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 2.7 Release Notes</title>
11 <div class="doc_title">LLVM 2.7 Release Notes</div>
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 2.7</a></li>
20 <li><a href="#whatsnew">What's New in LLVM 2.7?</a></li>
21 <li><a href="GettingStarted.html">Installation Instructions</a></li>
22 <li><a href="#portability">Portability and Supported Platforms</a></li>
23 <li><a href="#knownproblems">Known Problems</a></li>
24 <li><a href="#additionalinfo">Additional Information</a></li>
27 <div class="doc_author">
28 <p>Written by the <a href="http://llvm.org">LLVM Team</a></p>
32 <h1 style="color:red">These are in-progress notes for the upcoming LLVM 2.8
35 <a href="http://llvm.org/releases/2.6/docs/ReleaseNotes.html">LLVM 2.7
36 Release Notes</a>.</h1>-->
38 <!-- *********************************************************************** -->
39 <div class="doc_section">
40 <a name="intro">Introduction</a>
42 <!-- *********************************************************************** -->
44 <div class="doc_text">
46 <p>This document contains the release notes for the LLVM Compiler
47 Infrastructure, release 2.7. 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 the <a
50 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
54 web site</a>. If you have questions or comments, the <a
55 href="http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev">LLVM Developer's
56 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
59 main 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>
68 include/llvm/Analysis/LiveValues.h => Dan
69 lib/Transforms/IPO/MergeFunctions.cpp => consider for 2.8.
70 llvm/Analysis/PointerTracking.h => Edwin wants this, consider for 2.8.
73 lib/Transforms/Utils/SSI.cpp -> ABCD depends on it.
77 <!-- Features that need text if they're finished for 2.7:
80 llvm.dbg.value: variable debug info for optimized code
81 loop dependence analysis
84 <!-- for announcement email:
86 Many new papers added to /pubs/
89 <!-- *********************************************************************** -->
90 <div class="doc_section">
91 <a name="subproj">Sub-project Status Update</a>
93 <!-- *********************************************************************** -->
95 <div class="doc_text">
97 The LLVM 2.7 distribution currently consists of code from the core LLVM
98 repository (which roughly includes the LLVM optimizers, code generators
99 and supporting tools), the Clang repository and the llvm-gcc repository. In
100 addition to this code, the LLVM Project includes other sub-projects that are in
101 development. Here we include updates on these subprojects.
107 <!--=========================================================================-->
108 <div class="doc_subsection">
109 <a name="clang">Clang: C/C++/Objective-C Frontend Toolkit</a>
112 <div class="doc_text">
114 <p><a href="http://clang.llvm.org/">Clang</a> is an LLVM front end for the C,
115 C++, and Objective-C languages. Clang aims to provide a better user experience
116 through expressive diagnostics, a high level of conformance to language
117 standards, fast compilation, and low memory use. Like LLVM, Clang provides a
118 modular, library-based architecture that makes it suitable for creating or
119 integrating with other development tools. Clang is considered a
120 production-quality compiler for C and Objective-C on x86 (32- and 64-bit).</p>
122 <p>In the LLVM 2.7 time-frame, the Clang team has made many improvements:</p>
126 <li>C++ Support: Clang is now capable of self-hosting! While still
127 alpha-quality, Clang's C++ support has matured enough to build LLVM and Clang,
128 and C++ is now enabled by default. See the <a
129 href="http://clang.llvm.org/cxx_compatibility.html">Clang C++ compatibility
130 page</a> for common C++ migration issues.</li>
132 <li>Objective-C: Clang now includes experimental support for an updated
133 Objective-C ABI on non-Darwin platforms. This includes support for non-fragile
134 instance variables and accelerated proxies, as well as greater potential for
135 future optimisations. The new ABI is used when compiling with the
136 -fobjc-nonfragile-abi and -fgnu-runtime options. Code compiled with these
137 options may be mixed with code compiled with GCC or clang using the old GNU ABI,
138 but requires the libobjc2 runtime from the GNUstep project.</li>
140 <li>New warnings: Clang contains a number of new warnings, including
141 control-flow warnings (unreachable code, missing return statements in a
142 non-<code>void</code> function, etc.), sign-comparison warnings, and improved
143 format-string warnings.</li>
145 <li>CIndex API and Python bindings: Clang now includes a C API as part of the
146 CIndex library. Although we may make some changes to the API in the future, it
147 is intended to be stable and has been designed for use by external projects. See
149 doxygen <a href="http://clang.llvm.org/doxygen/group__CINDEX.html">CIndex</a>
150 documentation for more details. The CIndex API also includes a preliminary
151 set of Python bindings.</li>
153 <li>ARM Support: Clang now has ABI support for both the Darwin and Linux ARM
154 ABIs. Coupled with many improvements to the LLVM ARM backend, Clang is now
155 suitable for use as a beta quality ARM compiler.</li>
160 <!--=========================================================================-->
161 <div class="doc_subsection">
162 <a name="clangsa">Clang Static Analyzer</a>
165 <div class="doc_text">
167 <p>The <a href="http://clang-analyzer.llvm.org/">Clang Static Analyzer</a>
168 project is an effort to use static source code analysis techniques to
169 automatically find bugs in C and Objective-C programs (and hopefully <a
170 href="http://clang-analyzer.llvm.org/dev_cxx.html">C++ in the
171 future</a>!). The tool is very good at finding bugs that occur on specific
172 paths through code, such as on error conditions.</p>
174 <p>In the LLVM 2.7 time-frame, the analyzer core has made several major and
175 minor improvements, including better support for tracking the fields of
176 structures, initial support (not enabled by default yet) for doing
177 interprocedural (cross-function) analysis, and new checks have been added.
182 <!--=========================================================================-->
183 <div class="doc_subsection">
184 <a name="vmkit">VMKit: JVM/CLI Virtual Machine Implementation</a>
187 <div class="doc_text">
189 The <a href="http://vmkit.llvm.org/">VMKit project</a> is an implementation of
190 a JVM and a CLI Virtual Machine (Microsoft .NET is an
191 implementation of the CLI) using LLVM for static and just-in-time
195 With the release of LLVM 2.7, VMKit has shifted to a great framework for writing
196 virtual machines. VMKit now offers precise and efficient garbage collection with
197 multi-threading support, thanks to the MMTk memory management toolkit, as well
198 as just in time and ahead of time compilation with LLVM. The major changes in
203 <li>Garbage collection: VMKit now uses the MMTk toolkit for garbage collectors.
204 The first collector to be ported is the MarkSweep collector, which is precise,
205 and drastically improves the performance of VMKit.</li>
206 <li>Line number information in the JVM: by using the debug metadata of LLVM, the
207 JVM now supports precise line number information, useful when printing a stack
209 <li>Interface calls in the JVM: we implemented a variant of the Interface Method
210 Table technique for interface calls in the JVM.
217 <!--=========================================================================-->
218 <div class="doc_subsection">
219 <a name="compiler-rt">compiler-rt: Compiler Runtime Library</a>
222 <div class="doc_text">
224 The new LLVM <a href="http://compiler-rt.llvm.org/">compiler-rt project</a>
225 is a simple library that provides an implementation of the low-level
226 target-specific hooks required by code generation and other runtime components.
227 For example, when compiling for a 32-bit target, converting a double to a 64-bit
228 unsigned integer is compiled into a runtime call to the "__fixunsdfdi"
229 function. The compiler-rt library provides highly optimized implementations of
230 this and other low-level routines (some are 3x faster than the equivalent
231 libgcc routines).</p>
234 All of the code in the compiler-rt project is available under the standard LLVM
235 License, a "BSD-style" license. New in LLVM 2.7: compiler_rt now
236 supports ARM targets.</p>
240 <!--=========================================================================-->
241 <div class="doc_subsection">
242 <a name="dragonegg">DragonEgg: llvm-gcc ported to gcc-4.5</a>
245 <div class="doc_text">
247 <a href="http://dragonegg.llvm.org/">DragonEgg</a> is a port of llvm-gcc to
248 gcc-4.5. Unlike llvm-gcc, which makes many intrusive changes to the underlying
249 gcc-4.2 code, dragonegg in theory does not require any gcc-4.5 modifications
250 whatsoever (currently one small patch is needed). This is thanks to the new
251 <a href="http://gcc.gnu.org/wiki/plugins">gcc plugin architecture</a>, which
252 makes it possible to modify the behaviour of gcc at runtime by loading a plugin,
253 which is nothing more than a dynamic library which conforms to the gcc plugin
254 interface. DragonEgg is a gcc plugin that causes the LLVM optimizers to be run
255 instead of the gcc optimizers, and the LLVM code generators instead of the gcc
256 code generators, just like llvm-gcc. To use it, you add
257 "-fplugin=path/dragonegg.so" to the gcc-4.5 command line, and gcc-4.5 magically
258 becomes llvm-gcc-4.5!
262 DragonEgg is still a work in progress. Currently C works very well, while C++,
263 Ada and Fortran work fairly well. All other languages either don't work at all,
264 or only work poorly. For the moment only the x86-32 and x86-64 targets are
265 supported, and only on linux and darwin (darwin needs an additional gcc patch).
269 DragonEgg is a new project which is seeing its first release with llvm-2.7.
275 <!--=========================================================================-->
276 <div class="doc_subsection">
277 <a name="mc">llvm-mc: Machine Code Toolkit</a>
280 <div class="doc_text">
282 The LLVM Machine Code (aka MC) sub-project of LLVM was created to solve a number
283 of problems in the realm of assembly, disassembly, object file format handling,
284 and a number of other related areas that CPU instruction-set level tools work
285 in. It is a sub-project of LLVM which provides it with a number of advantages
286 over other compilers that do not have tightly integrated assembly-level tools.
287 For a gentle introduction, please see the <a
288 href="http://blog.llvm.org/2010/04/intro-to-llvm-mc-project.html">Intro to the
289 LLVM MC Project Blog Post</a>.
292 <p>2.7 includes major parts of the work required by the new MC Project. A few
293 targets have been refactored to support it, and work is underway to support a
294 native assembler in LLVM. This work is not complete in LLVM 2.7, but it has
295 made substantially more progress on LLVM mainline.</p>
297 <p>One minor example of what MC can do is to transcode an AT&T syntax
298 X86 .s file into intel syntax. You can do this with something like:</p>
300 llvm-mc foo.s -output-asm-variant=1 -o foo-intel.s
306 <!-- *********************************************************************** -->
307 <div class="doc_section">
308 <a name="externalproj">External Open Source Projects Using LLVM 2.7</a>
310 <!-- *********************************************************************** -->
312 <div class="doc_text">
314 <p>An exciting aspect of LLVM is that it is used as an enabling technology for
315 a lot of other language and tools projects. This section lists some of the
316 projects that have already been updated to work with LLVM 2.7.</p>
319 <!--=========================================================================-->
320 <div class="doc_subsection">
321 <a name="pure">Pure</a>
324 <div class="doc_text">
326 <a href="http://pure-lang.googlecode.com/">Pure</a>
327 is an algebraic/functional programming language based on term rewriting.
328 Programs are collections of equations which are used to evaluate expressions in
329 a symbolic fashion. Pure offers dynamic typing, eager and lazy evaluation,
330 lexical closures, a hygienic macro system (also based on term rewriting),
331 built-in list and matrix support (including list and matrix comprehensions) and
332 an easy-to-use C interface. The interpreter uses LLVM as a backend to
333 JIT-compile Pure programs to fast native code.</p>
335 <p>Pure versions 0.43 and later have been tested and are known to work with
336 LLVM 2.7 (and continue to work with older LLVM releases >= 2.5).</p>
340 <!--=========================================================================-->
341 <div class="doc_subsection">
342 <a name="RoadsendPHP">Roadsend PHP</a>
345 <div class="doc_text">
347 <a href="http://code.roadsend.com/rphp">Roadsend PHP</a> (rphp) is an open
348 source implementation of the PHP programming
349 language that uses LLVM for its optimizer, JIT and static compiler. This is a
350 reimplementation of an earlier project that is now based on LLVM.
354 <!--=========================================================================-->
355 <div class="doc_subsection">
356 <a name="UnladenSwallow">Unladen Swallow</a>
359 <div class="doc_text">
361 <a href="http://code.google.com/p/unladen-swallow/">Unladen Swallow</a> is a
362 branch of <a href="http://python.org/">Python</a> intended to be fully
363 compatible and significantly faster. It uses LLVM's optimization passes and JIT
368 <!--=========================================================================-->
369 <div class="doc_subsection">
370 <a name="tce">TTA-based Codesign Environment (TCE)</a>
373 <div class="doc_text">
375 <a href="http://tce.cs.tut.fi/">TCE</a> is a toolset for designing
376 application-specific processors (ASP) based on the Transport triggered
377 architecture (TTA). The toolset provides a complete co-design flow from C/C++
378 programs down to synthesizable VHDL and parallel program binaries. Processor
379 customization points include the register files, function units, supported
380 operations, and the interconnection network.</p>
382 <p>TCE uses llvm-gcc/Clang and LLVM for C/C++ language support, target
383 independent optimizations and also for parts of code generation. It generates
384 new LLVM-based code generators "on the fly" for the designed TTA processors and
385 loads them in to the compiler backend as runtime libraries to avoid per-target
386 recompilation of larger parts of the compiler chain.</p>
390 <!--=========================================================================-->
391 <div class="doc_subsection">
392 <a name="safecode">SAFECode Compiler</a>
395 <div class="doc_text">
397 <a href="http://safecode.cs.illinois.edu">SAFECode</a> is a memory safe C
398 compiler built using LLVM. It takes standard, unannotated C code, analyzes the
399 code to ensure that memory accesses and array indexing operations are safe, and
400 instruments the code with run-time checks when safety cannot be proven
405 <!--=========================================================================-->
406 <div class="doc_subsection">
407 <a name="icedtea">IcedTea Java Virtual Machine Implementation</a>
410 <div class="doc_text">
412 <a href="http://icedtea.classpath.org/wiki/Main_Page">IcedTea</a> provides a
413 harness to build OpenJDK using only free software build tools and to provide
414 replacements for the not-yet free parts of OpenJDK. One of the extensions that
415 IcedTea provides is a new JIT compiler named <a
416 href="http://icedtea.classpath.org/wiki/ZeroSharkFaq">Shark</a> which uses LLVM
417 to provide native code generation without introducing processor-dependent
420 <p>Icedtea6 1.8 and later have been tested and are known to work with
421 LLVM 2.7 (and continue to work with older LLVM releases >= 2.6 as well).
425 <!--=========================================================================-->
426 <div class="doc_subsection">
427 <a name="llvm-lua">LLVM-Lua</a>
430 <div class="doc_text">
432 <a href="http://code.google.com/p/llvm-lua/">LLVM-Lua</a> uses LLVM
433 to add JIT and static compiling support to the Lua VM. Lua
434 bytecode is analyzed to remove type checks, then LLVM is used to compile the
435 bytecode down to machine code.
437 <p>LLVM-Lua 1.2.0 have been tested and is known to work with LLVM 2.7.
441 <!--=========================================================================-->
442 <div class="doc_subsection">
443 <a name="MacRuby">MacRuby</a>
446 <div class="doc_text">
448 <a href="http://macruby.org">MacRuby</a> is an implementation of Ruby based on
449 core Mac OS technologies, sponsored by Apple Inc. It uses LLVM at runtime for
450 optimization passes, JIT compilation and exception handling. It also allows
451 static (ahead-of-time) compilation of Ruby code straight to machine code.
453 <p>The upcoming MacRuby 0.6 release works with LLVM 2.7.
457 <!--=========================================================================-->
458 <div class="doc_subsection">
459 <a name="GHC">Glasgow Haskell Compiler (GHC)</a>
462 <div class="doc_text">
464 <a href="http://www.haskell.org/ghc/">GHC</a> is an open source,
465 state-of-the-art programming suite for Haskell, a standard lazy
466 functional programming language. It includes an optimizing static
467 compiler generating good code for a variety of platforms, together
468 with an interactive system for convenient, quick development.</p>
470 <p>In addition to the existing C and native code generators, GHC now
472 href="http://hackage.haskell.org/trac/ghc/wiki/Commentary/Compiler/Backends/LLVM">LLVM
473 code generator</a>. GHC supports LLVM 2.7.</p>
478 <!-- *********************************************************************** -->
479 <div class="doc_section">
480 <a name="whatsnew">What's New in LLVM 2.7?</a>
482 <!-- *********************************************************************** -->
484 <div class="doc_text">
486 <p>This release includes a huge number of bug fixes, performance tweaks and
487 minor improvements. Some of the major improvements and new features are listed
493 <!--=========================================================================-->
494 <div class="doc_subsection">
495 <a name="orgchanges">LLVM Community Changes</a>
498 <div class="doc_text">
500 <p>In addition to changes to the code, between LLVM 2.6 and 2.7, a number of
501 organization changes have happened:
505 <li>LLVM has a new <a href="http://llvm.org/Logo.html">official logo</a>!</li>
507 <li>Ted Kremenek and Doug Gregor have stepped forward as <a
508 href="http://llvm.org/docs/DeveloperPolicy.html#owners">Code Owners</a> of the
509 Clang static analyzer and the Clang frontend, respectively.</li>
511 <li>LLVM now has an <a href="http://blog.llvm.org">official Blog</a> at
512 <a href="http://blog.llvm.org">http://blog.llvm.org</a>. This is a great way
513 to learn about new LLVM-related features as they are implemented. Several
514 features in this release are already explained on the blog.</li>
516 <li>The LLVM web pages are now checked into the SVN server, in the "www",
517 "www-pubs" and "www-releases" SVN modules. Previously they were hidden in a
518 largely inaccessible old CVS server.</li>
520 <li><a href="http://llvm.org">llvm.org</a> is now hosted on a new (and much
521 faster) server. It is still graciously hosted at the University of Illinois
522 of Urbana Champaign.</li>
526 <!--=========================================================================-->
527 <div class="doc_subsection">
528 <a name="majorfeatures">Major New Features</a>
531 <div class="doc_text">
533 <p>LLVM 2.7 includes several major new capabilities:</p>
536 <li>2.7 includes initial support for the <a
537 href="http://en.wikipedia.org/wiki/MicroBlaze">MicroBlaze</a> target.
538 MicroBlaze is a soft processor core designed for Xilinx FPGAs.</li>
540 <li>2.7 includes a new LLVM IR "extensible metadata" feature. This feature
541 supports many different use cases, including allowing front-end authors to
542 encode source level information into LLVM IR, which is consumed by later
543 language-specific passes. This is a great way to do high-level optimizations
544 like devirtualization, type-based alias analysis, etc. See the <a
545 href="http://blog.llvm.org/2010/04/extensible-metadata-in-llvm-ir.html">
546 Extensible Metadata Blog Post</a> for more information.</li>
548 <li>2.7 encodes <a href="SourceLevelDebugging.html">debug information</a>
549 in a completely new way, built on extensible metadata. The new implementation
550 is much more memory efficient and paves the way for improvements to optimized
551 code debugging experience.</li>
553 <li>2.7 now directly supports taking the address of a label and doing an
554 indirect branch through a pointer. This is particularly useful for
555 interpreter loops, and is used to implement the GCC "address of label"
556 extension. For more information, see the <a
557 href="http://blog.llvm.org/2010/01/address-of-label-and-indirect-branches.html">
558 Address of Label and Indirect Branches in LLVM IR Blog Post</a>.
560 <li>2.7 is the first release to start supporting APIs for assembling and
561 disassembling target machine code. These APIs are useful for a variety of
562 low level clients, and are surfaced in the new "enhanced disassembly" API.
563 For more information see the <a
564 href="http://blog.llvm.org/2010/01/x86-disassembler.html">The X86
565 Disassembler Blog Post</a> for more information.</li>
567 <li>2.7 includes major parts of the work required by the new MC Project,
568 see the <a href="#mc">MC update above</a> for more information.</li>
574 <!--=========================================================================-->
575 <div class="doc_subsection">
576 <a name="coreimprovements">LLVM IR and Core Improvements</a>
579 <div class="doc_text">
580 <p>LLVM IR has several new features for better support of new targets and that
581 expose new optimization opportunities:</p>
584 <li>LLVM IR now supports a 16-bit "half float" data type through <a
585 href="LangRef.html#int_fp16">two new intrinsics</a> and APFloat support.</li>
586 <li>LLVM IR supports two new <a href="LangRef.html#fnattrs">function
587 attributes</a>: inlinehint and alignstack(n). The former is a hint to the
588 optimizer that a function was declared 'inline' and thus the inliner should
589 weight it higher when considering inlining it. The later
590 indicates to the code generator that the function diverges from the platform
591 ABI on stack alignment.</li>
592 <li>The new <a href="LangRef.html#int_objectsize">llvm.objectsize</a> intrinsic
593 allows the optimizer to infer the sizes of memory objects in some cases.
594 This intrinsic is used to implement the GCC <tt>__builtin_object_size</tt>
596 <li>LLVM IR now supports marking load and store instructions with <a
597 href="LangRef.html#i_load">"non-temporal" hints</a> (building on the new
598 metadata feature). This hint encourages the code
599 generator to generate non-temporal accesses when possible, which are useful
600 for code that is carefully managing cache behavior. Currently, only the
601 X86 backend provides target support for this feature.</li>
603 <li>LLVM 2.7 has pre-alpha support for <a
604 href="LangRef.html#t_union">unions in LLVM IR</a>.
605 Unfortunately, this support is not really usable in 2.7, so if you're
606 interested in pushing it forward, please help contribute to LLVM mainline.</li>
612 <!--=========================================================================-->
613 <div class="doc_subsection">
614 <a name="optimizer">Optimizer Improvements</a>
617 <div class="doc_text">
619 <p>In addition to a large array of minor performance tweaks and bug fixes, this
620 release includes a few major enhancements and additions to the optimizers:</p>
624 <li>The inliner now merges arrays stack objects in different callees when
625 inlining multiple call sites into one function. This reduces the stack size
626 of the resultant function.</li>
627 <li>The -basicaa alias analysis pass (which is the default) has been improved to
628 be less dependent on "type safe" pointers. It can now look through bitcasts
629 and other constructs more aggressively, allowing better load/store
631 <li>The load elimination optimization in the GVN Pass [<a
632 href="http://blog.llvm.org/2009/12/introduction-to-load-elimination-in-gvn.html">intro
633 blog post</a>] has been substantially improved to be more aggressive about
634 partial redundancy elimination and do more aggressive phi translation. Please
636 href="http://blog.llvm.org/2009/12/advanced-topics-in-redundant-load.html">
637 Advanced Topics in Redundant Load Elimination with a Focus on PHI Translation
638 Blog Post</a> for more details.</li>
639 <li>The module <a href="LangRef.html#datalayout">target data string</a> now
640 includes a notion of 'native' integer data types for the target. This
641 helps mid-level optimizations avoid promoting complex sequences of
642 operations to data types that are not natively supported (e.g. converting
643 i32 operations to i64 on 32-bit chips).</li>
644 <li>The mid-level optimizer is now conservative when operating on a module with
645 no target data. Previously, it would default to SparcV9 settings, which is
646 not what most people expected.</li>
647 <li>Jump threading is now much more aggressive at simplifying correlated
648 conditionals and threading blocks with otherwise complex logic. It has
649 subsumed the old "Conditional Propagation" pass, and -condprop has been
650 removed from LLVM 2.7.</li>
651 <li>The -instcombine pass has been refactored from being one huge file to being
652 a library of its own. Internally, it uses a customized IRBuilder to clean
653 it up and simplify it.</li>
655 <li>The optimal edge profiling pass is reliable and much more complete than in
656 2.6. It can be used with the llvm-prof tool but isn't wired up to the
657 llvm-gcc and clang command line options yet.</li>
659 <li>A new experimental alias analysis implementation, -scev-aa, has been added.
660 It uses LLVM's Scalar Evolution implementation to do symbolic analysis of
661 pointer offset expressions to disambiguate pointers. It can catch a few
662 cases that basicaa cannot, particularly in complex loop nests.</li>
664 <li>The default pass ordering has been tweaked for improved optimization
672 <!--=========================================================================-->
673 <div class="doc_subsection">
674 <a name="executionengine">Interpreter and JIT Improvements</a>
677 <div class="doc_text">
680 <li>The JIT now supports generating debug information and is compatible with
681 the new GDB 7.0 (and later) interfaces for registering dynamically generated
684 <li>The JIT now <a href="http://llvm.org/PR5184">defaults
685 to compiling eagerly</a> to avoid a race condition in the lazy JIT.
686 Clients that still want the lazy JIT can switch it on by calling
687 <tt>ExecutionEngine::DisableLazyCompilation(false)</tt>.</li>
689 <li>It is now possible to create more than one JIT instance in the same process.
690 These JITs can generate machine code in parallel,
691 although <a href="http://llvm.org/docs/ProgrammersManual.html#jitthreading">you
692 still have to obey the other threading restrictions</a>.</li>
698 <!--=========================================================================-->
699 <div class="doc_subsection">
700 <a name="codegen">Target Independent Code Generator Improvements</a>
703 <div class="doc_text">
705 <p>We have put a significant amount of work into the code generator
706 infrastructure, which allows us to implement more aggressive algorithms and make
710 <li>The 'llc -asm-verbose' option (which is now the default) has been enhanced
711 to emit many useful comments to .s files indicating information about spill
712 slots and loop nest structure. This should make it much easier to read and
713 understand assembly files. This is wired up in llvm-gcc and clang to
714 the <tt>-fverbose-asm</tt> option.</li>
716 <li>New LSR with "full strength reduction" mode, which can reduce address
717 register pressure in loops where address generation is important.</li>
719 <li>A new codegen level Common Subexpression Elimination pass (MachineCSE)
720 is available and enabled by default. It catches redundancies exposed by
722 <li>A new pre-register-allocation tail duplication pass is available and enabled
723 by default, it can substantially improve branch prediction quality in some
725 <li>A new sign and zero extension optimization pass (OptimizeExtsPass)
726 is available and enabled by default. This pass can takes advantage
727 architecture features like x86-64 implicit zero extension behavior and
729 <li>The code generator now supports a mode where it attempts to preserve the
730 order of instructions in the input code. This is important for source that
731 is hand scheduled and extremely sensitive to scheduling. It is compatible
732 with the GCC <tt>-fno-schedule-insns</tt> option.</li>
733 <li>The target-independent code generator now supports generating code with
734 arbitrary numbers of result values. Returning more values than was
735 previously supported is handled by returning through a hidden pointer. In
736 2.7, only the X86 and XCore targets have adopted support for this
738 <li>The code generator now supports generating code that follows the
739 <a href="LangRef.html#callingconv">Glasgow Haskell Compiler Calling
740 Convention</a> and ABI.</li>
741 <li>The "<a href="CodeGenerator.html#selectiondag_select">DAG instruction
742 selection</a>" phase of the code generator has been largely rewritten for
743 2.7. Previously, tblgen spit out tons of C++ code which was compiled and
744 linked into the target to do the pattern matching, now it emits a much
745 smaller table which is read by the target-independent code. The primary
746 advantages of this approach is that the size and compile time of various
747 targets is much improved. The X86 code generator shrunk by 1.5MB of code,
749 <li>Almost the entire code generator has switched to emitting code through the
750 MC interfaces instead of printing textually to the .s file. This led to a
751 number of cleanups and speedups. In 2.7, debug an exception handling
752 information does not go through MC yet.</li>
756 <!--=========================================================================-->
757 <div class="doc_subsection">
758 <a name="x86">X86-32 and X86-64 Target Improvements</a>
761 <div class="doc_text">
762 <p>New features of the X86 target include:
766 <li>The X86 backend now optimizes tails calls much more aggressively for
767 functions that use the standard C calling convention.</li>
768 <li>The X86 backend now models scalar SSE registers as subregs of the SSE vector
769 registers, making the code generator more aggressive in cases where scalars
770 and vector types are mixed.</li>
776 <!--=========================================================================-->
777 <div class="doc_subsection">
778 <a name="ARM">ARM Target Improvements</a>
781 <div class="doc_text">
782 <p>New features of the ARM target include:
787 <li>The ARM backend now generates instructions in unified assembly syntax.</li>
789 <li>llvm-gcc now has complete support for the ARM v7 NEON instruction set. This
790 support differs slightly from the GCC implementation. Please see the
792 href="http://blog.llvm.org/2010/04/arm-advanced-simd-neon-intrinsics-and.html">
793 ARM Advanced SIMD (NEON) Intrinsics and Types in LLVM Blog Post</a> for
794 helpful information if migrating code from GCC to LLVM-GCC.</li>
796 <li>The ARM and Thumb code generators now use register scavenging for stack
797 object address materialization. This allows the use of R3 as a general
798 purpose register in Thumb1 code, as it was previous reserved for use in
799 stack address materialization. Secondly, sequential uses of the same
800 value will now re-use the materialized constant.</li>
802 <li>The ARM backend now has good support for ARMv4 targets and has been tested
803 on StrongARM hardware. Previously, LLVM only supported ARMv4T and
806 <li>Atomic builtins are now supported for ARMv6 and ARMv7 (__sync_synchronize,
807 __sync_fetch_and_add, etc.).</li>
814 <!--=========================================================================-->
815 <div class="doc_subsection">
816 <a name="newapis">New Useful APIs</a>
819 <div class="doc_text">
821 <p>This release includes a number of new APIs that are used internally, which
822 may also be useful for external clients.
826 <li>The optimizer uses the new CodeMetrics class to measure the size of code.
827 Various passes (like the inliner, loop unswitcher, etc) all use this to make
828 more accurate estimates of the code size impact of various
830 <li>A new <a href="http://llvm.org/doxygen/InstructionSimplify_8h-source.html">
831 llvm/Analysis/InstructionSimplify.h</a> interface is available for doing
832 symbolic simplification of instructions (e.g. <tt>a+0</tt> -> <tt>a</tt>)
833 without requiring the instruction to exist. This centralizes a lot of
834 ad-hoc symbolic manipulation code scattered in various passes.</li>
835 <li>The optimizer now uses a new <a
836 href="http://llvm.org/doxygen/SSAUpdater_8h-source.html">SSAUpdater</a>
837 class which efficiently supports
838 doing unstructured SSA update operations. This centralized a bunch of code
839 scattered throughout various passes (e.g. jump threading, lcssa,
840 loop rotate, etc) for doing this sort of thing. The code generator has a
841 similar <a href="http://llvm.org/doxygen/MachineSSAUpdater_8h-source.html">
842 MachineSSAUpdater</a> class.</li>
843 <li>The <a href="http://llvm.org/doxygen/Regex_8h-source.html">
844 llvm/Support/Regex.h</a> header exposes a platform independent regular
845 expression API. Building on this, the <a
846 href="TestingGuide.html#FileCheck">FileCheck</a> utility now supports
847 regular exressions.</li>
848 <li>raw_ostream now supports a circular "debug stream" accessed with "dbgs()".
849 By default, this stream works the same way as "errs()", but if you pass
850 <tt>-debug-buffer-size=1000</tt> to opt, the debug stream is capped to a
851 fixed sized circular buffer and the output is printed at the end of the
852 program's execution. This is helpful if you have a long lived compiler
853 process and you're interested in seeing snapshots in time.</li>
859 <!--=========================================================================-->
860 <div class="doc_subsection">
861 <a name="otherimprovements">Other Improvements and New Features</a>
864 <div class="doc_text">
865 <p>Other miscellaneous features include:</p>
868 <li>You can now build LLVM as a big dynamic library (e.g. "libllvm2.7.so"). To
869 get this, configure LLVM with the --enable-shared option.</li>
871 <li>LLVM command line tools now overwrite their output by default. Previously,
872 they would only do this with -f. This makes them more convenient to use, and
873 behave more like standard unix tools.</li>
875 <li>The opt and llc tools now autodetect whether their input is a .ll or .bc
876 file, and automatically do the right thing. This means you don't need to
877 explicitly use the llvm-as tool for most things.</li>
883 <!--=========================================================================-->
884 <div class="doc_subsection">
885 <a name="changes">Major Changes and Removed Features</a>
888 <div class="doc_text">
890 <p>If you're already an LLVM user or developer with out-of-tree changes based
891 on LLVM 2.6, this section lists some "gotchas" that you may run into upgrading
892 from the previous release.</p>
897 The Andersen's alias analysis ("anders-aa") pass, the Predicate Simplifier
898 ("predsimplify") pass, the LoopVR pass, the GVNPRE pass, and the random sampling
899 profiling ("rsprofiling") passes have all been removed. They were not being
900 actively maintained and had substantial problems. If you are interested in
901 these components, you are welcome to ressurect them from SVN, fix the
902 correctness problems, and resubmit them to mainline.</li>
904 <li>LLVM now defaults to building most libraries with RTTI turned off, providing
905 a code size reduction. Packagers who are interested in building LLVM to support
906 plugins that require RTTI information should build with "make REQUIRE_RTTI=1"
907 and should read the new <a href="Packaging.html">Advice on Packaging LLVM</a>
910 <li>The LLVM interpreter now defaults to <em>not</em> using <tt>libffi</tt> even
911 if you have it installed. This makes it more likely that an LLVM built on one
912 system will work when copied to a similar system. To use <tt>libffi</tt>,
913 configure with <tt>--enable-libffi</tt>.</li>
915 <li>Debug information uses a completely different representation, an LLVM 2.6
916 .bc file should work with LLVM 2.7, but debug info won't come forward.</li>
918 <li>The LLVM 2.6 (and earlier) "malloc" and "free" instructions got removed,
919 along with LowerAllocations pass. Now you should just use a call to the
920 malloc and free functions in libc. These calls are optimized as well as
921 the old instructions were.</li>
924 <p>In addition, many APIs have changed in this release. Some of the major LLVM
928 <li>Just about everything has been converted to use <tt>raw_ostream</tt> instead of
929 <tt>std::ostream</tt>.</li>
930 <li><tt>llvm/ADT/iterator.h</tt> has been removed, just use <tt><iterator></tt>
932 <li>The <tt>Streams.h</tt> file and <tt>DOUT</tt> got removed, use <tt>DEBUG(errs() << ...);</tt>
934 <li>The <tt>TargetAsmInfo</tt> interface was renamed to <tt>MCAsmInfo</tt>.</li>
935 <li><tt>ModuleProvider</tt> has been <a
936 href="http://llvm.org/viewvc/llvm-project?view=rev&revision=94686">removed</a>
937 and its methods moved to <tt>Module</tt> and <tt>GlobalValue</tt>.
938 Most clients can remove uses of <tt>ExistingModuleProvider</tt>,
939 replace <tt>getBitcodeModuleProvider</tt> with
940 <tt>getLazyBitcodeModule</tt>, and pass their <tt>Module</tt> to
941 functions that used to accept <tt>ModuleProvider</tt>. Clients who
942 wrote their own <tt>ModuleProvider</tt>s will need to derive from
943 <tt>GVMaterializer</tt> instead and use
944 <tt>Module::setMaterializer</tt> to attach it to a
945 <tt>Module</tt>.</li>
947 <li><tt>GhostLinkage</tt> has given up the ghost.
948 <tt>GlobalValue</tt>s that have not yet been read from their backing
949 storage have the same linkage they will have after being read in.
950 Clients must replace calls to
951 <tt>GlobalValue::hasNotBeenReadFromBitcode</tt> with
952 <tt>GlobalValue::isMaterializable</tt>.</li>
954 <li>The <tt>isInteger</tt>, <tt>isIntOrIntVector</tt>, <tt>isFloatingPoint</tt>,
955 <tt>isFPOrFPVector</tt> and <tt>isFPOrFPVector</tt> methods have been renamed
956 <tt>isIntegerTy</tt>, <tt>isIntOrIntVectorTy</tt>, <tt>isFloatingPointTy</tt>,
957 <tt>isFPOrFPVectorTy</tt> and <tt>isFPOrFPVectorTy</tt> respectively.</li>
959 <li><tt>llvm::Instruction::clone()</tt> no longer takes argument.</li>
960 <li><tt>raw_fd_ostream</tt>'s constructor now takes a flag argument, not individual
961 booleans (see <tt>include/llvm/Support/raw_ostream.h</tt> for details).</li>
962 <li>Some header files have been renamed:
964 <li><tt>llvm/Support/AIXDataTypesFix.h</tt> to
965 <tt>llvm/System/AIXDataTypesFix.h</tt></li>
966 <li><tt>llvm/Support/DataTypes.h</tt> to <tt>llvm/System/DataTypes.h</tt></li>
967 <li><tt>llvm/Transforms/Utils/InlineCost.h</tt> to
968 <tt>llvm/Analysis/InlineCost.h</tt></li>
969 <li><tt>llvm/Support/Mangler.h</tt> to <tt>llvm/Target/Mangler.h</tt></li>
970 <li><tt>llvm/Analysis/Passes.h</tt> to <tt>llvm/CodeGen/Passes.h</tt></li>
978 <!-- *********************************************************************** -->
979 <div class="doc_section">
980 <a name="portability">Portability and Supported Platforms</a>
982 <!-- *********************************************************************** -->
984 <div class="doc_text">
986 <p>LLVM is known to work on the following platforms:</p>
989 <li>Intel and AMD machines (IA32, X86-64, AMD64, EMT-64) running Red Hat
990 Linux, Fedora Core, FreeBSD and AuroraUX (and probably other unix-like
992 <li>PowerPC and X86-based Mac OS X systems, running 10.4 and above in 32-bit
993 and 64-bit modes.</li>
994 <li>Intel and AMD machines running on Win32 using MinGW libraries (native).</li>
995 <li>Intel and AMD machines running on Win32 with the Cygwin libraries (limited
996 support is available for native builds with Visual C++).</li>
997 <li>Sun x86 and AMD64 machines running Solaris 10, OpenSolaris 0906.</li>
998 <li>Alpha-based machines running Debian GNU/Linux.</li>
1001 <p>The core LLVM infrastructure uses GNU autoconf to adapt itself
1002 to the machine and operating system on which it is built. However, minor
1003 porting may be required to get LLVM to work on new platforms. We welcome your
1004 portability patches and reports of successful builds or error messages.</p>
1008 <!-- *********************************************************************** -->
1009 <div class="doc_section">
1010 <a name="knownproblems">Known Problems</a>
1012 <!-- *********************************************************************** -->
1014 <div class="doc_text">
1016 <p>This section contains significant known problems with the LLVM system,
1017 listed by component. If you run into a problem, please check the <a
1018 href="http://llvm.org/bugs/">LLVM bug database</a> and submit a bug if
1019 there isn't already one.</p>
1022 <li>LLVM will not correctly compile on Solaris and/or OpenSolaris
1023 using the stock GCC 3.x.x series 'out the box',
1024 See: <a href="GettingStarted.html#brokengcc">Broken versions of GCC and other tools</a>.
1025 However, A <a href="http://pkg.auroraux.org/GCC">Modern GCC Build</a>
1026 for x86/x86-64 has been made available from the third party AuroraUX Project
1027 that has been meticulously tested for bootstrapping LLVM & Clang.</li>
1032 <!-- ======================================================================= -->
1033 <div class="doc_subsection">
1034 <a name="experimental">Experimental features included with this release</a>
1037 <div class="doc_text">
1039 <p>The following components of this LLVM release are either untested, known to
1040 be broken or unreliable, or are in early development. These components should
1041 not be relied on, and bugs should not be filed against them, but they may be
1042 useful to some people. In particular, if you would like to work on one of these
1043 components, please contact us on the <a
1044 href="http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev">LLVMdev list</a>.</p>
1047 <li>The MSIL, Alpha, SPU, MIPS, PIC16, Blackfin, MSP430, SystemZ and MicroBlaze
1048 backends are experimental.</li>
1049 <li><tt>llc</tt> "<tt>-filetype=asm</tt>" (the default) is the only
1050 supported value for this option. The MachO writer is experimental, and
1051 works much better in mainline SVN.</li>
1056 <!-- ======================================================================= -->
1057 <div class="doc_subsection">
1058 <a name="x86-be">Known problems with the X86 back-end</a>
1061 <div class="doc_text">
1064 <li>The X86 backend does not yet support
1065 all <a href="http://llvm.org/PR879">inline assembly that uses the X86
1066 floating point stack</a>. It supports the 'f' and 't' constraints, but not
1068 <li>The X86 backend generates inefficient floating point code when configured
1069 to generate code for systems that don't have SSE2.</li>
1070 <li>Win64 code generation wasn't widely tested. Everything should work, but we
1071 expect small issues to happen. Also, llvm-gcc cannot build the mingw64
1072 runtime currently due to lack of support for the 'u' inline assembly
1073 constraint and for X87 floating point inline assembly.</li>
1074 <li>The X86-64 backend does not yet support the LLVM IR instruction
1075 <tt>va_arg</tt>. Currently, front-ends support variadic
1076 argument constructs on X86-64 by lowering them manually.</li>
1081 <!-- ======================================================================= -->
1082 <div class="doc_subsection">
1083 <a name="ppc-be">Known problems with the PowerPC back-end</a>
1086 <div class="doc_text">
1089 <li>The Linux PPC32/ABI support needs testing for the interpreter and static
1090 compilation, and lacks support for debug information.</li>
1095 <!-- ======================================================================= -->
1096 <div class="doc_subsection">
1097 <a name="arm-be">Known problems with the ARM back-end</a>
1100 <div class="doc_text">
1103 <li>Thumb mode works only on ARMv6 or higher processors. On sub-ARMv6
1104 processors, thumb programs can crash or produce wrong
1105 results (<a href="http://llvm.org/PR1388">PR1388</a>).</li>
1106 <li>Compilation for ARM Linux OABI (old ABI) is supported but not fully tested.
1112 <!-- ======================================================================= -->
1113 <div class="doc_subsection">
1114 <a name="sparc-be">Known problems with the SPARC back-end</a>
1117 <div class="doc_text">
1120 <li>The SPARC backend only supports the 32-bit SPARC ABI (-m32); it does not
1121 support the 64-bit SPARC ABI (-m64).</li>
1126 <!-- ======================================================================= -->
1127 <div class="doc_subsection">
1128 <a name="mips-be">Known problems with the MIPS back-end</a>
1131 <div class="doc_text">
1134 <li>64-bit MIPS targets are not supported yet.</li>
1139 <!-- ======================================================================= -->
1140 <div class="doc_subsection">
1141 <a name="alpha-be">Known problems with the Alpha back-end</a>
1144 <div class="doc_text">
1148 <li>On 21164s, some rare FP arithmetic sequences which may trap do not have the
1149 appropriate nops inserted to ensure restartability.</li>
1154 <!-- ======================================================================= -->
1155 <div class="doc_subsection">
1156 <a name="c-be">Known problems with the C back-end</a>
1159 <div class="doc_text">
1162 <li><a href="http://llvm.org/PR802">The C backend has only basic support for
1163 inline assembly code</a>.</li>
1164 <li><a href="http://llvm.org/PR1658">The C backend violates the ABI of common
1165 C++ programs</a>, preventing intermixing between C++ compiled by the CBE and
1166 C++ code compiled with <tt>llc</tt> or native compilers.</li>
1167 <li>The C backend does not support all exception handling constructs.</li>
1168 <li>The C backend does not support arbitrary precision integers.</li>
1174 <!-- ======================================================================= -->
1175 <div class="doc_subsection">
1176 <a name="c-fe">Known problems with the llvm-gcc C and C++ front-end</a>
1179 <div class="doc_text">
1181 <p>The only major language feature of GCC not supported by llvm-gcc is
1182 the <tt>__builtin_apply</tt> family of builtins. However, some extensions
1183 are only supported on some targets. For example, trampolines are only
1184 supported on some targets (these are used when you take the address of a
1185 nested function).</p>
1189 <!-- ======================================================================= -->
1190 <div class="doc_subsection">
1191 <a name="fortran-fe">Known problems with the llvm-gcc Fortran front-end</a>
1194 <div class="doc_text">
1196 <li>Fortran support generally works, but there are still several unresolved bugs
1197 in <a href="http://llvm.org/bugs/">Bugzilla</a>. Please see the
1198 tools/gfortran component for details.</li>
1202 <!-- ======================================================================= -->
1203 <div class="doc_subsection">
1204 <a name="ada-fe">Known problems with the llvm-gcc Ada front-end</a>
1207 <div class="doc_text">
1208 The llvm-gcc 4.2 Ada compiler works fairly well; however, this is not a mature
1209 technology, and problems should be expected.
1211 <li>The Ada front-end currently only builds on X86-32. This is mainly due
1212 to lack of trampoline support (pointers to nested functions) on other platforms.
1213 However, it <a href="http://llvm.org/PR2006">also fails to build on X86-64</a>
1214 which does support trampolines.</li>
1215 <li>The Ada front-end <a href="http://llvm.org/PR2007">fails to bootstrap</a>.
1216 This is due to lack of LLVM support for <tt>setjmp</tt>/<tt>longjmp</tt> style
1217 exception handling, which is used internally by the compiler.
1218 Workaround: configure with <tt>--disable-bootstrap</tt>.</li>
1219 <li>The c380004, <a href="http://llvm.org/PR2010">c393010</a>
1220 and <a href="http://llvm.org/PR2421">cxg2021</a> ACATS tests fail
1221 (c380004 also fails with gcc-4.2 mainline).
1222 If the compiler is built with checks disabled then <a href="http://llvm.org/PR2010">c393010</a>
1223 causes the compiler to go into an infinite loop, using up all system memory.</li>
1224 <li>Some GCC specific Ada tests continue to crash the compiler.</li>
1225 <li>The <tt>-E</tt> binder option (exception backtraces)
1226 <a href="http://llvm.org/PR1982">does not work</a> and will result in programs
1227 crashing if an exception is raised. Workaround: do not use <tt>-E</tt>.</li>
1228 <li>Only discrete types <a href="http://llvm.org/PR1981">are allowed to start
1229 or finish at a non-byte offset</a> in a record. Workaround: do not pack records
1230 or use representation clauses that result in a field of a non-discrete type
1231 starting or finishing in the middle of a byte.</li>
1232 <li>The <tt>lli</tt> interpreter <a href="http://llvm.org/PR2009">considers
1233 'main' as generated by the Ada binder to be invalid</a>.
1234 Workaround: hand edit the file to use pointers for <tt>argv</tt> and
1235 <tt>envp</tt> rather than integers.</li>
1236 <li>The <tt>-fstack-check</tt> option <a href="http://llvm.org/PR2008">is
1241 <!-- *********************************************************************** -->
1242 <div class="doc_section">
1243 <a name="additionalinfo">Additional Information</a>
1245 <!-- *********************************************************************** -->
1247 <div class="doc_text">
1249 <p>A wide variety of additional information is available on the <a
1250 href="http://llvm.org">LLVM web page</a>, in particular in the <a
1251 href="http://llvm.org/docs/">documentation</a> section. The web page also
1252 contains versions of the API documentation which is up-to-date with the
1253 Subversion version of the source code.
1254 You can access versions of these documents specific to this release by going
1255 into the "<tt>llvm/doc/</tt>" directory in the LLVM tree.</p>
1257 <p>If you have any questions or comments about LLVM, please feel free to contact
1258 us via the <a href="http://llvm.org/docs/#maillist"> mailing
1263 <!-- *********************************************************************** -->
1267 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
1268 src="http://jigsaw.w3.org/css-validator/images/vcss-blue" alt="Valid CSS"></a>
1269 <a href="http://validator.w3.org/check/referer"><img
1270 src="http://www.w3.org/Icons/valid-html401-blue" alt="Valid HTML 4.01"></a>
1272 <a href="http://llvm.org/">LLVM Compiler Infrastructure</a><br>
1273 Last modified: $Date$