[ Doc ] add how to run android application

Add documentation to build and run in android

**Self evaluation:**
1. Build test:	 [X]Passed [ ]Failed [ ]Skipped
2. Run test:	 [X]Passed [ ]Failed [ ]Skipped

Signed-off-by: jijoong.moon <jijoong.moon@samsung.com>

diff --git a/docs/how-to-run-example-android.md b/docs/how-to-run-example-android.md
new file mode 100644 (file)
index 0000000..f9cf78f
--- /dev/null
+++ b/docs/how-to-run-example-android.md
@@ -0,0 +1,86 @@
+---
+title: How to run examples - Android
+...
+
+# How to run examples - Android
+
+### Install NDK-Build
+
+Prepare NDK tool chain to build.
+Download the ndk tool chain from official site of android which is https://developer.android.com/ndk/downloads
+
+Choose the ndk package which is proper to you platform. In this guide, we use Linux 64-bit(x86). We choose latest version (currently r21d).
+Once the download is finished, decompress the package and set the library path properly. You can also set the LD_LIBRARAY_PATH in bashrc.
+
+```bash
+$ ls
+android-ndk-r21d-linux-x86_64.zip
+$ unzip android-ndk-r21d-linux-86_64.zip
+$ ls
+android-ndk-r21d
+$ export LD_LIBRARY_PATH = $LD_LIBRARY_PATH:${PWD}/android-ndk-r21d/
+```
+
+### Build NNTrainer with NDK-Build
+Once you install NDK package, you can build the nntrainer.so for android as below.
+
+```bash
+$ ls
+api                 CONTRIBUTING.md  index.md  MAINTAINERS.md     nnstreamer       packaging
+Applications        debian           jni       meson.build        nntrainer        README.md
+CODE_OF_CONDUCT.md  docs             LICENSE   meson_options.txt  nntrainer.pc.in  test
+$
+$ export NNTRAINER_ROOT=${PWD}
+$
+$ cd ./jni
+Android.mk      prepare_iniparser.sh  prepare_tflite.sh
+Application.mk  prepare_openblas.sh
+$ ndk-build NDK_PROJECT_PATH=./ APP_BUILD_SCRIPT=./Android.mk NDK_APPLICATION_MK=./Application.mk -j $(nproc)
+Android.mk:19: INIPARSER_ROOT is not defined!
+Android.mk:20: INIPARSER SRC is going to be downloaded!
+Cloning into 'iniparser'...
+~/nntrainer_new/jni ~/nntrainer_new/jni PREPARING ini parser at . ~/nntrainer_new/jni
+Android.mk:35: BUILDING TFLITE BACKBONE !
+Android.mk:40: TENSORFLOW_ROOT is not defined!
+Android.mk:41: TENSORFLOW SRC is going to be downloaded!
+PREPARING TENSORFLOW 1.13.1 at . ~/nntrainer_new/jni ~/nntrainer_new/jni [TENSORFLOW-LITE] Download tensorflow-1.13.1 [TENSORFLOW-LITE] Finish downloading tensorflow-1.13.1 [TENSORFLOW-LITE] untar tensorflow-1.13.1 ~/nntrainer_new/jni
+[arm64-v8a] Compile++      : nntrainer <= tflite_layer.cpp
+In file included from ././../nntrainer/layers/tflite_layer.cpp:15:
+In file included from ./../nntrainer/layers/tflite_layer.h:18:
+In file included from ./../nntrainer/layers/layer_internal.h:31:
+...
+$ ls
+Android.mk      iniparser  obj                   prepare_openblas.sh  tensorflow-1.13.1
+Application.mk  libs       prepare_iniparser.sh  prepare_tflite.sh
+$ ls libs/arm64-v8a
+libcapi-nntrainer.so  libccapi-nntrainer.so  libc++_shared.so  libnntrainer.so
+$ mv libs ../
+```
+
+### Build NNTrainer Applications with NDK-Build
+Now you are ready to build nntrainer application in ${NNTRAINER_ROOT}/Applications
+If you want to build Application/LogisticRegration,
+
+```bash
+$ cd ${NNTRAINER_ROOT}
+$ cd Applications/LogisticRegression/jni
+$ ls
+Android.mk  Application.mk  CMakeLists.txt  main.cpp  meson.build
+$ ndk-build NDK_PROJECT_PATH=./ APP_BUILD_SCRIPT=./Android.mk NDK_APPLICATION_MK=./Application.mk -j $(nproc)
+[arm64-v8a] Prebuilt       : libnntrainer.so <= /libs/arm64-v8a/
+[arm64-v8a] Prebuilt       : libc++_shared.so <= <NDK>/sources/cxx-stl/llvm-libc++/libs/arm64-v8a/
+[arm64-v8a] Install        : libnntrainer.so => libs/arm64-v8a/libnntrainer.so
+[arm64-v8a] Install        : libc++_shared.so => libs/arm64-v8a/libc++_shared.so
+[arm64-v8a] Compile++      : nntrainer_logistic <= main.cpp
+In file included from ./main.cpp:36:
+...
+$ ls
+Android.mk  Application.mk  CMakeLists.txt  libs  main.cpp  meson.build  obj
+$ ls libs/arm64-v8a
+libc++_shared.so  libnntrainer.so  nntrainer_logistic
+```
+
+Now you can find the execution binary of nntrainer_logistic.
+You can copy execution binary and nntrainer.so libraries in a proper place of your android device using ADB.
+
+Then you can run it using ADB shell.