Add instructions for building

README.md

@@ -44,6 +44,12 @@ The following is the comparison between our dabnn and [Caffe](http://caffe.berke
 
 ![Comparison](images/comparison_en.png)
 
+## Build
+
+We provide pre-built onnx2bnn and also dabnn Android package. However, you need to build it if you want to deploy BNNs on non-Android ARM devices.
+
+We use CMake build system like most C++ projects. Check out [docs/build.md](docs/build.md) for the detail instructions.
+
 ## Convert ONNX Model
 
 We provide a conversion tool, named onnx2bnn, to convert an ONNX model to a dabnn model. We provide onnx2bnn pre-built binaries for all platforms in [GitHub Releases](https://github.com/JDAI-CV/dabnn/releases). For Linux users, the onnx2bnn pre-built binary is [AppImage](https://appimage.org) format, see https://appimage.org for details.

README_CN.md

@@ -46,6 +46,12 @@ dabnn_bireal18_imagenet_stem   43279353 ns   41533009 ns         14       <--- 
 
 ![Comparison](images/comparison_cn.png)
 
+## 如何编译
+
+我们已经提供预编译的 onnx2bnn 和 dabnn 的 Android package，不过如果需要在非 Android 的 ARM 设备上使用 dabnn，就需要自己编译它了。
+
+如同大部分 C++ 项目，我们使用 CMake 构建系统。具体请看 [docs/build.md](docs/build.md)。
+
 ## 如何转换 ONNX 模型
 
 我们提供模型转换工具 onnx2bnn 将 ONNX 模型转换为 dabnn 格式的模型。在 [GitHub Releases](https://github.com/JDAI-CV/dabnn/releases) 里有各个平台的 onnx2bnn 预编译二进制文件，可以直接下载运行。Linux 用户我们提供的是 AppImage 格式的二进制文件，AppImage 的使用方法和其它相关信息请参考 https://appimage.org/。

docs/build.md

@@ -0,0 +1,51 @@
+## Build
+
+We use CMake build system like most C++ projects. 
+
+### Build dabnn
+
+Cross-compiling for ARMv8 Android:
+
+1. Download and unzip Android NDK from https://developer.android.com/ndk/downloads.
+
+1. Run cmake with the toolchain file, which determine the proper compiling toolchains.
+
+```bash
+mkdir build-dabnn
+cd build-dabnn
+cmake -DCMAKE_TOOLCHAIN_FILE=the_path_to_android_ndk/build/cmake/android.toolchain.cmake -DANDROID_ABI=arm64-v8a -DCMAKE_BUILD_TYPE=Release ..
+```
+
+**If this step fails, please check whether the toolchain file really exists on the path you set.**
+
+1. Build
+
+```bash
+cmake --build .
+```
+
+For ARMv7, just replace `-DANDROID_ABI=arm64-v8a` with `-DANDROID_ABI=armeabi-v7a` in step 2.
+
+For non-Android ARM devices, use the proper toolchain file for your device instead of the Android NDK toolchain file, or compile natively on your ARM device.
+
+For non-ARM devices, only the unoptimized code will work. If you still want to build dabnn for non-ARM devices, pass `-DBNN_BUILD_MAIN_LIB=ON` in step 2.
+
+### Build onnx2bnn
+
+On non-ARM devices, 
+
+1. Run cmake
+
+```bash
+mkdir build-onnx2bnn
+cd build-onnx2bnn
+cmake ..
+```
+
+1. Build
+
+```bash
+cmake --build .
+```
+
+On ARM devices, pass `-DBNN_BUILD_MAIN_LIB=OFF` in step 1.