<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01//EN" "http://www.w3.org/TR/html4/strict.dtd">
<html>
<head>
<meta http-equiv="Content-Type" content="text/html; charset=utf-8">
<title>The LLVM Compiler Driver (llvmc)</title>
<link rel="stylesheet" href="llvm.css" type="text/css">
<style type="text/css">
TR, TD { border: 2px solid gray; padding: 4pt 4pt 2pt 2pt; }
TH { border: 2px solid gray; font-weight: bold; font-size: 105%; }
TABLE { text-align: center; border: 2px solid black;
border-collapse: collapse; margin-top: 1em; margin-left: 1em;
margin-right: 1em; margin-bottom: 1em; }
.td_left { border: 2px solid gray; text-align: left; }
</style>
<meta name="author" content="Reid Spencer" name="author">
<meta name="description"
content="A description of the use and design of the LLVM Compiler Driver.">
</head>
<body>
<div class="doc_title">The LLVM Compiler Driver (llvmc)</div>
<p class="doc_warning">NOTE: This document is a work in progress!</p>
<ol>
<li><a href="#abstract">Abstract</a></li>
<li><a href="#introduction">Introduction</a>
<ol>
<li><a href="#purpose">Purpose</a></li>
<li><a href="#operation">Operation</a></li>
<li><a href="#phases">Phases</a></li>
<li><a href="#actions">Actions</a></li>
</ol>
</li>
<li><a href="#details">Details</a>
<li><a href="#configuration">Configuration</a>
<li><a href="#glossary">Glossary</a>
</ol>
<div class="doc_author">
<p>Written by <a href="mailto:rspencer@x10sys.com">Reid Spencer</a>
</p>
</div>
<!-- *********************************************************************** -->
<div class="doc_section"> <a name="abstract">Abstract</a></div>
<!-- *********************************************************************** -->
<div class="doc_text">
<p>This document describes the requirements, design, and configuration of the
LLVM compiler driver, <tt>llvmc</tt>. The compiler driver knows about LLVM's
tool set and can be configured to know about a variety of compilers for
source languages. It uses this knowledge to execute the tools necessary
to accomplish general compilation, optimization, and linking tasks. The main
purpose of <tt>llvmc</tt> is to provide a simple and consistent interface to
all compilation tasks. This reduces the burden on the end user who can just
learn to use <tt>llvmc</tt> instead of the entire LLVM tool set and all the
source language compilers compatible with LLVM.</p>
</div>
<!-- *********************************************************************** -->
<div class="doc_section"> <a name="introduction">Introduction</a></div>
<!-- *********************************************************************** -->
<div class="doc_text">
<p>The <tt>llvmc</tt> <a href="def_tool">tool</a> is a configurable compiler
<a href="def_driver">driver</a>. As such, it isn't the compiler, optimizer,
or linker itself but it drives (invokes) other software that perform those
tasks. If you are familiar with the GNU Compiler Collection's <tt>gcc</tt>
tool, <tt>llvmc</tt> is very similar.</p>
<p>The following introductory sections will help you understand why this tool
is necessary and what it does.</p>
</div>
<!-- _______________________________________________________________________ -->
<div class="doc_subsection"><a name="purpose">Purpose</a></div>
<div class="doc_text">
<p><tt>llvmc</tt> was invented to make compilation with LLVM based compilers
easier. To accomplish this, <tt>llvmc</tt> strives to:</p>
<ul>
<li>Be the single point of access to most of the LLVM tool set.</li>
<li>Hide the complexities of the LLVM tools through a single interface.</li>
<li>Provide a consistent interface for compiling all languages.</li>
</ul>
<p>Additionally, <tt>llvmc</tt> makes it easier to write a compiler for use
with LLVM, because it:</p>
<ul>
<li>Makes integration of existing non-LLVM tools simple.</li>
<li>Extends the capabilities of minimal front ends by optimizing their
output.</li>
<li>Reduces the number of interfaces a compiler writer must know about
before a working compiler can be completed (essentially only the VMCore
interfaces need to be understood).</li>
<li>Supports source language translator invocation via both dynamically
loadable shared objects and invocation of an executable.</li>
</ol>
</p>
</div>
<!-- _______________________________________________________________________ -->
<div class="doc_subsection"><a name="operation">Operation</a></div>
<div class="doc_text">
<p>At a high level, <tt>llvmc</tt> operation is very simple. The basic action
taken by <tt>llvmc</tt> is to simply invoke some tool or set of tools to fill
the user's request for compilation. Every execution of <tt>llvmc</tt>takes the
following sequence of steps:<br/>
<dl>
<dt><b>Collect Command Line Options</b></dt>
<dd>The command line options provide the marching orders to <tt>llvmc</tt>
on what actions it should perform. This is the request the user is making
of <tt>llvmc</tt> and it is interpreted first. See the <tt>llvmc</tt>
<a href="CommandGuide/html/llvmc.html">manual page</a> for details on the
options.</dd>
<dt><b>Read Configuration Files</b></dt>
<dd>Based on the options and the suffixes of the filenames presented, a set
