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">
7 <link rel="stylesheet" href="llvm.css" type="text/css">
8 <title>LLVM 2.8 Release Notes</title>
12 <div class="doc_title">LLVM 2.8 Release Notes</div>
14 <img align=right src="http://llvm.org/img/DragonSmall.png"
15 width="136" height="136" alt="LLVM Dragon Logo">
18 <li><a href="#intro">Introduction</a></li>
19 <li><a href="#subproj">Sub-project Status Update</a></li>
20 <li><a href="#externalproj">External Projects Using LLVM 2.8</a></li>
21 <li><a href="#whatsnew">What's New in LLVM 2.8?</a></li>
22 <li><a href="GettingStarted.html">Installation Instructions</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/2.7/docs/ReleaseNotes.html">LLVM 2.7
36 Release Notes</a>.</h1>
39 <!-- *********************************************************************** -->
40 <div class="doc_section">
41 <a name="intro">Introduction</a>
43 <!-- *********************************************************************** -->
45 <div class="doc_text">
47 <p>This document contains the release notes for the LLVM Compiler
48 Infrastructure, release 2.8. Here we describe the status of LLVM, including
49 major improvements from the previous release and significant known problems.
50 All LLVM releases may be downloaded from the <a
51 href="http://llvm.org/releases/">LLVM releases web site</a>.</p>
53 <p>For more information about LLVM, including information about the latest
54 release, please check out the <a href="http://llvm.org/">main LLVM
55 web site</a>. If you have questions or comments, the <a
56 href="http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev">LLVM Developer's
57 Mailing List</a> is a good place to send them.</p>
59 <p>Note that if you are reading this file from a Subversion checkout or the
60 main LLVM web page, this document applies to the <i>next</i> release, not the
61 current one. To see the release notes for a specific release, please see the
62 <a href="http://llvm.org/releases/">releases page</a>.</p>
69 include/llvm/Analysis/LiveValues.h => Dan
70 lib/Transforms/IPO/MergeFunctions.cpp => consider for 2.8.
75 <!-- Features that need text if they're finished for 2.9:
78 loop dependence analysis
80 CorrelatedValuePropagation
83 <!-- Announcement, lldb, libc++ -->
86 <!-- *********************************************************************** -->
87 <div class="doc_section">
88 <a name="subproj">Sub-project Status Update</a>
90 <!-- *********************************************************************** -->
92 <div class="doc_text">
94 The LLVM 2.8 distribution currently consists of code from the core LLVM
95 repository (which roughly includes the LLVM optimizers, code generators
96 and supporting tools), the Clang repository and the llvm-gcc repository. In
97 addition to this code, the LLVM Project includes other sub-projects that are in
98 development. Here we include updates on these subprojects.
104 <!--=========================================================================-->
105 <div class="doc_subsection">
106 <a name="clang">Clang: C/C++/Objective-C Frontend Toolkit</a>
109 <div class="doc_text">
111 <p><a href="http://clang.llvm.org/">Clang</a> is an LLVM front end for the C,
112 C++, and Objective-C languages. Clang aims to provide a better user experience
113 through expressive diagnostics, a high level of conformance to language
114 standards, fast compilation, and low memory use. Like LLVM, Clang provides a
115 modular, library-based architecture that makes it suitable for creating or
116 integrating with other development tools. Clang is considered a
117 production-quality compiler for C, Objective-C, C++ and Objective-C++ on x86
118 (32- and 64-bit), and for darwin-arm targets.</p>
120 <p>In the LLVM 2.8 time-frame, the Clang team has made many improvements:</p>
123 <li>Clang C++ is now feature-complete with respect to the ISO C++ 1998 and 2003 standards.</li>
124 <li>Added support for Objective-C++.</li>
125 <li>Clang now uses LLVM-MC to directly generate object code and to parse inline assembly (on Darwin).</li>
126 <li>Introduced many new warnings, including <code>-Wmissing-field-initializers</code>, <code>-Wshadow</code>, <code>-Wno-protocol</code>, <code>-Wtautological-compare</code>, <code>-Wstrict-selector-match</code>, <code>-Wcast-align</code>, <code>-Wunused</code> improvements, and greatly improved format-string checking.</li>
127 <li>Introduced the "libclang" library, a C interface to Clang intended to support IDE clients.</li>
128 <li>Added support for <code>#pragma GCC visibility</code>, <code>#pragma align</code>, and others.</li>
129 <li>Added support for SSE, ARM NEON, and Altivec.</li>
130 <li>Improved support for many Microsoft extensions.</li>
131 <li>Implemented support for blocks in C++.</li>
132 <li>Implemented precompiled headers for C++.</li>
133 <li>Improved abstract syntax trees to retain more accurate source information.</li>
134 <li>Added driver support for handling LLVM IR and bitcode files directly.</li>
135 <li>Major improvements to compiler correctness for exception handling.</li>
136 <li>Improved generated code quality in some areas:
138 <li>Good code generation for X86-32 and X86-64 ABI handling.</li>
139 <li>Improved code generation for bit-fields, although important work remains.</li>
145 <!--=========================================================================-->
146 <div class="doc_subsection">
147 <a name="clangsa">Clang Static Analyzer</a>
150 <div class="doc_text">
152 <p>The <a href="http://clang-analyzer.llvm.org/">Clang Static Analyzer</a>
153 project is an effort to use static source code analysis techniques to
154 automatically find bugs in C and Objective-C programs (and hopefully <a
155 href="http://clang-analyzer.llvm.org/dev_cxx.html">C++ in the
156 future</a>!). The tool is very good at finding bugs that occur on specific
157 paths through code, such as on error conditions.</p>
159 <p>The LLVM 2.8 release fixes a number of bugs and slightly improves precision
160 over 2.7, but there are no major new features in the release.
