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:
88 KLEE web page at klee.llvm.org
89 Many new papers added to /pubs/
93 <!-- *********************************************************************** -->
94 <div class="doc_section">
95 <a name="subproj">Sub-project Status Update</a>
97 <!-- *********************************************************************** -->
99 <div class="doc_text">
101 The LLVM 2.7 distribution currently consists of code from the core LLVM
102 repository (which roughly includes the LLVM optimizers, code generators
103 and supporting tools), the Clang repository and the llvm-gcc repository. In
104 addition to this code, the LLVM Project includes other sub-projects that are in
105 development. Here we include updates on these subprojects.
111 <!--=========================================================================-->
112 <div class="doc_subsection">
113 <a name="clang">Clang: C/C++/Objective-C Frontend Toolkit</a>
116 <div class="doc_text">
118 <p>The <a href="http://clang.llvm.org/">Clang project</a> is ...</p>
120 <p>In the LLVM 2.7 time-frame, the Clang team has made many improvements:</p>
123 <li>FIXME: C++! Include a link to cxx_compatibility.html Clang 2.7 can
126 <li>CIndex API and Python bindings: Clang now includes a C API as part of the
127 CIndex library. Although we make make some changes to the API in the future, it
128 is intended to be stable and has been designed for use by external projects. See
130 doxygen <a href="http://clang.llvm.org/doxygen/group__CINDEX.html">CIndex</a>
131 documentation for more details. The CIndex API also includes a preliminary
132 set of Python bindings.</li>
134 <li>ARM Support: Clang now has ABI support for both the Darwin and Linux ARM
135 ABIs. Coupled with many improvements to the LLVM ARM backend, Clang is now
136 suitable for use as a a beta quality ARM compiler.</li>
140 <!--=========================================================================-->
141 <div class="doc_subsection">
142 <a name="clangsa">Clang Static Analyzer</a>
145 <div class="doc_text">
147 <p>The <a href="http://clang-analyzer.llvm.org/">Clang Static Analyzer</a>
148 project is an effort to use static source code analysis techniques to
149 automatically find bugs in C and Objective-C programs (and hopefully <a
150 href="http://clang-analyzer.llvm.org/dev_cxx.html">C++ in the
151 future</a>!). The tool is very good at finding bugs that occur on specific
152 paths through code, such as on error conditions.</p>
154 <p>In the LLVM 2.7 time-frame, the analyzer core has made several major and
155 minor improvements, including better support for tracking the fields of
156 structures, initial support (not enabled by default yet) for doing
157 interprocedural (cross-function) analysis, and new checks have been added.
162 <!--=========================================================================-->
163 <div class="doc_subsection">
164 <a name="vmkit">VMKit: JVM/CLI Virtual Machine Implementation</a>
167 <div class="doc_text">
169 The <a href="http://vmkit.llvm.org/">VMKit project</a> is an implementation of
170 a JVM and a CLI Virtual Machine (Microsoft .NET is an
171 implementation of the CLI) using LLVM for static and just-in-time
175 With the release of LLVM 2.7, VMKit has shifted to a great framework for writing
176 virtual machines. VMKit now offers precise and efficient garbage collection with
177 multi-threading support, thanks to the MMTk memory management toolkit, as well
178 as just in time and ahead of time compilation with LLVM. The major changes in
183 <li>Garbage collection: VMKit now uses the MMTk toolkit for garbage collectors.
184 The first collector to be ported is the MarkSweep collector, which is precise,
185 and drastically improves the performance of VMKit.</li>
186 <li>Line number information in the JVM: by using the debug metadata of LLVM, the
187 JVM now supports precise line number information, useful when printing a stack
189 <li>Interface calls in the JVM: we implemented a variant of the Interface Method
190 Table technique for interface calls in the JVM.
197 <!--=========================================================================-->
198 <div class="doc_subsection">
199 <a name="compiler-rt">compiler-rt: Compiler Runtime Library</a>
202 <div class="doc_text">
204 The new LLVM <a href="http://compiler-rt.llvm.org/">compiler-rt project</a>
205 is a simple library that provides an implementation of the low-level
206 target-specific hooks required by code generation and other runtime components.
207 For example, when compiling for a 32-bit target, converting a double to a 64-bit
208 unsigned integer is compiled into a runtime call to the "__fixunsdfdi"
209 function. The compiler-rt library provides highly optimized implementations of
210 this and other low-level routines (some are 3x faster than the equivalent
211 libgcc routines).</p>
214 All of the code in the compiler-rt project is available under the standard LLVM
215 License, a "BSD-style" license. New in LLVM 2.7: compiler_rt now
216 supports ARM targets.</p>
220 <!--=========================================================================-->
221 <div class="doc_subsection">
222 <a name="dragonegg">DragonEgg: llvm-gcc ported to gcc-4.5</a>
225 <div class="doc_text">
227 <a href="http://dragonegg.llvm.org/">DragonEgg</a> is a port of llvm-gcc to
228 gcc-4.5. Unlike llvm-gcc, which makes many intrusive changes to the underlying
229 gcc-4.2 code, dragonegg in theory does not require any gcc-4.5 modifications
230 whatsoever (currently one small patch is needed). This is thanks to the new
231 <a href="http://gcc.gnu.org/wiki/plugins">gcc plugin architecture</a>, which
232 makes it possible to modify the behaviour of gcc at runtime by loading a plugin,
233 which is nothing more than a dynamic library which conforms to the gcc plugin
234 interface. DragonEgg is a gcc plugin that causes the LLVM optimizers to be run
235 instead of the gcc optimizers, and the LLVM code generators instead of the gcc
236 code generators, just like llvm-gcc. To use it, you add
237 "-fplugin=path/dragonegg.so" to the gcc-4.5 command line, and gcc-4.5 magically
238 becomes llvm-gcc-4.5!
