xref: /openbsd-src/gnu/llvm/llvm/docs/GettingStarted.rst (revision d415bd752c734aee168c4ee86ff32e8cc249eb16)
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30The LLVM Getting Started documentation may be out of date.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
110Stand-alone Builds
111------------------
112
113Stand-alone builds allow you to build a sub-project against a pre-built
114version of the clang or llvm libraries that is already present on your
115system.
116
117You can use the source code from a standard checkout of the llvm-project
118(as described above) to do stand-alone builds, but you may also build
119from a :ref:`sparse checkout<workflow-multicheckout-nocommit>` or from the
120tarballs available on the `releases <https://github.com/llvm/llvm-project/releases/>`_
121page.
122
123For stand-alone builds, you must have an llvm install that is configured
124properly to be consumable by stand-alone builds of the other projects.
125This could be a distro provided LLVM install, or you can build it yourself,
126like this:
127
128.. code-block:: console
129
130  cmake -G Ninja -S path/to/llvm-project/llvm -B $builddir \
131        -DLLVM_INSTALL_UTILS=ON \
132        -DCMAKE_INSTALL_PREFIX=/path/to/llvm/install/prefix \
133        < other options >
134
135  ninja -C $builddir install
136
137Once llvm is installed, to configure a project for a stand-alone build, invoke CMake like this:
138
139.. code-block:: console
140
141  cmake -G Ninja -S path/to/llvm-project/$subproj \
142        -B $buildir_subproj \
143        -DLLVM_EXTERNAL_LIT=/path/to/lit \
144        -DLLVM_ROOT=/path/to/llvm/install/prefix
145
146Notice that:
147
148* The stand-alone build needs to happen in a folder that is not the
149  original folder where LLVMN was built
150  (`$builddir!=$builddir_subproj`).
151* ``LLVM_ROOT`` should point to the prefix of your llvm installation,
152  so for example, if llvm is installed into ``/usr/bin`` and
153  ``/usr/lib64``, then you should pass ``-DLLVM_ROOT=/usr/``.
154* Both the ``LLVM_ROOT`` and ``LLVM_EXTERNAL_LIT`` options are
155  required to do stand-alone builds for all sub-projects.  Additional
156  required options for each sub-project can be found in the table
157  below.
158
159The ``check-$subproj`` and ``install`` build targets are supported for the
160sub-projects listed in the table below.
161
162============ ======================== ======================
163Sub-Project  Required Sub-Directories Required CMake Options
164============ ======================== ======================
165llvm         llvm, cmake, third-party LLVM_INSTALL_UTILS=ON
166clang        clang, cmake             CLANG_INCLUDE_TESTS=ON (Required for check-clang only)
167lld          lld, cmake
168============ ======================== ======================
169
170Example for building stand-alone `clang`:
171
172.. code-block:: console
173
174   #!/bin/sh
175
176   build_llvm=`pwd`/build-llvm
177   build_clang=`pwd`/build-clang
178   installprefix=`pwd`/install
179   llvm=`pwd`/llvm-project
180   mkdir -p $build_llvm
181   mkdir -p $installprefix
182
183   cmake -G Ninja -S $llvm/llvm -B $build_llvm \
184         -DLLVM_INSTALL_UTILS=ON \
185         -DCMAKE_INSTALL_PREFIX=$installprefix \
186         -DCMAKE_BUILD_TYPE=Release
187
188   ninja -C $build_llvm install
189
190   cmake -G Ninja -S $llvm/clang -B $build_clang \
191         -DLLVM_EXTERNAL_LIT=$build_llvm/utils/lit \
192         -DLLVM_ROOT=$installprefix
193
194   ninja -C $build_clang
195
196Requirements
197============
198
199Before you begin to use the LLVM system, review the requirements given below.
200This may save you some trouble by knowing ahead of time what hardware and
201software you will need.
202
203Hardware
204--------
205
206LLVM is known to work on the following host platforms:
207
208================== ===================== =============
209OS                 Arch                  Compilers
210================== ===================== =============
211Linux              x86\ :sup:`1`         GCC, Clang
212Linux              amd64                 GCC, Clang
213Linux              ARM                   GCC, Clang
214Linux              Mips                  GCC, Clang
215Linux              PowerPC               GCC, Clang
216Linux              SystemZ               GCC, Clang
217Solaris            V9 (Ultrasparc)       GCC
218DragonFlyBSD       amd64                 GCC, Clang
219FreeBSD            x86\ :sup:`1`         GCC, Clang
220FreeBSD            amd64                 GCC, Clang
221NetBSD             x86\ :sup:`1`         GCC, Clang
222NetBSD             amd64                 GCC, Clang
223OpenBSD            x86\ :sup:`1`         GCC, Clang
224OpenBSD            amd64                 GCC, Clang
225macOS\ :sup:`2`    PowerPC               GCC
226macOS              x86                   GCC, Clang
227Cygwin/Win32       x86\ :sup:`1, 3`      GCC
228Windows            x86\ :sup:`1`         Visual Studio
229Windows x64        x86-64                Visual Studio
230================== ===================== =============
231
232.. note::
233
234  #. Code generation supported for Pentium processors and up
235  #. Code generation supported for 32-bit ABI only
236  #. To use LLVM modules on Win32-based system, you may configure LLVM
237     with ``-DBUILD_SHARED_LIBS=On``.
238
239Note that Debug builds require a lot of time and disk space.  An LLVM-only build
240will need about 1-3 GB of space.  A full build of LLVM and Clang will need around
24115-20 GB of disk space.  The exact space requirements will vary by system.  (It
242is so large because of all the debugging information and the fact that the
243libraries are statically linked into multiple tools).
244
245If you are space-constrained, you can build only selected tools or only
246selected targets.  The Release build requires considerably less space.
