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>Previously announced in the 2.4, 2.5, and 2.6 LLVM releases, the Clang project also
149 includes an early stage static source code analysis tool for <a
150 href="http://clang.llvm.org/StaticAnalysis.html">automatically finding bugs</a>
151 in C and Objective-C programs. The tool performs checks to find
152 bugs that occur on a specific path within a program.</p>
154 <p>In the LLVM 2.7 time-frame, the analyzer core has sprouted legs and...</p>
158 <!--=========================================================================-->
159 <div class="doc_subsection">
160 <a name="vmkit">VMKit: JVM/CLI Virtual Machine Implementation</a>
163 <div class="doc_text">
165 The <a href="http://vmkit.llvm.org/">VMKit project</a> is an implementation of
166 a JVM and a CLI Virtual Machine (Microsoft .NET is an
167 implementation of the CLI) using LLVM for static and just-in-time
171 With the release of LLVM 2.7, VMKit has shifted to a great framework for writing
172 virtual machines. VMKit now offers precise and efficient garbage collection with
173 multi-threading support, thanks to the MMTk memory management toolkit, as well
174 as just in time and ahead of time compilation with LLVM. The major changes in
179 <li>Garbage collection: VMKit now uses the MMTk toolkit for garbage collectors.
180 The first collector to be ported is the MarkSweep collector, which is precise,
181 and drastically improves the performance of VMKit.</li>
182 <li>Line number information in the JVM: by using the debug metadata of LLVM, the
183 JVM now supports precise line number information, useful when printing a stack
185 <li>Interface calls in the JVM: we implemented a variant of the Interface Method
186 Table technique for interface calls in the JVM.
193 <!--=========================================================================-->
194 <div class="doc_subsection">
195 <a name="compiler-rt">compiler-rt: Compiler Runtime Library</a>
198 <div class="doc_text">
200 The new LLVM <a href="http://compiler-rt.llvm.org/">compiler-rt project</a>
201 is a simple library that provides an implementation of the low-level
202 target-specific hooks required by code generation and other runtime components.
203 For example, when compiling for a 32-bit target, converting a double to a 64-bit
204 unsigned integer is compiled into a runtime call to the "__fixunsdfdi"
205 function. The compiler-rt library provides highly optimized implementations of
206 this and other low-level routines (some are 3x faster than the equivalent
207 libgcc routines).</p>
210 All of the code in the compiler-rt project is available under the standard LLVM
211 License, a "BSD-style" license. New in LLVM 2.7: compiler_rt now
212 supports ARM targets.</p>
216 <!--=========================================================================-->
217 <div class="doc_subsection">
218 <a name="dragonegg">DragonEgg: llvm-gcc ported to gcc-4.5</a>
221 <div class="doc_text">
223 <a href="http://dragonegg.llvm.org/">DragonEgg</a> is a port of llvm-gcc to
224 gcc-4.5. Unlike llvm-gcc, which makes many intrusive changes to the underlying
225 gcc-4.2 code, dragonegg in theory does not require any gcc-4.5 modifications
226 whatsoever (currently one small patch is needed). This is thanks to the new
227 <a href="http://gcc.gnu.org/wiki/plugins">gcc plugin architecture</a>, which
228 makes it possible to modify the behaviour of gcc at runtime by loading a plugin,
229 which is nothing more than a dynamic library which conforms to the gcc plugin
230 interface. DragonEgg is a gcc plugin that causes the LLVM optimizers to be run
231 instead of the gcc optimizers, and the LLVM code generators instead of the gcc
232 code generators, just like llvm-gcc. To use it, you add
233 "-fplugin=path/dragonegg.so" to the gcc-4.5 command line, and gcc-4.5 magically
234 becomes llvm-gcc-4.5!
238 DragonEgg is still a work in progress. Currently C works very well, while C++,
239 Ada and Fortran work fairly well. All other languages either don't work at all,
240 or only work poorly. For the moment only the x86-32 and x86-64 targets are
241 supported, and only on linux and darwin (darwin needs an additional gcc patch).
245 DragonEgg is a new project which is seeing its first release with llvm-2.7.
251 <!--=========================================================================-->
252 <div class="doc_subsection">
253 <a name="mc">llvm-mc: Machine Code Toolkit</a>
256 <div class="doc_text">
258 The LLVM Machine Code (MC) Toolkit project is ...
262 MC Disassembler (with blog post), MCInstPrinter. Many X86 backend and AsmPrinter simplifications
263 Can transcode from GAS to intel syntax with "llvm-mc foo.s -output-asm-variant=1"
269 <!-- *********************************************************************** -->
270 <div class="doc_section">
271 <a name="externalproj">External Open Source Projects Using LLVM 2.7</a>
273 <!-- *********************************************************************** -->
275 <div class="doc_text">
277 <p>An exciting aspect of LLVM is that it is used as an enabling technology for
278 a lot of other language and tools projects. This section lists some of the
279 projects that have already been updated to work with LLVM 2.7.</p>
282 <!--=========================================================================-->
283 <div class="doc_subsection">
284 <a name="pure">Pure</a>
287 <div class="doc_text">
289 <a href="http://pure-lang.googlecode.com/">Pure</a>
290 is an algebraic/functional programming language based on term rewriting.
