ndarray.cc 10.9 KB
Newer Older
Minjie Wang's avatar
Minjie Wang committed
1
2
3
4
5
/*!
 *  Copyright (c) 2017 by Contributors
 * \file ndarray.cc
 * \brief NDArray container infratructure.
 */
6
#include <string.h>
Minjie Wang's avatar
Minjie Wang committed
7
8
9
10
11
12
13
14
15
#include <dmlc/logging.h>
#include <dgl/runtime/ndarray.h>
#include <dgl/runtime/c_runtime_api.h>
#include <dgl/runtime/device_api.h>
#include "runtime_base.h"

// deleter for arrays used by DLPack exporter
extern "C" void NDArrayDLPackDeleter(DLManagedTensor* tensor);

16
namespace dgl {
Minjie Wang's avatar
Minjie Wang committed
17
18
19
20
21
22
23
24
25
26
27
28
29
30
namespace runtime {

inline void VerifyDataType(DLDataType dtype) {
  CHECK_GE(dtype.lanes, 1);
  if (dtype.code == kDLFloat) {
    CHECK_EQ(dtype.bits % 8, 0);
  } else {
    CHECK_EQ(dtype.bits % 8, 0);
  }
  CHECK_EQ(dtype.bits & (dtype.bits - 1), 0);
}

inline size_t GetDataSize(const DLTensor& arr) {
  size_t size = 1;
31
  for (dgl_index_t i = 0; i < arr.ndim; ++i) {
Minjie Wang's avatar
Minjie Wang committed
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
    size *= arr.shape[i];
  }
  size *= (arr.dtype.bits * arr.dtype.lanes + 7) / 8;
  return size;
}

inline size_t GetDataAlignment(const DLTensor& arr) {
  size_t align = (arr.dtype.bits / 8) * arr.dtype.lanes;
  if (align < kAllocAlignment) return kAllocAlignment;
  return align;
}

struct NDArray::Internal {
  // Default deleter for the container
  static void DefaultDeleter(NDArray::Container* ptr) {
47
    using dgl::runtime::NDArray;
Minjie Wang's avatar
Minjie Wang committed
48
49
    if (ptr->manager_ctx != nullptr) {
      static_cast<NDArray::Container*>(ptr->manager_ctx)->DecRef();
50
51
52
53
#ifndef _WIN32
    } else if (ptr->mem) {
      ptr->mem = nullptr;
#endif  // _WIN32
Minjie Wang's avatar
Minjie Wang committed
54
    } else if (ptr->dl_tensor.data != nullptr) {
55
      dgl::runtime::DeviceAPI::Get(ptr->dl_tensor.ctx)->FreeDataSpace(
Minjie Wang's avatar
Minjie Wang committed
56
57
58
59
60
61
          ptr->dl_tensor.ctx, ptr->dl_tensor.data);
    }
    delete ptr;
  }
  // Deleter for NDArray converted from DLPack
  // This is used from data which is passed from external DLPack(DLManagedTensor)
62
  // that are not allocated inside of DGL.
Minjie Wang's avatar
Minjie Wang committed
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
  // This enables us to create NDArray from memory allocated by other
  // frameworks that are DLPack compatible
  static void DLPackDeleter(NDArray::Container* ptr) {
    DLManagedTensor* tensor = static_cast<DLManagedTensor*>(ptr->manager_ctx);
    if (tensor->deleter != nullptr) {
      (*tensor->deleter)(tensor);
    }
    delete ptr;
  }
  // Local create function which allocates tensor metadata
  // but does not allocate space for the data.
  static NDArray Create(std::vector<int64_t> shape,
                        DLDataType dtype,
                        DLContext ctx) {
    VerifyDataType(dtype);
    // critical zone
    NDArray::Container* data = new NDArray::Container();
    data->deleter = DefaultDeleter;
    NDArray ret(data);
    ret.data_ = data;
    // RAII now in effect
    // setup shape
    data->shape_ = std::move(shape);
    data->dl_tensor.shape = dmlc::BeginPtr(data->shape_);
    data->dl_tensor.ndim = static_cast<int>(data->shape_.size());
88
89
90
91
92
93
94
    // setup stride (this should be optional, but some framework
    //   does not support NULL stride and thus will crash the program).
    data->stride_.resize(data->dl_tensor.ndim, 1);
    for (int i = data->dl_tensor.ndim - 2; i >= 0; --i) {
      data->stride_[i] = data->shape_[i+1] * data->stride_[i+1];
    }
    data->dl_tensor.strides = dmlc::BeginPtr(data->stride_);
Minjie Wang's avatar
Minjie Wang committed
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
    // setup dtype
    data->dl_tensor.dtype = dtype;
    // setup ctx
    data->dl_tensor.ctx = ctx;
    return ret;
  }
  // Implementation of API function
  static DLTensor* MoveAsDLTensor(NDArray arr) {
    DLTensor* tensor = const_cast<DLTensor*>(arr.operator->());
    CHECK(reinterpret_cast<DLTensor*>(arr.data_) == tensor);
    arr.data_ = nullptr;
    return tensor;
  }
  // Container to DLManagedTensor
  static DLManagedTensor* ToDLPack(NDArray::Container* from) {
    CHECK(from != nullptr);
    DLManagedTensor* ret = new DLManagedTensor();
    ret->dl_tensor = from->dl_tensor;
    ret->manager_ctx = from;
    from->IncRef();
    ret->deleter = NDArrayDLPackDeleter;
    return ret;
  }
};

