intel/llvm
1
0
Fork 0
mirror of https://github.com/intel/llvm.git synced 2026-01-22 15:41:35 +08:00
Code Issues Packages Projects Releases Wiki Activity

Add some documentation about cross-compilation to www.

Browse Source 
Reviewers: zturner, vharron, tberghammer, omjavaid

Subscribers: tberghammer, lldb-commits

Differential Revision: http://reviews.llvm.org/D8610

llvm-svn: 234317
This commit is contained in:
Pavel Labath
2015-04-07 14:36:23 +00:00
parent f25949b588
commit 67239b2f73
1 changed files with 172 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

172
lldb/www/build.html
View File

@@ -425,6 +425,178 @@
is built correctly and is available to the default Python interpreter, run:
</p>
<code>&gt; python -c 'import lldb'</code></p>
<h2>Cross-compiling LLDB</h2>
<p>
In order to debug remote targets running different architectures than your host, you
will need to compile LLDB (or at least the server component) for the target. While
the easiest solution is to just compile it locally on the target, this is often not
feasable, and in these cases you will need to cross-compile LLDB on your host.
</p>
<p>
Cross-compilation is often a daunting task and has a lot of quirks which depend on
the exact host and target architectures, so it is not possible to give a universal
guide which will work on all platforms. However, here we try to provide an overview
of the cross-compilation process along with the main things you should look out for.
</p>
<p>
First, you will need a working toolchain which is capable of producing binaries for
the target architecture. Since you already have a checkout of clang and lldb, you
can compile a host version of clang in a separate folder and use that.
Alternatively you can use system clang or even cross-gcc if your distribution
provides such packages (e.g., <code>g++-aarch64-linux-gnu</code> on Ubuntu). On
Android, a working toolchain can be produced by downloading the Android NDK and
running the contained <code>make-standalone-toolchain.sh</code> script.
</p>
<p>
Next, you will need a copy of the required target headers and libraries on your
host. The libraries can be usually obtained by copying from the target machine,
however the headers are often not found there, especially in case of embedded
platforms. In this case, you will need to obtain them from another source, either
a cross-package if one is available, or cross-compiling the respective library from
source. Fortunately the list of LLDB dependencies is not big and if you are only
interested in the server component, you can reduce this even further by passing the
appropriate cmake options, such as:
</p>
<code>
-DLLDB_DISABLE_LIBEDIT=1<br/>
-DLLDB_DISABLE_CURSES=1<br/>
-DLLDB_DISABLE_PYTHON=1<br/>
-DLLVM_ENABLE_TERMINFO=0
</code>
<p>
In this case you, will often not need anything other than the standard C and C++
libraries.
</p>
<p>
In the case of Android, all required headers and libraries are provided by the
aforementioned <code>make-standalone-toolchain.sh</code> script.
</p>
<p>
Once all of the dependencies are in place, it's just a matter of configuring the
build system with the locations and arguments of all the necessary tools. The most
important cmake options here are:
</p>
<dl>
<dt>CMAKE_CROSSCOMPILING</dt>
<dd>Set to 1 to enable cross-compilation.</dd>
<dt>CMAKE_LIBRARY_ARCHITECTURE</dt>
<dd>Affects the cmake search path when looking for libraries. You may need to set
this to your architecture triple if you do not specify all your include and
library paths explicitly.</dd>
<dt>CMAKE_C_COMPILER, CMAKE_CXX_COMPILER</dt>
<dd>C and C++ compilers for the target architecture</dd>
<dt>CMAKE_C_FLAGS, CMAKE_CXX_FLAGS</dt>
<dd>The flags for the C and C++ target compilers. You may need to specify the
exact target cpu and abi besides the include paths for the target headers.</dd>
<dt>CMAKE_EXE_LINKER_FLAGS</dt>
<dd>The flags to be passed to the linker. Usually just a list of library search
