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">
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>FIXME: Static Analyzer improvements?</li>
127 <li>CIndex API and Python bindings: Clang now includes a C API as part of the
128 CIndex library. Although we make make some changes to the API in the future, it
129 is intended to be stable and has been designed for use by external projects. See
131 doxygen <a href="http://clang.llvm.org/doxygen/group__CINDEX.html">CIndex</a>
132 documentation for more details. The CIndex API also includings an preliminary
133 set of Python bindings.</li>
135 <li>ARM Support: Clang now has ABI support for both the Darwin and Linux ARM
136 ABIs. Coupled with many improvements to the LLVM ARM backend, Clang is now
137 suitable for use as a a beta quality ARM compiler.</li>
141 <!--=========================================================================-->
142 <div class="doc_subsection">
143 <a name="clangsa">Clang Static Analyzer</a>
146 <div class="doc_text">
148 <p>The <a href="http://clang-analyzer.llvm.org/">Clang Static Analyzer</a>
149 project is an effort to use static source code analysis techniques to
150 automatically find bugs in C and Objective-C programs (and hopefully <a
151 href="http://clang-analyzer.llvm.org/dev_cxx.html">C++ in the
152 future</a>!). The tool is very good at finding bugs that occur on specific
153 paths through code, such as on error conditions.</p>
155 <p>In the LLVM 2.7 time-frame, the analyzer core has made several major and
156 minor improvements, including better support for tracking the fields of
157 structures, initial support (not enabled by default yet) for doing
158 interprocedural (cross-function) analysis, and new checks have been added.
163 <!--=========================================================================-->
164 <div class="doc_subsection">
165 <a name="vmkit">VMKit: JVM/CLI Virtual Machine Implementation</a>
168 <div class="doc_text">
170 The <a href="http://vmkit.llvm.org/">VMKit project</a> is an implementation of
171 a JVM and a CLI Virtual Machine (Microsoft .NET is an
172 implementation of the CLI) using LLVM for static and just-in-time
176 With the release of LLVM 2.7, VMKit has shifted to a great framework for writing
177 virtual machines. VMKit now offers precise and efficient garbage collection with
178 multi-threading support, thanks to the MMTk memory management toolkit, as well
179 as just in time and ahead of time compilation with LLVM. The major changes in
184 <li>Garbage collection: VMKit now uses the MMTk toolkit for garbage collectors.
185 The first collector to be ported is the MarkSweep collector, which is precise,
186 and drastically improves the performance of VMKit.</li>
187 <li>Line number information in the JVM: by using the debug metadata of LLVM, the
188 JVM now supports precise line number information, useful when printing a stack
190 <li>Interface calls in the JVM: we implemented a variant of the Interface Method
191 Table technique for interface calls in the JVM.
198 <!--=========================================================================-->
199 <div class="doc_subsection">
200 <a name="compiler-rt">compiler-rt: Compiler Runtime Library</a>
203 <div class="doc_text">
205 The new LLVM <a href="http://compiler-rt.llvm.org/">compiler-rt project</a>
206 is a simple library that provides an implementation of the low-level
207 target-specific hooks required by code generation and other runtime components.
208 For example, when compiling for a 32-bit target, converting a double to a 64-bit
209 unsigned integer is compiled into a runtime call to the "__fixunsdfdi"
210 function. The compiler-rt library provides highly optimized implementations of
211 this and other low-level routines (some are 3x faster than the equivalent
212 libgcc routines).</p>
215 All of the code in the compiler-rt project is available under the standard LLVM
216 License, a "BSD-style" license. New in LLVM 2.7: compiler_rt now
217 supports ARM targets.</p>
221 <!--=========================================================================-->
222 <div class="doc_subsection">
223 <a name="dragonegg">DragonEgg: llvm-gcc ported to gcc-4.5</a>
226 <div class="doc_text">
228 <a href="http://dragonegg.llvm.org/">DragonEgg</a> is a port of llvm-gcc to
229 gcc-4.5. Unlike llvm-gcc, which makes many intrusive changes to the underlying
230 gcc-4.2 code, dragonegg in theory does not require any gcc-4.5 modifications
231 whatsoever (currently one small patch is needed). This is thanks to the new
232 <a href="http://gcc.gnu.org/wiki/plugins">gcc plugin architecture</a>, which
233 makes it possible to modify the behaviour of gcc at runtime by loading a plugin,
234 which is nothing more than a dynamic library which conforms to the gcc plugin
235 interface. DragonEgg is a gcc plugin that causes the LLVM optimizers to be run
236 instead of the gcc optimizers, and the LLVM code generators instead of the gcc
237 code generators, just like llvm-gcc. To use it, you add
238 "-fplugin=path/dragonegg.so" to the gcc-4.5 command line, and gcc-4.5 magically
239 becomes llvm-gcc-4.5!
243 DragonEgg is still a work in progress. Currently C works very well, while C++,
244 Ada and Fortran work fairly well. All other languages either don't work at all,
245 or only work poorly. For the moment only the x86-32 and x86-64 targets are
246 supported, and only on linux and darwin (darwin needs an additional gcc patch).
