* Skip --exclude-libs linker flag on macOS

* Change test to if(LINUX)

* Add "Any" op and AnyAllReplacement pass

* Add AnyAllReplacement to IGPU backend

* Stub (error-out) handlers for GPU and INTELGPU

* Add 'All' op

* Add AnyAllInsertion pass, deprecate deprecable ops, add stubs for INTELGPU

* Add failing unit tests to INTELGPU manifest

* Reduce boilerplate

* Reduce more boilerplate

* Add static keywords

* embedding fprop

* add a new line

* type prop tests

* rename

* add a stub handler for embeddinglookup on intelgpu

* rename embedding.* to embedding_lookup

* rename tests in manifest files

* move embeddinglookup to catchall case

* fix test case breaks after merge

* add a negative test, pull up an assertion

* fix test failures

* add empty framework for hybrid GPU, or GPUH

* move placement to the runtime directory

* wip

* skeleton for hybrid GPU backend. most unit tests pass.

* cleanup

* move hybrid code into hybrid dir/namespace

* move hybrid functions

* move more hybrid functions to hybrid directory

* fix placement after compile. All unit tests passing

* fix gpu backend ctor

* files pulled from bob/winbuild

* fix compile problems

* fix a few windows build errors

* add windows file to exclude from git

* add comment why change was made

* revert obsolete change

* more cleanup

* building interpreter and unit test on windows with DLLs

* Add flag for windows to export all symbols. Short term fix.

* enable MD build

* address warnings

* dump all windows build results to a single directory

* fix windows backend dll open issue

* remove debug

* fix file iterator for windows

* fix merge error

* fix test failure

* change header from h to hpp in hopes of making python happy

* address more linux build issues

* fix visibility enable

* add some helper function

* update with new helper function

* update reduce to nd with new helper function

* update float sum to stable sum

* fix bug

* update all reduce to stable sum for float

* fix bug and pass the sum stable test

* remove debug info

* style

* update with shape

* fix bug

* add host parameters to cuda_emitter

* clang format

* fix bugs

* add element::type support

* format

* add a cached value with datatype name

* add init_reduce_value

* unroll loop

* optimization

* remove the need for init_value

* add memset kernel

* add memcpy

* working version

* remove debug info

* add comments, clean up code.

* change in_idx to input_idx

* fix bug

* change args name for memset in emitter

* pass element::Type instead of string

* the op::reduce come with init value, add support

* resolve codacy-bot comment

* fix bug

* resove codacy-bot comment

* add soft_max_block_reduce kernel

* fix bugs

* add softmax_block_reduce to cuda_emitter

* compiing ok, result wrong

* fix bug in kernel

* working version

* removed unused code

* remove unused comments, resolve comments

* cuda reduce for max, min, mul, reduce op init value, format

* use type::info

* use type info for numeric_limits

* remove code from gpu_host_parameters

* header

* remvoe outdated comments

* add helper to check if stable sum is needed

* add stable sum test for double

* remove extra line

* consolidate helper functions

* no need list now.

* remove extra ;

* clang format

* style

* add skip test for cpu and intelGPU side

* resolve more conflict

* update comment

* fix a warning

* Update src/ngraph/runtime/gpu/gpu_cuda_kernel_builder.cpp

using load.
Co-Authored-By: fengleitian <35274053+fengleitian@users.noreply.github.com>

* using WARPSIZE instead of 32, using lambda

* more WARPSIZE instead of 32

* fix block_size_x bug

* using __expf

* fix crash in ngraph-tf test conv_ops_test.Conv2DTest.testConv2DKernelSmallerThanStrideSame

* fix file perms

* correct checks

* Bind cuda context to thread prior to compilation. Small refactoring.

* bind_cuda_context_to_thread in source

* bind_cuda_context_to_thread header

* [Py]Add version to ngraph python

* FIX

* reshape softmax reshape

* add new line

* add new line

* fix style errors

* Update cpu_external_function.cpp

* fix test case failures

* env var to abort matching

* Update matcher.cpp

* Update matcher.cpp

* add a comment

* give an env var a better name

* Fix installing numpy dependency on CentOS.

* Check whether nGraph library directory exists.

* Bug fixes to unordered map checks

* No in-place slice for non-native MKLDNN layouts

* is_op

* Removed winml from stack diagram

* Removed winml from full stack diagram

* Update README.md

* update the diagram without winml

* Changed sentence about WinML

* Removed duplication

* change all_close tests to return gtest AssertionResult instead of bool to allow for better error messages

* change throw to return error

* address PR comments and fix compile error

* reenable mkldnn convolution for large padding

* specify precision tolerance to unit test

* pass tolerance values to all_close

* move GPU specific test to GPU only

* fix unit test invocation

* fix compile error

* fix compile error

* style

* fix runtime error

* make GOE extend from util::Op

* fix build breaks

* initial commit for update slice op

* Finished up update_slice fusion and added codegen support

* style fixes

* Added unit test for in-place update-slice strided

* change pattern name

* change compile call to return Handle

* make CPU require compile() before call()

* fix unit tests to call compile() before call()

* fix failing ops

* update unit test

* revert some changes

* more fixups

* more diff cleanup

* a few more issues addressed

* more fixes

* update API

* more updates

* fix test_ops.py

* fix

* another attempt to fix

* fix unit test

* fix test error

* Double support for all_close_f

* all_close_f uses fixed number of mantissa bits now. Simplified testing code.

* Initialize test data members in constructor to values which will cause test failure. Setup then sets them correctly.

* Reduce info printed out during all_close_f unit tests.

