op_registry.h 12.6 KB
Newer Older
1 2
#pragma once

3
#include <algorithm>
4
#include <atomic>
Y
Yu Yang 已提交
5
#include <type_traits>
6 7
#include <unordered_map>
#include <unordered_set>
Q
Qiao Longfei 已提交
8
#include "paddle/framework/attr_checker.h"
9 10
#include "paddle/framework/op_desc.pb.h"
#include "paddle/framework/op_proto.pb.h"
Q
Qiao Longfei 已提交
11
#include "paddle/framework/operator.h"
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

namespace paddle {
namespace framework {

// helper class to set attribute type
struct AttrTypeHelper {
  template <typename T>
  static void SetAttrType(AttrProto* attr);

  static Attribute GetAttrValue(const AttrDesc& attr_desc) {
    switch (attr_desc.type()) {
      case paddle::framework::AttrType::INT: {
        return attr_desc.i();
      }
      case paddle::framework::AttrType::FLOAT: {
        return attr_desc.f();
      }
      case paddle::framework::AttrType::STRING: {
        return attr_desc.s();
      }
      case paddle::framework::AttrType::INTS: {
        std::vector<int> val(attr_desc.ints_size());
        for (int i = 0; i < attr_desc.ints_size(); ++i) {
          val[i] = attr_desc.ints(i);
        }
        return val;
      }
      case paddle::framework::AttrType::FLOATS: {
        std::vector<float> val(attr_desc.floats_size());
        for (int i = 0; i < attr_desc.floats_size(); ++i) {
          val[i] = attr_desc.floats(i);
        }
        return val;
      }
      case paddle::framework::AttrType::STRINGS: {
        std::vector<std::string> val(attr_desc.strings_size());
        for (int i = 0; i < attr_desc.strings_size(); ++i) {
          val[i] = attr_desc.strings(i);
        }
        return val;
      }
    }
    PADDLE_ENFORCE(false, "Unknown OpDesc::AttrDesc::type !");
    return boost::blank();
  }
};

// this class not only make proto but also init attribute checkers.
class OpProtoAndCheckerMaker {
 public:
  OpProtoAndCheckerMaker(OpProto* proto, OpAttrChecker* op_checker)
      : proto_(proto), op_checker_(op_checker) {}

65 66 67 68 69 70 71 72
  ~OpProtoAndCheckerMaker() {
    PADDLE_ENFORCE(validated_, "should call Validate after build");
  }

  void Validate() {
    validated_ = true;
    CheckNoDuplicatedInOutAttrs();
  }
73

74
 protected:
75 76
  void AddInput(const std::string& name, const std::string& comment,
                bool multiple = false) {
77
    auto input = proto_->mutable_inputs()->Add();
78 79
    *input->mutable_name() = name;
    *input->mutable_comment() = comment;
80 81 82 83 84 85 86 87
    input->set_multiple(multiple);
    if (multiple) {
      SetHasMultipleInput();
    }
  }

  void AddInputs(const std::string& name, const std::string& comment) {
    AddInput(name, comment, true);
88 89
  }

90 91
  void AddOutput(const std::string& name, const std::string& comment,
                 bool temporary = false, bool multiple = false) {
92
    auto output = proto_->mutable_outputs()->Add();
93 94
    *output->mutable_name() = name;
    *output->mutable_comment() = comment;
95 96 97 98 99 100 101 102 103 104 105 106 107
    output->set_multiple(multiple);
    if (multiple) {
      SetHasMultipleOutput();
    }
    output->set_temporary(temporary);
    if (temporary) {
      SetHasTemporaryOutput();
    }
  }

  void AddOutputs(const std::string& name, const std::string& comment,
                  bool temporary = false) {
    AddOutput(name, comment, temporary, true);
108 109 110 111
  }

  template <typename T>
  TypedAttrChecker<T>& AddAttr(const std::string& name,
112 113
                               const std::string& comment,
                               bool generated = false) {
114
    auto attr = proto_->mutable_attrs()->Add();
115 116
    *attr->mutable_name() = name;
    *attr->mutable_comment() = comment;
117
    attr->set_generated(generated);
118 119 120 121 122 123 124 125
    AttrTypeHelper::SetAttrType<T>(attr);
    return op_checker_->AddAttrChecker<T>(name);
  }

  void AddComment(const std::string& comment) {
    *(proto_->mutable_comment()) = comment;
  }