of configuration files are read to configure the actions <tt>llvmc</tt> will
take. Configuration files are provided by either LLVM or the front end
compiler tools that B<llvmc> invokes. These files determine what actions
<tt>llvmc</tt> will take in response to the user's request. See the section
on <a href="#configuration">configuration</a> for more details.</dd>
<dt><b>Determine Phases To Execute</b></dt>
<dd>Based on the command line options and configuration files,
<tt>llvmc</tt> determines the compilation <a href="#phases">phases</a> that
must be executed by the user's request. This is the primary work of
<tt>llvmc</tt>.</dd>
<dt><b>Determine Actions To Execute</b></dt>
<dd>Each <a href="#phases">phase</a> to be executed can result in the
invocation of one or more <a href="#actions">actions</a>. An action is
either a whole program or a function in a dynamically linked shared library.
In this step, <tt>llvmc</tt> determines the sequence of actions that must be
executed. Actions will always be executed in a deterministic order.</dd>
<dt><b>Execute Actions</b></dt>
<dd>The <a href="#actions">actions</a> necessary to support the user's
original request are executed sequentially and deterministically. All
actions result in either the invocation of a whole program to perform the
action or the loading of a dynamically linkable shared library and invocation
of a standard interface function within that library.</dd>
<dt><b>Termination</b></dt>
<dd>If any action fails (returns a non-zero result code), <tt>llvmc</tt>
also fails and returns the result code from the failing action. If
everything succeeds, <tt>llvmc</tt> will return a zero result code.</dd>
</dl></p>
<p><tt>llvmc</tt>'s operation must be simple, regular and predictable.
Developers need to be able to rely on it to take a consistent approach to
compilation. For example, the invocation:</p>
<tt><pre>
llvmc -O2 x.c y.c z.c -o xyz</pre></tt>
<p>must produce <i>exactly</i> the same results as:</p>
<tt><pre>
llvmc -O2 x.c
llvmc -O2 y.c
llvmc -O2 z.c
llvmc -O2 x.o y.o z.o -o xyz</pre></tt>
<p>To accomplish this, <tt>llvmc</tt> uses a very simple goal oriented
procedure to do its work. The overall goal is to produce a functioning
executable. To accomplish this, <tt>llvmc</tt> always attempts to execute a
series of compilation <a href="#def_phase">phases</a> in the same sequence.
However, the user's options to <tt>llvmc</tt> can cause the sequence of phases
to start in the middle or finish early.</p>
</div>
<!-- _______________________________________________________________________ -->
<div class="doc_subsection"><a name="phases"></a>Phases </div>
<div class="doc_text">
<p><tt>llvmc</tt> breaks every compilation task into the following five
distinct phases:</p>
<dl><dt><b>Preprocessing</b></dt><dd>Not all languages support preprocessing;
but for those that do, this phase can be invoked. This phase is for
languages that provide combining, filtering, or otherwise altering with the
source language input before the translator parses it. Although C and C++
are the most common users of this phase, other languages may provide their
own preprocessor (whether its the C pre-processor or not).</dd>
</dl>
<dl><dt><b>Translation</b></dt><dd>The translation phase converts the source
language input into something that LLVM can interpret and use for
downstream phases. The translation is essentially from "non-LLVM form" to
"LLVM form".</dd>
</dl>
<dl><dt><b>Optimization</b></dt><dd>Once an LLVM Module has been obtained from
the translation phase, the program enters the optimization phase. This phase
attempts to optimize all of the input provided on the command line according
to the options provided.</dd>
</dl>
<dl><dt><b>Linking</b></dt><dd>The inputs are combined to form a complete
program.</dd>
</dl>
<p>The following table shows the inputs, outputs, and command line options
applicabe to each phase.</p>
<table>
<tr>
<th style="width: 10%">Phase</th>
<th style="width: 25%">Inputs</th>
<th style="width: 25%">Outputs</th>
<th style="width: 40%">Options</th>
</tr>
<tr><td><b>Preprocessing</b></td>
<td class="td_left"><ul><li>Source Language File</li></ul></td>
<td class="td_left"><ul><li>Source Language File</li></ul></td>
<td class="td_left"><dl>
<dt><tt>-E</tt></dt>
<dd>Stops the compilation after preprocessing</dd>
</dl></td>
</tr>
<tr>
<td><b>Translation</b></td>
<td class="td_left"><ul>
<li>Source Language File</li>
</ul></td>
<td class="td_left"><ul>
<li>LLVM Assembly</li>
<li>LLVM Bytecode</li>
<li>LLVM C++ IR</li>
</ul></td>
<td class="td_left"><dl>
<dt><tt>-c</tt></dt>
<dd>Stops the compilation after translation so that optimization and
linking are not done.</dd>
<dt><tt>-S</tt></dt>
<dd>Stops the compilation before object code is written so that only
assembly code remains.</dd>
</dl></td>
</tr>
<tr>
<td><b>Optimization</b></td>
<td class="td_left"><ul>
<li>LLVM Assembly</li>
<li>LLVM Bytecode</li>
</ul></td>
<td class="td_left"><ul>
<li>LLVM Bytecode</li>
</ul></td>
<td class="td_left"><dl>
<dt><tt>-Ox</tt>
<dd>This group of options affects the amount of optimization
