blob: cc1414bb6073eadee647af108ae738f49f664809 [file] [log] [blame]
<meta http-equiv="Content-Type" content="text/html; charset=utf-8">
<link rel="stylesheet" href="llvm.css" type="text/css">
<title>LLVM 3.0 Release Notes</title>
<h1>LLVM 3.0 Release Notes</h1>
<img align=right src=""
width="136" height="136" alt="LLVM Dragon Logo">
<li><a href="#intro">Introduction</a></li>
<li><a href="#subproj">Sub-project Status Update</a></li>
<li><a href="#externalproj">External Projects Using LLVM 3.0</a></li>
<li><a href="#whatsnew">What's New in LLVM 3.0?</a></li>
<li><a href="GettingStarted.html">Installation Instructions</a></li>
<li><a href="#knownproblems">Known Problems</a></li>
<li><a href="#additionalinfo">Additional Information</a></li>
<div class="doc_author">
<p>Written by the <a href="">LLVM Team</a></p>
<h1 style="color:red">These are in-progress notes for the upcoming LLVM 3.0
You may prefer the
<a href="">LLVM 2.9
Release Notes</a>.</h1>
<!-- *********************************************************************** -->
<a name="intro">Introduction</a>
<!-- *********************************************************************** -->
<p>This document contains the release notes for the LLVM Compiler
Infrastructure, release 3.0. Here we describe the status of LLVM, including
major improvements from the previous release, improvements in various
subprojects of LLVM, and some of the current users of the code.
All LLVM releases may be downloaded from
the <a href="">LLVM releases web site</a>.</p>
<p>For more information about LLVM, including information about the latest
release, please check out the <a href="">main LLVM web
site</a>. If you have questions or comments,
the <a href="">LLVM
Developer's Mailing List</a> is a good place to send them.</p>
<p>Note that if you are reading this file from a Subversion checkout or the main
LLVM web page, this document applies to the <i>next</i> release, not the
current one. To see the release notes for a specific release, please see the
<a href="">releases page</a>.</p>
<!-- *********************************************************************** -->
<a name="subproj">Sub-project Status Update</a>
<!-- *********************************************************************** -->
<p>The LLVM 3.0 distribution currently consists of code from the core LLVM
repository (which roughly includes the LLVM optimizers, code generators and
supporting tools), and the Clang repository. In
addition to this code, the LLVM Project includes other sub-projects that are
in development. Here we include updates on these subprojects.</p>
<a name="clang">Clang: C/C++/Objective-C Frontend Toolkit</a>
<p><a href="">Clang</a> is an LLVM front end for the C,
C++, and Objective-C languages. Clang aims to provide a better user
experience through expressive diagnostics, a high level of conformance to
language standards, fast compilation, and low memory use. Like LLVM, Clang
provides a modular, library-based architecture that makes it suitable for
creating or integrating with other development tools. Clang is considered a
production-quality compiler for C, Objective-C, C++ and Objective-C++ on x86
(32- and 64-bit), and for Darwin/ARM targets.</p>
<p>In the LLVM 3.0 time-frame, the Clang team has made many improvements:
<li>Greatly improved support for building C++ applications, with greater
stability and better diagnostics.</li>
<li><a href="">Improved support</a> for
the <a href="">C++
2011</a> standard (aka "C++'0x"), including implementations of non-static data member
initializers, alias templates, delegating constructors, range-based
for loops, and implicitly-generated move constructors and move assignment
operators, among others.</li>
<li>Implemented support for some features of the upcoming C1x standard,
including static assertions and generic selections.</li>
<li>Better detection of include and linking paths for system headers and
libraries, especially for Linux distributions.</li>
<li>Several improvements to Objective-C support, including:
<li><a href="">
Automatic Reference Counting</a> (ARC) and an improved memory model
cleanly separating object and C memory.</li>
<li>A migration tool for moving manual retain/release code to ARC</li>
<li>Better support for data hiding, allowing instance variables to be
declared in implementation contexts or class extensions</li>
<li>Weak linking support for Objective-C classes</li>
<li>Improved static type checking by inferring the return type of methods
such as +alloc and -init.</li>
Some new Objective-C features require either the Mac OS X 10.7 / iOS 5
Objective-C runtime, or version 1.6 or later of the GNUstep Objective-C
runtime version.</li>
<li>Implemented a number of optimizations in <tt>libclang</tt>, the Clang C
interface, to improve the performance of code completion and the mapping
from source locations to abstract syntax tree nodes.</li>
For more details about the changes to Clang since the 2.9 release, see the
<a href="">Clang release notes</a>
<p>If Clang rejects your code but another compiler accepts it, please take a
look at the <a href="">language
compatibility</a> guide to make sure this is not intentional or a known
<a name="dragonegg">DragonEgg: GCC front-ends, LLVM back-end</a>
<p><a href="">DragonEgg</a> is a
<a href="">gcc plugin</a> that replaces GCC's
optimizers and code generators with LLVM's. It works with gcc-4.5 or gcc-4.6,
targets the x86-32 and x86-64 processor families, and has been successfully
used on the Darwin, FreeBSD, KFreeBSD, Linux and OpenBSD platforms. It fully
