1 ===================================================================
2 How To Cross-Compile Clang/LLVM using Clang/LLVM
3 ===================================================================
8 This document contains information about building LLVM and
9 Clang on host machine, targeting another platform.
11 For more information on how to use Clang as a cross-compiler,
12 please check https://clang.llvm.org/docs/CrossCompilation.html.
14 TODO: Add MIPS and other platforms to this document.
16 Cross-Compiling from x86_64 to ARM
17 ==================================
19 In this use case, we'll be using CMake and Ninja, on a Debian-based Linux
20 system, cross-compiling from an x86_64 host (most Intel and AMD chips
21 nowadays) to a hard-float ARM target (most ARM targets nowadays).
23 The packages you'll need are:
26 * ``ninja-build`` (from backports in Ubuntu)
27 * ``gcc-4.7-arm-linux-gnueabihf``
28 * ``gcc-4.7-multilib-arm-linux-gnueabihf``
29 * ``binutils-arm-linux-gnueabihf``
30 * ``libgcc1-armhf-cross``
31 * ``libsfgcc1-armhf-cross``
32 * ``libstdc++6-armhf-cross``
33 * ``libstdc++6-4.7-dev-armhf-cross``
38 For more information on how to configure CMake for LLVM/Clang,
41 The CMake options you need to add are:
43 * ``-DCMAKE_SYSTEM_NAME=<target-system>``
44 * ``-DCMAKE_INSTALL_PREFIX=<install-dir>``
45 * ``-DLLVM_TABLEGEN=<path-to-host-bin>/llvm-tblgen``
46 * ``-DCLANG_TABLEGEN=<path-to-host-bin>/clang-tblgen``
47 * ``-DLLVM_DEFAULT_TARGET_TRIPLE=arm-linux-gnueabihf``
48 * ``-DLLVM_TARGET_ARCH=ARM``
49 * ``-DLLVM_TARGETS_TO_BUILD=ARM``
51 Note: ``CMAKE_CROSSCOMPILING`` is always set automatically when ``CMAKE_SYSTEM_NAME`` is set. Don't put ``-DCMAKE_CROSSCOMPILING=TRUE`` in your options.
53 If you're compiling with GCC, you can use architecture options for your target,
54 and the compiler driver will detect everything that it needs:
56 * ``-DCMAKE_CXX_FLAGS='-march=armv7-a -mcpu=cortex-a9 -mfloat-abi=hard'``
58 However, if you're using Clang, the driver might not be up-to-date with your
59 specific Linux distribution, version or GCC layout, so you'll need to fudge.
61 In addition to the ones above, you'll also need:
63 * ``--target=arm-linux-gnueabihf`` or whatever is the triple of your cross GCC.
64 * ``'--sysroot=/usr/arm-linux-gnueabihf'``, ``'--sysroot=/opt/gcc/arm-linux-gnueabihf'``
65 or whatever is the location of your GCC's sysroot (where /lib, /bin etc are).
66 * Appropriate use of ``-I`` and ``-L``, depending on how the cross GCC is installed,
67 and where are the libraries and headers.
69 The TableGen options are required to compile it with the host compiler,
70 so you'll need to compile LLVM (or at least ``llvm-tblgen``) to your host
71 platform before you start. The CXX flags define the target, cpu (which in this case
72 defaults to ``fpu=VFP3`` with NEON), and forcing the hard-float ABI. If you're
73 using Clang as a cross-compiler, you will *also* have to set ``--sysroot``
74 to make sure it picks the correct linker.
76 When using Clang, it's important that you choose the triple to be *identical*
77 to the GCC triple and the sysroot. This will make it easier for Clang to
78 find the correct tools and include headers. But that won't mean all headers and
79 libraries will be found. You'll still need to use ``-I`` and ``-L`` to locate
80 those extra ones, depending on your distribution.
82 Most of the time, what you want is to have a native compiler to the
83 platform itself, but not others. So there's rarely a point in compiling
84 all back-ends. For that reason, you should also set the
85 ``TARGETS_TO_BUILD`` to only build the back-end you're targeting to.
