1 <!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01//EN"
2 "http://www.w3.org/TR/html4/strict.dtd">
5 <meta http-equiv="Content-Type" content="text/html; charset=utf-8">
6 <link rel="stylesheet" href="llvm.css" type="text/css">
7 <title>LLVM 3.0 Release Notes</title>
11 <h1>LLVM 3.0 Release Notes</h1>
13 <img align=right src="http://llvm.org/img/DragonSmall.png"
14 width="136" height="136" alt="LLVM Dragon Logo">
17 <li><a href="#intro">Introduction</a></li>
18 <li><a href="#subproj">Sub-project Status Update</a></li>
19 <li><a href="#externalproj">External Projects Using LLVM 3.0</a></li>
20 <li><a href="#whatsnew">What's New in LLVM 3.0?</a></li>
21 <li><a href="GettingStarted.html">Installation Instructions</a></li>
22 <li><a href="#knownproblems">Known Problems</a></li>
23 <li><a href="#additionalinfo">Additional Information</a></li>
26 <div class="doc_author">
27 <p>Written by the <a href="http://llvm.org/">LLVM Team</a></p>
31 <h1 style="color:red">These are in-progress notes for the upcoming LLVM 3.0
34 <a href="http://llvm.org/releases/2.9/docs/ReleaseNotes.html">LLVM 2.9
35 Release Notes</a>.</h1>
38 <!-- *********************************************************************** -->
40 <a name="intro">Introduction</a>
42 <!-- *********************************************************************** -->
46 <p>This document contains the release notes for the LLVM Compiler
47 Infrastructure, release 3.0. Here we describe the status of LLVM, including
48 major improvements from the previous release and significant known problems.
49 All LLVM releases may be downloaded from
50 the <a href="http://llvm.org/releases/">LLVM releases web site</a>.</p>
52 <p>For more information about LLVM, including information about the latest
53 release, please check out the <a href="http://llvm.org/">main LLVM web
54 site</a>. If you have questions or comments,
55 the <a href="http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev">LLVM
56 Developer's Mailing List</a> is a good place to send them.</p>
58 <p>Note that if you are reading this file from a Subversion checkout or the main
59 LLVM web page, this document applies to the <i>next</i> release, not the
60 current one. To see the release notes for a specific release, please see the
61 <a href="http://llvm.org/releases/">releases page</a>.</p>
65 <!-- Features that need text if they're finished for 3.1:
69 loop dependence analysis
70 CorrelatedValuePropagation
71 lib/Transforms/IPO/MergeFunctions.cpp => consider for 3.1.
74 <!-- *********************************************************************** -->
76 <a name="subproj">Sub-project Status Update</a>
78 <!-- *********************************************************************** -->
82 <p>The LLVM 3.0 distribution currently consists of code from the core LLVM
83 repository (which roughly includes the LLVM optimizers, code generators and
84 supporting tools), the Clang repository and the llvm-gcc repository. In
85 addition to this code, the LLVM Project includes other sub-projects that are
86 in development. Here we include updates on these subprojects.</p>
88 <!--=========================================================================-->
90 <a name="clang">Clang: C/C++/Objective-C Frontend Toolkit</a>
95 <p><a href="http://clang.llvm.org/">Clang</a> is an LLVM front end for the C,
96 C++, and Objective-C languages. Clang aims to provide a better user
97 experience through expressive diagnostics, a high level of conformance to
98 language standards, fast compilation, and low memory use. Like LLVM, Clang
99 provides a modular, library-based architecture that makes it suitable for
100 creating or integrating with other development tools. Clang is considered a
101 production-quality compiler for C, Objective-C, C++ and Objective-C++ on x86
102 (32- and 64-bit), and for darwin/arm targets.</p>
104 <p>In the LLVM 3.0 time-frame, the Clang team has made many improvements:</p>
107 <li>Greatly improved support for building C++ applications, with greater
108 stability and better diagnostics.</li>
110 <li><a href="http://clang.llvm.org/cxx_status.html">Improved support</a> for
111 the <a href="http://www.iso.org/iso/iso_catalogue/catalogue_tc/catalogue_detail.htm?csnumber=50372">C++
112 2011</a> standard, including implementations of non-static data member
113 initializers, alias templates, delegating constructors, the range-based
114 for loop, and implicitly-generated move constructors and move assignment
115 operators, among others.</li>
117 <li>Implemented support for some features of the upcoming C1x standard,
118 including static assertions and generic selections.</li>
120 <li>Better detection of include and linking paths for system headers and
121 libraries, especially for Linux distributions.</li>
123 <li>Implemented support
124 for <a href="http://clang.llvm.org/docs/AutomaticReferenceCounting.html">Automatic
125 Reference Counting</a> for Objective-C.</li>
127 <li>Implemented a number of optimizations in <tt>libclang</tt>, the Clang C
128 interface, to improve the performance of code completion and the mapping
129 from source locations to abstract syntax tree nodes.</li>
133 <p>If Clang rejects your code but another compiler accepts it, please take a
134 look at the <a href="http://clang.llvm.org/compatibility.html">language
135 compatibility</a> guide to make sure this is not intentional or a known
140 <!--=========================================================================-->
142 <a name="dragonegg">DragonEgg: GCC front-ends, LLVM back-end</a>
146 <p><a href="http://dragonegg.llvm.org/">DragonEgg</a> is a
147 <a href="http://gcc.gnu.org/wiki/plugins">gcc plugin</a> that replaces GCC's
148 optimizers and code generators with LLVM's. It works with gcc-4.5 or gcc-4.6,
149 targets the x86-32 and x86-64 processor families, and has been successfully
150 used on the Darwin, FreeBSD, KFreeBSD, Linux and OpenBSD platforms. It fully
151 supports Ada, C, C++ and Fortran. It has partial support for Go, Java, Obj-C
154 <p>The 3.0 release has the following notable changes:</p>
156 <li>GCC version 4.6 is now fully supported.</li>
158 <li>Patching and building GCC is no longer required: the plugin should work
159 with your system GCC (version 4.5 or 4.6; on Debian/Ubuntu systems the
160 gcc-4.5-plugin-dev or gcc-4.6-plugin-dev package is also needed).</li>
162 <li>The <tt>-fplugin-arg-dragonegg-enable-gcc-optzns</tt> option, which runs
163 GCC's optimizers as well as LLVM's, now works much better. This is the
164 option to use if you want ultimate performance! It not yet completely
165 stable: it may cause the plugin to crash.</li>
167 <li>The type and constant conversion logic has been almost entirely rewritten,
