* Add the first implememtation of fusion_group op #19621 (#3)

* Add the dynamic load of nvrtc, and support runtime compiling of CUDA kernel using nvrtc.
test=develop

* Call CUDA driver api to launch the kernel compiled by nvrtc.
test=develop

* Disable for mac and windows.
test=develop

* Refine the codes to support manually specified num_threads and workload_per_thread.
test=develop

* Refine the CUDA kernel to support large dims.
test=develop

* Add DeviceCodePool to manage all device codes.

* Add the first implementation fusion_group op.

* Add unit-test for fusion_group op.

* Add the check of result.

* Add the check of nvrtc in unit-test.
test=develop

* Add comment to explain the inputs, outputs and features of fusion_group op.
test=develop

* Disable fusion_group op for mac and windows.
test=develop

* Make the compiling of device code return status instead of hanging up.
test=develop

* Add the check of whether there is CUDA driver library, and do not core dump when failing to call the CUDA driver API.

* Unify fusion_group_op's input and output names.
test=develop

* Add the check of CUDA driver library in unittest.
test=develop

* Enable generating code for a given subgraph. #21126 (#4)

* Enable generating code for a given subgraph.

* Support sorting the subgraph.

* Remove the rearange of expressions because we use the sorted subgraph directly.

* Enable generating code for a subgraph which is composed of grad ops.

* Use expression information to check the accuracy in unittest.

* Separate load and store from computation expressions.
test=develop

* Improve the loading statements in generated codes.
test=develop

* Remove unused arguments from formal list.
test=develop

* Enable the detection of subgraph of grad ops.

* Generate code for detected subgraph in fusion_group_pass.

* Add an option in BuildStrategy to enable fusion_group_pass and add unittest.
test=develop

* Fix a bug when checking whether the shape of all inputs are the same.

* Add debug information.

* Remove subgraph_detector from inference/analysis to the common framework/ir directory. (#5)

test=develop

* Call subgraph_detector in fusion_group pass.
test=develop

* Disable fusion_group when WITH_GPU is OFF.
test=develop

* Refine all PADDLE_ENFORCE message.
test=develop

* Fix the case that some inputs are not defined in grad ops, and set op_role for fused op.
test=develop

* Follow review comments.
test=develop

* add bn and relu fuse pass

* add op attr assert and dtype assert

* fix some inputs&&outputs bugs for the fused op and pattern.

* add the unittest for fuse_bn_act_pass. test=develop

* use normative enforce statements. test=develop

* add the cpu test. test=develop

* add the support of batch_size=1 for the bn with relu op. test=develop

* add the error type for paddle throws. test=develop

* add fused_batch_norm_act and fused_batch_norm_act_grad to op_has_unsed_vars_white_list. test=develop

Add examples for error message writing specification - PreconditionNotMet, Unimplemented, Unavailable (#21137)

* add examples for error spec, test=develop

* change ENFORCE to ENFORCE_**, test=develop

* fix warning info
test=develop

* fix bug of all_reduce_deps_pass
test=develop

* Open fuse all reduce op
test=develop

* Add Fuse optimization op log

* Add log in fuse_optimizer op pass and fuse all_reduce op pass

* replace with boost::optional<bool>
test=develop

* Polish code
test=develop

* fix code coverage
test=develop

* fix vlog level and fuse option type
test=develop

* Enable fused_all_reduce_op_handle support GPU and CPU Gradients

* enable ngraph throught build_strategy test=develop

* add unittest test=develop

* put use_ngraph unconditional test=develop

* remove paddle_enforce test=develop

* remove paddle_enforce test=develop

* fix copyright test=develop

* limit for ngraph only test=develop

* Disable fuse optimization
test=develop

* remove legacy memory optimization codes, test=develop

* follow huihuang's comments,test=develop

* follow luotao's comments, test=develop

* open fuse optimization ops
test=develop

* first version memory optimize pass, test=develop

* remove move_tensor_sharing_pass, test=develop

* refine code comments, add unittests, test=develop

* turn off memory_optimize by default, test=develop

* follow huihuang's comments, test=develop

* follow chengduoZH's comments, test=develop

* fix grammar error, add const qualifier, fix pass_test exception message, test=develop

* follow chengduoZH's comments 2nd, test=develop

* support sparse gradients
test=develop

* feature/buffer_shared_inplace, test=develop

* refine code, test=develop

* fix elementwise_add op cpu inplace and sum inplace bug, test=develop

* add unittest and debug log, test=develop

* fix parallel_executor scope bug, polish code, test=develop

* fix sum op, activation op, single_in_place_inference bug, test=develop

* remove kLocalExecScopeName, test=develop

* fix unittest,test=develop

* fix out_var first version bug, test=develop

* follow comments,test=develop

test=develop

* make parallel_executor support FLAGS_use_mkldnn

test=develop

* add warning when set mkldnn_enabled_op_types_ in non-mkldnn env

test=develop

* move pass to ir

* polish code
test=develop

* fix dependency
test=develop

* add use_cuda to inplace pass,test=develop

* add test softmax_with_xe_inplace test,test=develop

* Add fuse momenutum ops

* speedup gc and inplace softmax_with_cross_entropy_grad
test=develop

* refine models gpu mem
Merge skip vars and warning messages of mem opt
remove relu mem opt
test=develop

* follow comments
test=develop

* fix bug of fuse optimize ops

* Add an option to enable the cache of expected kernel in train phase.
test=develop

* Change the default value of cache_expected_kernel to true.

* Try to enable the runtime_context_cache pass in train phase.

* Put the append of runtime_context_cache pass ahead of multi_dev passes.
test=develop

* fuse optimizer

* modify the interface of Pass::Allay
test=develop

* Polish code
test=develop

* Fix Travis CI
test=develop

* fix Pass::Apply interface
test=develop

* Fix Travis CI
test=develop

* refine parallelExecutor
test=develop

* Polish op_handle
test=develop

* Remove unnecessary op_handle
test=develop

* Fix Travis CI
test=develop

* Fix fetch bug
test=develop

* Remove WaitInputVarGenerated

* Fix OpHandleBase::Run
test=develop

* debug
test=develop

* use origin fetch_op_handle
test=develop

* Revert op_handle_base.cc
test=develop

* Polish code
test=develop

* Fix OpHandleBase::Run
test=develop

* code refine

* test CI and CE
test=develop

* fix OpHandle::Run
test=develop

* refine AllReduceOpHandle
test=develop

* Polish code
test=develop

* fuse all_reduce
test=develop

* add fuse_parameter_groups_size
test=develop

* Polish code
test=develop

* Fix travis-ci
test=develop

* Add SetGroupAccordingToLayers and SetGroupAccordingToGroupSize
test=develop

* Add SetGroupAccordingToMemorySize
test=develop

* fix multi_devices_graph
test=develop

* reset params_grads
test=develop

* Polish code
test=develop

* Support Sync Batch Norm.
* Note, do not enable it in one device.

Usage:

build_strategy = fluid.BuildStrategy()
build_strategy.sync_batch_norm = True
binary = fluid.compiler.CompiledProgram(tp).with_data_parallel(
        loss_name=loss_mean.name,
        build_strategy=build_strategy)

test=develop

test=develop

test=develop