247
248The LLVM suite *may* compile on other platforms, but it is not guaranteed to do
249so.  If compilation is successful, the LLVM utilities should be able to
250assemble, disassemble, analyze, and optimize LLVM bitcode.  Code generation
251should work as well, although the generated native code may not work on your
252platform.
253
254Software
255--------
256
257Compiling LLVM requires that you have several software packages installed. The
258table below lists those required packages. The Package column is the usual name
259for the software package that LLVM depends on. The Version column provides
260"known to work" versions of the package. The Notes column describes how LLVM
261uses the package and provides other details.
262
263=========================================================== ============ ==========================================
264Package                                                     Version      Notes
265=========================================================== ============ ==========================================
266`CMake <http://cmake.org/>`__                               >=3.13.4     Makefile/workspace generator
267`GCC <http://gcc.gnu.org/>`_                                >=7.1.0      C/C++ compiler\ :sup:`1`
268`python <http://www.python.org/>`_                          >=3.6        Automated test suite\ :sup:`2`
269`zlib <http://zlib.net>`_                                   >=1.2.3.4    Compression library\ :sup:`3`
270`GNU Make <http://savannah.gnu.org/projects/make>`_         3.79, 3.79.1 Makefile/build processor\ :sup:`4`
271=========================================================== ============ ==========================================
272
273.. note::
274
275   #. Only the C and C++ languages are needed so there's no need to build the
276      other languages for LLVM's purposes. See `below` for specific version
277      info.
278   #. Only needed if you want to run the automated test suite in the
279      ``llvm/test`` directory.
280   #. Optional, adds compression / uncompression capabilities to selected LLVM
281      tools.
282   #. Optional, you can use any other build tool supported by CMake.
283
284Additionally, your compilation host is expected to have the usual plethora of
285Unix utilities. Specifically:
286
287* **ar** --- archive library builder
288* **bzip2** --- bzip2 command for distribution generation
289* **bunzip2** --- bunzip2 command for distribution checking
290* **chmod** --- change permissions on a file
291* **cat** --- output concatenation utility
292* **cp** --- copy files
293* **date** --- print the current date/time
294* **echo** --- print to standard output
295* **egrep** --- extended regular expression search utility
296* **find** --- find files/dirs in a file system
297* **grep** --- regular expression search utility
298* **gzip** --- gzip command for distribution generation
299* **gunzip** --- gunzip command for distribution checking
300* **install** --- install directories/files
301* **mkdir** --- create a directory
302* **mv** --- move (rename) files
303* **ranlib** --- symbol table builder for archive libraries
304* **rm** --- remove (delete) files and directories
305* **sed** --- stream editor for transforming output
306* **sh** --- Bourne shell for make build scripts
307* **tar** --- tape archive for distribution generation
308* **test** --- test things in file system
309* **unzip** --- unzip command for distribution checking
310* **zip** --- zip command for distribution generation
311
312.. _below:
313.. _check here:
314
315Host C++ Toolchain, both Compiler and Standard Library
316------------------------------------------------------
317
318LLVM is very demanding of the host C++ compiler, and as such tends to expose
319bugs in the compiler. We also attempt to follow improvements and developments in
320the C++ language and library reasonably closely. As such, we require a modern
321host C++ toolchain, both compiler and standard library, in order to build LLVM.
322
323LLVM is written using the subset of C++ documented in :doc:`coding
324standards<CodingStandards>`. To enforce this language version, we check the most
325popular host toolchains for specific minimum versions in our build systems:
326
327* Clang 5.0
328* Apple Clang 10.0
329* GCC 7.1
330* Visual Studio 2019 16.7
331
332Anything older than these toolchains *may* work, but will require forcing the
333build system with a special option and is not really a supported host platform.
334Also note that older versions of these compilers have often crashed or
335miscompiled LLVM.
336
337For less widely used host toolchains such as ICC or xlC, be aware that a very
338recent version may be required to support all of the C++ features used in LLVM.
339
340We track certain versions of software that are *known* to fail when used as
341part of the host toolchain. These even include linkers at times.
342
343**GNU ld 2.16.X**. Some 2.16.X versions of the ld linker will produce very long
344warning messages complaining that some "``.gnu.linkonce.t.*``" symbol was
345defined in a discarded section. You can safely ignore these messages as they are
346erroneous and the linkage is correct.  These messages disappear using ld 2.17.
347
348**GNU binutils 2.17**: Binutils 2.17 contains `a bug
349<http://sourceware.org/bugzilla/show_bug.cgi?id=3111>`__ which causes huge link
350times (minutes instead of seconds) when building LLVM.  We recommend upgrading
351to a newer version (2.17.50.0.4 or later).
352
353**GNU Binutils 2.19.1 Gold**: This version of Gold contained `a bug
354<http://sourceware.org/bugzilla/show_bug.cgi?id=9836>`__ which causes
355intermittent failures when building LLVM with position independent code.  The
356symptom is an error about cyclic dependencies.  We recommend upgrading to a
357newer version of Gold.
358
359Getting a Modern Host C++ Toolchain
360^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
361
362This section mostly applies to Linux and older BSDs. On macOS, you should
363have a sufficiently modern Xcode, or you will likely need to upgrade until you
364do. Windows does not have a "system compiler", so you must install either Visual
365Studio 2019 (or later), or a recent version of mingw64. FreeBSD 10.0 and newer
366have a modern Clang as the system compiler.
367
368However, some Linux distributions and some other or older BSDs sometimes have
369extremely old versions of GCC. These steps attempt to help you upgrade you
370compiler even on such a system. However, if at all possible, we encourage you
371to use a recent version of a distribution with a modern system compiler that
372meets these requirements. Note that it is tempting to install a prior
373version of Clang and libc++ to be the host compiler, however libc++ was not
374well tested or set up to build on Linux until relatively recently. As
375a consequence, this guide suggests just using libstdc++ and a modern GCC as the
376initial host in a bootstrap, and then using Clang (and potentially libc++).