291 Programs are collections of equations which are used to evaluate expressions in
292 a symbolic fashion. Pure offers dynamic typing, eager and lazy evaluation,
293 lexical closures, a hygienic macro system (also based on term rewriting),
294 built-in list and matrix support (including list and matrix comprehensions) and
295 an easy-to-use C interface. The interpreter uses LLVM as a backend to
296 JIT-compile Pure programs to fast native code.</p>
298 <p>Pure versions 0.43 and later have been tested and are known to work with
299 LLVM 2.7 (and continue to work with older LLVM releases >= 2.5).</p>
303 <!--=========================================================================-->
304 <div class="doc_subsection">
305 <a name="RoadsendPHP">Roadsend PHP</a>
308 <div class="doc_text">
310 <a href="http://code.roadsend.com/rphp">Roadsend PHP</a> (rphp) is an open
311 source implementation of the PHP programming
312 language that uses LLVM for its optimizer, JIT and static compiler. This is a
313 reimplementation of an earlier project that is now based on LLVM.
317 <!--=========================================================================-->
318 <div class="doc_subsection">
319 <a name="UnladenSwallow">Unladen Swallow</a>
322 <div class="doc_text">
324 <a href="http://code.google.com/p/unladen-swallow/">Unladen Swallow</a> is a
325 branch of <a href="http://python.org/">Python</a> intended to be fully
326 compatible and significantly faster. It uses LLVM's optimization passes and JIT
331 <!--=========================================================================-->
332 <div class="doc_subsection">
333 <a name="tce">TTA-based Codesign Environment (TCE)</a>
336 <div class="doc_text">
338 <a href="http://tce.cs.tut.fi/">TCE</a> is a toolset for designing
339 application-specific processors (ASP) based on the Transport triggered
340 architecture (TTA). The toolset provides a complete co-design flow from C/C++
341 programs down to synthesizable VHDL and parallel program binaries. Processor
342 customization points include the register files, function units, supported
343 operations, and the interconnection network.</p>
345 <p>TCE uses llvm-gcc/Clang and LLVM for C/C++ language support, target
346 independent optimizations and also for parts of code generation. It generates
347 new LLVM-based code generators "on the fly" for the designed TTA processors and
348 loads them in to the compiler backend as runtime libraries to avoid per-target
349 recompilation of larger parts of the compiler chain.</p>
353 <!--=========================================================================-->
354 <div class="doc_subsection">
355 <a name="safecode">SAFECode Compiler</a>
358 <div class="doc_text">
360 <a href="http://safecode.cs.illinois.edu">SAFECode</a> is a memory safe C
361 compiler built using LLVM. It takes standard, unannotated C code, analyzes the
362 code to ensure that memory accesses and array indexing operations are safe, and
363 instruments the code with run-time checks when safety cannot be proven
369 <!-- *********************************************************************** -->
370 <div class="doc_section">
371 <a name="whatsnew">What's New in LLVM 2.7?</a>
373 <!-- *********************************************************************** -->
375 <div class="doc_text">
377 <p>This release includes a huge number of bug fixes, performance tweaks and
378 minor improvements. Some of the major improvements and new features are listed
384 <!--=========================================================================-->
385 <div class="doc_subsection">
386 <a name="orgchanges">LLVM Community Changes</a>
389 <div class="doc_text">
391 <p>In addition to changes to the code, between LLVM 2.6 and 2.7, a number of
392 organization changes have happened:
396 <li>LLVM has a new <a href="http://llvm.org/Logo.html">official logo</a>!</li>
398 <li>Ted Kremenek and Doug Gregor have stepped forward as <a
399 href="http://llvm.org/docs/DeveloperPolicy.html#owners">Code Owners</a> of the
400 Clang static analyzer and the Clang frontend, respectively.</li>
402 <li>LLVM now has an <a href="http://blog.llvm.org">official Blog</a> at
403 <a href="http://blog.llvm.org">http://blog.llvm.org</a>. This is a great way
404 to learn about new LLVM-related features as they are implemented. Several
405 features in this release are already explained on the blog.</li>
407 <li>The LLVM web pages are now checked into the SVN server, in the "www",
408 "www-pubs" and "www-releases" SVN modules. Previously they were hidden in a
409 largely inaccessible old CVS server.</li>
411 <li><a href="http://llvm.org">llvm.org</a> is now hosted on a new (and much
412 faster) server. It is still graciously hosted at the University of Illinois
413 of Urbana Champaign.</li>
417 <!--=========================================================================-->
418 <div class="doc_subsection">
419 <a name="majorfeatures">Major New Features</a>
422 <div class="doc_text">
424 <p>LLVM 2.7 includes several major new capabilities:</p>
427 <li>2.7 includes initial support for the <a
428 href="http://en.wikipedia.org/wiki/MicroBlaze">MicroBlaze</a> target.
