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, improvements in various
49 subprojects of LLVM, and some of the current users of the code.
50 All LLVM releases may be downloaded from
51 the <a href="http://llvm.org/releases/">LLVM releases web site</a>.</p>
53 <p>For more information about LLVM, including information about the latest
54 release, please check out the <a href="http://llvm.org/">main LLVM web
55 site</a>. If you have questions or comments,
56 the <a href="http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev">LLVM
57 Developer's Mailing List</a> is a good place to send them.</p>
59 <p>Note that if you are reading this file from a Subversion checkout or the main
60 LLVM web page, this document applies to the <i>next</i> release, not the
61 current one. To see the release notes for a specific release, please see the
62 <a href="http://llvm.org/releases/">releases page</a>.</p>
67 <!-- *********************************************************************** -->
69 <a name="subproj">Sub-project Status Update</a>
71 <!-- *********************************************************************** -->
75 <p>The LLVM 3.0 distribution currently consists of code from the core LLVM
76 repository (which roughly includes the LLVM optimizers, code generators and
77 supporting tools), and the Clang repository. In
78 addition to this code, the LLVM Project includes other sub-projects that are
79 in development. Here we include updates on these subprojects.</p>
81 <!--=========================================================================-->
83 <a name="clang">Clang: C/C++/Objective-C Frontend Toolkit</a>
88 <p><a href="http://clang.llvm.org/">Clang</a> is an LLVM front end for the C,
89 C++, and Objective-C languages. Clang aims to provide a better user
90 experience through expressive diagnostics, a high level of conformance to
91 language standards, fast compilation, and low memory use. Like LLVM, Clang
92 provides a modular, library-based architecture that makes it suitable for
93 creating or integrating with other development tools. Clang is considered a
94 production-quality compiler for C, Objective-C, C++ and Objective-C++ on x86
95 (32- and 64-bit), and for Darwin/ARM targets.</p>
97 <p>In the LLVM 3.0 time-frame, the Clang team has made many improvements:</p>
100 <li>Greatly improved support for building C++ applications, with greater
101 stability and better diagnostics.</li>
103 <li><a href="http://clang.llvm.org/cxx_status.html">Improved support</a> for
104 the <a href="http://www.iso.org/iso/iso_catalogue/catalogue_tc/catalogue_detail.htm?csnumber=50372">C++
105 2011</a> standard (aka "C++'0x"), including implementations of non-static data member
106 initializers, alias templates, delegating constructors, range-based
107 for loops, and implicitly-generated move constructors and move assignment
108 operators, among others.</li>
110 <li>Implemented support for some features of the upcoming C1x standard,
111 including static assertions and generic selections.</li>
113 <li>Better detection of include and linking paths for system headers and
114 libraries, especially for Linux distributions.</li>
116 <li>Several improvements to Objective-C support, including:
119 <li><a href="http://clang.llvm.org/docs/AutomaticReferenceCounting.html">
120 Automatic Reference Counting</a> (ARC) and an improved memory model
121 cleanly separating object and C memory.</li>
123 <li>A migration tool for moving manual retain/release code to ARC</li>
125 <li>Better support for data hiding, allowing instance variables to be
126 declared in implementation contexts or class extensions</li>
127 <li>Weak linking support for Objective-C classes</li>
128 <li>Improved static type checking by inferring the return type of methods
129 such as +alloc and -init.</li>
132 Some new Objective-C features require either the Mac OS X 10.7 / iOS 5
133 Objective-C runtime, or version 1.6 or later of the GNUstep Objective-C
134 runtime version.</li>
136 <li>Implemented a number of optimizations in <tt>libclang</tt>, the Clang C
137 interface, to improve the performance of code completion and the mapping
138 from source locations to abstract syntax tree nodes.</li>
142 <p>If Clang rejects your code but another compiler accepts it, please take a
143 look at the <a href="http://clang.llvm.org/compatibility.html">language
144 compatibility</a> guide to make sure this is not intentional or a known
149 <!--=========================================================================-->
151 <a name="dragonegg">DragonEgg: GCC front-ends, LLVM back-end</a>
155 <p><a href="http://dragonegg.llvm.org/">DragonEgg</a> is a
156 <a href="http://gcc.gnu.org/wiki/plugins">gcc plugin</a> that replaces GCC's
157 optimizers and code generators with LLVM's. It works with gcc-4.5 or gcc-4.6,
158 targets the x86-32 and x86-64 processor families, and has been successfully
159 used on the Darwin, FreeBSD, KFreeBSD, Linux and OpenBSD platforms. It fully
160 supports Ada, C, C++ and Fortran. It has partial support for Go, Java, Obj-C
163 <p>The 3.0 release has the following notable changes:</p>
166 <li>GCC version 4.6 is now fully supported.</li>
168 <li>Patching and building GCC is no longer required: the plugin should work
169 with your system GCC (version 4.5 or 4.6; on Debian/Ubuntu systems the
170 gcc-4.5-plugin-dev or gcc-4.6-plugin-dev package is also needed).</li>
172 <li>The <tt>-fplugin-arg-dragonegg-enable-gcc-optzns</tt> option, which runs
173 GCC's optimizers as well as LLVM's, now works much better. This is the
174 option to use if you want ultimate performance! It is still experimental
175 though: it may cause the plugin to crash.</li>
177 <li>The type and constant conversion logic has been almost entirely rewritten,
178 fixing a multitude of obscure bugs.</li>
184 <!--=========================================================================-->
186 <a name="compiler-rt">compiler-rt: Compiler Runtime Library</a>
191 <p>The new LLVM <a href="http://compiler-rt.llvm.org/">compiler-rt project</a>
192 is a simple library that provides an implementation of the low-level
193 target-specific hooks required by code generation and other runtime
194 components. For example, when compiling for a 32-bit target, converting a
195 double to a 64-bit unsigned integer is compiled into a runtime call to the
196 "__fixunsdfdi" function. The compiler-rt library provides highly optimized
197 implementations of this and other low-level routines (some are 3x faster than
198 the equivalent libgcc routines).</p>
200 <p>In the LLVM 3.0 timeframe, the target specific ARM code has converted to
201 "unified" assembly syntax, and several new functions have been added to the
206 <!--=========================================================================-->
208 <a name="lldb">LLDB: Low Level Debugger</a>
213 <p>LLDB has advanced by leaps and bounds in the 3.0 timeframe. It is
214 dramatically more stable and useful, and includes both a
215 new <a href="http://lldb.llvm.org/tutorial.html">tutorial</a> and
216 a <a href="http://lldb.llvm.org/lldb-gdb.html">side-by-side comparison with
221 <!--=========================================================================-->
223 <a name="libc++">libc++: C++ Standard Library</a>
228 <p>Like compiler_rt, libc++ is now <a href="DeveloperPolicy.html#license">dual
229 licensed</a> under the MIT and UIUC license, allowing it to be used more
232 <p>Libc++ has been ported to FreeBSD and imported into the base system. It is
233 planned to be the default STL implementation for FreeBSD 10.</p>
238 <!--=========================================================================-->
240 <a name="LLBrowse">LLBrowse: IR Browser</a>
245 <p><a href="http://llvm.org/svn/llvm-project/llbrowse/trunk/doc/LLBrowse.html">
246 LLBrowse</a> is an interactive viewer for LLVM modules. It can load any LLVM
247 module and displays its contents as an expandable tree view, facilitating an
248 easy way to inspect types, functions, global variables, or metadata nodes. It
249 is fully cross-platform, being based on the popular wxWidgets GUI
254 <!--=========================================================================-->
256 <a name="vmkit">VMKit</a>
261 <p>The <a href="http://vmkit.llvm.org/">VMKit project</a> is an
262 implementation of a Java Virtual Machine (Java VM or JVM) that uses LLVM for
263 static and just-in-time compilation.
