Merge branch 'dist_train_benchmark_vgg16' of https://github.com/typhoonzero/Paddle into dist_train_benchmark_vgg16

fix copyright

Move registery.py to layers module

Make compare_op reuse elemwise_op_funcs

Add WMT16 dataset.

* add memory optimization transpiler demo

* add memory benchmark compile option

* add gflags instead of macro

* refine code

Gradient clip by global norm

Add math operator patches

Fix priorbox layer when multiple values given in min_size

Add math function for sampling integers

refine openblas install lib when CAPI

[WIP] Split SelectedRows to multiple pservers

* init complete data layout transform

* can compile

* test passed

* optimize code

* fix while_grad_op first step loss lod problem

* optimize in out ptr for transform

* add check

* update copyright

* clean code

* add NeedTransformLayout

* add comment

* change the interface of data_type_transform

* init data_type_transform_test

* complete data_type_transform_test

* add TransDataType to data_transform

* Add fluid_compiler.md

* Paragraphing

* "fix decode bug"

* "follow commnet"

* "fix error"

* "fix hook bug"

* fix based comment

* fix copyright

* fix based on comment

Enhance matmul_op to support 4-D inputs.

remove unused vars which block the build

Recv OP: use BlockDesc* instread of ProgramDesc proto as Attribute

Fix the description of default scale attribute for scale op.