* Use CBLAS for SelectedRows elementwise add operation. It's faster.

* template compilation fix

* reverted template compilation fix

* slimmed template compilation fix
Co-authored-by: NAdam Osewski <adam.osewski@intel.com>

* Add oneDNN AXPY handler.

* Add fallback for small tensors.

* Fix ifdefs

* Remove unnecessary namespace prefixes and add missing headers.

* Guard handler_axpy with proper ifdefs.

* Compilation of this function is possible only when Paddle is not build
with CUDA nor HIP.

* Move AXPY handler code to separate files.

* Use oneDNN AXPY handler in SGD op.

* Use axpy handler only when Paddle is built with oneDNN.

* Add test for SUM BF16 with big rows.

* Fix SFINAE rules for elementwise_add_to.

* Add test case for SGD with big rows.

* update

* update
Co-authored-by: NAdam Osewski <adam.osewski@intel.com>

* modify matmul Op to complex template types

* remove complex64/128 head file

* use iwyu clean include second time, test=develop

* add support for complex grad accumulated

* add unittest for coverage

* update test dtype

* remove useless blank line

update error info for selected_rows_functor

* add new macro BOOST_GET_SAFELY & unittests, test=develop

* add different macro type, test=develop

* fix get macro type in executor, test=develop

* four macro part change backup

* using one macro for all case, test=develop

* revert attribute change, test=develop

* change to three func to solve gcc4.8 bug, test=develop

* polish some details, test=develop

* test=develop,  communicator merge add => merge average

test=develop

test=develop

test=develop

test=develop

Fix code to support cpplint syntax check

test=develop

test=develop

test=develop

  1. Accelerate SelectedRows MergeAdd functor

  2. Add SelectedRowsSumTo functor to support MergeAdd multiple SelectedRows into one

test=develop

* Fix CPPLint errors

* Fix CPPLint errors in sequence2batch

* Fix compilation

* Fix LSTM op and GRU op

* Fix LSTMP op

* Fix more cpplint errors in operators/math

* Address Code review feedback

There are mainly following fixes:

- take `DeviceContext` as the template parameter of math functors and OpKernel instead of `Place`
- remove `eigen_device` interface in base class  `DeviceContext`
- remove `GetEigenDevice` interface in `ExecutionContext` and base class `DeviceContext`
- remove unused `platform::EigenDeviceConverter`
- rename `REGISTER_OP_GPU_KERNEL` to `REGISTER_OP_CUDA_KERNEL`
- rename `USE_GPU_ONLY_OP` to `USE_CUDA_ONLY_OP`