87 You must set the ``CMAKE_INSTALL_PREFIX``, otherwise a ``ninja install``
88 will copy ARM binaries to your root filesystem, which is not what you
94 There are some bugs in current LLVM, which require some fiddling before
97 #. If you're using Clang as the cross-compiler, there is a problem in
98 the LLVM ARM back-end that is producing absolute relocations on
99 position-independent code (``R_ARM_THM_MOVW_ABS_NC``), so for now, you
104 -DLLVM_ENABLE_PIC=False
106 This is not a problem, since Clang/LLVM libraries are statically
107 linked anyway, it shouldn't affect much.
109 #. The ARM libraries won't be installed in your system.
110 But the CMake prepare step, which checks for
111 dependencies, will check the *host* libraries, not the *target*
112 ones. Below there's a list of some dependencies, but your project could
113 have more, or this document could be outdated. You'll see the errors
114 while linking as an indication of that.
116 Debian based distros have a way to add ``multiarch``, which adds
117 a new architecture and allows you to install packages for those
118 systems. See https://wiki.debian.org/Multiarch/HOWTO for more info.
120 But not all distros will have that, and possibly not an easy way to
121 install them in any anyway, so you'll have to build/download
124 A quick way of getting the libraries is to download them from
125 a distribution repository, like Debian (http://packages.debian.org/jessie/),
126 and download the missing libraries. Note that the ``libXXX``
127 will have the shared objects (``.so``) and the ``libXXX-dev`` will
128 give you the headers and the static (``.a``) library. Just in
131 The ones you need for ARM are: ``libtinfo``, ``zlib1g``,
132 ``libxml2`` and ``liblzma``. In the Debian repository you'll
133 find downloads for all architectures.
135 After you download and unpack all ``.deb`` packages, copy all
136 ``.so`` and ``.a`` to a directory, make the appropriate
137 symbolic links (if necessary), and add the relevant ``-L``
138 and ``-I`` paths to ``-DCMAKE_CXX_FLAGS`` above.
141 Running CMake and Building
142 --------------------------
144 Finally, if you're using your platform compiler, run:
148 $ cmake -G Ninja <source-dir> -DCMAKE_BUILD_TYPE=<type> <options above>
150 If you're using Clang as the cross-compiler, run:
154 $ CC='clang' CXX='clang++' cmake -G Ninja <source-dir> -DCMAKE_BUILD_TYPE=<type> <options above>
156 If you have ``clang``/``clang++`` on the path, it should just work, and special
157 Ninja files will be created in the build directory. I strongly suggest
158 you to run ``cmake`` on a separate build directory, *not* inside the
161 To build, simply type:
167 It should automatically find out how many cores you have, what are
168 the rules that needs building and will build the whole thing.
170 You can't run ``ninja check-all`` on this tree because the created
171 binaries are targeted to ARM, not x86_64.
176 After the LLVM/Clang has built successfully, you should install it
183 which will create a sysroot on the install-dir. You can then tar
184 that directory into a binary with the full triple name (for easy
185 identification), like:
189 $ ln -sf <install-dir> arm-linux-gnueabihf-clang
190 $ tar zchf arm-linux-gnueabihf-clang.tar.gz arm-linux-gnueabihf-clang
192 If you copy that tarball to your target board, you'll be able to use
193 it for running the test-suite, for example. Follow the guidelines at
194 https://llvm.org/docs/lnt/quickstart.html, unpack the tarball in the
195 test directory, and use options:
199 $ ./sandbox/bin/python sandbox/bin/lnt runtest nt \
201 --test-suite `pwd`/test-suite \
202 --cc `pwd`/arm-linux-gnueabihf-clang/bin/clang \
203 --cxx `pwd`/arm-linux-gnueabihf-clang/bin/clang++
205 Remember to add the ``-jN`` options to ``lnt`` to the number of CPUs
206 on your board. Also, the path to your clang has to be absolute, so
207 you'll need the `pwd` trick above.