429 MicroBlaze is a soft processor core designed for Xilinx FPGAs.</li>
431 <li>2.7 includes a new LLVM IR "extensible metadata" feature. This feature
432 supports many different use cases, including allowing front-end authors to
433 encode source level information into LLVM IR, which is consumed by later
434 language-specific passes. This is a great way to do high-level optimizations
435 like devirtualization, type-based alias analysis, etc. See the <a
436 href="http://blog.llvm.org/2010/04/extensible-metadata-in-llvm-ir.html">
437 Extensible Metadata Blog Post</a> for more information.</li>
439 <li>2.7 encodes <a href="SourceLevelDebugging.html">debug information</a>
440 in a completely new way, built on extensible metadata. The new implementation
441 is much more memory efficient and paves the way for improvements to optimized
442 code debugging experience.</li>
444 <li>2.7 now directly supports taking the address of a label and doing an
445 indirect branch through a pointer. This is particularly useful for
446 interpreter loops, and is used to implement the GCC "address of label"
447 extension. For more information, see the <a
448 href="http://blog.llvm.org/2010/01/address-of-label-and-indirect-branches.html">
449 Address of Label and Indirect Branches in LLVM IR Blog Post</a>.
451 <li>2.7 is the first release to start supporting APIs for assembling and
452 disassembling target machine code. These APIs are useful for a variety of
453 low level clients, and are surfaced in the new "enhanced disassembly" API.
454 For more information see the <a
455 href="http://blog.llvm.org/2010/01/x86-disassembler.html">The X86
456 Disassembler Blog Post</a> for more information.</li>
458 <li>2.7 includes major parts of the work required by the new MC Project,
459 which aims to rework our handling of low-level machine code. A few targets
460 have been refactored to support it, and work is underway to support a native
461 assembler in LLVM. This work is not complete in LLVM 2.7, but you has made
462 substantially more progress on LLVM mainline. You can read more about this
463 in the <a href="http://blog.llvm.org/2010/04/intro-to-llvm-mc-project.html">
464 Intro to the LLVM MC Project Blog Post</a>.
471 <!--=========================================================================-->
472 <div class="doc_subsection">
473 <a name="coreimprovements">LLVM IR and Core Improvements</a>
476 <div class="doc_text">
477 <p>LLVM IR has several new features for better support of new targets and that
478 expose new optimization opportunities:</p>
481 <li>LLVM IR now supports a 16-bit "half float" data type through <a
482 href="LangRef.html#int_fp16">two new intrinsics</a> and APFloat support.</li>
483 <li>LLVM IR supports two new <a href="LangRef.html#fnattrs">function
484 attributes</a>: inlinehint and alignstack(n). The former is a hint to the
485 optimizer that a function was declared 'inline' and thus the inliner should
486 weight it higher when considering inlining it. The later
487 indicates to the code generator that the function diverges from the platform
488 ABI on stack alignment.</li>
489 <li>The new <a href="LangRef.html#int_objectsize">llvm.objectsize</a> intrinsic
490 allows the optimizer to infer the sizes of memory objects in some cases.
491 This intrinsic is used to implement the GCC <tt>__builtin_object_size</tt>
493 <li>LLVM IR now supports marking load and store instructions with <a
494 href="LangRef.html#i_load">"non-temporal" hints</a> (building on the new
495 metadata feature). This hint encourages the code
496 generator to generate non-temporal accesses when possible, which are useful
497 for code that is carefully managing cache behavior. Currently, only the
498 X86 backend provides target support for this feature.</li>
500 <li>LLVM 2.7 has pre-alpha support for <a
501 href="http://llvm.org/docs/LangRef.html#t_union">unions in LLVM IR</a>.
