1====================================
2Getting Started with the LLVM System
3====================================
4
5.. contents::
6   :local:
7
8Overview
9========
10
11Welcome to the LLVM project!
12
13The LLVM project has multiple components. The core of the project is
14itself called "LLVM". This contains all of the tools, libraries, and header
15files needed to process intermediate representations and converts it into
16object files.  Tools include an assembler, disassembler, bitcode analyzer, and
17bitcode optimizer.  It also contains basic regression tests.
18
19C-like languages use the `Clang <https://clang.llvm.org/>`_ front end.  This
20component compiles C, C++, Objective C, and Objective C++ code into LLVM bitcode
21-- and from there into object files, using LLVM.
22
23Other components include:
24the `libc++ C++ standard library <https://libcxx.llvm.org>`_,
25the `LLD linker <https://lld.llvm.org>`_, and more.
26
27Getting the Source Code and Building LLVM
28=========================================
29
30The LLVM Getting Started documentation may be out of date.  The `Clang
31Getting Started <https://clang.llvm.org/get_started.html>`_ page might have more
32accurate information.
33
34This is an example workflow and configuration to get and build the LLVM source:
35
36#. Checkout LLVM (including related subprojects like Clang):
37
38   * ``git clone https://github.com/llvm/llvm-project.git``
39   * Or, on windows, ``git clone --config core.autocrlf=false
40     https://github.com/llvm/llvm-project.git``
41   * To save storage and speed-up the checkout time, you may want to do a
42     `shallow clone <https://git-scm.com/docs/git-clone#Documentation/git-clone.txt---depthltdepthgt>`_.
43     For example, to get the latest revision of the LLVM project, use
44     ``git clone --depth 1 https://github.com/llvm/llvm-project.git``
45
46#. Configure and build LLVM and Clang:
47
48   * ``cd llvm-project``
49   * ``mkdir build``
50   * ``cd build``
51   * ``cmake -G <generator> -DCMAKE_BUILD_TYPE=<type> [options] ../llvm``
52
53     Some common build system generators are:
54
55     * ``Ninja`` --- for generating `Ninja <https://ninja-build.org>`_
56       build files. Most llvm developers use Ninja.
57     * ``Unix Makefiles`` --- for generating make-compatible parallel makefiles.
58     * ``Visual Studio`` --- for generating Visual Studio projects and
59       solutions.
60     * ``Xcode`` --- for generating Xcode projects.
61
62     Some Common options:
63
64     * ``-DLLVM_ENABLE_PROJECTS='...'`` --- semicolon-separated list of the LLVM
65       subprojects you'd like to additionally build. Can include any of: clang,
66       clang-tools-extra, lldb, compiler-rt, lld, polly, or cross-project-tests.
67
68       For example, to build LLVM, Clang, libcxx, and libcxxabi, use
69       ``-DLLVM_ENABLE_PROJECTS="clang" -DLLVM_ENABLE_RUNTIMES="libcxx;libcxxabi"``.
70
71     * ``-DCMAKE_INSTALL_PREFIX=directory`` --- Specify for *directory* the full
72       pathname of where you want the LLVM tools and libraries to be installed
73       (default ``/usr/local``).
74
75     * ``-DCMAKE_BUILD_TYPE=type`` --- Controls optimization level and debug information
76       of the build. The default value is ``Debug`` which fits people who want
77       to work on LLVM or its libraries. ``Release`` is a better fit for most
78       users of LLVM and Clang. For more detailed information see
79       :ref:`CMAKE_BUILD_TYPE <cmake_build_type>`.
80
81     * ``-DLLVM_ENABLE_ASSERTIONS=On`` --- Compile with assertion checks enabled
82       (default is Yes for Debug builds, No for all other build types).
83
84   * ``cmake --build . [--target <target>]`` or the build system specified
85     above directly.
86
87     * The default target (i.e. ``cmake --build .`` or ``make``) will build all of
88       LLVM.
89
90     * The ``check-all`` target (i.e. ``ninja check-all``) will run the
91       regression tests to ensure everything is in working order.
92
93     * CMake will generate build targets for each tool and library, and most
94       LLVM sub-projects generate their own ``check-<project>`` target.
95
96     * Running a serial build will be **slow**.  To improve speed, try running a
97       parallel build. That's done by default in Ninja; for ``make``, use the
98       option ``-j NN``, where ``NN`` is the number of parallel jobs, e.g. the
99       number of available CPUs.
100
101   * For more information see `CMake <CMake.html>`__
102
103   * If you get an "internal compiler error (ICE)" or test failures, see
104     `below`_.
105
106Consult the `Getting Started with LLVM`_ section for detailed information on
107configuring and compiling LLVM.  Go to `Directory Layout`_ to learn about the
108layout of the source code tree.
109
110Requirements
111============
112
113Before you begin to use the LLVM system, review the requirements given below.
114This may save you some trouble by knowing ahead of time what hardware and
115software you will need.
116
117Hardware
118--------
119
120LLVM is known to work on the following host platforms:
121
122================== ===================== =============
123OS                 Arch                  Compilers
124================== ===================== =============
125Linux              x86\ :sup:`1`         GCC, Clang
126Linux              amd64                 GCC, Clang
127Linux              ARM                   GCC, Clang
128Linux              Mips                  GCC, Clang
129Linux              PowerPC               GCC, Clang
130Linux              SystemZ               GCC, Clang
131Solaris            V9 (Ultrasparc)       GCC
132DragonFlyBSD       amd64                 GCC, Clang
133FreeBSD            x86\ :sup:`1`         GCC, Clang
134FreeBSD            amd64                 GCC, Clang
135NetBSD             x86\ :sup:`1`         GCC, Clang
136NetBSD             amd64                 GCC, Clang
137OpenBSD            x86\ :sup:`1`         GCC, Clang
138OpenBSD            amd64                 GCC, Clang
139macOS\ :sup:`2`    PowerPC               GCC
140macOS              x86                   GCC, Clang
141Cygwin/Win32       x86\ :sup:`1, 3`      GCC
142Windows            x86\ :sup:`1`         Visual Studio
143Windows x64        x86-64                Visual Studio
144================== ===================== =============
145
146.. note::
147
148  #. Code generation supported for Pentium processors and up
149  #. Code generation supported for 32-bit ABI only
150  #. To use LLVM modules on Win32-based system, you may configure LLVM
151     with ``-DBUILD_SHARED_LIBS=On``.
