You can use the toolchains provided with the Android NDK independently, or as plug-ins with an existing IDE. This flexibility can be useful if you already have your own build system, and only need the ability to invoke the cross-compiler in order to add support to Android for it.
A typical use case is invoking the configure script of an open-source library that expects a cross-compiler in the CC
environment variable.
Note: This page assumes significant understanding of compiling, linking, and low-level architecture. In addition, the techniques it describes are unnecessary for most use cases. In most cases, we recommend that you forego using a standalone toolchain, and instead stick to the NDK build system.
Before anything else, you need to decide which processing architecture your standalone toolchain is going to target. Each architecture corresponds to a different toolchain name, as Table 1 shows.
Architecture | Toolchain name |
---|---|
ARM-based | arm-linux-androideabi-<gcc-version> |
x86-based | x86-<gcc-version> |
MIPS-based | mipsel-linux-android-<gcc-version> |
ARM64-based | aarch64-linux-android-<gcc-version> |
X86-64-based | x86_64-<gcc-version> |
MIPS64-based | mips64el-linux-android--<gcc-version> |
The next thing you need to do is define your sysroot (A sysroot is a directory containing the system headers and libraries for your target). To define the sysroot, you must must know the Android API level you want to target for native support; available native APIs vary by Android API level.
Native APIs for the respective Android API levels reside under $NDK/platforms/
; each API-level directory, in turn, contains subdirectories for the various CPUs and architectures. The following example shows how to define a sysroot for a build targeting Android 5.0 (API level 21), for ARM architecture:
SYSROOT=$NDK/platforms/android-21/arch-armFor more detail about the Android API levels and the respective native APIs they support, see Android NDK Native APIs .
There are two ways to invoke the compiler. One method is simple, and leaves most of the lifting to the build system. The other is more advanced, but provides more flexibility.
The simplest way to build is by invoking the appropriate compiler directly from the command line, using the --sysroot
option to indicate the location of the system files for the platform you're targeting. For example:
export CC="$NDK/toolchains/arm-linux-androideabi-4.8/prebuilt/ \ linux-x86/bin/arm-linux-androideabi-gcc-4.8 --sysroot=$SYSROOT" $CC -o foo.o -c foo.c
While this method is simple, it lacks in flexibility: It does not allow you to use any C++ STL (STLport, libc++, or the GNU libstdc++) with it. It also does not support exceptions or RTTI.
For Clang, you need to perform an additional two steps:
-target
for the target architecture, as Table 2 shows.
Architecture | Value |
---|---|
armeabi | -target armv5te-none-linux-androideabi |
armeabi-v7a | -target armv7-none-linux-androideabi |
arm64-v8a | -target aarch64-none-linux-android |
x86 | -target i686-none-linux-android |
x86_64 | -target x86_64-none-linux-android |
mips | -target mipsel-none-linux-android |
-gcc-toolchain
option, as in the following example: -gcc-toolchain $NDK/toolchains/arm-linux-androideabi-4.8/prebuilt/linux-x86_64
export CC="$NDK/toolchains/arm-linux-androideabi-4.8/prebuilt/ \ linux-x86/bin/arm-linux-androideabi-gcc-4.8 --sysroot=$SYSROOT" -target \ armv7-none-linux-androideabi \ -gcc-toolchain $NDK/toolchains/arm-linux-androideabi-4.8/prebuilt/linux-x86_64" $CC -o foo.o -c foo.c
The NDK provides the make-standalone-toolchain.sh
shell script to allow you to perform a customized toolchain installation from the command line. This approach affords you more flexibility than the procedure described in Simple method.
The script is located in the $NDK/build/tools/
directory, where $NDK
is the installation root for the NDK. An example of the use of this script appears below:
$NDK/build/tools/make-standalone-toolchain.sh \ --arch=arm --platform=android-21 --install-dir=/tmp/my-android-toolchain
This command creates a directory named /tmp/my-android-toolchain/
, containing a copy of the android-21/arch-arm
sysroot, and of the toolchain binaries for a 32-bit ARM architecture.
Note that the toolchain binaries do not depend on or contain host-specific paths, in other words, you can install them in any location, or even move them if you need to.