502 Unfortunately, this support is not really usable in 2.7, so if you're
503 interested in pushing it forward, please help contribute to LLVM mainline.</li>
509 <!--=========================================================================-->
510 <div class="doc_subsection">
511 <a name="optimizer">Optimizer Improvements</a>
514 <div class="doc_text">
516 <p>In addition to a large array of minor performance tweaks and bug fixes, this
517 release includes a few major enhancements and additions to the optimizers:</p>
521 <li>The inliner reuses now merges arrays stack objects in different callees when
522 inlining multiple call sites into one function. This reduces the stack size
523 of the resultant function.</li>
524 <li>The -basicaa alias analysis pass (which is the default) has been improved to
525 be less dependent on "type safe" pointers. It can now look through bitcasts
526 and other constructs more aggressively, allowing better load/store
528 <li>The load elimination optimization in the GVN Pass [<a
529 href="http://blog.llvm.org/2009/12/introduction-to-load-elimination-in-gvn.html">intro
530 blog post</a>] has been substantially improved to be more aggressive about
531 partial redundancy elimination and do more aggressive phi translation. Please
533 href="http://blog.llvm.org/2009/12/advanced-topics-in-redundant-load.html">
534 Advanced Topics in Redundant Load Elimination with a Focus on PHI Translation
535 Blog Post</a> for more details.</li>
536 <li>The module <a href="LangRef.html#datalayout">target data string</a> now
537 includes a notion of what the 'native' integer data types a for the target,
538 which allows various optimizations to use. This helps mid-level
539 optimizations avoid promoting complex sequences of operations to data types
540 that are not natively supported (e.g. converting i32 operations to i64 on
542 <li>The mid-level optimizer is now conservative when operating on a module with
543 no target data. Previously, it would default to SparcV9 settings, which is
544 not what most people expected.</li>
545 <li>Jump threading is now much more aggressive at simplifying correlated
546 conditionals and threading blocks with otherwise complex logic. It has
547 subsumed the old "Conditional Propagation" pass, and -condprop has been
548 removed from LLVM 2.7.</li>
549 <li>The -instcombine pass has been refactored from being one huge file to being
550 a library of its own. Internally, it uses a customized IRBuilder to clean
551 it up and simplify it.</li>
553 <li>The optimal edge profiling pass is reliable and much more complete than in
554 2.6. It can be used with the llvm-prof tool but isn't wired up to the
555 llvm-gcc and clang command line options yet.</li>
557 <li>A new experimental alias analysis implementation, -scev-aa, has been added.
558 It uses LLVM's Scalar Evolution implementation to do symbolic analysis of
559 pointer offset expressions to disambiguate pointers. It can catch a few
560 cases that basicaa cannot, particularly in complex loop nests.</li>
562 <li>The default pass ordering has been tweaked for improved optimization
570 <!--=========================================================================-->
571 <div class="doc_subsection">
572 <a name="executionengine">Interpreter and JIT Improvements</a>
575 <div class="doc_text">
578 <li>The JIT now supports generating debug information, which is compatible with
579 the new GDB 7.0 (and later) interfaces for registering debug info for
580 dynamically generated code.</li>
582 <li>The JIT now <a href="http://llvm.org/PR5184">defaults
583 to compiling eagerly</a> to avoid a race condition in the lazy JIT.
584 Clients that still want the lazy JIT can switch it on by calling
585 <tt>ExecutionEngine::DisableLazyCompilation(false)</tt>.</li>
587 <li>It is now possible to create more than one JIT instance in the same process.
588 These JITs can generate machine code in parallel,
589 although <a href="http://llvm.org/docs/ProgrammersManual.html#jitthreading">you
590 still have to obey the other threading restrictions</a>.</li>
596 <!--=========================================================================-->
597 <div class="doc_subsection">
598 <a name="codegen">Target Independent Code Generator Improvements</a>
601 <div class="doc_text">
603 <p>We have put a significant amount of work into the code generator
604 infrastructure, which allows us to implement more aggressive algorithms and make
608 <li>The 'llc -asm-verbose' option (which is now the default) has been enhanced
609 to emit many useful comments to .s files indicating information about spill
610 slots and loop nest structure. This should make it much easier to read and
611 understand assembly files. This is wired up in llvm-gcc and clang to
612 the <tt>-fverbose-asm</tt> option.</li>
614 <li>New LSR with "full strength reduction" mode. FIXME: Description?</li>
616 <li>A new codegen level Common Subexpression Elimination pass (MachineCSE)
617 is available and enabled by default. It catches redundancies exposed by
619 <li>A new pre-register-allocation tail duplication pass is available and enabled
620 by default, it can substantially improve branch prediction quality in some
622 <li>A new sign and zero extension optimization pass (OptimizeExtsPass)