supports Ada, C, C++ and Fortran. It has partial support for Go, Java, Obj-C
and Obj-C++.</p>
<p>The 3.0 release has the following notable changes:</p>
<li>GCC version 4.6 is now fully supported.</li>
<li>Patching and building GCC is no longer required: the plugin should work
with your system GCC (version 4.5 or 4.6; on Debian/Ubuntu systems the
gcc-4.5-plugin-dev or gcc-4.6-plugin-dev package is also needed).</li>
<li>The <tt>-fplugin-arg-dragonegg-enable-gcc-optzns</tt> option, which runs
GCC's optimizers as well as LLVM's, now works much better. This is the
option to use if you want ultimate performance! It is still experimental
though: it may cause the plugin to crash. Setting the optimization level
to <tt>-O4</tt> when using this option will optimize even harder, though
this usually doesn't result in any improvement over <tt>-O3<tt>.</li>
<li>The type and constant conversion logic has been almost entirely rewritten,
fixing a multitude of obscure bugs.</li>
<a name="compiler-rt">compiler-rt: Compiler Runtime Library</a>
<p>The new LLVM <a href="">compiler-rt project</a>
is a simple library that provides an implementation of the low-level
target-specific hooks required by code generation and other runtime
components. For example, when compiling for a 32-bit target, converting a
double to a 64-bit unsigned integer is compiled into a runtime call to the
"__fixunsdfdi" function. The compiler-rt library provides highly optimized
implementations of this and other low-level routines (some are 3x faster than
the equivalent libgcc routines).</p>
<p>In the LLVM 3.0 timeframe, the target specific ARM code has converted to
"unified" assembly syntax, and several new functions have been added to the
<a name="lldb">LLDB: Low Level Debugger</a>
<p>LLDB is a ground-up implementation of a command line debugger, as well as a
debugger API that can be used from other applications. LLDB makes use of the
Clang parser to provide high-fidelity expression parsing (particularly for
C++) and uses the LLVM JIT for target support.</p>
<p>LLDB has advanced by leaps and bounds in the 3.0 timeframe. It is
dramatically more stable and useful, and includes both a
new <a href="">tutorial</a> and
a <a href="">side-by-side comparison with
<a name="libc++">libc++: C++ Standard Library</a>
<p>Like compiler_rt, libc++ is now <a href="DeveloperPolicy.html#license">dual
licensed</a> under the MIT and UIUC license, allowing it to be used more
<p>Libc++ has been ported to FreeBSD and imported into the base system. It is
planned to be the default STL implementation for FreeBSD 10.</p>
<a name="vmkit">VMKit</a>
<p>The <a href="">VMKit project</a> is an
implementation of a Java Virtual Machine (Java VM or JVM) that uses LLVM for
static and just-in-time compilation.
<p>In the LLVM 3.0 time-frame, VMKit has had significant improvements on both
runtime and startup performance:</p>
<li>Precompilation: by compiling ahead of time a small subset of Java's core
library, the startup performance have been highly optimized to the point that
running a 'Hello World' program takes less than 30 milliseconds.</li>
<li>Customization: by customizing virtual methods for individual classes,
the VM can statically determine the target of a virtual call, and decide to
inline it.</li>
<li>Inlining: the VM does more inlining than it did before, by allowing more
bytecode instructions to be inlined, and thanks to customization. It also
inlines GC barriers, and object allocations.</li>
<li>New exception model: the generated code for a method that does not do
any try/catch is not penalized anymore by the eventuality of calling a
method that throws an exception. Instead, the method that throws the
exception jumps directly to the method that could catch it.</li>
<a name="LLBrowse">LLBrowse: IR Browser</a>
<p><a href="">
LLBrowse</a> is an interactive viewer for LLVM modules. It can load any LLVM
module and displays its contents as an expandable tree view, facilitating an
easy way to inspect types, functions, global variables, or metadata nodes. It
is fully cross-platform, being based on the popular wxWidgets GUI
<a name="klee">KLEE: A Symbolic Execution Virtual Machine</a>
<a href="">KLEE</a> is a symbolic execution framework for
programs in LLVM bitcode form. KLEE tries to symbolically evaluate "all" paths
through the application and records state transitions that lead to fault
states. This allows it to construct testcases that lead to faults and can even
be used to verify some algorithms.
<!-- *********************************************************************** -->
<a name="externalproj">External Open Source Projects Using LLVM 3.0</a>
<!-- *********************************************************************** -->
<p>An exciting aspect of LLVM is that it is used as an enabling technology for
a lot of other language and tools projects. This section lists some of the
projects that have already been updated to work with LLVM 3.0.</p>
<p><a href="">AddressSanitizer</a>
uses compiler instrumentation and a specialized malloc library to find C/C++
bugs such as use-after-free and out-of-bound accesses to heap, stack, and
globals. The key feature of the tool is speed: the average slowdown
introduced by AddressSanitizer is less than 2x.</p>
<p><a href="">Clam AntiVirus</a> is an open source (GPL)
anti-virus toolkit for UNIX, designed especially for e-mail scanning on mail
<p>Since version 0.96 it
has <a href="">bytecode
signatures</a> that allow writing detections for complex malware.