165 <!--=========================================================================-->
166 <div class="doc_subsection">
167 <a name="dragonegg">DragonEgg: llvm-gcc ported to gcc-4.5</a>
170 <div class="doc_text">
172 <a href="http://dragonegg.llvm.org/">DragonEgg</a> is a port of llvm-gcc to
173 gcc-4.5. Unlike llvm-gcc, dragonegg in theory does not require any gcc-4.5
174 modifications whatsoever (currently one small patch is needed) thanks to the
175 new <a href="http://gcc.gnu.org/wiki/plugins">gcc plugin architecture</a>.
176 DragonEgg is a gcc plugin that makes gcc-4.5 use the LLVM optimizers and code
177 generators instead of gcc's, just like with llvm-gcc.
181 DragonEgg is still a work in progress, but it is able to compile a lot of code,
182 for example all of gcc, LLVM and clang. Currently Ada, C, C++ and Fortran work
183 well, while all other languages either don't work at all or only work poorly.
184 For the moment only the x86-32 and x86-64 targets are supported, and only on
185 linux and darwin (darwin may need additional gcc patches).
189 The 2.8 release has the following notable changes:
191 <li>The plugin loads faster due to exporting fewer symbols.</li>
192 <li>Additional vector operations such as addps256 are now supported.</li>
193 <li>Ada global variables with no initial value are no longer zero initialized,
194 resulting in better optimization.</li>
195 <li>The '-fplugin-arg-dragonegg-enable-gcc-optzns' flag now runs all gcc
196 optimizers, rather than just a handful.</li>
197 <li>Fortran programs using common variables now link correctly.</li>
198 <li>GNU OMP constructs no longer crash the compiler.</li>
204 <!--=========================================================================-->
205 <div class="doc_subsection">
206 <a name="vmkit">VMKit: JVM/CLI Virtual Machine Implementation</a>
209 <div class="doc_text">
211 The <a href="http://vmkit.llvm.org/">VMKit project</a> is an implementation of
212 a Java Virtual Machine (Java VM or JVM) that uses LLVM for static and
213 just-in-time compilation. As of LLVM 2.8, VMKit now supports copying garbage
214 collectors, and can be configured to use MMTk's copy mark-sweep garbage
215 collector. In LLVM 2.8, the VMKit .NET VM is no longer being maintained.
219 <!--=========================================================================-->
220 <div class="doc_subsection">
221 <a name="compiler-rt">compiler-rt: Compiler Runtime Library</a>
224 <div class="doc_text">
226 The new LLVM <a href="http://compiler-rt.llvm.org/">compiler-rt project</a>
227 is a simple library that provides an implementation of the low-level
228 target-specific hooks required by code generation and other runtime components.
229 For example, when compiling for a 32-bit target, converting a double to a 64-bit
230 unsigned integer is compiled into a runtime call to the "__fixunsdfdi"
231 function. The compiler-rt library provides highly optimized implementations of
232 this and other low-level routines (some are 3x faster than the equivalent
233 libgcc routines).</p>
236 All of the code in the compiler-rt project is available under the standard LLVM
237 License, a "BSD-style" license. New in LLVM 2.8, compiler_rt now supports
238 soft floating point (for targets that don't have a real floating point unit),
239 and includes an extensive testsuite for the "blocks" language feature and the
240 blocks runtime included in compiler_rt.</p>
244 <!--=========================================================================-->
245 <div class="doc_subsection">
246 <a name="lldb">LLDB: Low Level Debugger</a>
249 <div class="doc_text">
251 <a href="http://lldb.llvm.org/">LLDB</a> is a brand new member of the LLVM
252 umbrella of projects. LLDB is a next generation, high-performance debugger. It
253 is built as a set of reusable components which highly leverage existing
254 libraries in the larger LLVM Project, such as the Clang expression parser, the
255 LLVM disassembler and the LLVM JIT.</p>
258 LLDB is in early development and not included as part of the LLVM 2.8 release,
259 but is mature enough to support basic debugging scenarios on Mac OS X in C,
260 Objective-C and C++. We'd really like help extending and expanding LLDB to
261 support new platforms, new languages, new architectures, and new features.
266 <!--=========================================================================-->
267 <div class="doc_subsection">
268 <a name="libc++">libc++: C++ Standard Library</a>
271 <div class="doc_text">
273 <a href="http://libc++.llvm.org/">libc++</a> is another new member of the LLVM
274 family. It is an implementation of the C++ standard library, written from the
275 ground up to specifically target the forthcoming C++'0X standard and focus on
276 delivering great performance.</p>
279 As of the LLVM 2.8 release, libc++ is virtually feature complete, but would
280 benefit from more testing and better integration with Clang++. It is also
281 looking forward to the C++ committee finalizing the C++'0x standard.