250 DragonEgg is a new project which is seeing its first release with llvm-2.7.
256 <!--=========================================================================-->
257 <div class="doc_subsection">
258 <a name="mc">llvm-mc: Machine Code Toolkit</a>
261 <div class="doc_text">
263 The LLVM Machine Code (aka MC) sub-project of LLVM was created to solve a number
264 of problems in the realm of assembly, disassembly, object file format handling,
265 and a number of other related areas that CPU instruction-set level tools work
266 in. It is a sub-project of LLVM which provides it with a number of advantages
267 over other compilers that do not have tightly integrated assembly-level tools.
268 For a gentle introduction, please see the <a
269 href="http://blog.llvm.org/2010/04/intro-to-llvm-mc-project.html">Intro to the
270 LLVM MC Project Blog Post</a>.
273 <p>2.7 includes major parts of the work required by the new MC Project. A few
274 targets have been refactored to support it, and work is underway to support a
275 native assembler in LLVM. This work is not complete in LLVM 2.7, but you has
276 made substantially more progress on LLVM mainline.</p>
278 <p>One minor example of what MC can do is to transcode an AT&T syntax
279 X86 .s file into intel syntax. You can do this with something like:</p>
282 llvm-mc foo.s -output-asm-variant=1 -o foo-intel.s
289 <!-- *********************************************************************** -->
290 <div class="doc_section">
291 <a name="externalproj">External Open Source Projects Using LLVM 2.7</a>
293 <!-- *********************************************************************** -->
295 <div class="doc_text">
297 <p>An exciting aspect of LLVM is that it is used as an enabling technology for
298 a lot of other language and tools projects. This section lists some of the
299 projects that have already been updated to work with LLVM 2.7.</p>
302 <!--=========================================================================-->
303 <div class="doc_subsection">
304 <a name="pure">Pure</a>
307 <div class="doc_text">
309 <a href="http://pure-lang.googlecode.com/">Pure</a>
310 is an algebraic/functional programming language based on term rewriting.
311 Programs are collections of equations which are used to evaluate expressions in
312 a symbolic fashion. Pure offers dynamic typing, eager and lazy evaluation,
313 lexical closures, a hygienic macro system (also based on term rewriting),
314 built-in list and matrix support (including list and matrix comprehensions) and
315 an easy-to-use C interface. The interpreter uses LLVM as a backend to
316 JIT-compile Pure programs to fast native code.</p>
318 <p>Pure versions 0.43 and later have been tested and are known to work with
319 LLVM 2.7 (and continue to work with older LLVM releases >= 2.5).</p>
323 <!--=========================================================================-->
324 <div class="doc_subsection">
325 <a name="RoadsendPHP">Roadsend PHP</a>
328 <div class="doc_text">
330 <a href="http://code.roadsend.com/rphp">Roadsend PHP</a> (rphp) is an open
331 source implementation of the PHP programming
332 language that uses LLVM for its optimizer, JIT and static compiler. This is a
333 reimplementation of an earlier project that is now based on LLVM.
337 <!--=========================================================================-->
338 <div class="doc_subsection">
339 <a name="UnladenSwallow">Unladen Swallow</a>
342 <div class="doc_text">
344 <a href="http://code.google.com/p/unladen-swallow/">Unladen Swallow</a> is a
345 branch of <a href="http://python.org/">Python</a> intended to be fully
346 compatible and significantly faster. It uses LLVM's optimization passes and JIT
351 <!--=========================================================================-->
352 <div class="doc_subsection">
353 <a name="tce">TTA-based Codesign Environment (TCE)</a>
356 <div class="doc_text">
358 <a href="http://tce.cs.tut.fi/">TCE</a> is a toolset for designing
359 application-specific processors (ASP) based on the Transport triggered
360 architecture (TTA). The toolset provides a complete co-design flow from C/C++
361 programs down to synthesizable VHDL and parallel program binaries. Processor
362 customization points include the register files, function units, supported
363 operations, and the interconnection network.</p>
365 <p>TCE uses llvm-gcc/Clang and LLVM for C/C++ language support, target
366 independent optimizations and also for parts of code generation. It generates
367 new LLVM-based code generators "on the fly" for the designed TTA processors and
368 loads them in to the compiler backend as runtime libraries to avoid per-target
369 recompilation of larger parts of the compiler chain.</p>
373 <!--=========================================================================-->
374 <div class="doc_subsection">
375 <a name="safecode">SAFECode Compiler</a>
378 <div class="doc_text">
380 <a href="http://safecode.cs.illinois.edu">SAFECode</a> is a memory safe C
381 compiler built using LLVM. It takes standard, unannotated C code, analyzes the
382 code to ensure that memory accesses and array indexing operations are safe, and
383 instruments the code with run-time checks when safety cannot be proven
389 <!-- *********************************************************************** -->
390 <div class="doc_section">
391 <a name="whatsnew">What's New in LLVM 2.7?</a>
