Change LoDTensorArray to DenseTensorArray (#6946)

Author: co63oc

docs/api/paddle/static/nn/while_loop_cn.rst

@@ -16,8 +16,8 @@ ____________________________________
 :::::::::
 
     - **cond** (callable) - 返回 boolean 类型 Tensor 的可调用函数，用以判断循环是否继续执行。``cond`` 的参数和 ``loop_vars`` 相对应。
-    - **body** (callable) - 循环执行的结构体。其返回一个包含 tensor 或 LoDTensorArray 的列表或元组，且这些 tensor 或 LoDTensorArray 的长度，结构，类型和 ``loop_vars`` 中的相同。且``body`` 的参数与 ``loop_vars`` 相对应。
-    - **loop_vars** (list|tuple) - 包含 tensor 或 LoDTensorArray 的列表或是元组，将其传入至 ``cond`` 和 ``body`` 中，得到循环条件和输出值。
+    - **body** (callable) - 循环执行的结构体。其返回一个包含 tensor 或 DenseTensorArray 的列表或元组，且这些 tensor 或 DenseTensorArray 的长度，结构，类型和 ``loop_vars`` 中的相同。且``body`` 的参数与 ``loop_vars`` 相对应。
+    - **loop_vars** (list|tuple) - 包含 tensor 或 DenseTensorArray 的列表或是元组，将其传入至 ``cond`` 和 ``body`` 中，得到循环条件和输出值。
     - **is_test** (bool，可选) - 用于表明是否在测试阶段执行，默认值为 False。
     - **name** (str，可选) - 具体用法请参见 :ref:`api_guide_Name`，一般无需设置，默认值为 None。
 

docs/design/concepts/python_data_feeding.md

@@ -6,7 +6,7 @@ In the former implementation of Paddle Fluid, there are two ways to feed data:
 
 - Feed data directly using `DataFeeder.feed()` in Python codes. It is more flexible than the first way. Many kinds of preprocessing steps can be performed before feeding using Python or any other languages, instead of adding many uncommon `operators` in C++ side. But this method is less efficient: the program cannot read the next mini-batch data before `Executor::Run()` ends. Moreover, `decorated_readers` such as `double_buffer_reader` cannot be used for better performance.
 
-In this document, we design a Python Data Feeding process combining the efficiency of the first way and the flexibility of the second way. A data queue `LoDTensorBlockingQueue` is designed to be shared by the Python and C++ side, while `LoDTensorArray` is pushed into the queue in Python side and `reader_op` in C++ side reads out the data from the queue.
+In this document, we design a Python Data Feeding process combining the efficiency of the first way and the flexibility of the second way. A data queue `LoDTensorBlockingQueue` is designed to be shared by the Python and C++ side, while `DenseTensorArray` is pushed into the queue in Python side and `reader_op` in C++ side reads out the data from the queue.
 
 
 ## Design of LoDTensorBlockingQueue

docs/design/concurrent/channel.md

@@ -41,7 +41,7 @@ fluid.channel_close(ch)
 ### Send data to a channel
 
 Sends a variable to a channel.  Currently, variables of dtype `LoDTensor`,
-`LoDRankTable`, `LoDTensorArray`, `SelectedRows`, `ReaderHolder`, and
+`LoDRankTable`, `DenseTensorArray`, `SelectedRows`, `ReaderHolder`, and
 `ChannelHolder` are supported.
 
 By default, the data of the Variable is moved from the sender to the receiver,

docs/guides/jit/case_analysis_cn.md

@@ -344,12 +344,12 @@ out = convert_ifelse(paddle.mean(x) > 5.0, true_fn_0, false_fn_0, (x,), (x,), (o
 ```
 
 
-### 8.1 list 与 LoDTensorArray
+### 8.1 list 与 DenseTensorArray
 
 当控制流中，出现了 ``list.append`` 类似语法时，情况会有一点点特殊。
 
 Paddle 框架中的 ``cond_op`` 和 [``while_loop``](https://www.paddlepaddle.org.cn/documentation/docs/zh/api/paddle/static/nn/while_loop_cn.html#while-loop) 对输入和返回类型有一个要求：
-> 输入或者返回类型必须是：LoDTensor 或者 LoDTensorArray <br><br>
+> 输入或者返回类型必须是：LoDTensor 或者 DenseTensorArray <br><br>
 > 即：不支持其他非 LoDTensor 类型
 
 因此控制流中类似：
@@ -370,7 +370,7 @@ def forward(self, x)：
 
 def forward(x):
     bs = paddle.shape(x)[0]    # <---- bs 是个静态图 Variable, shape = (1, )
-    outs = paddle.tensor.create_array(dtype='float32')    # <--- list 转为 LoDTensorArray
+    outs = paddle.tensor.create_array(dtype='float32')    # <--- list 转为 DenseTensorArray
     i = 0
 
     def for_loop_condition_0(outs, bs, i, x):
@@ -391,8 +391,8 @@ def forward(x):
 
 关于 控制流中包含 ``list`` 相关操作的几点说明：
 
-+ **并非所有**的 list 都会转为 ``LoDTensorArray``
-> 只有在此控制流语句是依赖 ``Tensor`` 时，才会触发 ``list`` &rarr; ``LoDTensorArray`` 的转换
++ **并非所有**的 list 都会转为 ``DenseTensorArray``
+> 只有在此控制流语句是依赖 ``Tensor`` 时，才会触发 ``list`` &rarr; ``DenseTensorArray`` 的转换
 
 + 暂不支持依赖 Tensor 的控制流中，使用多层嵌套的 ``list.append`` 操作
 
@@ -407,7 +407,7 @@ def forward(x):
         return outs
     ```
 
-> 因为框架底层的 ``LoDTensorArray = std::vector< LoDTensor >`` ，不支持两层以上 ``vector`` 嵌套
+> 因为框架底层的 ``DenseTensorArray = std::vector< LoDTensor >`` ，不支持两层以上 ``vector`` 嵌套
 
 
 ### 8.2 x.shape 与 paddle.shape(x)