288 <!--=========================================================================-->
289 <div class="doc_subsection">
290 <a name="klee">KLEE: A Symbolic Execution Virtual Machine</a>
293 <div class="doc_text">
295 <a href="http://klee.llvm.org/">KLEE</a> is a symbolic execution framework for
296 programs in LLVM bitcode form. KLEE tries to symbolically evaluate "all" paths
297 through the application and records state transitions that lead to fault
298 states. This allows it to construct testcases that lead to faults and can even
299 be used to verify some algorithms.
302 <p>Although KLEE does not have any major new features as of 2.8, we have made
303 various minor improvements, particular to ease development:</p>
305 <li>Added support for LLVM 2.8. KLEE currently maintains compatibility with
306 LLVM 2.6, 2.7, and 2.8.</li>
307 <li>Added a buildbot for 2.6, 2.7, and trunk. A 2.8 buildbot will be coming
308 soon following release.</li>
309 <li>Fixed many C++ code issues to allow building with Clang++. Mostly
310 complete, except for the version of MiniSAT which is inside the KLEE STP
312 <li>Improved support for building with separate source and build
314 <li>Added support for "long double" on x86.</li>
315 <li>Initial work on KLEE support for using 'lit' test runner instead of
317 <li>Added <tt>configure</tt> support for using an external version of
324 <!-- *********************************************************************** -->
325 <div class="doc_section">
326 <a name="externalproj">External Open Source Projects Using LLVM 2.8</a>
328 <!-- *********************************************************************** -->
330 <div class="doc_text">
332 <p>An exciting aspect of LLVM is that it is used as an enabling technology for
333 a lot of other language and tools projects. This section lists some of the
334 projects that have already been updated to work with LLVM 2.8.</p>
337 <!--=========================================================================-->
338 <div class="doc_subsection">
339 <a name="tce">TTA-based Codesign Environment (TCE)</a>
342 <div class="doc_text">
344 <a href="http://tce.cs.tut.fi/">TCE</a> is a toolset for designing
345 application-specific processors (ASP) based on the Transport triggered
346 architecture (TTA). The toolset provides a complete co-design flow from C/C++
347 programs down to synthesizable VHDL and parallel program binaries. Processor
348 customization points include the register files, function units, supported
349 operations, and the interconnection network.</p>
351 <p>TCE uses llvm-gcc/Clang and LLVM for C/C++ language support, target
352 independent optimizations and also for parts of code generation. It generates
353 new LLVM-based code generators "on the fly" for the designed TTA processors and
354 loads them in to the compiler backend as runtime libraries to avoid per-target
355 recompilation of larger parts of the compiler chain.</p>
359 <!--=========================================================================-->
360 <div class="doc_subsection">
361 <a name="Horizon">Horizon Bytecode Compiler</a>
364 <div class="doc_text">
366 <a href="http://www.quokforge.org/projects/horizon">Horizon</a> is a bytecode
367 language and compiler written on top of LLVM, intended for producing
368 single-address-space managed code operating systems that
369 run faster than the equivalent multiple-address-space C systems.
370 More in-depth blurb is available on the <a
371 href="http://www.quokforge.org/projects/horizon/wiki/Wiki">wiki</a>.</p>
375 <!--=========================================================================-->
376 <div class="doc_subsection">
377 <a name="clamav">Clam AntiVirus</a>
380 <div class="doc_text">
382 <a href=http://www.clamav.net>Clam AntiVirus</a> is an open source (GPL)
383 anti-virus toolkit for UNIX, designed especially for e-mail scanning on mail
384 gateways. Since version 0.96 it has <a
385 href="http://vrt-sourcefire.blogspot.com/2010/09/introduction-to-clamavs-low-level.html">bytecode
