* ADD NV JETSON SUPPORT
test=release/1.6

* CHERRY_PICK: specify the auto growth allocator for inference.
test=release/1.6

* add support to gcc8, add docker env
* remove the warning issue

* remove -Wmaybe-uninitialized warning

test=develop

* remove uninitialized op_handle_ in scale_loss_grad_op_handle.cc

test=develop

* Add support for dynamic code analysis (Sanitizers)
test=develop

* Move options to one option

test=develop

* Missing check

test=develop

test=develop

* HIP cmake.
Enable whole archieve build for pybind library.

Disable two warning.

Rollback to C++11.

Link RCCL to WA gpu kernel loading issue.

Update eigen to fix build failure.

Add more include directories.

Fix O3 build failure.

Update eigen.

fix tensor_util_test segment fault issue

add more macro check in hip.cmake.
we may consider refine hip.cmake to inherit all add_definitions() in parrent scope, in the future.

Fix rocRAND load.

Update eigen to fix gru_unit_op and reduce_op.

Add HIP support to testing.

Update eigen to support int16 and int8 in arg min and arg max.

* add rocprim as cub library used by nv implementation

* Reduce build time in rocprim.

* Add rocprim introduction, remove useless cmake code.

* Remove useless flags and format cmake file.

* flags

* "fix ci"

* cudnn widndows

* "add comment"

* "windows support"

* "fix cmake error"

1. Automatically detect GPU arch by default.
2. Specify -DCUDA_ARCH_NAME=All when releasing PaddlePaddle new version

to create a gradient machine for inference using merged model with parameters
which is genearted by `paddle merge_model`.