623 is available and enabled by default. This pass can takes advantage
624 architecture features like x86-64 implicit zero extension behavior and
626 <li>The code generator now supports a mode where it attempts to preserve the
627 order of instructions in the input code. This is important for source that
628 is hand scheduled and extremely sensitive to scheduling. It is compatible
629 with the GCC <tt>-fno-schedule-insns</tt> option.</li>
630 <li>The target-independent code generator now supports generating code with
631 arbitrary numbers of result values. Returning more values than was
632 previously supported is handled by returning through a hidden pointer. In
633 2.7, only the X86 and XCore targets have adopted support for this
635 <li>The code generator now supports generating code that follows the
636 <a href="LangRef.html#callingconv">Glasgow Haskell Compiler Calling
637 Convention</a> and ABI.</li>
638 <li>The "<a href="CodeGenerator.html#selectiondag_select">DAG instruction
639 selection</a>" phase of the code generator has been largely rewritten for
640 2.7. Previously, tblgen spit out tons of C++ code which was compiled and
641 linked into the target to do the pattern matching, now it emits a much
642 smaller table which is read by the target-independent code. The primary
643 advantages of this approach is that the size and compile time of various
644 targets is much improved. The X86 code generator shrunk by 1.5MB of code,
646 <li>Almost the entire code generator has switched to emitting code through the
647 MC interfaces instead of printing textually to the .s file. This led to a
648 number of cleanups and speedups. In 2.7, debug an exception handling
649 information does not go through MC yet.</li>
653 <!--=========================================================================-->
654 <div class="doc_subsection">
655 <a name="x86">X86-32 and X86-64 Target Improvements</a>
658 <div class="doc_text">
659 <p>New features of the X86 target include:
663 <li>The X86 backend now optimizes tails calls much more aggressively for
664 functions that use the standard C calling convention.</li>
665 <li>The X86 backend now models scalar SSE registers as subregs of the SSE vector
666 registers, making the code generator more aggressive in cases where scalars
667 and vector types are mixed.</li>
669 <li>PostRA scheduler for X86? FIXME: is this on by default in 2.7?</li>
675 <!--=========================================================================-->
676 <div class="doc_subsection">
677 <a name="ARM">ARM Target Improvements</a>
680 <div class="doc_text">
681 <p>New features of the ARM target include:
686 <li>The ARM backend now generates instructions in unified assembly syntax.</li>
688 <li>llvm-gcc now has complete support for the ARM v7 NEON instruction set. This
689 support differs slightly from the GCC implementation. Please see the
691 href="http://blog.llvm.org/2010/04/arm-advanced-simd-neon-intrinsics-and.html">
692 ARM Advanced SIMD (NEON) Intrinsics and Types in LLVM Blog Post</a> for
693 helpful information if migrating code from GCC to LLVM-GCC.</li>
695 <li>The ARM and Thumb code generators now using register scavenging for stack
696 object address materialization.(FIXME: WHAT BENEFIT DOES THIS PROVIDE?)</li>
698 <li>The ARM backend now has good support for ARMv4 targets, and has been tested
699 on StrongARM hardware. Previously, LLVM only supported ARMv4T and
706 <!--=========================================================================-->
707 <div class="doc_subsection">
708 <a name="newapis">New Useful APIs</a>
711 <div class="doc_text">
713 <p>This release includes a number of new APIs that are used internally, which
714 may also be useful for external clients.
718 <li>The optimizer uses the new CodeMetrics class to measure the size of code.
719 Various passes that use thing (like the inliner, loop unswitcher, etc) all
720 use this to make more accurate estimates of the code size impact of various
722 <li>A new <a href="http://llvm.org/doxygen/InstructionSimplify_8h-source.html">
723 llvm/Analysis/InstructionSimplify.h</a> interface available for doing
724 symbolic simplification of instructions (e.g. <tt>a+0</tt> -> <tt>a</tt>)
725 without requiring the instruction to exist. This centralizes a lot of
726 ad-hoc symbolic manipulation code scattered in various passes.</li>
727 <li>The optimizer now uses a new <a
728 href="http://llvm.org/doxygen/SSAUpdater_8h-source.html">SSAUpdater</a>
729 class which efficiently supports
730 doing unstructured SSA update operations. This centralized a bunch of code
731 scattered through various passes (e.g. jump threading, lcssa, loop rotate,
732 etc) for doing this sort of thing. The code generator has an similar
733 <a href="http://llvm.org/doxygen/MachineSSAUpdater_8h-source.html">
734 MachineSSAUpdater</a> class.</li>
735 <li>The <a href="http://llvm.org/doxygen/Regex_8h-source.html">
736 llvm/Support/Regex.h</a> header exposes a platform independent regular
737 expression API. Building on this, the <a
738 href="TestingGuide.html#FileCheck">FileCheck</a> utility now supports
739 regular exressions.</li>
740 <li>raw_ostream now supports a circular "debug stream" accessed with "dbgs()".