168 fixing a multitude of obscure bugs.</li>
178 <!--=========================================================================-->
180 <a name="compiler-rt">compiler-rt: Compiler Runtime Library</a>
185 <p>The new LLVM <a href="http://compiler-rt.llvm.org/">compiler-rt project</a>
186 is a simple library that provides an implementation of the low-level
187 target-specific hooks required by code generation and other runtime
188 components. For example, when compiling for a 32-bit target, converting a
189 double to a 64-bit unsigned integer is compiled into a runtime call to the
190 "__fixunsdfdi" function. The compiler-rt library provides highly optimized
191 implementations of this and other low-level routines (some are 3x faster than
192 the equivalent libgcc routines).</p>
194 <p>In the LLVM 3.0 timeframe,</p>
198 <!--=========================================================================-->
200 <a name="lldb">LLDB: Low Level Debugger</a>
205 <p>LLDB has advanced by leaps and bounds in the 3.0 timeframe. It is
206 dramatically more stable and useful, and includes both a
207 new <a href="http://lldb.llvm.org/tutorial.html">tutorial</a> and
208 a <a href="http://lldb.llvm.org/lldb-gdb.html">side-by-side comparison with
213 <!--=========================================================================-->
215 <a name="libc++">libc++: C++ Standard Library</a>
220 <p>Like compiler_rt, libc++ is now <a href="DeveloperPolicy.html#license">dual
221 licensed</a> under the MIT and UIUC license, allowing it to be used more
227 <!--=========================================================================-->
229 <a name="LLBrowse">LLBrowse: IR Browser</a>
234 <p><a href="http://llvm.org/svn/llvm-project/llbrowse/trunk/doc/LLBrowse.html">
235 LLBrowse</a> is an interactive viewer for LLVM modules. It can load any LLVM
236 module and displays its contents as an expandable tree view, facilitating an
237 easy way to inspect types, functions, global variables, or metadata nodes. It
238 is fully cross-platform, being based on the popular wxWidgets GUI
243 <!--=========================================================================-->
245 <a name="vmkit">VMKit</a>
250 <p>The <a href="http://vmkit.llvm.org/">VMKit project</a> is an implementation
251 of a Java Virtual Machine (Java VM or JVM) that uses LLVM for static and
252 just-in-time compilation. As of LLVM 3.0, VMKit now supports generational
253 garbage collectors. The garbage collectors are provided by the MMTk
254 framework, and VMKit can be configured to use one of the numerous implemented
255 collectors of MMTk.</p>
260 <!--=========================================================================-->
263 <a name="klee">KLEE: A Symbolic Execution Virtual Machine</a>
268 <a href="http://klee.llvm.org/">KLEE</a> is a symbolic execution framework for
269 programs in LLVM bitcode form. KLEE tries to symbolically evaluate "all" paths
270 through the application and records state transitions that lead to fault
271 states. This allows it to construct testcases that lead to faults and can even
272 be used to verify some algorithms.
280 <!-- *********************************************************************** -->
282 <a name="externalproj">External Open Source Projects Using LLVM 3.0</a>
284 <!-- *********************************************************************** -->
288 <p>An exciting aspect of LLVM is that it is used as an enabling technology for
289 a lot of other language and tools projects. This section lists some of the
290 projects that have already been updated to work with LLVM 3.0.</p>
292 <!--=========================================================================-->
293 <h3>AddressSanitizer</h3>
297 <p><a href="http://code.google.com/p/address-sanitizer/">AddressSanitizer</a>
298 uses compiler instrumentation and a specialized malloc library to find C/C++
299 bugs such as use-after-free and out-of-bound accesses to heap, stack, and
300 globals. The key feature of the tool is speed: the average slowdown
301 introduced by AddressSanitizer is less than 2x.</p>
305 <!--=========================================================================-->
310 <p><a href="http://www.clamav.net">Clam AntiVirus</a> is an open source (GPL)
311 anti-virus toolkit for UNIX, designed especially for e-mail scanning on mail
314 <p>Since version 0.96 it
315 has <a href="http://vrt-sourcefire.blogspot.com/2010/09/introduction-to-clamavs-low-level.html">bytecode
316 signatures</a> that allow writing detections for complex malware.</p>
318 <p>It uses LLVM's JIT to speed up the execution of bytecode on X86, X86-64,
319 PPC32/64, falling back to its own interpreter otherwise. The git version was
320 updated to work with LLVM 3.0.</p>
324 <!--=========================================================================-->
329 <p><a href="https://bitbucket.org/dwilliamson/clreflect">clReflect</a> is a C++
330 parser that uses clang/LLVM to derive a light-weight reflection database
331 suitable for use in game development. It comes with a very simple runtime
332 library for loading and querying the database, requiring no external
333 dependencies (including CRT), and an additional utility library for object
334 management and serialisation.</p>
338 <!--=========================================================================-->
339 <h3>Cling C++ Interpreter</h3>
343 <p><a href="http://cern.ch/cling">Cling</a> is an interactive compiler interface
344 (aka C++ interpreter). It uses LLVM's JIT and clang; it currently supports
345 C++ and C. It has a prompt interface, runs source files, calls into shared
346 libraries, prints the value of expressions, even does runtime lookup of
347 identifiers (dynamic scopes). And it just behaves like one would expect from
352 <!--=========================================================================-->
353 <h3>Crack Programming Language</h3>
357 <p><a href="http://code.google.com/p/crack-language/">Crack</a> aims to provide
358 the ease of development of a scripting language with the performance of a
359 compiled language. The language derives concepts from C++, Java and Python,
360 incorporating object-oriented programming, operator overloading and strong
365 <!--=========================================================================-->
370 <p><a href="http://eerolanguage.org/">Eero</a> is a fully
371 header-and-binary-compatible dialect of Objective-C 2.0, implemented with a
372 patched version of the Clang/LLVM compiler. It features a streamlined syntax,
373 Python-like indentation, and new operators, for improved readability and
374 reduced code clutter. It also has new features such as limited forms of