performed.</dd>
</dl></td>
</tr>
<tr>
<td><b>Linking</b></td>
<td class="td_left"><ul>
<li>LLVM Bytecode</li>
<li>Native Object Code</li>
<li>LLVM Library</li>
<li>Native Library</li>
</ul></td>
<td class="td_left"><ul>
<li>LLVM Bytecode Executable</li>
<li>Native Executable</li>
</ul></td>
<td class="td_left"><dl>
<dt><tt>-L</tt></dt><dd>Specifies a path for library search.</dd>
<dt><tt>-l</tt></dt><dd>Specifies a library to link in.</dd>
</dl></td>
</tr>
</table>
</div>
<!-- _______________________________________________________________________ -->
<div class="doc_subsection"><a name="actions"></a>Actions</div>
<div class="doc_text">
<p>An action, with regard to <tt>llvmc</tt> is a basic operation that it takes
in order to fulfill the user's request. Each phase of compilation will invoke
zero or more actions in order to accomplish that phase.</p>
<p>Actions come in two forms:<ol>
<li>Invokable Executables</li>
<li>Functions in a shared library</li>
</ul></p>
</div>
<!-- *********************************************************************** -->
<div class="doc_section"><a name="details">Details</a></div>
<!-- *********************************************************************** -->
<div class="doc_text">
</div>
<!-- *********************************************************************** -->
<div class="doc_section"><a name="configuration">Configuration</a></div>
<!-- *********************************************************************** -->
<div class="doc_text">
<p>This section of the document describes the configuration files used by
<tt>llvmc</tt>. Configuration information is relatively static for a
given release of LLVM and a front end compiler. However, the details may
change from release to release of either. Users are encouraged to simply use
the various options of the B<llvmc> command and ignore the configuration of
the tool. These configuration files are for compiler writers and LLVM
developers. Those wishing to simply use B<llvmc> don't need to understand
this section but it may be instructive on how the tool works.</p>
</div>
<!-- _______________________________________________________________________ -->
<div class="doc_subsection"><a name="overview"></a>Overview</div>
<div class="doc_text">
<p><tt>llvmc</tt> is highly configurable both on the command line and in
configuration files. The options it understands are generic, consistent and
simple by design. Furthermore, the <tt>llvmc</tt> options apply to the
compilation of any LLVM enabled programming language. To be enabled as a
supported source language compiler, a compiler writer must provide a
configuration file that tells <tt>llvmc</tt> how to invoke the compiler
and what its capabilities are. The purpose of the configuration files then
is to allow compiler writers to specify to <tt>llvmc</tt> how the compiler
should be invoked. Users may but are not advised to alter the compiler's
<tt>llvmc</tt> configuration.</p>
<p>Because <tt>llvmc</tt> just invokes other programs, it must deal with the
available command line options for those programs regardless of whether they
were written for LLVM or not. Furthermore, not all compilation front ends will
have the same capabilities. Some front ends will simply generate LLVM assembly
code, others will be able to generate fully optimized byte code. In general,
<tt>llvmc</tt> doesn't make any assumptions about the capabilities or command
line options of a sub-tool. It simply uses the details found in the configuration
files and leaves it to the compiler writer to specify the configuration
correctly.</p>
<p>This approach means that new compiler front ends can be up and working very
quickly. As a first cut, a front end can simply compile its source to raw
(unoptimized) bytecode or LLVM assembly and <tt>llvmc</tt> can be configured
to pick up the slack (translate LLVM assembly to bytecode, optimize the
bytecode, generate native assembly, link, etc.). In fact, the front end need
not use any LLVM libraries, and it could be written in any language (instead of
C++). The configuration data will allow the full range of optimization,
assembly, and linking capabilities that LLVM provides to be added to these kinds
of tools. Enabling the rapid development of front-ends is one of the primary
goals of <tt>llvmc</tt>.</p>
<p>As a compiler front end matures, it may utilize the LLVM libraries and tools
to more efficiently produce optimized bytecode directly in a single compilation
and optimization program. In these cases, multiple tools would not be needed
and the configuration data for the compiler would change.</p>
<p>Configuring <tt>llvmc</tt> to the needs and capabilities of a source language
compiler is relatively straight forward. A compiler writer must provide a
definition of what to do for each of the five compilation phases for each of
the optimization levels. The specification consists simply of prototypical
command lines into which <tt>llvmc</tt> can substitute command line
arguments and file names. Note that any given phase can be completely blank if
the source language's compiler combines multiple phases into a single program.