393 <!-- *********************************************************************** -->
395 <div class="doc_text">
397 <p>This release includes a huge number of bug fixes, performance tweaks and
398 minor improvements. Some of the major improvements and new features are listed
404 <!--=========================================================================-->
405 <div class="doc_subsection">
406 <a name="orgchanges">LLVM Community Changes</a>
409 <div class="doc_text">
411 <p>In addition to changes to the code, between LLVM 2.6 and 2.7, a number of
412 organization changes have happened:
416 <li>LLVM has a new <a href="http://llvm.org/Logo.html">official logo</a>!</li>
418 <li>Ted Kremenek and Doug Gregor have stepped forward as <a
419 href="http://llvm.org/docs/DeveloperPolicy.html#owners">Code Owners</a> of the
420 Clang static analyzer and the Clang frontend, respectively.</li>
422 <li>LLVM now has an <a href="http://blog.llvm.org">official Blog</a> at
423 <a href="http://blog.llvm.org">http://blog.llvm.org</a>. This is a great way
424 to learn about new LLVM-related features as they are implemented. Several
425 features in this release are already explained on the blog.</li>
427 <li>The LLVM web pages are now checked into the SVN server, in the "www",
428 "www-pubs" and "www-releases" SVN modules. Previously they were hidden in a
429 largely inaccessible old CVS server.</li>
431 <li><a href="http://llvm.org">llvm.org</a> is now hosted on a new (and much
432 faster) server. It is still graciously hosted at the University of Illinois
433 of Urbana Champaign.</li>
437 <!--=========================================================================-->
438 <div class="doc_subsection">
439 <a name="majorfeatures">Major New Features</a>
442 <div class="doc_text">
444 <p>LLVM 2.7 includes several major new capabilities:</p>
447 <li>2.7 includes initial support for the <a
448 href="http://en.wikipedia.org/wiki/MicroBlaze">MicroBlaze</a> target.
449 MicroBlaze is a soft processor core designed for Xilinx FPGAs.</li>
451 <li>2.7 includes a new LLVM IR "extensible metadata" feature. This feature
452 supports many different use cases, including allowing front-end authors to
453 encode source level information into LLVM IR, which is consumed by later
454 language-specific passes. This is a great way to do high-level optimizations
455 like devirtualization, type-based alias analysis, etc. See the <a
456 href="http://blog.llvm.org/2010/04/extensible-metadata-in-llvm-ir.html">
457 Extensible Metadata Blog Post</a> for more information.</li>
459 <li>2.7 encodes <a href="SourceLevelDebugging.html">debug information</a>
460 in a completely new way, built on extensible metadata. The new implementation
461 is much more memory efficient and paves the way for improvements to optimized
462 code debugging experience.</li>
464 <li>2.7 now directly supports taking the address of a label and doing an
465 indirect branch through a pointer. This is particularly useful for
466 interpreter loops, and is used to implement the GCC "address of label"
467 extension. For more information, see the <a
468 href="http://blog.llvm.org/2010/01/address-of-label-and-indirect-branches.html">
469 Address of Label and Indirect Branches in LLVM IR Blog Post</a>.
471 <li>2.7 is the first release to start supporting APIs for assembling and
472 disassembling target machine code. These APIs are useful for a variety of
473 low level clients, and are surfaced in the new "enhanced disassembly" API.
474 For more information see the <a
475 href="http://blog.llvm.org/2010/01/x86-disassembler.html">The X86
476 Disassembler Blog Post</a> for more information.</li>
478 <li>2.7 includes major parts of the work required by the new MC Project,
479 see the <a href="#mc">MC update above</a> for more information.</li>
486 <!--=========================================================================-->
487 <div class="doc_subsection">
488 <a name="coreimprovements">LLVM IR and Core Improvements</a>
491 <div class="doc_text">
492 <p>LLVM IR has several new features for better support of new targets and that
493 expose new optimization opportunities:</p>
496 <li>LLVM IR now supports a 16-bit "half float" data type through <a
497 href="LangRef.html#int_fp16">two new intrinsics</a> and APFloat support.</li>
498 <li>LLVM IR supports two new <a href="LangRef.html#fnattrs">function
499 attributes</a>: inlinehint and alignstack(n). The former is a hint to the
500 optimizer that a function was declared 'inline' and thus the inliner should
501 weight it higher when considering inlining it. The later
502 indicates to the code generator that the function diverges from the platform
503 ABI on stack alignment.</li>
504 <li>The new <a href="LangRef.html#int_objectsize">llvm.objectsize</a> intrinsic
505 allows the optimizer to infer the sizes of memory objects in some cases.
506 This intrinsic is used to implement the GCC <tt>__builtin_object_size</tt>
508 <li>LLVM IR now supports marking load and store instructions with <a
509 href="LangRef.html#i_load">"non-temporal" hints</a> (building on the new
510 metadata feature). This hint encourages the code
511 generator to generate non-temporal accesses when possible, which are useful
512 for code that is carefully managing cache behavior. Currently, only the
513 X86 backend provides target support for this feature.</li>
515 <li>LLVM 2.7 has pre-alpha support for <a
516 href="http://llvm.org/docs/LangRef.html#t_union">unions in LLVM IR</a>.
517 Unfortunately, this support is not really usable in 2.7, so if you're
518 interested in pushing it forward, please help contribute to LLVM mainline.</li>
524 <!--=========================================================================-->
525 <div class="doc_subsection">
526 <a name="optimizer">Optimizer Improvements</a>
529 <div class="doc_text">
531 <p>In addition to a large array of minor performance tweaks and bug fixes, this
532 release includes a few major enhancements and additions to the optimizers:</p>
536 <li>The inliner reuses now merges arrays stack objects in different callees when
537 inlining multiple call sites into one function. This reduces the stack size
538 of the resultant function.</li>
539 <li>The -basicaa alias analysis pass (which is the default) has been improved to
540 be less dependent on "type safe" pointers. It can now look through bitcasts
541 and other constructs more aggressively, allowing better load/store
543 <li>The load elimination optimization in the GVN Pass [<a
544 href="http://blog.llvm.org/2009/12/introduction-to-load-elimination-in-gvn.html">intro
545 blog post</a>] has been substantially improved to be more aggressive about
546 partial redundancy elimination and do more aggressive phi translation. Please
548 href="http://blog.llvm.org/2009/12/advanced-topics-in-redundant-load.html">
549 Advanced Topics in Redundant Load Elimination with a Focus on PHI Translation
550 Blog Post</a> for more details.</li>
551 <li>The module <a href="LangRef.html#datalayout">target data string</a> now
552 includes a notion of what the 'native' integer data types a for the target,
553 which allows various optimizations to use. This helps mid-level
554 optimizations avoid promoting complex sequences of operations to data types
555 that are not natively supported (e.g. converting i32 operations to i64 on
557 <li>The mid-level optimizer is now conservative when operating on a module with
558 no target data. Previously, it would default to SparcV9 settings, which is
559 not what most people expected.</li>
560 <li>Jump threading is now much more aggressive at simplifying correlated
561 conditionals and threading blocks with otherwise complex logic. It has
562 subsumed the old "Conditional Propagation" pass, and -condprop has been
563 removed from LLVM 2.7.</li>
564 <li>The -instcombine pass has been refactored from being one huge file to being
565 a library of its own. Internally, it uses a customized IRBuilder to clean
566 it up and simplify it.</li>
568 <li>The optimal edge profiling pass is reliable and much more complete than in
569 2.6. It can be used with the llvm-prof tool but isn't wired up to the
570 llvm-gcc and clang command line options yet.</li>
572 <li>A new experimental alias analysis implementation, -scev-aa, has been added.
573 It uses LLVM's Scalar Evolution implementation to do symbolic analysis of
574 pointer offset expressions to disambiguate pointers. It can catch a few
575 cases that basicaa cannot, particularly in complex loop nests.</li>
577 <li>The default pass ordering has been tweaked for improved optimization
585 <!--=========================================================================-->
586 <div class="doc_subsection">
587 <a name="executionengine">Interpreter and JIT Improvements</a>
590 <div class="doc_text">
593 <li>The JIT now supports generating debug information, which is compatible with
594 the new GDB 7.0 (and later) interfaces for registering debug info for
595 dynamically generated code.</li>
597 <li>The JIT now <a href="http://llvm.org/PR5184">defaults
598 to compiling eagerly</a> to avoid a race condition in the lazy JIT.
599 Clients that still want the lazy JIT can switch it on by calling
600 <tt>ExecutionEngine::DisableLazyCompilation(false)</tt>.</li>
602 <li>It is now possible to create more than one JIT instance in the same process.
603 These JITs can generate machine code in parallel,
604 although <a href="http://llvm.org/docs/ProgrammersManual.html#jitthreading">you
605 still have to obey the other threading restrictions</a>.</li>
611 <!--=========================================================================-->
612 <div class="doc_subsection">
613 <a name="codegen">Target Independent Code Generator Improvements</a>
616 <div class="doc_text">
618 <p>We have put a significant amount of work into the code generator
619 infrastructure, which allows us to implement more aggressive algorithms and make
623 <li>The 'llc -asm-verbose' option (which is now the default) has been enhanced
624 to emit many useful comments to .s files indicating information about spill
625 slots and loop nest structure. This should make it much easier to read and
626 understand assembly files. This is wired up in llvm-gcc and clang to
627 the <tt>-fverbose-asm</tt> option.</li>
629 <li>New LSR with "full strength reduction" mode. FIXME: Description?</li>
631 <li>A new codegen level Common Subexpression Elimination pass (MachineCSE)
632 is available and enabled by default. It catches redundancies exposed by
634 <li>A new pre-register-allocation tail duplication pass is available and enabled
635 by default, it can substantially improve branch prediction quality in some
637 <li>A new sign and zero extension optimization pass (OptimizeExtsPass)
638 is available and enabled by default. This pass can takes advantage
639 architecture features like x86-64 implicit zero extension behavior and
641 <li>The code generator now supports a mode where it attempts to preserve the
642 order of instructions in the input code. This is important for source that
643 is hand scheduled and extremely sensitive to scheduling. It is compatible
644 with the GCC <tt>-fno-schedule-insns</tt> option.</li>
645 <li>The target-independent code generator now supports generating code with
646 arbitrary numbers of result values. Returning more values than was
647 previously supported is handled by returning through a hidden pointer. In
648 2.7, only the X86 and XCore targets have adopted support for this
650 <li>The code generator now supports generating code that follows the
651 <a href="LangRef.html#callingconv">Glasgow Haskell Compiler Calling
652 Convention</a> and ABI.</li>
653 <li>The "<a href="CodeGenerator.html#selectiondag_select">DAG instruction
654 selection</a>" phase of the code generator has been largely rewritten for
655 2.7. Previously, tblgen spit out tons of C++ code which was compiled and
656 linked into the target to do the pattern matching, now it emits a much
657 smaller table which is read by the target-independent code. The primary
658 advantages of this approach is that the size and compile time of various
659 targets is much improved. The X86 code generator shrunk by 1.5MB of code,
661 <li>Almost the entire code generator has switched to emitting code through the
662 MC interfaces instead of printing textually to the .s file. This led to a
663 number of cleanups and speedups. In 2.7, debug an exception handling
664 information does not go through MC yet.</li>
668 <!--=========================================================================-->
669 <div class="doc_subsection">
670 <a name="x86">X86-32 and X86-64 Target Improvements</a>
673 <div class="doc_text">
674 <p>New features of the X86 target include:
678 <li>The X86 backend now optimizes tails calls much more aggressively for
679 functions that use the standard C calling convention.</li>
680 <li>The X86 backend now models scalar SSE registers as subregs of the SSE vector
681 registers, making the code generator more aggressive in cases where scalars
682 and vector types are mixed.</li>
684 <li>PostRA scheduler for X86? FIXME: is this on by default in 2.7?</li>
690 <!--=========================================================================-->
691 <div class="doc_subsection">
692 <a name="ARM">ARM Target Improvements</a>
695 <div class="doc_text">
696 <p>New features of the ARM target include:
701 <li>The ARM backend now generates instructions in unified assembly syntax.</li>
703 <li>llvm-gcc now has complete support for the ARM v7 NEON instruction set. This
704 support differs slightly from the GCC implementation. Please see the
706 href="http://blog.llvm.org/2010/04/arm-advanced-simd-neon-intrinsics-and.html">
707 ARM Advanced SIMD (NEON) Intrinsics and Types in LLVM Blog Post</a> for
708 helpful information if migrating code from GCC to LLVM-GCC.</li>
710 <li>The ARM and Thumb code generators now using register scavenging for stack
711 object address materialization.(FIXME: WHAT BENEFIT DOES THIS PROVIDE?)</li>
713 <li>The ARM backend now has good support for ARMv4 targets, and has been tested
714 on StrongARM hardware. Previously, LLVM only supported ARMv4T and
721 <!--=========================================================================-->
722 <div class="doc_subsection">
723 <a name="newapis">New Useful APIs</a>
726 <div class="doc_text">
728 <p>This release includes a number of new APIs that are used internally, which
729 may also be useful for external clients.
733 <li>The optimizer uses the new CodeMetrics class to measure the size of code.
734 Various passes that use thing (like the inliner, loop unswitcher, etc) all
735 use this to make more accurate estimates of the code size impact of various
737 <li>A new <a href="http://llvm.org/doxygen/InstructionSimplify_8h-source.html">
738 llvm/Analysis/InstructionSimplify.h</a> interface available for doing
739 symbolic simplification of instructions (e.g. <tt>a+0</tt> -> <tt>a</tt>)
740 without requiring the instruction to exist. This centralizes a lot of
741 ad-hoc symbolic manipulation code scattered in various passes.</li>
742 <li>The optimizer now uses a new <a
743 href="http://llvm.org/doxygen/SSAUpdater_8h-source.html">SSAUpdater</a>
744 class which efficiently supports
745 doing unstructured SSA update operations. This centralized a bunch of code
746 scattered through various passes (e.g. jump threading, lcssa, loop rotate,
747 etc) for doing this sort of thing. The code generator has an similar
748 <a href="http://llvm.org/doxygen/MachineSSAUpdater_8h-source.html">
749 MachineSSAUpdater</a> class.</li>
750 <li>The <a href="http://llvm.org/doxygen/Regex_8h-source.html">
751 llvm/Support/Regex.h</a> header exposes a platform independent regular
752 expression API. Building on this, the <a
753 href="TestingGuide.html#FileCheck">FileCheck</a> utility now supports
754 regular exressions.</li>
755 <li>raw_ostream now supports a circular "debug stream" accessed with "dbgs()".
756 By default, this stream works the same way as "errs()", but if you pass
757 <tt>-debug-buffer-size=1000</tt> to opt, the debug stream is capped to a
758 fixed sized circular buffer and the output is printed at the end of the
759 program's execution. This is helpful if you have a long lived compiler
760 process and you're interested in seeing snapshots in time.</li>
766 <!--=========================================================================-->
767 <div class="doc_subsection">
768 <a name="otherimprovements">Other Improvements and New Features</a>
771 <div class="doc_text">
772 <p>Other miscellaneous features include:</p>
775 <li>You can now build LLVM as a big dynamic library (e.g. "libllvm2.7.so"). To
776 get this, configure LLVM with the --enable-shared option.</li>
778 <li>LLVM command line tools now overwrite their output by default, before they
779 would only do this with -f. This makes them more convenient to use, and
780 behave more like standard unix tools.</li>
782 <li>The opt and llc tools now autodetect whether their input is a .ll or .bc
783 file, and automatically do the right thing. This means you don't need to
784 explicitly use the llvm-as tool for most things.</li>
790 <!--=========================================================================-->
791 <div class="doc_subsection">
792 <a name="changes">Major Changes and Removed Features</a>
795 <div class="doc_text">
797 <p>If you're already an LLVM user or developer with out-of-tree changes based
798 on LLVM 2.6, this section lists some "gotchas" that you may run into upgrading
799 from the previous release.</p>
804 The Andersen's alias analysis ("anders-aa") pass, the Predicate Simplifier
805 ("predsimplify") pass, the LoopVR pass, the GVNPRE pass, and the random sampling
806 profiling ("rsprofiling") passes have all been removed. They were not being
807 actively maintained and had substantial problems. If you are interested in
808 these components, you are welcome to ressurect them from SVN, fix the
809 correctness problems, and resubmit them to mainline.</li>
811 <li>LLVM now defaults to building most libraries with RTTI turned off, providing
812 a code size reduction. Packagers who are interested in building LLVM to support
813 plugins that require RTTI information should build with "make REQUIRE_RTTI=1"
814 and should read the new <a href="Packaging.html">Advice on Packaging LLVM</a>
817 <li>The LLVM interpreter now defaults to <em>not</em> using <tt>libffi</tt> even
818 if you have it installed. This makes it more likely that an LLVM built on one
819 system will work when copied to a similar system. To use <tt>libffi</tt>,
820 configure with <tt>--enable-libffi</tt>.</li>
822 <li>Debug information uses a completely different representation, an LLVM 2.6
823 .bc file should work with LLVM 2.7, but debug info won't come forward.</li>
825 <li>The LLVM 2.6 (and earlier) "malloc" and "free" instructions got removed,
826 along with LowerAllocations pass. Now you should just use a call to the
827 malloc and free functions in libc. These calls are optimized as well as
828 the old instructions were.</li>
831 <p>In addition, many APIs have changed in this release. Some of the major LLVM
835 <li>Just about everything has been converted to use raw_ostream instead of
837 <li>llvm/ADT/iterator.h has been removed, just use <iterator>
839 <li>The Streams.h file and "DOUT" got removed, use "DEBUG(errs() << ...);"
841 <li><tt>ModuleProvider</tt> has been <a
842 href="http://llvm.org/viewvc/llvm-project?view=rev&revision=94686">removed</a>
843 and its methods moved to <tt>Module</tt> and <tt>GlobalValue</tt>.
844 Most clients can remove uses of <tt>ExistingModuleProvider</tt>,
845 replace <tt>getBitcodeModuleProvider</tt> with
846 <tt>getLazyBitcodeModule</tt>, and pass their <tt>Module</tt> to
847 functions that used to accept <tt>ModuleProvider</tt>. Clients who
848 wrote their own <tt>ModuleProvider</tt>s will need to derive from
849 <tt>GVMaterializer</tt> instead and use
850 <tt>Module::setMaterializer</tt> to attach it to a
851 <tt>Module</tt>.</li>
853 <li><tt>GhostLinkage</tt> has given up the ghost.
854 <tt>GlobalValue</tt>s that have not yet been read from their backing
855 storage have the same linkage they will have after being read in.
856 Clients must replace calls to
857 <tt>GlobalValue::hasNotBeenReadFromBitcode</tt> with
858 <tt>GlobalValue::isMaterializable</tt>.</li>
860 <li>The <tt>llvm/Support/DataTypes.h</tt> header has moved
861 to <tt>llvm/System/DataTypes.h</tt>.</li>
863 <li>The <tt>isInteger</tt>, <tt>isIntOrIntVector</tt>, <tt>isFloatingPoint</tt>,
864 <tt>isFPOrFPVector</tt> and <tt>isFPOrFPVector</tt> methods have been renamed
865 <tt>isIntegerTy</tt>, <tt>isIntOrIntVectorTy</tt>, <tt>isFloatingPointTy</tt>,
866 <tt>isFPOrFPVectorTy</tt> and <tt>isFPOrFPVectorTy</tt> respectively.</li>
873 <!-- *********************************************************************** -->
874 <div class="doc_section">
875 <a name="portability">Portability and Supported Platforms</a>
877 <!-- *********************************************************************** -->
879 <div class="doc_text">
881 <p>LLVM is known to work on the following platforms:</p>
884 <li>Intel and AMD machines (IA32, X86-64, AMD64, EMT-64) running Red Hat
885 Linux, Fedora Core, FreeBSD and AuroraUX (and probably other unix-like
887 <li>PowerPC and X86-based Mac OS X systems, running 10.4 and above in 32-bit
888 and 64-bit modes.</li>
889 <li>Intel and AMD machines running on Win32 using MinGW libraries (native).</li>
890 <li>Intel and AMD machines running on Win32 with the Cygwin libraries (limited
891 support is available for native builds with Visual C++).</li>
892 <li>Sun x86 and AMD64 machines running Solaris 10, OpenSolaris 0906.</li>
893 <li>Alpha-based machines running Debian GNU/Linux.</li>
896 <p>The core LLVM infrastructure uses GNU autoconf to adapt itself
897 to the machine and operating system on which it is built. However, minor
898 porting may be required to get LLVM to work on new platforms. We welcome your
899 portability patches and reports of successful builds or error messages.</p>
903 <!-- *********************************************************************** -->
904 <div class="doc_section">
905 <a name="knownproblems">Known Problems</a>
907 <!-- *********************************************************************** -->
909 <div class="doc_text">
911 <p>This section contains significant known problems with the LLVM system,
912 listed by component. If you run into a problem, please check the <a
913 href="http://llvm.org/bugs/">LLVM bug database</a> and submit a bug if
914 there isn't already one.</p>
917 <li>LLVM will not correctly compile on Solaris and/or OpenSolaris
918 using the stock GCC 3.x.x series 'out the box',
919 See: <a href="GettingStarted.html#brokengcc">Broken versions of GCC and other tools</a>.
920 However, A <a href="http://pkg.auroraux.org/GCC">Modern GCC Build</a>
921 for x86/x86-64 has been made available from the third party AuroraUX Project
922 that has been meticulously tested for bootstrapping LLVM & Clang.</li>
927 <!-- ======================================================================= -->
928 <div class="doc_subsection">
929 <a name="experimental">Experimental features included with this release</a>
932 <div class="doc_text">
934 <p>The following components of this LLVM release are either untested, known to
935 be broken or unreliable, or are in early development. These components should
936 not be relied on, and bugs should not be filed against them, but they may be
937 useful to some people. In particular, if you would like to work on one of these
938 components, please contact us on the <a
939 href="http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev">LLVMdev list</a>.</p>
942 <li>The MSIL, Alpha, SPU, MIPS, PIC16, Blackfin, MSP430, SystemZ and MicroBlaze
943 backends are experimental.</li>
944 <li><tt>llc</tt> "<tt>-filetype=asm</tt>" (the default) is the only
945 supported value for this option. The MachO writer is experimental, and
946 works much better in mainline SVN.</li>
951 <!-- ======================================================================= -->
952 <div class="doc_subsection">
953 <a name="x86-be">Known problems with the X86 back-end</a>
956 <div class="doc_text">
959 <li>The X86 backend does not yet support
960 all <a href="http://llvm.org/PR879">inline assembly that uses the X86
961 floating point stack</a>. It supports the 'f' and 't' constraints, but not
963 <li>The X86 backend generates inefficient floating point code when configured
964 to generate code for systems that don't have SSE2.</li>
965 <li>Win64 code generation wasn't widely tested. Everything should work, but we
966 expect small issues to happen. Also, llvm-gcc cannot build the mingw64
967 runtime currently due to lack of support for the 'u' inline assembly
968 constraint and for X87 floating point inline assembly.</li>
969 <li>The X86-64 backend does not yet support the LLVM IR instruction
970 <tt>va_arg</tt>. Currently, front-ends support variadic
971 argument constructs on X86-64 by lowering them manually.</li>
976 <!-- ======================================================================= -->
977 <div class="doc_subsection">
978 <a name="ppc-be">Known problems with the PowerPC back-end</a>
981 <div class="doc_text">
984 <li>The Linux PPC32/ABI support needs testing for the interpreter and static
985 compilation, and lacks support for debug information.</li>
990 <!-- ======================================================================= -->
991 <div class="doc_subsection">
992 <a name="arm-be">Known problems with the ARM back-end</a>
995 <div class="doc_text">
998 <li>Thumb mode works only on ARMv6 or higher processors. On sub-ARMv6
999 processors, thumb programs can crash or produce wrong
1000 results (<a href="http://llvm.org/PR1388">PR1388</a>).</li>
1001 <li>Compilation for ARM Linux OABI (old ABI) is supported but not fully tested.
1007 <!-- ======================================================================= -->
1008 <div class="doc_subsection">
1009 <a name="sparc-be">Known problems with the SPARC back-end</a>
1012 <div class="doc_text">
1015 <li>The SPARC backend only supports the 32-bit SPARC ABI (-m32); it does not
1016 support the 64-bit SPARC ABI (-m64).</li>
1021 <!-- ======================================================================= -->
1022 <div class="doc_subsection">
1023 <a name="mips-be">Known problems with the MIPS back-end</a>
1026 <div class="doc_text">
1029 <li>64-bit MIPS targets are not supported yet.</li>
1034 <!-- ======================================================================= -->
1035 <div class="doc_subsection">
1036 <a name="alpha-be">Known problems with the Alpha back-end</a>
1039 <div class="doc_text">
1043 <li>On 21164s, some rare FP arithmetic sequences which may trap do not have the
1044 appropriate nops inserted to ensure restartability.</li>
1049 <!-- ======================================================================= -->
1050 <div class="doc_subsection">
1051 <a name="c-be">Known problems with the C back-end</a>
1054 <div class="doc_text">
1057 <li><a href="http://llvm.org/PR802">The C backend has only basic support for
1058 inline assembly code</a>.</li>
1059 <li><a href="http://llvm.org/PR1658">The C backend violates the ABI of common
1060 C++ programs</a>, preventing intermixing between C++ compiled by the CBE and
1061 C++ code compiled with <tt>llc</tt> or native compilers.</li>
1062 <li>The C backend does not support all exception handling constructs.</li>
1063 <li>The C backend does not support arbitrary precision integers.</li>
1069 <!-- ======================================================================= -->
1070 <div class="doc_subsection">
1071 <a name="c-fe">Known problems with the llvm-gcc C and C++ front-end</a>
1074 <div class="doc_text">
1076 <p>The only major language feature of GCC not supported by llvm-gcc is
1077 the <tt>__builtin_apply</tt> family of builtins. However, some extensions
1078 are only supported on some targets. For example, trampolines are only
1079 supported on some targets (these are used when you take the address of a
1080 nested function).</p>
1084 <!-- ======================================================================= -->
1085 <div class="doc_subsection">
1086 <a name="fortran-fe">Known problems with the llvm-gcc Fortran front-end</a>
1089 <div class="doc_text">
1091 <li>Fortran support generally works, but there are still several unresolved bugs
1092 in <a href="http://llvm.org/bugs/">Bugzilla</a>. Please see the
1093 tools/gfortran component for details.</li>
1097 <!-- ======================================================================= -->
1098 <div class="doc_subsection">
1099 <a name="ada-fe">Known problems with the llvm-gcc Ada front-end</a>
1102 <div class="doc_text">
1103 The llvm-gcc 4.2 Ada compiler works fairly well; however, this is not a mature
1104 technology, and problems should be expected.
1106 <li>The Ada front-end currently only builds on X86-32. This is mainly due
1107 to lack of trampoline support (pointers to nested functions) on other platforms.
1108 However, it <a href="http://llvm.org/PR2006">also fails to build on X86-64</a>
1109 which does support trampolines.</li>
1110 <li>The Ada front-end <a href="http://llvm.org/PR2007">fails to bootstrap</a>.
1111 This is due to lack of LLVM support for <tt>setjmp</tt>/<tt>longjmp</tt> style
1112 exception handling, which is used internally by the compiler.
1113 Workaround: configure with <tt>--disable-bootstrap</tt>.</li>
1114 <li>The c380004, <a href="http://llvm.org/PR2010">c393010</a>
1115 and <a href="http://llvm.org/PR2421">cxg2021</a> ACATS tests fail
1116 (c380004 also fails with gcc-4.2 mainline).
1117 If the compiler is built with checks disabled then <a href="http://llvm.org/PR2010">c393010</a>
1118 causes the compiler to go into an infinite loop, using up all system memory.</li>
1119 <li>Some GCC specific Ada tests continue to crash the compiler.</li>
1120 <li>The <tt>-E</tt> binder option (exception backtraces)
1121 <a href="http://llvm.org/PR1982">does not work</a> and will result in programs
1122 crashing if an exception is raised. Workaround: do not use <tt>-E</tt>.</li>
1123 <li>Only discrete types <a href="http://llvm.org/PR1981">are allowed to start
1124 or finish at a non-byte offset</a> in a record. Workaround: do not pack records
1125 or use representation clauses that result in a field of a non-discrete type
1126 starting or finishing in the middle of a byte.</li>
1127 <li>The <tt>lli</tt> interpreter <a href="http://llvm.org/PR2009">considers
1128 'main' as generated by the Ada binder to be invalid</a>.
1129 Workaround: hand edit the file to use pointers for <tt>argv</tt> and
1130 <tt>envp</tt> rather than integers.</li>
1131 <li>The <tt>-fstack-check</tt> option <a href="http://llvm.org/PR2008">is
1136 <!-- *********************************************************************** -->
1137 <div class="doc_section">
1138 <a name="additionalinfo">Additional Information</a>
1140 <!-- *********************************************************************** -->
1142 <div class="doc_text">
1144 <p>A wide variety of additional information is available on the <a
1145 href="http://llvm.org">LLVM web page</a>, in particular in the <a
1146 href="http://llvm.org/docs/">documentation</a> section. The web page also
1147 contains versions of the API documentation which is up-to-date with the
1148 Subversion version of the source code.
1149 You can access versions of these documents specific to this release by going
1150 into the "<tt>llvm/doc/</tt>" directory in the LLVM tree.</p>
1152 <p>If you have any questions or comments about LLVM, please feel free to contact
1153 us via the <a href="http://llvm.org/docs/#maillist"> mailing
1158 <!-- *********************************************************************** -->
1162 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
1163 src="http://jigsaw.w3.org/css-validator/images/vcss-blue" alt="Valid CSS"></a>
1164 <a href="http://validator.w3.org/check/referer"><img
1165 src="http://www.w3.org/Icons/valid-html401-blue" alt="Valid HTML 4.01"></a>
1167 <a href="http://llvm.org/">LLVM Compiler Infrastructure</a><br>
1168 Last modified: $Date$