375 operator overloading and namespaces, and strict (type-and-operator-safe)
376 enumerations. It is inspired by languages such as Smalltalk, Python, and
381 <!--=========================================================================-->
382 <h3>Glasgow Haskell Compiler (GHC)</h3>
386 <p>GHC is an open source, state-of-the-art programming suite for Haskell, a
387 standard lazy functional programming language. It includes an optimizing
388 static compiler generating good code for a variety of platforms, together
389 with an interactive system for convenient, quick development.</p>
391 <p>GHC 7.0 and onwards include an LLVM code generator, supporting LLVM 2.8 and
392 later. Since LLVM 2.9, GHC now includes experimental support for the ARM
393 platform with LLVM 3.0.</p>
397 <!--=========================================================================-->
402 <p><a href="http://botwars.tk/gwscript/">gwXscript</a> is an object oriented,
403 aspect oriented programming language which can create both executables (ELF,
404 EXE) and shared libraries (DLL, SO, DYNLIB). The compiler is implemented in
405 its own language and translates scripts into LLVM-IR which can be optimized
406 and translated into native code by the LLVM framework. Source code in
407 gwScript contains definitions that expand the namespaces. So you can build
408 your project and simply 'plug out' features by removing a file. The remaining
409 project does not leave scars since you directly separate concerns by the
410 'template' feature of gwX. It is also possible to add new features to a
411 project by just adding files and without editing the original project. This
412 language is used for example to create games or content management systems
413 that should be extendable.</p>
415 <p>gwXscript is strongly typed and offers comfort with its native types string,
416 hash and array. You can easily write new libraries in gwXscript or native
417 code. gwXscript is type safe and users should not be able to crash your
418 program or execute malicious code except code that is eating CPU time.</p>
422 <!--=========================================================================-->
423 <h3>include-what-you-use</h3>
427 <p><a href="http://code.google.com/p/include-what-you-use">include-what-you-use</a>
428 is a tool to ensure that a file directly <code>#include</code>s
429 all <code>.h</code> files that provide a symbol that the file uses. It also
430 removes superfluous <code>#include</code>s from source files.</p>
434 <!--=========================================================================-->
435 <h3>ispc: The Intel SPMD Program Compiler</h3>
439 <p><a href="http://ispc.github.com">ispc</a> is a compiler for "single program,
440 multiple data" (SPMD) programs. It compiles a C-based SPMD programming
441 language to run on the SIMD units of CPUs; it often delivers 5-6x speedups on
442 a single core of a CPU with an 8-wide SIMD unit compared to serial code,
443 while still providing a clean and easy-to-understand programming model. For
444 an introduction to the language and its performance,
445 see <a href="http://ispc.github.com/example.html">the walkthrough of a short
446 example program. ispc is licensed under the BSD license.</p>
450 <!--=========================================================================-->
451 <h3>LanguageKit and Pragmatic Smalltalk</h3>
455 <p><a href="http://etoileos.com/etoile/features/languagekit/">LanguageKit</a> is
456 a framework for implementing dynamic languages sharing an object model with
457 Objective-C. It provides static and JIT compilation using LLVM along with
458 its own interpreter. Pragmatic Smalltalk is a dialect of Smalltalk, built on
459 top of LanguageKit, that interfaces directly with Objective-C, sharing the
460 same object representation and message sending behaviour. These projects are
461 developed as part of the Étoié desktop environment.</p>
465 <!--=========================================================================-->
470 <p><a href="http://lua-av.mat.ucsb.edu/blog/">LuaAV</a> is a real-time
471 audiovisual scripting environment based around the Lua language and a
472 collection of libraries for sound, graphics, and other media protocols. LuaAV
473 uses LLVM and Clang to JIT compile efficient user-defined audio synthesis
474 routines specified in a declarative syntax.</p>
478 <!--=========================================================================-->
483 <p>An open source, cross-platform implementation of C# and the CLR that is
484 binary compatible with Microsoft.NET. Has an optional, dynamically-loaded
485 LLVM code generation backend in Mini, the JIT compiler.</p>
487 <p>Note that we use a Git mirror of LLVM with some patches. See:
488 https://github.com/mono/llvm</p>
492 <!--=========================================================================-->
493 <h3>Portable OpenCL (pocl)</h3>
497 <p>Portable OpenCL is an open source implementation of the OpenCL standard which
498 can be easily adapted for new targets. One of the goals of the project is
499 improving performance portability of OpenCL programs, avoiding the need for
500 target-dependent manual optimizations. A "native" target is included, which
501 allows running OpenCL kernels on the host (CPU).</p>
505 <!--=========================================================================-->
509 <p><a href="http://pure-lang.googlecode.com/">Pure</a> is an
510 algebraic/functional programming language based on term rewriting. Programs
511 are collections of equations which are used to evaluate expressions in a
512 symbolic fashion. The interpreter uses LLVM as a backend to JIT-compile Pure
513 programs to fast native code. Pure offers dynamic typing, eager and lazy
514 evaluation, lexical closures, a hygienic macro system (also based on term
515 rewriting), built-in list and matrix support (including list and matrix
516 comprehensions) and an easy-to-use interface to C and other programming
517 languages (including the ability to load LLVM bitcode modules, and inline C,
518 C++, Fortran and Faust code in Pure programs if the corresponding LLVM-enabled
519 compilers are installed).</p>
521 <p>Pure version 0.48 has been tested and is known to work with LLVM 3.0
522 (and continues to work with older LLVM releases >= 2.5).</p>
526 <!--=========================================================================-->
527 <h3>Renderscript</h3>
531 <p><a href="http://developer.android.com/guide/topics/renderscript/index.html">Renderscript</a>
532 is Android's advanced 3D graphics rendering and compute API. It provides a
533 portable C99-based language with extensions to facilitate common use cases
534 for enhancing graphics and thread level parallelism. The Renderscript