741 By default, this stream works the same way as "errs()", but if you pass
742 <tt>-debug-buffer-size=1000</tt> to opt, the debug stream is capped to a
743 fixed sized circular buffer and the output is printed at the end of the
744 program's execution. This is helpful if you have a long lived compiler
745 process and you're interested in seeing snapshots in time.</li>
751 <!--=========================================================================-->
752 <div class="doc_subsection">
753 <a name="otherimprovements">Other Improvements and New Features</a>
756 <div class="doc_text">
757 <p>Other miscellaneous features include:</p>
760 <li>You can now build LLVM as a big dynamic library (e.g. "libllvm2.7.so"). To
761 get this, configure LLVM with the --enable-shared option.</li>
763 <li>LLVM command line tools now overwrite their output by default, before they
764 would only do this with -f. This makes them more convenient to use, and
765 behave more like standard unix tools.</li>
767 <li>The opt and llc tools now autodetect whether their input is a .ll or .bc
768 file, and automatically do the right thing. This means you don't need to
769 explicitly use the llvm-as tool for most things.</li>
775 <!--=========================================================================-->
776 <div class="doc_subsection">
777 <a name="changes">Major Changes and Removed Features</a>
780 <div class="doc_text">
782 <p>If you're already an LLVM user or developer with out-of-tree changes based
783 on LLVM 2.6, this section lists some "gotchas" that you may run into upgrading
784 from the previous release.</p>
789 The Andersen's alias analysis ("anders-aa") pass, the Predicate Simplifier
790 ("predsimplify") pass, the LoopVR pass, the GVNPRE pass, and the random sampling
791 profiling ("rsprofiling") passes have all been removed. They were not being
792 actively maintained and had substantial problems. If you are interested in
793 these components, you are welcome to ressurect them from SVN, fix the
794 correctness problems, and resubmit them to mainline.</li>
796 <li>LLVM now defaults to building most libraries with RTTI turned off, providing
797 a code size reduction. Packagers who are interested in building LLVM to support
798 plugins that require RTTI information should build with "make REQUIRE_RTTI=1"
799 and should read the new <a href="Packaging.html">Advice on Packaging LLVM</a>
802 <li>The LLVM interpreter now defaults to <em>not</em> using <tt>libffi</tt> even
803 if you have it installed. This makes it more likely that an LLVM built on one
804 system will work when copied to a similar system. To use <tt>libffi</tt>,
805 configure with <tt>--enable-libffi</tt>.</li>
807 <li>Debug information uses a completely different representation, an LLVM 2.6
808 .bc file should work with LLVM 2.7, but debug info won't come forward.</li>
810 <li>The LLVM 2.6 (and earlier) "malloc" and "free" instructions got removed,
811 along with LowerAllocations pass. Now you should just use a call to the
812 malloc and free functions in libc. These calls are optimized as well as
813 the old instructions were.</li>
816 <p>In addition, many APIs have changed in this release. Some of the major LLVM
820 <li>Just about everything has been converted to use raw_ostream instead of
822 <li>llvm/ADT/iterator.h has been removed, just use <iterator>
824 <li>The Streams.h file and "DOUT" got removed, use "DEBUG(errs() << ...);"
826 <li><tt>ModuleProvider</tt> has been <a
827 href="http://llvm.org/viewvc/llvm-project?view=rev&revision=94686">removed</a>
828 and its methods moved to <tt>Module</tt> and <tt>GlobalValue</tt>.
829 Most clients can remove uses of <tt>ExistingModuleProvider</tt>,
830 replace <tt>getBitcodeModuleProvider</tt> with
831 <tt>getLazyBitcodeModule</tt>, and pass their <tt>Module</tt> to
832 functions that used to accept <tt>ModuleProvider</tt>. Clients who
833 wrote their own <tt>ModuleProvider</tt>s will need to derive from
834 <tt>GVMaterializer</tt> instead and use
835 <tt>Module::setMaterializer</tt> to attach it to a
836 <tt>Module</tt>.</li>
838 <li><tt>GhostLinkage</tt> has given up the ghost.
839 <tt>GlobalValue</tt>s that have not yet been read from their backing
840 storage have the same linkage they will have after being read in.
841 Clients must replace calls to
842 <tt>GlobalValue::hasNotBeenReadFromBitcode</tt> with
843 <tt>GlobalValue::isMaterializable</tt>.</li>
845 <li>The <tt>llvm/Support/DataTypes.h</tt> header has moved
846 to <tt>llvm/System/DataTypes.h</tt>.</li>
848 <li>The <tt>isInteger</tt>, <tt>isIntOrIntVector</tt>, <tt>isFloatingPoint</tt>,
849 <tt>isFPOrFPVector</tt> and <tt>isFPOrFPVector</tt> methods have been renamed
850 <tt>isIntegerTy</tt>, <tt>isIntOrIntVectorTy</tt>, <tt>isFloatingPointTy</tt>,
851 <tt>isFPOrFPVectorTy</tt> and <tt>isFPOrFPVectorTy</tt> respectively.</li>
858 <!-- *********************************************************************** -->
859 <div class="doc_section">
860 <a name="portability">Portability and Supported Platforms</a>
862 <!-- *********************************************************************** -->
864 <div class="doc_text">
866 <p>LLVM is known to work on the following platforms:</p>
869 <li>Intel and AMD machines (IA32, X86-64, AMD64, EMT-64) running Red Hat
870 Linux, Fedora Core, FreeBSD and AuroraUX (and probably other unix-like
872 <li>PowerPC and X86-based Mac OS X systems, running 10.4 and above in 32-bit
873 and 64-bit modes.</li>
874 <li>Intel and AMD machines running on Win32 using MinGW libraries (native).</li>
875 <li>Intel and AMD machines running on Win32 with the Cygwin libraries (limited
876 support is available for native builds with Visual C++).</li>
877 <li>Sun x86 and AMD64 machines running Solaris 10, OpenSolaris 0906.</li>
878 <li>Alpha-based machines running Debian GNU/Linux.</li>
881 <p>The core LLVM infrastructure uses GNU autoconf to adapt itself
882 to the machine and operating system on which it is built. However, minor
883 porting may be required to get LLVM to work on new platforms. We welcome your
884 portability patches and reports of successful builds or error messages.</p>
888 <!-- *********************************************************************** -->
889 <div class="doc_section">
890 <a name="knownproblems">Known Problems</a>
892 <!-- *********************************************************************** -->