377
378The first step is to get a recent GCC toolchain installed. The most common
379distribution on which users have struggled with the version requirements is
380Ubuntu Precise, 12.04 LTS. For this distribution, one easy option is to install
381the `toolchain testing PPA`_ and use it to install a modern GCC. There is
382a really nice discussions of this on the `ask ubuntu stack exchange`_ and a
383`github gist`_ with updated commands. However, not all users can use PPAs and
384there are many other distributions, so it may be necessary (or just useful, if
385you're here you *are* doing compiler development after all) to build and install
386GCC from source. It is also quite easy to do these days.
387
388.. _toolchain testing PPA:
389  https://launchpad.net/~ubuntu-toolchain-r/+archive/test
390.. _ask ubuntu stack exchange:
391  https://askubuntu.com/questions/466651/how-do-i-use-the-latest-gcc-on-ubuntu/581497#58149
392.. _github gist:
393  https://gist.github.com/application2000/73fd6f4bf1be6600a2cf9f56315a2d91
394
395Easy steps for installing GCC 7.1.0:
396
397.. code-block:: console
398
399  % gcc_version=7.1.0
400  % wget https://ftp.gnu.org/gnu/gcc/gcc-${gcc_version}/gcc-${gcc_version}.tar.bz2
401  % wget https://ftp.gnu.org/gnu/gcc/gcc-${gcc_version}/gcc-${gcc_version}.tar.bz2.sig
402  % wget https://ftp.gnu.org/gnu/gnu-keyring.gpg
403  % signature_invalid=`gpg --verify --no-default-keyring --keyring ./gnu-keyring.gpg gcc-${gcc_version}.tar.bz2.sig`
404  % if [ $signature_invalid ]; then echo "Invalid signature" ; exit 1 ; fi
405  % tar -xvjf gcc-${gcc_version}.tar.bz2
406  % cd gcc-${gcc_version}
407  % ./contrib/download_prerequisites
408  % cd ..
409  % mkdir gcc-${gcc_version}-build
410  % cd gcc-${gcc_version}-build
411  % $PWD/../gcc-${gcc_version}/configure --prefix=$HOME/toolchains --enable-languages=c,c++
412  % make -j$(nproc)
413  % make install
414
415For more details, check out the excellent `GCC wiki entry`_, where I got most
416of this information from.
417
418.. _GCC wiki entry:
419  https://gcc.gnu.org/wiki/InstallingGCC
420
421Once you have a GCC toolchain, configure your build of LLVM to use the new
422toolchain for your host compiler and C++ standard library. Because the new
423version of libstdc++ is not on the system library search path, you need to pass
424extra linker flags so that it can be found at link time (``-L``) and at runtime
425(``-rpath``). If you are using CMake, this invocation should produce working
426binaries:
427
428.. code-block:: console
429
430  % mkdir build
431  % cd build
432  % CC=$HOME/toolchains/bin/gcc CXX=$HOME/toolchains/bin/g++ \
433    cmake .. -DCMAKE_CXX_LINK_FLAGS="-Wl,-rpath,$HOME/toolchains/lib64 -L$HOME/toolchains/lib64"
434
435If you fail to set rpath, most LLVM binaries will fail on startup with a message
436from the loader similar to ``libstdc++.so.6: version `GLIBCXX_3.4.20' not
437found``. This means you need to tweak the -rpath linker flag.
438
439This method will add an absolute path to the rpath of all executables. That's
440fine for local development. If you want to distribute the binaries you build
441so that they can run on older systems, copy ``libstdc++.so.6`` into the
442``lib/`` directory.  All of LLVM's shipping binaries have an rpath pointing at
443``$ORIGIN/../lib``, so they will find ``libstdc++.so.6`` there.  Non-distributed
444binaries don't have an rpath set and won't find ``libstdc++.so.6``. Pass
445``-DLLVM_LOCAL_RPATH="$HOME/toolchains/lib64"`` to cmake to add an absolute
446path to ``libstdc++.so.6`` as above. Since these binaries are not distributed,
447having an absolute local path is fine for them.
448
449When you build Clang, you will need to give *it* access to modern C++
450standard library in order to use it as your new host in part of a bootstrap.
451There are two easy ways to do this, either build (and install) libc++ along
452with Clang and then use it with the ``-stdlib=libc++`` compile and link flag,
453or install Clang into the same prefix (``$HOME/toolchains`` above) as GCC.
454Clang will look within its own prefix for libstdc++ and use it if found. You
455can also add an explicit prefix for Clang to look in for a GCC toolchain with
456the ``--gcc-toolchain=/opt/my/gcc/prefix`` flag, passing it to both compile and
457link commands when using your just-built-Clang to bootstrap.
458
459.. _Getting Started with LLVM:
460
461Getting Started with LLVM
462=========================
463
464The remainder of this guide is meant to get you up and running with LLVM and to
465give you some basic information about the LLVM environment.
466
467The later sections of this guide describe the `general layout`_ of the LLVM
468source tree, a `simple example`_ using the LLVM tool chain, and `links`_ to find
469more information about LLVM or to get help via e-mail.