120
121
122
123
size_t NDArray::GetSize() const {
  return GetDataSize(data_->dl_tensor);
}

124
125
126
127
128
129
130
131
132
133
134
135
bool NDArray::IsContiguous() const {
  CHECK(data_ != nullptr);
  if (data_->dl_tensor.strides == nullptr)
    return true;
  for (int i = 0; i < data_->dl_tensor.ndim - 1; ++i) {
    if (data_->dl_tensor.strides[i] !=
        data_->dl_tensor.shape[i+1] * data_->dl_tensor.strides[i+1])
      return false;
  }
  return data_->dl_tensor.strides[data_->dl_tensor.ndim - 1] == 1;
}

Minjie Wang's avatar
Minjie Wang committed
136
NDArray NDArray::CreateView(std::vector<int64_t> shape,
137
138
                            DLDataType dtype,
                            int64_t offset) {
Minjie Wang's avatar
Minjie Wang committed
139
  CHECK(data_ != nullptr);
140
  CHECK(IsContiguous()) << "Can only create view for compact tensor";
Minjie Wang's avatar
Minjie Wang committed
141
142
143
144
145
146
147
148
149
150
  NDArray ret = Internal::Create(shape, dtype, data_->dl_tensor.ctx);
  ret.data_->dl_tensor.byte_offset =
      this->data_->dl_tensor.byte_offset;
  size_t curr_size = GetDataSize(this->data_->dl_tensor);
  size_t view_size = GetDataSize(ret.data_->dl_tensor);
  CHECK_LE(view_size, curr_size)
      << "Tries to create a view that has bigger memory than current one";
  // increase ref count
  this->data_->IncRef();
  ret.data_->manager_ctx = this->data_;
151
152
  ret.data_->dl_tensor.data =
    static_cast<char*>(this->data_->dl_tensor.data) + offset;
Minjie Wang's avatar
Minjie Wang committed
153
154
155
156
157
158
159
  return ret;
}

DLManagedTensor* NDArray::ToDLPack() const {
  return Internal::ToDLPack(data_);
}

160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
NDArray NDArray::EmptyShared(const std::string &name,
                       std::vector<int64_t> shape,
                       DLDataType dtype,
                       DLContext ctx, bool is_create) {
  NDArray ret = Internal::Create(shape, dtype, ctx);
  // setup memory content
  size_t size = GetDataSize(ret.data_->dl_tensor);
#ifndef _WIN32
  auto mem = std::make_shared<SharedMemory>(name);
  if (is_create) {
    ret.data_->dl_tensor.data = mem->create_new(size);
  } else {
    ret.data_->dl_tensor.data = mem->open(size);
  }

  ret.data_->mem = mem;
#else
  LOG(FATAL) << "Windows doesn't support NDArray with shared memory";
#endif  // _WIN32
  return ret;
}