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
 private:
  void SetHasMultiple(const std::string& in_out, bool* flag) {
    if (!*flag) {
      AddAttr<std::vector<int>>(in_out + "_format",
                                "The multiple index of " + in_out +
                                    "\n"
                                    R"DOC(
This attribute is used by Paddle core framework. Paddle's Op support each input
or output could be a list of variable. This attribute is used to show how that
list organized.

e.g.
  input = ["a", "b", "c", "d", "e", "f"]
  input_format = [0, 4, 5, 6]

means
  The number of all input variables this op is six, and they are segmented into
  three inputs.

  The first input is input[0:4], second is input[4:5], third is input[5:6].
)DOC",
                                /*generated*/ true);
      *flag = true;
    }
  }

  void SetHasMultipleInput() { SetHasMultiple("input", &has_multiple_input_); }
  void SetHasMultipleOutput() {
    SetHasMultiple("output", &has_multiple_output_);
  }

  void SetHasTemporaryOutput() {
    if (!has_temporary_output_) {
      AddAttr<std::vector<int>>("temporary_index",
                                R"DOC(The temporary index of output.

Not all output of Paddle Op is used by user. For faster computation, each op
could output some its internal state to other op, other op could take that
output to make compute faster.

Add a mark to which output is temporary is helpful for future optimization.
)DOC",
                                /*generated*/ true)
          .SetDefault(std::vector<int>());
      has_temporary_output_ = true;
    }
  }

174
  void CheckNoDuplicatedInOutAttrs() {
175
    std::unordered_set<std::string> names;
176 177 178 179
    auto checker = [&](const std::string& name) {
      PADDLE_ENFORCE(!names.count(name), "[%s] is duplicated", name);
      names.insert(name);
    };
180
    for (auto& attr : proto_->attrs()) {
181 182 183 184 185 186 187
      checker(attr.name());
    }
    for (auto& input : proto_->inputs()) {
      checker(input.name());
    }
    for (auto& output : proto_->outputs()) {
      checker(output.name());
188 189 190
    }
  }

191 192
  OpProto* proto_;
  OpAttrChecker* op_checker_;
193
  bool validated_{false};
194 195 196
  bool has_multiple_input_{false};
  bool has_multiple_output_{false};
  bool has_temporary_output_{false};
197 198 199
};

class OpRegistry {
Q
Qiao Longfei 已提交
200
  using OpCreator = std::function<OperatorBase*()>;
201 202 203 204

 public:
  template <typename OpType, typename ProtoMakerType>
  static void RegisterOp(const std::string& op_type) {
205 206 207
    creators()[op_type] = [] { return new OpType; };
    OpProto& op_proto = protos()[op_type];
    OpAttrChecker& op_checker = op_checkers()[op_type];
208 209
    auto maker = ProtoMakerType(&op_proto, &op_checker);
    maker.Validate();
Y
Yu Yang 已提交
210 211 212 213 214
    *op_proto.mutable_type() = op_type;
    PADDLE_ENFORCE(
        op_proto.IsInitialized(),
        "Fail to initialize %s's OpProto, because %s is not initialized",
        op_type, op_proto.InitializationErrorString());
215 216
  }