152
153Note that Debug builds require a lot of time and disk space.  An LLVM-only build
154will need about 1-3 GB of space.  A full build of LLVM and Clang will need around
15515-20 GB of disk space.  The exact space requirements will vary by system.  (It
156is so large because of all the debugging information and the fact that the
157libraries are statically linked into multiple tools).
158
159If you are space-constrained, you can build only selected tools or only
160selected targets.  The Release build requires considerably less space.
161
162The LLVM suite *may* compile on other platforms, but it is not guaranteed to do
163so.  If compilation is successful, the LLVM utilities should be able to
164assemble, disassemble, analyze, and optimize LLVM bitcode.  Code generation
165should work as well, although the generated native code may not work on your
166platform.
167
168Software
169--------
170
171Compiling LLVM requires that you have several software packages installed. The
172table below lists those required packages. The Package column is the usual name
173for the software package that LLVM depends on. The Version column provides
174"known to work" versions of the package. The Notes column describes how LLVM
175uses the package and provides other details.
176
177=========================================================== ============ ==========================================
178Package                                                     Version      Notes
179=========================================================== ============ ==========================================
180`CMake <http://cmake.org/>`__                               >=3.13.4     Makefile/workspace generator
181`GCC <http://gcc.gnu.org/>`_                                >=7.1.0      C/C++ compiler\ :sup:`1`
182`python <http://www.python.org/>`_                          >=3.6        Automated test suite\ :sup:`2`
183`zlib <http://zlib.net>`_                                   >=1.2.3.4    Compression library\ :sup:`3`
184`GNU Make <http://savannah.gnu.org/projects/make>`_         3.79, 3.79.1 Makefile/build processor\ :sup:`4`
185=========================================================== ============ ==========================================
186
187.. note::
188
189   #. Only the C and C++ languages are needed so there's no need to build the
190      other languages for LLVM's purposes. See `below` for specific version
191      info.
192   #. Only needed if you want to run the automated test suite in the
193      ``llvm/test`` directory.
194   #. Optional, adds compression / uncompression capabilities to selected LLVM
195      tools.
196   #. Optional, you can use any other build tool supported by CMake.
197
198Additionally, your compilation host is expected to have the usual plethora of
199Unix utilities. Specifically:
200
201* **ar** --- archive library builder
202* **bzip2** --- bzip2 command for distribution generation
203* **bunzip2** --- bunzip2 command for distribution checking
204* **chmod** --- change permissions on a file
205* **cat** --- output concatenation utility
206* **cp** --- copy files
207* **date** --- print the current date/time
208* **echo** --- print to standard output
209* **egrep** --- extended regular expression search utility
210* **find** --- find files/dirs in a file system
211* **grep** --- regular expression search utility
212* **gzip** --- gzip command for distribution generation
213* **gunzip** --- gunzip command for distribution checking
214* **install** --- install directories/files
215* **mkdir** --- create a directory
216* **mv** --- move (rename) files
217* **ranlib** --- symbol table builder for archive libraries
218* **rm** --- remove (delete) files and directories
219* **sed** --- stream editor for transforming output
220* **sh** --- Bourne shell for make build scripts
221* **tar** --- tape archive for distribution generation
222* **test** --- test things in file system
223* **unzip** --- unzip command for distribution checking
224* **zip** --- zip command for distribution generation
225
226.. _below:
227.. _check here:
228
229Host C++ Toolchain, both Compiler and Standard Library
230------------------------------------------------------
231
232LLVM is very demanding of the host C++ compiler, and as such tends to expose
233bugs in the compiler. We also attempt to follow improvements and developments in
234the C++ language and library reasonably closely. As such, we require a modern
235host C++ toolchain, both compiler and standard library, in order to build LLVM.
236
237LLVM is written using the subset of C++ documented in :doc:`coding
238standards<CodingStandards>`. To enforce this language version, we check the most
239popular host toolchains for specific minimum versions in our build systems:
240
241* Clang 5.0
242* Apple Clang 9.3
243* GCC 7.1
244* Visual Studio 2019 16.7
245
246Anything older than these toolchains *may* work, but will require forcing the
247build system with a special option and is not really a supported host platform.
248Also note that older versions of these compilers have often crashed or
249miscompiled LLVM.
250
251For less widely used host toolchains such as ICC or xlC, be aware that a very
252recent version may be required to support all of the C++ features used in LLVM.
253
254We track certain versions of software that are *known* to fail when used as
255part of the host toolchain. These even include linkers at times.
256
257**GNU ld 2.16.X**. Some 2.16.X versions of the ld linker will produce very long
258warning messages complaining that some "``.gnu.linkonce.t.*``" symbol was
259defined in a discarded section. You can safely ignore these messages as they are
260erroneous and the linkage is correct.  These messages disappear using ld 2.17.
261
262**GNU binutils 2.17**: Binutils 2.17 contains `a bug
263<http://sourceware.org/bugzilla/show_bug.cgi?id=3111>`__ which causes huge link
264times (minutes instead of seconds) when building LLVM.  We recommend upgrading
265to a newer version (2.17.50.0.4 or later).
266
267**GNU Binutils 2.19.1 Gold**: This version of Gold contained `a bug
268<http://sourceware.org/bugzilla/show_bug.cgi?id=9836>`__ which causes
269intermittent failures when building LLVM with position independent code.  The
270symptom is an error about cyclic dependencies.  We recommend upgrading to a
271newer version of Gold.