535 compiler frontend is based on Clang/LLVM. It emits a portable bitcode format
536 for the actual compiled script code, as well as reflects a Java interface for
537 developers to control the execution of the compiled bitcode. Executable
538 machine code is then generated from this bitcode by an LLVM backend on the
539 device. Renderscript is thus able to provide a mechanism by which Android
540 developers can improve performance of their applications while retaining
545 <!--=========================================================================-->
550 <p><a href="http://safecode.cs.illinois.edu">SAFECode</a> is a memory safe C/C++
551 compiler built using LLVM. It takes standard, unannotated C/C++ code,
552 analyzes the code to ensure that memory accesses and array indexing
553 operations are safe, and instruments the code with run-time checks when
554 safety cannot be proven statically. SAFECode can be used as a debugging aid
555 (like Valgrind) to find and repair memory safety bugs. It can also be used
556 to protect code from security attacks at run-time.</p>
560 <!--=========================================================================-->
561 <h3>The Stupid D Compiler (SDC)</h3>
565 <p><a href="https://github.com/bhelyer/SDC">The Stupid D Compiler</a> is a
566 project seeking to write a self-hosting compiler for the D programming
567 language without using the frontend of the reference compiler (DMD).</p>
571 <!--=========================================================================-->
572 <h3>TTA-based Co-design Environment (TCE)</h3>
576 <p>TCE is a toolset for designing application-specific processors (ASP) based on
577 the Transport triggered architecture (TTA). The toolset provides a complete
578 co-design flow from C/C++ programs down to synthesizable VHDL and parallel
579 program binaries. Processor customization points include the register files,
580 function units, supported operations, and the interconnection network.</p>
582 <p>TCE uses Clang and LLVM for C/C++ language support, target independent
583 optimizations and also for parts of code generation. It generates new
584 LLVM-based code generators "on the fly" for the designed TTA processors and
585 loads them in to the compiler backend as runtime libraries to avoid
586 per-target recompilation of larger parts of the compiler chain.</p>
590 <!--=========================================================================-->
591 <h3>Tart Programming Language</h3>
595 <p><a href="http://code.google.com/p/tart/">Tart</a> is a general-purpose,
596 strongly typed programming language designed for application
597 developers. Strongly inspired by Python and C#, Tart focuses on practical
598 solutions for the professional software developer, while avoiding the clutter
599 and boilerplate of legacy languages like Java and C++. Although Tart is still
600 in development, the current implementation supports many features expected of
601 a modern programming language, such as garbage collection, powerful
602 bidirectional type inference, a greatly simplified syntax for template
603 metaprogramming, closures and function literals, reflection, operator
604 overloading, explicit mutability and immutability, and much more. Tart is
605 flexible enough to accommodate a broad range of programming styles and
606 philosophies, while maintaining a strong commitment to simplicity, minimalism
607 and elegance in design.</p>
611 <!--=========================================================================-->
612 <h3>ThreadSanitizer</h3>
616 <p><a href="http://code.google.com/p/data-race-test/">ThreadSanitizer</a> is a
617 data race detector for (mostly) C and C++ code, available for Linux, Mac OS
618 and Windows. On different systems, we use binary instrumentation frameworks
619 (Valgrind and Pin) as frontends that generate the program events for the race
620 detection algorithm. On Linux, there's an option of using LLVM-based
621 compile-time instrumentation.</p>
625 <!--=========================================================================-->
626 <h3>The ZooLib C++ Cross-Platform Application Framework</h3>
630 <p><a href="http://www.zoolib.org/">ZooLib</a> is Open Source under the MIT
631 License. It provides GUI, filesystem access, TCP networking, thread-safe
632 memory management, threading and locking for Mac OS X, Classic Mac OS,
633 Microsoft Windows, POSIX operating systems with X11, BeOS, Haiku, Apple's iOS
634 and Research in Motion's BlackBerry.</p>
636 <p>My current work is to use CLang's static analyzer to improve ZooLib's code
637 quality. I also plan to set up LLVM compiles of the demo programs and test
638 programs using CLang and LLVM on all the platforms that CLang, LLVM and
639 ZooLib all support.</p>
643 <!--=========================================================================-->
648 <p><a href="http://gitorious.org/pinavm/pages/Home">PinaVM</a> is an open
649 source, <a href="http://www.systemc.org/">SystemC</a> front-end. Unlike many
650 other front-ends, PinaVM actually executes the elaboration of the
651 program analyzed using LLVM's JIT infrastructure. It later enriches the
652 bitcode with SystemC-specific information.</p>
657 <!--=========================================================================-->
659 <h3 id="icedtea">IcedTea Java Virtual Machine Implementation</h3>
663 <a href="http://icedtea.classpath.org/wiki/Main_Page">IcedTea</a> provides a
664 harness to build OpenJDK using only free software build tools and to provide
665 replacements for the not-yet free parts of OpenJDK. One of the extensions that
666 IcedTea provides is a new JIT compiler named <a
667 href="http://icedtea.classpath.org/wiki/ZeroSharkFaq">Shark</a> which uses LLVM
668 to provide native code generation without introducing processor-dependent
672 <p> OpenJDK 7 b112, IcedTea6 1.9 and IcedTea7 1.13 and later have been tested
673 and are known to work with LLVM 3.0 (and continue to work with older LLVM
674 releases >= 2.6 as well).</p>
678 <!--=========================================================================-->
680 <h3>Polly - Polyhedral optimizations for LLVM</h3>
683 <p>Polly is a project that aims to provide advanced memory access optimizations
684 to better take advantage of SIMD units, cache hierarchies, multiple cores or
685 even vector accelerators for LLVM. Built around an abstract mathematical
686 description based on Z-polyhedra, it provides the infrastructure to develop
687 advanced optimizations in LLVM and to connect complex external optimizers. In
688 its first year of existence Polly already provides an exact value-based
689 dependency analysis as well as basic SIMD and OpenMP code generation support.