* re-enable quantize_clamp_int32 test on CPU

* MLKDNN typo

* Quantize the bias to int32

* Bias scale fix

* mnist works

* Quantize Bias

* Introduce Quantize op in the graph to quantize bias & feedback

* Add QuantizedConvBiasAdd

* Comments and some refactoring

* Add test case with float bias and enable int32 as quantized type in ngraph

* Change shape of scale from Shape{} to Shape{1} in the backend

* Add QuantizedConvBiasSignedAdd

* Fix Layouts, clean up and a test case for QCBA

* Test case for QCBSA

* cleanup mkldnn_emitter.hpp

* fix build error

* Constant fold

* one output

passing tests

clean up

fix build breaks

* move generators into a separate file

* an env var to disable individual fusions

* fix env var name

* give fusions names

* fix build breaks

* fix perms

*  abort messages in matcher.cpp

* style fixes

* Isolated per op testing when comparing graphs for better determination of source of accuracy divergence.

* Improve clarity of comment

* Update README for PyPI

* Update README for PyPI

* Remove redundant newlines

* Fix links

* Uniform quotes style .

* Fix comment style.

* Check setup.py with flake8.

- Fix flake8 errors.

* Move function out of class scope.

* Fix function paramter list

* Fix formatting.

* add some helper function

* update with new helper function

* update reduce to nd with new helper function

* update float sum to stable sum

* fix bug

* update all reduce to stable sum for float

* fix bug and pass the sum stable test

* remove debug info

* style

* update with shape

* fix bug

* add host parameters to cuda_emitter

* clang format

* fix bugs

* add element::type support

* format

* add a cached value with datatype name

* add init_reduce_value

* unroll loop

* optimization

* remove the need for init_value

* add memset kernel

* add memcpy

* working version

* remove debug info

* add comments, clean up code.

* change in_idx to input_idx

* fix bug

* change args name for memset in emitter

* pass element::Type instead of string

* the op::reduce come with init value, add support

* resolve codacy-bot comment

* fix bug

* resove codacy-bot comment

* remove unused comments, resolve comments

* cuda reduce for max, min, mul, reduce op init value, format

* use type::info

* use type info for numeric_limits

* remove code from gpu_host_parameters

* header

* remvoe outdated comments

* add helper to check if stable sum is needed

* add stable sum test for double

* remove extra line

* consolidate helper functions

* no need list now.

* remove extra ;

* clang format

* style

* add skip test for cpu and intelGPU side

* add line between groups of headers

* add two simple stable sum test for float and double

* skip test for intelGPU

* - Added reorder support for rnn weights_layer/iter

* i) fixed compilation issues ii) working but still observing precision error

* i) fixed failing rnn unit test for DEX ii) refactored workspace in RNN mkldnn emitter

* i) added support for src reorder to TNC from NTC

* reorder support for rnn output fron NTC to TNC

* - added support for rnn weight reorder ldgoi -> ldigo
- code refactor for lstm/rnn kernel in mkldnn emitter

* - refactor rnn mkldnnn kernel, change variable names

* fix RNN codegen kernel

* disbale layer rnn fusion pass, to test CI

* method to validate recurrent rnn inputs

* add correlated macthes for Recurrent RNN PM

* - simplify reorder logic for rnn_weights
- fix graph pattern for fusing rnn cell across time steps

* do weights reorders in rnn timesteps fusion

* refactored LSTM graph pass

* - Bug fix for finding the lstm inputs determenstically
- Refactored LSTM graph pass to single pass
- made changes to LSTM RNN time step fusion graph pass

* - use replace_node instead of replace_output in Lstm_step_wise fusion graph pass

* fix compilation error

* Fix GNMT rnn fusion

* check if the node is in use before replacing in RNN graph passes

*  i) fix style ii) fix topo sort issue in RNN graph pass

* style fix

* fix bug in simplify_concat pass

* replaces Lstm1 -> {GOE1, GOE2} -> {Slice1, Slice2} -> Concat -> Lstm2 with Lstm1 -> Lstm2

* cse for convert layout

* addressed PR comments

* - optimization pass to remove  Lstm1 -> {GOE1, GOE2} -> {Slice1, Slice2} -> Lstm2
- conditional fusing of LSTM cells only for the decoder

* made changes to multi layer RNN fusion callback

* fix asserts in RNN op

* - added support to fuse layers when slc=dlc for RNN cells
- bug fix on the sanity checks for RNN Op

* - support RNN layer fusion till slc = dlc
- bug fixes in multi layer rnn fusion call back

* capture reshape in the RNN weights

* Addressed PR comments

* - added comments in multi layer PM call back
- fuse only if slc == DLC across layers

* restore deleted 3_lstm_cell_forward.json file

* fix typo

* fix failing unit tets

* When processing in place slice, do not change the offset of the slice node if the argument pointer comes from function input.

* Address PR feedback: process in place slice after propagating in place input.

* Set INTERMEDIATE role before propagating in place input.

* Do not add temporaries to the variable name map before propagating in place input in codegen.

* Fix a bug in codegen.

* Fix a bug in codegen slice.

* reenable disabled rnn unit test

* fix compiler error

* - bug fix in the slicing logic for the layer fused rnn cell
- fix failing rnn unit test

* - Addressed PR comments
- removed redundant checks from the rnn graph pass
- simplified rnn call back replace node logic

* - added new multilayer rnn *.json file
- fix test case

* [PRIVATE BRANCH] Style fixes (#2080)

* Style fixes

* change order of lstm gates

* [PRIVATE BRANCH] Jbobba/rnn fusion review (#2113)

* Style fixes for single-layer RNN fusion

* Style fixes to multi-layer RNN

* style fix

* disable GPU test

* fix fialing bn test

* fix style

* fix expand layout for padded dimensions

* enable squeeze padded layouts