272
273Getting a Modern Host C++ Toolchain
274^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
275
276This section mostly applies to Linux and older BSDs. On macOS, you should
277have a sufficiently modern Xcode, or you will likely need to upgrade until you
278do. Windows does not have a "system compiler", so you must install either Visual
279Studio 2019 (or later), or a recent version of mingw64. FreeBSD 10.0 and newer
280have a modern Clang as the system compiler.
281
282However, some Linux distributions and some other or older BSDs sometimes have
283extremely old versions of GCC. These steps attempt to help you upgrade you
284compiler even on such a system. However, if at all possible, we encourage you
285to use a recent version of a distribution with a modern system compiler that
286meets these requirements. Note that it is tempting to install a prior
287version of Clang and libc++ to be the host compiler, however libc++ was not
288well tested or set up to build on Linux until relatively recently. As
289a consequence, this guide suggests just using libstdc++ and a modern GCC as the
290initial host in a bootstrap, and then using Clang (and potentially libc++).
291
292The first step is to get a recent GCC toolchain installed. The most common
293distribution on which users have struggled with the version requirements is
294Ubuntu Precise, 12.04 LTS. For this distribution, one easy option is to install
295the `toolchain testing PPA`_ and use it to install a modern GCC. There is
296a really nice discussions of this on the `ask ubuntu stack exchange`_ and a
297`github gist`_ with updated commands. However, not all users can use PPAs and
298there are many other distributions, so it may be necessary (or just useful, if
299you're here you *are* doing compiler development after all) to build and install
300GCC from source. It is also quite easy to do these days.
301
302.. _toolchain testing PPA:
303  https://launchpad.net/~ubuntu-toolchain-r/+archive/test
304.. _ask ubuntu stack exchange:
305  https://askubuntu.com/questions/466651/how-do-i-use-the-latest-gcc-on-ubuntu/581497#58149
306.. _github gist:
307  https://gist.github.com/application2000/73fd6f4bf1be6600a2cf9f56315a2d91
308
309Easy steps for installing GCC 7.1.0:
310
311.. code-block:: console
312
313  % gcc_version=7.1.0
314  % wget https://ftp.gnu.org/gnu/gcc/gcc-${gcc_version}/gcc-${gcc_version}.tar.bz2
315  % wget https://ftp.gnu.org/gnu/gcc/gcc-${gcc_version}/gcc-${gcc_version}.tar.bz2.sig
316  % wget https://ftp.gnu.org/gnu/gnu-keyring.gpg
317  % signature_invalid=`gpg --verify --no-default-keyring --keyring ./gnu-keyring.gpg gcc-${gcc_version}.tar.bz2.sig`
318  % if [ $signature_invalid ]; then echo "Invalid signature" ; exit 1 ; fi
319  % tar -xvjf gcc-${gcc_version}.tar.bz2
320  % cd gcc-${gcc_version}
321  % ./contrib/download_prerequisites
322  % cd ..
323  % mkdir gcc-${gcc_version}-build
324  % cd gcc-${gcc_version}-build
325  % $PWD/../gcc-${gcc_version}/configure --prefix=$HOME/toolchains --enable-languages=c,c++
326  % make -j$(nproc)
327  % make install
328
329For more details, check out the excellent `GCC wiki entry`_, where I got most
330of this information from.
331
332.. _GCC wiki entry:
333  https://gcc.gnu.org/wiki/InstallingGCC
334
335Once you have a GCC toolchain, configure your build of LLVM to use the new
336toolchain for your host compiler and C++ standard library. Because the new
337version of libstdc++ is not on the system library search path, you need to pass
338extra linker flags so that it can be found at link time (``-L``) and at runtime
339(``-rpath``). If you are using CMake, this invocation should produce working
340binaries:
341
342.. code-block:: console
343
344  % mkdir build
345  % cd build
346  % CC=$HOME/toolchains/bin/gcc CXX=$HOME/toolchains/bin/g++ \
347    cmake .. -DCMAKE_CXX_LINK_FLAGS="-Wl,-rpath,$HOME/toolchains/lib64 -L$HOME/toolchains/lib64"
348
349If you fail to set rpath, most LLVM binaries will fail on startup with a message
350from the loader similar to ``libstdc++.so.6: version `GLIBCXX_3.4.20' not
351found``. This means you need to tweak the -rpath linker flag.
352
353This method will add an absolute path to the rpath of all executables. That's
354fine for local development. If you want to distribute the binaries you build
355so that they can run on older systems, copy ``libstdc++.so.6`` into the
356``lib/`` directory.  All of LLVM's shipping binaries have an rpath pointing at
357``$ORIGIN/../lib``, so they will find ``libstdc++.so.6`` there.  Non-distributed
358binaries don't have an rpath set and won't find ``libstdc++.so.6``. Pass
359``-DLLVM_LOCAL_RPATH="$HOME/toolchains/lib64"`` to cmake to add an absolute
360path to ``libstdc++.so.6`` as above. Since these binaries are not distributed,
361having an absolute local path is fine for them.
362
363When you build Clang, you will need to give *it* access to modern C++
364standard library in order to use it as your new host in part of a bootstrap.
365There are two easy ways to do this, either build (and install) libc++ along
366with Clang and then use it with the ``-stdlib=libc++`` compile and link flag,
367or install Clang into the same prefix (``$HOME/toolchains`` above) as GCC.
368Clang will look within its own prefix for libstdc++ and use it if found. You
369can also add an explicit prefix for Clang to look in for a GCC toolchain with
370the ``--gcc-toolchain=/opt/my/gcc/prefix`` flag, passing it to both compile and
371link commands when using your just-built-Clang to bootstrap.