690 Furthermore, Polly can use PoCC(Pluto) an advanced optimizer for data-locality
695 <!--=========================================================================-->
700 <p><a href="http://github.com/evanphx/rubinius">Rubinius</a> is an environment
701 for running Ruby code which strives to write as much of the implementation in
702 Ruby as possible. Combined with a bytecode interpreting VM, it uses LLVM to
703 optimize and compile ruby code down to machine code. Techniques such as type
704 feedback, method inlining, and deoptimization are all used to remove dynamism
705 from ruby execution and increase performance.</p>
709 <!--=========================================================================-->
712 <a name="FAUST">FAUST Real-Time Audio Signal Processing Language</a>
717 <a href="http://faust.grame.fr">FAUST</a> is a compiled language for real-time
718 audio signal processing. The name FAUST stands for Functional AUdio STream. Its
719 programming model combines two approaches: functional programming and block
720 diagram composition. In addition with the C, C++, JAVA output formats, the
721 Faust compiler can now generate LLVM bitcode, and works with LLVM 2.7-3.0.</p>
728 <!-- *********************************************************************** -->
730 <a name="whatsnew">What's New in LLVM 3.0?</a>
732 <!-- *********************************************************************** -->
736 <p>This release includes a huge number of bug fixes, performance tweaks and
737 minor improvements. Some of the major improvements and new features are
738 listed in this section.</p>
740 <!--=========================================================================-->
742 <a name="majorfeatures">Major New Features</a>
747 <p>LLVM 3.0 includes several major new capabilities:</p>
759 <!--=========================================================================-->
761 <a name="coreimprovements">LLVM IR and Core Improvements</a>
766 <p>LLVM IR has several new features for better support of new targets and that
767 expose new optimization opportunities:</p>
769 <p>One of the biggest changes is that 3.0 has a new exception handling
770 system. The old system used LLVM intrinsics to convey the exception handling
771 information to the code generator. It worked in most cases, but not
772 all. Inlining was especially difficult to get right. Also, the intrinsics
773 could be moved away from the <code>invoke</code> instruction, making it hard
774 to recover that information.</p>
776 <p>The new EH system makes exception handling a first-class member of the IR. It
777 adds two new instructions:</p>
780 <li><a href="LangRef.html#i_landingpad"><code>landingpad</code></a> —
781 this instruction defines a landing pad basic block. It contains all of the
782 information that's needed by the code generator. It's also required to be
783 the first non-PHI instruction in the landing pad. In addition, a landing
784 pad may be jumped to only by the unwind edge of an <code>invoke</code>
787 <li><a href="LangRef.html#i_resume"><code>resume</code></a> — this
788 instruction causes the current exception to resume traveling up the
789 stack. It replaces the <code>@llvm.eh.resume</code> intrinsic.</li>
792 <p>Converting from the old EH API to the new EH API is rather simple, because a
793 lot of complexity has been removed. The two intrinsics,
794 <code>@llvm.eh.exception</code> and <code>@llvm.eh.selector</code> have been
795 superceded by the <code>landingpad</code> instruction. Instead of generating
796 a call to <code>@llvm.eh.exception</code> and <code>@llvm.eh.selector</code>:
798 <div class="doc_code">
800 Function *ExcIntr = Intrinsic::getDeclaration(TheModule,
801 Intrinsic::eh_exception);
802 Function *SlctrIntr = Intrinsic::getDeclaration(TheModule,
803 Intrinsic::eh_selector);
805 // The exception pointer.
806 Value *ExnPtr = Builder.CreateCall(ExcIntr, "exc_ptr");
808 std::vector<Value*> Args;
809 Args.push_back(ExnPtr);
810 Args.push_back(Builder.CreateBitCast(Personality,
811 Type::getInt8PtrTy(Context)));
813 <i>// Add selector clauses to Args.</i>
815 // The selector call.
816 Builder.CreateCall(SlctrIntr, Args, "exc_sel");
820 <p>You should instead generate a <code>landingpad</code> instruction, that
821 returns an exception object and selector value:</p>
823 <div class="doc_code">
825 LandingPadInst *LPadInst =
826 Builder.CreateLandingPad(StructType::get(Int8PtrTy, Int32Ty, NULL),
829 Value *LPadExn = Builder.CreateExtractValue(LPadInst, 0);
830 Builder.CreateStore(LPadExn, getExceptionSlot());
832 Value *LPadSel = Builder.CreateExtractValue(LPadInst, 1);
833 Builder.CreateStore(LPadSel, getEHSelectorSlot());
837 <p>It's now trivial to add the individual clauses to the <code>landingpad</code>
840 <div class="doc_code">
842 <i><b>// Adding a catch clause</b></i>
843 Constant *TypeInfo = getTypeInfo();
844 LPadInst->addClause(TypeInfo);
846 <i><b>// Adding a C++ catch-all</b></i>
847 LPadInst->addClause(Constant::getNullValue(Builder.getInt8PtrTy()));
849 <i><b>// Adding a cleanup</b></i>
850 LPadInst->setCleanup(true);
852 <i><b>// Adding a filter clause</b></i>
853 std::vector<Constant*> TypeInfos;
854 Constant *TypeInfo = getFilterTypeInfo();
855 TypeInfos.push_back(Builder.CreateBitCast(TypeInfo, Builder.getInt8PtrTy()));
857 ArrayType *FilterTy = ArrayType::get(Int8PtrTy, TypeInfos.size());
