* [MLIR] Update MLIR/LLVM repos

* Move MLIR/LLVM repos forward

This includes fix to affine fusion algorithm.

* Fix issues after merge

* Fix lit test

* [MKLDNN] Emit dgemm for 2D DP FP Dot op

Add support for emitting MKLDNN's double precision FP gemm from a 2D double
precision floating point Dot operation.

* Removed unnecessarily duplicated pattern

* Add f64 matmul support to CPU Emitter + unit test

* Add check for DP unsupported bias in cpu_fusion.cpp

* fix layernorm flatten issue

* update ut

* checkout output val

* fix style

* apply tolerance

Update jenkins-trigger.groovy

Moving to gitlab.

* Make sure examples compile

* Resolve doc build error due to opset versioning and align dynamic tensor doc to cpp example

* Add latest rc

* Remove deprecated API

* Update brief link summary

* Dist example

* update doc for cpp code examples folder

* Fix typo and toc index

* Build config for example, deprecation in dist test

* style

* Fix serializer op name

* cleanup

* [ONNX] Added gatherND op to ONNX importer.

* Added tests.

* Removed new line.

* Update onnx_import.in.cpp

* Changed tests.

* add ut to reproduce issue from PR#3931

* disable ut for plaidml

* [MLIR] Update MLIR/LLVM repos

* Move MLIR/LLVM repos forward

This includes fix to affine fusion algorithm.

* Fix issues after merge

* Fix lit test

* Add exports

* Work-around windows issues

* windows

* Avoid vectors

* disable group conv if data dilation is not 1s

* Enable group conv decomposition if data or window dilation is enabled. Re-enable a unit-test

* PR fix

* PR fix

* Fix fails

* fused::v1::Split op skeleton & serialization

* Shape inference for Split and upgrade/downgrade passes

* Upgrade pass for the Split op

* Disable the failing onnx_importer Split op tests on interpreter

* Opset1 test correction for v1::Split

* Disable the failing Split op unit tests on PlaidML

* Fix of the upgrade pass for Split

* Remove one of the obsolete v0::Split constructors

Don't imbue locale to cout, imbue a stringstream instead so that it does not contaminate ngraph output (#3989)

* Added v1::Select op with support for implicit broadcasting

* Addressed PR feedback

* Constant folding support for v1::Select op

* Remove commented-out code

* More shape inference tests

This reverts commit 4a1e308a.

* Fix some opset bugs
Typo in opset0
Include ops.hpp rather than ngraph.hpp

* Opset op insertion

* Make opsets extendable, able to create instances

* Update like replacement

* Review comments

* Fixed group convolution infer function

* style

* Add Fluid reduce_sum and reduce_sum_grad

* Temp save.

* Try cmake target_sources

* Make files compile.

* add fluid pool op placeholder

* can compile

* add global pooling attribute

* implement fprop

* implement bprop

* remove unused

* Add FLuid test.

* Fix typo.

* Fix Typo.

* Added reproducer for specialize_function

* Remove GOEs that might have been introduced while cloning nodes in function specialization

* Address PR feedback

* Fix incorrect merge

Typo in opset0
Include ops.hpp rather than ngraph.hpp

* NonMaxSuppression op skeleton

* Validation of the NonMaxSuppresion op

* Correct last 'boxes' dimention check

* onnx_importer support for NonMaxSuppression

* Code formatting

* Type and shape inference for NonMaxSuppression

* Different initialization of NMS inputs in onnx_importer

* Code formatting

* Basic type_prop tests for NonMaxSuppression

* More type_prop validation for NMS

* [SPEC] Add new v1::GroupConvolution and v1::GroupConvolutionBackpropData ops.

* Sort downgrade passes

* Add downgrade pass fro GroupConvolution

* WIP I

* Fix GroupConvolution validate and infer types

* Add upgrade pass for GroupConvolution

* Remove unnecesary get_static_groups() method

* Disable pass for cases when groups are in filters

* Review Fix I

* Move ops to fused/group_conv.hpp

* Use Op instead of FusedOp again

* Move v1::GroupConvolution and v1::GroupConvolutionBackpropData to FusedOp but temporarily disable decomposition

* Enable multitple int types for OneHot 'depth' param

* Move validation logic out of helper

* Make helper more generic

* Convert slope to data type

* UPdated Mod decomposition to use V1 Subtract

* Fixed Selu op decomposition

* Updated Selu op

* Add Fluid reduce_sum and reduce_sum_grad

* Temp save.

* Try cmake target_sources

* Make files compile.

* Add fprop and build Fluid ops by default.

* Some fixes.

* Fix shape inference issue.

* Add bprop.

* Fix bprop.

* Bug fixes.