* Collapse dimensions for inputs to Dot

* Remove eigen kernels for higher dimension dots since they will collapse to cblas_gemm kernels

* Moved collapse dims pass after the fusion passes to prevent interference with fusion patterns. Use cblas_gemm for 2D dot

* Refactor include_helpers into an nvrtc specific helper file. Add templated define functions for coherent and noncoherent memory loads.

* Format

* const refs.

* Remove cast of zero.

* add nop backend

* nop backend

* fix flag name

* add new switch to cmake output of switch settings

* add new unit test to igpu manifest

* remove redundant test

* address issues which surface with clang 6.0

* revert changes due to new clang warning and disable new warning

* quantize scale passing unit tests

* epsilon bump

* finished with quantization scale

* unit tests passing with convolution scale as builder

* broadcasted constants and cleanup

* api consistency for quant builders

* code style

* cleanup

* newline at EOF

* use requantization_scale

* drop TF license as we are no longer using TF code directly

* Do not fuse nodes if one node is predecessor of another node in horizontal fusion.

* Add dead node check and remove predecessor check in horizontal fusion.

* wip

* wip

* wip

* move hybrid wrapper to hybrid backend dir

* move hybrid wrapper to correct namespace

* wip

* sorta working

* remove debug from sorta working homogeneous hybrid backend

* is_supported is supported for GPU

* cleanup debug

* more progress

* remove debug

* cleanup

* turn off hybrid by default

* revert change

* revert

* rename wrapper to backend

* revert

* address review comments

* style

* Add missing halide dependency

* Bug fix in halide op creation

* Localize halide/llvm to cpu backend

* Added comments

* Pass NGRAPH_HALIDE to tests

* Resolve merge conflicts

* Update CODEOWNERS for /licenses and /LICENSE

* Review comments

* Minor formatting

* onnx: enable external weights to enable Caffe2 support
Signed-off-by: Artur Wojcik <artur.wojcik@intel.com>

* onnx: update ONNX importer interface documentation
Signed-off-by: Artur Wojcik <artur.wojcik@intel.com>

* onnx: after review updates
Signed-off-by: Artur Wojcik <artur.wojcik@intel.com>

* added tests for malloc mode and graph transform

* Comment incorporation

* changed comparing backend to INTERPRETER

* COmments resolved+clang

* Adressed all comments

* IntelGPU does not support topk

* size_t for placement in node

* enable hybrid backend test

* style

* cp placement functions

* placement size_t based functions

* placement based on backends

* add placement based on size_t

* backend size_t based placement

* call

* update

* resolve bug

* format

* revert cmake changes

* address PR comments

* ci error

* pr comments

* extend cse to handle backend ops

* revert back to static casts

* Make debug logging threadsafe

* Add nil stream comments

Primitive CPU Executor for managing thread pools and executing kernels across partitioned resources (#1858)

* Add arena support to DEX kernels

* Add CPUExecutionContext to pass executor information to kernels

* Removed eigen_thread_pool. Functionality moved to cpu_executor

* Moved MKLDNN engine to CPU executor

* Added TBB arenas for node affinitization

* Fixed missing arena in one hot kernel

* Style fixes

* remove 5arg BNTraining

* Remove 5-arg BNWithStats from gpu/op.

* Change GetOutputElement to use delayed validation

* Cleanup

* Update "JSON for Modern C++" to a version that officially supports gcc 4.8

* "JSON for Modern C++" is header only and does not need to be configured.

* Implement partial shape/type propagation for Convolution; fail for want of unit tests

* Implement unit tests for partial shapes/types for Convolution

* Implement partial shape/type validation for BatchNormInference, BatchNormTraining, BatchNormTrainingBackprop with unit tests

* Formatting

* Update CPU and GPU backends to deal with elimination of BatchNormBase

* Update BatchNormTrainingWithStats to use templated emit_BatchNorm function

* Restore five-argument BatchNormTraining ctor for now; #1901 will eliminate it

* Replace enum for input indices with symbolic constants

* Update intelgpu/visualize_tree.cpp to deal with removal of BatchNormBase

* generate mask

* add codegen+dex

* states in context

* CPU dropout

* remove dead code

* remove dead code

* remove state.cpp

* change perms, add state.cpp

* address bobs feedback

* restore igpu unit-test manifest after a bad merge

* better error msgs

* throw on GPUs to keep a compiler happy

* address more feedback

* fix tests

* graph_rewriter until fixed point

* fix a bug + comment

* move comment

* move rewritten=true into process_node conditional

* create unique_ptr backend

* unit test cleanup

* address more code that was recently added

* change from reference to pointer when passing backend to reduce the number of lines changed.

* fix build error

* fix python wrapper

* style

* more specific treatment for unique_ptr