858 LPadInst->addClause(ConstantArray::get(FilterTy, TypeInfos));
862 <p>Converting from using the <code>@llvm.eh.resume</code> intrinsic to
863 the <code>resume</code> instruction is trivial. It takes the exception
864 pointer and exception selector values returned by
865 the <code>landingpad</code> instruction:</p>
867 <div class="doc_code">
869 Type *UnwindDataTy = StructType::get(Builder.getInt8PtrTy(),
870 Builder.getInt32Ty(), NULL);
871 Value *UnwindData = UndefValue::get(UnwindDataTy);
872 Value *ExcPtr = Builder.CreateLoad(getExceptionObjSlot());
873 Value *ExcSel = Builder.CreateLoad(getExceptionSelSlot());
874 UnwindData = Builder.CreateInsertValue(UnwindData, ExcPtr, 0, "exc_ptr");
875 UnwindData = Builder.CreateInsertValue(UnwindData, ExcSel, 1, "exc_sel");
876 Builder.CreateResume(UnwindData);
882 <!--=========================================================================-->
884 <a name="loopoptimization">Loop Optimization Improvements</a>
888 <p>The induction variable simplification pass in 3.0 only modifies
889 induction variables when profitable. Sign and zero extension
890 elimination, linear function test replacement, loop unrolling, and
891 other simplifications that require induction variable analysis have
892 been generalized so they no longer require loops to be rewritten in a
893 typically suboptimal form prior to optimization. This new design
894 preserves more IR level information, avoids undoing earlier loop
895 optimizations (particularly hand-optimized loops), and no longer
896 strongly depends on the code generator rewriting loops a second time
897 in a now optimal form--an intractable problem.</p>
899 <p>The original behavior can be restored with -mllvm -enable-iv-rewrite;
900 however, support for this mode will be short lived. As such, bug
901 reports should be filed for any significant performance regressions
902 when moving from -mllvm -enable-iv-rewrite to the 3.0 default mode.</p>
905 <!--=========================================================================-->
907 <a name="optimizer">Optimizer Improvements</a>
912 <p>In addition to a large array of minor performance tweaks and bug fixes, this
913 release includes a few major enhancements and additions to the
926 <!--=========================================================================-->
928 <a name="mc">MC Level Improvements</a>
933 <p>The LLVM Machine Code (aka MC) subsystem was created to solve a number of
934 problems in the realm of assembly, disassembly, object file format handling,
935 and a number of other related areas that CPU instruction-set level tools work
944 <p>For more information, please see
945 the <a href="http://blog.llvm.org/2010/04/intro-to-llvm-mc-project.html">Intro
946 to the LLVM MC Project Blog Post</a>.</p>
950 <!--=========================================================================-->
952 <a name="codegen">Target Independent Code Generator Improvements</a>
957 <p>We have put a significant amount of work into the code generator
958 infrastructure, which allows us to implement more aggressive algorithms and
959 make it run faster:</p>
968 <!--=========================================================================-->
970 <a name="x86">X86-32 and X86-64 Target Improvements</a>
975 <p>New features and major changes in the X86 target include:</p>
979 <li>The CRC32 intrinsics have been renamed. The intrinsics were previously
980 <code>@llvm.x86.sse42.crc32.[8|16|32]</code>
981 and <code>@llvm.x86.sse42.crc64.[8|64]</code>. They have been renamed to
982 <code>@llvm.x86.sse42.crc32.32.[8|16|32]</code> and
983 <code>@llvm.x86.sse42.crc32.64.[8|64]</code>.</li>
989 <!--=========================================================================-->
991 <a name="ARM">ARM Target Improvements</a>
996 <p>New features of the ARM target include:</p>
1005 <!--=========================================================================-->
1007 <a name="OtherTS">Other Target Specific Improvements</a>
1010 <p>PPC32/ELF va_arg was implemented.</p>
1011 <p>PPC32 initial support for .o file writing was implemented.</p>
1023 <!--=========================================================================-->
1025 <a name="changes">Major Changes and Removed Features</a>
1030 <p>If you're already an LLVM user or developer with out-of-tree changes based on
1031 LLVM 2.9, this section lists some "gotchas" that you may run into upgrading
1032 from the previous release.</p>
1035 <li>The <code>LLVMC</code> front end code was removed while separating
1036 out language independence.</li>
1037 <li>The <code>LowerSetJmp</code> pass wasn't used effectively by any
1038 target and has been removed.</li>
1039 <li>The old <code>TailDup</code> pass was not used in the standard pipeline
1040 and was unable to update ssa form, so it has been removed.
1041 <li>The syntax of volatile loads and stores in IR has been changed to
1042 "<code>load volatile</code>"/"<code>store volatile</code>". The old
1043 syntax ("<code>volatile load</code>"/"<code>volatile store</code>")
1044 is still accepted, but is now considered deprecated.</li>
1045 <li>The old atomic intrinscs (<code>llvm.memory.barrier</code> and
1046 <code>llvm.atomic.*</code>) are now gone. Please use the new atomic
1047 instructions, described in the <a href="Atomics.html">atomics guide</a>.
1050 <h4>Windows (32-bit)</h4>
1054 <li>On Win32(MinGW32 and MSVC), Windows 2000 will not be supported.