470
471Terminology and Notation
472------------------------
473
474Throughout this manual, the following names are used to denote paths specific to
475the local system and working environment.  *These are not environment variables
476you need to set but just strings used in the rest of this document below*.  In
477any of the examples below, simply replace each of these names with the
478appropriate pathname on your local system.  All these paths are absolute:
479
480``SRC_ROOT``
481
482  This is the top level directory of the LLVM source tree.
483
484``OBJ_ROOT``
485
486  This is the top level directory of the LLVM object tree (i.e. the tree where
487  object files and compiled programs will be placed.  It can be the same as
488  SRC_ROOT).
489
490Unpacking the LLVM Archives
491---------------------------
492
493If you have the LLVM distribution, you will need to unpack it before you can
494begin to compile it.  LLVM is distributed as a number of different
495subprojects. Each one has its own download which is a TAR archive that is
496compressed with the gzip program.
497
498The files are as follows, with *x.y* marking the version number:
499
500``llvm-x.y.tar.gz``
501
502  Source release for the LLVM libraries and tools.
503
504``cfe-x.y.tar.gz``
505
506  Source release for the Clang frontend.
507
508.. _checkout:
509
510Checkout LLVM from Git
511----------------------
512
513You can also checkout the source code for LLVM from Git.
514
515.. note::
516
517  Passing ``--config core.autocrlf=false`` should not be required in
518  the future after we adjust the .gitattribute settings correctly, but
519  is required for Windows users at the time of this writing.
520
521Simply run:
522
523.. code-block:: console
524
525  % git clone https://github.com/llvm/llvm-project.git
526
527or on Windows,
528
529.. code-block:: console
530
531  % git clone --config core.autocrlf=false https://github.com/llvm/llvm-project.git
532
533This will create an '``llvm-project``' directory in the current directory and
534fully populate it with all of the source code, test directories, and local
535copies of documentation files for LLVM and all the related subprojects. Note
536that unlike the tarballs, which contain each subproject in a separate file, the
537git repository contains all of the projects together.
538
539If you want to get a specific release (as opposed to the most recent revision),
540you can check out a tag after cloning the repository. E.g., `git checkout
541llvmorg-6.0.1` inside the ``llvm-project`` directory created by the above
542command.  Use `git tag -l` to list all of them.
543
544Sending patches
545^^^^^^^^^^^^^^^
546
547See :ref:`Contributing <submit_patch>`.
548
549Bisecting commits
550^^^^^^^^^^^^^^^^^
551
552See `Bisecting LLVM code <GitBisecting.html>`_ for how to use ``git bisect``
553on LLVM.
554
555Reverting a change
556^^^^^^^^^^^^^^^^^^
557
558When reverting changes using git, the default message will say "This reverts
559commit XYZ". Leave this at the end of the commit message, but add some details
560before it as to why the commit is being reverted. A brief explanation and/or
561links to bots that demonstrate the problem are sufficient.
562
563Local LLVM Configuration
564------------------------
565
566Once checked out repository, the LLVM suite source code must be configured
567before being built. This process uses CMake.  Unlinke the normal ``configure``
568script, CMake generates the build files in whatever format you request as well
569as various ``*.inc`` files, and ``llvm/include/Config/config.h``.
570
571Variables are passed to ``cmake`` on the command line using the format
572``-D<variable name>=<value>``. The following variables are some common options
573used by people developing LLVM.
574
575+-------------------------+----------------------------------------------------+
576| Variable                | Purpose                                            |
577+=========================+====================================================+
578| CMAKE_C_COMPILER        | Tells ``cmake`` which C compiler to use. By        |
579|                         | default, this will be /usr/bin/cc.                 |
580+-------------------------+----------------------------------------------------+
581| CMAKE_CXX_COMPILER      | Tells ``cmake`` which C++ compiler to use. By      |
582|                         | default, this will be /usr/bin/c++.                |
583+-------------------------+----------------------------------------------------+
584| CMAKE_BUILD_TYPE        | Tells ``cmake`` what type of build you are trying  |
585|                         | to generate files for. Valid options are Debug,    |
586|                         | Release, RelWithDebInfo, and MinSizeRel. Default   |
587|                         | is Debug.                                          |
588+-------------------------+----------------------------------------------------+
589| CMAKE_INSTALL_PREFIX    | Specifies the install directory to target when     |
590|                         | running the install action of the build files.     |
591+-------------------------+----------------------------------------------------+
592| Python3_EXECUTABLE      | Forces CMake to use a specific Python version by   |
593|                         | passing a path to a Python interpreter. By default |
594|                         | the Python version of the interpreter in your PATH |
595|                         | is used.                                           |
596+-------------------------+----------------------------------------------------+
597| LLVM_TARGETS_TO_BUILD   | A semicolon delimited list controlling which       |
598|                         | targets will be built and linked into llvm.        |
599|                         | The default list is defined as                     |
600|                         | ``LLVM_ALL_TARGETS``, and can be set to include    |
601|                         | out-of-tree targets. The default value includes:   |
602|                         | ``AArch64, AMDGPU, ARM, AVR, BPF, Hexagon, Lanai,  |
603|                         | Mips, MSP430, NVPTX, PowerPC, RISCV, Sparc,        |
604|                         | SystemZ, WebAssembly, X86, XCore``.                |
605|                         |                                                    |
606+-------------------------+----------------------------------------------------+
607| LLVM_ENABLE_DOXYGEN     | Build doxygen-based documentation from the source  |
608|                         | code This is disabled by default because it is     |
609|                         | slow and generates a lot of output.                |
610+-------------------------+----------------------------------------------------+
