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 2.6 Release Notes</title>
11 <div class="doc_title">LLVM 2.6 Release Notes</div>
14 <li><a href="#intro">Introduction</a></li>
15 <li><a href="#subproj">Sub-project Status Update</a></li>
16 <li><a href="#externalproj">External Projects Using LLVM 2.6</a></li>
17 <li><a href="#whatsnew">What's New in LLVM 2.6?</a></li>
18 <li><a href="GettingStarted.html">Installation Instructions</a></li>
19 <li><a href="#portability">Portability and Supported Platforms</a></li>
20 <li><a href="#knownproblems">Known Problems</a></li>
21 <li><a href="#additionalinfo">Additional Information</a></li>
24 <div class="doc_author">
25 <p>Written by the <a href="http://llvm.org">LLVM Team</a></p>
28 <!-- *********************************************************************** -->
29 <div class="doc_section">
30 <a name="intro">Introduction</a>
32 <!-- *********************************************************************** -->
34 <div class="doc_text">
36 <p>This document contains the release notes for the LLVM Compiler
37 Infrastructure, release 2.6. Here we describe the status of LLVM, including
38 major improvements from the previous release and significant known problems.
39 All LLVM releases may be downloaded from the <a
40 href="http://llvm.org/releases/">LLVM releases web site</a>.</p>
42 <p>For more information about LLVM, including information about the latest
43 release, please check out the <a href="http://llvm.org/">main LLVM
44 web site</a>. If you have questions or comments, the <a
45 href="http://mail.cs.uiuc.edu/mailman/listinfo/llvmdev">LLVM Developer's Mailing
46 List</a> is a good place to send them.</p>
48 <p>Note that if you are reading this file from a Subversion checkout or the
49 main LLVM web page, this document applies to the <i>next</i> release, not the
50 current one. To see the release notes for a specific release, please see the
51 <a href="http://llvm.org/releases/">releases page</a>.</p>
58 include/llvm/Analysis/LiveValues.h => Dan
59 lib/Transforms/IPO/MergeFunctions.cpp => consider for 2.8.
60 llvm/Analysis/PointerTracking.h => Edwin wants this, consider for 2.8.
61 ABCD, SCCVN, GEPSplitterPass
65 <!-- Unfinished features in 2.6:
68 variable debug info for optimized code
69 postalloc scheduler: anti dependence breaking, hazard recognizer?
71 loop dependence analysis
72 ELF Writer? How stable?
73 <li>PostRA scheduler improvements, ARM adoption (David Goodwin).</li>
74 2.7 supports the GDB 7.0 jit interfaces for debug info.
75 2.7 eliminates ADT/iterator.h
78 <!-- for announcement email:
82 KLEE web page at klee.llvm.org
83 Many new papers added to /pubs/
88 <!-- *********************************************************************** -->
89 <div class="doc_section">
90 <a name="subproj">Sub-project Status Update</a>
92 <!-- *********************************************************************** -->
94 <div class="doc_text">
96 The LLVM 2.6 distribution currently consists of code from the core LLVM
97 repository (which roughly includes the LLVM optimizers, code generators
98 and supporting tools), the Clang repository and the llvm-gcc repository. In
99 addition to this code, the LLVM Project includes other sub-projects that are in
100 development. Here we include updates on these subprojects.
106 <!--=========================================================================-->
107 <div class="doc_subsection">
108 <a name="clang">Clang: C/C++/Objective-C Frontend Toolkit</a>
111 <div class="doc_text">
113 <p>The <a href="http://clang.llvm.org/">Clang project</a> is an effort to build
114 a set of new 'LLVM native' front-end technologies for the C family of languages.
115 LLVM 2.6 is the first release to officially include Clang, and it provides a
116 production quality C and Objective-C compiler. If you are interested in <a
117 href="http://clang.llvm.org/performance.html">fast compiles</a> and
118 <a href="http://clang.llvm.org/diagnostics.html">good diagnostics</a>, we
119 encourage you to try it out. Clang currently compiles typical Objective-C code
120 3x faster than GCC and compiles C code about 30% faster than GCC at -O0 -g
121 (which is when the most pressure is on the frontend).</p>
123 <p>In addition to supporting these languages, C++ support is also <a
124 href="http://clang.llvm.org/cxx_status.html">well under way</a>, and mainline
125 Clang is able to parse the libstdc++ 4.2 headers and even codegen simple apps.
126 If you are interested in Clang C++ support or any other Clang feature, we
127 strongly encourage you to get involved on the <a
128 href="http://lists.cs.uiuc.edu/mailman/listinfo/cfe-dev">Clang front-end mailing
131 <p>In the LLVM 2.6 time-frame, the Clang team has made many improvements:</p>
134 <li>C and Objective-C support are now considered production quality.</li>
135 <li>AuroraUX, FreeBSD and OpenBSD are now supported.</li>
136 <li>Most of Objective-C 2.0 is now supported with the GNU runtime.</li>
137 <li>Many many bugs are fixed and lots of features have been added.</li>
141 <!--=========================================================================-->
142 <div class="doc_subsection">
143 <a name="clangsa">Clang Static Analyzer</a>
146 <div class="doc_text">
148 <p>Previously announced in the 2.4 and 2.5 LLVM releases, the Clang project also
149 includes an early stage static source code analysis tool for <a
150 href="http://clang.llvm.org/StaticAnalysis.html">automatically finding bugs</a>
151 in C and Objective-C programs. The tool performs checks to find
152 bugs that occur on a specific path within a program.</p>
154 <p>In the LLVM 2.6 time-frame, the analyzer core has undergone several important
155 improvements and cleanups and now includes a new <em>Checker</em> interface that
156 is intended to eventually serve as a basis for domain-specific checks. Further,
157 in addition to generating HTML files for reporting analysis results, the
158 analyzer can now also emit bug reports in a structured XML format that is
159 intended to be easily readable by other programs.</p>
161 <p>The set of checks performed by the static analyzer continues to expand, and
162 future plans for the tool include full source-level inter-procedural analysis
163 and deeper checks such as buffer overrun detection. There are many opportunities
164 to extend and enhance the static analyzer, and anyone interested in working on
165 this project is encouraged to get involved!</p>
169 <!--=========================================================================-->
170 <div class="doc_subsection">
171 <a name="vmkit">VMKit: JVM/CLI Virtual Machine Implementation</a>
174 <div class="doc_text">
176 The <a href="http://vmkit.llvm.org/">VMKit project</a> is an implementation of
177 a JVM and a CLI Virtual Machine (Microsoft .NET is an
178 implementation of the CLI) using LLVM for static and just-in-time
182 VMKit version 0.26 builds with LLVM 2.6 and you can find it on its
183 <a href="http://vmkit.llvm.org/releases/">web page</a>. The release includes
184 bug fixes, cleanup and new features. The major changes are:</p>
188 <li>A new llcj tool to generate shared libraries or executables of Java
190 <li>Cooperative garbage collection. </li>
191 <li>Fast subtype checking (paper from Click et al [JGI'02]). </li>
192 <li>Implementation of a two-word header for Java objects instead of the original
193 three-word header. </li>
194 <li>Better Java specification-compliance: division by zero checks, stack
195 overflow checks, finalization and references support. </li>
201 <!--=========================================================================-->
202 <div class="doc_subsection">
203 <a name="compiler-rt">compiler-rt: Compiler Runtime Library</a>
206 <div class="doc_text">
208 The new LLVM <a href="http://compiler-rt.llvm.org/">compiler-rt project</a>
209 is a simple library that provides an implementation of the low-level
210 target-specific hooks required by code generation and other runtime components.
211 For example, when compiling for a 32-bit target, converting a double to a 64-bit
212 unsigned integer is compiled into a runtime call to the "__fixunsdfdi"
213 function. The compiler-rt library provides highly optimized implementations of
214 this and other low-level routines (some are 3x faster than the equivalent
215 libgcc routines).</p>
218 All of the code in the compiler-rt project is available under the standard LLVM
219 License, a "BSD-style" license.</p>
223 <!--=========================================================================-->
224 <div class="doc_subsection">
225 <a name="klee">KLEE: Symbolic Execution and Automatic Test Case Generator</a>
228 <div class="doc_text">
230 The new LLVM <a href="http://klee.llvm.org/">KLEE project</a> is a symbolic
231 execution framework for programs in LLVM bitcode form. KLEE tries to
232 symbolically evaluate "all" paths through the application and records state
233 transitions that lead to fault states. This allows it to construct testcases
234 that lead to faults and can even be used to verify algorithms. For more
235 details, please see the <a
236 href="http://llvm.org/pubs/2008-12-OSDI-KLEE.html">OSDI 2008 paper</a> about
241 <!--=========================================================================-->
242 <div class="doc_subsection">
243 <a name="dragonegg">DragonEgg: GCC-4.5 as an LLVM frontend</a>
246 <div class="doc_text">
248 The goal of <a href="http://dragonegg.llvm.org/">DragonEgg</a> is to make
249 gcc-4.5 act like llvm-gcc without requiring any gcc modifications whatsoever.
250 <a href="http://dragonegg.llvm.org/">DragonEgg</a> is a shared library (dragonegg.so)
251 that is loaded by gcc at runtime. It uses the new gcc plugin architecture to
252 disable the GCC optimizers and code generators, and schedule the LLVM optimizers
253 and code generators (or direct output of LLVM IR) instead. Currently only Linux
254 and Darwin are supported, and only on x86-32 and x86-64. It should be easy to
255 add additional unix-like architectures and other processor families. In theory
256 it should be possible to use <a href="http://dragonegg.llvm.org/">DragonEgg</a>
257 with any language supported by gcc, however only C and Fortran work well for the
258 moment. Ada and C++ work to some extent, while Java, Obj-C and Obj-C++ are so
259 far entirely untested. Since gcc-4.5 has not yet been released, neither has
260 <a href="http://dragonegg.llvm.org/">DragonEgg</a>. To build
261 <a href="http://dragonegg.llvm.org/">DragonEgg</a> you will need to check out the
262 development versions of <a href="http://gcc.gnu.org/svn.html/"> gcc</a>,
263 <a href="http://llvm.org/docs/GettingStarted.html#checkout">llvm</a> and
264 <a href="http://dragonegg.llvm.org/">DragonEgg</a> from their respective
265 subversion repositories, and follow the instructions in the
266 <a href="http://dragonegg.llvm.org/">DragonEgg</a> README.
272 <!--=========================================================================-->
273 <div class="doc_subsection">
274 <a name="mc">llvm-mc: Machine Code Toolkit</a>
277 <div class="doc_text">
279 The LLVM Machine Code (MC) Toolkit project is a (very early) effort to build
280 better tools for dealing with machine code, object file formats, etc. The idea
281 is to be able to generate most of the target specific details of assemblers and
282 disassemblers from existing LLVM target .td files (with suitable enhancements),
283 and to build infrastructure for reading and writing common object file formats.
284 One of the first deliverables is to build a full assembler and integrate it into
285 the compiler, which is predicted to substantially reduce compile time in some
289 <p>In the LLVM 2.6 timeframe, the MC framework has grown to the point where it
290 can reliably parse and pretty print (with some encoding information) a
291 darwin/x86 .s file successfully, and has the very early phases of a Mach-O
292 assembler in progress. Beyond the MC framework itself, major refactoring of the
293 LLVM code generator has started. The idea is to make the code generator reason
294 about the code it is producing in a much more semantic way, rather than a
295 textual way. For example, the code generator now uses MCSection objects to
296 represent section assignments, instead of text strings that print to .section
299 <p>MC is an early and ongoing project that will hopefully continue to lead to
300 many improvements in the code generator and build infrastructure useful for many
307 <!-- *********************************************************************** -->
308 <div class="doc_section">
309 <a name="externalproj">External Open Source Projects Using LLVM 2.6</a>
311 <!-- *********************************************************************** -->
313 <div class="doc_text">
315 <p>An exciting aspect of LLVM is that it is used as an enabling technology for
316 a lot of other language and tools projects. This section lists some of the
317 projects that have already been updated to work with LLVM 2.6.</p>
321 <!--=========================================================================-->
322 <div class="doc_subsection">
323 <a name="Rubinius">Rubinius</a>
326 <div class="doc_text">
327 <p><a href="http://github.com/evanphx/rubinius">Rubinius</a> is an environment
328 for running Ruby code which strives to write as much of the core class
329 implementation in Ruby as possible. Combined with a bytecode interpreting VM, it
330 uses LLVM to optimize and compile ruby code down to machine code. Techniques
331 such as type feedback, method inlining, and uncommon traps are all used to
332 remove dynamism from ruby execution and increase performance.</p>
334 <p>Since LLVM 2.5, Rubinius has made several major leaps forward, implementing
335 a counter based JIT, type feedback and speculative method inlining.
340 <!--=========================================================================-->
341 <div class="doc_subsection">
342 <a name="macruby">MacRuby</a>
345 <div class="doc_text">
348 <a href="http://macruby.org">MacRuby</a> is an implementation of Ruby on top of
349 core Mac OS X technologies, such as the Objective-C common runtime and garbage
350 collector and the CoreFoundation framework. It is principally developed by
351 Apple and aims at enabling the creation of full-fledged Mac OS X applications.
355 MacRuby uses LLVM for optimization passes, JIT and AOT compilation of Ruby
356 expressions. It also uses zero-cost DWARF exceptions to implement Ruby exception
362 <!--=========================================================================-->
363 <div class="doc_subsection">
364 <a name="pure">Pure</a>
367 <div class="doc_text">
369 <a href="http://pure-lang.googlecode.com/">Pure</a>
370 is an algebraic/functional programming language based on term rewriting.
371 Programs are collections of equations which are used to evaluate expressions in
372 a symbolic fashion. Pure offers dynamic typing, eager and lazy evaluation,
373 lexical closures, a hygienic macro system (also based on term rewriting),
374 built-in list and matrix support (including list and matrix comprehensions) and
375 an easy-to-use C interface. The interpreter uses LLVM as a backend to
376 JIT-compile Pure programs to fast native code.</p>
378 <p>Pure versions 0.31 and later have been tested and are known to work with
379 LLVM 2.6 (and continue to work with older LLVM releases >= 2.3 as well).
384 <!--=========================================================================-->
385 <div class="doc_subsection">
386 <a name="ldc">LLVM D Compiler</a>
389 <div class="doc_text">
391 <a href="http://www.dsource.org/projects/ldc">LDC</a> is an implementation of
392 the D Programming Language using the LLVM optimizer and code generator.
393 The LDC project works great with the LLVM 2.6 release. General improvements in
395 cycle have included new inline asm constraint handling, better debug info
396 support, general bug fixes and better x86-64 support. This has allowed
397 some major improvements in LDC, getting it much closer to being as
398 fully featured as the original DMD compiler from DigitalMars.
402 <!--=========================================================================-->
403 <div class="doc_subsection">
404 <a name="RoadsendPHP">Roadsend PHP</a>
407 <div class="doc_text">
409 <a href="http://code.roadsend.com/rphp">Roadsend PHP</a> (rphp) is an open
410 source implementation of the PHP programming
411 language that uses LLVM for its optimizer, JIT and static compiler. This is a
412 reimplementation of an earlier project that is now based on LLVM.</p>
415 <!--=========================================================================-->
416 <div class="doc_subsection">
417 <a name="UnladenSwallow">Unladen Swallow</a>
420 <div class="doc_text">
422 <a href="http://code.google.com/p/unladen-swallow/">Unladen Swallow</a> is a
423 branch of <a href="http://python.org/">Python</a> intended to be fully
424 compatible and significantly faster. It uses LLVM's optimization passes and JIT
428 <!--=========================================================================-->
429 <div class="doc_subsection">
430 <a name="llvm-lua">llvm-lua</a>
433 <div class="doc_text">
435 <a href="http://code.google.com/p/llvm-lua/">LLVM-Lua</a> uses LLVM to add JIT
436 and static compiling support to the Lua VM. Lua bytecode is analyzed to
437 remove type checks, then LLVM is used to compile the bytecode down to machine
441 <!--=========================================================================-->
442 <div class="doc_subsection">
443 <a name="icedtea">IcedTea Java Virtual Machine Implementation</a>
446 <div class="doc_text">
448 <a href="http://icedtea.classpath.org/wiki/Main_Page">IcedTea</a> provides a
449 harness to build OpenJDK using only free software build tools and to provide
450 replacements for the not-yet free parts of OpenJDK. One of the extensions that
451 IcedTea provides is a new JIT compiler named <a
452 href="http://icedtea.classpath.org/wiki/ZeroSharkFaq">Shark</a> which uses LLVM
453 to provide native code generation without introducing processor-dependent
460 <!-- *********************************************************************** -->
461 <div class="doc_section">
462 <a name="whatsnew">What's New in LLVM 2.6?</a>
464 <!-- *********************************************************************** -->
466 <div class="doc_text">
468 <p>This release includes a huge number of bug fixes, performance tweaks and
469 minor improvements. Some of the major improvements and new features are listed
475 <!--=========================================================================-->
476 <div class="doc_subsection">
477 <a name="majorfeatures">Major New Features</a>
480 <div class="doc_text">
482 <p>LLVM 2.6 includes several major new capabilities:</p>
485 <li>New <a href="#compiler-rt">compiler-rt</a>, <A href="#klee">KLEE</a>
486 and <a href="#mc">machine code toolkit</a> sub-projects.</li>
487 <li>Debug information now includes line numbers when optimizations are enabled.
488 This allows statistical sampling tools like OProfile and Shark to map
489 samples back to source lines.</li>
490 <li>LLVM now includes new experimental backends to support the MSP430, SystemZ
491 and BlackFin architectures.</li>
492 <li>LLVM supports a new <a href="GoldPlugin.html">Gold Linker Plugin</a> which
493 enables support for <a href="LinkTimeOptimization.html">transparent
494 link-time optimization</a> on ELF targets when used with the Gold binutils
496 <li>LLVM now supports doing optimization and code generation on multiple
497 threads. Please see the <a href="ProgrammersManual.html#threading">LLVM
498 Programmer's Manual</a> for more information.</li>
499 <li>LLVM now has experimental support for <a
500 href="http://nondot.org/~sabre/LLVMNotes/EmbeddedMetadata.txt">embedded
501 metadata</a> in LLVM IR, though the implementation is not guaranteed to be
502 final and the .bc file format may change in future releases. Debug info
503 does not yet use this format in LLVM 2.6.</li>
508 <!--=========================================================================-->
509 <div class="doc_subsection">
510 <a name="coreimprovements">LLVM IR and Core Improvements</a>
513 <div class="doc_text">
514 <p>LLVM IR has several new features for better support of new targets and that
515 expose new optimization opportunities:</p>
518 <li>The <a href="LangRef.html#i_add">add</a>, <a
519 href="LangRef.html#i_sub">sub</a> and <a href="LangRef.html#i_mul">mul</a>
520 instructions have been split into integer and floating point versions (like
521 divide and remainder), introducing new <a
522 href="LangRef.html#i_fadd">fadd</a>, <a href="LangRef.html#i_fsub">fsub</a>,
523 and <a href="LangRef.html#i_fmul">fmul</a> instructions.</li>
524 <li>The <a href="LangRef.html#i_add">add</a>, <a
525 href="LangRef.html#i_sub">sub</a> and <a href="LangRef.html#i_mul">mul</a>
526 instructions now support optional "nsw" and "nuw" bits which indicate that
527 the operation is guaranteed to not overflow (in the signed or
528 unsigned case, respectively). This gives the optimizer more information and
529 can be used for things like C signed integer values, which are undefined on
531 <li>The <a href="LangRef.html#i_sdiv">sdiv</a> instruction now supports an
532 optional "exact" flag which indicates that the result of the division is
533 guaranteed to have a remainder of zero. This is useful for optimizing pointer
534 subtraction in C.</li>
535 <li>The <a href="LangRef.html#i_getelementptr">getelementptr</a> instruction now
536 supports arbitrary integer index values for array/pointer indices. This
537 allows for better code generation on 16-bit pointer targets like PIC16.</li>
538 <li>The <a href="LangRef.html#i_getelementptr">getelementptr</a> instruction now
539 supports an "inbounds" optimization hint that tells the optimizer that the
540 pointer is guaranteed to be within its allocated object.</li>
541 <li>LLVM now support a series of new linkage types for global values which allow
542 for better optimization and new capabilities:
544 <li><a href="LangRef.html#linkage_linkonce">linkonce_odr</a> and
545 <a href="LangRef.html#linkage_weak">weak_odr</a> have the same linkage
546 semantics as the non-"odr" linkage types. The difference is that these
547 linkage types indicate that all definitions of the specified function
548 are guaranteed to have the same semantics. This allows inlining
549 templates functions in C++ but not inlining weak functions in C,
550 which previously both got the same linkage type.</li>
551 <li><a href="LangRef.html#linkage_available_externally">available_externally
552 </a> is a new linkage type that gives the optimizer visibility into the
553 definition of a function (allowing inlining and side effect analysis)
554 but that does not cause code to be generated. This allows better
555 optimization of "GNU inline" functions, extern templates, etc.</li>
556 <li><a href="LangRef.html#linkage_linker_private">linker_private</a> is a
557 new linkage type (which is only useful on Mac OS X) that is used for
558 some metadata generation and other obscure things.</li>
560 <li>Finally, target-specific intrinsics can now return multiple values, which
561 is useful for modeling target operations with multiple results.</li>
566 <!--=========================================================================-->
567 <div class="doc_subsection">
568 <a name="optimizer">Optimizer Improvements</a>
571 <div class="doc_text">
573 <p>In addition to a large array of minor performance tweaks and bug fixes, this
574 release includes a few major enhancements and additions to the optimizers:</p>
578 <li>The <a href="Passes.html#scalarrepl">Scalar Replacement of Aggregates</a>
579 pass has many improvements that allow it to better promote vector unions,
580 variables which are memset, and much more strange code that can happen to
581 do bitfield accesses to register operations. An interesting change is that
582 it now produces "unusual" integer sizes (like i1704) in some cases and lets
583 other optimizers clean things up.</li>
584 <li>The <a href="Passes.html#loop-reduce">Loop Strength Reduction</a> pass now
585 promotes small integer induction variables to 64-bit on 64-bit targets,
586 which provides a major performance boost for much numerical code. It also
587 promotes shorts to int on 32-bit hosts, etc. LSR now also analyzes pointer
588 expressions (e.g. getelementptrs), as well as integers.</li>
589 <li>The <a href="Passes.html#gvn">GVN</a> pass now eliminates partial
590 redundancies of loads in simple cases.</li>
591 <li>The <a href="Passes.html#inline">Inliner</a> now reuses stack space when
592 inlining similar arrays from multiple callees into one caller.</li>
593 <li>LLVM includes a new experimental Static Single Information (SSI)
594 construction pass.</li>
601 <!--=========================================================================-->
602 <div class="doc_subsection">
603 <a name="executionengine">Interpreter and JIT Improvements</a>
606 <div class="doc_text">
609 <li>LLVM has a new "EngineBuilder" class which makes it more obvious how to
610 set up and configure an ExecutionEngine (a JIT or interpreter).</li>
611 <li>The JIT now supports generating more than 16M of code.</li>
612 <li>When configured with <tt>--with-oprofile</tt>, the JIT can now inform
613 OProfile about JIT'd code, allowing OProfile to get line number and function
614 name information for JIT'd functions.</li>
615 <li>When "libffi" is available, the LLVM interpreter now uses it, which supports
616 calling almost arbitrary external (natively compiled) functions.</li>
617 <li>Clients of the JIT can now register a 'JITEventListener' object to receive
618 callbacks when the JIT emits or frees machine code. The OProfile support
619 uses this mechanism.</li>
624 <!--=========================================================================-->
625 <div class="doc_subsection">
626 <a name="codegen">Target Independent Code Generator Improvements</a>
629 <div class="doc_text">
631 <p>We have put a significant amount of work into the code generator
632 infrastructure, which allows us to implement more aggressive algorithms and make
637 <li>The <tt>llc -asm-verbose</tt> option (exposed from llvm-gcc as <tt>-dA</tt>
638 and clang as <tt>-fverbose-asm</tt> or <tt>-dA</tt>) now adds a lot of
639 useful information in comments to
640 the generated .s file. This information includes location information (if
641 built with <tt>-g</tt>) and loop nest information.</li>
642 <li>The code generator now supports a new MachineVerifier pass which is useful
643 for finding bugs in targets and codegen passes.</li>
644 <li>The Machine LICM is now enabled by default. It hoists instructions out of
645 loops (such as constant pool loads, loads from read-only stubs, vector
646 constant synthesization code, etc.) and is currently configured to only do
647 so when the hoisted operation can be rematerialized.</li>
648 <li>The Machine Sinking pass is now enabled by default. This pass moves
649 side-effect free operations down the CFG so that they are executed on fewer
650 paths through a function.</li>
651 <li>The code generator now performs "stack slot coloring" of register spills,
652 which allows spill slots to be reused. This leads to smaller stack frames
653 in cases where there are lots of register spills.</li>
654 <li>The register allocator has many improvements to take better advantage of
655 commutable operations, various spiller peephole optimizations, and can now
656 coalesce cross-register-class copies.</li>
657 <li>Tblgen now supports multiclass inheritance and a number of new string and
658 list operations like <tt>!(subst)</tt>, <tt>!(foreach)</tt>, <tt>!car</tt>,
659 <tt>!cdr</tt>, <tt>!null</tt>, <tt>!if</tt>, <tt>!cast</tt>.
660 These make the .td files more expressive and allow more aggressive factoring
661 of duplication across instruction patterns.</li>
662 <li>Target-specific intrinsics can now be added without having to hack VMCore to
663 add them. This makes it easier to maintain out-of-tree targets.</li>
664 <li>The instruction selector is better at propagating information about values
665 (such as whether they are sign/zero extended etc.) across basic block
667 <li>The SelectionDAG datastructure has new nodes for representing buildvector
668 and <a href="http://llvm.org/PR2957">vector shuffle</a> operations. This
669 makes operations and pattern matching more efficient and easier to get
671 <li>The Prolog/Epilog Insertion Pass now has experimental support for performing
672 the "shrink wrapping" optimization, which moves spills and reloads around in
673 the CFG to avoid doing saves on paths that don't need them.</li>
674 <li>LLVM includes new experimental support for writing ELF .o files directly
675 from the compiler. It works well for many simple C testcases, but doesn't
676 support exception handling, debug info, inline assembly, etc.</li>
677 <li>Targets can now specify register allocation hints through
678 <tt>MachineRegisterInfo::setRegAllocationHint</tt>. A regalloc hint consists
679 of hint type and physical register number. A hint type of zero specifies a
680 register allocation preference. Other hint type values are target specific
681 which are resolved by <tt>TargetRegisterInfo::ResolveRegAllocHint</tt>. An
682 example is the ARM target which uses register hints to request that the
683 register allocator provide an even / odd register pair to two virtual
688 <!--=========================================================================-->
689 <div class="doc_subsection">
690 <a name="x86">X86-32 and X86-64 Target Improvements</a>
693 <div class="doc_text">
694 <p>New features of the X86 target include:
699 <li>SSE 4.2 builtins are now supported.</li>
700 <li>GCC-compatible soft float modes are now supported, which are typically used
702 <li>X86-64 now models implicit zero extensions better, which allows the code
703 generator to remove a lot of redundant zexts. It also models the 8-bit "H"
704 registers as subregs, which allows them to be used in some tricky
706 <li>X86-64 now supports the "local exec" and "initial exec" thread local storage
708 <li>The vector forms of the <a href="LangRef.html#i_icmp">icmp</a> and <a
709 href="LangRef.html#i_fcmp">fcmp</a> instructions now select to efficient
711 <li>Support for the win64 calling conventions have improved. The primary
712 missing feature is support for varargs function definitions. It seems to
713 work well for many win64 JIT purposes.</li>
714 <li>The X86 backend has preliminary support for <a
715 href="CodeGenerator.html#x86_memory">mapping address spaces to segment
716 register references</a>. This allows you to write GS or FS relative memory
717 accesses directly in LLVM IR for cases where you know exactly what you're
718 doing (such as in an OS kernel). There are some known problems with this
719 support, but it works in simple cases.</li>
720 <li>The X86 code generator has been refactored to move all global variable
721 reference logic to one place
722 (<tt>X86Subtarget::ClassifyGlobalReference</tt>) which
723 makes it easier to reason about.</li>
729 <!--=========================================================================-->
730 <div class="doc_subsection">
731 <a name="pic16">PIC16 Target Improvements</a>
734 <div class="doc_text">
735 <p>New features of the PIC16 target include:
739 <li>Support for floating-point, indirect function calls, and
740 passing/returning aggregate types to functions.
741 <li>The code generator is able to generate debug info into output COFF files.
742 <li>Support for placing an object into a specific section or at a specific
743 address in memory.</li>
746 <p>Things not yet supported:</p>
749 <li>Variable arguments.</li>
750 <li>Interrupts/programs.</li>
755 <!--=========================================================================-->
756 <div class="doc_subsection">
757 <a name="ARM">ARM Target Improvements</a>
760 <div class="doc_text">
761 <p>New features of the ARM target include:
766 <li>Preliminary support for processors, such as the Cortex-A8 and Cortex-A9,
767 that implement version v7-A of the ARM architecture. The ARM backend now
768 supports both the Thumb2 and Advanced SIMD (Neon) instruction sets.</li>
770 <li>The AAPCS-VFP "hard float" calling conventions are also supported with the
771 <tt>-float-abi=hard</tt> flag.</li>
773 <li>The ARM calling convention code is now tblgen generated instead of resorting
777 <p>These features are still somewhat experimental
778 and subject to change. The Neon intrinsics, in particular, may change in future
779 releases of LLVM. ARMv7 support has progressed a lot on top of tree since 2.6
785 <!--=========================================================================-->
786 <div class="doc_subsection">
787 <a name="OtherTarget">Other Target Specific Improvements</a>
790 <div class="doc_text">
791 <p>New features of other targets include:
795 <li>Mips now supports O32 Calling Convention.</li>
796 <li>Many improvements to the 32-bit PowerPC SVR4 ABI (used on powerpc-linux)
797 support, lots of bugs fixed.</li>
798 <li>Added support for the 64-bit PowerPC SVR4 ABI (used on powerpc64-linux).
799 Needs more testing.</li>
804 <!--=========================================================================-->
805 <div class="doc_subsection">
806 <a name="newapis">New Useful APIs</a>
809 <div class="doc_text">
811 <p>This release includes a number of new APIs that are used internally, which
812 may also be useful for external clients.
816 <li>New <a href="http://llvm.org/doxygen/PrettyStackTrace_8h-source.html">
817 <tt>PrettyStackTrace</tt> class</a> allows crashes of llvm tools (and applications
818 that integrate them) to provide more detailed indication of what the
819 compiler was doing at the time of the crash (e.g. running a pass).
820 At the top level for each LLVM tool, it includes the command line arguments.
822 <li>New <a href="http://llvm.org/doxygen/StringRef_8h-source.html">StringRef</a>
823 and <a href="http://llvm.org/doxygen/Twine_8h-source.html">Twine</a> classes
824 make operations on character ranges and
825 string concatenation to be more efficient. <tt>StringRef</tt> is just a <tt>const
826 char*</tt> with a length, <tt>Twine</tt> is a light-weight rope.</li>
827 <li>LLVM has new <tt>WeakVH</tt>, <tt>AssertingVH</tt> and <tt>CallbackVH</tt>
828 classes, which make it easier to write LLVM IR transformations. <tt>WeakVH</tt>
829 is automatically drops to null when the referenced <tt>Value</tt> is deleted,
830 and is updated across a <tt>replaceAllUsesWith</tt> operation.
831 <tt>AssertingVH</tt> aborts the program if the
832 referenced value is destroyed while it is being referenced. <tt>CallbackVH</tt>
833 is a customizable class for handling value references. See <a
834 href="http://llvm.org/doxygen/ValueHandle_8h-source.html">ValueHandle.h</a>
835 for more information.</li>
836 <li>The new '<a href="http://llvm.org/doxygen/Triple_8h-source.html">Triple
837 </a>' class centralizes a lot of logic that reasons about target
839 <li>The new '<a href="http://llvm.org/doxygen/ErrorHandling_8h-source.html">
840 llvm_report_error()</a>' set of APIs allows tools to embed the LLVM
841 optimizer and backend and recover from previously unrecoverable errors.</li>
842 <li>LLVM has new abstractions for <a
843 href="http://llvm.org/doxygen/Atomic_8h-source.html">atomic operations</a>
844 and <a href="http://llvm.org/doxygen/RWMutex_8h-source.html">reader/writer
846 <li>LLVM has new <a href="http://llvm.org/doxygen/SourceMgr_8h-source.html">
847 <tt>SourceMgr</tt> and <tt>SMLoc</tt> classes</a> which implement caret
848 diagnostics and basic include stack processing for simple parsers. It is
849 used by tablegen, llvm-mc, the .ll parser and FileCheck.</li>
855 <!--=========================================================================-->
856 <div class="doc_subsection">
857 <a name="otherimprovements">Other Improvements and New Features</a>
860 <div class="doc_text">
861 <p>Other miscellaneous features include:</p>
864 <li>LLVM now includes a new internal '<a
865 href="http://llvm.org/cmds/FileCheck.html">FileCheck</a>' tool which allows
866 writing much more accurate regression tests that run faster. Please see the
867 <a href="TestingGuide.html#FileCheck">FileCheck section of the Testing
868 Guide</a> for more information.</li>
869 <li>LLVM profile information support has been significantly improved to produce
870 correct use counts, and has support for edge profiling with reduced runtime
871 overhead. Combined, the generated profile information is both more correct and
872 imposes about half as much overhead (2.6. from 12% to 6% overhead on SPEC
874 <li>The C bindings (in the llvm/include/llvm-c directory) include many newly
876 <li>LLVM 2.6 includes a brand new experimental LLVM bindings to the Ada2005
877 programming language.</li>
879 <li>The LLVMC driver has several new features:
881 <li>Dynamic plugins now work on Windows.</li>
882 <li>New option property: init. Makes possible to provide default values for
883 options defined in plugins (interface to <tt>cl::init</tt>).</li>
884 <li>New example: Skeleton, shows how to create a standalone LLVMC-based
886 <li>New example: mcc16, a driver for the PIC16 toolchain.</li>
895 <!--=========================================================================-->
896 <div class="doc_subsection">
897 <a name="changes">Major Changes and Removed Features</a>
900 <div class="doc_text">
902 <p>If you're already an LLVM user or developer with out-of-tree changes based
903 on LLVM 2.5, this section lists some "gotchas" that you may run into upgrading
904 from the previous release.</p>
907 <li>The Itanium (IA64) backend has been removed. It was not actively supported
908 and had bitrotted.</li>
909 <li>The BigBlock register allocator has been removed, it had also bitrotted.</li>
910 <li>The C Backend (<tt>-march=c</tt>) is no longer considered part of the LLVM release
911 criteria. We still want it to work, but no one is maintaining it and it lacks
912 support for arbitrary precision integers and other important IR features.</li>
914 <li>All LLVM tools now default to overwriting their output file, behaving more
915 like standard unix tools. Previously, this only happened with the '<tt>-f</tt>'
917 <li>LLVM build now builds all libraries as .a files instead of some
918 libraries as relinked .o files. This requires some APIs like
919 InitializeAllTargets.h.
924 <p>In addition, many APIs have changed in this release. Some of the major LLVM
928 <li>All uses of <tt>hash_set</tt> and <tt>hash_map</tt> have been removed from
929 the LLVM tree and the wrapper headers have been removed.</li>
930 <li>The llvm/Streams.h and <tt>DOUT</tt> member of Debug.h have been removed. The
931 <tt>llvm::Ostream</tt> class has been completely removed and replaced with
932 uses of <tt>raw_ostream</tt>.</li>
933 <li>LLVM's global uniquing tables for <tt>Type</tt>s and <tt>Constant</tt>s have
934 been privatized into members of an <tt>LLVMContext</tt>. A number of APIs
935 now take an <tt>LLVMContext</tt> as a parameter. To smooth the transition
936 for clients that will only ever use a single context, the new
937 <tt>getGlobalContext()</tt> API can be used to access a default global
938 context which can be passed in any and all cases where a context is
940 <li>The <tt>getABITypeSize</tt> methods are now called <tt>getAllocSize</tt>.</li>
941 <li>The <tt>Add</tt>, <tt>Sub</tt> and <tt>Mul</tt> operators are no longer
942 overloaded for floating-point types. Floating-point addition, subtraction
943 and multiplication are now represented with new operators <tt>FAdd</tt>,
944 <tt>FSub</tt> and <tt>FMul</tt>. In the <tt>IRBuilder</tt> API,
945 <tt>CreateAdd</tt>, <tt>CreateSub</tt>, <tt>CreateMul</tt> and
946 <tt>CreateNeg</tt> should only be used for integer arithmetic now;
947 <tt>CreateFAdd</tt>, <tt>CreateFSub</tt>, <tt>CreateFMul</tt> and
948 <tt>CreateFNeg</tt> should now be used for floating-point arithmetic.</li>
949 <li>The <tt>DynamicLibrary</tt> class can no longer be constructed, its functionality has
950 moved to static member functions.</li>
951 <li><tt>raw_fd_ostream</tt>'s constructor for opening a given filename now
952 takes an extra <tt>Force</tt> argument. If <tt>Force</tt> is set to
953 <tt>false</tt>, an error will be reported if a file with the given name
954 already exists. If <tt>Force</tt> is set to <tt>true</tt>, the file will
955 be silently truncated (which is the behavior before this flag was
957 <li><tt>SCEVHandle</tt> no longer exists, because reference counting is no
958 longer done for <tt>SCEV*</tt> objects, instead <tt>const SCEV*</tt>
961 <li>Many APIs, notably <tt>llvm::Value</tt>, now use the <tt>StringRef</tt>
962 and <tt>Twine</tt> classes instead of passing <tt>const char*</tt>
963 or <tt>std::string</tt>, as described in
964 the <a href="ProgrammersManual.html#string_apis">Programmer's Manual</a>. Most
965 clients should be unaffected by this transition, unless they are used to
966 <tt>Value::getName()</tt> returning a string. Here are some tips on updating to
969 <li><tt>getNameStr()</tt> is still available, and matches the old
970 behavior. Replacing <tt>getName()</tt> calls with this is an safe option,
971 although more efficient alternatives are now possible.</li>
973 <li>If you were just relying on <tt>getName()</tt> being able to be sent to
974 a <tt>std::ostream</tt>, consider migrating
975 to <tt>llvm::raw_ostream</tt>.</li>
977 <li>If you were using <tt>getName().c_str()</tt> to get a <tt>const
978 char*</tt> pointer to the name, you can use <tt>getName().data()</tt>.
979 Note that this string (as before), may not be the entire name if the
980 name contains embedded null characters.</li>
982 <li>If you were using <tt>operator +</tt> on the result of <tt>getName()</tt> and
983 treating the result as an <tt>std::string</tt>, you can either
984 use <tt>Twine::str</tt> to get the result as an <tt>std::string</tt>, or
985 could move to a <tt>Twine</tt> based design.</li>
987 <li><tt>isName()</tt> should be replaced with comparison
988 against <tt>getName()</tt> (this is now efficient).
992 <li>The registration interfaces for backend Targets has changed (what was
993 previously <tt>TargetMachineRegistry</tt>). For backend authors, see the <a
994 href="WritingAnLLVMBackend.html#TargetRegistration">Writing An LLVM Backend</a>
995 guide. For clients, the notable API changes are:
997 <li><tt>TargetMachineRegistry</tt> has been renamed
998 to <tt>TargetRegistry</tt>.</li>
1000 <li>Clients should move to using the <tt>TargetRegistry::lookupTarget()</tt>
1001 function to find targets.</li>
1010 <!-- *********************************************************************** -->
1011 <div class="doc_section">
1012 <a name="portability">Portability and Supported Platforms</a>
1014 <!-- *********************************************************************** -->
1016 <div class="doc_text">
1018 <p>LLVM is known to work on the following platforms:</p>
1021 <li>Intel and AMD machines (IA32, X86-64, AMD64, EMT-64) running Red Hat
1022 Linux, Fedora Core, FreeBSD and AuroraUX (and probably other unix-like
1024 <li>PowerPC and X86-based Mac OS X systems, running 10.3 and above in 32-bit
1025 and 64-bit modes.</li>
1026 <li>Intel and AMD machines running on Win32 using MinGW libraries (native).</li>
1027 <li>Intel and AMD machines running on Win32 with the Cygwin libraries (limited
1028 support is available for native builds with Visual C++).</li>
1029 <li>Sun x86 and AMD64 machines running Solaris 10, OpenSolaris 0906.</li>
1030 <li>Alpha-based machines running Debian GNU/Linux.</li>
1033 <p>The core LLVM infrastructure uses GNU autoconf to adapt itself
1034 to the machine and operating system on which it is built. However, minor
1035 porting may be required to get LLVM to work on new platforms. We welcome your
1036 portability patches and reports of successful builds or error messages.</p>
1040 <!-- *********************************************************************** -->
1041 <div class="doc_section">
1042 <a name="knownproblems">Known Problems</a>
1044 <!-- *********************************************************************** -->
1046 <div class="doc_text">
1048 <p>This section contains significant known problems with the LLVM system,
1049 listed by component. If you run into a problem, please check the <a
1050 href="http://llvm.org/bugs/">LLVM bug database</a> and submit a bug if
1051 there isn't already one.</p>
1054 <li>The llvm-gcc bootstrap will fail with some versions of binutils (e.g. 2.15)
1055 with a message of "<tt><a href="http://llvm.org/PR5004">Error: can not do 8
1056 byte pc-relative relocation</a></tt>" when building C++ code. We intend to
1057 fix this on mainline, but a workaround for 2.6 is to upgrade to binutils
1060 <li>LLVM will not correctly compile on Solaris and/or OpenSolaris
1061 using the stock GCC 3.x.x series 'out the box',
1062 See: <a href="GettingStarted.html#brokengcc">Broken versions of GCC and other tools</a>.
1063 However, A <a href="http://pkg.auroraux.org/GCC">Modern GCC Build</a>
1064 for x86/x86-64 has been made available from the third party AuroraUX Project
1065 that has been meticulously tested for bootstrapping LLVM & Clang.</li>
1070 <!-- ======================================================================= -->
1071 <div class="doc_subsection">
1072 <a name="experimental">Experimental features included with this release</a>
1075 <div class="doc_text">
1077 <p>The following components of this LLVM release are either untested, known to
1078 be broken or unreliable, or are in early development. These components should
1079 not be relied on, and bugs should not be filed against them, but they may be
1080 useful to some people. In particular, if you would like to work on one of these
1081 components, please contact us on the <a
1082 href="http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev">LLVMdev list</a>.</p>
1085 <li>The MSIL, Alpha, SPU, MIPS, PIC16, Blackfin, MSP430 and SystemZ backends are
1087 <li>The <tt>llc</tt> "<tt>-filetype=asm</tt>" (the default) is the only
1088 supported value for this option. The ELF writer is experimental.</li>
1089 <li>The implementation of Andersen's Alias Analysis has many known bugs.</li>
1094 <!-- ======================================================================= -->
1095 <div class="doc_subsection">
1096 <a name="x86-be">Known problems with the X86 back-end</a>
1099 <div class="doc_text">
1102 <li>The X86 backend does not yet support
1103 all <a href="http://llvm.org/PR879">inline assembly that uses the X86
1104 floating point stack</a>. It supports the 'f' and 't' constraints, but not
1106 <li>The X86 backend generates inefficient floating point code when configured
1107 to generate code for systems that don't have SSE2.</li>
1108 <li>Win64 code generation wasn't widely tested. Everything should work, but we
1109 expect small issues to happen. Also, llvm-gcc cannot build the mingw64
1110 runtime currently due
1111 to <a href="http://llvm.org/PR2255">several</a>
1112 <a href="http://llvm.org/PR2257">bugs</a> and due to lack of support for
1114 'u' inline assembly constraint and for X87 floating point inline assembly.</li>
1115 <li>The X86-64 backend does not yet support the LLVM IR instruction
1116 <tt>va_arg</tt>. Currently, the llvm-gcc and front-ends support variadic
1117 argument constructs on X86-64 by lowering them manually.</li>
1122 <!-- ======================================================================= -->
1123 <div class="doc_subsection">
1124 <a name="ppc-be">Known problems with the PowerPC back-end</a>
1127 <div class="doc_text">
1130 <li>The Linux PPC32/ABI support needs testing for the interpreter and static
1131 compilation, and lacks support for debug information.</li>
1136 <!-- ======================================================================= -->
1137 <div class="doc_subsection">
1138 <a name="arm-be">Known problems with the ARM back-end</a>
1141 <div class="doc_text">
1144 <li>Support for the Advanced SIMD (Neon) instruction set is still incomplete
1145 and not well tested. Some features may not work at all, and the code quality
1146 may be poor in some cases.</li>
1147 <li>Thumb mode works only on ARMv6 or higher processors. On sub-ARMv6
1148 processors, thumb programs can crash or produce wrong
1149 results (<a href="http://llvm.org/PR1388">PR1388</a>).</li>
1150 <li>Compilation for ARM Linux OABI (old ABI) is supported but not fully tested.
1156 <!-- ======================================================================= -->
1157 <div class="doc_subsection">
1158 <a name="sparc-be">Known problems with the SPARC back-end</a>
1161 <div class="doc_text">
1164 <li>The SPARC backend only supports the 32-bit SPARC ABI (-m32); it does not
1165 support the 64-bit SPARC ABI (-m64).</li>
1170 <!-- ======================================================================= -->
1171 <div class="doc_subsection">
1172 <a name="mips-be">Known problems with the MIPS back-end</a>
1175 <div class="doc_text">
1178 <li>64-bit MIPS targets are not supported yet.</li>
1183 <!-- ======================================================================= -->
1184 <div class="doc_subsection">
1185 <a name="alpha-be">Known problems with the Alpha back-end</a>
1188 <div class="doc_text">
1192 <li>On 21164s, some rare FP arithmetic sequences which may trap do not have the
1193 appropriate nops inserted to ensure restartability.</li>
1198 <!-- ======================================================================= -->
1199 <div class="doc_subsection">
1200 <a name="c-be">Known problems with the C back-end</a>
1203 <div class="doc_text">
1206 <li><a href="http://llvm.org/PR802">The C backend has only basic support for
1207 inline assembly code</a>.</li>
1208 <li><a href="http://llvm.org/PR1658">The C backend violates the ABI of common
1209 C++ programs</a>, preventing intermixing between C++ compiled by the CBE and
1210 C++ code compiled with <tt>llc</tt> or native compilers.</li>
1211 <li>The C backend does not support all exception handling constructs.</li>
1212 <li>The C backend does not support arbitrary precision integers.</li>
1218 <!-- ======================================================================= -->
1219 <div class="doc_subsection">
1220 <a name="c-fe">Known problems with the llvm-gcc C front-end</a>
1223 <div class="doc_text">
1225 <p>The only major language feature of GCC not supported by llvm-gcc is
1226 the <tt>__builtin_apply</tt> family of builtins. However, some extensions
1227 are only supported on some targets. For example, trampolines are only
1228 supported on some targets (these are used when you take the address of a
1229 nested function).</p>
1231 <p>If you run into GCC extensions which are not supported, please let us know.
1236 <!-- ======================================================================= -->
1237 <div class="doc_subsection">
1238 <a name="c++-fe">Known problems with the llvm-gcc C++ front-end</a>
1241 <div class="doc_text">
1243 <p>The C++ front-end is considered to be fully
1244 tested and works for a number of non-trivial programs, including LLVM
1245 itself, Qt, Mozilla, etc.</p>
1248 <li>Exception handling works well on the X86 and PowerPC targets. Currently
1249 only Linux and Darwin targets are supported (both 32 and 64 bit).</li>
1254 <!-- ======================================================================= -->
1255 <div class="doc_subsection">
1256 <a name="fortran-fe">Known problems with the llvm-gcc Fortran front-end</a>
1259 <div class="doc_text">
1261 <li>Fortran support generally works, but there are still several unresolved bugs
1262 in <a href="http://llvm.org/bugs/">Bugzilla</a>. Please see the
1263 tools/gfortran component for details.</li>
1267 <!-- ======================================================================= -->
1268 <div class="doc_subsection">
1269 <a name="ada-fe">Known problems with the llvm-gcc Ada front-end</a>
1272 <div class="doc_text">
1273 The llvm-gcc 4.2 Ada compiler works fairly well; however, this is not a mature
1274 technology, and problems should be expected.
1276 <li>The Ada front-end currently only builds on X86-32. This is mainly due
1277 to lack of trampoline support (pointers to nested functions) on other platforms.
1278 However, it <a href="http://llvm.org/PR2006">also fails to build on X86-64</a>
1279 which does support trampolines.</li>
1280 <li>The Ada front-end <a href="http://llvm.org/PR2007">fails to bootstrap</a>.
1281 This is due to lack of LLVM support for <tt>setjmp</tt>/<tt>longjmp</tt> style
1282 exception handling, which is used internally by the compiler.
1283 Workaround: configure with <tt>--disable-bootstrap</tt>.</li>
1284 <li>The c380004, <a href="http://llvm.org/PR2010">c393010</a>
1285 and <a href="http://llvm.org/PR2421">cxg2021</a> ACATS tests fail
1286 (c380004 also fails with gcc-4.2 mainline).
1287 If the compiler is built with checks disabled then <a href="http://llvm.org/PR2010">c393010</a>
1288 causes the compiler to go into an infinite loop, using up all system memory.</li>
1289 <li>Some GCC specific Ada tests continue to crash the compiler.</li>
1290 <li>The <tt>-E</tt> binder option (exception backtraces)
1291 <a href="http://llvm.org/PR1982">does not work</a> and will result in programs
1292 crashing if an exception is raised. Workaround: do not use <tt>-E</tt>.</li>
1293 <li>Only discrete types <a href="http://llvm.org/PR1981">are allowed to start
1294 or finish at a non-byte offset</a> in a record. Workaround: do not pack records
1295 or use representation clauses that result in a field of a non-discrete type
1296 starting or finishing in the middle of a byte.</li>
1297 <li>The <tt>lli</tt> interpreter <a href="http://llvm.org/PR2009">considers
1298 'main' as generated by the Ada binder to be invalid</a>.
1299 Workaround: hand edit the file to use pointers for <tt>argv</tt> and
1300 <tt>envp</tt> rather than integers.</li>
1301 <li>The <tt>-fstack-check</tt> option <a href="http://llvm.org/PR2008">is
1306 <!-- ======================================================================= -->
1307 <div class="doc_subsection">
1308 <a name="ocaml-bindings">Known problems with the O'Caml bindings</a>
1311 <div class="doc_text">
1313 <p>The <tt>Llvm.Linkage</tt> module is broken, and has incorrect values. Only
1314 <tt>Llvm.Linkage.External</tt>, <tt>Llvm.Linkage.Available_externally</tt>, and
1315 <tt>Llvm.Linkage.Link_once</tt> will be correct. If you need any of the other linkage
1316 modes, you'll have to write an external C library in order to expose the
1317 functionality. This has been fixed in the trunk.</p>
1320 <!-- *********************************************************************** -->
1321 <div class="doc_section">
1322 <a name="additionalinfo">Additional Information</a>
1324 <!-- *********************************************************************** -->
1326 <div class="doc_text">
1328 <p>A wide variety of additional information is available on the <a
1329 href="http://llvm.org">LLVM web page</a>, in particular in the <a
1330 href="http://llvm.org/docs/">documentation</a> section. The web page also
1331 contains versions of the API documentation which is up-to-date with the
1332 Subversion version of the source code.
1333 You can access versions of these documents specific to this release by going
1334 into the "<tt>llvm/doc/</tt>" directory in the LLVM tree.</p>
1336 <p>If you have any questions or comments about LLVM, please feel free to contact
1337 us via the <a href="http://llvm.org/docs/#maillist"> mailing
1342 <!-- *********************************************************************** -->
1346 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
1347 src="http://jigsaw.w3.org/css-validator/images/vcss-blue" alt="Valid CSS"></a>
1348 <a href="http://validator.w3.org/check/referer"><img
1349 src="http://www.w3.org/Icons/valid-html401-blue" alt="Valid HTML 4.01"></a>
1351 <a href="http://llvm.org/">LLVM Compiler Infrastructure</a><br>
1352 Last modified: $Date$