* onnx: fix build for CentOS 7.4
Signed-off-by: Artur Wojcik <artur.wojcik@intel.com>

* onnx: fix build for CentOS 7.4
Signed-off-by: Artur Wojcik <artur.wojcik@intel.com>

* onnx: compile fix for CentOS 7
Signed-off-by: Artur Wojcik <artur.wojcik@intel.com>

* onnx: unit-test compilation fixes for CentOS 7
Signed-off-by: Artur Wojcik <artur.wojcik@intel.com>

* onnx: unit-test compilation fixes for CentOS 7
Signed-off-by: Artur Wojcik <artur.wojcik@intel.com>

* onnx: fix 'add abc initializer' unit test
Signed-off-by: Artur Wojcik <artur.wojcik@intel.com>

* onnx: fix code style
Signed-off-by: Artur Wojcik <artur.wojcik@intel.com>

* argmax

* manifests and serailizer

* get_default_order

* add a newline

* address scotts feedback

* add numeric header

* IntelGPU backend: Broadcast operation optimization

* PR1450 style check fixed

* add EXCLUDE_FROM_ALL to codegen library so we don't need to using conditionals in build. If codegen is needed it will be built.

* update llvm dependencies

This reverts commit 5ab5a129.

* static link cpu library to ngraph

* remove debug

* link ngraph and cpu backend into a single shared object

* add -fPIC and whole-archive for CPU backend

* Added conditional for --whole-archive for Mac OS.

* Added more conditonal for MacOS.

* fix linking problem and unit test failures caused by multiple copies of the same function in CPU backend and INTERPRETER

* fix nbench build

* add nbench to unit test build

* add version number to libngraph

* - update MKLDNN version from v0.14 -> v0.15

* - added support to query tensor size from mkldnn desc for MKLDNN Op's
- moved tensor size calculation to tensor_view_layout
- populate mkldnn layout desc witk MKLDNN queried size in the cpu_layout pass

* style fix

* - Fixed failing test cases

(cherry picked from commit d68d3738673c2081f458f9aa458c16361cc47657)

* style fix

* renamed tensor->logical_size() member function to tensor->size()

* Addressed PR comments

* - update mkldnn from v0.15 to v0.16
- update mkl SHA1 hash for lnx, max, win

* style fix

* Addressed Scott's PR comments

* merge branch master with pruthvi/mkldnn_v0.15

* argmin

* address feedbacka argmin

* add new lines

*  addnew lines

* address adam's nitpicks

* scott's feedback

* fix unit tests

* optimize replace slice

* fix bugs

* fix bug

* optimize pad dynamic

* fix bug

* fix bug

* fix bug

* remove *

* add gpu_assignment to pass

* refactor cuda replace slice.

* fix bug

* refactor replace slice

* working version

* clang format

* us layout instead of assignment

* us layout instead of assignment in cmakelist

* update gpu_layout

* fix bugs

* resolve conflict

* GPUShape to NVShape

* using kernel args

* using kernel args

* fix bugs

* fix bugs

* fix bug, remove mkldnn.h from gpu_layout.cpp

* fix bug for pad_below

* remove cast to rep_slice

* fix bugs

* clang format

* change add_in_place_oi_pair({0, 0, false} to add_in_place_oi_pair({0, 0, true};

* [ONNX] Add Relu op

* Refactoring

* Refactoring - move op::Constant implementation to .cpp file

* cmake file list order

* Code review

* [ONNX] Add Relu op

* IntelGPU backend: BatchNormBackprop operation

* PR1443. Requested refactoring done

* Collapse dims when possible to facilitate faster kernels

* Minor debug change

* Define and track tensor roles through graph optimizations

* Backend test: MaxPoolBackprop test strided and padded

* PR1428. Test excluded from CPU backend

* Optionally get rid of codegen from the CPU backend

* Rename option variable

* Merge fixes

* Merge

* Remove extra changes

* remove dex only exclusions (#1429)

* Unconditionally pick  m_direct_execution if NGRAPH_DEX_ONLY

* Style fix

* IntelGPU backend: Code refactoring and optimization

* Update gen_func_def()

* dex sigmoid multiply

* sigmoid multiply

* refactor compute logic into standalone kernels

* address jayaram's feedback

* onnx: pull Protobuf 3.5.x from GitHub
Signed-off-by: Artur Wojcik <artur.wojcik@intel.com>

* onnx: fix Travis build configuration issue
Signed-off-by: Artur Wojcik <artur.wojcik@intel.com>

* onnx: additional fixes for Travis
Signed-off-by: Artur Wojcik <artur.wojcik@intel.com>

* onnx: update documentation - build requirements for MacOS
Signed-off-by: Artur Wojcik <artur.wojcik@intel.com>

* onnx: update documentation - build requirements for Ubuntu
Signed-off-by: Artur Wojcik <artur.wojcik@intel.com>

* Merge

* Remove extra changes

* IntelGPU backend: Minimum, Not, OneHot operations

* Code format update

* Refactor OneHot op

* PR1414. exclude one test

* Fix errors

* IntelGPU backend: MaxPoolBackprop operation

* PR1427: Initialization loop added

* Define and track tensor roles through graph optimizations

* style fix

* Address PR feedback

* [Py] Wrapper for LRN

* Add missing header

* Add default param values, docs and some unit tests

* Fixes

* clang-format

* Fold affine transformations on 4d convolution

* Handle more cases for affine parameters

* Style fix

* Temporary broadcast optimizations till halide support is merged

* Use computed broadcast axes

* Move shape squeeze

* Review feedback: Add comments

* Patch broadcast axes after the last refactor

* Squeeze input shape