* elementwise_add with bcast: Brian's implementation by Brian added, with default bcasts

* elementwise_add with bcast: GetExpectedKernelType added to elementwise_op

* elementwise_add with bcast: use_mkldnn attribute added

* elementwise_add with bcast: changes after review and some formatting

* elementwise_add with bcast: changes after style check

* elementwise_add with bcast: changes after style check cont.

* elementwise_add with bcast: MKLDNN unittests added

* elementwise_add with bcast: original unittests with use_mkldnn flag

* elementwise_add with bcast: handling of MKLDNN format corrected

* elementwise_add with bcast: setting MKLDNN format turned into lambda

* elementwise_add with bcast: MKDNN format setting turned into separate function

* elementwise_add with bcast: condition for choosing MKLDNN simplified

* elementwise_add with bcast: fix for MKLDNN format set incorrectly in bcasts

* elementwise_add with bcast: changes in unittests for broadcasts

* elementwise_add with bcast: fixes in unittests regarding dimensions

* elementwise_add with bcast: bring back correct format setting in mklml grad path

* elementwise_add with bcast: fixed compilation error

* Add MKLDNN layout support in Paddle

Add MKLDNN layout in Paddle so that MKLDNN friendly memory layout
can be used in MKLDNN enabled OP kernel. Before this commit, NCHW
is hardcode to be used in all MKLDNN op kernels. As a result,
non-optimized execution path is selected in MKLDNN primitive which
bring worse performance.
Besides framework change, three MKLDNN OP kernels were updated
for using new MKLDNN layout. They are conv/pool2d/batch_norm.
Other MKLDNN OP kernels need be also updated in similar way to
achieve best performance.

* Add MKLDNN layout support in activation OP

* Don't populate layout from input to output when kMKLDNN in

* Refine pool mkldnn op kernel

* MKLDNN layout

* Remove the inferitance from tensor file

* MKLDNN layout: refactoring

* Remove additional #define to register new operator

* Prepare mkldnn tests to work with layout

* Fic CPPLint issues with data_transform

* Fic CPPLint issues with prune.cc

* test cpu float16 data transform

* add isnan etc

* small fix

* fix containsNAN test error

* add data_type transform GPU test

* add float16 GPU example

* fix error

* fix GPU test error

* add context wait

* 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 data layout transform and optimize the implementation of data_transform

* "unified operators"

* "add CUDNN register"

* "add use cudnn attribute"

* "add attribute"

* "test conv tranpose op"

* "remove duplicated attr"

* "fix op test"

* "add attribute to set cudnn"

* "add more log"

* "need layout op register support"

* "add more log"

* "change GetExpectedKernelType "

* "fix Get attr in conv_op"

* "fix CI"

* "fix tests"

* "removed kernel priority fallback"

* "fix CI"

* "fix stack pointer bug"

* "refine buggy interface"

* "add const cast to save life"

* "fix get_output_with_grad"

* "fix op test with dataformat"

* ""fix pooling

* "fix pooling test"

* "fix CI"

* "fix with_gpu error"

* "add transform needed functional check"

* "fix unpack list error"

* "comment out parallel.do temporary"

* "fix CI"

* "fix compile doc error"

* "make threshold larger"

* init data_type_transform

* split data_layout_transform

* tmp rm data_transform_test

* change device_data_transform to data_device_transform

* clean code

* clean code

* "reuse ShareLoD with no regret"

* "removed base class shareLayout"

* "fix CI"

* add rename guard

* add device_data_transform

* add device_data_transform_test

* modify GetExpectedKernelType

* update operator.run

* support test test_label_semantic_roles

* optimize code

* optimize code

* rename GetActualKernelType to GetExpectedKernelType

* fix chunk_eval_op and device_data_transform_test

* add is_same_place to place

* optimize code, refine rename_guard

* refine rename guard, add GetKernelTypeForVar

* optimize code

* add some log

* rename guard

* use sub scope to create var

* fix compile

* add IsInitialized for Tensor

* add VarIsTensor

* fix op_registry_test

* test

* tmp disable priority

* restore switch_kernel.md

* code clean

* "add c++ side kernel selection"

* "add multiple kernel op test"

* "kernel selection only support cudnn"

* "better formatter"

* "small fix with UseCPU"

* "depends on change interface Get(Place, Library)"

* "fix CI"

* "fix python cudnn test"

* "leave the register cudnn op to another PR"

* "fix CI"

* "use all kernel by default"

* "fix CI"

* "fix typo"

* "fix based on comments"

* "follow gogle style"

* "fix based on comemnts"

* "fix data transform"

* "data transformer"

* "add device pool"

* "add test"

* "fix ci"

* "fix datalayout implementation "

* "fix based on comment"

* "fix data transform"

* "split into next PR"

* init data_transform

* complete DataTransform

* fix build error

* add data_transform_test

* add a register test for data_transform_fn

* use function to simulate registration macro

* add register macro

* update test

* clean code

* restore unrelated code

* update data transform test

* generate unique name for REGISTER_DATA_TRANSFORM_FN

* add const

* follow comment

* update KernelTypePair hash function

* "fix threadpool style"

* "remove header"

* Feed/Fetch op just plain operator, not a OpWithKernel
* Do not register OpInfoMaker since Feed/Fetch will never be
  configured by users
* Feed/Fetch op has empty gradient
* Feed/Fetch op do not hard code `feed_variable`, `fetch_variable` as
  its input and output, make it as a plain Operator input/output

* add NDEBUG switch to PADDLE_ENFORCE

* Free will be added soon