It uses LLVM's JIT to speed up the execution of bytecode on X86, X86-64,
PPC32/64, falling back to its own interpreter otherwise. The git version was
updated to work with LLVM 3.0.</p>
<h3>clang_complete for VIM</h3>
<p><a href="">clang_complete</a> is a
VIM plugin, that provides accurate C/C++ autocompletion using the clang front
end. The development version of clang complete, can directly use libclang
which can maintain a cache to speed up auto completion.</p>
<p><a href="">clReflect</a> is a C++
parser that uses clang/LLVM to derive a light-weight reflection database
suitable for use in game development. It comes with a very simple runtime
library for loading and querying the database, requiring no external
dependencies (including CRT), and an additional utility library for object
management and serialisation.</p>
<h3>Cling C++ Interpreter</h3>
<p><a href="">Cling</a> is an interactive compiler interface
(aka C++ interpreter). It supports C++ and C, and uses LLVM's JIT and the
Clang parser. It has a prompt interface, runs source files, calls into shared
libraries, prints the value of expressions, even does runtime lookup of
identifiers (dynamic scopes). And it just behaves like one would expect from
an interpreter.</p>
<h3>Crack Programming Language</h3>
<p><a href="">Crack</a> aims to provide
the ease of development of a scripting language with the performance of a
compiled language. The language derives concepts from C++, Java and Python,
incorporating object-oriented programming, operator overloading and strong
<p><a href="">Eero</a> is a fully
header-and-binary-compatible dialect of Objective-C 2.0, implemented with a
patched version of the Clang/LLVM compiler. It features a streamlined syntax,
Python-like indentation, and new operators, for improved readability and
reduced code clutter. It also has new features such as limited forms of
operator overloading and namespaces, and strict (type-and-operator-safe)
enumerations. It is inspired by languages such as Smalltalk, Python, and
<h3>FAUST Real-Time Audio Signal Processing Language</h3>
<p><a href="">FAUST</a> is a compiled language for
real-time audio signal processing. The name FAUST stands for Functional
AUdio STream. Its programming model combines two approaches: functional
programming and block diagram composition. In addition with the C, C++, Java
output formats, the Faust compiler can now generate LLVM bitcode, and works
with LLVM 2.7-3.0.
<h3>Glasgow Haskell Compiler (GHC)</h3>
<p>GHC is an open source, state-of-the-art programming suite for Haskell, a
standard lazy functional programming language. It includes an optimizing
static compiler generating good code for a variety of platforms, together
with an interactive system for convenient, quick development.</p>
<p>GHC 7.0 and onwards include an LLVM code generator, supporting LLVM 2.8 and
later. Since LLVM 2.9, GHC now includes experimental support for the ARM
platform with LLVM 3.0.</p>
<p><a href="">gwXscript</a> is an object oriented,
aspect oriented programming language which can create both executables (ELF,
EXE) and shared libraries (DLL, SO, DYNLIB). The compiler is implemented in
its own language and translates scripts into LLVM-IR which can be optimized
and translated into native code by the LLVM framework. Source code in
gwScript contains definitions that expand the namespaces. So you can build
your project and simply 'plug out' features by removing a file. The remaining
project does not leave scars since you directly separate concerns by the
'template' feature of gwX. It is also possible to add new features to a
project by just adding files and without editing the original project. This
language is used for example to create games or content management systems
that should be extendable.</p>
<p>gwXscript is strongly typed and offers comfort with its native types string,
hash and array. You can easily write new libraries in gwXscript or native
code. gwXscript is type safe and users should not be able to crash your
program or execute malicious code except code that is eating CPU time.</p>
<p><a href="">include-what-you-use</a>
is a tool to ensure that a file directly <code>#include</code>s
all <code>.h</code> files that provide a symbol that the file uses. It also
removes superfluous <code>#include</code>s from source files.</p>
<h3>ispc: The Intel SPMD Program Compiler</h3>
<p><a href="">ispc</a> is a compiler for "single program,
multiple data" (SPMD) programs. It compiles a C-based SPMD programming
language to run on the SIMD units of CPUs; it often delivers 5-6x speedups on
a single core of a CPU with an 8-wide SIMD unit compared to serial code,
while still providing a clean and easy-to-understand programming model. For
an introduction to the language and its performance,
see <a href="">the walkthrough</a> of a short
example program. ispc is licensed under the BSD license.</p>
<h3>The Julia Programming Language</h3>
<p><a href="">Julia</a> is a high-level,
high-performance dynamic language for technical
computing. It provides a sophisticated compiler, distributed parallel
execution, numerical accuracy, and an extensive mathematical function
library. The compiler uses type inference to generate fast code
without any type declarations, and uses LLVM's optimization passes and
JIT compiler. The language is designed around multiple dispatch,
giving programs a large degree of flexibility. It is ready for use on many
kinds of problems.</p>
<h3>LanguageKit and Pragmatic Smalltalk</h3>
<p><a href="">LanguageKit</a> is
a framework for implementing dynamic languages sharing an object model with
Objective-C. It provides static and JIT compilation using LLVM along with
its own interpreter. Pragmatic Smalltalk is a dialect of Smalltalk, built on
top of LanguageKit, that interfaces directly with Objective-C, sharing the
same object representation and message sending behaviour. These projects are
developed as part of the &Eacute;toil&eacute; desktop environment.</p>
<p><a href="">LuaAV</a> is a real-time
audiovisual scripting environment based around the Lua language and a
collection of libraries for sound, graphics, and other media protocols. LuaAV
uses LLVM and Clang to JIT compile efficient user-defined audio synthesis
routines specified in a declarative syntax.</p>
<p>An open source, cross-platform implementation of C# and the CLR that is
binary compatible with Microsoft.NET. Has an optional, dynamically-loaded
LLVM code generation backend in Mini, the JIT compiler.</p>
<p>Note that we use a Git mirror of LLVM <a
href="">with some patches</a>.</p>
<p><a href="">Polly</a> is an advanced data-locality
