Bootstrapping the LLVM C/C++ Front-End

This document is intended to explain the process of building the LLVM C/C++ front-end from its source code. You have to do this, for example, if you are porting LLVM to a new architecture or operating system, if you are working from Top-Of-Tree CVS/SVN, or if there is no precompiled snapshot available.

NOTE: This is currently a somewhat fragile, error-prone process, and you should only try to do it if:

you really, really, really can't use the binaries we distribute
you are an elite GCC hacker.
you want to use the latest bits from CVS.

We welcome patches to help make this process simpler.

If you are building LLVM and the GCC front-end under Cygwin, please note that the LLVM and GCC makefiles do not correctly handle spaces in paths. To deal with this issue, make sure that your LLVM and GCC source and build trees are located in a top-level directory (like /cygdrive/c/llvm and /cygdrive/c/llvm-cfrontend), not in a directory that contains a space (which includes your "home directory", because it lives under the "Documents and Settings" directory). We welcome patches to fix this issue.

It has been found that the GCC 3.3.3 compiler provided with recent Cygwin versions is incapable of compiling the LLVM GCC front-end correctly. If your Cygwin installation includes GCC 3.3.3, we strongly recommend that you download GCC 3.4.3, build it separately, and use it for compiling the LLVM GCC front-end. This has been shown to work correctly.

Some versions of Cygwin utilize an experimental version of GNU binutils that will cause the GNU ld linker to fail an assertion when linking components of the libstdc++. It is recommended that you replace the entire binutils package with version 2.15 such that "ld --version" responds with

GNU ld version 2.15

not with:

GNU ld version 2.15.91 20040725

If you are building LLVM and the GCC front-end under AIX, do NOT use GNU Binutils. They are not stable under AIX and may produce incorrect and/or invalid code. Instead, use the system assembler and linker.

This section describes how to aquire and build llvm-gcc4, which is based on the GCC 4.0.1 front-end. This front-end supports C, C++, Objective-C, and Objective-C++. Note that the instructions for building this front-end are completely different than those for building llvm-gcc3.

Retrieve the appropriate llvm-gcc4-x.y.source.tar.gz archive from the llvm web site.

It is also possible to download the sources of the llvm-gcc4 front end from a read-only mirror using subversion. To check out the code the first time use:
```
svn co svn://anonsvn.opensource.apple.com/svn/llvm/trunk dst-directory
```
After that, the code can be be updated in the destination directory using:
```
svn update
```
The mirror is brought up to date every evening.
Follow the directions in the top-level README.LLVM file for up-to-date instructions on how to build llvm-gcc4.

Aquire llvm-gcc3 from LLVM CVS or from a release tarball.
Configure and build the LLVM libraries and tools. There are two ways to do this: either with objdir == srcdir or objdir != srcdir. It is recommended that srcdir be the same as objdir for your LLVM tree (but note that you should always use srcdir != objdir for llvm-gcc):
- With objdir != srcdir:
```
% cd objdir
% srcdir/configure --prefix=/some/path/you/can/install/to [options...]
% gmake tools-only
```
- With objdir == srcdir:
```
% cd llvm
% ./configure --prefix=/some/path/you/can/install/to [options...]
% gmake tools-only
```
This will build all of the LLVM tools and libraries. The --prefix option defaults to /usr/local (per configure standards) but unless you are a system administrator, you probably won't be able to install LLVM there because of permissions. Specify a path into which LLVM can be installed (e.g. --prefix=/home/user/llvm).

Add the directory containing the tools to your PATH.

csh:

  % set path = ( `cd llvm/Debug/bin && pwd` $path )

sh:

  % export PATH=`cd llvm/Debug/bin && pwd`:$PATH

Unpack the C/C++ front-end source, either by untar'ing/unzipping a tar.gz file or checking out CVS into this directory.