611| LLVM_ENABLE_PROJECTS    | A semicolon-delimited list selecting which of the  |
612|                         | other LLVM subprojects to additionally build. (Only|
613|                         | effective when using a side-by-side project layout |
614|                         | e.g. via git). The default list is empty. Can      |
615|                         | include: clang, clang-tools-extra,                 |
616|                         | cross-project-tests, flang, libc, libclc, lld,     |
617|                         | lldb, mlir, openmp, polly, or pstl.                |
618+-------------------------+----------------------------------------------------+
619| LLVM_ENABLE_RUNTIMES    | A semicolon-delimited list selecting which of the  |
620|                         | runtimes to build. (Only effective when using the  |
621|                         | full monorepo layout). The default list is empty.  |
622|                         | Can include: compiler-rt, libc, libcxx, libcxxabi, |
623|                         | libunwind, or openmp.                              |
624+-------------------------+----------------------------------------------------+
625| LLVM_ENABLE_SPHINX      | Build sphinx-based documentation from the source   |
626|                         | code. This is disabled by default because it is    |
627|                         | slow and generates a lot of output. Sphinx version |
628|                         | 1.5 or later recommended.                          |
629+-------------------------+----------------------------------------------------+
630| LLVM_BUILD_LLVM_DYLIB   | Generate libLLVM.so. This library contains a       |
631|                         | default set of LLVM components that can be         |
632|                         | overridden with ``LLVM_DYLIB_COMPONENTS``. The     |
633|                         | default contains most of LLVM and is defined in    |
634|                         | ``tools/llvm-shlib/CMakelists.txt``. This option is|
635|                         | not available on Windows.                          |
636+-------------------------+----------------------------------------------------+
637| LLVM_OPTIMIZED_TABLEGEN | Builds a release tablegen that gets used during    |
638|                         | the LLVM build. This can dramatically speed up     |
639|                         | debug builds.                                      |
640+-------------------------+----------------------------------------------------+
641
642To configure LLVM, follow these steps:
643
644#. Change directory into the object root directory:
645
646   .. code-block:: console
647
648     % cd OBJ_ROOT
649
650#. Run the ``cmake``:
651
652   .. code-block:: console
653
654     % cmake -G "Unix Makefiles" -DCMAKE_BUILD_TYPE=<type> -DCMAKE_INSTALL_PREFIX=/install/path
655       [other options] SRC_ROOT
656
657Compiling the LLVM Suite Source Code
658------------------------------------
659
660Unlike with autotools, with CMake your build type is defined at configuration.
661If you want to change your build type, you can re-run cmake with the following
662invocation:
663
664   .. code-block:: console
665
666     % cmake -G "Unix Makefiles" -DCMAKE_BUILD_TYPE=<type> -DCMAKE_BUILD_TYPE=type SRC_ROOT
667
668Between runs, CMake preserves the values set for all options. CMake has the
669following build types defined:
670
671Debug
672
673  These builds are the default. The build system will compile the tools and
674  libraries unoptimized, with debugging information, and asserts enabled.
675
676Release
677
678  For these builds, the build system will compile the tools and libraries
679  with optimizations enabled and not generate debug info. CMakes default
680  optimization level is -O3. This can be configured by setting the
681  ``CMAKE_CXX_FLAGS_RELEASE`` variable on the CMake command line.
682
683RelWithDebInfo
684
685  These builds are useful when debugging. They generate optimized binaries with
686  debug information. CMakes default optimization level is -O2. This can be
687  configured by setting the ``CMAKE_CXX_FLAGS_RELWITHDEBINFO`` variable on the
688  CMake command line.
689
690Once you have LLVM configured, you can build it by entering the *OBJ_ROOT*
691directory and issuing the following command:
692
693.. code-block:: console
694
695  % make
696
697If the build fails, please `check here`_ to see if you are using a version of
698GCC that is known not to compile LLVM.
699
700If you have multiple processors in your machine, you may wish to use some of the
701parallel build options provided by GNU Make.  For example, you could use the
702command:
703
704.. code-block:: console
705
706  % make -j2
707
708There are several special targets which are useful when working with the LLVM
709source code:
710
711``make clean``
712
713  Removes all files generated by the build.  This includes object files,
714  generated C/C++ files, libraries, and executables.
715
716``make install``
717
718  Installs LLVM header files, libraries, tools, and documentation in a hierarchy
719  under ``$PREFIX``, specified with ``CMAKE_INSTALL_PREFIX``, which
720  defaults to ``/usr/local``.
721
722``make docs-llvm-html``
723
724  If configured with ``-DLLVM_ENABLE_SPHINX=On``, this will generate a directory
725  at ``OBJ_ROOT/docs/html`` which contains the HTML formatted documentation.
726
727Cross-Compiling LLVM
728--------------------
729
730It is possible to cross-compile LLVM itself. That is, you can create LLVM
731executables and libraries to be hosted on a platform different from the platform
732where they are built (a Canadian Cross build). To generate build files for
733cross-compiling CMake provides a variable ``CMAKE_TOOLCHAIN_FILE`` which can
734define compiler flags and variables used during the CMake test operations.
735
736The result of such a build is executables that are not runnable on the build
737host but can be executed on the target. As an example the following CMake
738invocation can generate build files targeting iOS. This will work on macOS
739with the latest Xcode:
740
741.. code-block:: console
742
743  % cmake -G "Ninja" -DCMAKE_OSX_ARCHITECTURES="armv7;armv7s;arm64"
744    -DCMAKE_TOOLCHAIN_FILE=<PATH_TO_LLVM>/cmake/platforms/iOS.cmake
745    -DCMAKE_BUILD_TYPE=Release -DLLVM_BUILD_RUNTIME=Off -DLLVM_INCLUDE_TESTS=Off
746    -DLLVM_INCLUDE_EXAMPLES=Off -DLLVM_ENABLE_BACKTRACES=Off [options]
747    <PATH_TO_LLVM>
748
749Note: There are some additional flags that need to be passed when building for
750iOS due to limitations in the iOS SDK.
751
752Check :doc:`HowToCrossCompileLLVM` and `Clang docs on how to cross-compile in general
753<https://clang.llvm.org/docs/CrossCompilation.html>`_ for more information
754about cross-compiling.