optimizer and automatic parallelizer. It uses an advanced, mathematical
model to calculate detailed data dependency information which it uses to
optimize the loop structure of a program. Polly can speed up sequential code
by improving memory locality and consequently the cache use. Furthermore,
Polly is able to expose different kind of parallelism which it exploits by
introducing (basic) OpenMP and SIMD code. A mid-term goal of Polly is to
automatically create optimized GPU code.</p>
<h3>Portable OpenCL (pocl)</h3>
<p>Portable OpenCL is an open source implementation of the OpenCL standard which
can be easily adapted for new targets. One of the goals of the project is
improving performance portability of OpenCL programs, avoiding the need for
target-dependent manual optimizations. A "native" target is included, which
allows running OpenCL kernels on the host (CPU).</p>
<p><a href="">Pure</a> is an
algebraic/functional programming language based on term rewriting. Programs
are collections of equations which are used to evaluate expressions in a
symbolic fashion. The interpreter uses LLVM as a backend to JIT-compile Pure
programs to fast native code. Pure offers dynamic typing, eager and lazy
evaluation, lexical closures, a hygienic macro system (also based on term
rewriting), built-in list and matrix support (including list and matrix
comprehensions) and an easy-to-use interface to C and other programming
languages (including the ability to load LLVM bitcode modules, and inline C,
C++, Fortran and Faust code in Pure programs if the corresponding LLVM-enabled
compilers are installed).</p>
<p>Pure version 0.48 has been tested and is known to work with LLVM 3.0
(and continues to work with older LLVM releases &gt;= 2.5).</p>
<p><a href="">Renderscript</a>
is Android's advanced 3D graphics rendering and compute API. It provides a
portable C99-based language with extensions to facilitate common use cases
for enhancing graphics and thread level parallelism. The Renderscript
compiler frontend is based on Clang/LLVM. It emits a portable bitcode format
for the actual compiled script code, as well as reflects a Java interface for
developers to control the execution of the compiled bitcode. Executable
machine code is then generated from this bitcode by an LLVM backend on the
device. Renderscript is thus able to provide a mechanism by which Android
developers can improve performance of their applications while retaining
<p><a href="">SAFECode</a> is a memory safe C/C++
compiler built using LLVM. It takes standard, unannotated C/C++ code,
analyzes the code to ensure that memory accesses and array indexing
operations are safe, and instruments the code with run-time checks when
safety cannot be proven statically. SAFECode can be used as a debugging aid
(like Valgrind) to find and repair memory safety bugs. It can also be used
to protect code from security attacks at run-time.</p>
<h3>The Stupid D Compiler (SDC)</h3>
<p><a href="">The Stupid D Compiler</a> is a
project seeking to write a self-hosting compiler for the D programming
language without using the frontend of the reference compiler (DMD).</p>
<h3>TTA-based Co-design Environment (TCE)</h3>
<p>TCE is a toolset for designing application-specific processors (ASP) based on
the Transport triggered architecture (TTA). The toolset provides a complete
co-design flow from C/C++ programs down to synthesizable VHDL and parallel
program binaries. Processor customization points include the register files,
function units, supported operations, and the interconnection network.</p>
<p>TCE uses Clang and LLVM for C/C++ language support, target independent
optimizations and also for parts of code generation. It generates new
LLVM-based code generators "on the fly" for the designed TTA processors and
loads them in to the compiler backend as runtime libraries to avoid
per-target recompilation of larger parts of the compiler chain.</p>
<h3>Tart Programming Language</h3>
<p><a href="">Tart</a> is a general-purpose,
strongly typed programming language designed for application
developers. Strongly inspired by Python and C#, Tart focuses on practical
solutions for the professional software developer, while avoiding the clutter
and boilerplate of legacy languages like Java and C++. Although Tart is still
in development, the current implementation supports many features expected of
a modern programming language, such as garbage collection, powerful
bidirectional type inference, a greatly simplified syntax for template
metaprogramming, closures and function literals, reflection, operator
overloading, explicit mutability and immutability, and much more. Tart is
flexible enough to accommodate a broad range of programming styles and
philosophies, while maintaining a strong commitment to simplicity, minimalism
and elegance in design.</p>
<p><a href="">ThreadSanitizer</a> is a
data race detector for (mostly) C and C++ code, available for Linux, Mac OS
and Windows. On different systems, we use binary instrumentation frameworks
(Valgrind and Pin) as frontends that generate the program events for the race
detection algorithm. On Linux, there's an option of using LLVM-based
compile-time instrumentation.</p>
<!-- *********************************************************************** -->
<a name="whatsnew">What's New in LLVM 3.0?</a>
<!-- *********************************************************************** -->
<p>This release includes a huge number of bug fixes, performance tweaks and
minor improvements. Some of the major improvements and new features are
listed in this section.</p>
<a name="majorfeatures">Major New Features</a>
<!-- Features that need text if they're finished for 3.1:
strong phi elim
loop dependence analysis
lib/Transforms/IPO/MergeFunctions.cpp => consider for 3.1.
Integrated assembler on by default for arm/thumb?
<!-- Near dead:
Analysis/RegionInfo.h + Dom Frontiers
SparseBitVector: used in LiveVar.
llvm/lib/Archive - replace with lib object?
<p>LLVM 3.0 includes several major changes and big features:</p>
<li>llvm-gcc is no longer supported, and not included in the release. We
recommend switching to <a
href="">Clang</a> or <a
<li>The linear scan register allocator has been replaced with a new "greedy"
register allocator, enabling live range splitting and many other
optimizations that lead to better code quality. Please see its <a
href="">blog post</a> or its talk at the <a
href="">Developer Meeting</a>
for more information.</li>
<li>LLVM IR now includes full support for <a href="Atomics.html">atomics
memory operations</a> intended to support the C++'11 and C'1x memory models.