By default, the build system uses the 32-bit, ARM-based GCC 4.8 toolchain. You can specify a different value, however, by specifying --arch=<toolchain>
as an option. Table 3 shows the values to use for other toolchains:
Toolchain | Value |
---|---|
mips64 compiler | --arch=mips64 |
mips GCC 4.8 compiler | --arch=mips |
x86 GCC 4.8 compiler | --arch=x86 |
x86_64 GCC 4.8 compiler | --arch=x86_64 |
mips GCC 4.8 compiler | --arch=mips |
Alternatively, you can use the --toolchain=<toolchain>
option. Table 4 shows the values you can specify for<toolchain>
:
Toolchain | Value |
---|---|
arm | --toolchain=arm-linux-androideabi-4.8 --toolchain=arm-linux-androideabi-4.9 --toolchain=arm-linux-android-clang3.5 --toolchain=arm-linux-android-clang3.6 |
x86 | --toolchain=x86-linux-android-4.8 --toolchain=x86-linux-android-4.9 --toolchain=x86-linux-android-clang3.5 --toolchain=x86-linux-android-clang3.6 |
mips | --toolchain=mips-linux-android-4.8 --toolchain=mips-linux-android-4.9 --toolchain=mips-linux-android-clang3.5 --toolchain=mips-linux-android-clang3.6 |
arm64 | --toolchain=aarch64-linux-android-4.9 --toolchain=aarch64-linux-android-clang3.5 --toolchain=aarch64-linux-android-clang3.6 |
x86_64 | --toolchain=x86_64-linux-android-4.9 --toolchain=x86_64-linux-android-clang3.5 --toolchain=x86_64-linux-android-clang3.6 |
mips64 | --toolchain=mips64el-linux-android-4.9 --toolchain=mips64el-linux-android-clang3.5 --toolchain=mips64el-linux-android-clang3.6 |
Note: Table 4 is not an exhaustive list. Other combinations may also be valid, but are unverified.
You can also copy Clang/LLVM 3.6, using one of two methods: You can append -clang3.6
to the --toolchain
option, so that the --toolchain
option looks like the following example:
--toolchain=arm-linux-androideabi-clang3.6
You can also add -llvm-version=3.6
as a separate option on the command line.
Note: Instead of specifying a specific version, you can also use <version>
, which defaults to the highest available version of Clang.
By default, the build system builds for a 32-bit host toolchain. You can specify a 64-bit host toolchain instead. Table 5 shows the value to use with -system
for different platforms.
Host toolchain | Value |
---|---|
64-bit Linux | -system=linux-x86_64 |
64-bit MacOSX | -system=darwin-x86_64 |
64-bit Windows | -system=windows-x86_64 |
You may specify --stl=stlport
to copy libstlport
instead of the default libgnustl
. If you do so, and you wish to link against the shared library, you must explicitly use -lstlport_shared
. This requirement is similar to having to use -lgnustl_shared
for GNU libstdc++
.
Similarly, you can specify --stl=libc++
to copy the LLVM libc++ headers and libraries. To link against the shared library, you must explicitly use -lc++_shared.
You can make these settings directly, as in the following example:
export PATH=/tmp/my-android-toolchain/bin:$PATH export CC=arm-linux-androideabi-gcc # or export CC=clang export CXX=arm-linux-androideabi-g++ # or export CXX=clang++
Note that if you omit the -install-dir
option, the make-standalone-toolchain.sh
shell script creates a tarball in tmp/ndk/<toolchain-name>.tar.bz2
. This tarball makes it easy to archive, as well as to redistribute the binaries.
This standalone toolchain provides an additional benefit, as well, in that it contains a working copy of a C++ STL library, with working exceptions and RTTI support.
For more options and details, use --help
.
You can install Clang binaries in the standalone installation by using the --llvm-version=<version>
option.<version>
is a LLVM/Clang version number, such as 3.5
or 3.6
. For example:
build/tools/make-standalone-toolchain.sh \ --install-dir=/tmp/mydir \ --toolchain=arm-linux-androideabi-4.8 \ --llvm-version=3.6
Note that Clang binaries are copied along with the GCC ones, because they rely on the same assembler, linker, headers, libraries, and C++ STL implementation.
This operation also installs two scripts, named clang
and clang++
, under <install-dir>/bin/@
. These scripts invoke the real clang
binary with default target architecture flags. In other words, they should work without any modification, and you should be able to use them in your own builds by just setting the CC
andCXX
environment variables to point to them.
In an ARM standalone installation built with llvm-version=3.6
, invoking Clang on a Unix system takes the form of a single line. For instance:
`dirname $0`/clang36 -target armv5te-none-linux-androideabi "$@"
clang++
invokes clang++31
in the same way.
