* fuse optimizer

* modify the interface of Pass::Allay
test=develop

* Polish code
test=develop

* Fix Travis CI
test=develop

* fix Pass::Apply interface
test=develop

* Fix Travis CI
test=develop

* fuse all_reduce
test=develop

* add fuse_parameter_groups_size
test=develop

* Polish code
test=develop

* Fix travis-ci
test=develop

* Add SetGroupAccordingToLayers and SetGroupAccordingToGroupSize
test=develop

* Add SetGroupAccordingToMemorySize
test=develop

* fix multi_devices_graph
test=develop

* reset params_grads
test=develop

* Polish code
test=develop

* Support Sync Batch Norm.
* Note, do not enable it in one device.

Usage:

build_strategy = fluid.BuildStrategy()
build_strategy.sync_batch_norm = True
binary = fluid.compiler.CompiledProgram(tp).with_data_parallel(
        loss_name=loss_mean.name,
        build_strategy=build_strategy)

* add ir memory optimize test case for IfElse op, test=develop

* fix some unitttest failure by force using the python memory_optimize, test=develop

* tweak comments, test=develop

* fix unittest, test=develop

* fix unittest, test=develop

* add ir memory optimize test case for IfElse op, test=develop

* fix some unitttest failure by force using the python memory_optimize, test=develop

* tweak comments, test=develop

* fix unittest, test=develop

* fix unittest, test=develop

test=develop

test=develop

* mem opt

* test=develop

* test=develop

* test=develop

* test=develop

* test=develop

* test=develop

* test=develop

* refine code  test=develop

* refine code  test=develop

* refine code  test=develop

* refine code  test=develop

* refine with cub test=develop

* fix mkldnn test && remove comments && test=develop

* polish code && test=develop

* add only_forward test && test=develop

* Refactor ParallelExecutor
test=develop

* extract Reduce and AllReduce mode from MultiDevSSAGraphBuilder
test=develop

* Refactor MultiDevSSAGraphBuilder
test=developt

* Remove enable_data_balance
test=develop

* code refine
test=develop

* remove data balance
test=develop

* refine ScaleLossGradOp
test=develop

* remove uncessary file
test=develop

* code refine
test=develop

* modify  function name
test=develop

* follow comments
test=develop

* add is_distribution field
test=develop

* set is_distribution
test=develop

* fix DistSSAGraphBuilder
test=develop

* refine parallel_executor

* remove uncessary code
test=develop

* follow comments. test=develop

* Fix typo

* fix compile error. test=develop

* merge develop branch. test=develop

* Remove set_equal

* Polish code

* Delete unused functions

test=develop

* polish code. test=develop

* follow comment

* polish code.

* fix windows compile error. test=develop

* fix op handle.

* rerun ci. test=develop

* rerun ci. test=develop

* rerun macci. test=develop

* polish code. test=develop

* rewrite sort code. test=develop

* remove unused code. test=develop

* fix tests. test=develop

* fix conflict. test=develop

* follow comment. test=develop

* merge develop branch. test=develop

* fix tests. test=develop

* remove ToTypeIndex. test=develop

* rerun ci. test=develop

test=develop

test=develop

test=develop

* fix nccl2 lars dist support

* put lars in momentum op

* add tests lars

* fix ci

* fix cpu kernel

* soft warning

* remove lars in test_recognize_digits.py

* move to another op

* add file

* update api.spec test=develop

* update test=develop

* fix api.spec test=develop

* wip

* wip, finish grad merge ops

* wip, finish graph build

* wip test running

* work on 1 gpu

* workable version

* update

* fix tests

* fuse broadcast op

* fix compile failed

* refine

* add batch merge test mnist

* fix CI test=develop

* fix build

* use independent bn params for batch merge test=develop

* update api.spec

* follow comments and for test

* wip

* refine tests test=develop

* follow comments test=develop

* remove startup bn modify test=develop

* follow comments test=develop

* fix merge test=develop

* Add Preface

* Add demo code

* Save file

* Refine code

* seems can work

* use elementwise strategy

* Use ElementwiseComputeEx

* Add comments

* extract functions from operator

* Refine code

* Follow comment

* code refine

* add op_fuse  pass

* add backward

* code refine

* use TopologySortOperations

* follow comments

* refine IsFusible

* code enhance

* fix op_fusion_pass

* refine code

* refine fuse_elemwise_act_op

* adjust the input and output

* refine logic

* add intermediate_edge

* disable inplace

* follow comments

* refine logic

* follow comments

* Remove the removable IntermediateOut

* change strategy

* code refine

* enable fuse backward

* code refine

* code refine

* rename unit test

* follow comments

* fix the order of sum

* add doc

* check whether need to copy

* follow comments

* Use Builder Chain to decorate new builders. It is easy to extend
  builders.
* Make graphviz path as a build strategy, not a FLAGS.