386 signatures</a> that allow writing detections for complex malware. It
387 uses LLVM's JIT to speed up the execution of bytecode on
388 X86, X86-64, PPC32/64, falling back to its own interpreter otherwise.
389 The git version was updated to work with LLVM 2.8.
393 href="http://git.clamav.net/gitweb?p=clamav-bytecode-compiler.git;a=blob_plain;f=docs/user/clambc-user.pdf">
394 ClamAV bytecode compiler</a> uses Clang and LLVM to compile a C-like
395 language, insert runtime checks, and generate ClamAV bytecode.</p>
399 <!--=========================================================================-->
400 <div class="doc_subsection">
401 <a name="pure">Pure</a>
404 <div class="doc_text">
406 <a href="http://pure-lang.googlecode.com/">Pure</a>
407 is an algebraic/functional
408 programming language based on term rewriting. Programs are collections
409 of equations which are used to evaluate expressions in a symbolic
410 fashion. Pure offers dynamic typing, eager and lazy evaluation, lexical
411 closures, a hygienic macro system (also based on term rewriting),
412 built-in list and matrix support (including list and matrix
413 comprehensions) and an easy-to-use C interface. The interpreter uses
414 LLVM as a backend to JIT-compile Pure programs to fast native code.</p>
416 <p>Pure versions 0.44 and later have been tested and are known to work with
417 LLVM 2.8 (and continue to work with older LLVM releases >= 2.5).</p>
421 <!--=========================================================================-->
422 <div class="doc_subsection">
423 <a name="GHC">Glasgow Haskell Compiler (GHC)</a>
426 <div class="doc_text">
428 <a href="http://www.haskell.org/ghc/">GHC</a> is an open source,
429 state-of-the-art programming suite for
430 Haskell, a standard lazy functional programming language. It includes
431 an optimizing static compiler generating good code for a variety of
432 platforms, together with an interactive system for convenient, quick
435 <p>In addition to the existing C and native code generators, GHC 7.0 now
437 href="http://hackage.haskell.org/trac/ghc/wiki/Commentary/Compiler/Backends/LLVM">LLVM
438 code generator</a>. GHC supports LLVM 2.7 and later.</p>
442 <!--=========================================================================-->
443 <div class="doc_subsection">
444 <a name="Clay">Clay Programming Language</a>
447 <div class="doc_text">
449 <a href="http://tachyon.in/clay/">Clay</a> is a new systems programming
450 language that is specifically designed for generic programming. It makes
451 generic programming very concise thanks to whole program type propagation. It
452 uses LLVM as its backend.</p>
456 <!--=========================================================================-->
457 <div class="doc_subsection">
458 <a name="llvm-py">llvm-py Python Bindings for LLVM</a>
461 <div class="doc_text">
463 <a href="http://www.mdevan.org/llvm-py/">llvm-py</a> has been updated to work
464 with LLVM 2.8. llvm-py provides Python bindings for LLVM, allowing you to write a
465 compiler backend or a VM in Python.</p>
470 <!--=========================================================================-->
471 <div class="doc_subsection">
472 <a name="FAUST">FAUST Real-Time Audio Signal Processing Language</a>
475 <div class="doc_text">
477 <a href="http://faust.grame.fr">FAUST</a> is a compiled language for real-time
478 audio signal processing. The name FAUST stands for Functional AUdio STream. Its
479 programming model combines two approaches: functional programming and block
480 diagram composition. In addition with the C, C++, JAVA output formats, the
481 Faust compiler can now generate LLVM bitcode, and works with LLVM 2.7 and
486 <!--=========================================================================-->
487 <div class="doc_subsection">
488 <a name="jade">Jade Just-in-time Adaptive Decoder Engine</a>
491 <div class="doc_text">
493 href="http://sourceforge.net/apps/trac/orcc/wiki/JadeDocumentation">Jade</a>
494 (Just-in-time Adaptive Decoder Engine) is a generic video decoder engine using
495 LLVM for just-in-time compilation of video decoder configurations. Those
496 configurations are designed by MPEG Reconfigurable Video Coding (RVC) committee.
497 MPEG RVC standard is built on a stream-based dataflow representation of
498 decoders. It is composed of a standard library of coding tools written in
499 RVC-CAL language and a dataflow configuration — block diagram —
502 <p>Jade project is hosted as part of the <a href="http://orcc.sf.net">Open
503 RVC-CAL Compiler</a> and requires it to translate the RVC-CAL standard library
504 of video coding tools into an LLVM assembly code.</p>
508 <!--=========================================================================-->
509 <div class="doc_subsection">
510 <a name="neko_llvm_jit">LLVM JIT for Neko VM</a>
513 <div class="doc_text">
514 <p><a href="http://github.com/vava/neko_llvm_jit">Neko LLVM JIT</a>
515 replaces the standard Neko JIT with an LLVM-based implementation. While not
516 fully complete, it is already providing a 1.5x speedup on 64-bit systems.
517 Neko LLVM JIT requires LLVM 2.8 or later.</p>
521 <!--=========================================================================-->
522 <div class="doc_subsection">
523 <a name="crack">Crack Scripting Language</a>
526 <div class="doc_text">
528 <a href="http://code.google.com/p/crack-language/">Crack</a> aims to provide
529 the ease of development of a scripting language with the performance of a
530 compiled language. The language derives concepts from C++, Java and Python,
531 incorporating object-oriented programming, operator overloading and strong
532 typing. Crack 0.2 works with LLVM 2.7, and the forthcoming Crack 0.2.1 release
533 builds on LLVM 2.8.</p>
537 <!--=========================================================================-->
538 <div class="doc_subsection">
539 <a name="DresdenTM">Dresden TM Compiler (DTMC)</a>
542 <div class="doc_text">
544 <a href="http://tm.inf.tu-dresden.de">DTMC</a> provides support for
545 Transactional Memory, which is an easy-to-use and efficient way to synchronize
546 accesses to shared memory. Transactions can contain normal C/C++ code (e.g.,
547 <code>__transaction { list.remove(x); x.refCount--; }</code>) and will be executed
548 virtually atomically and isolated from other transactions.</p>
552 <!--=========================================================================-->
553 <div class="doc_subsection">
554 <a name="Kai">Kai Programming Language</a>
557 <div class="doc_text">
559 <a href="http://www.oriontransfer.co.nz/research/kai">Kai</a> (Japanese 会 for
560 meeting/gathering) is an experimental interpreter that provides a highly
561 extensible runtime environment and explicit control over the compilation
562 process. Programs are defined using nested symbolic expressions, which are all
563 parsed into first-class values with minimal intrinsic semantics. Kai can
564 generate optimised code at run-time (using LLVM) in order to exploit the nature
565 of the underlying hardware and to integrate with external software libraries.