This includes <a href="LangRef.html#memoryops">atomic load and store,
compare and exchange, and read/modify/write instructions</a> as well as a
full set of <a href="LangRef.html#ordering">memory ordering constraints</a>.
Please see the <a href="Atomics.html">Atomics Guide</a> for more
<li>The LLVM IR exception handling representation has been redesigned and
reimplemented, making it more elegant, fixing a huge number of bugs, and
enabling inlining and other optimizations. Please see its <a href=
post</a> and the <a href="ExceptionHandling.html">Exception Handling
documentation</a> for more information.</li>
<li>The LLVM IR Type system has been redesigned and reimplemented, making it
faster and solving some long-standing problems.
Please see its <a
post</a> for more information.</li>
<li>The MIPS backend has made major leaps in this release, going from an
experimental target to being virtually production quality and supporting a
wide variety of MIPS subtargets. See the <a href="#MIPS">MIPS section</a>
below for more information.</li>
<li>The optimizer and code generator now supports gprof and gcov-style coverage
and profiling information, and includes a new llvm-cov tool (but also works
with gcov). Clang exposes coverage and profiling through GCC-compatible
command line options.</li>
<a name="coreimprovements">LLVM IR and Core Improvements</a>
<p>LLVM IR has several new features for better support of new targets and that
expose new optimization opportunities:</p>
<li><a href="Atomics.html">Atomic memory accesses and memory ordering</a> are
now directly expressible in the IR.</li>
<li>A new <a href="LangRef.html#int_fma">llvm.fma intrinsic</a> directly
represents floating point multiply accumulate operations without an
intermediate rounding stage.</li>
<li>A new llvm.expect intrinsic allows a frontend to express expected control
flow (and the __builtin_expect builtin from GNU C).</li>
<li>The <a href="LangRef.html#int_prefetch">llvm.prefetch intrinsic</a> now
takes a 4th argument that specifies whether the prefetch happens from the
icache or dcache.</li>
<li>The new <a href="LangRef.html#uwtable">uwtable function attribute</a>
allows a frontend to control emission of unwind tables.</li>
<li>The new <a href="LangRef.html#fnattrs">nonlazybind function
attribute</a> allow optimization of Global Offset Table (GOT) accesses.</li>
<li>The new <a href="LangRef.html#returns_twice">returns_twice attribute</a>
allows better modeling of functions like setjmp.</li>
<li>The <a href="LangRef.html#datalayout">target datalayout</a> string can now
encode the natural alignment of the target's stack for better optimization.
<a name="optimizer">Optimizer Improvements</a>
<p>In addition to many minor performance tweaks and bug fixes, this
release includes a few major enhancements and additions to the
<li>The pass manager now has an extension API that allows front-ends and plugins
to insert their own optimizations in the well-known places in the standard
pass optimization pipeline.</li>
<li>Information about <a href="BranchWeightMetadata.html">branch probability</a>
and basic block frequency is now available within LLVM, based on a
combination of static branch prediction heuristics and
<code>__builtin_expect</code> calls. That information is currently used for
register spill placement and if-conversion, with additional optimizations
planned for future releases. The same framework is intended for eventual
use with profile-guided optimization.</li>
<li>The "-indvars" induction variable simplification pass only modifies
induction variables when profitable. Sign and zero extension
elimination, linear function test replacement, loop unrolling, and
other simplifications that require induction variable analysis have
been generalized so they no longer require loops to be rewritten into
canonical form prior to optimization. This new design
preserves more IR level information, avoids undoing earlier loop
optimizations (particularly hand-optimized loops), and no longer
requires the code generator to reconstruct loops into an optimal form -
an intractable problem.</li>
<li>LLVM now includes a pass to optimize retain/release calls for the
<a href="">Automatic
Reference Counting</a> (ARC) Objective-C language feature (in
lib/Transforms/Scalar/ObjCARC.cpp). It is a decent example of implementing
a source-language-specific optimization in LLVM.</li>
<a name="mc">MC Level Improvements</a>
<p>The LLVM Machine Code (aka MC) subsystem was created to solve a number of
problems in the realm of assembly, disassembly, object file format handling,
and a number of other related areas that CPU instruction-set level tools work
in. For more information, please see
the <a href="">Intro
to the LLVM MC Project Blog Post</a>.</p>
<li>The MC layer has undergone significant refactoring to eliminate layering
violations that caused it to pull in the LLVM compiler backend code.</li>
<li>The ELF object file writers are much more full featured.</li>
<li>The integrated assembler now supports #line directives.</li>
<li>An early implementation of a JIT built on top of the MC framework (known
as MC-JIT) has been implemented and will eventually replace the old JIT.
It emits object files direct to memory and uses a runtime dynamic linker to
resolve references and drive lazy compilation. The MC-JIT enables much
greater code reuse between the JIT and the static compiler and provides
better integration with the platform ABI as a result.
<li>The assembly printer now makes uses of assemblers instruction aliases
(InstAliases) to print simplified mneumonics when possible.</li>
<li>TableGen can now autogenerate MC expansion logic for pseudo
instructions that expand to multiple MC instructions (through the
PseudoInstExpansion class).</li>
<li>A new llvm-dwarfdump tool provides a start of a drop-in
replacement for the corresponding tool that use LLVM libraries. As part of
this, LLVM has the beginnings of a dwarf parsing library.</li>
<li>llvm-objdump has more output including, symbol by symbol disassembly,
inline relocations, section headers, symbol tables, and section contents.