Make "build" and "install" directories as siblings of the "src" tree:

csh:

  % pwd
  /usr/local/example/llvm-gcc3.4/src
  % cd ..
  % mkdir build install
  % set CFEINSTALL = `pwd`/install

sh:

  % pwd
  /usr/local/example/llvm-gcc3.4/src
  % cd ..
  % mkdir build install
  % export CFEINSTALL=`pwd`/install

Configure, build, and install the GCC front-end:

Linux/x86:
Linux/IA-64:
MacOS X/PowerPC (requires dlcompat library):
AIX/PowerPC:
```
% cd build
% ../src/configure --prefix=$CFEINSTALL --disable-threads --disable-nls \
  --disable-shared --enable-languages=c,c++ --program-prefix=llvm-
% gmake all; gmake install
```
Cygwin/x86:
```
% cd build
% ../src/configure --prefix=$CFEINSTALL --disable-threads --disable-nls \
  --disable-shared --enable-languages=c,c++ --disable-c-mbchar \
  --program-prefix=llvm-
% gmake all; gmake install
```
Solaris/SPARC:

The GCC front-end can be configured for either SPARC V8 (32 bit) or SPARC V9 (64 bit). This changes, among other things, the sizes of integer types and the macros defined for conditional compilation.

The SPARC V8 ABI support is more robust than the V9 ABI support and can generate SPARC V9 code. It is highly recommended that you use the V8 ABI with LLVM, as shown below. Also, note that Solaris has trouble with various wide (multibyte) character functions from C as referenced from C++, so we typically configure with --disable-c-mbchar (cf. Bug 206).
```
% cd build
% ../src/configure --prefix=$CFEINSTALL --disable-threads --disable-nls \
  --disable-shared --enable-languages=c,c++ --host=sparc-sun-solaris2.8 \
  --disable-c-mbchar --program-prefix=llvm-
% gmake all; gmake install
```
Common Problem: You may get error messages regarding the fact that LLVM does not support inline assembly. Here are two common fixes:
- Fix 1: If you have system header files that include inline assembly, you may have to modify them to remove the inline assembly and install the modified versions in $CFEINSTALL/lib/gcc/target-triplet/3.4-llvm/include.
- Fix 2: If you are building the C++ front-end on a CPU we haven't tried yet, you will probably have to edit the appropriate version of atomicity.h under src/libstdc++-v3/config/cpu/name-of-cpu/atomicity.h and apply a patch so that it does not use inline assembly.
Porting to a new architecture: If you are porting the front-end to a new architecture or compiling in a configuration that we have not tried previously, there are probably several changes you will have to make to the GCC target to get it to work correctly. These include:
- Often targets include special assembler or linker flags which llvm-as, opt, or llvm-ld do not understand. In general, these can just be removed.
- LLVM currently does not support any floating point values other than 32-bit and 64-bit IEEE floating point. The primary effect of this is that you may have to map "long double" onto "double".
- The profiling hooks in GCC do not apply at all to the LLVM front-end. These may need to be disabled.
- No inline assembly for position independent code. At the LLVM level, everything is position independent.
- We handle .init and .fini differently.
- You may have to disable multilib support in your target. Using multilib support causes the GCC compiler driver to add a lot of "-L" options to the link line, which do not relate to LLVM. To disable multilibs, delete any MULTILIB_OPTIONS lines from your target files.
- Did we mention that we don't support inline assembly? You'll probably have to add some fixinclude hacks to disable it in the system headers.

Put $CFEINSTALL/bin into your PATH environment variable.

csh:

  % setenv PATH $CFEINSTALL/bin:$PATH

sh:

  % export PATH=$CFEINSTALL/bin:$PATH

Go back into the LLVM source tree proper. Rerun configure, using the same options as the last time. This will cause the configuration to now find the newly built llvm-gcc and llvm-g++ executables.
Rebuild your CVS tree. This shouldn't cause the whole thing to be rebuilt, but it should build the runtime libraries. After the tree is built, install the runtime libraries into your GCC front-end build tree. These are the commands you need:
```
% gmake
% gmake -C runtime install-bytecode
```
Optionally, build a symbol table for the newly installed runtime libraries. Although this step is optional, you are strongly encouraged to do this as the symbol tables will make a significant difference in your link times. Use the llvm-ranlib tool to do this, as follows:
```
% cd $CFEINSTALL/lib
% llvm-ranlib libiberty.a
% llvm-ranlib libstdc++.a
% llvm-ranlib libsupc++.a
% cd $CFEINSTALL/lib/gcc/target-triplet/3.4-llvm
% llvm-ranlib libgcc.a
% llvm-ranlib libgcov.a
```
Test the newly-installed C frontend by one or more of the following means:
- running the feature & regression tests via make check
- compiling and running a "hello, LLVM" program in C and C++.
- running the tests found in the llvm-test CVS module

The LLVM GCC frontend is licensed to you under the GNU General Public License and the GNU Lesser General Public License. Please see the files COPYING and COPYING.LIB for more details.

More information is available in the FAQ.