242 DragonEgg is still a work in progress. Currently C works very well, while C++,
243 Ada and Fortran work fairly well. All other languages either don't work at all,
244 or only work poorly. For the moment only the x86-32 and x86-64 targets are
245 supported, and only on linux and darwin (darwin needs an additional gcc patch).
249 DragonEgg is a new project which is seeing its first release with llvm-2.7.
255 <!--=========================================================================-->
256 <div class="doc_subsection">
257 <a name="mc">llvm-mc: Machine Code Toolkit</a>
260 <div class="doc_text">
262 The LLVM Machine Code (aka MC) sub-project of LLVM was created to solve a number
263 of problems in the realm of assembly, disassembly, object file format handling,
264 and a number of other related areas that CPU instruction-set level tools work
265 in. It is a sub-project of LLVM which provides it with a number of advantages
266 over other compilers that do not have tightly integrated assembly-level tools.
267 For a gentle introduction, please see the <a
268 href="http://blog.llvm.org/2010/04/intro-to-llvm-mc-project.html">Intro to the
269 LLVM MC Project Blog Post</a>.
272 <p>2.7 includes major parts of the work required by the new MC Project. A few
273 targets have been refactored to support it, and work is underway to support a
274 native assembler in LLVM. This work is not complete in LLVM 2.7, but you has
275 made substantially more progress on LLVM mainline.</p>
277 <p>One minor example of what MC can do is to transcode an AT&T syntax
278 X86 .s file into intel syntax. You can do this with something like:</p>
280 llvm-mc foo.s -output-asm-variant=1 -o foo-intel.s
286 <!-- *********************************************************************** -->
287 <div class="doc_section">
288 <a name="externalproj">External Open Source Projects Using LLVM 2.7</a>
290 <!-- *********************************************************************** -->
292 <div class="doc_text">
294 <p>An exciting aspect of LLVM is that it is used as an enabling technology for
295 a lot of other language and tools projects. This section lists some of the
296 projects that have already been updated to work with LLVM 2.7.</p>
299 <!--=========================================================================-->
300 <div class="doc_subsection">
301 <a name="pure">Pure</a>
304 <div class="doc_text">
306 <a href="http://pure-lang.googlecode.com/">Pure</a>
307 is an algebraic/functional programming language based on term rewriting.
308 Programs are collections of equations which are used to evaluate expressions in
309 a symbolic fashion. Pure offers dynamic typing, eager and lazy evaluation,
310 lexical closures, a hygienic macro system (also based on term rewriting),
311 built-in list and matrix support (including list and matrix comprehensions) and
312 an easy-to-use C interface. The interpreter uses LLVM as a backend to
313 JIT-compile Pure programs to fast native code.</p>
315 <p>Pure versions 0.43 and later have been tested and are known to work with
316 LLVM 2.7 (and continue to work with older LLVM releases >= 2.5).</p>
320 <!--=========================================================================-->
321 <div class="doc_subsection">
322 <a name="RoadsendPHP">Roadsend PHP</a>
325 <div class="doc_text">
327 <a href="http://code.roadsend.com/rphp">Roadsend PHP</a> (rphp) is an open
328 source implementation of the PHP programming
329 language that uses LLVM for its optimizer, JIT and static compiler. This is a
330 reimplementation of an earlier project that is now based on LLVM.
334 <!--=========================================================================-->
335 <div class="doc_subsection">
336 <a name="UnladenSwallow">Unladen Swallow</a>
339 <div class="doc_text">
341 <a href="http://code.google.com/p/unladen-swallow/">Unladen Swallow</a> is a
342 branch of <a href="http://python.org/">Python</a> intended to be fully
343 compatible and significantly faster. It uses LLVM's optimization passes and JIT
348 <!--=========================================================================-->
349 <div class="doc_subsection">
350 <a name="tce">TTA-based Codesign Environment (TCE)</a>
353 <div class="doc_text">
355 <a href="http://tce.cs.tut.fi/">TCE</a> is a toolset for designing
356 application-specific processors (ASP) based on the Transport triggered
357 architecture (TTA). The toolset provides a complete co-design flow from C/C++
358 programs down to synthesizable VHDL and parallel program binaries. Processor
359 customization points include the register files, function units, supported
360 operations, and the interconnection network.</p>
362 <p>TCE uses llvm-gcc/Clang and LLVM for C/C++ language support, target
363 independent optimizations and also for parts of code generation. It generates
364 new LLVM-based code generators "on the fly" for the designed TTA processors and
365 loads them in to the compiler backend as runtime libraries to avoid per-target
366 recompilation of larger parts of the compiler chain.</p>
370 <!--=========================================================================-->
371 <div class="doc_subsection">
372 <a name="safecode">SAFECode Compiler</a>
375 <div class="doc_text">
377 <a href="http://safecode.cs.illinois.edu">SAFECode</a> is a memory safe C
378 compiler built using LLVM. It takes standard, unannotated C code, analyzes the
379 code to ensure that memory accesses and array indexing operations are safe, and
380 instruments the code with run-time checks when safety cannot be proven
385 <!--=========================================================================-->
386 <div class="doc_subsection">
387 <a name="icedtea">IcedTea Java Virtual Machine Implementation</a>
390 <div class="doc_text">
392 <a href="http://icedtea.classpath.org/wiki/Main_Page">IcedTea</a> provides a
393 harness to build OpenJDK using only free software build tools and to provide
394 replacements for the not-yet free parts of OpenJDK. One of the extensions that
395 IcedTea provides is a new JIT compiler named <a
396 href="http://icedtea.classpath.org/wiki/ZeroSharkFaq">Shark</a> which uses LLVM