372
373.. _Getting Started with LLVM:
374
375Getting Started with LLVM
376=========================
377
378The remainder of this guide is meant to get you up and running with LLVM and to
379give you some basic information about the LLVM environment.
380
381The later sections of this guide describe the `general layout`_ of the LLVM
382source tree, a `simple example`_ using the LLVM tool chain, and `links`_ to find
383more information about LLVM or to get help via e-mail.
384
385Terminology and Notation
386------------------------
387
388Throughout this manual, the following names are used to denote paths specific to
389the local system and working environment.  *These are not environment variables
390you need to set but just strings used in the rest of this document below*.  In
391any of the examples below, simply replace each of these names with the
392appropriate pathname on your local system.  All these paths are absolute:
393
394``SRC_ROOT``
395
396  This is the top level directory of the LLVM source tree.
397
398``OBJ_ROOT``
399
400  This is the top level directory of the LLVM object tree (i.e. the tree where
401  object files and compiled programs will be placed.  It can be the same as
402  SRC_ROOT).
403
404Unpacking the LLVM Archives
405---------------------------
406
407If you have the LLVM distribution, you will need to unpack it before you can
408begin to compile it.  LLVM is distributed as a number of different
409subprojects. Each one has its own download which is a TAR archive that is
410compressed with the gzip program.
411
412The files are as follows, with *x.y* marking the version number:
413
414``llvm-x.y.tar.gz``
415
416  Source release for the LLVM libraries and tools.
417
418``cfe-x.y.tar.gz``
419
420  Source release for the Clang frontend.
421
422.. _checkout:
423
424Checkout LLVM from Git
425----------------------
426
427You can also checkout the source code for LLVM from Git.
428
429.. note::
430
431  Passing ``--config core.autocrlf=false`` should not be required in
432  the future after we adjust the .gitattribute settings correctly, but
433  is required for Windows users at the time of this writing.
434
435Simply run:
436
437.. code-block:: console
438
439  % git clone https://github.com/llvm/llvm-project.git
440
441or on Windows,
442
443.. code-block:: console
444
445  % git clone --config core.autocrlf=false https://github.com/llvm/llvm-project.git
446
447This will create an '``llvm-project``' directory in the current directory and
448fully populate it with all of the source code, test directories, and local
449copies of documentation files for LLVM and all the related subprojects. Note
450that unlike the tarballs, which contain each subproject in a separate file, the
451git repository contains all of the projects together.
452
453If you want to get a specific release (as opposed to the most recent revision),
454you can check out a tag after cloning the repository. E.g., `git checkout
455llvmorg-6.0.1` inside the ``llvm-project`` directory created by the above
456command.  Use `git tag -l` to list all of them.
457
458Sending patches
459^^^^^^^^^^^^^^^
460
461See :ref:`Contributing <submit_patch>`.
462
463Bisecting commits
464^^^^^^^^^^^^^^^^^
465
466See `Bisecting LLVM code <GitBisecting.html>`_ for how to use ``git bisect``
467on LLVM.
468
469Reverting a change
470^^^^^^^^^^^^^^^^^^
471
472When reverting changes using git, the default message will say "This reverts
473commit XYZ". Leave this at the end of the commit message, but add some details
474before it as to why the commit is being reverted. A brief explanation and/or
475links to bots that demonstrate the problem are sufficient.
476
477Local LLVM Configuration
478------------------------
479
480Once checked out repository, the LLVM suite source code must be configured
481before being built. This process uses CMake.  Unlinke the normal ``configure``
482script, CMake generates the build files in whatever format you request as well
483as various ``*.inc`` files, and ``llvm/include/Config/config.h``.
484
485Variables are passed to ``cmake`` on the command line using the format
486``-D<variable name>=<value>``. The following variables are some common options
487used by people developing LLVM.
488
489+-------------------------+----------------------------------------------------+
490| Variable                | Purpose                                            |
491+=========================+====================================================+
492| CMAKE_C_COMPILER        | Tells ``cmake`` which C compiler to use. By        |
493|                         | default, this will be /usr/bin/cc.                 |
494+-------------------------+----------------------------------------------------+
495| CMAKE_CXX_COMPILER      | Tells ``cmake`` which C++ compiler to use. By      |
496|                         | default, this will be /usr/bin/c++.                |
497+-------------------------+----------------------------------------------------+
498| CMAKE_BUILD_TYPE        | Tells ``cmake`` what type of build you are trying  |
499|                         | to generate files for. Valid options are Debug,    |
500|                         | Release, RelWithDebInfo, and MinSizeRel. Default   |
501|                         | is Debug.                                          |
502+-------------------------+----------------------------------------------------+
503| CMAKE_INSTALL_PREFIX    | Specifies the install directory to target when     |
504|                         | running the install action of the build files.     |
505+-------------------------+----------------------------------------------------+
506| PYTHON_EXECUTABLE       | Forces CMake to use a specific Python version by   |
507|                         | passing a path to a Python interpreter. By default |
508|                         | the Python version of the interpreter in your PATH |
509|                         | is used.                                           |
510+-------------------------+----------------------------------------------------+
511| LLVM_TARGETS_TO_BUILD   | A semicolon delimited list controlling which       |
512|                         | targets will be built and linked into llvm.        |
513|                         | The default list is defined as                     |
514|                         | ``LLVM_ALL_TARGETS``, and can be set to include    |
515|                         | out-of-tree targets. The default value includes:   |
516|                         | ``AArch64, AMDGPU, ARM, AVR, BPF, Hexagon, Lanai,  |
517|                         | Mips, MSP430, NVPTX, PowerPC, RISCV, Sparc,        |
518|                         | SystemZ, WebAssembly, X86, XCore``.                |
519|                         |                                                    |
520+-------------------------+----------------------------------------------------+
