Import HPC linear algebra libs as cblas target for cc_library deps

Use PaddlePaddle fork of any

Refine unit test in op_test_util

"change device context to pointer"

It can be run both CPU/GPU. configure attributes are:

* min: the min value of uniform random
* max: the max value of uniform random
* dims: the dimension of output tensor
* seed: the random seed of uniform random. 0 means generate a seed each
        time.

remove global linker and exe from mkldnn iomp

Remove paddle/utils/Flags.h warnings

Support groups in NNPACKFunction.

Fix comparison between signed and unsigned values

* init cross entropy graident

* add crossentropy grad op

* remove details

* fix static compile

Add environment variable WITH_DEB to trigger the build of .deb file

add gradient check for sequence softmax activation.

Only make install when it necessary in build.sh

* move namespace rnn to a directory

Faster Python unittest

* User must install Paddle python package before unittest.
* Or use docker to build Paddle

ClangFormat for proto and cuda

Simplify CrossMapNormalOpTest and BlockExpandOpTest.

use isclose instead of assert_almost_equal