Support for archive files has also been added.</li>
<li>llvm-nm has gained support for archives of binary files.</li>
<li>llvm-size has been added. This tool prints out section sizes.</li>
<a name="codegen">Target Independent Code Generator Improvements</a>
<p>We have put a significant amount of work into the code generator
infrastructure, which allows us to implement more aggressive algorithms and
make it run faster:</p>
<li>LLVM can now produce code that works with libgcc
to <a href="SegmentedStacks.html">dynamically allocate stack
segments</a>, as opposed to allocating a worst-case chunk of
virtual memory for each thread.</li>
<li>LLVM generates substantially better code for indirect gotos due to a new
tail duplication pass, which can be a substantial performance win for
interpreter loops that use them.</li>
<li>Exception handling and debug frame information is now emitted with CFI
directives. This lets the assembler produce more compact info as it knows
the final offsets, yielding <a href="">much smaller executables</a> for some C++ applications.
If the system assembler doesn't support it, MC exands the directives when
the integrated assembler is not used.
<li>The code generator now supports vector "select" operations on vector
comparisons, turning them into various optimized code sequences (e.g.
using the SSE4/AVX "blend" instructions).</li>
<li>The SSE execution domain fix pass and the ARM NEON move fix pass have been
merged to a target independent execution dependency fix pass. This pass is
used to select alternative equivalent opcodes in a way that minimizes
execution domain crossings. Closely connected instructions are moved to
the same execution domain when possible. Targets can override the
<code>getExecutionDomain</code> and <code>setExecutionDomain</code> hooks
to use the pass.</li>
<a name="x86">X86-32 and X86-64 Target Improvements</a>
<p>New features and major changes in the X86 target include:</p>
<li>The X86 backend, assembler and disassembler now have full support for AVX 1.
To enable it pass <code>-mavx</code> to the compiler. AVX2 implementation is
underway on mainline.</li>
<li>The integrated assembler and disassembler now support a broad range of new
instructions including Atom, Ivy Bridge, <a
href="">SSE4a/BMI</a> instructions, <a
href="">rdrand</a> and many others.</li>
<li>The X86 backend now fully supports the <a href="">X87
floating point stack inline assembly constraints</a>.</li>
<li>The integrated assembler now supports the <tt>.code32</tt> and
<tt>.code64</tt> directives to switch between 32-bit and 64-bit
<li>The X86 backend now synthesizes horizontal add/sub instructions from generic
vector code when the appropriate instructions are enabled.</li>
<li>The X86-64 backend generates smaller and faster code at -O0 due to
improvements in fast instruction selection.</li>
<li><a href="">Native Client</a>
subtarget support has been added.</li>
<li>The CRC32 intrinsics have been renamed. The intrinsics were previously
and <code>@llvm.x86.sse42.crc64.[8|64]</code>. They have been renamed to
<code>@llvm.x86.sse42.crc32.32.[8|16|32]</code> and
<a name="ARM">ARM Target Improvements</a>
<p>New features of the ARM target include:</p>
<li>The ARM backend generates much faster code for Cortex-A9 chips.</li>
<li>The ARM backend has improved support for Cortex-M series processors.</li>
<li>The ARM inline assembly constraints have been implemented and are now fully
<li>NEON code produced by Clang often runs much faster due to improvements in
the Scalar Replacement of Aggregates pass.</li>
<li>The old ARM disassembler is replaced with a new one based on autogenerated
encoding information from ARM .td files.</li>
<li>The integrated assembler has made major leaps forward, but is still beta quality in LLVM 3.0.</li>
<a name="MIPS">MIPS Target Improvements</a>
<p>This release has seen major new work on just about every aspect of the MIPS
backend. Some of the major new features include:</p>
<li>Most MIPS32r1 and r2 instructions are now supported.</li>
<li>LE/BE MIPS32r1/r2 has been tested extensively.</li>
<li>O32 ABI has been fully tested.</li>
<li>MIPS backend has migrated to using the MC infrastructure for assembly printing. Initial support for direct object code emission has been implemented too.</li>
<li>Delay slot filler has been updated. Now it tries to fill delay slots with useful instructions instead of always filling them with NOPs.</li>
<li>Support for old-style JIT is complete.</li>
<li>Support for old architectures (MIPS1 and MIPS2) has been removed.</li>
<li>Initial support for MIPS64 has been added.</li>
<a name="PTX">PTX Target Improvements</a>
The PTX back-end is still experimental, but is fairly usable for compute kernels
in LLVM 3.0. Most scalar arithmetic is implemented, as well as intrinsics to
access the special PTX registers and sync instructions. The major missing
pieces are texture/sampler support and some vector operations.</p>
<p>That said, the backend is already being used for domain-specific languages
and can be used by Clang to
<a href="">compile OpenCL
C code</a> into PTX.</p>
<a name="OtherTS">Other Target Specific Improvements</a>
<li>Many PowerPC improvements have been implemented for ELF targets, including
support for varargs and initial support for direct .o file emission.</li>
<li>MicroBlaze scheduling itineraries were added that model the
3-stage and the 5-stage pipeline architectures. The 3-stage
pipeline model can be selected with <code>-mcpu=mblaze3</code>
and the 5-stage pipeline model can be selected with
<a name="changes">Major Changes and Removed Features</a>
<p>If you're already an LLVM user or developer with out-of-tree changes based on
LLVM 2.9, this section lists some "gotchas" that you may run into upgrading
from the previous release.</p>
<li>LLVM 3.0 removes support for reading LLVM 2.8 and earlier files, and LLVM
3.1 will eliminate support for reading LLVM 2.9 files. Going forward, we
aim for all future versions of LLVM to read bitcode files and .ll files
produced by LLVM 3.0.</li>
<li>Tablegen has been split into a library, allowing the clang tblgen pieces
to now live in the clang tree. The llvm version has been renamed to
llvm-tblgen instead of tblgen.</li>
<li>The <code>LLVMC</code> meta compiler driver was removed.</li>
<li>The unused PostOrder Dominator Frontiers and LowerSetJmp passes were removed.</li>
<li>The old <code>TailDup</code> pass was not used in the standard pipeline
and was unable to update ssa form, so it has been removed.