1055 Windows XP or higher is required.</li>
1062 <!--=========================================================================-->
1064 <a name="api_changes">Internal API Changes</a>
1069 <p>In addition, many APIs have changed in this release. Some of the major
1070 LLVM API changes are:</p>
1073 <li>The biggest and most pervasive change is that llvm::Type's are no longer
1074 returned or accepted as 'const' values. Instead, just pass around
1075 non-const Type's.</li>
1077 <li><code>PHINode::reserveOperandSpace</code> has been removed. Instead, you
1078 must specify how many operands to reserve space for when you create the
1079 PHINode, by passing an extra argument
1080 into <code>PHINode::Create</code>.</li>
1082 <li>PHINodes no longer store their incoming BasicBlocks as operands. Instead,
1083 the list of incoming BasicBlocks is stored separately, and can be accessed
1084 with new functions <code>PHINode::block_begin</code>
1085 and <code>PHINode::block_end</code>.</li>
1087 <li>Various functions now take an <code>ArrayRef</code> instead of either a
1088 pair of pointers (or iterators) to the beginning and end of a range, or a
1089 pointer and a length. Others now return an <code>ArrayRef</code> instead
1090 of a reference to a <code>SmallVector</code>
1091 or <code>std::vector</code>. These include:
1093 <!-- Please keep this list sorted. -->
1094 <li><code>CallInst::Create</code></li>
1095 <li><code>ComputeLinearIndex</code> (in <code>llvm/CodeGen/Analysis.h</code>)</li>
1096 <li><code>ConstantArray::get</code></li>
1097 <li><code>ConstantExpr::getExtractElement</code></li>
1098 <li><code>ConstantExpr::getGetElementPtr</code></li>
1099 <li><code>ConstantExpr::getInBoundsGetElementPtr</code></li>
1100 <li><code>ConstantExpr::getIndices</code></li>
1101 <li><code>ConstantExpr::getInsertElement</code></li>
1102 <li><code>ConstantExpr::getWithOperands</code></li>
1103 <li><code>ConstantFoldCall</code> (in <code>llvm/Analysis/ConstantFolding.h</code>)</li>
1104 <li><code>ConstantFoldInstOperands</code> (in <code>llvm/Analysis/ConstantFolding.h</code>)</li>
1105 <li><code>ConstantVector::get</code></li>
1106 <li><code>DIBuilder::createComplexVariable</code></li>
1107 <li><code>DIBuilder::getOrCreateArray</code></li>
1108 <li><code>ExtractValueInst::Create</code></li>
1109 <li><code>ExtractValueInst::getIndexedType</code></li>
1110 <li><code>ExtractValueInst::getIndices</code></li>
1111 <li><code>FindInsertedValue</code> (in <code>llvm/Analysis/ValueTracking.h</code>)</li>
1112 <li><code>gep_type_begin</code> (in <code>llvm/Support/GetElementPtrTypeIterator.h</code>)</li>
1113 <li><code>gep_type_end</code> (in <code>llvm/Support/GetElementPtrTypeIterator.h</code>)</li>
1114 <li><code>GetElementPtrInst::Create</code></li>
1115 <li><code>GetElementPtrInst::CreateInBounds</code></li>
1116 <li><code>GetElementPtrInst::getIndexedType</code></li>
1117 <li><code>InsertValueInst::Create</code></li>
1118 <li><code>InsertValueInst::getIndices</code></li>
1119 <li><code>InvokeInst::Create</code></li>
1120 <li><code>IRBuilder::CreateCall</code></li>
1121 <li><code>IRBuilder::CreateExtractValue</code></li>
1122 <li><code>IRBuilder::CreateGEP</code></li>
1123 <li><code>IRBuilder::CreateInBoundsGEP</code></li>
1124 <li><code>IRBuilder::CreateInsertValue</code></li>
1125 <li><code>IRBuilder::CreateInvoke</code></li>
1126 <li><code>MDNode::get</code></li>
1127 <li><code>MDNode::getIfExists</code></li>
1128 <li><code>MDNode::getTemporary</code></li>
1129 <li><code>MDNode::getWhenValsUnresolved</code></li>
1130 <li><code>SimplifyGEPInst</code> (in <code>llvm/Analysis/InstructionSimplify.h</code>)</li>
1131 <li><code>TargetData::getIndexedOffset</code></li>
1134 <li>All forms of <code>StringMap::getOrCreateValue</code> have been remove
1135 except for the one which takes a <code>StringRef</code>.</li>
1137 <li>The <code>LLVMBuildUnwind</code> function from the C API was removed. The
1138 LLVM <code>unwind</code> instruction has been deprecated for a long time
1139 and isn't used by the current front-ends. So this was removed during the
1140 exception handling rewrite.</li>
1142 <li>The <code>LLVMAddLowerSetJmpPass</code> function from the C API was
1143 removed because the <code>LowerSetJmp</code> pass was removed.</li>
1145 <li>The <code>DIBuilder</code> interface used by front ends to encode
1146 debugging information in the LLVM IR now expects clients to
1147 use <code>DIBuilder::finalize()</code> at the end of translation unit to
1148 complete debugging information encoding.</li>
1150 <li>The way the type system works has been
1151 rewritten: <code>PATypeHolder</code> and <code>OpaqueType</code> are gone,
1152 and all APIs deal with <code>Type*</code> instead of <code>const
1153 Type*</code>. If you need to create recursive structures, then create a
1154 named structure, and use <code>setBody()</code> when all its elements are
1155 built. Type merging and refining is gone too: named structures are not
1156 merged with other structures, even if their layout is identical. (of
1157 course anonymous structures are still uniqued by layout).</li>
1159 <li>TargetSelect.h moved to Support/ from Target/</li>
1161 <li>UpgradeIntrinsicCall no longer upgrades pre-2.9 intrinsic calls (for
1162 example <code>llvm.memset.i32</code>).</li>
1164 <li>It is mandatory to initialize all out-of-tree passes too and their dependencies now with
1165 <code>INITIALIZE_PASS{BEGIN,END,}</code>
1166 and <code>INITIALIZE_{PASS,AG}_DEPENDENCY</code>.</li>
1168 <li>The interface for MemDepResult in MemoryDependenceAnalysis has been
1169 enhanced with new return types Unknown and NonFuncLocal, in addition to
1170 the existing types Clobber, Def, and NonLocal.</li>
1177 <!-- *********************************************************************** -->
1179 <a name="knownproblems">Known Problems</a>
1181 <!-- *********************************************************************** -->
1185 <p>This section contains significant known problems with the LLVM system, listed
1186 by component. If you run into a problem, please check
1187 the <a href="http://llvm.org/bugs/">LLVM bug database</a> and submit a bug if
1188 there isn't already one.</p>
1190 <!-- ======================================================================= -->
1192 <a name="experimental">Experimental features included with this release</a>
1197 <p>The following components of this LLVM release are either untested, known to
1198 be broken or unreliable, or are in early development. These components
1199 should not be relied on, and bugs should not be filed against them, but they
1200 may be useful to some people. In particular, if you would like to work on
1201 one of these components, please contact us on
1202 the <a href="http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev">LLVMdev
1206 <li>The Alpha, Blackfin, CellSPU, MicroBlaze, MSP430, MIPS, PTX, SystemZ and
1207 XCore backends are experimental.</li>
1209 <li><tt>llc</tt> "<tt>-filetype=obj</tt>" is experimental on all targets other
1210 than darwin and ELF X86 systems.</li>
1215 <!-- ======================================================================= -->
1217 <a name="x86-be">Known problems with the X86 back-end</a>
1223 <li>The X86 backend does not yet support
1224 all <a href="http://llvm.org/PR879">inline assembly that uses the X86
1225 floating point stack</a>. It supports the 'f' and 't' constraints, but
1228 <li>The X86-64 backend does not yet support the LLVM IR instruction
1229 <tt>va_arg</tt>. Currently, front-ends support variadic argument
1230 constructs on X86-64 by lowering them manually.</li>
1232 <li>Windows x64 (aka Win64) code generator has a few issues.
1234 <li>llvm-gcc cannot build the mingw-w64 runtime currently due to lack of
1235 support for the 'u' inline assembly constraint and for X87 floating
1236 point inline assembly.</li>
1238 <li>On mingw-w64, you will see unresolved symbol <tt>__chkstk</tt> due
1239 to <a href="http://llvm.org/bugs/show_bug.cgi?id=8919">Bug 8919</a>.
1241 in <a href="http://lists.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20110321/118499.html">r128206</a>.</li>
1243 <li>Miss-aligned MOVDQA might crash your program. It is due to
1244 <a href="http://llvm.org/bugs/show_bug.cgi?id=9483">Bug 9483</a>, lack
1245 of handling aligned internal globals.</li>
1253 <!-- ======================================================================= -->
1255 <a name="ppc-be">Known problems with the PowerPC back-end</a>
1261 <li>The PPC32/ELF support lacks PIC support.</li>
1266 <!-- ======================================================================= -->
1268 <a name="arm-be">Known problems with the ARM back-end</a>
1274 <li>Thumb mode works only on ARMv6 or higher processors. On sub-ARMv6
1275 processors, thumb programs can crash or produce wrong results
1276 (<a href="http://llvm.org/PR1388">PR1388</a>).</li>
1278 <li>Compilation for ARM Linux OABI (old ABI) is supported but not fully
1284 <!-- ======================================================================= -->
1286 <a name="sparc-be">Known problems with the SPARC back-end</a>
1292 <li>The SPARC backend only supports the 32-bit SPARC ABI (-m32); it does not
1293 support the 64-bit SPARC ABI (-m64).</li>
1298 <!-- ======================================================================= -->
1300 <a name="mips-be">Known problems with the MIPS back-end</a>
1306 <li>64-bit MIPS targets are not supported yet.</li>
1311 <!-- ======================================================================= -->
1313 <a name="alpha-be">Known problems with the Alpha back-end</a>
1319 <li>On 21164s, some rare FP arithmetic sequences which may trap do not have
1320 the appropriate nops inserted to ensure restartability.</li>
1325 <!-- ======================================================================= -->
1327 <a name="c-be">Known problems with the C back-end</a>
1332 <p>The C backend has numerous problems and is not being actively maintained.
1333 Depending on it for anything serious is not advised.</p>
1336 <li><a href="http://llvm.org/PR802">The C backend has only basic support for
1337 inline assembly code</a>.</li>
1339 <li><a href="http://llvm.org/PR1658">The C backend violates the ABI of common
1340 C++ programs</a>, preventing intermixing between C++ compiled by the CBE
1341 and C++ code compiled with <tt>llc</tt> or native compilers.</li>
1343 <li>The C backend does not support all exception handling constructs.</li>
1345 <li>The C backend does not support arbitrary precision integers.</li>
1351 <!-- ======================================================================= -->
1353 <a name="llvm-gcc">Known problems with the llvm-gcc front-end</a>
1358 <p><b>LLVM 2.9 was the last release of llvm-gcc.</b></p>
1360 <p>llvm-gcc is generally very stable for the C family of languages. The only
1361 major language feature of GCC not supported by llvm-gcc is the
1362 <tt>__builtin_apply</tt> family of builtins. However, some extensions
1363 are only supported on some targets. For example, trampolines are only
1364 supported on some targets (these are used when you take the address of a
1365 nested function).</p>
1367 <p>Fortran support generally works, but there are still several unresolved bugs
1368 in <a href="http://llvm.org/bugs/">Bugzilla</a>. Please see the
1369 tools/gfortran component for details. Note that llvm-gcc is missing major
1370 Fortran performance work in the frontend and library that went into GCC after
1371 4.2. If you are interested in Fortran, we recommend that you consider using
1372 <a href="#dragonegg">dragonegg</a> instead.</p>
1374 <p>The llvm-gcc 4.2 Ada compiler has basic functionality, but is no longer being
1375 actively maintained. If you are interested in Ada, we recommend that you
1376 consider using <a href="#dragonegg">dragonegg</a> instead.</p>
1382 <!-- *********************************************************************** -->
1384 <a name="additionalinfo">Additional Information</a>
1386 <!-- *********************************************************************** -->
1390 <p>A wide variety of additional information is available on
1391 the <a href="http://llvm.org/">LLVM web page</a>, in particular in
1392 the <a href="http://llvm.org/docs/">documentation</a> section. The web page
1393 also contains versions of the API documentation which is up-to-date with the
1394 Subversion version of the source code. You can access versions of these
1395 documents specific to this release by going into the "<tt>llvm/doc/</tt>"
1396 directory in the LLVM tree.</p>
1398 <p>If you have any questions or comments about LLVM, please feel free to contact
1399 us via the <a href="http://llvm.org/docs/#maillist"> mailing lists</a>.</p>
1403 <!-- *********************************************************************** -->
1407 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
1408 src="http://jigsaw.w3.org/css-validator/images/vcss-blue" alt="Valid CSS"></a>
1409 <a href="http://validator.w3.org/check/referer"><img
1410 src="http://www.w3.org/Icons/valid-html401-blue" alt="Valid HTML 4.01"></a>
1412 <a href="http://llvm.org/">LLVM Compiler Infrastructure</a><br>
1413 Last modified: $Date$