566 It is a unique exploration into world of dynamic code compilation, and the
567 interaction between high level and low level semantics.</p>
571 <!--=========================================================================-->
572 <div class="doc_subsection">
573 <a name="OSL">OSL: Open Shading Language</a>
576 <div class="doc_text">
578 <a href="http://code.google.com/p/openshadinglanguage/">OSL</a> is a shading
579 language designed for use in physically based renderers and in particular
580 production rendering. By using LLVM instead of the interpreter, it was able to
581 meet its performance goals (>= C-code) while retaining the benefits of
582 runtime specialization and a portable high-level language.
589 <!-- *********************************************************************** -->
590 <div class="doc_section">
591 <a name="whatsnew">What's New in LLVM 2.8?</a>
593 <!-- *********************************************************************** -->
595 <div class="doc_text">
597 <p>This release includes a huge number of bug fixes, performance tweaks and
598 minor improvements. Some of the major improvements and new features are listed
604 <!--=========================================================================-->
605 <div class="doc_subsection">
606 <a name="majorfeatures">Major New Features</a>
609 <div class="doc_text">
611 <p>LLVM 2.8 includes several major new capabilities:</p>
614 <li>As mentioned above, <a href="#libc++">libc++</a> and <a
615 href="#lldb">LLDB</a> are major new additions to the LLVM collective.</li>
616 <li>LLVM 2.8 now has pretty decent support for debugging optimized code. You
617 should be able to reliably get debug info for function arguments, assuming
618 that the value is actually available where you have stopped.</li>
620 <li>A new 'llvm-diff' tool is available that does a semantic diff of .ll
622 <li>The <a href="#mc">MC subproject</a> has made major progress in this release.
623 Direct .o file writing support for darwin/x86[-64] is now reliable and
624 support for other targets and object file formats are in progress.</li>
629 <!--=========================================================================-->
630 <div class="doc_subsection">
631 <a name="coreimprovements">LLVM IR and Core Improvements</a>
634 <div class="doc_text">
635 <p>LLVM IR has several new features for better support of new targets and that
636 expose new optimization opportunities:</p>
639 <li>The <a href="LangRef.html#int_libc">memcpy, memmove, and memset</a>
640 intrinsics now take address space qualified pointers and a bit to indicate
641 whether the transfer is "<a href="LangRef.html#volatile">volatile</a>" or not.
643 <li>Per-instruction debug info metadata is much faster and uses less memory by
644 using the new DebugLoc class.</li>
645 <li>LLVM IR now has a more formalized concept of "<a
646 href="LangRef.html#trapvalues">trap values</a>", which allow the optimizer
647 to optimize more aggressively in the presence of undefined behavior, while
648 still producing predictable results.</li>
649 <li>LLVM IR now supports two new <a href="LangRef.html#linkage">linkage
650 types</a> (linker_private_weak and linker_private_weak_def_auto) which map
651 onto some obscure MachO concepts.</li>
656 <!--=========================================================================-->
657 <div class="doc_subsection">
658 <a name="optimizer">Optimizer Improvements</a>
661 <div class="doc_text">
663 <p>In addition to a large array of minor performance tweaks and bug fixes, this
664 release includes a few major enhancements and additions to the optimizers:</p>
667 <li>As mentioned above, the optimizer now has support for updating debug
668 information as it goes. A key aspect of this is the new <a
669 href="SourceLevelDebugging.html#format_common_value">llvm.dbg.value</a>
670 intrinsic. This intrinsic represents debug info for variables that are
671 promoted to SSA values (typically by mem2reg or the -scalarrepl passes).</li>
673 <li>The JumpThreading pass is now much more aggressive about implied value
674 relations, allowing it to thread conditions like "a == 4" when a is known to
675 be 13 in one of the predecessors of a block. It does this in conjunction
676 with the new LazyValueInfo analysis pass.</li>
677 <li>The new RegionInfo analysis pass identifies single-entry single-exit regions
678 in the CFG. You can play with it with the "opt -regions analyze" or
679 "opt -view-regions" commands.</li>
680 <li>The loop optimizer has significantly improved strength reduction and analysis
681 capabilities. Notably it is able to build on the trap value and signed
682 integer overflow information to optimize <= and >= loops.</li>
683 <li>The CallGraphSCCPassManager now has some basic support for iterating within
684 an SCC when a optimizer devirtualizes a function call. This allows inlining
685 through indirect call sites that are devirtualized by store-load forwarding
686 and other optimizations.</li>
687 <li>The new <A href="Passes.html#loweratomic">-loweratomic</a> pass is available