397 to provide native code generation without introducing processor-dependent
400 <p>Icedtea6 1.8 and later have been tested and are known to work with
401 LLVM 2.7 (and continue to work with older LLVM releases >= 2.6 as well).
405 <!--=========================================================================-->
406 <div class="doc_subsection">
407 <a name="llvm-lua">LLVM-Lua</a>
410 <div class="doc_text">
412 <a href="http://code.google.com/p/llvm-lua/">LLVM-Lua</a> uses LLVM
413 to add JIT and static compiling support to the Lua VM. Lua
414 bytecode is analyzed to remove type checks, then LLVM is used to compile the
415 bytecode down to machine code.
417 <p>LLVM-Lua 1.2.0 have been tested and is known to work with LLVM 2.7.
420 <!-- *********************************************************************** -->
421 <div class="doc_section">
422 <a name="whatsnew">What's New in LLVM 2.7?</a>
424 <!-- *********************************************************************** -->
426 <div class="doc_text">
428 <p>This release includes a huge number of bug fixes, performance tweaks and
429 minor improvements. Some of the major improvements and new features are listed
435 <!--=========================================================================-->
436 <div class="doc_subsection">
437 <a name="orgchanges">LLVM Community Changes</a>
440 <div class="doc_text">
442 <p>In addition to changes to the code, between LLVM 2.6 and 2.7, a number of
443 organization changes have happened:
447 <li>LLVM has a new <a href="http://llvm.org/Logo.html">official logo</a>!</li>
449 <li>Ted Kremenek and Doug Gregor have stepped forward as <a
450 href="http://llvm.org/docs/DeveloperPolicy.html#owners">Code Owners</a> of the
451 Clang static analyzer and the Clang frontend, respectively.</li>
453 <li>LLVM now has an <a href="http://blog.llvm.org">official Blog</a> at
454 <a href="http://blog.llvm.org">http://blog.llvm.org</a>. This is a great way
455 to learn about new LLVM-related features as they are implemented. Several
456 features in this release are already explained on the blog.</li>
458 <li>The LLVM web pages are now checked into the SVN server, in the "www",
459 "www-pubs" and "www-releases" SVN modules. Previously they were hidden in a
460 largely inaccessible old CVS server.</li>
462 <li><a href="http://llvm.org">llvm.org</a> is now hosted on a new (and much
463 faster) server. It is still graciously hosted at the University of Illinois
464 of Urbana Champaign.</li>
468 <!--=========================================================================-->
469 <div class="doc_subsection">
470 <a name="majorfeatures">Major New Features</a>
473 <div class="doc_text">
475 <p>LLVM 2.7 includes several major new capabilities:</p>
478 <li>2.7 includes initial support for the <a
479 href="http://en.wikipedia.org/wiki/MicroBlaze">MicroBlaze</a> target.
480 MicroBlaze is a soft processor core designed for Xilinx FPGAs.</li>
482 <li>2.7 includes a new LLVM IR "extensible metadata" feature. This feature
483 supports many different use cases, including allowing front-end authors to
484 encode source level information into LLVM IR, which is consumed by later
485 language-specific passes. This is a great way to do high-level optimizations
486 like devirtualization, type-based alias analysis, etc. See the <a
487 href="http://blog.llvm.org/2010/04/extensible-metadata-in-llvm-ir.html">
488 Extensible Metadata Blog Post</a> for more information.</li>
490 <li>2.7 encodes <a href="SourceLevelDebugging.html">debug information</a>
491 in a completely new way, built on extensible metadata. The new implementation
492 is much more memory efficient and paves the way for improvements to optimized
493 code debugging experience.</li>
495 <li>2.7 now directly supports taking the address of a label and doing an
496 indirect branch through a pointer. This is particularly useful for
497 interpreter loops, and is used to implement the GCC "address of label"
498 extension. For more information, see the <a
499 href="http://blog.llvm.org/2010/01/address-of-label-and-indirect-branches.html">
500 Address of Label and Indirect Branches in LLVM IR Blog Post</a>.
502 <li>2.7 is the first release to start supporting APIs for assembling and
503 disassembling target machine code. These APIs are useful for a variety of
504 low level clients, and are surfaced in the new "enhanced disassembly" API.
505 For more information see the <a
506 href="http://blog.llvm.org/2010/01/x86-disassembler.html">The X86
507 Disassembler Blog Post</a> for more information.</li>
509 <li>2.7 includes major parts of the work required by the new MC Project,
510 see the <a href="#mc">MC update above</a> for more information.</li>
516 <!--=========================================================================-->
517 <div class="doc_subsection">
518 <a name="coreimprovements">LLVM IR and Core Improvements</a>
521 <div class="doc_text">
522 <p>LLVM IR has several new features for better support of new targets and that
523 expose new optimization opportunities:</p>
526 <li>LLVM IR now supports a 16-bit "half float" data type through <a
527 href="LangRef.html#int_fp16">two new intrinsics</a> and APFloat support.</li>
528 <li>LLVM IR supports two new <a href="LangRef.html#fnattrs">function
529 attributes</a>: inlinehint and alignstack(n). The former is a hint to the
530 optimizer that a function was declared 'inline' and thus the inliner should
531 weight it higher when considering inlining it. The later
532 indicates to the code generator that the function diverges from the platform
533 ABI on stack alignment.</li>
534 <li>The new <a href="LangRef.html#int_objectsize">llvm.objectsize</a> intrinsic
535 allows the optimizer to infer the sizes of memory objects in some cases.
536 This intrinsic is used to implement the GCC <tt>__builtin_object_size</tt>
538 <li>LLVM IR now supports marking load and store instructions with <a
539 href="LangRef.html#i_load">"non-temporal" hints</a> (building on the new
540 metadata feature). This hint encourages the code
541 generator to generate non-temporal accesses when possible, which are useful
542 for code that is carefully managing cache behavior. Currently, only the
543 X86 backend provides target support for this feature.</li>
545 <li>LLVM 2.7 has pre-alpha support for <a
546 href="LangRef.html#t_union">unions in LLVM IR</a>.
547 Unfortunately, this support is not really usable in 2.7, so if you're
548 interested in pushing it forward, please help contribute to LLVM mainline.</li>
554 <!--=========================================================================-->
555 <div class="doc_subsection">
556 <a name="optimizer">Optimizer Improvements</a>
559 <div class="doc_text">
561 <p>In addition to a large array of minor performance tweaks and bug fixes, this
562 release includes a few major enhancements and additions to the optimizers:</p>
566 <li>The inliner reuses now merges arrays stack objects in different callees when
567 inlining multiple call sites into one function. This reduces the stack size
568 of the resultant function.</li>
569 <li>The -basicaa alias analysis pass (which is the default) has been improved to
570 be less dependent on "type safe" pointers. It can now look through bitcasts
571 and other constructs more aggressively, allowing better load/store
573 <li>The load elimination optimization in the GVN Pass [<a
574 href="http://blog.llvm.org/2009/12/introduction-to-load-elimination-in-gvn.html">intro
575 blog post</a>] has been substantially improved to be more aggressive about
576 partial redundancy elimination and do more aggressive phi translation. Please
578 href="http://blog.llvm.org/2009/12/advanced-topics-in-redundant-load.html">
579 Advanced Topics in Redundant Load Elimination with a Focus on PHI Translation
580 Blog Post</a> for more details.</li>
581 <li>The module <a href="LangRef.html#datalayout">target data string</a> now
582 includes a notion of 'native' integer data types for the target. This
583 helps mid-level optimizations avoid promoting complex sequences of
584 operations to data types that are not natively supported (e.g. converting
585 i32 operations to i64 on 32-bit chips).</li>
586 <li>The mid-level optimizer is now conservative when operating on a module with
587 no target data. Previously, it would default to SparcV9 settings, which is
588 not what most people expected.</li>
589 <li>Jump threading is now much more aggressive at simplifying correlated
590 conditionals and threading blocks with otherwise complex logic. It has
591 subsumed the old "Conditional Propagation" pass, and -condprop has been
592 removed from LLVM 2.7.</li>
593 <li>The -instcombine pass has been refactored from being one huge file to being
594 a library of its own. Internally, it uses a customized IRBuilder to clean
595 it up and simplify it.</li>
597 <li>The optimal edge profiling pass is reliable and much more complete than in
598 2.6. It can be used with the llvm-prof tool but isn't wired up to the
599 llvm-gcc and clang command line options yet.</li>
601 <li>A new experimental alias analysis implementation, -scev-aa, has been added.
602 It uses LLVM's Scalar Evolution implementation to do symbolic analysis of
603 pointer offset expressions to disambiguate pointers. It can catch a few
604 cases that basicaa cannot, particularly in complex loop nests.</li>
606 <li>The default pass ordering has been tweaked for improved optimization
614 <!--=========================================================================-->
615 <div class="doc_subsection">
616 <a name="executionengine">Interpreter and JIT Improvements</a>
619 <div class="doc_text">
622 <li>The JIT now supports generating debug information and is compatible with
623 the new GDB 7.0 (and later) interfaces for registering dynamically generated
626 <li>The JIT now <a href="http://llvm.org/PR5184">defaults
627 to compiling eagerly</a> to avoid a race condition in the lazy JIT.
628 Clients that still want the lazy JIT can switch it on by calling
629 <tt>ExecutionEngine::DisableLazyCompilation(false)</tt>.</li>
631 <li>It is now possible to create more than one JIT instance in the same process.
632 These JITs can generate machine code in parallel,
633 although <a href="http://llvm.org/docs/ProgrammersManual.html#jitthreading">you
634 still have to obey the other threading restrictions</a>.</li>
640 <!--=========================================================================-->
641 <div class="doc_subsection">
642 <a name="codegen">Target Independent Code Generator Improvements</a>
645 <div class="doc_text">
647 <p>We have put a significant amount of work into the code generator
648 infrastructure, which allows us to implement more aggressive algorithms and make
652 <li>The 'llc -asm-verbose' option (which is now the default) has been enhanced
653 to emit many useful comments to .s files indicating information about spill
654 slots and loop nest structure. This should make it much easier to read and
655 understand assembly files. This is wired up in llvm-gcc and clang to
656 the <tt>-fverbose-asm</tt> option.</li>
658 <li>New LSR with "full strength reduction" mode. FIXME: Description?</li>
660 <li>A new codegen level Common Subexpression Elimination pass (MachineCSE)
661 is available and enabled by default. It catches redundancies exposed by
663 <li>A new pre-register-allocation tail duplication pass is available and enabled
664 by default, it can substantially improve branch prediction quality in some
666 <li>A new sign and zero extension optimization pass (OptimizeExtsPass)
667 is available and enabled by default. This pass can takes advantage
668 architecture features like x86-64 implicit zero extension behavior and
670 <li>The code generator now supports a mode where it attempts to preserve the
671 order of instructions in the input code. This is important for source that
672 is hand scheduled and extremely sensitive to scheduling. It is compatible
673 with the GCC <tt>-fno-schedule-insns</tt> option.</li>
674 <li>The target-independent code generator now supports generating code with
675 arbitrary numbers of result values. Returning more values than was
676 previously supported is handled by returning through a hidden pointer. In
677 2.7, only the X86 and XCore targets have adopted support for this
679 <li>The code generator now supports generating code that follows the
680 <a href="LangRef.html#callingconv">Glasgow Haskell Compiler Calling
681 Convention</a> and ABI.</li>
682 <li>The "<a href="CodeGenerator.html#selectiondag_select">DAG instruction
683 selection</a>" phase of the code generator has been largely rewritten for
684 2.7. Previously, tblgen spit out tons of C++ code which was compiled and
685 linked into the target to do the pattern matching, now it emits a much
686 smaller table which is read by the target-independent code. The primary
687 advantages of this approach is that the size and compile time of various
688 targets is much improved. The X86 code generator shrunk by 1.5MB of code,
690 <li>Almost the entire code generator has switched to emitting code through the
691 MC interfaces instead of printing textually to the .s file. This led to a
692 number of cleanups and speedups. In 2.7, debug an exception handling
693 information does not go through MC yet.</li>
697 <!--=========================================================================-->
698 <div class="doc_subsection">
699 <a name="x86">X86-32 and X86-64 Target Improvements</a>
702 <div class="doc_text">
703 <p>New features of the X86 target include:
707 <li>The X86 backend now optimizes tails calls much more aggressively for
708 functions that use the standard C calling convention.</li>
709 <li>The X86 backend now models scalar SSE registers as subregs of the SSE vector
710 registers, making the code generator more aggressive in cases where scalars
711 and vector types are mixed.</li>
713 <li>PostRA scheduler for X86? FIXME: is this on by default in 2.7?</li>
719 <!--=========================================================================-->
720 <div class="doc_subsection">
721 <a name="ARM">ARM Target Improvements</a>
724 <div class="doc_text">
725 <p>New features of the ARM target include:
730 <li>The ARM backend now generates instructions in unified assembly syntax.</li>
732 <li>llvm-gcc now has complete support for the ARM v7 NEON instruction set. This
733 support differs slightly from the GCC implementation. Please see the
735 href="http://blog.llvm.org/2010/04/arm-advanced-simd-neon-intrinsics-and.html">
736 ARM Advanced SIMD (NEON) Intrinsics and Types in LLVM Blog Post</a> for
737 helpful information if migrating code from GCC to LLVM-GCC.</li>
739 <li>The ARM and Thumb code generators now use register scavenging for stack
740 object address materialization.(FIXME: WHAT BENEFIT DOES THIS PROVIDE?)</li>
742 <li>The ARM backend now has good support for ARMv4 targets and has been tested
743 on StrongARM hardware. Previously, LLVM only supported ARMv4T and
750 <!--=========================================================================-->
751 <div class="doc_subsection">
752 <a name="newapis">New Useful APIs</a>
755 <div class="doc_text">
757 <p>This release includes a number of new APIs that are used internally, which
758 may also be useful for external clients.
762 <li>The optimizer uses the new CodeMetrics class to measure the size of code.
763 Various passes that use thing (like the inliner, loop unswitcher, etc) all
764 use this to make more accurate estimates of the code size impact of various
766 <li>A new <a href="http://llvm.org/doxygen/InstructionSimplify_8h-source.html">
767 llvm/Analysis/InstructionSimplify.h</a> interface is available for doing
768 symbolic simplification of instructions (e.g. <tt>a+0</tt> -> <tt>a</tt>)
769 without requiring the instruction to exist. This centralizes a lot of
770 ad-hoc symbolic manipulation code scattered in various passes.</li>
771 <li>The optimizer now uses a new <a
772 href="http://llvm.org/doxygen/SSAUpdater_8h-source.html">SSAUpdater</a>
773 class which efficiently supports
774 doing unstructured SSA update operations. This centralized a bunch of code
775 scattered throughout various passes (e.g. jump threading, lcssa,
776 loop rotate, etc) for doing this sort of thing. The code generator has a
777 similar <a href="http://llvm.org/doxygen/MachineSSAUpdater_8h-source.html">
778 MachineSSAUpdater</a> class.</li>
779 <li>The <a href="http://llvm.org/doxygen/Regex_8h-source.html">
780 llvm/Support/Regex.h</a> header exposes a platform independent regular
781 expression API. Building on this, the <a
782 href="TestingGuide.html#FileCheck">FileCheck</a> utility now supports
783 regular exressions.</li>
784 <li>raw_ostream now supports a circular "debug stream" accessed with "dbgs()".
785 By default, this stream works the same way as "errs()", but if you pass
786 <tt>-debug-buffer-size=1000</tt> to opt, the debug stream is capped to a
787 fixed sized circular buffer and the output is printed at the end of the
788 program's execution. This is helpful if you have a long lived compiler
789 process and you're interested in seeing snapshots in time.</li>
795 <!--=========================================================================-->
796 <div class="doc_subsection">
797 <a name="otherimprovements">Other Improvements and New Features</a>
800 <div class="doc_text">
801 <p>Other miscellaneous features include:</p>
804 <li>You can now build LLVM as a big dynamic library (e.g. "libllvm2.7.so"). To
805 get this, configure LLVM with the --enable-shared option.</li>
807 <li>LLVM command line tools now overwrite their output by default. Previously,
808 they would only do this with -f. This makes them more convenient to use, and
809 behave more like standard unix tools.</li>
811 <li>The opt and llc tools now autodetect whether their input is a .ll or .bc
812 file, and automatically do the right thing. This means you don't need to
813 explicitly use the llvm-as tool for most things.</li>
819 <!--=========================================================================-->
820 <div class="doc_subsection">
821 <a name="changes">Major Changes and Removed Features</a>
824 <div class="doc_text">
826 <p>If you're already an LLVM user or developer with out-of-tree changes based
827 on LLVM 2.6, this section lists some "gotchas" that you may run into upgrading
828 from the previous release.</p>
833 The Andersen's alias analysis ("anders-aa") pass, the Predicate Simplifier
834 ("predsimplify") pass, the LoopVR pass, the GVNPRE pass, and the random sampling
835 profiling ("rsprofiling") passes have all been removed. They were not being
836 actively maintained and had substantial problems. If you are interested in
837 these components, you are welcome to ressurect them from SVN, fix the
838 correctness problems, and resubmit them to mainline.</li>
840 <li>LLVM now defaults to building most libraries with RTTI turned off, providing
841 a code size reduction. Packagers who are interested in building LLVM to support
842 plugins that require RTTI information should build with "make REQUIRE_RTTI=1"
843 and should read the new <a href="Packaging.html">Advice on Packaging LLVM</a>
846 <li>The LLVM interpreter now defaults to <em>not</em> using <tt>libffi</tt> even
847 if you have it installed. This makes it more likely that an LLVM built on one
848 system will work when copied to a similar system. To use <tt>libffi</tt>,
849 configure with <tt>--enable-libffi</tt>.</li>
851 <li>Debug information uses a completely different representation, an LLVM 2.6
852 .bc file should work with LLVM 2.7, but debug info won't come forward.</li>
854 <li>The LLVM 2.6 (and earlier) "malloc" and "free" instructions got removed,
855 along with LowerAllocations pass. Now you should just use a call to the
856 malloc and free functions in libc. These calls are optimized as well as
857 the old instructions were.</li>
860 <p>In addition, many APIs have changed in this release. Some of the major LLVM
864 <li>Just about everything has been converted to use raw_ostream instead of
866 <li>llvm/ADT/iterator.h has been removed, just use <iterator>
868 <li>The Streams.h file and "DOUT" got removed, use "DEBUG(errs() << ...);"
870 <li>The TargetAsmInfo interface was renamed to MCAsmInfo.</li>
871 <li><tt>ModuleProvider</tt> has been <a
872 href="http://llvm.org/viewvc/llvm-project?view=rev&revision=94686">removed</a>
873 and its methods moved to <tt>Module</tt> and <tt>GlobalValue</tt>.
874 Most clients can remove uses of <tt>ExistingModuleProvider</tt>,
875 replace <tt>getBitcodeModuleProvider</tt> with
876 <tt>getLazyBitcodeModule</tt>, and pass their <tt>Module</tt> to
877 functions that used to accept <tt>ModuleProvider</tt>. Clients who
878 wrote their own <tt>ModuleProvider</tt>s will need to derive from
879 <tt>GVMaterializer</tt> instead and use
880 <tt>Module::setMaterializer</tt> to attach it to a
881 <tt>Module</tt>.</li>
883 <li><tt>GhostLinkage</tt> has given up the ghost.
884 <tt>GlobalValue</tt>s that have not yet been read from their backing
885 storage have the same linkage they will have after being read in.
886 Clients must replace calls to
887 <tt>GlobalValue::hasNotBeenReadFromBitcode</tt> with
888 <tt>GlobalValue::isMaterializable</tt>.</li>
890 <li>The <tt>llvm/Support/DataTypes.h</tt> header has moved
891 to <tt>llvm/System/DataTypes.h</tt>.</li>
893 <li>The <tt>isInteger</tt>, <tt>isIntOrIntVector</tt>, <tt>isFloatingPoint</tt>,
894 <tt>isFPOrFPVector</tt> and <tt>isFPOrFPVector</tt> methods have been renamed
895 <tt>isIntegerTy</tt>, <tt>isIntOrIntVectorTy</tt>, <tt>isFloatingPointTy</tt>,
896 <tt>isFPOrFPVectorTy</tt> and <tt>isFPOrFPVectorTy</tt> respectively.</li>
898 <li><tt>llvm::Instruction::clone()</tt> no longer takes argument.</li>
899 <li><tt>raw_fd_ostream</tt>'s constructor now takes a flag argument, not individual
900 booleans (see <tt>include/llvm/Support/raw_ostream.h</tt> for details).</li>
901 <li>Some header files have been renamed:
903 <li><tt>llvm/Support/AIXDataTypesFix.h</tt> to
904 <tt>llvm/System/AIXDataTypesFix.h</tt></li>
905 <li><tt>llvm/Support/DataTypes.h</tt> to <tt>llvm/System/DataTypes.h</tt></li>
906 <li><tt>llvm/Transforms/Utils/InlineCost.h</tt> to
907 <tt>llvm/Analysis/InlineCost.h</tt></li>
908 <li><tt>llvm/Support/Mangler.h</tt> to <tt>llvm/Target/Mangler.h</tt></li>
909 <li><tt>llvm/Analysis/Passes.h</tt> to <tt>llvm/CodeGen/Passes.h</tt></li>
917 <!-- *********************************************************************** -->
918 <div class="doc_section">
919 <a name="portability">Portability and Supported Platforms</a>
921 <!-- *********************************************************************** -->
923 <div class="doc_text">
925 <p>LLVM is known to work on the following platforms:</p>
928 <li>Intel and AMD machines (IA32, X86-64, AMD64, EMT-64) running Red Hat
929 Linux, Fedora Core, FreeBSD and AuroraUX (and probably other unix-like
931 <li>PowerPC and X86-based Mac OS X systems, running 10.4 and above in 32-bit
932 and 64-bit modes.</li>
933 <li>Intel and AMD machines running on Win32 using MinGW libraries (native).</li>
934 <li>Intel and AMD machines running on Win32 with the Cygwin libraries (limited
935 support is available for native builds with Visual C++).</li>
936 <li>Sun x86 and AMD64 machines running Solaris 10, OpenSolaris 0906.</li>
937 <li>Alpha-based machines running Debian GNU/Linux.</li>
940 <p>The core LLVM infrastructure uses GNU autoconf to adapt itself
941 to the machine and operating system on which it is built. However, minor
942 porting may be required to get LLVM to work on new platforms. We welcome your
943 portability patches and reports of successful builds or error messages.</p>
947 <!-- *********************************************************************** -->
948 <div class="doc_section">
949 <a name="knownproblems">Known Problems</a>
951 <!-- *********************************************************************** -->
953 <div class="doc_text">
955 <p>This section contains significant known problems with the LLVM system,
956 listed by component. If you run into a problem, please check the <a
957 href="http://llvm.org/bugs/">LLVM bug database</a> and submit a bug if
958 there isn't already one.</p>
961 <li>LLVM will not correctly compile on Solaris and/or OpenSolaris
962 using the stock GCC 3.x.x series 'out the box',
963 See: <a href="GettingStarted.html#brokengcc">Broken versions of GCC and other tools</a>.
964 However, A <a href="http://pkg.auroraux.org/GCC">Modern GCC Build</a>
965 for x86/x86-64 has been made available from the third party AuroraUX Project
966 that has been meticulously tested for bootstrapping LLVM & Clang.</li>
971 <!-- ======================================================================= -->
972 <div class="doc_subsection">
973 <a name="experimental">Experimental features included with this release</a>
976 <div class="doc_text">
978 <p>The following components of this LLVM release are either untested, known to
979 be broken or unreliable, or are in early development. These components should
980 not be relied on, and bugs should not be filed against them, but they may be
981 useful to some people. In particular, if you would like to work on one of these
982 components, please contact us on the <a
983 href="http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev">LLVMdev list</a>.</p>
986 <li>The MSIL, Alpha, SPU, MIPS, PIC16, Blackfin, MSP430, SystemZ and MicroBlaze
987 backends are experimental.</li>
988 <li><tt>llc</tt> "<tt>-filetype=asm</tt>" (the default) is the only
989 supported value for this option. The MachO writer is experimental, and
990 works much better in mainline SVN.</li>
995 <!-- ======================================================================= -->
996 <div class="doc_subsection">
997 <a name="x86-be">Known problems with the X86 back-end</a>
1000 <div class="doc_text">
1003 <li>The X86 backend does not yet support
1004 all <a href="http://llvm.org/PR879">inline assembly that uses the X86
1005 floating point stack</a>. It supports the 'f' and 't' constraints, but not
1007 <li>The X86 backend generates inefficient floating point code when configured
1008 to generate code for systems that don't have SSE2.</li>
1009 <li>Win64 code generation wasn't widely tested. Everything should work, but we
1010 expect small issues to happen. Also, llvm-gcc cannot build the mingw64
1011 runtime currently due to lack of support for the 'u' inline assembly
1012 constraint and for X87 floating point inline assembly.</li>
1013 <li>The X86-64 backend does not yet support the LLVM IR instruction
1014 <tt>va_arg</tt>. Currently, front-ends support variadic
1015 argument constructs on X86-64 by lowering them manually.</li>
1020 <!-- ======================================================================= -->
1021 <div class="doc_subsection">
1022 <a name="ppc-be">Known problems with the PowerPC back-end</a>
1025 <div class="doc_text">
1028 <li>The Linux PPC32/ABI support needs testing for the interpreter and static
1029 compilation, and lacks support for debug information.</li>
1034 <!-- ======================================================================= -->
1035 <div class="doc_subsection">
1036 <a name="arm-be">Known problems with the ARM back-end</a>
1039 <div class="doc_text">
1042 <li>Thumb mode works only on ARMv6 or higher processors. On sub-ARMv6
1043 processors, thumb programs can crash or produce wrong
1044 results (<a href="http://llvm.org/PR1388">PR1388</a>).</li>
1045 <li>Compilation for ARM Linux OABI (old ABI) is supported but not fully tested.
1051 <!-- ======================================================================= -->
1052 <div class="doc_subsection">
1053 <a name="sparc-be">Known problems with the SPARC back-end</a>
1056 <div class="doc_text">
1059 <li>The SPARC backend only supports the 32-bit SPARC ABI (-m32); it does not
1060 support the 64-bit SPARC ABI (-m64).</li>
1065 <!-- ======================================================================= -->
1066 <div class="doc_subsection">
1067 <a name="mips-be">Known problems with the MIPS back-end</a>
1070 <div class="doc_text">
1073 <li>64-bit MIPS targets are not supported yet.</li>
1078 <!-- ======================================================================= -->
1079 <div class="doc_subsection">
1080 <a name="alpha-be">Known problems with the Alpha back-end</a>
1083 <div class="doc_text">
1087 <li>On 21164s, some rare FP arithmetic sequences which may trap do not have the
1088 appropriate nops inserted to ensure restartability.</li>
1093 <!-- ======================================================================= -->
1094 <div class="doc_subsection">
1095 <a name="c-be">Known problems with the C back-end</a>
1098 <div class="doc_text">
1101 <li><a href="http://llvm.org/PR802">The C backend has only basic support for
1102 inline assembly code</a>.</li>
1103 <li><a href="http://llvm.org/PR1658">The C backend violates the ABI of common
1104 C++ programs</a>, preventing intermixing between C++ compiled by the CBE and
1105 C++ code compiled with <tt>llc</tt> or native compilers.</li>
1106 <li>The C backend does not support all exception handling constructs.</li>
1107 <li>The C backend does not support arbitrary precision integers.</li>
1113 <!-- ======================================================================= -->
1114 <div class="doc_subsection">
1115 <a name="c-fe">Known problems with the llvm-gcc C and C++ front-end</a>
1118 <div class="doc_text">
1120 <p>The only major language feature of GCC not supported by llvm-gcc is
1121 the <tt>__builtin_apply</tt> family of builtins. However, some extensions
1122 are only supported on some targets. For example, trampolines are only
1123 supported on some targets (these are used when you take the address of a
1124 nested function).</p>
1128 <!-- ======================================================================= -->
1129 <div class="doc_subsection">
1130 <a name="fortran-fe">Known problems with the llvm-gcc Fortran front-end</a>
1133 <div class="doc_text">
1135 <li>Fortran support generally works, but there are still several unresolved bugs
1136 in <a href="http://llvm.org/bugs/">Bugzilla</a>. Please see the
1137 tools/gfortran component for details.</li>
1141 <!-- ======================================================================= -->
1142 <div class="doc_subsection">
1143 <a name="ada-fe">Known problems with the llvm-gcc Ada front-end</a>
1146 <div class="doc_text">
1147 The llvm-gcc 4.2 Ada compiler works fairly well; however, this is not a mature
1148 technology, and problems should be expected.
1150 <li>The Ada front-end currently only builds on X86-32. This is mainly due
1151 to lack of trampoline support (pointers to nested functions) on other platforms.
1152 However, it <a href="http://llvm.org/PR2006">also fails to build on X86-64</a>
1153 which does support trampolines.</li>
1154 <li>The Ada front-end <a href="http://llvm.org/PR2007">fails to bootstrap</a>.
1155 This is due to lack of LLVM support for <tt>setjmp</tt>/<tt>longjmp</tt> style
1156 exception handling, which is used internally by the compiler.
1157 Workaround: configure with <tt>--disable-bootstrap</tt>.</li>
1158 <li>The c380004, <a href="http://llvm.org/PR2010">c393010</a>
1159 and <a href="http://llvm.org/PR2421">cxg2021</a> ACATS tests fail
1160 (c380004 also fails with gcc-4.2 mainline).
1161 If the compiler is built with checks disabled then <a href="http://llvm.org/PR2010">c393010</a>
1162 causes the compiler to go into an infinite loop, using up all system memory.</li>
1163 <li>Some GCC specific Ada tests continue to crash the compiler.</li>
1164 <li>The <tt>-E</tt> binder option (exception backtraces)
1165 <a href="http://llvm.org/PR1982">does not work</a> and will result in programs
1166 crashing if an exception is raised. Workaround: do not use <tt>-E</tt>.</li>
1167 <li>Only discrete types <a href="http://llvm.org/PR1981">are allowed to start
1168 or finish at a non-byte offset</a> in a record. Workaround: do not pack records
1169 or use representation clauses that result in a field of a non-discrete type
1170 starting or finishing in the middle of a byte.</li>
1171 <li>The <tt>lli</tt> interpreter <a href="http://llvm.org/PR2009">considers
1172 'main' as generated by the Ada binder to be invalid</a>.
1173 Workaround: hand edit the file to use pointers for <tt>argv</tt> and
1174 <tt>envp</tt> rather than integers.</li>
1175 <li>The <tt>-fstack-check</tt> option <a href="http://llvm.org/PR2008">is
1180 <!-- *********************************************************************** -->
1181 <div class="doc_section">
1182 <a name="additionalinfo">Additional Information</a>
1184 <!-- *********************************************************************** -->
1186 <div class="doc_text">
1188 <p>A wide variety of additional information is available on the <a
1189 href="http://llvm.org">LLVM web page</a>, in particular in the <a
1190 href="http://llvm.org/docs/">documentation</a> section. The web page also
1191 contains versions of the API documentation which is up-to-date with the
1192 Subversion version of the source code.
1193 You can access versions of these documents specific to this release by going
1194 into the "<tt>llvm/doc/</tt>" directory in the LLVM tree.</p>
1196 <p>If you have any questions or comments about LLVM, please feel free to contact
1197 us via the <a href="http://llvm.org/docs/#maillist"> mailing
1202 <!-- *********************************************************************** -->
1206 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
1207 src="http://jigsaw.w3.org/css-validator/images/vcss-blue" alt="Valid CSS"></a>
1208 <a href="http://validator.w3.org/check/referer"><img
1209 src="http://www.w3.org/Icons/valid-html401-blue" alt="Valid HTML 4.01"></a>
1211 <a href="http://llvm.org/">LLVM Compiler Infrastructure</a><br>
1212 Last modified: $Date$