Feature/copytensor (#5455)

* "make global tensor function independently"

* "replace functor"

* "fix inline template error"

* "fix tensor array with CopyFrom"

* "fix other case use CopyFrom"

* "move the op interface hardly"

* "fix operators"

* "fix typo"

* "delete dynamic recurrent rnn and fix gru_unit in debugmode"

* "fix unique_ptr copy"

* "fix cuda copy"

* "fix namespace error"

* "removed nccl python test"

* "fix include error"

* "fix typo"

* fix copy util test

Author: dzhwinter

paddle/framework/CMakeLists.txt

@@ -6,7 +6,10 @@ cc_test(ddim_test SRCS ddim_test.cc DEPS ddim)
 nv_test(dim_test SRCS dim_test.cu DEPS ddim)
 
 cc_library(tensor SRCS tensor.cc DEPS ddim place paddle_memory device_context)
+
 cc_test(tensor_test SRCS tensor_test.cc DEPS tensor)
+cc_test(tensor_util_test SRCS tensor_util_test.cc DEPS tensor)
+
 cc_test(eigen_test SRCS eigen_test.cc DEPS tensor)
 
 cc_library(lod_tensor SRCS lod_tensor.cc DEPS ddim place tensor framework_proto)
@@ -51,10 +54,6 @@ cc_library(executor SRCS executor.cc DEPS op_registry device_context scope frame
 
 cc_library(prune SRCS prune.cc DEPS framework_proto)
 cc_test(prune_test SRCS prune_test.cc DEPS op_info prune recurrent_op device_context)
-
-cc_library(tensor_array SRCS tensor_array.cc DEPS lod_tensor)
-cc_test(tensor_array_test SRCS tensor_array_test.cc DEPS tensor_array place)
-
 cc_test(var_type_inference_test SRCS var_type_inference_test.cc DEPS op_registry
         proto_desc)
 cc_library(selected_rows SRCS selected_rows.cc DEPS tensor)

paddle/framework/backward.cc

@@ -22,7 +22,6 @@
 
 #include "paddle/framework/block_desc.h"
 #include "paddle/framework/op_registry.h"
-#include "paddle/operators/dynamic_recurrent_op.h"
 #include "paddle/operators/net_op.h"
 
 namespace paddle {
@@ -218,21 +217,6 @@ static std::unique_ptr<OperatorBase> BackwardRecursive(
                      return false;
                    });
 
-    // process recurrent gradient op as a special operator.
-    if (forwardOp.Type() == "dynamic_recurrent") {
-      // NOTE clean up cycle call somewhere (RNN's stepnet constains itself),
-      // or this will result in infinite loop.
-      const auto& rnnop =
-          *static_cast<const operators::DynamicRecurrentOp*>(&forwardOp);
-      auto rnn_grad_op =
-          static_cast<operators::DynamicRecurrentGradientOp*>(grad_op.get());
-      const auto& stepnet_op =
-          *static_cast<const OperatorBase*>(&rnnop.rnn.GetStepUnit());
-      // create stepnet's gradient op
-      rnn_grad_op->rnn.SetStepUnit(
-          BackwardRecursive(stepnet_op, no_grad_names, grad_to_var, uniq_id));
-    }
-
     if (net->ops_.empty()) {  // Current no aux op is added to network
       return grad_op;
     }

paddle/framework/lod_tensor.h

@@ -24,6 +24,7 @@
 #include <glog/logging.h>
 #include "paddle/framework/ddim.h"
 #include "paddle/framework/tensor.h"
+#include "paddle/framework/tensor_util.h"
 #include "paddle/platform/enforce.h"
 #include "paddle/platform/place.h"
 
@@ -175,9 +176,9 @@ LoDTensor LodExpand(const LoDTensor& source, const LoD& lod, size_t level,
   PADDLE_ENFORCE_EQ(num_instances, lod_level.size() - 1);
   for (size_t ins = 0; ins < num_instances; ins++) {
     for (size_t elem = lod_level[ins]; elem < lod_level[ins + 1]; elem++) {
-      tensor.Slice(elem, elem + 1)
-          .CopyFrom(source.Slice(ins, ins + 1), platform::CPUPlace(),
-                    platform::CPUDeviceContext());
+      auto slice = tensor.Slice(elem, elem + 1);
+      CopyFrom(source.Slice(ins, ins + 1), platform::CPUPlace(),
+               platform::CPUDeviceContext(), &slice);
     }
   }
   return tensor;

paddle/framework/tensor.h

@@ -89,34 +89,6 @@ class Tensor {
   /*! The internal of two tensors share the same memory block. */
   inline Tensor& ShareDataWith(const Tensor& src);
 
-  /**
-   * @brief   Copy the content of external tensor to a new place.
-   *
-   * @param[in] src        The external tensor.
-   * @param[in] dst_place  The dst place.
-   * @param[in] ctx        The device context contains device resources.
-   *
-   * @note    CopyFrom supports CPU <-> GPU, GPU <-> GPU.
-   */
-  // TODO(qijun): https://github.com/PaddlePaddle/Paddle/issues/4647
-  // Remove `CopyFrom` and `CopyFromVector` from Tensor interface
-  // and make them global functions
-  inline void CopyFrom(const Tensor& src, const platform::Place& dst_place,
-                       const platform::DeviceContext& ctx);
-
-  /**
-   * @brief   Copy the content of an external vector to a tensor.
-   *
-   * @param[in] src        The external tensor.
-   * @param[in] ctx        The device context contains device resources.
-   *
-   * * @note    CopyFromVector assumes that the tensor has been resized
-   *            before invoking.
-   */
-  template <typename T>
-  inline void CopyFromVector(const std::vector<T>& src,
-                             const platform::DeviceContext& ctx);
-
   /**
    * @brief  Return a sub-tensor of the given tensor.
    *
@@ -141,7 +113,6 @@ class Tensor {
 
   size_t memory_size() const;
 
- private:
   inline void check_memory_size() const;
 
  private: