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</li>
125 <li>CIndex API and Python bindings: Clang now includes a C API as part of the
126 CIndex library. Although we make make some changes to the API in the future, it
127 is intended to be stable and has been designed for use by external projects. See
129 doxygen <a href="http://clang.llvm.org/doxygen/group__CINDEX.html">CIndex</a>
130 documentation for more details. The CIndex API also includings an preliminary
131 set of Python bindings.</li>
133 <li>ARM Support: Clang now has ABI support for both the Darwin and Linux ARM
134 ABIs. Coupled with many improvements to the LLVM ARM backend, Clang is now
135 suitable for use as a a beta quality ARM compiler.</li>
139 <!--=========================================================================-->
140 <div class="doc_subsection">
141 <a name="clangsa">Clang Static Analyzer</a>
144 <div class="doc_text">
146 <p>The <a href="http://clang-analyzer.llvm.org/">Clang Static Analyzer</a>
147 project is an effort to use static source code analysis techniques to
148 automatically find bugs in C and Objective-C programs (and hopefully <a
149 href="http://clang-analyzer.llvm.org/dev_cxx.html">C++ in the
150 future</a>!). The tool is very good at finding bugs that occur on specific
151 paths through code, such as on error conditions.</p>
153 <p>In the LLVM 2.7 time-frame, the analyzer core has made several major and
154 minor improvements, including better support for tracking the fields of
155 structures, initial support (not enabled by default yet) for doing
156 interprocedural (cross-function) analysis, and new checks have been added.
161 <!--=========================================================================-->
162 <div class="doc_subsection">
163 <a name="vmkit">VMKit: JVM/CLI Virtual Machine Implementation</a>
166 <div class="doc_text">
168 The <a href="http://vmkit.llvm.org/">VMKit project</a> is an implementation of
169 a JVM and a CLI Virtual Machine (Microsoft .NET is an
170 implementation of the CLI) using LLVM for static and just-in-time
174 With the release of LLVM 2.7, VMKit has shifted to a great framework for writing
175 virtual machines. VMKit now offers precise and efficient garbage collection with
176 multi-threading support, thanks to the MMTk memory management toolkit, as well
177 as just in time and ahead of time compilation with LLVM. The major changes in
182 <li>Garbage collection: VMKit now uses the MMTk toolkit for garbage collectors.
183 The first collector to be ported is the MarkSweep collector, which is precise,
184 and drastically improves the performance of VMKit.</li>
185 <li>Line number information in the JVM: by using the debug metadata of LLVM, the
186 JVM now supports precise line number information, useful when printing a stack
188 <li>Interface calls in the JVM: we implemented a variant of the Interface Method
189 Table technique for interface calls in the JVM.
196 <!--=========================================================================-->
197 <div class="doc_subsection">
198 <a name="compiler-rt">compiler-rt: Compiler Runtime Library</a>
201 <div class="doc_text">
203 The new LLVM <a href="http://compiler-rt.llvm.org/">compiler-rt project</a>
204 is a simple library that provides an implementation of the low-level
205 target-specific hooks required by code generation and other runtime components.
206 For example, when compiling for a 32-bit target, converting a double to a 64-bit
207 unsigned integer is compiled into a runtime call to the "__fixunsdfdi"
208 function. The compiler-rt library provides highly optimized implementations of
209 this and other low-level routines (some are 3x faster than the equivalent
210 libgcc routines).</p>
213 All of the code in the compiler-rt project is available under the standard LLVM
214 License, a "BSD-style" license. New in LLVM 2.7: compiler_rt now
215 supports ARM targets.</p>
219 <!--=========================================================================-->
220 <div class="doc_subsection">
221 <a name="dragonegg">DragonEgg: llvm-gcc ported to gcc-4.5</a>
224 <div class="doc_text">
226 <a href="http://dragonegg.llvm.org/">DragonEgg</a> is a port of llvm-gcc to
227 gcc-4.5. Unlike llvm-gcc, which makes many intrusive changes to the underlying
228 gcc-4.2 code, dragonegg in theory does not require any gcc-4.5 modifications
229 whatsoever (currently one small patch is needed). This is thanks to the new
230 <a href="http://gcc.gnu.org/wiki/plugins">gcc plugin architecture</a>, which
231 makes it possible to modify the behaviour of gcc at runtime by loading a plugin,
232 which is nothing more than a dynamic library which conforms to the gcc plugin
233 interface. DragonEgg is a gcc plugin that causes the LLVM optimizers to be run
234 instead of the gcc optimizers, and the LLVM code generators instead of the gcc
235 code generators, just like llvm-gcc. To use it, you add
236 "-fplugin=path/dragonegg.so" to the gcc-4.5 command line, and gcc-4.5 magically
237 becomes llvm-gcc-4.5!
241 DragonEgg is still a work in progress. Currently C works very well, while C++,
242 Ada and Fortran work fairly well. All other languages either don't work at all,
243 or only work poorly. For the moment only the x86-32 and x86-64 targets are
244 supported, and only on linux and darwin (darwin needs an additional gcc patch).
248 DragonEgg is a new project which is seeing its first release with llvm-2.7.
254 <!--=========================================================================-->
255 <div class="doc_subsection">
256 <a name="mc">llvm-mc: Machine Code Toolkit</a>
259 <div class="doc_text">
261 The LLVM Machine Code (aka MC) sub-project of LLVM was created to solve a number
262 of problems in the realm of assembly, disassembly, object file format handling,
263 and a number of other related areas that CPU instruction-set level tools work
264 in. It is a sub-project of LLVM which provides it with a number of advantages
265 over other compilers that do not have tightly integrated assembly-level tools.
266 For a gentle introduction, please see the <a
267 href="http://blog.llvm.org/2010/04/intro-to-llvm-mc-project.html">Intro to the
268 LLVM MC Project Blog Post</a>.
271 <p>2.7 includes major parts of the work required by the new MC Project. A few
272 targets have been refactored to support it, and work is underway to support a
273 native assembler in LLVM. This work is not complete in LLVM 2.7, but you has
274 made substantially more progress on LLVM mainline.</p>
276 <p>One minor example of what MC can do is to transcode an AT&T syntax
277 X86 .s file into intel syntax. You can do this with something like:</p>
279 llvm-mc foo.s -output-asm-variant=1 -o foo-intel.s
285 <!-- *********************************************************************** -->
286 <div class="doc_section">
287 <a name="externalproj">External Open Source Projects Using LLVM 2.7</a>
289 <!-- *********************************************************************** -->
291 <div class="doc_text">
293 <p>An exciting aspect of LLVM is that it is used as an enabling technology for
294 a lot of other language and tools projects. This section lists some of the
295 projects that have already been updated to work with LLVM 2.7.</p>
298 <!--=========================================================================-->
299 <div class="doc_subsection">
300 <a name="pure">Pure</a>
303 <div class="doc_text">
305 <a href="http://pure-lang.googlecode.com/">Pure</a>
306 is an algebraic/functional programming language based on term rewriting.
307 Programs are collections of equations which are used to evaluate expressions in
308 a symbolic fashion. Pure offers dynamic typing, eager and lazy evaluation,
309 lexical closures, a hygienic macro system (also based on term rewriting),
310 built-in list and matrix support (including list and matrix comprehensions) and
311 an easy-to-use C interface. The interpreter uses LLVM as a backend to
312 JIT-compile Pure programs to fast native code.</p>
314 <p>Pure versions 0.43 and later have been tested and are known to work with
315 LLVM 2.7 (and continue to work with older LLVM releases >= 2.5).</p>
319 <!--=========================================================================-->
320 <div class="doc_subsection">
321 <a name="RoadsendPHP">Roadsend PHP</a>
324 <div class="doc_text">
326 <a href="http://code.roadsend.com/rphp">Roadsend PHP</a> (rphp) is an open
327 source implementation of the PHP programming
328 language that uses LLVM for its optimizer, JIT and static compiler. This is a
329 reimplementation of an earlier project that is now based on LLVM.
333 <!--=========================================================================-->
334 <div class="doc_subsection">
335 <a name="UnladenSwallow">Unladen Swallow</a>
338 <div class="doc_text">
340 <a href="http://code.google.com/p/unladen-swallow/">Unladen Swallow</a> is a
341 branch of <a href="http://python.org/">Python</a> intended to be fully
342 compatible and significantly faster. It uses LLVM's optimization passes and JIT
347 <!--=========================================================================-->
348 <div class="doc_subsection">
349 <a name="tce">TTA-based Codesign Environment (TCE)</a>
352 <div class="doc_text">
354 <a href="http://tce.cs.tut.fi/">TCE</a> is a toolset for designing
355 application-specific processors (ASP) based on the Transport triggered
356 architecture (TTA). The toolset provides a complete co-design flow from C/C++
357 programs down to synthesizable VHDL and parallel program binaries. Processor
358 customization points include the register files, function units, supported
359 operations, and the interconnection network.</p>
361 <p>TCE uses llvm-gcc/Clang and LLVM for C/C++ language support, target
362 independent optimizations and also for parts of code generation. It generates
363 new LLVM-based code generators "on the fly" for the designed TTA processors and
364 loads them in to the compiler backend as runtime libraries to avoid per-target
365 recompilation of larger parts of the compiler chain.</p>
369 <!--=========================================================================-->
370 <div class="doc_subsection">
371 <a name="safecode">SAFECode Compiler</a>
374 <div class="doc_text">
376 <a href="http://safecode.cs.illinois.edu">SAFECode</a> is a memory safe C
377 compiler built using LLVM. It takes standard, unannotated C code, analyzes the
378 code to ensure that memory accesses and array indexing operations are safe, and
379 instruments the code with run-time checks when safety cannot be proven
385 <!-- *********************************************************************** -->
386 <div class="doc_section">
387 <a name="whatsnew">What's New in LLVM 2.7?</a>
389 <!-- *********************************************************************** -->
391 <div class="doc_text">
393 <p>This release includes a huge number of bug fixes, performance tweaks and
394 minor improvements. Some of the major improvements and new features are listed
400 <!--=========================================================================-->
401 <div class="doc_subsection">
402 <a name="orgchanges">LLVM Community Changes</a>
405 <div class="doc_text">
407 <p>In addition to changes to the code, between LLVM 2.6 and 2.7, a number of
408 organization changes have happened:
412 <li>LLVM has a new <a href="http://llvm.org/Logo.html">official logo</a>!</li>
414 <li>Ted Kremenek and Doug Gregor have stepped forward as <a
415 href="http://llvm.org/docs/DeveloperPolicy.html#owners">Code Owners</a> of the
416 Clang static analyzer and the Clang frontend, respectively.</li>
418 <li>LLVM now has an <a href="http://blog.llvm.org">official Blog</a> at
419 <a href="http://blog.llvm.org">http://blog.llvm.org</a>. This is a great way
420 to learn about new LLVM-related features as they are implemented. Several
421 features in this release are already explained on the blog.</li>
423 <li>The LLVM web pages are now checked into the SVN server, in the "www",
424 "www-pubs" and "www-releases" SVN modules. Previously they were hidden in a
425 largely inaccessible old CVS server.</li>
427 <li><a href="http://llvm.org">llvm.org</a> is now hosted on a new (and much
428 faster) server. It is still graciously hosted at the University of Illinois
429 of Urbana Champaign.</li>
433 <!--=========================================================================-->
434 <div class="doc_subsection">
435 <a name="majorfeatures">Major New Features</a>
438 <div class="doc_text">
440 <p>LLVM 2.7 includes several major new capabilities:</p>
443 <li>2.7 includes initial support for the <a
444 href="http://en.wikipedia.org/wiki/MicroBlaze">MicroBlaze</a> target.
445 MicroBlaze is a soft processor core designed for Xilinx FPGAs.</li>
447 <li>2.7 includes a new LLVM IR "extensible metadata" feature. This feature
448 supports many different use cases, including allowing front-end authors to
449 encode source level information into LLVM IR, which is consumed by later
450 language-specific passes. This is a great way to do high-level optimizations
451 like devirtualization, type-based alias analysis, etc. See the <a
452 href="http://blog.llvm.org/2010/04/extensible-metadata-in-llvm-ir.html">
453 Extensible Metadata Blog Post</a> for more information.</li>
455 <li>2.7 encodes <a href="SourceLevelDebugging.html">debug information</a>
456 in a completely new way, built on extensible metadata. The new implementation
457 is much more memory efficient and paves the way for improvements to optimized
458 code debugging experience.</li>
460 <li>2.7 now directly supports taking the address of a label and doing an
461 indirect branch through a pointer. This is particularly useful for
462 interpreter loops, and is used to implement the GCC "address of label"
463 extension. For more information, see the <a
464 href="http://blog.llvm.org/2010/01/address-of-label-and-indirect-branches.html">
465 Address of Label and Indirect Branches in LLVM IR Blog Post</a>.
467 <li>2.7 is the first release to start supporting APIs for assembling and
468 disassembling target machine code. These APIs are useful for a variety of
469 low level clients, and are surfaced in the new "enhanced disassembly" API.
470 For more information see the <a
471 href="http://blog.llvm.org/2010/01/x86-disassembler.html">The X86
472 Disassembler Blog Post</a> for more information.</li>
474 <li>2.7 includes major parts of the work required by the new MC Project,
475 see the <a href="#mc">MC update above</a> for more information.</li>
481 <!--=========================================================================-->
482 <div class="doc_subsection">
483 <a name="coreimprovements">LLVM IR and Core Improvements</a>
486 <div class="doc_text">
487 <p>LLVM IR has several new features for better support of new targets and that
488 expose new optimization opportunities:</p>
491 <li>LLVM IR now supports a 16-bit "half float" data type through <a
492 href="LangRef.html#int_fp16">two new intrinsics</a> and APFloat support.</li>
493 <li>LLVM IR supports two new <a href="LangRef.html#fnattrs">function
494 attributes</a>: inlinehint and alignstack(n). The former is a hint to the
495 optimizer that a function was declared 'inline' and thus the inliner should
496 weight it higher when considering inlining it. The later
497 indicates to the code generator that the function diverges from the platform
498 ABI on stack alignment.</li>
499 <li>The new <a href="LangRef.html#int_objectsize">llvm.objectsize</a> intrinsic
500 allows the optimizer to infer the sizes of memory objects in some cases.
501 This intrinsic is used to implement the GCC <tt>__builtin_object_size</tt>
503 <li>LLVM IR now supports marking load and store instructions with <a
504 href="LangRef.html#i_load">"non-temporal" hints</a> (building on the new
505 metadata feature). This hint encourages the code
506 generator to generate non-temporal accesses when possible, which are useful
507 for code that is carefully managing cache behavior. Currently, only the
508 X86 backend provides target support for this feature.</li>
510 <li>LLVM 2.7 has pre-alpha support for <a
511 href="LangRef.html#t_union">unions in LLVM IR</a>.
512 Unfortunately, this support is not really usable in 2.7, so if you're
513 interested in pushing it forward, please help contribute to LLVM mainline.</li>
519 <!--=========================================================================-->
520 <div class="doc_subsection">
521 <a name="optimizer">Optimizer Improvements</a>
524 <div class="doc_text">
526 <p>In addition to a large array of minor performance tweaks and bug fixes, this
527 release includes a few major enhancements and additions to the optimizers:</p>
531 <li>The inliner reuses now merges arrays stack objects in different callees when
532 inlining multiple call sites into one function. This reduces the stack size
533 of the resultant function.</li>
534 <li>The -basicaa alias analysis pass (which is the default) has been improved to
535 be less dependent on "type safe" pointers. It can now look through bitcasts
536 and other constructs more aggressively, allowing better load/store
538 <li>The load elimination optimization in the GVN Pass [<a
539 href="http://blog.llvm.org/2009/12/introduction-to-load-elimination-in-gvn.html">intro
540 blog post</a>] has been substantially improved to be more aggressive about
541 partial redundancy elimination and do more aggressive phi translation. Please
543 href="http://blog.llvm.org/2009/12/advanced-topics-in-redundant-load.html">
544 Advanced Topics in Redundant Load Elimination with a Focus on PHI Translation
545 Blog Post</a> for more details.</li>
546 <li>The module <a href="LangRef.html#datalayout">target data string</a> now
547 includes a notion of 'native' integer data types for the target. This
548 helps mid-level optimizations avoid promoting complex sequences of
549 operations to data types that are not natively supported (e.g. converting
550 i32 operations to i64 on 32-bit chips).</li>
551 <li>The mid-level optimizer is now conservative when operating on a module with
552 no target data. Previously, it would default to SparcV9 settings, which is
553 not what most people expected.</li>
554 <li>Jump threading is now much more aggressive at simplifying correlated
555 conditionals and threading blocks with otherwise complex logic. It has
556 subsumed the old "Conditional Propagation" pass, and -condprop has been
557 removed from LLVM 2.7.</li>
558 <li>The -instcombine pass has been refactored from being one huge file to being
559 a library of its own. Internally, it uses a customized IRBuilder to clean
560 it up and simplify it.</li>
562 <li>The optimal edge profiling pass is reliable and much more complete than in
563 2.6. It can be used with the llvm-prof tool but isn't wired up to the
564 llvm-gcc and clang command line options yet.</li>
566 <li>A new experimental alias analysis implementation, -scev-aa, has been added.
567 It uses LLVM's Scalar Evolution implementation to do symbolic analysis of
568 pointer offset expressions to disambiguate pointers. It can catch a few
569 cases that basicaa cannot, particularly in complex loop nests.</li>
571 <li>The default pass ordering has been tweaked for improved optimization
579 <!--=========================================================================-->
580 <div class="doc_subsection">
581 <a name="executionengine">Interpreter and JIT Improvements</a>
584 <div class="doc_text">
587 <li>The JIT now supports generating debug information and is compatible with
588 the new GDB 7.0 (and later) interfaces for registering dynamically generated
591 <li>The JIT now <a href="http://llvm.org/PR5184">defaults
592 to compiling eagerly</a> to avoid a race condition in the lazy JIT.
593 Clients that still want the lazy JIT can switch it on by calling
594 <tt>ExecutionEngine::DisableLazyCompilation(false)</tt>.</li>
596 <li>It is now possible to create more than one JIT instance in the same process.
597 These JITs can generate machine code in parallel,
598 although <a href="http://llvm.org/docs/ProgrammersManual.html#jitthreading">you
599 still have to obey the other threading restrictions</a>.</li>
605 <!--=========================================================================-->
606 <div class="doc_subsection">
607 <a name="codegen">Target Independent Code Generator Improvements</a>
610 <div class="doc_text">
612 <p>We have put a significant amount of work into the code generator
613 infrastructure, which allows us to implement more aggressive algorithms and make
617 <li>The 'llc -asm-verbose' option (which is now the default) has been enhanced
618 to emit many useful comments to .s files indicating information about spill
619 slots and loop nest structure. This should make it much easier to read and
620 understand assembly files. This is wired up in llvm-gcc and clang to
621 the <tt>-fverbose-asm</tt> option.</li>
623 <li>New LSR with "full strength reduction" mode. FIXME: Description?</li>
625 <li>A new codegen level Common Subexpression Elimination pass (MachineCSE)
626 is available and enabled by default. It catches redundancies exposed by
628 <li>A new pre-register-allocation tail duplication pass is available and enabled
629 by default, it can substantially improve branch prediction quality in some
631 <li>A new sign and zero extension optimization pass (OptimizeExtsPass)
632 is available and enabled by default. This pass can takes advantage
633 architecture features like x86-64 implicit zero extension behavior and
635 <li>The code generator now supports a mode where it attempts to preserve the
636 order of instructions in the input code. This is important for source that
637 is hand scheduled and extremely sensitive to scheduling. It is compatible
638 with the GCC <tt>-fno-schedule-insns</tt> option.</li>
639 <li>The target-independent code generator now supports generating code with
640 arbitrary numbers of result values. Returning more values than was
641 previously supported is handled by returning through a hidden pointer. In
642 2.7, only the X86 and XCore targets have adopted support for this
644 <li>The code generator now supports generating code that follows the
645 <a href="LangRef.html#callingconv">Glasgow Haskell Compiler Calling
646 Convention</a> and ABI.</li>
647 <li>The "<a href="CodeGenerator.html#selectiondag_select">DAG instruction
648 selection</a>" phase of the code generator has been largely rewritten for
649 2.7. Previously, tblgen spit out tons of C++ code which was compiled and
650 linked into the target to do the pattern matching, now it emits a much
651 smaller table which is read by the target-independent code. The primary
652 advantages of this approach is that the size and compile time of various
653 targets is much improved. The X86 code generator shrunk by 1.5MB of code,
655 <li>Almost the entire code generator has switched to emitting code through the
656 MC interfaces instead of printing textually to the .s file. This led to a
657 number of cleanups and speedups. In 2.7, debug an exception handling
658 information does not go through MC yet.</li>
662 <!--=========================================================================-->
663 <div class="doc_subsection">
664 <a name="x86">X86-32 and X86-64 Target Improvements</a>
667 <div class="doc_text">
668 <p>New features of the X86 target include:
672 <li>The X86 backend now optimizes tails calls much more aggressively for
673 functions that use the standard C calling convention.</li>
674 <li>The X86 backend now models scalar SSE registers as subregs of the SSE vector
675 registers, making the code generator more aggressive in cases where scalars
676 and vector types are mixed.</li>
678 <li>PostRA scheduler for X86? FIXME: is this on by default in 2.7?</li>
684 <!--=========================================================================-->
685 <div class="doc_subsection">
686 <a name="ARM">ARM Target Improvements</a>
689 <div class="doc_text">
690 <p>New features of the ARM target include:
695 <li>The ARM backend now generates instructions in unified assembly syntax.</li>
697 <li>llvm-gcc now has complete support for the ARM v7 NEON instruction set. This
698 support differs slightly from the GCC implementation. Please see the
700 href="http://blog.llvm.org/2010/04/arm-advanced-simd-neon-intrinsics-and.html">
701 ARM Advanced SIMD (NEON) Intrinsics and Types in LLVM Blog Post</a> for
702 helpful information if migrating code from GCC to LLVM-GCC.</li>
704 <li>The ARM and Thumb code generators now use register scavenging for stack
705 object address materialization.(FIXME: WHAT BENEFIT DOES THIS PROVIDE?)</li>
707 <li>The ARM backend now has good support for ARMv4 targets and has been tested
708 on StrongARM hardware. Previously, LLVM only supported ARMv4T and
715 <!--=========================================================================-->
716 <div class="doc_subsection">
717 <a name="newapis">New Useful APIs</a>
720 <div class="doc_text">
722 <p>This release includes a number of new APIs that are used internally, which
723 may also be useful for external clients.
727 <li>The optimizer uses the new CodeMetrics class to measure the size of code.
728 Various passes that use thing (like the inliner, loop unswitcher, etc) all
729 use this to make more accurate estimates of the code size impact of various
731 <li>A new <a href="http://llvm.org/doxygen/InstructionSimplify_8h-source.html">
732 llvm/Analysis/InstructionSimplify.h</a> interface is available for doing
733 symbolic simplification of instructions (e.g. <tt>a+0</tt> -> <tt>a</tt>)
734 without requiring the instruction to exist. This centralizes a lot of
735 ad-hoc symbolic manipulation code scattered in various passes.</li>
736 <li>The optimizer now uses a new <a
737 href="http://llvm.org/doxygen/SSAUpdater_8h-source.html">SSAUpdater</a>
738 class which efficiently supports
739 doing unstructured SSA update operations. This centralized a bunch of code
740 scattered throughout various passes (e.g. jump threading, lcssa,
741 loop rotate, etc) for doing this sort of thing. The code generator has a
742 similar <a href="http://llvm.org/doxygen/MachineSSAUpdater_8h-source.html">
743 MachineSSAUpdater</a> class.</li>
744 <li>The <a href="http://llvm.org/doxygen/Regex_8h-source.html">
745 llvm/Support/Regex.h</a> header exposes a platform independent regular
746 expression API. Building on this, the <a
747 href="TestingGuide.html#FileCheck">FileCheck</a> utility now supports
748 regular exressions.</li>
749 <li>raw_ostream now supports a circular "debug stream" accessed with "dbgs()".
750 By default, this stream works the same way as "errs()", but if you pass
751 <tt>-debug-buffer-size=1000</tt> to opt, the debug stream is capped to a
752 fixed sized circular buffer and the output is printed at the end of the
753 program's execution. This is helpful if you have a long lived compiler
754 process and you're interested in seeing snapshots in time.</li>
760 <!--=========================================================================-->
761 <div class="doc_subsection">
762 <a name="otherimprovements">Other Improvements and New Features</a>
765 <div class="doc_text">
766 <p>Other miscellaneous features include:</p>
769 <li>You can now build LLVM as a big dynamic library (e.g. "libllvm2.7.so"). To
770 get this, configure LLVM with the --enable-shared option.</li>
772 <li>LLVM command line tools now overwrite their output by default, before they
773 would only do this with -f. This makes them more convenient to use, and
774 behave more like standard unix tools.</li>
776 <li>The opt and llc tools now autodetect whether their input is a .ll or .bc
777 file, and automatically do the right thing. This means you don't need to
778 explicitly use the llvm-as tool for most things.</li>
784 <!--=========================================================================-->
785 <div class="doc_subsection">
786 <a name="changes">Major Changes and Removed Features</a>
789 <div class="doc_text">
791 <p>If you're already an LLVM user or developer with out-of-tree changes based
792 on LLVM 2.6, this section lists some "gotchas" that you may run into upgrading
793 from the previous release.</p>
798 The Andersen's alias analysis ("anders-aa") pass, the Predicate Simplifier
799 ("predsimplify") pass, the LoopVR pass, the GVNPRE pass, and the random sampling
800 profiling ("rsprofiling") passes have all been removed. They were not being
801 actively maintained and had substantial problems. If you are interested in
802 these components, you are welcome to ressurect them from SVN, fix the
803 correctness problems, and resubmit them to mainline.</li>
805 <li>LLVM now defaults to building most libraries with RTTI turned off, providing
806 a code size reduction. Packagers who are interested in building LLVM to support
807 plugins that require RTTI information should build with "make REQUIRE_RTTI=1"
808 and should read the new <a href="Packaging.html">Advice on Packaging LLVM</a>
811 <li>The LLVM interpreter now defaults to <em>not</em> using <tt>libffi</tt> even
812 if you have it installed. This makes it more likely that an LLVM built on one
813 system will work when copied to a similar system. To use <tt>libffi</tt>,
814 configure with <tt>--enable-libffi</tt>.</li>
816 <li>Debug information uses a completely different representation, an LLVM 2.6
817 .bc file should work with LLVM 2.7, but debug info won't come forward.</li>
819 <li>The LLVM 2.6 (and earlier) "malloc" and "free" instructions got removed,
820 along with LowerAllocations pass. Now you should just use a call to the
821 malloc and free functions in libc. These calls are optimized as well as
822 the old instructions were.</li>
825 <p>In addition, many APIs have changed in this release. Some of the major LLVM
829 <li>Just about everything has been converted to use raw_ostream instead of
831 <li>llvm/ADT/iterator.h has been removed, just use <iterator>
833 <li>The Streams.h file and "DOUT" got removed, use "DEBUG(errs() << ...);"
835 <li><tt>ModuleProvider</tt> has been <a
836 href="http://llvm.org/viewvc/llvm-project?view=rev&revision=94686">removed</a>
837 and its methods moved to <tt>Module</tt> and <tt>GlobalValue</tt>.
838 Most clients can remove uses of <tt>ExistingModuleProvider</tt>,
839 replace <tt>getBitcodeModuleProvider</tt> with
840 <tt>getLazyBitcodeModule</tt>, and pass their <tt>Module</tt> to
841 functions that used to accept <tt>ModuleProvider</tt>. Clients who
842 wrote their own <tt>ModuleProvider</tt>s will need to derive from
843 <tt>GVMaterializer</tt> instead and use
844 <tt>Module::setMaterializer</tt> to attach it to a
845 <tt>Module</tt>.</li>
847 <li><tt>GhostLinkage</tt> has given up the ghost.
848 <tt>GlobalValue</tt>s that have not yet been read from their backing
849 storage have the same linkage they will have after being read in.
850 Clients must replace calls to
851 <tt>GlobalValue::hasNotBeenReadFromBitcode</tt> with
852 <tt>GlobalValue::isMaterializable</tt>.</li>
854 <li>The <tt>llvm/Support/DataTypes.h</tt> header has moved
855 to <tt>llvm/System/DataTypes.h</tt>.</li>
857 <li>The <tt>isInteger</tt>, <tt>isIntOrIntVector</tt>, <tt>isFloatingPoint</tt>,
858 <tt>isFPOrFPVector</tt> and <tt>isFPOrFPVector</tt> methods have been renamed
859 <tt>isIntegerTy</tt>, <tt>isIntOrIntVectorTy</tt>, <tt>isFloatingPointTy</tt>,
860 <tt>isFPOrFPVectorTy</tt> and <tt>isFPOrFPVectorTy</tt> respectively.</li>
867 <!-- *********************************************************************** -->
868 <div class="doc_section">
869 <a name="portability">Portability and Supported Platforms</a>
871 <!-- *********************************************************************** -->
873 <div class="doc_text">
875 <p>LLVM is known to work on the following platforms:</p>
878 <li>Intel and AMD machines (IA32, X86-64, AMD64, EMT-64) running Red Hat
879 Linux, Fedora Core, FreeBSD and AuroraUX (and probably other unix-like
881 <li>PowerPC and X86-based Mac OS X systems, running 10.4 and above in 32-bit
882 and 64-bit modes.</li>
883 <li>Intel and AMD machines running on Win32 using MinGW libraries (native).</li>
884 <li>Intel and AMD machines running on Win32 with the Cygwin libraries (limited
885 support is available for native builds with Visual C++).</li>
886 <li>Sun x86 and AMD64 machines running Solaris 10, OpenSolaris 0906.</li>
887 <li>Alpha-based machines running Debian GNU/Linux.</li>
890 <p>The core LLVM infrastructure uses GNU autoconf to adapt itself
891 to the machine and operating system on which it is built. However, minor
892 porting may be required to get LLVM to work on new platforms. We welcome your
893 portability patches and reports of successful builds or error messages.</p>
897 <!-- *********************************************************************** -->
898 <div class="doc_section">
899 <a name="knownproblems">Known Problems</a>
901 <!-- *********************************************************************** -->
903 <div class="doc_text">
905 <p>This section contains significant known problems with the LLVM system,
906 listed by component. If you run into a problem, please check the <a
907 href="http://llvm.org/bugs/">LLVM bug database</a> and submit a bug if
908 there isn't already one.</p>
911 <li>LLVM will not correctly compile on Solaris and/or OpenSolaris
912 using the stock GCC 3.x.x series 'out the box',
913 See: <a href="GettingStarted.html#brokengcc">Broken versions of GCC and other tools</a>.
914 However, A <a href="http://pkg.auroraux.org/GCC">Modern GCC Build</a>
915 for x86/x86-64 has been made available from the third party AuroraUX Project
916 that has been meticulously tested for bootstrapping LLVM & Clang.</li>
921 <!-- ======================================================================= -->
922 <div class="doc_subsection">
923 <a name="experimental">Experimental features included with this release</a>
926 <div class="doc_text">
928 <p>The following components of this LLVM release are either untested, known to
929 be broken or unreliable, or are in early development. These components should
930 not be relied on, and bugs should not be filed against them, but they may be
931 useful to some people. In particular, if you would like to work on one of these
932 components, please contact us on the <a
933 href="http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev">LLVMdev list</a>.</p>
936 <li>The MSIL, Alpha, SPU, MIPS, PIC16, Blackfin, MSP430, SystemZ and MicroBlaze
937 backends are experimental.</li>
938 <li><tt>llc</tt> "<tt>-filetype=asm</tt>" (the default) is the only
939 supported value for this option. The MachO writer is experimental, and
940 works much better in mainline SVN.</li>
945 <!-- ======================================================================= -->
946 <div class="doc_subsection">
947 <a name="x86-be">Known problems with the X86 back-end</a>
950 <div class="doc_text">
953 <li>The X86 backend does not yet support
954 all <a href="http://llvm.org/PR879">inline assembly that uses the X86
955 floating point stack</a>. It supports the 'f' and 't' constraints, but not
957 <li>The X86 backend generates inefficient floating point code when configured
958 to generate code for systems that don't have SSE2.</li>
959 <li>Win64 code generation wasn't widely tested. Everything should work, but we
960 expect small issues to happen. Also, llvm-gcc cannot build the mingw64
961 runtime currently due to lack of support for the 'u' inline assembly
962 constraint and for X87 floating point inline assembly.</li>
963 <li>The X86-64 backend does not yet support the LLVM IR instruction
964 <tt>va_arg</tt>. Currently, front-ends support variadic
965 argument constructs on X86-64 by lowering them manually.</li>
970 <!-- ======================================================================= -->
971 <div class="doc_subsection">
972 <a name="ppc-be">Known problems with the PowerPC back-end</a>
975 <div class="doc_text">
978 <li>The Linux PPC32/ABI support needs testing for the interpreter and static
979 compilation, and lacks support for debug information.</li>
984 <!-- ======================================================================= -->
985 <div class="doc_subsection">
986 <a name="arm-be">Known problems with the ARM back-end</a>
989 <div class="doc_text">
992 <li>Thumb mode works only on ARMv6 or higher processors. On sub-ARMv6
993 processors, thumb programs can crash or produce wrong
994 results (<a href="http://llvm.org/PR1388">PR1388</a>).</li>
995 <li>Compilation for ARM Linux OABI (old ABI) is supported but not fully tested.
1001 <!-- ======================================================================= -->
1002 <div class="doc_subsection">
1003 <a name="sparc-be">Known problems with the SPARC back-end</a>
1006 <div class="doc_text">
1009 <li>The SPARC backend only supports the 32-bit SPARC ABI (-m32); it does not
1010 support the 64-bit SPARC ABI (-m64).</li>
1015 <!-- ======================================================================= -->
1016 <div class="doc_subsection">
1017 <a name="mips-be">Known problems with the MIPS back-end</a>
1020 <div class="doc_text">
1023 <li>64-bit MIPS targets are not supported yet.</li>
1028 <!-- ======================================================================= -->
1029 <div class="doc_subsection">
1030 <a name="alpha-be">Known problems with the Alpha back-end</a>
1033 <div class="doc_text">
1037 <li>On 21164s, some rare FP arithmetic sequences which may trap do not have the
1038 appropriate nops inserted to ensure restartability.</li>
1043 <!-- ======================================================================= -->
1044 <div class="doc_subsection">
1045 <a name="c-be">Known problems with the C back-end</a>
1048 <div class="doc_text">
1051 <li><a href="http://llvm.org/PR802">The C backend has only basic support for
1052 inline assembly code</a>.</li>
1053 <li><a href="http://llvm.org/PR1658">The C backend violates the ABI of common
1054 C++ programs</a>, preventing intermixing between C++ compiled by the CBE and
1055 C++ code compiled with <tt>llc</tt> or native compilers.</li>
1056 <li>The C backend does not support all exception handling constructs.</li>
1057 <li>The C backend does not support arbitrary precision integers.</li>
1063 <!-- ======================================================================= -->
1064 <div class="doc_subsection">
1065 <a name="c-fe">Known problems with the llvm-gcc C and C++ front-end</a>
1068 <div class="doc_text">
1070 <p>The only major language feature of GCC not supported by llvm-gcc is
1071 the <tt>__builtin_apply</tt> family of builtins. However, some extensions
1072 are only supported on some targets. For example, trampolines are only
1073 supported on some targets (these are used when you take the address of a
1074 nested function).</p>
1078 <!-- ======================================================================= -->
1079 <div class="doc_subsection">
1080 <a name="fortran-fe">Known problems with the llvm-gcc Fortran front-end</a>
1083 <div class="doc_text">
1085 <li>Fortran support generally works, but there are still several unresolved bugs
1086 in <a href="http://llvm.org/bugs/">Bugzilla</a>. Please see the
1087 tools/gfortran component for details.</li>
1091 <!-- ======================================================================= -->
1092 <div class="doc_subsection">
1093 <a name="ada-fe">Known problems with the llvm-gcc Ada front-end</a>
1096 <div class="doc_text">
1097 The llvm-gcc 4.2 Ada compiler works fairly well; however, this is not a mature
1098 technology, and problems should be expected.
1100 <li>The Ada front-end currently only builds on X86-32. This is mainly due
1101 to lack of trampoline support (pointers to nested functions) on other platforms.
1102 However, it <a href="http://llvm.org/PR2006">also fails to build on X86-64</a>
1103 which does support trampolines.</li>
1104 <li>The Ada front-end <a href="http://llvm.org/PR2007">fails to bootstrap</a>.
1105 This is due to lack of LLVM support for <tt>setjmp</tt>/<tt>longjmp</tt> style
1106 exception handling, which is used internally by the compiler.
1107 Workaround: configure with <tt>--disable-bootstrap</tt>.</li>
1108 <li>The c380004, <a href="http://llvm.org/PR2010">c393010</a>
1109 and <a href="http://llvm.org/PR2421">cxg2021</a> ACATS tests fail
1110 (c380004 also fails with gcc-4.2 mainline).
1111 If the compiler is built with checks disabled then <a href="http://llvm.org/PR2010">c393010</a>
1112 causes the compiler to go into an infinite loop, using up all system memory.</li>
1113 <li>Some GCC specific Ada tests continue to crash the compiler.</li>
1114 <li>The <tt>-E</tt> binder option (exception backtraces)
1115 <a href="http://llvm.org/PR1982">does not work</a> and will result in programs
1116 crashing if an exception is raised. Workaround: do not use <tt>-E</tt>.</li>
1117 <li>Only discrete types <a href="http://llvm.org/PR1981">are allowed to start
1118 or finish at a non-byte offset</a> in a record. Workaround: do not pack records
1119 or use representation clauses that result in a field of a non-discrete type
1120 starting or finishing in the middle of a byte.</li>
1121 <li>The <tt>lli</tt> interpreter <a href="http://llvm.org/PR2009">considers
1122 'main' as generated by the Ada binder to be invalid</a>.
1123 Workaround: hand edit the file to use pointers for <tt>argv</tt> and
1124 <tt>envp</tt> rather than integers.</li>
1125 <li>The <tt>-fstack-check</tt> option <a href="http://llvm.org/PR2008">is
1130 <!-- *********************************************************************** -->
1131 <div class="doc_section">
1132 <a name="additionalinfo">Additional Information</a>
1134 <!-- *********************************************************************** -->
1136 <div class="doc_text">
1138 <p>A wide variety of additional information is available on the <a
1139 href="http://llvm.org">LLVM web page</a>, in particular in the <a
1140 href="http://llvm.org/docs/">documentation</a> section. The web page also
1141 contains versions of the API documentation which is up-to-date with the
1142 Subversion version of the source code.
1143 You can access versions of these documents specific to this release by going
1144 into the "<tt>llvm/doc/</tt>" directory in the LLVM tree.</p>
1146 <p>If you have any questions or comments about LLVM, please feel free to contact
1147 us via the <a href="http://llvm.org/docs/#maillist"> mailing
1152 <!-- *********************************************************************** -->
1156 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
1157 src="http://jigsaw.w3.org/css-validator/images/vcss-blue" alt="Valid CSS"></a>
1158 <a href="http://validator.w3.org/check/referer"><img
1159 src="http://www.w3.org/Icons/valid-html401-blue" alt="Valid HTML 4.01"></a>
1161 <a href="http://llvm.org/">LLVM Compiler Infrastructure</a><br>
1162 Last modified: $Date$