755
756The Location of LLVM Object Files
757---------------------------------
758
759The LLVM build system is capable of sharing a single LLVM source tree among
760several LLVM builds.  Hence, it is possible to build LLVM for several different
761platforms or configurations using the same source tree.
762
763* Change directory to where the LLVM object files should live:
764
765  .. code-block:: console
766
767    % cd OBJ_ROOT
768
769* Run ``cmake``:
770
771  .. code-block:: console
772
773    % cmake -G "Unix Makefiles" -DCMAKE_BUILD_TYPE=Release SRC_ROOT
774
775The LLVM build will create a structure underneath *OBJ_ROOT* that matches the
776LLVM source tree. At each level where source files are present in the source
777tree there will be a corresponding ``CMakeFiles`` directory in the *OBJ_ROOT*.
778Underneath that directory there is another directory with a name ending in
779``.dir`` under which you'll find object files for each source.
780
781For example:
782
783  .. code-block:: console
784
785    % cd llvm_build_dir
786    % find lib/Support/ -name APFloat*
787    lib/Support/CMakeFiles/LLVMSupport.dir/APFloat.cpp.o
788
789Optional Configuration Items
790----------------------------
791
792If you're running on a Linux system that supports the `binfmt_misc
793<http://en.wikipedia.org/wiki/binfmt_misc>`_
794module, and you have root access on the system, you can set your system up to
795execute LLVM bitcode files directly. To do this, use commands like this (the
796first command may not be required if you are already using the module):
797
798.. code-block:: console
799
800  % mount -t binfmt_misc none /proc/sys/fs/binfmt_misc
801  % echo ':llvm:M::BC::/path/to/lli:' > /proc/sys/fs/binfmt_misc/register
802  % chmod u+x hello.bc   (if needed)
803  % ./hello.bc
804
805This allows you to execute LLVM bitcode files directly.  On Debian, you can also
806use this command instead of the 'echo' command above:
807
808.. code-block:: console
809
810  % sudo update-binfmts --install llvm /path/to/lli --magic 'BC'
811
812.. _Program Layout:
813.. _general layout:
814
815Directory Layout
816================
817
818One useful source of information about the LLVM source base is the LLVM `doxygen
819<http://www.doxygen.org/>`_ documentation available at
820`<https://llvm.org/doxygen/>`_.  The following is a brief introduction to code
821layout:
822
823``llvm/cmake``
824--------------
825Generates system build files.
826
827``llvm/cmake/modules``
828  Build configuration for llvm user defined options. Checks compiler version and
829  linker flags.
830
831``llvm/cmake/platforms``
832  Toolchain configuration for Android NDK, iOS systems and non-Windows hosts to
833  target MSVC.
834
835``llvm/examples``
836-----------------
837
838- Some simple examples showing how to use LLVM as a compiler for a custom
839  language - including lowering, optimization, and code generation.
840
841- Kaleidoscope Tutorial: Kaleidoscope language tutorial run through the
842  implementation of a nice little compiler for a non-trivial language
843  including a hand-written lexer, parser, AST, as well as code generation
844  support using LLVM- both static (ahead of time) and various approaches to
845  Just In Time (JIT) compilation.
846  `Kaleidoscope Tutorial for complete beginner
847  <https://llvm.org/docs/tutorial/MyFirstLanguageFrontend/index.html>`_.
848
849- BuildingAJIT: Examples of the `BuildingAJIT tutorial
850  <https://llvm.org/docs/tutorial/BuildingAJIT1.html>`_ that shows how LLVM’s
851  ORC JIT APIs interact with other parts of LLVM. It also, teaches how to
852  recombine them to build a custom JIT that is suited to your use-case.
853
854``llvm/include``
855----------------
856
857Public header files exported from the LLVM library. The three main subdirectories:
858
859``llvm/include/llvm``
860
861  All LLVM-specific header files, and  subdirectories for different portions of
862  LLVM: ``Analysis``, ``CodeGen``, ``Target``, ``Transforms``, etc...
863
864``llvm/include/llvm/Support``
865
866  Generic support libraries provided with LLVM but not necessarily specific to
867  LLVM. For example, some C++ STL utilities and a Command Line option processing
868  library store header files here.
869
870``llvm/include/llvm/Config``
871
872  Header files configured by ``cmake``.  They wrap "standard" UNIX and
873  C header files.  Source code can include these header files which
874  automatically take care of the conditional #includes that ``cmake``
875  generates.
876
877``llvm/lib``
878------------
879
880Most source files are here. By putting code in libraries, LLVM makes it easy to
881share code among the `tools`_.
882
883``llvm/lib/IR/``
884
885  Core LLVM source files that implement core classes like Instruction and
886  BasicBlock.
887
888``llvm/lib/AsmParser/``
889
890  Source code for the LLVM assembly language parser library.
891
892``llvm/lib/Bitcode/``
893
894  Code for reading and writing bitcode.
895
896``llvm/lib/Analysis/``
897
898  A variety of program analyses, such as Call Graphs, Induction Variables,
899  Natural Loop Identification, etc.
900
901``llvm/lib/Transforms/``
902
903  IR-to-IR program transformations, such as Aggressive Dead Code Elimination,
904  Sparse Conditional Constant Propagation, Inlining, Loop Invariant Code Motion,
905  Dead Global Elimination, and many others.
906
907``llvm/lib/Target/``
908
909  Files describing target architectures for code generation.  For example,
910  ``llvm/lib/Target/X86`` holds the X86 machine description.
911
912``llvm/lib/CodeGen/``
913
914  The major parts of the code generator: Instruction Selector, Instruction
915  Scheduling, and Register Allocation.
916
917``llvm/lib/MC/``
918
919  The libraries represent and process code at machine code level. Handles
920  assembly and object-file emission.