688 to lower atomic instructions into their non-atomic form. This can be useful
689 to optimize generic code that expects to run in a single-threaded
694 <p>In addition to these features that are done in 2.8, there is preliminary
695 support in the release for Type Based Alias Analysis
696 Preliminary work on TBAA but not usable in 2.8.
697 New CorrelatedValuePropagation pass, not on by default in 2.8 yet.
702 <!--=========================================================================-->
703 <div class="doc_subsection">
704 <a name="mc">MC Level Improvements</a>
707 <div class="doc_text">
709 The LLVM Machine Code (aka MC) subsystem was created to solve a number
710 of problems in the realm of assembly, disassembly, object file format handling,
711 and a number of other related areas that CPU instruction-set level tools work
714 <p>The MC subproject has made great leaps in LLVM 2.8. For example, support for
715 directly writing .o files from LLC (and clang) now works reliably for
716 darwin/x86[-64] (including inline assembly support) and the integrated
717 assembler is turned on by default in Clang for these targets. This provides
718 improved compile times among other things.</p>
721 <li>The entire compiler has converted over to using the MCStreamer assembler API
722 instead of writing out a .s file textually.</li>
723 <li>The "assembler parser" is far more mature than in 2.7, supporting a full
724 complement of directives, now supports assembler macros, etc.</li>
725 <li>The "assembler backend" has been completed, including support for relaxation
726 relocation processing and all the other things that an assembler does.</li>
727 <li>The MachO file format support is now fully functional and works.</li>
728 <li>The MC disassembler now fully supports ARM and Thumb. ARM assembler support
729 is still in early development though.</li>
730 <li>The X86 MC assembler now supports the X86 AES and AVX instruction set.</li>
731 <li>Work on ELF and COFF object files and ARM target support is well underway,
732 but isn't useful yet in LLVM 2.8. Please contact the llvmdev mailing list
733 if you're interested in this.</li>
736 <p>For more information, please see the <a
737 href="http://blog.llvm.org/2010/04/intro-to-llvm-mc-project.html">Intro to the
738 LLVM MC Project Blog Post</a>.
744 <!--=========================================================================-->
745 <div class="doc_subsection">
746 <a name="codegen">Target Independent Code Generator Improvements</a>
749 <div class="doc_text">
751 <p>We have put a significant amount of work into the code generator
752 infrastructure, which allows us to implement more aggressive algorithms and make
756 <li>The clang/gcc -momit-leaf-frame-pointer argument is now supported.</li>
757 <li>The clang/gcc -ffunction-sections and -fdata-sections arguments are now
758 supported on ELF targets (like GCC).</li>
759 <li>The MachineCSE pass is now tuned and on by default. It eliminates common
760 subexpressions that are exposed when lowering to machine instructions.</li>
761 <li>The "local" register allocator was replaced by a new "fast" register
762 allocator. This new allocator (which is often used at -O0) is substantially
763 faster and produces better code than the old local register allocator.</li>
764 <li>A new LLC "-regalloc=default" option is available, which automatically
765 chooses a register allocator based on the -O optimization level.</li>
766 <li>The common code generator code was modified to promote illegal argument and
767 return value vectors to wider ones when possible instead of scalarizing
768 them. For example, <3 x float> will now pass in one SSE register
769 instead of 3 on X86. This generates substantially better code since the
770 rest of the code generator was already expecting this.</li>
771 <li>The code generator uses a new "COPY" machine instruction. This speeds up
772 the code generator and eliminates the need for targets to implement the
773 isMoveInstr hook. Also, the copyRegToReg hook was renamed to copyPhysReg
775 <li>The code generator now has a "LocalStackSlotPass", which optimizes stack
776 slot access for targets (like ARM) that have limited stack displacement
778 <li>A new "PeepholeOptimizer" is available, which eliminates sign and zero
779 extends, and optimizes away compare instructions when the condition result
780 is available from a previous instruction.</li>
781 <li>Atomic operations now get legalized into simpler atomic operations if not
782 natively supported, easing the implementation burden on targets.</li>
783 <li>We have added two new bottom-up pre-allocation register pressure aware schedulers:
785 <li>The hybrid scheduler schedules aggressively to minimize schedule length when registers are available and avoid overscheduling in high pressure situations.</li>
786 <li>The instruction-level-parallelism scheduler schedules for maximum ILP when registers are available and avoid overscheduling in high pressure situations.</li>
788 <li>The tblgen type inference algorithm was rewritten to be more consistent and
789 diagnose more target bugs. If you have an out-of-tree backend, you may
790 find that it finds bugs in your target description. This support also
791 allows limited support for writing patterns for instructions that return
792 multiple results (e.g. a virtual register and a flag result). The
793 'parallel' modifier in tblgen was removed, you should use the new support
794 for multiple results instead.</li>
795 <li>A new (experimental) "-rendermf" pass is available which renders a
796 MachineFunction into HTML, showing live ranges and other useful
798 <li>The new SubRegIndex tablegen class allows subregisters to be indexed