<li>The syntax of volatile loads and stores in IR has been changed to
"<code>load volatile</code>"/"<code>store volatile</code>". The old
syntax ("<code>volatile load</code>"/"<code>volatile store</code>")
is still accepted, but is now considered deprecated and will be removed in
<li>llvm-gcc's frontend tests have been removed from llvm/test/Frontend*, sunk
into the clang and dragonegg testsuites.</li>
<li>The old atomic intrinsics (<code>llvm.memory.barrier</code> and
<code>llvm.atomic.*</code>) are now gone. Please use the new atomic
instructions, described in the <a href="Atomics.html">atomics guide</a>.
<li>LLVM's configure script doesn't depend on llvm-gcc anymore, eliminating a
strange circular dependence between projects.</li>
<h4>Windows (32-bit)</h4>
<li>On Win32(MinGW32 and MSVC), Windows 2000 will not be supported.
Windows XP or higher is required.</li>
<a name="api_changes">Internal API Changes</a>
<p>In addition, many APIs have changed in this release. Some of the major
LLVM API changes are:</p>
<li>The biggest and most pervasive change is that the type system has been
rewritten: <code>PATypeHolder</code> and <code>OpaqueType</code> are gone,
and all APIs deal with <code>Type*</code> instead of <code>const
Type*</code>. If you need to create recursive structures, then create a
named structure, and use <code>setBody()</code> when all its elements are
built. Type merging and refining is gone too: named structures are not
merged with other structures, even if their layout is identical. (of
course anonymous structures are still uniqued by layout).</li>
<li><code>PHINode::reserveOperandSpace</code> has been removed. Instead, you
must specify how many operands to reserve space for when you create the
PHINode, by passing an extra argument
into <code>PHINode::Create</code>.</li>
<li>PHINodes no longer store their incoming BasicBlocks as operands. Instead,
the list of incoming BasicBlocks is stored separately, and can be accessed
with new functions <code>PHINode::block_begin</code>
and <code>PHINode::block_end</code>.</li>
<li>Various functions now take an <code>ArrayRef</code> instead of either a
pair of pointers (or iterators) to the beginning and end of a range, or a
pointer and a length. Others now return an <code>ArrayRef</code> instead
of a reference to a <code>SmallVector</code>
or <code>std::vector</code>. These include:
<!-- Please keep this list sorted. -->
<li><code>ComputeLinearIndex</code> (in <code>llvm/CodeGen/Analysis.h</code>)</li>
<li><code>ConstantFoldCall</code> (in <code>llvm/Analysis/ConstantFolding.h</code>)</li>
<li><code>ConstantFoldInstOperands</code> (in <code>llvm/Analysis/ConstantFolding.h</code>)</li>
<li><code>FindInsertedValue</code> (in <code>llvm/Analysis/ValueTracking.h</code>)</li>
<li><code>gep_type_begin</code> (in <code>llvm/Support/GetElementPtrTypeIterator.h</code>)</li>
<li><code>gep_type_end</code> (in <code>llvm/Support/GetElementPtrTypeIterator.h</code>)</li>
<li><code>SimplifyGEPInst</code> (in <code>llvm/Analysis/InstructionSimplify.h</code>)</li>
<li>All forms of <code>StringMap::getOrCreateValue</code> have been remove
except for the one which takes a <code>StringRef</code>.</li>
<li>The <code>LLVMBuildUnwind</code> function from the C API was removed. The
LLVM <code>unwind</code> instruction has been deprecated for a long time
and isn't used by the current front-ends. So this was removed during the
exception handling rewrite.</li>
<li>The <code>LLVMAddLowerSetJmpPass</code> function from the C API was
removed because the <code>LowerSetJmp</code> pass was removed.</li>
<li>The <code>DIBuilder</code> interface used by front ends to encode
debugging information in the LLVM IR now expects clients to
use <code>DIBuilder::finalize()</code> at the end of translation unit to
complete debugging information encoding.</li>
<li>TargetSelect.h moved to Support/ from Target/</li>
<li>UpgradeIntrinsicCall no longer upgrades pre-2.9 intrinsic calls (for
example <code>llvm.memset.i32</code>).</li>
<li>It is mandatory to initialize all out-of-tree passes too and their dependencies now with
and <code>INITIALIZE_{PASS,AG}_DEPENDENCY</code>.</li>
<li>The interface for MemDepResult in MemoryDependenceAnalysis has been
enhanced with new return types Unknown and NonFuncLocal, in addition to
the existing types Clobber, Def, and NonLocal.</li>
<!-- *********************************************************************** -->
<a name="knownproblems">Known Problems</a>
<!-- *********************************************************************** -->
<p>LLVM is generally a production quality compiler, and is used by a broad range
of applications and shipping in many products. That said, not every
subsystem is as mature as the aggregate, particularly the more obscure
targets. If you run into a problem, please check the <a
href="">LLVM bug database</a> and submit a bug if
there isn't already one or ask on the <a
<p>Known problem areas include:</p>
<li>The Alpha, Blackfin, CellSPU, MSP430, PTX, SystemZ and
XCore backends are experimental, and the Alpha, Blackfin and SystemZ
targets have already been removed from mainline.</li>
<li>The integrated assembler, disassembler, and JIT is not supported by
several targets. If an integrated assembler is not supported, then a
system assembler is required. For more details, see the <a
href="CodeGenerator.html#targetfeatures">Target Features Matrix</a>.