894 <div class="doc_text">
896 <p>This section contains significant known problems with the LLVM system,
897 listed by component. If you run into a problem, please check the <a
898 href="http://llvm.org/bugs/">LLVM bug database</a> and submit a bug if
899 there isn't already one.</p>
902 <li>LLVM will not correctly compile on Solaris and/or OpenSolaris
903 using the stock GCC 3.x.x series 'out the box',
904 See: <a href="GettingStarted.html#brokengcc">Broken versions of GCC and other tools</a>.
905 However, A <a href="http://pkg.auroraux.org/GCC">Modern GCC Build</a>
906 for x86/x86-64 has been made available from the third party AuroraUX Project
907 that has been meticulously tested for bootstrapping LLVM & Clang.</li>
912 <!-- ======================================================================= -->
913 <div class="doc_subsection">
914 <a name="experimental">Experimental features included with this release</a>
917 <div class="doc_text">
919 <p>The following components of this LLVM release are either untested, known to
920 be broken or unreliable, or are in early development. These components should
921 not be relied on, and bugs should not be filed against them, but they may be
922 useful to some people. In particular, if you would like to work on one of these
923 components, please contact us on the <a
924 href="http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev">LLVMdev list</a>.</p>
927 <li>The MSIL, Alpha, SPU, MIPS, PIC16, Blackfin, MSP430, SystemZ and MicroBlaze
928 backends are experimental.</li>
929 <li><tt>llc</tt> "<tt>-filetype=asm</tt>" (the default) is the only
930 supported value for this option. The MachO writer is experimental, and
931 works much better in mainline SVN.</li>
936 <!-- ======================================================================= -->
937 <div class="doc_subsection">
938 <a name="x86-be">Known problems with the X86 back-end</a>
941 <div class="doc_text">
944 <li>The X86 backend does not yet support
945 all <a href="http://llvm.org/PR879">inline assembly that uses the X86
946 floating point stack</a>. It supports the 'f' and 't' constraints, but not
948 <li>The X86 backend generates inefficient floating point code when configured
949 to generate code for systems that don't have SSE2.</li>
950 <li>Win64 code generation wasn't widely tested. Everything should work, but we
951 expect small issues to happen. Also, llvm-gcc cannot build the mingw64
952 runtime currently due to lack of support for the 'u' inline assembly
953 constraint and for X87 floating point inline assembly.</li>
954 <li>The X86-64 backend does not yet support the LLVM IR instruction
955 <tt>va_arg</tt>. Currently, front-ends support variadic
956 argument constructs on X86-64 by lowering them manually.</li>
961 <!-- ======================================================================= -->
962 <div class="doc_subsection">
963 <a name="ppc-be">Known problems with the PowerPC back-end</a>
966 <div class="doc_text">
969 <li>The Linux PPC32/ABI support needs testing for the interpreter and static
970 compilation, and lacks support for debug information.</li>
975 <!-- ======================================================================= -->
976 <div class="doc_subsection">
977 <a name="arm-be">Known problems with the ARM back-end</a>
980 <div class="doc_text">
983 <li>Thumb mode works only on ARMv6 or higher processors. On sub-ARMv6
984 processors, thumb programs can crash or produce wrong
985 results (<a href="http://llvm.org/PR1388">PR1388</a>).</li>
986 <li>Compilation for ARM Linux OABI (old ABI) is supported but not fully tested.