Q
Qiao Longfei 已提交
217
  static OperatorPtr CreateOp(const OpDesc& op_desc) {
218
    //! Create a OpPtr by type.
219
    std::string op_type = op_desc.type();
Q
Qiao Longfei 已提交
220
    OperatorPtr op(creators().at(op_type)());
221 222
    //! Fill op's data member. Not use constructor because it will be noising
    //! for Op developer.
Y
Yan Chunwei 已提交
223
    const OpProto& op_proto = protos().at(op_type);
Q
Qiao Longfei 已提交
224
    op->type_ = op_desc.type();
225
    // set op's inputs_ from desc.
226 227 228
    op->inputs_.reserve((size_t)op_desc.inputs_size());
    std::copy(op_desc.inputs().begin(), op_desc.inputs().end(),
              std::back_inserter(op->inputs_));
Y
Yan Chunwei 已提交
229
    // set op's outputs_ from desc.
230 231 232
    op->outputs_.reserve((size_t)op_desc.outputs_size());
    std::copy(op_desc.outputs().begin(), op_desc.outputs().end(),
              std::back_inserter(op->outputs_));
233 234

    //! Fill attrs, and validate attrs.
235
    for (auto& attr : op_desc.attrs()) {
Q
Qiao Longfei 已提交
236
      op->attrs_[attr.name()] = AttrTypeHelper::GetAttrValue(attr);
237
    }
Q
Qiao Longfei 已提交
238
    op_checkers().at(op_type).Check(op->attrs_);
239 240

    //! Convert Temporary variable name to an unique variable name.
241
    GenerateTempVariableName(op.get());
242

Y
Yan Chunwei 已提交
243 244
    // set argument offsets stored in op.
    CreateInOutOffsetMap(op, op_proto);
245 246
    //! Other op's custom Init for a complex Op. For simple Op, the Init
    //! method do nothing.
Q
Qiao Longfei 已提交
247
    op->Init();
248 249 250
    return op;
  }

Y
Yan Chunwei 已提交
251 252 253 254 255
  // init op.in_out_idxs_ to accelerate argument's offset lookup.
  static void CreateInOutOffsetMap(OperatorPtr op, const OpProto& proto) {
    op->CreateInOutOffsetMap(proto);
  }

Y
Yu Yang 已提交
256 257 258 259 260
  static std::unordered_map<std::string, OpProto>& protos() {
    static std::unordered_map<std::string, OpProto> protos_;
    return protos_;
  };

261
 private:
262
  static void GenerateTempVariableName(OperatorBase* op) {
263 264 265
    static std::atomic<size_t> gUniqId(0UL);
    for (auto& outname : op->outputs_) {
      if (outname == OperatorBase::TMP_VAR_NAME()) {
266
        outname += op->type_;
267 268 269 270 271 272
        outname += "@";
        outname += std::to_string(gUniqId.fetch_add(1));
      }
    }
  }

273 274 275 276
  static std::unordered_map<std::string, OpCreator>& creators() {
    static std::unordered_map<std::string, OpCreator> creators_;
    return creators_;
  }
277

278 279 280 281 282
  static std::unordered_map<std::string, OpAttrChecker>& op_checkers() {
    static std::unordered_map<std::string, OpAttrChecker> op_checkers_;
    return op_checkers_;
  };
};
283 284 285 286

template <typename OpType, typename ProtoMakerType>
class OpRegisterHelper {
 public:
Y
Yu Yang 已提交
287
  OpRegisterHelper(const char* op_type) {
288 289 290 291
    OpRegistry::RegisterOp<OpType, ProtoMakerType>(op_type);
  }
};

292 293 294
/**
 * check if MACRO is used in GLOBAL NAMESPACE.
 */
Y
Yu Yang 已提交
295 296 297 298 299 300
#define STATIC_ASSERT_GLOBAL_NAMESPACE(uniq_name, msg)                        \
  struct __test_global_namespace_##uniq_name##__ {};                          \
  static_assert(std::is_same<::__test_global_namespace_##uniq_name##__,       \
                             __test_global_namespace_##uniq_name##__>::value, \
                msg)

301 302 303
/**
 * Macro to Register Operator.
 */
Y
Yu Yang 已提交
304 305 306 307 308 309 310
#define REGISTER_OP(__op_type, __op_class, __op_maker_class)                 \
  STATIC_ASSERT_GLOBAL_NAMESPACE(__reg_op__##__op_type,                      \
                                 "REGISTER_OP must be in global namespace"); \
  static ::paddle::framework::OpRegisterHelper<__op_class, __op_maker_class> \
      __op_register_##__op_type##__(#__op_type);                             \
  int __op_register_##__op_type##_handle__() { return 0; }