When building for ARM, Clang changes the target based on the presence of the -march=armv7-a
and/or -mthumb
options:
-march value |
Resulting target |
---|---|
-march=armv7-a |
armv7-none-linux-androideabi |
-mthumb |
thumb-none-linux-androideabi |
Both -march=armv7-a and -mthumb |
thumbv7-none-linux-androideabi |
You may also override with your own -target
if you wish.
The -gcc-toolchain
option is unnecessary because, in a standalone package, Clang locates as
and ld
in a predefined relative location.
clang
and clang++
should be easy drop-in replacements for gcc
and g++
in a makefile. When in doubt, add the following options to verify that they are working properly:
-v
to dump commands associated with compiler driver issues-###
to dump command line options, including implicitly predefined ones.-x c < /dev/null -dM -E
to dump predefined preprocessor definitions-save-temps
to compare *.i
or *.ii
preprocessed files.For more information about Clang, see http://clang.llvm.org/, especially the GCC compatibility section.
The machine code that the ARM toolchain generates should be compatible with the official Android armeabi
ABI by default.
We recommend use of the -mthumb
compiler flag to force the generation of 16-bit Thumb-1 instructions (the default being 32-bit ARM instructions).
If you want to target the armeabi-v7a ABI, you must set the following flags:
CFLAGS= -march=armv7-a -mfloat-abi=softfp -mfpu=vfpv3-d16
The first flag enables Thumb-2 instructions. The second flag enables hardware-FPU instructions while ensuring that the system passes floating-point parameters in core registers, which is critical for ABI compatibility.
Note: In versions of the NDK prior to r9b, do not use these flags separately. You must set all or none of them. Otherwise, unpredictable behavior and crashes may result.
To use NEON instructions, you must change the -mfpu
compiler flag:
CFLAGS= -march=armv7-a -mfloat-abi=softfp -mfpu=neon
Note that this setting forces the use of VFPv3-D32
, per the ARM specification.
Also, make sure to provide the following two flags to the linker:
LDFLAGS= -march=armv7-a -Wl,--fix-cortex-a8
The first flag instructs the linker to pick libgcc.a
, libgcov.a
, and crt*.o
, which are tailored for armv7-a. The 2nd flag is required as a workaround for a CPU bug in some Cortex-A8 implementations.
Since NDK version r9b, all Android native APIs taking or returning double or float values haveattribute((pcs("aapcs")))
for ARM. This makes it possible to compile user code in -mhard-float
(which implies -mfloat-abi=hard
), and still link with the Android native APIs that comply with the softfp ABI. For more information on this, see the comments in $NDK/tests/device/hard-float/jni/Android.mk
.
If you want to use NEON intrinsics on x86, the build system can translate them to the native x86 SSE intrinsics using a special C/C++ language header with the same name, arm_neon.h
, as the standard ARM NEON intrinsics header.
By default, the x86 ABI supports SIMD up to SSSE3, and the header covers ~93% of (1869 of 2009) NEON functions.
You don't have to use any specific compiler flag when targeting the MIPS ABI.
To learn more about ABI support, see x86 Support.
The Windows binaries do not depend on Cygwin. This lack of dependency makes them faster. The cost, however, is that they do not understand Cygwin path specifications like cygdrive/c/foo/bar
, as opposed toC:/foo/bar
.
The NDK build system ensures that all paths passed to the compiler from Cygwin are automatically translated, and manages other complexities, as well. If you have a custom build system, you may need to resolve these complexities yourself.
For information on contributing to support for Cygwin/MSys, visit the android-ndk forum.
The Android platform did not really support wchar_t
until Android 2.3 (API level 9). This fact has several ramifications:
wchar_t
is 4 bytes, and most wide-char
functions are available in the C library (with the exception of multi-byte encoding/decoding functions andwsprintf
/wsscanf
).wchar_t
is 1 byte, and none of the wide-char functions works. We recommend that you get rid of any dependencies on the wchar_t
type, and switch to better representations. The support provided in Android is only there to help you migrate existing code.
The toolchain binaries support C++ exceptions and RTTI by default. To disable C++ exceptions and RTTI when building sources (to generate lighter-weight machine code, for example), use -fno-exceptions
and -fno-rtti
.
To use these features in conjunction with GNU libstdc++, you must explicitly link with libsupc++. To do so, use-lsupc++
when linking binaries. For example:
arm-linux-androideabi-g++ .... -lsupc++
You do not need to do this when using the STLport or libc++ library.
The standalone toolchain includes a copy of a C++ Standard Template Library implementation. This implementation is either for GNU libstdc++, STLport, or libc++, depending on what you specify for the --stl=<name>
option described previously. To use this implementation of STL, you need to link your project with the proper library: