* Modify relu native implementation

* fix GPU performance

* add  parameter in roi_align op

* enable exhaustive_search for input_grad when dtype is float16

* enable exhaustive_search for forward algos

* Add conv transpose BF16

* Share function GetWeightsTz

* Adjust to review and fix op compatibility

* Add bias to unique handler name

* Remove errors related to paddle enforce

* Add conv2d_transpose to bf16 list and kernel refator

* use iwyu clean include second time, test=develop

fix the broadcast for the large second input 

* update lite tag

* disable ut

* transformer predict

* trans bug fix

* added external reorder to profiler

* resolved conflict

* added enable_static

* initial version of lstm, not working yet

* added lstm to operators.cmake

* added vanilla lstm mkldnn op

* added peephole weights integration

* minor changes

* added formatting

* added fusion_lstm_mkldnn to static_whitelist

* added formatting

* removed comment

* moved use_peepholes attribute inside is_cached block

* reverted wrong changes

* minor formatting change

* minor changes

* changed stream handling

* minor change

* added datatype to GetExpectedKernelType()

* added reading stream from TLS

This reverts commit d834f4e6.

* added external reorder to profiler

* resolved conflict

* added enable_static

* initial version of lstm, not working yet

* added lstm to operators.cmake

* added vanilla lstm mkldnn op

* added peephole weights integration

* minor changes

* added formatting

* added fusion_lstm_mkldnn to static_whitelist

* added formatting

* removed comment

* moved use_peepholes attribute inside is_cached block

* reverted wrong changes

* minor formatting change

* minor changes

* More precise mkldnn kernel choice in GetExpectedKernelType

* Fixes after review

* Refresh develop for CI

* CI experiment

* get back from CI exper

* add abs test case

* use std::abs to fix abs bug

* fix the abs bug

* fix abs bug

* support reduce_max op on kunlun

* support reduce_max op on kunlun

* support reduce_max op on kunlun

* support reduce_max op on kunlun

* optimize unity build, test=develop

* fix code style error, test=develop

* fix code style error and test /MP settings, test=develop

* rewrite abs op

* rewrite abs op and remove abs in activation

* remove abs register in old codes

* fix abs_grad type error

* fix abs double_grad output name error

* modify abs_grad, abs_grad_grad functor for windows building

* format code style

* fix the bug of result is nan when the divisor is zero

* add missing abs attr and add abs for float16

* add rmsprop_op_xpu test=kunlun

* modified rmsprop_op_xpu error code. test=kunlun