799 symbolically instead of numerically. If your target uses subregisters you
800 will need to adapt to use SubRegIndex when you upgrade to 2.8.</li>
803 <li>The -fast-isel instruction selection path (used at -O0 on X86) was rewritten
804 to work bottom-up on basic blocks instead of top down. This makes it
805 slightly faster (because the MachineDCE pass is not needed any longer) and
806 allows it to generate better code in some cases.</li>
811 <!--=========================================================================-->
812 <div class="doc_subsection">
813 <a name="x86">X86-32 and X86-64 Target Improvements</a>
816 <div class="doc_text">
817 <p>New features and major changes in the X86 target include:
821 <li>The X86 backend now supports holding X87 floating point stack values
822 in registers across basic blocks, dramatically improving performance of code
823 that uses long double, and when targeting CPUs that don't support SSE.</li>
825 <li>The X86 backend now uses a SSEDomainFix pass to optimize SSE operations. On
826 Nehalem ("Core i7") and newer CPUs there is a 2 cycle latency penalty on
827 using a register in a different domain than where it was defined. This pass
828 optimizes away these stalls.</li>
830 <li>The X86 backend now promotes 16-bit integer operations to 32-bits when
831 possible. This avoids 0x66 prefixes, which are slow on some
832 microarchitectures and bloat the code on all of them.</li>
834 <li>The X86 backend now supports the Microsoft "thiscall" calling convention,
835 and a <a href="LangRef.html#callingconv">calling convention</a> to support
836 <a href="#GHC">ghc</a>.</li>
838 <li>The X86 backend supports a new "llvm.x86.int" intrinsic, which maps onto
839 the X86 "int $42" and "int3" instructions.</li>
841 <li>At the IR level, the <2 x float> datatype is now promoted and passed
842 around as a <4 x float> instead of being passed and returned as an MMX
843 vector. If you have a frontend that uses this, please pass and return a
844 <2 x i32> instead (using bitcasts).</li>
846 <li>When printing .s files in verbose assembly mode (the default for clang -S),
847 the X86 backend now decodes X86 shuffle instructions and prints human
848 readable comments after the most inscrutable of them, e.g.:
851 insertps $113, %xmm3, %xmm0 <i># xmm0 = zero,xmm0[1,2],xmm3[1]</i>
852 unpcklps %xmm1, %xmm0 <i># xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]</i>
853 pshufd $1, %xmm1, %xmm1 <i># xmm1 = xmm1[1,0,0,0]</i>
861 <!--=========================================================================-->
862 <div class="doc_subsection">
863 <a name="ARM">ARM Target Improvements</a>
866 <div class="doc_text">
867 <p>New features of the ARM target include:
871 <li>The ARM backend now optimizes tail calls into jumps.</li>
872 <li>Scheduling is improved through the new list-hybrid scheduler as well
873 as through better modeling of structural hazards.</li>
874 <li><a href="LangRef.html#int_fp16">Half float</a> instructions are now
876 <li>NEON support has been improved to model instructions which operate onto
877 multiple consecutive registers more aggressively. This avoids lots of
878 extraneous register copies.</li>
879 <li>The ARM backend now uses a new "ARMGlobalMerge" pass, which merges several
880 global variables into one, saving extra address computation (all the global
881 variables can be accessed via same base address) and potentially reducing
882 register pressure.</li>
884 <li>The ARM has received many minor improvements and tweaks which lead to
885 substantially better performance in a wide range of different scenarios.</li>
887 <li>The ARM NEON intrinsics have been substantially reworked to reduce
888 redundancy and improve code generation. Some of the major changes are:
891 All of the NEON load and store intrinsics (llvm.arm.neon.vld* and
892 llvm.arm.neon.vst*) take an extra parameter to specify the alignment in bytes
893 of the memory being accessed.
896 The llvm.arm.neon.vaba intrinsic (vector absolute difference and
897 accumulate) has been removed. This operation is now represented using
898 the llvm.arm.neon.vabd intrinsic (vector absolute difference) followed by a
902 The llvm.arm.neon.vabdl and llvm.arm.neon.vabal intrinsics (lengthening
903 vector absolute difference with and without accumulation) have been removed.
904 They are represented using the llvm.arm.neon.vabd intrinsic (vector absolute
905 difference) followed by a vector zero-extend operation, and for vabal,
909 The llvm.arm.neon.vmovn intrinsic has been removed. Calls of this intrinsic
910 are now replaced by vector truncate operations.
913 The llvm.arm.neon.vmovls and llvm.arm.neon.vmovlu intrinsics have been
914 removed. They are now represented as vector sign-extend (vmovls) and
915 zero-extend (vmovlu) operations.
918 The llvm.arm.neon.vaddl*, llvm.arm.neon.vaddw*, llvm.arm.neon.vsubl*, and
919 llvm.arm.neon.vsubw* intrinsics (lengthening vector add and subtract) have
920 been removed. They are replaced by vector add and vector subtract operations
921 where one (vaddw, vsubw) or both (vaddl, vsubl) of the operands are either
922 sign-extended or zero-extended.
925 The llvm.arm.neon.vmulls, llvm.arm.neon.vmullu, llvm.arm.neon.vmlal*, and
926 llvm.arm.neon.vmlsl* intrinsics (lengthening vector multiply with and without
927 accumulation and subtraction) have been removed. These operations are now
928 represented as vector multiplications where the operands are either
929 sign-extended or zero-extended, followed by a vector add for vmlal or a
930 vector subtract for vmlsl. Note that the polynomial vector multiply
931 intrinsic, llvm.arm.neon.vmullp, remains unchanged.