521| LLVM_ENABLE_DOXYGEN     | Build doxygen-based documentation from the source  |
522|                         | code This is disabled by default because it is     |
523|                         | slow and generates a lot of output.                |
524+-------------------------+----------------------------------------------------+
525| LLVM_ENABLE_PROJECTS    | A semicolon-delimited list selecting which of the  |
526|                         | other LLVM subprojects to additionally build. (Only|
527|                         | effective when using a side-by-side project layout |
528|                         | e.g. via git). The default list is empty. Can      |
529|                         | include: clang, clang-tools-extra,                 |
530|                         | cross-project-tests, flang, libc, libclc, lld,     |
531|                         | lldb, mlir, openmp, polly, or pstl.                |
532+-------------------------+----------------------------------------------------+
533| LLVM_ENABLE_RUNTIMES    | A semicolon-delimited list selecting which of the  |
534|                         | runtimes to build. (Only effective when using the  |
535|                         | full monorepo layout). The default list is empty.  |
536|                         | Can include: compiler-rt, libc, libcxx, libcxxabi, |
537|                         | libunwind, or openmp.                              |
538+-------------------------+----------------------------------------------------+
539| LLVM_ENABLE_SPHINX      | Build sphinx-based documentation from the source   |
540|                         | code. This is disabled by default because it is    |
541|                         | slow and generates a lot of output. Sphinx version |
542|                         | 1.5 or later recommended.                          |
543+-------------------------+----------------------------------------------------+
544| LLVM_BUILD_LLVM_DYLIB   | Generate libLLVM.so. This library contains a       |
545|                         | default set of LLVM components that can be         |
546|                         | overridden with ``LLVM_DYLIB_COMPONENTS``. The     |
547|                         | default contains most of LLVM and is defined in    |
548|                         | ``tools/llvm-shlib/CMakelists.txt``. This option is|
549|                         | not available on Windows.                          |
550+-------------------------+----------------------------------------------------+
551| LLVM_OPTIMIZED_TABLEGEN | Builds a release tablegen that gets used during    |
552|                         | the LLVM build. This can dramatically speed up     |
553|                         | debug builds.                                      |
554+-------------------------+----------------------------------------------------+
555
556To configure LLVM, follow these steps:
557
558#. Change directory into the object root directory:
559
560   .. code-block:: console
561
562     % cd OBJ_ROOT
563
564#. Run the ``cmake``:
565
566   .. code-block:: console
567
568     % cmake -G "Unix Makefiles" -DCMAKE_BUILD_TYPE=<type> -DCMAKE_INSTALL_PREFIX=/install/path
569       [other options] SRC_ROOT
570
571Compiling the LLVM Suite Source Code
572------------------------------------
573
574Unlike with autotools, with CMake your build type is defined at configuration.
575If you want to change your build type, you can re-run cmake with the following
576invocation:
577
578   .. code-block:: console
579
580     % cmake -G "Unix Makefiles" -DCMAKE_BUILD_TYPE=<type> -DCMAKE_BUILD_TYPE=type SRC_ROOT
581
582Between runs, CMake preserves the values set for all options. CMake has the
583following build types defined:
584
585Debug
586
587  These builds are the default. The build system will compile the tools and
588  libraries unoptimized, with debugging information, and asserts enabled.
589
590Release
591
592  For these builds, the build system will compile the tools and libraries
593  with optimizations enabled and not generate debug info. CMakes default
594  optimization level is -O3. This can be configured by setting the
595  ``CMAKE_CXX_FLAGS_RELEASE`` variable on the CMake command line.
596
597RelWithDebInfo
598
599  These builds are useful when debugging. They generate optimized binaries with
600  debug information. CMakes default optimization level is -O2. This can be
601  configured by setting the ``CMAKE_CXX_FLAGS_RELWITHDEBINFO`` variable on the
602  CMake command line.
603
604Once you have LLVM configured, you can build it by entering the *OBJ_ROOT*
605directory and issuing the following command:
606
607.. code-block:: console
608
609  % make
610
611If the build fails, please `check here`_ to see if you are using a version of
612GCC that is known not to compile LLVM.
613
614If you have multiple processors in your machine, you may wish to use some of the
615parallel build options provided by GNU Make.  For example, you could use the
616command:
617
618.. code-block:: console
619
620  % make -j2
621
622There are several special targets which are useful when working with the LLVM
623source code:
624
625``make clean``
626
627  Removes all files generated by the build.  This includes object files,
628  generated C/C++ files, libraries, and executables.
629
630``make install``
631
632  Installs LLVM header files, libraries, tools, and documentation in a hierarchy
633  under ``$PREFIX``, specified with ``CMAKE_INSTALL_PREFIX``, which
634  defaults to ``/usr/local``.
635
636``make docs-llvm-html``
637
638  If configured with ``-DLLVM_ENABLE_SPHINX=On``, this will generate a directory
639  at ``OBJ_ROOT/docs/html`` which contains the HTML formatted documentation.
640
641Cross-Compiling LLVM
642--------------------
643
644It is possible to cross-compile LLVM itself. That is, you can create LLVM
645executables and libraries to be hosted on a platform different from the platform
646where they are built (a Canadian Cross build). To generate build files for
647cross-compiling CMake provides a variable ``CMAKE_TOOLCHAIN_FILE`` which can
648define compiler flags and variables used during the CMake test operations.
649
650The result of such a build is executables that are not runnable on the build
651host but can be executed on the target. As an example the following CMake
652invocation can generate build files targeting iOS. This will work on macOS
653with the latest Xcode:
654
655.. code-block:: console
656
657  % cmake -G "Ninja" -DCMAKE_OSX_ARCHITECTURES="armv7;armv7s;arm64"
658    -DCMAKE_TOOLCHAIN_FILE=<PATH_TO_LLVM>/cmake/platforms/iOS.cmake
659    -DCMAKE_BUILD_TYPE=Release -DLLVM_BUILD_RUNTIME=Off -DLLVM_INCLUDE_TESTS=Off
660    -DLLVM_INCLUDE_EXAMPLES=Off -DLLVM_ENABLE_BACKTRACES=Off [options]
661    <PATH_TO_LLVM>
662
663Note: There are some additional flags that need to be passed when building for
664iOS due to limitations in the iOS SDK.