311 312 313
/**
 * Macro to Register OperatorKernel.
 */
314
#define REGISTER_OP_KERNEL(type, DEVICE_TYPE, PlaceType, ...)             \
Y
Yu Yang 已提交
315
  STATIC_ASSERT_GLOBAL_NAMESPACE(                                         \
316
      __reg_op_kernel_##type##_##DEVICE_TYPE##__,                         \
Y
Yu Yang 已提交
317 318 319 320 321 322
      "REGISTER_OP_KERNEL must be in global namespace");                  \
  struct __op_kernel_register__##type##__ {                               \
    __op_kernel_register__##type##__() {                                  \
      ::paddle::framework::OperatorWithKernel::OpKernelKey key;           \
      key.place_ = PlaceType();                                           \
      ::paddle::framework::OperatorWithKernel::AllOpKernels()[#type][key] \
323
          .reset(new __VA_ARGS__());                                      \
Y
Yu Yang 已提交
324 325 326
    }                                                                     \
  };                                                                      \
  static __op_kernel_register__##type##__ __reg_kernel_##type##__;        \
327
  int __op_kernel_register_##type##_handle_##DEVICE_TYPE##__() { return 0; }
Y
Yu Yang 已提交
328

329 330 331
// (type, KernelType)
#define REGISTER_OP_GPU_KERNEL(type, ...) \
  REGISTER_OP_KERNEL(type, GPU, ::paddle::platform::GPUPlace, __VA_ARGS__)
Y
Yu Yang 已提交
332

333 334 335
// (type, KernelType)
#define REGISTER_OP_CPU_KERNEL(type, ...) \
  REGISTER_OP_KERNEL(type, CPU, ::paddle::platform::CPUPlace, __VA_ARGS__)
Y
Yu Yang 已提交
336

337 338 339 340
/**
 * Macro to mark what Operator and Kernel we will use and tell the compiler to
 * link them into target.
 */
Y
Yu Yang 已提交
341 342 343 344 345 346 347 348
#define USE_OP_WITHOUT_KERNEL(op_type)                      \
  STATIC_ASSERT_GLOBAL_NAMESPACE(                           \
      __use_op_without_kernel_##op_type,                    \
      "USE_OP_WITHOUT_KERNEL must be in global namespace"); \
  extern int __op_register_##op_type##_handle__();          \
  static int __use_op_ptr_##op_type##_without_kernel__      \
      __attribute__((unused)) = __op_register_##op_type##_handle__()

Y
Yu Yang 已提交
349 350 351 352 353 354 355 356
#define USE_OP_KERNEL(op_type, DEVICE_TYPE)                               \
  STATIC_ASSERT_GLOBAL_NAMESPACE(                                         \
      __use_op_kernel_##op_type##_##DEVICE_TYPE##__,                      \
      "USE_OP_KERNEL must be in global namespace");                       \
  extern int __op_kernel_register_##op_type##_handle_##DEVICE_TYPE##__(); \
  static int __use_op_ptr_##op_type##_##DEVICE_TYPE##_kernel__            \
      __attribute__((unused)) =                                           \
          __op_kernel_register_##op_type##_handle_##DEVICE_TYPE##__()
Y
Yu Yang 已提交
357

358 359
// use Operator with only cpu kernel.
#define USE_OP_CPU(op_type)       \
Y
Yu Yang 已提交
360
  USE_OP_WITHOUT_KERNEL(op_type); \
361
  USE_OP_KERNEL(op_type, CPU)
Y
Yu Yang 已提交
362

363 364
#ifdef PADDLE_ONLY_CPU
#define USE_OP(op_type) USE_OP_CPU(op_type)
Y
Yu Yang 已提交
365
#else
366 367
#define USE_OP(op_type) \
  USE_OP_CPU(op_type);  \
Y
Yu Yang 已提交
368 369
  USE_OP_KERNEL(op_type, GPU)
#endif
370 371 372

}  // namespace framework
}  // namespace paddle