1 =============================
2 Advanced Build Configurations
3 =============================
11 `CMake <http://www.cmake.org/>`_ is a cross-platform build-generator tool. CMake
12 does not build the project, it generates the files needed by your build tool
13 (GNU make, Visual Studio, etc.) for building LLVM.
15 If **you are a new contributor**, please start with the :doc:`GettingStarted` or
16 :doc:`CMake` pages. This page is intended for users doing more complex builds.
18 Many of the examples below are written assuming specific CMake Generators.
19 Unless otherwise explicitly called out these commands should work with any CMake
25 The Clang CMake build system supports bootstrap (aka multi-stage) builds. At a
26 high level a multi-stage build is a chain of builds that pass data from one
27 stage into the next. The most common and simple version of this is a traditional
30 In a simple two-stage bootstrap build, we build clang using the system compiler,
31 then use that just-built clang to build clang again. In CMake this simplest form
32 of a bootstrap build can be configured with a single option,
33 CLANG_ENABLE_BOOTSTRAP.
35 .. code-block:: console
37 $ cmake -G Ninja -DCLANG_ENABLE_BOOTSTRAP=On <path to source>
40 This command itself isn't terribly useful because it assumes default
41 configurations for each stage. The next series of examples utilize CMake cache
42 scripts to provide more complex options.
44 By default, only a few CMake options will be passed between stages.
45 The list, called _BOOTSTRAP_DEFAULT_PASSTHROUGH, is defined in clang/CMakeLists.txt.
46 To force the passing of the variables between stages, use the -DCLANG_BOOTSTRAP_PASSTHROUGH
47 CMake option, each variable separated by a ";". As example:
49 .. code-block:: console
51 $ cmake -G Ninja -DCLANG_ENABLE_BOOTSTRAP=On -DCLANG_BOOTSTRAP_PASSTHROUGH="CMAKE_INSTALL_PREFIX;CMAKE_VERBOSE_MAKEFILE" <path to source>
54 CMake options starting by ``BOOTSTRAP_`` will be passed only to the stage2 build.
55 This gives the opportunity to use Clang specific build flags.
56 For example, the following CMake call will enabled '-fno-addrsig' only during
57 the stage2 build for C and C++.
59 .. code-block:: console
61 $ cmake [..] -DBOOTSTRAP_CMAKE_CXX_FLAGS='-fno-addrsig' -DBOOTSTRAP_CMAKE_C_FLAGS='-fno-addrsig' [..]
63 The clang build system refers to builds as stages. A stage1 build is a standard
64 build using the compiler installed on the host, and a stage2 build is built
65 using the stage1 compiler. This nomenclature holds up to more stages too. In
66 general a stage*n* build is built using the output from stage*n-1*.
68 Apple Clang Builds (A More Complex Bootstrap)
69 =============================================
71 Apple's Clang builds are a slightly more complicated example of the simple
72 bootstrapping scenario. Apple Clang is built using a 2-stage build.
74 The stage1 compiler is a host-only compiler with some options set. The stage1
75 compiler is a balance of optimization vs build time because it is a throwaway.
76 The stage2 compiler is the fully optimized compiler intended to ship to users.
78 Setting up these compilers requires a lot of options. To simplify the
79 configuration the Apple Clang build settings are contained in CMake Cache files.
80 You can build an Apple Clang compiler using the following commands:
82 .. code-block:: console
84 $ cmake -G Ninja -C <path to clang>/cmake/caches/Apple-stage1.cmake <path to source>
85 $ ninja stage2-distribution
87 This CMake invocation configures the stage1 host compiler, and sets
88 CLANG_BOOTSTRAP_CMAKE_ARGS to pass the Apple-stage2.cmake cache script to the
89 stage2 configuration step.
91 When you build the stage2-distribution target it builds the minimal stage1
92 compiler and required tools, then configures and builds the stage2 compiler
93 based on the settings in Apple-stage2.cmake.
95 This pattern of using cache scripts to set complex settings, and specifically to
96 make later stage builds include cache scripts is common in our more advanced
102 Profile-Guided Optimizations (PGO) is a really great way to optimize the code
103 clang generates. Our multi-stage PGO builds are a workflow for generating PGO
104 profiles that can be used to optimize clang.
106 At a high level, the way PGO works is that you build an instrumented compiler,
107 then you run the instrumented compiler against sample source files. While the
108 instrumented compiler runs it will output a bunch of files containing
109 performance counters (.profraw files). After generating all the profraw files
110 you use llvm-profdata to merge the files into a single profdata file that you
111 can feed into the LLVM_PROFDATA_FILE option.
113 Our PGO.cmake cache script automates that whole process. You can use it by
116 .. code-block:: console
118 $ cmake -G Ninja -C <path_to_clang>/cmake/caches/PGO.cmake <source dir>
119 $ ninja stage2-instrumented-generate-profdata
121 If you let that run for a few hours or so, it will place a profdata file in your
122 build directory. This takes a really long time because it builds clang twice,
123 and you *must* have compiler-rt in your build tree.
125 This process uses any source files under the perf-training directory as training
126 data as long as the source files are marked up with LIT-style RUN lines.
128 After it finishes you can use “find . -name clang.profdata” to find it, but it
129 should be at a path something like:
131 .. code-block:: console
133 <build dir>/tools/clang/stage2-instrumented-bins/utils/perf-training/clang.profdata
135 You can feed that file into the LLVM_PROFDATA_FILE option when you build your
138 The PGO came cache has a slightly different stage naming scheme than other
139 multi-stage builds. It generates three stages; stage1, stage2-instrumented, and
140 stage2. Both of the stage2 builds are built using the stage1 compiler.
142 The PGO came cache generates the following additional targets:
144 **stage2-instrumented**
145 Builds a stage1 x86 compiler, runtime, and required tools (llvm-config,
146 llvm-profdata) then uses that compiler to build an instrumented stage2 compiler.
148 **stage2-instrumented-generate-profdata**
149 Depends on "stage2-instrumented" and will use the instrumented compiler to
150 generate profdata based on the training files in <clang>/utils/perf-training
153 Depends of "stage2-instrumented-generate-profdata" and will use the stage1
154 compiler with the stage2 profdata to build a PGO-optimized compiler.
156 **stage2-check-llvm**
157 Depends on stage2 and runs check-llvm using the stage2 compiler.
159 **stage2-check-clang**
160 Depends on stage2 and runs check-clang using the stage2 compiler.
163 Depends on stage2 and runs check-all using the stage2 compiler.
165 **stage2-test-suite**
166 Depends on stage2 and runs the test-suite using the stage3 compiler (requires
169 3-Stage Non-Determinism
170 =======================
172 In the ancient lore of compilers non-determinism is like the multi-headed hydra.
173 Whenever its head pops up, terror and chaos ensue.
175 Historically one of the tests to verify that a compiler was deterministic would
176 be a three stage build. The idea of a three stage build is you take your sources
177 and build a compiler (stage1), then use that compiler to rebuild the sources
178 (stage2), then you use that compiler to rebuild the sources a third time
179 (stage3) with an identical configuration to the stage2 build. At the end of
180 this, you have a stage2 and stage3 compiler that should be bit-for-bit
183 You can perform one of these 3-stage builds with LLVM & clang using the
186 .. code-block:: console
188 $ cmake -G Ninja -C <path_to_clang>/cmake/caches/3-stage.cmake <source dir>
191 After the build you can compare the stage2 & stage3 compilers. We have a bot
192 setup `here <http://lab.llvm.org:8011/builders/clang-3stage-ubuntu>`_ that runs
193 this build and compare configuration.