665
666Check :doc:`HowToCrossCompileLLVM` and `Clang docs on how to cross-compile in general
667<https://clang.llvm.org/docs/CrossCompilation.html>`_ for more information
668about cross-compiling.
669
670The Location of LLVM Object Files
671---------------------------------
672
673The LLVM build system is capable of sharing a single LLVM source tree among
674several LLVM builds.  Hence, it is possible to build LLVM for several different
675platforms or configurations using the same source tree.
676
677* Change directory to where the LLVM object files should live:
678
679  .. code-block:: console
680
681    % cd OBJ_ROOT
682
683* Run ``cmake``:
684
685  .. code-block:: console
686
687    % cmake -G "Unix Makefiles" -DCMAKE_BUILD_TYPE=Release SRC_ROOT
688
689The LLVM build will create a structure underneath *OBJ_ROOT* that matches the
690LLVM source tree. At each level where source files are present in the source
691tree there will be a corresponding ``CMakeFiles`` directory in the *OBJ_ROOT*.
692Underneath that directory there is another directory with a name ending in
693``.dir`` under which you'll find object files for each source.
694
695For example:
696
697  .. code-block:: console
698
699    % cd llvm_build_dir
700    % find lib/Support/ -name APFloat*
701    lib/Support/CMakeFiles/LLVMSupport.dir/APFloat.cpp.o
702
703Optional Configuration Items
704----------------------------
705
706If you're running on a Linux system that supports the `binfmt_misc
707<http://en.wikipedia.org/wiki/binfmt_misc>`_
708module, and you have root access on the system, you can set your system up to
709execute LLVM bitcode files directly. To do this, use commands like this (the
710first command may not be required if you are already using the module):
711
712.. code-block:: console
713
714  % mount -t binfmt_misc none /proc/sys/fs/binfmt_misc
715  % echo ':llvm:M::BC::/path/to/lli:' > /proc/sys/fs/binfmt_misc/register
716  % chmod u+x hello.bc   (if needed)
717  % ./hello.bc
718
719This allows you to execute LLVM bitcode files directly.  On Debian, you can also
720use this command instead of the 'echo' command above:
721
722.. code-block:: console
723
724  % sudo update-binfmts --install llvm /path/to/lli --magic 'BC'
725
726.. _Program Layout:
727.. _general layout:
728
729Directory Layout
730================
731
732One useful source of information about the LLVM source base is the LLVM `doxygen
733<http://www.doxygen.org/>`_ documentation available at
734`<https://llvm.org/doxygen/>`_.  The following is a brief introduction to code
735layout:
736
737``llvm/cmake``
738--------------
739Generates system build files.
740
741``llvm/cmake/modules``
742  Build configuration for llvm user defined options. Checks compiler version and
743  linker flags.
744
745``llvm/cmake/platforms``
746  Toolchain configuration for Android NDK, iOS systems and non-Windows hosts to
747  target MSVC.
748
749``llvm/examples``
750-----------------
751
752- Some simple examples showing how to use LLVM as a compiler for a custom
753  language - including lowering, optimization, and code generation.
754
755- Kaleidoscope Tutorial: Kaleidoscope language tutorial run through the
756  implementation of a nice little compiler for a non-trivial language
757  including a hand-written lexer, parser, AST, as well as code generation
758  support using LLVM- both static (ahead of time) and various approaches to
759  Just In Time (JIT) compilation.
760  `Kaleidoscope Tutorial for complete beginner
761  <https://llvm.org/docs/tutorial/MyFirstLanguageFrontend/index.html>`_.
762
763- BuildingAJIT: Examples of the `BuildingAJIT tutorial
764  <https://llvm.org/docs/tutorial/BuildingAJIT1.html>`_ that shows how LLVM’s
765  ORC JIT APIs interact with other parts of LLVM. It also, teaches how to
766  recombine them to build a custom JIT that is suited to your use-case.
767
768``llvm/include``
769----------------
770
771Public header files exported from the LLVM library. The three main subdirectories:
772
773``llvm/include/llvm``
774
775  All LLVM-specific header files, and  subdirectories for different portions of
776  LLVM: ``Analysis``, ``CodeGen``, ``Target``, ``Transforms``, etc...
777
778``llvm/include/llvm/Support``
779
780  Generic support libraries provided with LLVM but not necessarily specific to
781  LLVM. For example, some C++ STL utilities and a Command Line option processing
782  library store header files here.
783
784``llvm/include/llvm/Config``
785
786  Header files configured by ``cmake``.  They wrap "standard" UNIX and
787  C header files.  Source code can include these header files which
788  automatically take care of the conditional #includes that ``cmake``
789  generates.
790
791``llvm/lib``
792------------
793
794Most source files are here. By putting code in libraries, LLVM makes it easy to
795share code among the `tools`_.
796
797``llvm/lib/IR/``
798
799  Core LLVM source files that implement core classes like Instruction and
800  BasicBlock.
801
802``llvm/lib/AsmParser/``
803
804  Source code for the LLVM assembly language parser library.
805
806``llvm/lib/Bitcode/``
807
808  Code for reading and writing bitcode.
809
810``llvm/lib/Analysis/``
811
812  A variety of program analyses, such as Call Graphs, Induction Variables,
813  Natural Loop Identification, etc.
814
815``llvm/lib/Transforms/``
816
817  IR-to-IR program transformations, such as Aggressive Dead Code Elimination,
818  Sparse Conditional Constant Propagation, Inlining, Loop Invariant Code Motion,
819  Dead Global Elimination, and many others.
820
821``llvm/lib/Target/``
822
823  Files describing target architectures for code generation.  For example,
824  ``llvm/lib/Target/X86`` holds the X86 machine description.