For example, quite often pre-processing, translation, and optimization are
combined into a single program. The specification for such a compiler would have
blank entries for pre-processing and translation but a full command line for
optimization.</p>
</div>
<!-- _______________________________________________________________________ -->
<div class="doc_subsection"><a name="filetypes"></a>Configuration Files</div>
<div class="doc_text">
<h3>File Types</h3>
<p>There are two types of configuration files: the master configuration file
and the language specific configuration file. The master configuration file
contains the general configuration of <tt>llvmc</tt> itself and is supplied
with the tool. It contains information that is source language agnostic.
Language specific configuration files tell <tt>llvmc</tt> how to invoke the
language's compiler for a variety of different tasks and what other tools
are needed to backfill the compiler's missing features (e.g.
optimization).</p>
<h3>Directory Search</h3>
<p><tt>llvmc</tt> always looks for files of a specific name. It uses the
first file with the name its looking for by searching directories in the
following order:<br/>
<ol>
<li>Any directory specified by the <tt>--config-dir</tt> option will be
checked first.</li>
<li>If the environment variable LLVM_CONFIG_DIR is set, and it contains
the name of a valid directory, that directory will be searched next.</li>
<li>If the user's home directory (typically <tt>/home/user</tt> contains
a sub-directory named <tt>.llvm</tt> and that directory contains a
sub-directory named <tt>etc</tt> then that directory will be tried
next.</li>
<li>If the LLVM installation directory (typically <tt>/usr/local/llvm</tt>
contains a sub-directory named <tt>etc</tt> then that directory will be
tried last.</li>
<li>If the configuration file sought still can't be found, <tt>llvmc</tt>
will print an error message and exit.</li>
</ol>
The first file found in this search will be used. Other files with the same
name will be ignored even if they exist in one of the subsequent search
locations.</p>
<h3>File Names</h3>
<p>In the directories searched, a file named <tt>master</tt> will be
recognized as the master configuration file for <tt>llvmc</tt>. Note that
users <i>may</i> override the master file with a copy in their home directory
but they are advised not to. This capability is only useful for compiler
implementers needing to alter the master configuration while developing
their compiler front end. When reading the configuration files, the master
files are always read first.</p>
<p>Language specific configuration files are given specific names to foster
faster lookup. The name of a given language specific configuration file is
the same as the suffix used to identify files containing source in that
language. For example, a configuration file for C++ source might be named
<tt>cpp</tt>, <tt>C</tt>, or <tt>cxx</tt>.</p>
<h3>What Gets Read</h3>
<p>The master configuration file is always read. Which language specific
configuration files are read depends on the command line options and the
suffixes of the file names provided on <tt>llvmc</tt>'s command line. Note
that the <tt>--x LANGUAGE</tt> option alters the language that <tt>llvmc</tt>
uses for the subsequent files on the command line. Only the language
specific configuration files actually needed to complete <tt>llvmc</tt>'s
task are read. Other language specific files will be ignored.</p>
</div>
<!-- _______________________________________________________________________ -->
<div class="doc_subsection"><a name="syntax"></a>Syntax</div>
<div class="doc_text">
<p>The syntax of the configuration files is yet to be determined. There are
two viable options remaining:<br/>
<ul>
<li>XML DTD Specific To <tt>llvmc</tt></li>
<li>Windows .ini style file with numerous sections</li>
</ul></p>
</div>
<!-- _______________________________________________________________________ -->
<div class="doc_subsection"><a name="master_items">Configuration Items</a></div>