921
922``llvm/lib/ExecutionEngine/``
923
924  Libraries for directly executing bitcode at runtime in interpreted and
925  JIT-compiled scenarios.
926
927``llvm/lib/Support/``
928
929  Source code that corresponding to the header files in ``llvm/include/ADT/``
930  and ``llvm/include/Support/``.
931
932``llvm/bindings``
933----------------------
934
935Contains bindings for the LLVM compiler infrastructure to allow
936programs written in languages other than C or C++ to take advantage of the LLVM
937infrastructure.
938LLVM project provides language bindings for OCaml and Python.
939
940``llvm/projects``
941-----------------
942
943Projects not strictly part of LLVM but shipped with LLVM. This is also the
944directory for creating your own LLVM-based projects which leverage the LLVM
945build system.
946
947``llvm/test``
948-------------
949
950Feature and regression tests and other sanity checks on LLVM infrastructure. These
951are intended to run quickly and cover a lot of territory without being exhaustive.
952
953``test-suite``
954--------------
955
956A comprehensive correctness, performance, and benchmarking test suite
957for LLVM.  This comes in a ``separate git repository
958<https://github.com/llvm/llvm-test-suite>``, because it contains a
959large amount of third-party code under a variety of licenses. For
960details see the :doc:`Testing Guide <TestingGuide>` document.
961
962.. _tools:
963
964``llvm/tools``
965--------------
966
967Executables built out of the libraries
968above, which form the main part of the user interface.  You can always get help
969for a tool by typing ``tool_name -help``.  The following is a brief introduction
970to the most important tools.  More detailed information is in
971the `Command Guide <CommandGuide/index.html>`_.
972
973``bugpoint``
974
975  ``bugpoint`` is used to debug optimization passes or code generation backends
976  by narrowing down the given test case to the minimum number of passes and/or
977  instructions that still cause a problem, whether it is a crash or
978  miscompilation. See `<HowToSubmitABug.html>`_ for more information on using
979  ``bugpoint``.
980
981``llvm-ar``
982
983  The archiver produces an archive containing the given LLVM bitcode files,
984  optionally with an index for faster lookup.
985
986``llvm-as``
987
988  The assembler transforms the human readable LLVM assembly to LLVM bitcode.
989
990``llvm-dis``
991
992  The disassembler transforms the LLVM bitcode to human readable LLVM assembly.
993
994``llvm-link``
995
996  ``llvm-link``, not surprisingly, links multiple LLVM modules into a single
997  program.
998
999``lli``
1000
1001  ``lli`` is the LLVM interpreter, which can directly execute LLVM bitcode
1002  (although very slowly...). For architectures that support it (currently x86,
1003  Sparc, and PowerPC), by default, ``lli`` will function as a Just-In-Time
1004  compiler (if the functionality was compiled in), and will execute the code
1005  *much* faster than the interpreter.
1006
1007``llc``
1008
1009  ``llc`` is the LLVM backend compiler, which translates LLVM bitcode to a
1010  native code assembly file.
1011
1012``opt``
1013
1014  ``opt`` reads LLVM bitcode, applies a series of LLVM to LLVM transformations
1015  (which are specified on the command line), and outputs the resultant
1016  bitcode.   '``opt -help``'  is a good way to get a list of the
1017  program transformations available in LLVM.
1018
1019  ``opt`` can also  run a specific analysis on an input LLVM bitcode
1020  file and print  the results.  Primarily useful for debugging
1021  analyses, or familiarizing yourself with what an analysis does.
1022
1023``llvm/utils``
1024--------------
1025
1026Utilities for working with LLVM source code; some are part of the build process
1027because they are code generators for parts of the infrastructure.
1028
1029
1030``codegen-diff``
1031
1032  ``codegen-diff`` finds differences between code that LLC
1033  generates and code that LLI generates. This is useful if you are
1034  debugging one of them, assuming that the other generates correct output. For
1035  the full user manual, run ```perldoc codegen-diff'``.
1036
1037``emacs/``
1038
1039   Emacs and XEmacs syntax highlighting  for LLVM   assembly files and TableGen
1040   description files.  See the ``README`` for information on using them.
1041
1042``getsrcs.sh``
1043
1044  Finds and outputs all non-generated source files,
1045  useful if one wishes to do a lot of development across directories
1046  and does not want to find each file. One way to use it is to run,
1047  for example: ``xemacs `utils/getsources.sh``` from the top of the LLVM source
1048  tree.
1049
1050``llvmgrep``
1051
1052  Performs an ``egrep -H -n`` on each source file in LLVM and
1053  passes to it a regular expression provided on ``llvmgrep``'s command
1054  line. This is an efficient way of searching the source base for a
1055  particular regular expression.
1056
1057``TableGen/``
1058
1059  Contains the tool used to generate register
1060  descriptions, instruction set descriptions, and even assemblers from common
1061  TableGen description files.
1062
1063``vim/``
1064
1065  vim syntax-highlighting for LLVM assembly files
1066  and TableGen description files. See the    ``README`` for how to use them.
1067
1068.. _simple example:
1069
1070An Example Using the LLVM Tool Chain
1071====================================
1072
1073This section gives an example of using LLVM with the Clang front end.
1074
1075Example with clang
1076------------------
1077
1078#. First, create a simple C file, name it 'hello.c':
1079
1080   .. code-block:: c
1081
1082     #include <stdio.h>
1083
1084     int main() {
1085       printf("hello world\n");
1086       return 0;
1087     }
1088
1089#. Next, compile the C file into a native executable:
1090
1091   .. code-block:: console
1092
1093     % clang hello.c -o hello
1094
1095   .. note::
1096
1097     Clang works just like GCC by default.  The standard -S and -c arguments
1098     work as usual (producing a native .s or .o file, respectively).