265 <p>In the LLVM 3.0 time-frame, VMKit has had significant improvements on both
266 runtime and startup performance:</p>
269 <li>Precompilation: by compiling ahead of time a small subset of Java's core
270 library, the startup performance have been highly optimized to the point that
271 running a 'Hello World' program takes less than 30 milliseconds.</li>
273 <li>Customization: by customizing virtual methods for individual classes,
274 the VM can statically determine the target of a virtual call, and decide to
277 <li>Inlining: the VM does more inlining than it did before, by allowing more
278 bytecode instructions to be inlined, and thanks to customization. It also
279 inlines GC barriers, and object allocations.</li>
281 <li>New exception model: the generated code for a method that does not do
282 any try/catch is not penalized anymore by the eventuality of calling a
283 method that throws an exception. Instead, the method that throws the
284 exception jumps directly to the method that could catch it.</li>
290 <!--=========================================================================-->
293 <a name="klee">KLEE: A Symbolic Execution Virtual Machine</a>
298 <a href="http://klee.llvm.org/">KLEE</a> is a symbolic execution framework for
299 programs in LLVM bitcode form. KLEE tries to symbolically evaluate "all" paths
300 through the application and records state transitions that lead to fault
301 states. This allows it to construct testcases that lead to faults and can even
302 be used to verify some algorithms.
310 <!-- *********************************************************************** -->
312 <a name="externalproj">External Open Source Projects Using LLVM 3.0</a>
314 <!-- *********************************************************************** -->
318 <p>An exciting aspect of LLVM is that it is used as an enabling technology for
319 a lot of other language and tools projects. This section lists some of the
320 projects that have already been updated to work with LLVM 3.0.</p>
322 <!--=========================================================================-->
323 <h3>AddressSanitizer</h3>
327 <p><a href="http://code.google.com/p/address-sanitizer/">AddressSanitizer</a>
328 uses compiler instrumentation and a specialized malloc library to find C/C++
329 bugs such as use-after-free and out-of-bound accesses to heap, stack, and
330 globals. The key feature of the tool is speed: the average slowdown
331 introduced by AddressSanitizer is less than 2x.</p>
335 <!--=========================================================================-->
340 <p><a href="http://www.clamav.net">Clam AntiVirus</a> is an open source (GPL)
341 anti-virus toolkit for UNIX, designed especially for e-mail scanning on mail
344 <p>Since version 0.96 it
345 has <a href="http://vrt-sourcefire.blogspot.com/2010/09/introduction-to-clamavs-low-level.html">bytecode
346 signatures</a> that allow writing detections for complex malware.</p>
348 <p>It uses LLVM's JIT to speed up the execution of bytecode on X86, X86-64,
349 PPC32/64, falling back to its own interpreter otherwise. The git version was
350 updated to work with LLVM 3.0.</p>
354 <!--=========================================================================-->
355 <h3>clang_complete for VIM</h3>
359 <p><a href="https://github.com/Rip-Rip/clang_complete">clang_complete</a> is a
360 VIM plugin, that provides accurate C/C++ autocompletion using the clang front
361 end. The development version of clang complete, can directly use libclang
362 which can maintain a cache to speed up auto completion.</p>
366 <!--=========================================================================-->
371 <p><a href="https://bitbucket.org/dwilliamson/clreflect">clReflect</a> is a C++
372 parser that uses clang/LLVM to derive a light-weight reflection database
373 suitable for use in game development. It comes with a very simple runtime
374 library for loading and querying the database, requiring no external
375 dependencies (including CRT), and an additional utility library for object
376 management and serialisation.</p>
380 <!--=========================================================================-->
381 <h3>Cling C++ Interpreter</h3>
385 <p><a href="http://cern.ch/cling">Cling</a> is an interactive compiler interface
386 (aka C++ interpreter). It uses LLVM's JIT and clang; it currently supports
387 C++ and C. It has a prompt interface, runs source files, calls into shared
388 libraries, prints the value of expressions, even does runtime lookup of
389 identifiers (dynamic scopes). And it just behaves like one would expect from
394 <!--=========================================================================-->
395 <h3>Crack Programming Language</h3>
399 <p><a href="http://code.google.com/p/crack-language/">Crack</a> aims to provide
400 the ease of development of a scripting language with the performance of a
401 compiled language. The language derives concepts from C++, Java and Python,
402 incorporating object-oriented programming, operator overloading and strong
407 <!--=========================================================================-->
412 <p><a href="http://eerolanguage.org/">Eero</a> is a fully
413 header-and-binary-compatible dialect of Objective-C 2.0, implemented with a
414 patched version of the Clang/LLVM compiler. It features a streamlined syntax,
415 Python-like indentation, and new operators, for improved readability and
416 reduced code clutter. It also has new features such as limited forms of
417 operator overloading and namespaces, and strict (type-and-operator-safe)
418 enumerations. It is inspired by languages such as Smalltalk, Python, and
423 <!--=========================================================================-->
424 <h3>FAUST Real-Time Audio Signal Processing Language</h3>
428 <p><a href="http://faust.grame.fr/">FAUST</a> is a compiled language for
429 real-time audio signal processing. The name FAUST stands for Functional
430 AUdio STream. Its programming model combines two approaches: functional
431 programming and block diagram composition. In addition with the C, C++, Java
432 output formats, the Faust compiler can now generate LLVM bitcode, and works
438 <!--=========================================================================-->
439 <h3>Glasgow Haskell Compiler (GHC)</h3>
443 <p>GHC is an open source, state-of-the-art programming suite for Haskell, a
444 standard lazy functional programming language. It includes an optimizing
445 static compiler generating good code for a variety of platforms, together
446 with an interactive system for convenient, quick development.</p>
448 <p>GHC 7.0 and onwards include an LLVM code generator, supporting LLVM 2.8 and
449 later. Since LLVM 2.9, GHC now includes experimental support for the ARM
450 platform with LLVM 3.0.</p>
454 <!--=========================================================================-->
459 <p><a href="http://botwars.tk/gwscript/">gwXscript</a> is an object oriented,
460 aspect oriented programming language which can create both executables (ELF,
461 EXE) and shared libraries (DLL, SO, DYNLIB). The compiler is implemented in
462 its own language and translates scripts into LLVM-IR which can be optimized
463 and translated into native code by the LLVM framework. Source code in
464 gwScript contains definitions that expand the namespaces. So you can build
465 your project and simply 'plug out' features by removing a file. The remaining
466 project does not leave scars since you directly separate concerns by the
467 'template' feature of gwX. It is also possible to add new features to a
468 project by just adding files and without editing the original project. This
469 language is used for example to create games or content management systems
470 that should be extendable.</p>
472 <p>gwXscript is strongly typed and offers comfort with its native types string,
473 hash and array. You can easily write new libraries in gwXscript or native
474 code. gwXscript is type safe and users should not be able to crash your
475 program or execute malicious code except code that is eating CPU time.</p>
479 <!--=========================================================================-->
480 <h3>include-what-you-use</h3>
484 <p><a href="http://code.google.com/p/include-what-you-use">include-what-you-use</a>
485 is a tool to ensure that a file directly <code>#include</code>s
486 all <code>.h</code> files that provide a symbol that the file uses. It also
487 removes superfluous <code>#include</code>s from source files.</p>
491 <!--=========================================================================-->
492 <h3>ispc: The Intel SPMD Program Compiler</h3>
496 <p><a href="http://ispc.github.com">ispc</a> is a compiler for "single program,
497 multiple data" (SPMD) programs. It compiles a C-based SPMD programming
498 language to run on the SIMD units of CPUs; it often delivers 5-6x speedups on
499 a single core of a CPU with an 8-wide SIMD unit compared to serial code,
500 while still providing a clean and easy-to-understand programming model. For
501 an introduction to the language and its performance,
502 see <a href="http://ispc.github.com/example.html">the walkthrough</a> of a short
503 example program. ispc is licensed under the BSD license.</p>
507 <!--=========================================================================-->
508 <h3>The Julia Programming Language</h3>
512 <p><a href="http://github.com/JuliaLang/julia">Julia</a> is a high-level,
513 high-performance dynamic language for technical
514 computing. It provides a sophisticated compiler, distributed parallel
515 execution, numerical accuracy, and an extensive mathematical function
516 library. The compiler uses type inference to generate fast code
517 without any type declarations, and uses LLVM's optimization passes and
518 JIT compiler. The language is designed around multiple dispatch,
519 giving programs a large degree of flexibility. It is ready for use on many
520 kinds of problems.</p>
523 <!--=========================================================================-->
524 <h3>LanguageKit and Pragmatic Smalltalk</h3>
528 <p><a href="http://etoileos.com/etoile/features/languagekit/">LanguageKit</a> is
529 a framework for implementing dynamic languages sharing an object model with
530 Objective-C. It provides static and JIT compilation using LLVM along with
531 its own interpreter. Pragmatic Smalltalk is a dialect of Smalltalk, built on
532 top of LanguageKit, that interfaces directly with Objective-C, sharing the
533 same object representation and message sending behaviour. These projects are
534 developed as part of the Étoilé desktop environment.</p>
538 <!--=========================================================================-->
543 <p><a href="http://lua-av.mat.ucsb.edu/blog/">LuaAV</a> is a real-time
544 audiovisual scripting environment based around the Lua language and a
545 collection of libraries for sound, graphics, and other media protocols. LuaAV
546 uses LLVM and Clang to JIT compile efficient user-defined audio synthesis
547 routines specified in a declarative syntax.</p>
551 <!--=========================================================================-->
556 <p>An open source, cross-platform implementation of C# and the CLR that is
557 binary compatible with Microsoft.NET. Has an optional, dynamically-loaded
558 LLVM code generation backend in Mini, the JIT compiler.</p>
560 <p>Note that we use a Git mirror of LLVM with some patches. See:
561 https://github.com/mono/llvm</p>
565 <!--=========================================================================-->
570 <p><a href="http://polly.grosser.es">Polly</a> is an advanced data-locality
571 optimizer and automatic parallelizer. It uses an advanced, mathematical
572 model to calculate detailed data dependency information which it uses to
573 optimize the loop structure of a program. Polly can speed up sequential code
574 by improving memory locality and consequently the cache use. Furthermore,
575 Polly is able to expose different kind of parallelism which it exploits by
576 introducing (basic) OpenMP and SIMD code. A mid-term goal of Polly is to
577 automatically create optimized GPU code.</p>
581 <!--=========================================================================-->
582 <h3>Portable OpenCL (pocl)</h3>
586 <p>Portable OpenCL is an open source implementation of the OpenCL standard which
587 can be easily adapted for new targets. One of the goals of the project is
588 improving performance portability of OpenCL programs, avoiding the need for
589 target-dependent manual optimizations. A "native" target is included, which
590 allows running OpenCL kernels on the host (CPU).</p>
594 <!--=========================================================================-->
598 <p><a href="http://pure-lang.googlecode.com/">Pure</a> is an
599 algebraic/functional programming language based on term rewriting. Programs
600 are collections of equations which are used to evaluate expressions in a
601 symbolic fashion. The interpreter uses LLVM as a backend to JIT-compile Pure
602 programs to fast native code. Pure offers dynamic typing, eager and lazy
603 evaluation, lexical closures, a hygienic macro system (also based on term
604 rewriting), built-in list and matrix support (including list and matrix
605 comprehensions) and an easy-to-use interface to C and other programming
606 languages (including the ability to load LLVM bitcode modules, and inline C,
607 C++, Fortran and Faust code in Pure programs if the corresponding LLVM-enabled
608 compilers are installed).</p>
610 <p>Pure version 0.48 has been tested and is known to work with LLVM 3.0
611 (and continues to work with older LLVM releases >= 2.5).</p>
615 <!--=========================================================================-->
616 <h3>Renderscript</h3>
620 <p><a href="http://developer.android.com/guide/topics/renderscript/index.html">Renderscript</a>
621 is Android's advanced 3D graphics rendering and compute API. It provides a
622 portable C99-based language with extensions to facilitate common use cases
623 for enhancing graphics and thread level parallelism. The Renderscript
624 compiler frontend is based on Clang/LLVM. It emits a portable bitcode format
625 for the actual compiled script code, as well as reflects a Java interface for
626 developers to control the execution of the compiled bitcode. Executable
627 machine code is then generated from this bitcode by an LLVM backend on the
628 device. Renderscript is thus able to provide a mechanism by which Android
629 developers can improve performance of their applications while retaining
634 <!--=========================================================================-->
639 <p><a href="http://safecode.cs.illinois.edu">SAFECode</a> is a memory safe C/C++
640 compiler built using LLVM. It takes standard, unannotated C/C++ code,
641 analyzes the code to ensure that memory accesses and array indexing
642 operations are safe, and instruments the code with run-time checks when
643 safety cannot be proven statically. SAFECode can be used as a debugging aid
644 (like Valgrind) to find and repair memory safety bugs. It can also be used
645 to protect code from security attacks at run-time.</p>
649 <!--=========================================================================-->
650 <h3>The Stupid D Compiler (SDC)</h3>
654 <p><a href="https://github.com/bhelyer/SDC">The Stupid D Compiler</a> is a
655 project seeking to write a self-hosting compiler for the D programming
656 language without using the frontend of the reference compiler (DMD).</p>
660 <!--=========================================================================-->
661 <h3>TTA-based Co-design Environment (TCE)</h3>
665 <p>TCE is a toolset for designing application-specific processors (ASP) based on
666 the Transport triggered architecture (TTA). The toolset provides a complete
667 co-design flow from C/C++ programs down to synthesizable VHDL and parallel
668 program binaries. Processor customization points include the register files,
669 function units, supported operations, and the interconnection network.</p>
671 <p>TCE uses Clang and LLVM for C/C++ language support, target independent
672 optimizations and also for parts of code generation. It generates new
673 LLVM-based code generators "on the fly" for the designed TTA processors and
674 loads them in to the compiler backend as runtime libraries to avoid
675 per-target recompilation of larger parts of the compiler chain.</p>
679 <!--=========================================================================-->
680 <h3>Tart Programming Language</h3>
684 <p><a href="http://code.google.com/p/tart/">Tart</a> is a general-purpose,
685 strongly typed programming language designed for application
686 developers. Strongly inspired by Python and C#, Tart focuses on practical
687 solutions for the professional software developer, while avoiding the clutter
688 and boilerplate of legacy languages like Java and C++. Although Tart is still
689 in development, the current implementation supports many features expected of
690 a modern programming language, such as garbage collection, powerful
691 bidirectional type inference, a greatly simplified syntax for template
692 metaprogramming, closures and function literals, reflection, operator
693 overloading, explicit mutability and immutability, and much more. Tart is
694 flexible enough to accommodate a broad range of programming styles and
695 philosophies, while maintaining a strong commitment to simplicity, minimalism
696 and elegance in design.</p>
700 <!--=========================================================================-->
701 <h3>ThreadSanitizer</h3>
705 <p><a href="http://code.google.com/p/data-race-test/">ThreadSanitizer</a> is a
706 data race detector for (mostly) C and C++ code, available for Linux, Mac OS
707 and Windows. On different systems, we use binary instrumentation frameworks
708 (Valgrind and Pin) as frontends that generate the program events for the race
709 detection algorithm. On Linux, there's an option of using LLVM-based
710 compile-time instrumentation.</p>
716 <!-- *********************************************************************** -->
718 <a name="whatsnew">What's New in LLVM 3.0?</a>
720 <!-- *********************************************************************** -->
724 <p>This release includes a huge number of bug fixes, performance tweaks and
725 minor improvements. Some of the major improvements and new features are
726 listed in this section.</p>
728 <!--=========================================================================-->
730 <a name="majorfeatures">Major New Features</a>
735 <!-- Features that need text if they're finished for 3.1:
739 loop dependence analysis
740 CorrelatedValuePropagation
741 lib/Transforms/IPO/MergeFunctions.cpp => consider for 3.1.
744 <p><b>llvm-gcc is gone</b>. LLVM's configure script doesn't depend on llvm-gcc anymore, clean layering.</p>
746 <p>LLVM 3.0 includes several major new capabilities:</p>
749 Analysis/RegionInfo.h + Dom Frontiers
750 SparseBitVector: used in LiveVar.
751 llvm/lib/Archive - replace with lib object?
755 Type system rewrite: http://blog.llvm.org/2011/11/llvm-30-type-system-rewrite.html
756 Better performance for Neon code in clang due to SRoA improvements.
757 New regalloc on by default. Lin scan going away in 3.1
758 PGO / builtin_expect improvements (summary needed)
760 AVX support, assembler, compiler and disassembler.
761 IndVar improvements: andy
762 PTX backend improvements: Justin
763 llvm-rtdyld & MC JIT: JimG
764 InstAliases now automatically used in the asmprinter where they are shorter.
765 Integrated assembler on by default for arm/thumb?
766 PostOrder Dominator frontiers were removed.
767 Line Profiling / gcov support
768 EH and debug information produced with CFI directives, yielding smaller executables: http://blog.mozilla.com/respindola/2011/05/12/cfi-directives/
769 X86-64 generates smaller and faster code at -O0 (fast isel improvements)
770 Better code generation for Cortex-A9
771 Many APIs take ArrayRef's now.
772 Pass manager extension API.
773 ARM inline asm constraints implemented.
774 LangRef.html#fnattrs uwtable attribute for asynch unwind tables.
775 better performance for indirect gotos.
776 llvm.prefetch now takes a 4th argument that specifies whether the prefetch happens to the icache or dcache.
777 New PackedVector, TinyPtrVector class (see Programmer's Manual)
778 New nonlazybind function attribute.
779 ARC language specific optimizer (Transforms/ObjCARC) a decent example of language-specific transformation.
780 LLVM 3.0 removes support for reading LLVM 2.8 and earlier files. Aim to maintain compatibility all the way back to 3.0 "forever".
782 New llvm.expect intrinsic.
783 Table generated MC expansion logic for pseudo instructions that expand to multiple MC instructions through the PseudoInstExpansion class. (JimG)
784 New llvm.fma intrinsic.
786 Euro dev meeting and main one too.
787 New atomics instructions, "#i_fence" instruction, cmpxchg, atomicrmw too. What target support (X86/ARM)? Also 'atomic load/store'. See Atomics.html
788 X86: inline assembler supports .code32 and .code64.
789 Exception handling rewrite: new landingpad and resume instruction. Unwind gone.
790 LowerSetJmp pass removed, unused.
791 llvm-objdump / dwarf parser library / llvm-dwarfdump (d0k)
792 object file parsing stuff and llvm-size (mspencer)
794 Old arm disassembler replaced with a new one based on autogenerated encoding information from ARM .td files.
795 Frontend tests removed from llvm/test/Frontend* (was this completed for 3.0?)
796 Segmented stack support (X86 only?) Rafael and Sanjoy Das: docs/SegmentedStacks.html should be in CodeGen.html status table?
797 X86 backend support for NaCl (David Meyer / Nick L)
798 Codegen now supports vector "select" operations on vector comparisons, turning
799 them into various optimized code sequences (e.g. using the SSE4/AVX "blend"
801 #line directives in integrated assembler
802 SSE domain fixing code enabled for AVX (Bruno/Jakob). Domain fixing pass is
803 now target independent (ExecutionDepsFix pass). (Jakob)
804 X86 backend synthesizes horizontal add/sub instructions from generic code.
805 returns_twice attribute (rafael)
806 Tablegen has been split into a library, clang tblgen pieces now live in clang.
807 The llvm version is now named llvm-tblgen instead of tblgen.
808 X86: Tons of encoding improvements and new instructions (e.g. Atom, Ivy Bridge,
809 and BMI instructions)
810 added to assembler and disassembler (Craig Topper)
811 data layout string can encode the natural alignment of the target's stack for better optimization (LangRef.html#datalayout)
824 <!--=========================================================================-->
826 <a name="coreimprovements">LLVM IR and Core Improvements</a>
831 <p>LLVM IR has several new features for better support of new targets and that
832 expose new optimization opportunities:</p>
834 <p>One of the biggest changes is that 3.0 has a new exception handling
835 system. The old system used LLVM intrinsics to convey the exception handling
836 information to the code generator. It worked in most cases, but not
837 all. Inlining was especially difficult to get right. Also, the intrinsics
838 could be moved away from the <code>invoke</code> instruction, making it hard
839 to recover that information.</p>
841 <p>The new EH system makes exception handling a first-class member of the IR. It
842 adds two new instructions:</p>
845 <li><a href="LangRef.html#i_landingpad"><code>landingpad</code></a> —
846 this instruction defines a landing pad basic block. It contains all of the
847 information that's needed by the code generator. It's also required to be
848 the first non-PHI instruction in the landing pad. In addition, a landing
849 pad may be jumped to only by the unwind edge of an <code>invoke</code>
852 <li><a href="LangRef.html#i_resume"><code>resume</code></a> — this
853 instruction causes the current exception to resume traveling up the
854 stack. It replaces the <code>@llvm.eh.resume</code> intrinsic.</li>
857 <p>Converting from the old EH API to the new EH API is rather simple, because a
858 lot of complexity has been removed. The two intrinsics,
859 <code>@llvm.eh.exception</code> and <code>@llvm.eh.selector</code> have been
860 superseded by the <code>landingpad</code> instruction. Instead of generating
861 a call to <code>@llvm.eh.exception</code> and <code>@llvm.eh.selector</code>:
863 <div class="doc_code">
865 Function *ExcIntr = Intrinsic::getDeclaration(TheModule,
866 Intrinsic::eh_exception);
867 Function *SlctrIntr = Intrinsic::getDeclaration(TheModule,
868 Intrinsic::eh_selector);
870 // The exception pointer.
871 Value *ExnPtr = Builder.CreateCall(ExcIntr, "exc_ptr");
873 std::vector<Value*> Args;
874 Args.push_back(ExnPtr);
875 Args.push_back(Builder.CreateBitCast(Personality,
876 Type::getInt8PtrTy(Context)));
878 <i>// Add selector clauses to Args.</i>
880 // The selector call.
881 Builder.CreateCall(SlctrIntr, Args, "exc_sel");
885 <p>You should instead generate a <code>landingpad</code> instruction, that
886 returns an exception object and selector value:</p>
888 <div class="doc_code">
890 LandingPadInst *LPadInst =
891 Builder.CreateLandingPad(StructType::get(Int8PtrTy, Int32Ty, NULL),
894 Value *LPadExn = Builder.CreateExtractValue(LPadInst, 0);
895 Builder.CreateStore(LPadExn, getExceptionSlot());
897 Value *LPadSel = Builder.CreateExtractValue(LPadInst, 1);
898 Builder.CreateStore(LPadSel, getEHSelectorSlot());
902 <p>It's now trivial to add the individual clauses to the <code>landingpad</code>
905 <div class="doc_code">
907 <i><b>// Adding a catch clause</b></i>
908 Constant *TypeInfo = getTypeInfo();
909 LPadInst->addClause(TypeInfo);
911 <i><b>// Adding a C++ catch-all</b></i>
912 LPadInst->addClause(Constant::getNullValue(Builder.getInt8PtrTy()));
914 <i><b>// Adding a cleanup</b></i>
915 LPadInst->setCleanup(true);
917 <i><b>// Adding a filter clause</b></i>
918 std::vector<Constant*> TypeInfos;
919 Constant *TypeInfo = getFilterTypeInfo();
920 TypeInfos.push_back(Builder.CreateBitCast(TypeInfo, Builder.getInt8PtrTy()));
922 ArrayType *FilterTy = ArrayType::get(Int8PtrTy, TypeInfos.size());
923 LPadInst->addClause(ConstantArray::get(FilterTy, TypeInfos));
927 <p>Converting from using the <code>@llvm.eh.resume</code> intrinsic to
928 the <code>resume</code> instruction is trivial. It takes the exception
929 pointer and exception selector values returned by
930 the <code>landingpad</code> instruction:</p>
932 <div class="doc_code">
934 Type *UnwindDataTy = StructType::get(Builder.getInt8PtrTy(),
935 Builder.getInt32Ty(), NULL);
936 Value *UnwindData = UndefValue::get(UnwindDataTy);
937 Value *ExcPtr = Builder.CreateLoad(getExceptionObjSlot());
938 Value *ExcSel = Builder.CreateLoad(getExceptionSelSlot());
939 UnwindData = Builder.CreateInsertValue(UnwindData, ExcPtr, 0, "exc_ptr");
940 UnwindData = Builder.CreateInsertValue(UnwindData, ExcSel, 1, "exc_sel");
941 Builder.CreateResume(UnwindData);
947 <!--=========================================================================-->
949 <a name="loopoptimization">Loop Optimization Improvements</a>
953 <p>The induction variable simplification pass in 3.0 only modifies
954 induction variables when profitable. Sign and zero extension
955 elimination, linear function test replacement, loop unrolling, and
956 other simplifications that require induction variable analysis have
957 been generalized so they no longer require loops to be rewritten in a
958 typically suboptimal form prior to optimization. This new design
959 preserves more IR level information, avoids undoing earlier loop
960 optimizations (particularly hand-optimized loops), and no longer
961 strongly depends on the code generator rewriting loops a second time
962 in a now optimal form--an intractable problem.</p>
964 <p>The original behavior can be restored with -mllvm -enable-iv-rewrite;
965 however, support for this mode will be short lived. As such, bug
966 reports should be filed for any significant performance regressions
967 when moving from -mllvm -enable-iv-rewrite to the 3.0 default mode.</p>
970 <!--=========================================================================-->
972 <a name="optimizer">Optimizer Improvements</a>
977 <p>In addition to a large array of minor performance tweaks and bug fixes, this
978 release includes a few major enhancements and additions to the
982 <li>Information about <a href="BranchWeightMetadata.html">branch probability</a>
983 and basic block frequency is now available within LLVM, based on a
984 combination of static branch prediction heuristics and
985 <code>__builtin_expect</code> calls. That information is currently used for
986 register spill placement and if-conversion, with additional optimizations
987 planned for future releases. The same framework is intended for eventual
988 use with profile-guided optimization.</li>
995 <!--=========================================================================-->
997 <a name="mc">MC Level Improvements</a>
1002 <p>The LLVM Machine Code (aka MC) subsystem was created to solve a number of
1003 problems in the realm of assembly, disassembly, object file format handling,
1004 and a number of other related areas that CPU instruction-set level tools work
1008 <li>The ELF object streamers are much more full featured.</li>
1009 <li>Target dependent relocation handling has been refactored into the Targets.</li>
1010 <li>Early stage MC-JIT infrastructure has been implemented.</li>
1013 <p>The MC-JIT is a major new feature for MC, and will eventually grow to replace
1014 the current JIT implementation. It emits object files direct to memory and
1015 uses a runtime dynamic linker to resolve references and drive lazy compilation.
1016 The MC-JIT enables much greater code reuse between the JIT and the static
1017 compiler and provides better integration with the platform ABI as a result.</p>
1019 <p>For more information, please see
1020 the <a href="http://blog.llvm.org/2010/04/intro-to-llvm-mc-project.html">Intro
1021 to the LLVM MC Project Blog Post</a>.</p>
1025 <!--=========================================================================-->
1027 <a name="codegen">Target Independent Code Generator Improvements</a>
1032 <p>We have put a significant amount of work into the code generator
1033 infrastructure, which allows us to implement more aggressive algorithms and
1034 make it run faster:</p>
1043 <!--=========================================================================-->
1045 <a name="x86">X86-32 and X86-64 Target Improvements</a>
1050 <p>New features and major changes in the X86 target include:</p>
1053 <li>The X86 backend, assembler and disassembler now completely support AVX.
1054 To enable it pass <code>-mavx</code> to the compiler.</li>
1056 <li>The X86 backend now supports
1057 all <a href="http://llvm.org/PR879">inline assembly that uses the X86
1058 floating point stack</a>.</li>
1060 <li>The CRC32 intrinsics have been renamed. The intrinsics were previously
1061 <code>@llvm.x86.sse42.crc32.[8|16|32]</code>
1062 and <code>@llvm.x86.sse42.crc64.[8|64]</code>. They have been renamed to
1063 <code>@llvm.x86.sse42.crc32.32.[8|16|32]</code> and
1064 <code>@llvm.x86.sse42.crc32.64.[8|64]</code>.</li>
1070 <!--=========================================================================-->
1072 <a name="ARM">ARM Target Improvements</a>
1077 <p>New features of the ARM target include:</p>
1080 <li>Reworked Set Jump Long Jump EH Lowering,</li>
1081 <li>improved support for Cortex-M series processors, and</li>
1082 <li>beta quality integrated assembler support.</li>
1087 <!--=========================================================================-->
1089 <a name="MIPS">MIPS Target Improvements</a>
1094 <p>New features and major changes in the MIPS target include:</p>
1097 <li>Most MIPS32r1 and r2 instructions are now supported.</li>
1098 <li>LE/BE MIPS32r1/r2 has been tested extensively.</li>
1099 <li>O32 ABI has been fully tested.</li>
1100 <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>
1101 <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>
1102 <li>Support for old-style JIT is complete.</li>
1103 <li>Support for old architectures (MIPS1 and MIPS2) has been removed.</li>
1104 <li>Initial support for MIPS64 has been added.</li>
1108 <!--=========================================================================-->
1110 <a name="PTX">PTX Target Improvements</a>
1116 The PTX back-end is still experimental, but is fairly usable for compute kernels
1117 in LLVM 3.0. Most scalar arithmetic is implemented, as well as intrinsics to
1118 access the special PTX registers and sync instructions. The major missing
1119 pieces are texture/sampler support and some vector operations.</p>
1121 <p>That said, the backend is already being used for domain-specific languages
1122 and works well with the <a href="http://www.pcc.me.uk/~peter/libclc/">libclc
1123 library</a> to supply OpenCL built-ins. With it, you can use Clang to compile
1124 OpenCL code into PTX and execute it by loading the resulting PTX as a binary
1125 blob using the nVidia OpenCL library. It has been tested with several OpenCL
1126 programs, including some from the nVidia GPU Computing SDK, and the performance
1127 is on par with the nVidia compiler.</p>
1131 <!--=========================================================================-->
1133 <a name="OtherTS">Other Target Specific Improvements</a>
1138 <p>PPC32/ELF va_arg was implemented.</p>
1139 <p>PPC32 initial support for .o file writing was implemented.</p>
1140 <p>MicroBlaze scheduling itineraries were added that model the
1141 3-stage and the 5-stage pipeline architectures. The 3-stage
1142 pipeline model can be selected with <code>-mcpu=mblaze3</code>
1143 and the 5-stage pipeline model can be selected with
1144 <code>-mcpu=mblaze5</code>.</p>
1154 <!--=========================================================================-->
1156 <a name="changes">Major Changes and Removed Features</a>
1161 <p>If you're already an LLVM user or developer with out-of-tree changes based on
1162 LLVM 2.9, this section lists some "gotchas" that you may run into upgrading
1163 from the previous release.</p>
1166 <li>The <code>LLVMC</code> meta compiler driver was removed.</li>
1167 <li>The <code>LowerSetJmp</code> pass wasn't used effectively by any
1168 target and has been removed.</li>
1169 <li>The old <code>TailDup</code> pass was not used in the standard pipeline
1170 and was unable to update ssa form, so it has been removed.