Minjie Wang's avatar
Minjie Wang committed
182
NDArray NDArray::Empty(std::vector<int64_t> shape,
183
184
                       DLDataType dtype,
                       DLContext ctx) {
Minjie Wang's avatar
Minjie Wang committed
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
  NDArray ret = Internal::Create(shape, dtype, ctx);
  // setup memory content
  size_t size = GetDataSize(ret.data_->dl_tensor);
  size_t alignment = GetDataAlignment(ret.data_->dl_tensor);
  ret.data_->dl_tensor.data =
      DeviceAPI::Get(ret->ctx)->AllocDataSpace(
          ret->ctx, size, alignment, ret->dtype);
  return ret;
}

NDArray NDArray::FromDLPack(DLManagedTensor* tensor) {
  NDArray::Container* data = new NDArray::Container();
  data->deleter = Internal::DLPackDeleter;
  data->manager_ctx = tensor;
  data->dl_tensor = tensor->dl_tensor;
  return NDArray(data);
}

void NDArray::CopyFromTo(DLTensor* from,
                         DLTensor* to,
205
                         DGLStreamHandle stream) {
Minjie Wang's avatar
Minjie Wang committed
206
207
208
  size_t from_size = GetDataSize(*from);
  size_t to_size = GetDataSize(*to);
  CHECK_EQ(from_size, to_size)
209
    << "DGLArrayCopyFromTo: The size must exactly match";
Minjie Wang's avatar
Minjie Wang committed
210
211
212
213
214
215
216
217

  CHECK(from->ctx.device_type == to->ctx.device_type
        || from->ctx.device_type == kDLCPU
        || to->ctx.device_type == kDLCPU)
    << "Can not copy across different ctx types directly";

  // Use the context that is *not* a cpu context to get the correct device
  // api manager.
218
  DGLContext ctx = from->ctx.device_type != kDLCPU ? from->ctx : to->ctx;
Minjie Wang's avatar
Minjie Wang committed
219
220
221
222
223
224
225
226

  DeviceAPI::Get(ctx)->CopyDataFromTo(
    from->data, static_cast<size_t>(from->byte_offset),
    to->data, static_cast<size_t>(to->byte_offset),
    from_size, from->ctx, to->ctx, from->dtype, stream);
}

}  // namespace runtime
227
}  // namespace dgl
Minjie Wang's avatar
Minjie Wang committed
228

229
using namespace dgl::runtime;
Minjie Wang's avatar
Minjie Wang committed
230
231
232
233
234
235

void NDArrayDLPackDeleter(DLManagedTensor* tensor) {
  static_cast<NDArray::Container*>(tensor->manager_ctx)->DecRef();
  delete tensor;
}

236
int DGLArrayAlloc(const dgl_index_t* shape,
Minjie Wang's avatar
Minjie Wang committed
237
238
239
240
241
242
                  int ndim,
                  int dtype_code,
                  int dtype_bits,
                  int dtype_lanes,
                  int device_type,
                  int device_id,
243
                  DGLArrayHandle* out) {
Minjie Wang's avatar
Minjie Wang committed
244
245
246
247
248
249
250
251
252
253
254
255
256
  API_BEGIN();
  DLDataType dtype;
  dtype.code = static_cast<uint8_t>(dtype_code);
  dtype.bits = static_cast<uint8_t>(dtype_bits);
  dtype.lanes = static_cast<uint16_t>(dtype_lanes);
  DLContext ctx;
  ctx.device_type = static_cast<DLDeviceType>(device_type);
  ctx.device_id = device_id;
  *out = NDArray::Internal::MoveAsDLTensor(
      NDArray::Empty(std::vector<int64_t>(shape, shape + ndim), dtype, ctx));
  API_END();
}

