Change the behaviour of thread pool. Thread pool will ignore the
thrown exception implicitly. It is hard to debug.
Corrently, ThreadPool::Run will invoke `LOG(FATAL)` if an exception
thrown.

Add a new type `Reader` for `VarDesc`, which can holds more than one
LoDTensor.

* refine buffer channel

*  refine Receive and Send

* follow comments

* Adding more receivers less senders

* Added more receivers less senders

* Added more send

* Updated comment

* Fixed code style

* Fixing review comments

Adding unit tests for checking that closing unbuffered channel closes all blocked senders and receivers (#8067)

* Adding unit tests for checking that closing unbuffered channel closes all blocked senders and receivers

* Fixing sleep time

* Add unite test with less receivers, more senders

* Fixed the check

* Adding an initial implementation for the unbuffered channel

* Including atomic header

* update comment

* Adding the closed attribute

* Add comments

* Updated locking mechanism

* Add simple unbuffered test

* Enhance unit test and fix bug

* Add details

* "clean code"

* "clean code"

* "Need to re-design LoD "

* "add lod design"

* "fix lod gpu ptr pointer"

* "removed commented code"

* "fix CI"

* "remove set lod in pybind"

* "fix style check"

* "fix CI"

* "fix long type template error"

* "pybind reorder to use Place"

* "fix ci"

* "fix ci"

* fix ci

* "sperate as a new file"

* "fix CI"

* "fix ci"

* small fix

* "add test"

* "fix adam op"

* "fix lstmp op"

* "fix adam op"

* "follow comments"

* "fix ci"

* refine channel test

* follow comments

* Add dependency enforce to threadpool

* Revert changes to channel_test.cc

* Revert changes to channel_test.cc

* Add #include "paddle/framework/macros.h"

* Add unit tests

* fix code format

* refine close channel

* follow comments

* use delete to destroy channel

* "fix gpu init"

* "set env variable default value for share gpu"

* "fix ci"

* "removed CUDA_VISIBLE_DEVICES default"

* "removed"

* "unify flags"

* "fix init"

* Initial commit

* Remove resolution bug

* Modify IsParam

* Remove small bugs

* First commit unifying Run and Load

* Fix bugs

* Fix Cmake

* Modify Cmake and dir structure

* Add io.* files to inference dir

* Fix include in example

* Address review comments: part 1

* Address review comments: round 2

* Address review comments: round 3

* Address review comments: round 4

* refine channel test

* follow comments

* Add dependency enforce to threadpool

* Revert changes to channel_test.cc

* Revert changes to channel_test.cc

* Add #include "paddle/framework/macros.h"

* Remove IsBounded as buffered channels have to be bounded

* Add derived classes Buffered and UnBuffered"

* Implement buffered and unbuffered channels

* Correct the syntax of Channel::Receive

* clang-format

* clang-format 3.8

* clang 3.8

* Polish threadpool

* Add #include <vector>

* Rename variables

* Rename variables

* clang-format

* initial commit

* add new executor run function

* fix bug

* fix multiple definition of feed_fetch_method issue

* fix cmake

* fix tensor copy error

* refine executor code

* add comments

* temporary modification

* address comments

* fix bug

neural machine translation model support beam search with while op

* limit variable type to lod tensor in memory optimization transpiler

* refine policy

* support while operator

* fix random seed and training data order

* refine get_cfgs method to support multi while operators

* refine codes

* add memory optimization transpiler demo

* add memory benchmark compile option

* add gflags instead of macro

* refine code