<div class="doc_text">
<p>The following description of configuration items is syntax-less and simply
uses a naming hierarchy to describe the configuration items. Whatever
syntax is chosen will need to map the hierarchy to the given syntax.</p>
<table>
<tr>
<th>Name</th>
<th>Value Type</th>
<th>Description</th>
</tr>
<tr>
<td><b>Capabilities.hasPreProcessor</b></td>
<td>boolean</td>
<td class="td_left">This item specifies whether the language has a
pre-processing phase or not. This controls whether the B<-E> option works
for the language or not.</td>
</tr>
<tr>
<td><b>Capabilities.outputFormat</b></td>
<td>"bc" or "ll"</td>
<td class="td_left">This item specifies the kind of output the language's
compiler generates. The choices are either bytecode (<tt>bc</tt>) or LLVM
assembly (<tt>ll</tt>).</td>
</tr>
<tr>
<td><b>Capabilities.understandsOptimization</b></td>
<td>boolean</td>
<td>Indicates whether the compiler for this language understands the
<tt>-O</tt> options or not</td>
</tr>
</table>
</div>
<!-- *********************************************************************** -->
<div class="doc_section"><a name="glossary">Glossary</a></div>
<!-- *********************************************************************** -->
<div class="doc_text">
<p>This document uses precise terms in reference to the various artifacts and
concepts related to compilation. The terms used throughout this document are
defined below.</p>
<dl>
<dt><a name="def_assembly"><b>assembly</b></a></dt>
<dd>A compilation <a href="#def_phase">phase</a> in which LLVM bytecode or
LLVM assembly code is assembled to a native code format (either target
specific aseembly language or the platform's native object file format).
</dd>
<dt><a name="def_compiler"><b>compiler</b></a></dt>
<dd>Refers to any program that can be invoked by <tt>llvmc</tt> to accomplish
the work of one or more compilation <a href="#def_phase">phases</a>.</dd>
<dt><a name="def_driver"><b>driver</b></a></dt>
<dd>Refers to <tt>llvmc</tt> itself.</dd>
<dt><a name="def_linking"><b>linking</b></a></dt>
<dd>A compilation <a href="#def_phase">phase</a> in which LLVM bytecode files
and (optionally) native system libraries are combined to form a complete
executable program.</dd>
<dt><a name="def_optimization"><b>optimization</b></a></dt>
<dd>A compilation <a href="#def_phase">phase</a> in which LLVM bytecode is
optimized.</dd>
<dt><a name="def_phase"><b>phase</b></a></dt>
<dd>Refers to any one of the five compilation phases that that
<tt>llvmc</tt> supports. The five phases are:
<a href="#def_preprocessing">preprocessing</a>,
<a href="#def_translation">translation</a>,
<a href="#def_optimization">optimization</a>,
<a href="#def_assembly">assembly</a>,
<a href="#def_linking">linking</a>.</dd>
<dt><a name="def_sourcelanguage"><b>source language</b></a></dt>
<dd>Any common programming language (e.g. C, C++, Java, Stacker, ML,
FORTRAN). These languages are distinguished from any of the lower level
languages (such as LLVM or native assembly), by the fact that a
<a href="#def_translation">translation</a> <a href="#def_phase">phase</a>
is required before LLVM can be applied.</dd>
<dt><a name="def_tool"><b>tool</b></a></dt>
<dd>Refers to any program in the LLVM tool set.</dd>
<dt><a name="def_translation"><b>translation</b></a></dt>
<dd>A compilation <a href="#def_phase">phase</a> in which
<a href="#def_sourcelanguage">source language</a> code is translated into
either LLVM assembly language or LLVM bytecode.</dd>
</dl>
</div>
<!-- *********************************************************************** -->
<hr>
<address> <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
src="http://jigsaw.w3.org/css-validator/images/vcss" alt="Valid CSS!"></a><a
href="http://validator.w3.org/check/referer"><img
src="http://www.w3.org/Icons/valid-html401" alt="Valid HTML 4.01!"></a><a
href="mailto:rspencer@x10sys.com">Reid Spencer</a><br>
<a href="http://llvm.cs.uiuc.edu">The LLVM Compiler Infrastructure</a><br>
Last modified: $Date$
</address>
<!-- vim: sw=2
-->
</body>
</html>