* Support for no-arg op construction

* Add a-ops.

* Two const init variations to try to solve windows link error

* Try export for windows

* Create outputs as needed, test

* Remove attribute helpers

* another windows try

* windows

* windows

* Partial node cleanup

* Review comments
Node destruction that doesn't blow out the stack on deep graphs

* Review comments

* Update src/ngraph/node.hpp
Co-Authored-By: Adam Procter <adam.m.procter@intel.com>

* Have huge test check that nodes were deleted

* Switch constructors to use Output instead of Node

* Review comments

* Add some comments

This patch introduces tablegen definitions for element-wise reduction
operations (only definitions, not building/lowering code). This includes
argmin, argmax, min, max, sum, prod, all and any.

From the design point of view, a single base class seems to be enough to
cover all the common ground of these operations. A different base class will
be necessary for avg pool reduction operations.

* Improve verification for unary, binary, cmp and select ops. Refactor .td file to enable that.

MLIR Compile only supported ops

Element type was not lowered.

It introduces end-to-end support for sub-graphs with multiple nodes. It
adds a test with a dot+add ops.
Co-authored-by: Diego Caballero <diego.caballero@intel.com>
Co-authored-by: Nagy Mostafa <nagy.mostafa@gmail.com>

It extends dot definition to be able to deal with operands that are not
square tensors. It also fixes a bug in the lowerer related to that.

* [MLIR] Replace MatmulBiasOp with DotOp

We disable CPUFusion if MLIR is enabled to avoid CPU specific ops to be
introduced for now.

* Upgrade MLIR. Several code fixes based on API changes

* Fixes due to DialectConv API changes

* style-apply

* PR fixes

MLIR is now initialize once once in CPU backend.

* Move dialect and types to mlir namespace

* PR fixes and some cleanup

* Merge fix

* Enable module verification. Fix FakeInput result type to memref

* style-apply

* Initial td file. Cmake changes

* Move all ops to .td file.

* Added few more opcodes to show-case. Fixed PR feedback

* Remove NG_ prefix of opcode records. Some fixes

* Added some doc

*  Adding back NG prefix

* Bug fix in MLIR gen

This patch leverages CompiledKernel to delimit sub-graphs to be compiled
with MLIR. It introduces a pass that creates a CompiledKernel for the
whole function (for now) and changes MLIRCompiler to align with this new
approach.

* Use NGRAPH export macros instead of CPU

* Move code to ngmlir namespace

* Move MLIR to src/contrib/mlir

This patch enables LLVM optimizations at -O3 level.

We want to use ComputedKernel for any target to delimit sub-graphs to be
compiled and executed with MLIR.

* Create MLIR as cmake external project. Clone and build via ngraph cmake

* Moved code and enabled compilation. Need to clone and build MLIR/LLVM during cmake step, so find_package can work

* clone and build llvm/mlir during configuration. Compiles now. Needs more testing

* Force DEX only if MLIR is ON

* Remove extra cmake file. Style-apply

* Remove redundant files in cpu/mlir

* Update CODEOWNERS. Check for ninja and fail if not found

* Fixing post merge issues

* Implements a simple memory manager that just does malloc for now. Pointers are freed during cleanup.
* Enable JIT call-back to memory manager to allocate temps.
* Memory manager pointer is passed to the JIT'ed code upon invocation. That makes the code re-entrant from different threads in case the code is shared among identical sub-graphs that are executed in parallel.

The following test should work now:
NGRAPH_MLIR_DUMP_ALL=1 NGRAPH_MLIR=1 test/unit-test '--gtest_filter=CPU.dot2d'

* Link MLIR static libs to cpu backend

* Use LLVMConfig.cmake

* Initial commit. Link fails with undefined reference to typeinfo for mlir::Dialect

* Added AddOp

* initial compiler class

* Initialize module/function, and map tensors to arguments

* Code compiles. Moved MLIR building to correct DEX handler

* NGDialect code-gen working

* Use vector instead of sets for i/o tensors. Use functor in executor

* Misc fixes

* style-apply

* WIP: Adding support for dialect lowering.

* WIP: Lowered to affine. Crash on constant ops have side effects in Constant Folding

* Fixed missing whole package linkage.

* Removed fake instruction and update func type

*  Enable lowering to LLVM dialect and IR

* Made loop nest builder handle any rank

* Fixes per PR feedback. Major ones:
- Removed ngdialect namespace
- renamed dialect classes to start with NG prefixwq:w

* Add unreachable assert

* Add reading of LLVM options from an env var MLIR_LLVM_OPTIONS (#5)

Avoid negative values in int64 initialization for cases where int64 parameters are used as indices (#3004)

* Link MLIR static libs to cpu backend

* Use LLVMConfig.cmake

* handle case where a node's output is connected multiple inputs of another node

* fix creation of the FunctionCall to have the correct outputs

* fix per review comment

* Cleanup how compile flags set and used by nGraph and external projects.
Set C++11 through CMake and pass it down to external projects.
Prefer CMake variables such as CMAKE_POSITION_INDEPENDENT_CODE and
CMAKE_CXX_STANDARD instead of explicitly setting compiler dependent
flags.
Create json compilation database for external projects.
CMAKE_CXX_FLAGS is used as common global options for nGraph and external
projects.
add_compile_options() is used for local options for current and sub
directories.
add_definitions() is used for setting definitions for current and sub
directories.
Note: Global options are not passed down to some external projects.
Note: mkl-dnn resets CMAKE_CXX_FLAGS internally.
Note: TBB and MLSL are not CMake based.
Noet: Eigen and json is header only library.

* Fix error.

* Fix error. (second attempt)

* Cleanup code.

* Allow check for undefined macro.

* Try to fix cldnn issue.

* Set type for CMake arguments.

* Pass C++ standard to protobuf.

* Pass C++ standard down to TBB.

* Change how Clang specific flags are handled.

* Fix error.

* Workaround for compile error on Baidu's PDPD docker.

* Fix windows build error.

* Initial implementation of implicit broadcasting for eltwise ops. Only Add supported

* Addressed PR feedback

* cleanup

* Rename Bcast to Broadcast

* Autobroadcasting support for rest of elementwise ops

* Serializer support for autobroadcast

* Added missing autob serialization for Minimum

* Added execution unit tests and more op types to implicit broadcast elimination

* Addressed PR feedback

* Fixes windows build issue

* RVO optimization per PR feedback

* serialize constant faster

* more speedup

* Draft of FakeQuantize operation along with UTs.

* Add FakeQuantize to implemented operators on IGPU.

* Get back FakeQuantize op case to switch.

* Fix compilation errors.

* Skip test for INTERPRETER backend and disable type_prop tests.

* Initial implementation covering the most basic case

* Cleanup of fake_quantize_with_clip UT

* Reformat the cpu unit tests manifest and unlock anothe fake quant UT

* Handle the clipping case by subtracting input_low from quantization input

* Clip the input data before quantization to avoid Selects

* UT manifest fix

* Obsolete comment removed

* Code formatting

* Broadcast input data for non-scalar in/out params

* Code formatting

* Enable the type prop tests for FakeQuantize

* Dequant the data without using the Dequantize op (fixes an edge case)