<li>The C backend has numerous problems and is not being actively maintained.
Depending on it for anything serious is not advised.</li>
<!-- *********************************************************************** -->
<a name="additionalinfo">Additional Information</a>
<!-- *********************************************************************** -->
<p>A wide variety of additional information is available on
the <a href="">LLVM web page</a>, in particular in
the <a href="">documentation</a> section. The web page
also contains versions of the API documentation which is up-to-date with the
Subversion version of the source code. You can access versions of these
documents specific to this release by going into the "<tt>llvm/doc/</tt>"
directory in the LLVM tree.</p>
<p>If you have any questions or comments about LLVM, please feel free to contact
us via the <a href=""> mailing lists</a>.</p>
<!-- EH details: to be moved to a blog post:
<p>One of the biggest changes is that 3.0 has a new exception handling
system. The old system used LLVM intrinsics to convey the exception handling
information to the code generator. It worked in most cases, but not
all. Inlining was especially difficult to get right. Also, the intrinsics
could be moved away from the <code>invoke</code> instruction, making it hard
to recover that information.</p>
<p>The new EH system makes exception handling a first-class member of the IR. It
adds two new instructions:</p>
<li><a href="LangRef.html#i_landingpad"><code>landingpad</code></a> &mdash;
this instruction defines a landing pad basic block. It contains all of the
information that's needed by the code generator. It's also required to be
the first non-PHI instruction in the landing pad. In addition, a landing
pad may be jumped to only by the unwind edge of an <code>invoke</code>
<li><a href="LangRef.html#i_resume"><code>resume</code></a> &mdash; this
instruction causes the current exception to resume traveling up the
stack. It replaces the <code></code> intrinsic.</li>
<p>Converting from the old EH API to the new EH API is rather simple, because a
lot of complexity has been removed. The two intrinsics,
<code></code> and <code></code> have been
superseded by the <code>landingpad</code> instruction. Instead of generating
a call to <code></code> and <code></code>:
<div class="doc_code">
Function *ExcIntr = Intrinsic::getDeclaration(TheModule,
Function *SlctrIntr = Intrinsic::getDeclaration(TheModule,
// The exception pointer.
Value *ExnPtr = Builder.CreateCall(ExcIntr, "exc_ptr");
std::vector&lt;Value*&gt; Args;
<i>// Add selector clauses to Args.</i>
// The selector call.
Builder.CreateCall(SlctrIntr, Args, "exc_sel");
<p>You should instead generate a <code>landingpad</code> instruction, that
returns an exception object and selector value:</p>
<div class="doc_code">
LandingPadInst *LPadInst =
Builder.CreateLandingPad(StructType::get(Int8PtrTy, Int32Ty, NULL),
Personality, 0);
Value *LPadExn = Builder.CreateExtractValue(LPadInst, 0);
Builder.CreateStore(LPadExn, getExceptionSlot());
Value *LPadSel = Builder.CreateExtractValue(LPadInst, 1);
Builder.CreateStore(LPadSel, getEHSelectorSlot());
<p>It's now trivial to add the individual clauses to the <code>landingpad</code>
<div class="doc_code">
<i><b>// Adding a catch clause</b></i>
Constant *TypeInfo = getTypeInfo();
<i><b>// Adding a C++ catch-all</b></i>
<i><b>// Adding a cleanup</b></i>
<i><b>// Adding a filter clause</b></i>
std::vector&lt;Constant*&gt; TypeInfos;
Constant *TypeInfo = getFilterTypeInfo();
TypeInfos.push_back(Builder.CreateBitCast(TypeInfo, Builder.getInt8PtrTy()));
ArrayType *FilterTy = ArrayType::get(Int8PtrTy, TypeInfos.size());
LPadInst-&gt;addClause(ConstantArray::get(FilterTy, TypeInfos));
<p>Converting from using the <code></code> intrinsic to
the <code>resume</code> instruction is trivial. It takes the exception
pointer and exception selector values returned by
the <code>landingpad</code> instruction:</p>
<div class="doc_code">
Type *UnwindDataTy = StructType::get(Builder.getInt8PtrTy(),
Builder.getInt32Ty(), NULL);
Value *UnwindData = UndefValue::get(UnwindDataTy);
Value *ExcPtr = Builder.CreateLoad(getExceptionObjSlot());
Value *ExcSel = Builder.CreateLoad(getExceptionSelSlot());
UnwindData = Builder.CreateInsertValue(UnwindData, ExcPtr, 0, "exc_ptr");
UnwindData = Builder.CreateInsertValue(UnwindData, ExcSel, 1, "exc_sel");
<!-- *********************************************************************** -->
<a href=""><img
src="" alt="Valid CSS"></a>
<a href=""><img
src="" alt="Valid HTML 4.01"></a>
<a href="">LLVM Compiler Infrastructure</a><br>
Last modified: $Date$