", "_cdata", nullptr};
if (!PyArg_ParseTupleAndKeywords(
args, kwargs, "|iK", kwlist, &priority, &cdata)) {
return nullptr;
}
THPObjectPtr ptr(type->tp_alloc(type, 0));
if (!ptr) {
return nullptr;
}
at::cuda::CUDAStream stream =
cdata ?
at::cuda::CUDAStream::unpack(cdata) :
at::cuda::getStreamFromPool(
/* isHighPriority */ priority < 0 ? true : false);
THCPStream* self = (THCPStream *)ptr.get();
self->cdata = stream.pack();
new (&self->cuda_stream) at::cuda::CUDAStream(stream);
return (PyObject *)ptr.release();
END_HANDLE_TH_ERRORS
}
static void THCPStream_dealloc(THCPStream *self) {
self->cuda_stream.~CUDAStream();
Py_TYPE(self)->tp_free((PyObject*)self);
}
static PyObject * THCPStream_get_device(THCPStream *self, void *unused) {
HANDLE_TH_ERRORS
return THPDevice_New(self->cuda_stream.device());
END_HANDLE_TH_ERRORS
}
static PyObject * THCPStream_get_cuda_stream(THCPStream *self, void *unused) {
HANDLE_TH_ERRORS
return PyLong_FromVoidPtr(self->cuda_stream.stream());
END_HANDLE_TH_ERRORS
}
static PyObject * THCPStream_get_priority(THCPStream *self, void *unused) {
HANDLE_TH_ERRORS
return PyLong_FromLong(self->cuda_stream.priority());
END_HANDLE_TH_ERRORS
}
static PyObject * THCPStream_priority_range() {
HANDLE_TH_ERRORS
int least_priority, greatest_priority;
std::tie(least_priority, greatest_priority) =
at::cuda::CUDAStream::priority_range();
return Py_BuildValue("(ii)", least_priority, greatest_priority);
END_HANDLE_TH_ERRORS
}
static PyObject * THCPStream_query(THCPStream *self, PyObject *noargs) {
HANDLE_TH_ERRORS
return PyBool_FromLong(self->cuda_stream.query());
END_HANDLE_TH_ERRORS
}
static PyObject * THCPStream_synchronize(THCPStream *self, PyObject *noargs) {
HANDLE_TH_ERRORS
{
pybind11::gil_scoped_release no_gil;
self->cuda_stream.synchronize();
}
Py_RETURN_NONE;
END_HANDLE_TH_ERRORS
}
static PyObject * THCPStream_eq(THCPStream *self, THCPStream *other) {
HANDLE_TH_ERRORS
return PyBool_FromLong(self->cuda_stream == other->cuda_stream);
END_HANDLE_TH_ERRORS
}
static struct PyMemberDef THCPStream_members[] = {
{(char*)"_cdata",
T_ULONGLONG, offsetof(THCPStream, cdata), READONLY, nullptr},
{nullptr}
};
static struct PyGetSetDef THCPStream_properties[] = {
{"device", (getter)THCPStream_get_device, nullptr, nullptr, nullptr},
{"cuda_stream",
(getter)THCPStream_get_cuda_stream, nullptr, nullptr, nullptr},
{"priority", (getter)THCPStream_get_priority, nullptr, nullptr, nullptr},
{nullptr}
};
static PyMethodDef THCPStream_methods[] = {
{(char*)"query", (PyCFunction)THCPStream_query, METH_NOARGS, nullptr},
{(char*)"synchronize",
(PyCFunction)THCPStream_synchronize, METH_NOARGS, nullptr},
{(char*)"priority_range",
(PyCFunction)(void(*)(void))THCPStream_priority_range, METH_STATIC | METH_NOARGS, nullptr},
{(char*)"__eq__", (PyCFunction)THCPStream_eq, METH_O, nullptr},
{nullptr}
};
PyTypeObject THCPStreamType = {
PyVarObject_HEAD_INIT(nullptr, 0)
"torch._C._CudaStreamBase", /* tp_name */
sizeof(THCPStream), /* tp_basicsize */
0, /* tp_itemsize */
(destructor)THCPStream_dealloc, /* tp_dealloc */
0, /* tp_vectorcall_offset */
0, /* tp_getattr */
0, /* tp_setattr */
0, /* tp_reserved */
0, /* tp_repr */
0, /* tp_as_number */
0, /* tp_as_sequence */
0, /* tp_as_mapping */
0, /* tp_hash */
0, /* tp_call */
0, /* tp_str */
0, /* tp_getattro */
0, /* tp_setattro */
0, /* tp_as_buffer */
Py_TPFLAGS_DEFAULT | Py_TPFLAGS_BASETYPE, /* tp_flags */
nullptr, /* tp_doc */
0, /* tp_traverse */
0, /* tp_clear */
0, /* tp_richcompare */
0, /* tp_weaklistoffset */
0, /* tp_iter */
0, /* tp_iternext */
THCPStream_methods, /* tp_methods */
THCPStream_members, /* tp_members */
THCPStream_properties, /* tp_getset */
0, /* tp_base */
0, /* tp_dict */
0, /* tp_descr_get */
0, /* tp_descr_set */
0, /* tp_dictoffset */
0, /* tp_init */
0, /* tp_alloc */
THCPStream_pynew, /* tp_new */
};
void THCPStream_init(PyObject *module)
{
THCPStreamClass = (PyObject*)&THCPStreamType;
if (PyType_Ready(&THCPStreamType) < 0) {
throw python_error();
}
Py_INCREF(&THCPStreamType);
if (PyModule_AddObject(
module, "_CudaStreamBase", (PyObject *)&THCPStreamType) < 0) {
throw python_error();
}
}
|