257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
int DGLArrayAllocSharedMem(const char *mem_name,
                           const dgl_index_t *shape,
                           int ndim,
                           int dtype_code,
                           int dtype_bits,
                           int dtype_lanes,
                           bool is_create,
                           DGLArrayHandle* out) {
  API_BEGIN();
  DLDataType dtype;
  dtype.code = static_cast<uint8_t>(dtype_code);
  dtype.bits = static_cast<uint8_t>(dtype_bits);
  dtype.lanes = static_cast<uint16_t>(dtype_lanes);
  std::vector<int64_t> shape_vec(shape, shape + ndim);
  NDArray arr = NDArray::EmptyShared(mem_name, shape_vec, dtype,
                                     DLContext{kDLCPU, 0}, is_create);
  *out = NDArray::Internal::MoveAsDLTensor(arr);
  API_END();
}

277
int DGLArrayFree(DGLArrayHandle handle) {
Minjie Wang's avatar
Minjie Wang committed
278
279
280
281
282
  API_BEGIN();
  reinterpret_cast<NDArray::Container*>(handle)->DecRef();
  API_END();
}

283
284
285
int DGLArrayCopyFromTo(DGLArrayHandle from,
                       DGLArrayHandle to,
                       DGLStreamHandle stream) {
Minjie Wang's avatar
Minjie Wang committed
286
287
288
289
290
  API_BEGIN();
  NDArray::CopyFromTo(from, to, stream);
  API_END();
}

291
292
int DGLArrayFromDLPack(DLManagedTensor* from,
                       DGLArrayHandle* out) {
Minjie Wang's avatar
Minjie Wang committed
293
294
295
296
297
  API_BEGIN();
  *out = NDArray::Internal::MoveAsDLTensor(NDArray::FromDLPack(from));
  API_END();
}

298
int DGLArrayToDLPack(DGLArrayHandle from,
Minjie Wang's avatar
Minjie Wang committed
299
300
301
302
303
304
                     DLManagedTensor** out) {
  API_BEGIN();
  *out = NDArray::Internal::ToDLPack(reinterpret_cast<NDArray::Container*>(from));
  API_END();
}

305
void DGLDLManagedTensorCallDeleter(DLManagedTensor* dltensor) {
Minjie Wang's avatar
Minjie Wang committed
306
307
308
  (*(dltensor->deleter))(dltensor);
}

309
int DGLArrayCopyFromBytes(DGLArrayHandle handle,
Minjie Wang's avatar
Minjie Wang committed
310
311
312
                          void* data,
                          size_t nbytes) {
  API_BEGIN();
313
  DGLContext cpu_ctx;
Minjie Wang's avatar
Minjie Wang committed
314
315
316
317
  cpu_ctx.device_type = kDLCPU;
  cpu_ctx.device_id = 0;
  size_t arr_size = GetDataSize(*handle);
  CHECK_EQ(arr_size, nbytes)
318
      << "DGLArrayCopyFromBytes: size mismatch";
Minjie Wang's avatar
Minjie Wang committed
319
320
321
322
323
324
325
  DeviceAPI::Get(handle->ctx)->CopyDataFromTo(
      data, 0,
      handle->data, static_cast<size_t>(handle->byte_offset),
      nbytes, cpu_ctx, handle->ctx, handle->dtype, nullptr);
  API_END();
}

326
int DGLArrayCopyToBytes(DGLArrayHandle handle,
Minjie Wang's avatar
Minjie Wang committed
327
328
329
                        void* data,
                        size_t nbytes) {
  API_BEGIN();
330
  DGLContext cpu_ctx;
Minjie Wang's avatar
Minjie Wang committed
331
332
333
334
  cpu_ctx.device_type = kDLCPU;
  cpu_ctx.device_id = 0;
  size_t arr_size = GetDataSize(*handle);
  CHECK_EQ(arr_size, nbytes)
335
      << "DGLArrayCopyToBytes: size mismatch";
Minjie Wang's avatar
Minjie Wang committed
336
337
338
339
340
341
  DeviceAPI::Get(handle->ctx)->CopyDataFromTo(
      handle->data, static_cast<size_t>(handle->byte_offset),
      data, 0,
      nbytes, handle->ctx, cpu_ctx, handle->dtype, nullptr);
  API_END();
}