1099
1100#. Next, compile the C file into an LLVM bitcode file:
1101
1102   .. code-block:: console
1103
1104     % clang -O3 -emit-llvm hello.c -c -o hello.bc
1105
1106   The -emit-llvm option can be used with the -S or -c options to emit an LLVM
1107   ``.ll`` or ``.bc`` file (respectively) for the code.  This allows you to use
1108   the `standard LLVM tools <CommandGuide/index.html>`_ on the bitcode file.
1109
1110#. Run the program in both forms. To run the program, use:
1111
1112   .. code-block:: console
1113
1114      % ./hello
1115
1116   and
1117
1118   .. code-block:: console
1119
1120     % lli hello.bc
1121
1122   The second examples shows how to invoke the LLVM JIT, :doc:`lli
1123   <CommandGuide/lli>`.
1124
1125#. Use the ``llvm-dis`` utility to take a look at the LLVM assembly code:
1126
1127   .. code-block:: console
1128
1129     % llvm-dis < hello.bc | less
1130
1131#. Compile the program to native assembly using the LLC code generator:
1132
1133   .. code-block:: console
1134
1135     % llc hello.bc -o hello.s
1136
1137#. Assemble the native assembly language file into a program:
1138
1139   .. code-block:: console
1140
1141     % /opt/SUNWspro/bin/cc -xarch=v9 hello.s -o hello.native   # On Solaris
1142
1143     % gcc hello.s -o hello.native                              # On others
1144
1145#. Execute the native code program:
1146
1147   .. code-block:: console
1148
1149     % ./hello.native
1150
1151   Note that using clang to compile directly to native code (i.e. when the
1152   ``-emit-llvm`` option is not present) does steps 6/7/8 for you.
1153
1154Common Problems
1155===============
1156
1157If you are having problems building or using LLVM, or if you have any other
1158general questions about LLVM, please consult the `Frequently Asked
1159Questions <FAQ.html>`_ page.
1160
1161If you are having problems with limited memory and build time, please try
1162building with ninja instead of make. Please consider configuring the
1163following options with cmake:
1164
1165 * -G Ninja
1166   Setting this option will allow you to build with ninja instead of make.
1167   Building with ninja significantly improves your build time, especially with
1168   incremental builds, and improves your memory usage.
1169
1170 * -DLLVM_USE_LINKER
1171   Setting this option to lld will significantly reduce linking time for LLVM
1172   executables on ELF-based platforms, such as Linux. If you are building LLVM
1173   for the first time and lld is not available to you as a binary package, then
1174   you may want to use the gold linker as a faster alternative to GNU ld.
1175
1176 * -DCMAKE_BUILD_TYPE
1177   Controls optimization level and debug information of the build.  This setting
1178   can affect RAM and disk usage, see :ref:`CMAKE_BUILD_TYPE <cmake_build_type>`
1179   for more information.
1180
1181 * -DLLVM_ENABLE_ASSERTIONS
1182   This option defaults to ON for Debug builds and defaults to OFF for Release
1183   builds. As mentioned in the previous option, using the Release build type and
1184   enabling assertions may be a good alternative to using the Debug build type.
1185
1186 * -DLLVM_PARALLEL_LINK_JOBS
1187   Set this equal to number of jobs you wish to run simultaneously. This is
1188   similar to the -j option used with make, but only for link jobs. This option
1189   can only be used with ninja. You may wish to use a very low number of jobs,
1190   as this will greatly reduce the amount of memory used during the build
1191   process. If you have limited memory, you may wish to set this to 1.
1192
1193 * -DLLVM_TARGETS_TO_BUILD
1194   Set this equal to the target you wish to build. You may wish to set this to
1195   X86; however, you will find a full list of targets within the
1196   llvm-project/llvm/lib/Target directory.
1197
1198 * -DLLVM_OPTIMIZED_TABLEGEN
1199   Set this to ON to generate a fully optimized tablegen during your build. This
1200   will significantly improve your build time. This is only useful if you are
1201   using the Debug build type.
1202
1203 * -DLLVM_ENABLE_PROJECTS
1204   Set this equal to the projects you wish to compile (e.g. clang, lld, etc.) If
1205   compiling more than one project, separate the items with a semicolon. Should
1206   you run into issues with the semicolon, try surrounding it with single quotes.
1207
1208 * -DLLVM_ENABLE_RUNTIMES
1209   Set this equal to the runtimes you wish to compile (e.g. libcxx, libcxxabi, etc.)
1210   If compiling more than one runtime, separate the items with a semicolon. Should
1211   you run into issues with the semicolon, try surrounding it with single quotes.
1212
1213 * -DCLANG_ENABLE_STATIC_ANALYZER
1214   Set this option to OFF if you do not require the clang static analyzer. This
1215   should improve your build time slightly.
1216
1217 * -DLLVM_USE_SPLIT_DWARF
1218   Consider setting this to ON if you require a debug build, as this will ease
1219   memory pressure on the linker. This will make linking much faster, as the
1220   binaries will not contain any of the debug information; however, this will
1221   generate the debug information in the form of a DWARF object file (with the
1222   extension .dwo). This only applies to host platforms using ELF, such as Linux.
1223
1224.. _links:
1225
1226Links
1227=====
1228
1229This document is just an **introduction** on how to use LLVM to do some simple
1230things... there are many more interesting and complicated things that you can do
1231that aren't documented here (but we'll gladly accept a patch if you want to
1232write something up!).  For more information about LLVM, check out:
1233
1234* `LLVM Homepage <https://llvm.org/>`_
1235* `LLVM Doxygen Tree <https://llvm.org/doxygen/>`_
1236* `Starting a Project that Uses LLVM <https://llvm.org/docs/Projects.html>`_
1237