992 <!-- ======================================================================= -->
993 <div class="doc_subsection">
994 <a name="sparc-be">Known problems with the SPARC back-end</a>
997 <div class="doc_text">
1000 <li>The SPARC backend only supports the 32-bit SPARC ABI (-m32); it does not
1001 support the 64-bit SPARC ABI (-m64).</li>
1006 <!-- ======================================================================= -->
1007 <div class="doc_subsection">
1008 <a name="mips-be">Known problems with the MIPS back-end</a>
1011 <div class="doc_text">
1014 <li>64-bit MIPS targets are not supported yet.</li>
1019 <!-- ======================================================================= -->
1020 <div class="doc_subsection">
1021 <a name="alpha-be">Known problems with the Alpha back-end</a>
1024 <div class="doc_text">
1028 <li>On 21164s, some rare FP arithmetic sequences which may trap do not have the
1029 appropriate nops inserted to ensure restartability.</li>
1034 <!-- ======================================================================= -->
1035 <div class="doc_subsection">
1036 <a name="c-be">Known problems with the C back-end</a>
1039 <div class="doc_text">
1042 <li><a href="http://llvm.org/PR802">The C backend has only basic support for
1043 inline assembly code</a>.</li>
1044 <li><a href="http://llvm.org/PR1658">The C backend violates the ABI of common
1045 C++ programs</a>, preventing intermixing between C++ compiled by the CBE and
1046 C++ code compiled with <tt>llc</tt> or native compilers.</li>
1047 <li>The C backend does not support all exception handling constructs.</li>
1048 <li>The C backend does not support arbitrary precision integers.</li>
1054 <!-- ======================================================================= -->
1055 <div class="doc_subsection">
1056 <a name="c-fe">Known problems with the llvm-gcc C and C++ front-end</a>
1059 <div class="doc_text">
1061 <p>The only major language feature of GCC not supported by llvm-gcc is
1062 the <tt>__builtin_apply</tt> family of builtins. However, some extensions
1063 are only supported on some targets. For example, trampolines are only
1064 supported on some targets (these are used when you take the address of a
1065 nested function).</p>
1069 <!-- ======================================================================= -->
1070 <div class="doc_subsection">
1071 <a name="fortran-fe">Known problems with the llvm-gcc Fortran front-end</a>
1074 <div class="doc_text">
1076 <li>Fortran support generally works, but there are still several unresolved bugs
1077 in <a href="http://llvm.org/bugs/">Bugzilla</a>. Please see the
1078 tools/gfortran component for details.</li>
1082 <!-- ======================================================================= -->
1083 <div class="doc_subsection">
1084 <a name="ada-fe">Known problems with the llvm-gcc Ada front-end</a>
1087 <div class="doc_text">
1088 The llvm-gcc 4.2 Ada compiler works fairly well; however, this is not a mature
1089 technology, and problems should be expected.
1091 <li>The Ada front-end currently only builds on X86-32. This is mainly due
1092 to lack of trampoline support (pointers to nested functions) on other platforms.
1093 However, it <a href="http://llvm.org/PR2006">also fails to build on X86-64</a>
1094 which does support trampolines.</li>
1095 <li>The Ada front-end <a href="http://llvm.org/PR2007">fails to bootstrap</a>.
1096 This is due to lack of LLVM support for <tt>setjmp</tt>/<tt>longjmp</tt> style
1097 exception handling, which is used internally by the compiler.
1098 Workaround: configure with <tt>--disable-bootstrap</tt>.</li>
1099 <li>The c380004, <a href="http://llvm.org/PR2010">c393010</a>
1100 and <a href="http://llvm.org/PR2421">cxg2021</a> ACATS tests fail
1101 (c380004 also fails with gcc-4.2 mainline).
1102 If the compiler is built with checks disabled then <a href="http://llvm.org/PR2010">c393010</a>
1103 causes the compiler to go into an infinite loop, using up all system memory.</li>
1104 <li>Some GCC specific Ada tests continue to crash the compiler.</li>
1105 <li>The <tt>-E</tt> binder option (exception backtraces)
1106 <a href="http://llvm.org/PR1982">does not work</a> and will result in programs
1107 crashing if an exception is raised. Workaround: do not use <tt>-E</tt>.</li>
1108 <li>Only discrete types <a href="http://llvm.org/PR1981">are allowed to start
1109 or finish at a non-byte offset</a> in a record. Workaround: do not pack records
1110 or use representation clauses that result in a field of a non-discrete type
1111 starting or finishing in the middle of a byte.</li>
1112 <li>The <tt>lli</tt> interpreter <a href="http://llvm.org/PR2009">considers
1113 'main' as generated by the Ada binder to be invalid</a>.
1114 Workaround: hand edit the file to use pointers for <tt>argv</tt> and
1115 <tt>envp</tt> rather than integers.</li>
1116 <li>The <tt>-fstack-check</tt> option <a href="http://llvm.org/PR2008">is
1121 <!-- *********************************************************************** -->
1122 <div class="doc_section">
1123 <a name="additionalinfo">Additional Information</a>
1125 <!-- *********************************************************************** -->
1127 <div class="doc_text">
1129 <p>A wide variety of additional information is available on the <a
1130 href="http://llvm.org">LLVM web page</a>, in particular in the <a
1131 href="http://llvm.org/docs/">documentation</a> section. The web page also
1132 contains versions of the API documentation which is up-to-date with the
1133 Subversion version of the source code.
1134 You can access versions of these documents specific to this release by going
1135 into the "<tt>llvm/doc/</tt>" directory in the LLVM tree.</p>
1137 <p>If you have any questions or comments about LLVM, please feel free to contact
1138 us via the <a href="http://llvm.org/docs/#maillist"> mailing
1143 <!-- *********************************************************************** -->
1147 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
1148 src="http://jigsaw.w3.org/css-validator/images/vcss-blue" alt="Valid CSS"></a>
1149 <a href="http://validator.w3.org/check/referer"><img
1150 src="http://www.w3.org/Icons/valid-html401-blue" alt="Valid HTML 4.01"></a>
1152 <a href="http://llvm.org/">LLVM Compiler Infrastructure</a><br>
1153 Last modified: $Date$