825
826``llvm/lib/CodeGen/``
827
828  The major parts of the code generator: Instruction Selector, Instruction
829  Scheduling, and Register Allocation.
830
831``llvm/lib/MC/``
832
833  The libraries represent and process code at machine code level. Handles
834  assembly and object-file emission.
835
836``llvm/lib/ExecutionEngine/``
837
838  Libraries for directly executing bitcode at runtime in interpreted and
839  JIT-compiled scenarios.
840
841``llvm/lib/Support/``
842
843  Source code that corresponding to the header files in ``llvm/include/ADT/``
844  and ``llvm/include/Support/``.
845
846``llvm/bindings``
847----------------------
848
849Contains bindings for the LLVM compiler infrastructure to allow
850programs written in languages other than C or C++ to take advantage of the LLVM
851infrastructure.
852LLVM project provides language bindings for Go, OCaml and Python.
853
854``llvm/projects``
855-----------------
856
857Projects not strictly part of LLVM but shipped with LLVM. This is also the
858directory for creating your own LLVM-based projects which leverage the LLVM
859build system.
860
861``llvm/test``
862-------------
863
864Feature and regression tests and other sanity checks on LLVM infrastructure. These
865are intended to run quickly and cover a lot of territory without being exhaustive.
866
867``test-suite``
868--------------
869
870A comprehensive correctness, performance, and benchmarking test suite
871for LLVM.  This comes in a ``separate git repository
872<https://github.com/llvm/llvm-test-suite>``, because it contains a
873large amount of third-party code under a variety of licenses. For
874details see the :doc:`Testing Guide <TestingGuide>` document.
875
876.. _tools:
877
878``llvm/tools``
879--------------
880
881Executables built out of the libraries
882above, which form the main part of the user interface.  You can always get help
883for a tool by typing ``tool_name -help``.  The following is a brief introduction
884to the most important tools.  More detailed information is in
885the `Command Guide <CommandGuide/index.html>`_.
886
887``bugpoint``
888
889  ``bugpoint`` is used to debug optimization passes or code generation backends
890  by narrowing down the given test case to the minimum number of passes and/or
891  instructions that still cause a problem, whether it is a crash or
892  miscompilation. See `<HowToSubmitABug.html>`_ for more information on using
893  ``bugpoint``.
894
895``llvm-ar``
896
897  The archiver produces an archive containing the given LLVM bitcode files,
898  optionally with an index for faster lookup.
899
900``llvm-as``
901
902  The assembler transforms the human readable LLVM assembly to LLVM bitcode.
903
904``llvm-dis``
905
906  The disassembler transforms the LLVM bitcode to human readable LLVM assembly.
907
908``llvm-link``
909
910  ``llvm-link``, not surprisingly, links multiple LLVM modules into a single
911  program.
912
913``lli``
914
915  ``lli`` is the LLVM interpreter, which can directly execute LLVM bitcode
916  (although very slowly...). For architectures that support it (currently x86,
917  Sparc, and PowerPC), by default, ``lli`` will function as a Just-In-Time
918  compiler (if the functionality was compiled in), and will execute the code
919  *much* faster than the interpreter.
920
921``llc``
922
923  ``llc`` is the LLVM backend compiler, which translates LLVM bitcode to a
924  native code assembly file.
925
926``opt``
927
928  ``opt`` reads LLVM bitcode, applies a series of LLVM to LLVM transformations
929  (which are specified on the command line), and outputs the resultant
930  bitcode.   '``opt -help``'  is a good way to get a list of the
931  program transformations available in LLVM.
932
933  ``opt`` can also  run a specific analysis on an input LLVM bitcode
934  file and print  the results.  Primarily useful for debugging
935  analyses, or familiarizing yourself with what an analysis does.
936
937``llvm/utils``
938--------------
939
940Utilities for working with LLVM source code; some are part of the build process
941because they are code generators for parts of the infrastructure.
942
943
944``codegen-diff``
945
946  ``codegen-diff`` finds differences between code that LLC
947  generates and code that LLI generates. This is useful if you are
948  debugging one of them, assuming that the other generates correct output. For
949  the full user manual, run ```perldoc codegen-diff'``.
950
951``emacs/``
952
953   Emacs and XEmacs syntax highlighting  for LLVM   assembly files and TableGen
954   description files.  See the ``README`` for information on using them.
955
956``getsrcs.sh``
957
958  Finds and outputs all non-generated source files,
959  useful if one wishes to do a lot of development across directories
960  and does not want to find each file. One way to use it is to run,
961  for example: ``xemacs `utils/getsources.sh``` from the top of the LLVM source
962  tree.
963
964``llvmgrep``
965
966  Performs an ``egrep -H -n`` on each source file in LLVM and
967  passes to it a regular expression provided on ``llvmgrep``'s command
968  line. This is an efficient way of searching the source base for a
969  particular regular expression.
970
971``TableGen/``
972
973  Contains the tool used to generate register
974  descriptions, instruction set descriptions, and even assemblers from common
975  TableGen description files.
976
977``vim/``
978
979  vim syntax-highlighting for LLVM assembly files
980  and TableGen description files. See the    ``README`` for how to use them.
981
982.. _simple example:
983
984An Example Using the LLVM Tool Chain
985====================================
986
987This section gives an example of using LLVM with the Clang front end.
988
989Example with clang
990------------------
991
992#. First, create a simple C file, name it 'hello.c':
993
994   .. code-block:: c
995
996     #include <stdio.h>
997
998     int main() {
999       printf("hello world\n");
1000       return 0;
1001     }
1002
1003#. Next, compile the C file into a native executable:
1004
1005   .. code-block:: console
1006
1007     % clang hello.c -o hello
1008
1009   .. note::
1010
1011     Clang works just like GCC by default.  The standard -S and -c arguments
1012     work as usual (producing a native .s or .o file, respectively).