1171 <li>The syntax of volatile loads and stores in IR has been changed to
1172 "<code>load volatile</code>"/"<code>store volatile</code>". The old
1173 syntax ("<code>volatile load</code>"/"<code>volatile store</code>")
1174 is still accepted, but is now considered deprecated.</li>
1175 <li>The old atomic intrinsics (<code>llvm.memory.barrier</code> and
1176 <code>llvm.atomic.*</code>) are now gone. Please use the new atomic
1177 instructions, described in the <a href="Atomics.html">atomics guide</a>.
1180 <h4>Windows (32-bit)</h4>
1184 <li>On Win32(MinGW32 and MSVC), Windows 2000 will not be supported.
1185 Windows XP or higher is required.</li>
1192 <!--=========================================================================-->
1194 <a name="api_changes">Internal API Changes</a>
1199 <p>In addition, many APIs have changed in this release. Some of the major
1200 LLVM API changes are:</p>
1203 <li>The biggest and most pervasive change is that llvm::Types are no longer
1204 returned or accepted as 'const' values. Instead, just pass around
1205 non-const Types.</li>
1207 <li><code>PHINode::reserveOperandSpace</code> has been removed. Instead, you
1208 must specify how many operands to reserve space for when you create the
1209 PHINode, by passing an extra argument
1210 into <code>PHINode::Create</code>.</li>
1212 <li>PHINodes no longer store their incoming BasicBlocks as operands. Instead,
1213 the list of incoming BasicBlocks is stored separately, and can be accessed
1214 with new functions <code>PHINode::block_begin</code>
1215 and <code>PHINode::block_end</code>.</li>
1217 <li>Various functions now take an <code>ArrayRef</code> instead of either a
1218 pair of pointers (or iterators) to the beginning and end of a range, or a
1219 pointer and a length. Others now return an <code>ArrayRef</code> instead
1220 of a reference to a <code>SmallVector</code>
1221 or <code>std::vector</code>. These include:
1223 <!-- Please keep this list sorted. -->
1224 <li><code>CallInst::Create</code></li>
1225 <li><code>ComputeLinearIndex</code> (in <code>llvm/CodeGen/Analysis.h</code>)</li>
1226 <li><code>ConstantArray::get</code></li>
1227 <li><code>ConstantExpr::getExtractElement</code></li>
1228 <li><code>ConstantExpr::getGetElementPtr</code></li>
1229 <li><code>ConstantExpr::getInBoundsGetElementPtr</code></li>
1230 <li><code>ConstantExpr::getIndices</code></li>
1231 <li><code>ConstantExpr::getInsertElement</code></li>
1232 <li><code>ConstantExpr::getWithOperands</code></li>
1233 <li><code>ConstantFoldCall</code> (in <code>llvm/Analysis/ConstantFolding.h</code>)</li>
1234 <li><code>ConstantFoldInstOperands</code> (in <code>llvm/Analysis/ConstantFolding.h</code>)</li>
1235 <li><code>ConstantVector::get</code></li>
1236 <li><code>DIBuilder::createComplexVariable</code></li>
1237 <li><code>DIBuilder::getOrCreateArray</code></li>
1238 <li><code>ExtractValueInst::Create</code></li>
1239 <li><code>ExtractValueInst::getIndexedType</code></li>
1240 <li><code>ExtractValueInst::getIndices</code></li>
1241 <li><code>FindInsertedValue</code> (in <code>llvm/Analysis/ValueTracking.h</code>)</li>
1242 <li><code>gep_type_begin</code> (in <code>llvm/Support/GetElementPtrTypeIterator.h</code>)</li>
1243 <li><code>gep_type_end</code> (in <code>llvm/Support/GetElementPtrTypeIterator.h</code>)</li>
1244 <li><code>GetElementPtrInst::Create</code></li>
1245 <li><code>GetElementPtrInst::CreateInBounds</code></li>
1246 <li><code>GetElementPtrInst::getIndexedType</code></li>
1247 <li><code>InsertValueInst::Create</code></li>
1248 <li><code>InsertValueInst::getIndices</code></li>
1249 <li><code>InvokeInst::Create</code></li>
1250 <li><code>IRBuilder::CreateCall</code></li>
1251 <li><code>IRBuilder::CreateExtractValue</code></li>
1252 <li><code>IRBuilder::CreateGEP</code></li>
1253 <li><code>IRBuilder::CreateInBoundsGEP</code></li>
1254 <li><code>IRBuilder::CreateInsertValue</code></li>
1255 <li><code>IRBuilder::CreateInvoke</code></li>
1256 <li><code>MDNode::get</code></li>
1257 <li><code>MDNode::getIfExists</code></li>
1258 <li><code>MDNode::getTemporary</code></li>
1259 <li><code>MDNode::getWhenValsUnresolved</code></li>
1260 <li><code>SimplifyGEPInst</code> (in <code>llvm/Analysis/InstructionSimplify.h</code>)</li>
1261 <li><code>TargetData::getIndexedOffset</code></li>
1264 <li>All forms of <code>StringMap::getOrCreateValue</code> have been remove
1265 except for the one which takes a <code>StringRef</code>.</li>
1267 <li>The <code>LLVMBuildUnwind</code> function from the C API was removed. The
1268 LLVM <code>unwind</code> instruction has been deprecated for a long time
1269 and isn't used by the current front-ends. So this was removed during the
1270 exception handling rewrite.</li>
1272 <li>The <code>LLVMAddLowerSetJmpPass</code> function from the C API was
1273 removed because the <code>LowerSetJmp</code> pass was removed.</li>
1275 <li>The <code>DIBuilder</code> interface used by front ends to encode
1276 debugging information in the LLVM IR now expects clients to
1277 use <code>DIBuilder::finalize()</code> at the end of translation unit to
1278 complete debugging information encoding.</li>
1280 <li>The way the type system works has been
1281 rewritten: <code>PATypeHolder</code> and <code>OpaqueType</code> are gone,
1282 and all APIs deal with <code>Type*</code> instead of <code>const
1283 Type*</code>. If you need to create recursive structures, then create a
1284 named structure, and use <code>setBody()</code> when all its elements are
1285 built. Type merging and refining is gone too: named structures are not
1286 merged with other structures, even if their layout is identical. (of
1287 course anonymous structures are still uniqued by layout).</li>
1289 <li>TargetSelect.h moved to Support/ from Target/</li>
1291 <li>UpgradeIntrinsicCall no longer upgrades pre-2.9 intrinsic calls (for
1292 example <code>llvm.memset.i32</code>).</li>
1294 <li>It is mandatory to initialize all out-of-tree passes too and their dependencies now with
1295 <code>INITIALIZE_PASS{BEGIN,END,}</code>
1296 and <code>INITIALIZE_{PASS,AG}_DEPENDENCY</code>.</li>
1298 <li>The interface for MemDepResult in MemoryDependenceAnalysis has been
1299 enhanced with new return types Unknown and NonFuncLocal, in addition to
1300 the existing types Clobber, Def, and NonLocal.</li>
1307 <!-- *********************************************************************** -->
1309 <a name="knownproblems">Known Problems</a>
1311 <!-- *********************************************************************** -->
1315 <p>This section contains significant known problems with the LLVM system, listed
1316 by component. If you run into a problem, please check
1317 the <a href="http://llvm.org/bugs/">LLVM bug database</a> and submit a bug if
1318 there isn't already one.</p>
1320 <!-- ======================================================================= -->
1322 <a name="experimental">Experimental features included with this release</a>
1327 <p>The following components of this LLVM release are either untested, known to
1328 be broken or unreliable, or are in early development. These components
1329 should not be relied on, and bugs should not be filed against them, but they
1330 may be useful to some people. In particular, if you would like to work on
1331 one of these components, please contact us on
1332 the <a href="http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev">LLVMdev
1336 <li>The Alpha, Blackfin, CellSPU, MicroBlaze, MSP430, MIPS, PTX, SystemZ and
1337 XCore backends are experimental.</li>
1339 <li><tt>llc</tt> "<tt>-filetype=obj</tt>" is experimental on all targets other
1340 than darwin and ELF X86 systems.</li>
1345 <!-- ======================================================================= -->
1347 <a name="x86-be">Known problems with the X86 back-end</a>
1353 <li>The X86-64 backend <a href="http://llvm.org/PR1740">does not yet support
1354 the <tt>va_arg</tt> LLVM IR instruction</a>. Currently, front-ends support
1355 variadic argument constructs on X86-64 by lowering them manually.</li>
1360 <!-- ======================================================================= -->
1362 <a name="ppc-be">Known problems with the PowerPC back-end</a>
1368 <li>The PPC32/ELF support lacks PIC support.</li>
1373 <!-- ======================================================================= -->
1375 <a name="arm-be">Known problems with the ARM back-end</a>
1381 <li>Thumb mode works only on ARMv6 or higher processors. On sub-ARMv6
1382 processors, thumb programs can crash or produce wrong results
1383 (<a href="http://llvm.org/PR1388">PR1388</a>).</li>
1385 <li>Compilation for ARM Linux OABI (old ABI) is supported but not fully
1391 <!-- ======================================================================= -->
1393 <a name="sparc-be">Known problems with the SPARC back-end</a>
1399 <li>The SPARC backend only supports the 32-bit SPARC ABI (-m32); it does not
1400 support the 64-bit SPARC ABI (-m64).</li>
1405 <!-- ======================================================================= -->
1407 <a name="mips-be">Known problems with the MIPS back-end</a>
1413 <li>64-bit MIPS targets are not supported yet.</li>
1418 <!-- ======================================================================= -->
1420 <a name="alpha-be">Known problems with the Alpha back-end</a>
1426 <li>On 21164s, some rare FP arithmetic sequences which may trap do not have
1427 the appropriate nops inserted to ensure restartability.</li>
1432 <!-- ======================================================================= -->
1434 <a name="c-be">Known problems with the C back-end</a>
1439 <p>The C backend has numerous problems and is not being actively maintained.
1440 Depending on it for anything serious is not advised.</p>
1443 <li><a href="http://llvm.org/PR802">The C backend has only basic support for
1444 inline assembly code</a>.</li>
1446 <li><a href="http://llvm.org/PR1658">The C backend violates the ABI of common
1447 C++ programs</a>, preventing intermixing between C++ compiled by the CBE
1448 and C++ code compiled with <tt>llc</tt> or native compilers.</li>
1450 <li>The C backend does not support all exception handling constructs.</li>
1452 <li>The C backend does not support arbitrary precision integers.</li>
1459 <!-- *********************************************************************** -->
1461 <a name="additionalinfo">Additional Information</a>
1463 <!-- *********************************************************************** -->
1467 <p>A wide variety of additional information is available on
1468 the <a href="http://llvm.org/">LLVM web page</a>, in particular in
1469 the <a href="http://llvm.org/docs/">documentation</a> section. The web page
1470 also contains versions of the API documentation which is up-to-date with the
1471 Subversion version of the source code. You can access versions of these
1472 documents specific to this release by going into the "<tt>llvm/doc/</tt>"
1473 directory in the LLVM tree.</p>
1475 <p>If you have any questions or comments about LLVM, please feel free to contact
1476 us via the <a href="http://llvm.org/docs/#maillist"> mailing lists</a>.</p>
1480 <!-- *********************************************************************** -->
1484 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
1485 src="http://jigsaw.w3.org/css-validator/images/vcss-blue" alt="Valid CSS"></a>
1486 <a href="http://validator.w3.org/check/referer"><img
1487 src="http://www.w3.org/Icons/valid-html401-blue" alt="Valid HTML 4.01"></a>
1489 <a href="http://llvm.org/">LLVM Compiler Infrastructure</a><br>
1490 Last modified: $Date$