paths referencing the target libraries.</dd>
<dt>LLVM_TABLEGEN, CLANG_TABLEGEN</dt>
<dd>Paths to llvm-tblgen and clang-tblgen for the <em>host</em> architecture. If
you already have built clang for the host, you can point these variables to the
executables in your build directory. If not, you will need to build the
llvm-tblgen and clang-tblgen host targets at least.<dd>
<dt>LLVM_HOST_TRIPLE</dt>
<dd>The triple of the system that lldb (or lldb-server) will run on. Not setting
this (or setting it incorrectly) can cause a lot of issues with remote debugging
as a lot of the choices lldb makes depend on the triple reported by the remote
platform.</dd>
</dl>
<p>
You can of course also specify the usual cmake options like CMAKE_BUILD_TYPE, etc.
</p>
<h3>Example 1: Cross-compiling for linux arm64 on Ubuntu host</h3>
<p>
Ubuntu already provides the packages necessary to cross-compile LLDB for arm64. It
is sufficient to install pacakges gcc-aarch64-linux-gnu, g++-aarch64-linux-gnu,
binutils-aarch64-linux-gnu. Then it is possible to prepare the cmake build with the
following parameters:
</p>
<code>
-DCMAKE_CROSSCOMPILING=1 \<br/>
-DCMAKE_C_COMPILER=aarch64-linux-gnu-gcc \<br/>
-DCMAKE_CXX_COMPILER=aarch64-linux-gnu-g++ \<br/>
-DLLVM_HOST_TRIPLE=aarch64-unknown-linux-gnu \<br/>
-DLLVM_TABLEGEN=&lt;path-to-host&gt;/bin/llvm-tblgen \<br/>
-DCLANG_TABLEGEN=&lt;path-to-host&gt;/bin/clang-tblgen \<br/>
-DLLDB_DISABLE_PYTHON=1 \<br/>
-DLLDB_DISABLE_LIBEDIT=1 \<br/>
-DLLDB_DISABLE_CURSES=1
</code>
<p>
An alternative (and recommended) way to compile LLDB is with clang. Unfortunately,
clang is not able to find all the include paths necessary for a successful
cross-compile, so we need to help it with a couple of CFLAGS options. In my case it
was sufficient to add the following arguments to CMAKE_C_FLAGS and CMAKE_CXX_FLAGS
(in addition to changing CMAKE_C(XX)_COMPILER to point to clang compilers):
</p>
<code>
-target aarch64-linux-gnu \<br/>
-I /usr/aarch64-linux-gnu/include/c++/4.8.2/aarch64-linux-gnu \<br/>
-I /usr/aarch64-linux-gnu/include
</code>
<p>
If you wanted to build a full version of LLDB and avoid passing
-DLLDB_DISABLE_PYTHON and other options, you would need to obtain the target
versions of the respective libraries. The easiest way to achive this is to use the
<code>qemu-debootstrap</code> utility, which can prepare a system image using qemu
and chroot to simulate the target environment. Then you can install the necessary
packages in this environment (python-dev, libedit-dev, etc.) and point your
compiler to use them using the correct -I and -L arguments.
</p>
<h3>Example 2: Cross-compiling for Android on Linux</h3>
<p>
All tools needed to build LLDB for android are available in the Android NDK. For
example, we can produce an x86 toolchain along with all the libraries and headers
by running
</p>
<code>
./build/tools/make-standalone-toolchain.sh \<br/>
--platform=android-21 \<br/>
--toolchain=x86-4.9 \<br/>
--install-dir=$HOME/Toolchains/x86-android-toolchain
</code>
<p>
from inside the unzipped NDK. Toolchains for other architectures can be produced in
a similar manner.
</p>
<p>
For Android we provide a Android.cmake script which sets a lot of the required
options automatically. A cmake build can therefore be prepared with the following parameters:
</p>
<code>
-DCMAKE_TOOLCHAIN_FILE=cmake/platforms/Android.cmake \<br/>
-DANDROID_TOOLCHAIN_DIR=$HOME/Toolchains/x86-android-toolchain \<br/>
-DANDROID_ABI=x86 \<br/>
-DLLVM_HOST_TRIPLE=i386-unknown-linux-android \<br/>
-DLLVM_TABLEGEN=&lt;path-to-host&gt;/bin/llvm-tblgen \<br/>
-DCLANG_TABLEGEN=&lt;path-to-host&gt;/bin/clang-tblgen
</code>
<p>
Note that the full LLVM build is not functional on android yet, so simply runing
<code>ninja</code> will not work. You will need to manually specify the target you
want to build: <code>lldb</code>, <code>lldb-server</code>, etc.
</p>
</div>
<div class="postfooter"></div>
</div>