1013
1014#. Next, compile the C file into an LLVM bitcode file:
1015
1016   .. code-block:: console
1017
1018     % clang -O3 -emit-llvm hello.c -c -o hello.bc
1019
1020   The -emit-llvm option can be used with the -S or -c options to emit an LLVM
1021   ``.ll`` or ``.bc`` file (respectively) for the code.  This allows you to use
1022   the `standard LLVM tools <CommandGuide/index.html>`_ on the bitcode file.
1023
1024#. Run the program in both forms. To run the program, use:
1025
1026   .. code-block:: console
1027
1028      % ./hello
1029
1030   and
1031
1032   .. code-block:: console
1033
1034     % lli hello.bc
1035
1036   The second examples shows how to invoke the LLVM JIT, :doc:`lli
1037   <CommandGuide/lli>`.
1038
1039#. Use the ``llvm-dis`` utility to take a look at the LLVM assembly code:
1040
1041   .. code-block:: console
1042
1043     % llvm-dis < hello.bc | less
1044
1045#. Compile the program to native assembly using the LLC code generator:
1046
1047   .. code-block:: console
1048
1049     % llc hello.bc -o hello.s
1050
1051#. Assemble the native assembly language file into a program:
1052
1053   .. code-block:: console
1054
1055     % /opt/SUNWspro/bin/cc -xarch=v9 hello.s -o hello.native   # On Solaris
1056
1057     % gcc hello.s -o hello.native                              # On others
1058
1059#. Execute the native code program:
1060
1061   .. code-block:: console
1062
1063     % ./hello.native
1064
1065   Note that using clang to compile directly to native code (i.e. when the
1066   ``-emit-llvm`` option is not present) does steps 6/7/8 for you.
1067
1068Common Problems
1069===============
1070
1071If you are having problems building or using LLVM, or if you have any other
1072general questions about LLVM, please consult the `Frequently Asked
1073Questions <FAQ.html>`_ page.
1074
1075If you are having problems with limited memory and build time, please try
1076building with ninja instead of make. Please consider configuring the
1077following options with cmake:
1078
1079 * -G Ninja
1080   Setting this option will allow you to build with ninja instead of make.
1081   Building with ninja significantly improves your build time, especially with
1082   incremental builds, and improves your memory usage.
1083
1084 * -DLLVM_USE_LINKER
1085   Setting this option to lld will significantly reduce linking time for LLVM
1086   executables on ELF-based platforms, such as Linux. If you are building LLVM
1087   for the first time and lld is not available to you as a binary package, then
1088   you may want to use the gold linker as a faster alternative to GNU ld.
1089
1090 * -DCMAKE_BUILD_TYPE
1091   Controls optimization level and debug information of the build.  This setting
1092   can affect RAM and disk usage, see :ref:`CMAKE_BUILD_TYPE <cmake_build_type>`
1093   for more information.
1094
1095 * -DLLVM_ENABLE_ASSERTIONS
1096   This option defaults to ON for Debug builds and defaults to OFF for Release
1097   builds. As mentioned in the previous option, using the Release build type and
1098   enabling assertions may be a good alternative to using the Debug build type.
1099
1100 * -DLLVM_PARALLEL_LINK_JOBS
1101   Set this equal to number of jobs you wish to run simultaneously. This is
1102   similar to the -j option used with make, but only for link jobs. This option
1103   can only be used with ninja. You may wish to use a very low number of jobs,
1104   as this will greatly reduce the amount of memory used during the build
1105   process. If you have limited memory, you may wish to set this to 1.
1106
1107 * -DLLVM_TARGETS_TO_BUILD
1108   Set this equal to the target you wish to build. You may wish to set this to
1109   X86; however, you will find a full list of targets within the
1110   llvm-project/llvm/lib/Target directory.
1111
1112 * -DLLVM_OPTIMIZED_TABLEGEN
1113   Set this to ON to generate a fully optimized tablegen during your build. This
1114   will significantly improve your build time. This is only useful if you are
1115   using the Debug build type.
1116
1117 * -DLLVM_ENABLE_PROJECTS
1118   Set this equal to the projects you wish to compile (e.g. clang, lld, etc.) If
1119   compiling more than one project, separate the items with a semicolon. Should
1120   you run into issues with the semicolon, try surrounding it with single quotes.
1121
1122 * -DLLVM_ENABLE_RUNTIMES
1123   Set this equal to the runtimes you wish to compile (e.g. libcxx, libcxxabi, etc.)
1124   If compiling more than one runtime, separate the items with a semicolon. Should
1125   you run into issues with the semicolon, try surrounding it with single quotes.
1126
1127 * -DCLANG_ENABLE_STATIC_ANALYZER
1128   Set this option to OFF if you do not require the clang static analyzer. This
1129   should improve your build time slightly.
1130
1131 * -DLLVM_USE_SPLIT_DWARF
1132   Consider setting this to ON if you require a debug build, as this will ease
1133   memory pressure on the linker. This will make linking much faster, as the
1134   binaries will not contain any of the debug information; however, this will
1135   generate the debug information in the form of a DWARF object file (with the
1136   extension .dwo). This only applies to host platforms using ELF, such as Linux.
1137
1138.. _links:
1139
1140Links
1141=====
1142
1143This document is just an **introduction** on how to use LLVM to do some simple
1144things... there are many more interesting and complicated things that you can do
1145that aren't documented here (but we'll gladly accept a patch if you want to
1146write something up!).  For more information about LLVM, check out:
1147
1148* `LLVM Homepage <https://llvm.org/>`_
1149* `LLVM Doxygen Tree <https://llvm.org/doxygen/>`_
1150* `Starting a Project that Uses LLVM <https://llvm.org/docs/Projects.html>`_
1151