940 <!--=========================================================================-->
941 <div class="doc_subsection">
942 <a name="changes">Major Changes and Removed Features</a>
945 <div class="doc_text">
947 <p>If you're already an LLVM user or developer with out-of-tree changes based
948 on LLVM 2.7, this section lists some "gotchas" that you may run into upgrading
949 from the previous release.</p>
952 <li>The build configuration machinery changed the output directory names. It
953 wasn't clear to many people that a "Release-Asserts" build was a release build
954 without asserts. To make this more clear, "Release" does not include
955 assertions and "Release+Asserts" does (likewise, "Debug" and
956 "Debug+Asserts").</li>
957 <li>The MSIL Backend was removed, it was unsupported and broken.</li>
958 <li>The ABCD, SSI, and SCCVN passes were removed. These were not fully
959 functional and their behavior has been or will be subsumed by the
960 LazyValueInfo pass.</li>
961 <li>The LLVM IR 'Union' feature was removed. While this is a desirable feature
962 for LLVM IR to support, the existing implementation was half baked and
963 barely useful. We'd really like anyone interested to resurrect the work and
964 finish it for a future release.</li>
965 <li>If you're used to reading .ll files, you'll probably notice that .ll file
966 dumps don't produce #uses comments anymore. To get them, run a .bc file
967 through "llvm-dis --show-annotations".</li>
968 <li>Target triples are now stored in a normalized form, and all inputs from
969 humans are expected to be normalized by Triple::normalize before being
970 stored in a module triple or passed to another library.</li>
975 <p>In addition, many APIs have changed in this release. Some of the major LLVM
978 <li>LLVM 2.8 changes the internal order of operands in <a
979 href="http://llvm.org/doxygen/classllvm_1_1InvokeInst.html"><tt>InvokeInst</tt></a>
980 and <a href="http://llvm.org/doxygen/classllvm_1_1CallInst.html"><tt>CallInst</tt></a>.
981 To be portable across releases, please use the <tt>CallSite</tt> class and the
982 high-level accessors, such as <tt>getCalledValue</tt> and
983 <tt>setUnwindDest</tt>.
986 You can no longer pass use_iterators directly to cast<> (and similar),
987 because these routines tend to perform costly dereference operations more
988 than once. You have to dereference the iterators yourself and pass them in.
991 llvm.memcpy.*, llvm.memset.*, llvm.memmove.* intrinsics take an extra
992 parameter now ("i1 isVolatile"), totaling 5 parameters, and the pointer
993 operands are now address-space qualified.
994 If you were creating these intrinsic calls and prototypes yourself (as opposed
995 to using Intrinsic::getDeclaration), you can use
996 UpgradeIntrinsicFunction/UpgradeIntrinsicCall to be portable across releases.
999 SetCurrentDebugLocation takes a DebugLoc now instead of a MDNode.
1000 Change your code to use
1001 SetCurrentDebugLocation(DebugLoc::getFromDILocation(...)).
1004 The <tt>RegisterPass</tt> and <tt>RegisterAnalysisGroup</tt> templates are
1005 considered deprecated, but continue to function in LLVM 2.8. Clients are
1006 strongly advised to use the upcoming <tt>INITIALIZE_PASS()</tt> and
1007 <tt>INITIALIZE_AG_PASS()</tt> macros instead.
1010 The constructor for the Triple class no longer tries to understand odd triple
1011 specifications. Frontends should ensure that they only pass valid triples to
1012 LLVM. The Triple::normalize utility method has been added to help front-ends
1013 deal with funky triples.
1017 Some APIs got renamed:
1019 <li>llvm_report_error -> report_fatal_error</li>
1020 <li>llvm_install_error_handler -> install_fatal_error_handler</li>
1021 <li>llvm::DwarfExceptionHandling -> llvm::JITExceptionHandling</li>
1022 <li>VISIBILITY_HIDDEN -> LLVM_LIBRARY_VISIBILITY</li>
1031 <!-- *********************************************************************** -->
1032 <div class="doc_section">
1033 <a name="knownproblems">Known Problems</a>
1035 <!-- *********************************************************************** -->
1037 <div class="doc_text">
1039 <p>This section contains significant known problems with the LLVM system,
1040 listed by component. If you run into a problem, please check the <a
1041 href="http://llvm.org/bugs/">LLVM bug database</a> and submit a bug if
1042 there isn't already one.</p>
1046 <!-- ======================================================================= -->
1047 <div class="doc_subsection">
1048 <a name="experimental">Experimental features included with this release</a>
1051 <div class="doc_text">
1053 <p>The following components of this LLVM release are either untested, known to
1054 be broken or unreliable, or are in early development. These components should
1055 not be relied on, and bugs should not be filed against them, but they may be
1056 useful to some people. In particular, if you would like to work on one of these
1057 components, please contact us on the <a
1058 href="http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev">LLVMdev list</a>.</p>
1061 <li>The Alpha, Blackfin, CellSPU, MicroBlaze, MSP430, MIPS, PIC16, SystemZ
1062 and XCore backends are experimental.</li>
1063 <li><tt>llc</tt> "<tt>-filetype=obj</tt>" is experimental on all targets
1064 other than darwin-i386 and darwin-x86_64.</li>
1069 <!-- ======================================================================= -->
1070 <div class="doc_subsection">
1071 <a name="x86-be">Known problems with the X86 back-end</a>
1074 <div class="doc_text">
1077 <li>The X86 backend does not yet support
1078 all <a href="http://llvm.org/PR879">inline assembly that uses the X86
1079 floating point stack</a>. It supports the 'f' and 't' constraints, but not
1081 <li>Win64 code generation wasn't widely tested. Everything should work, but we
1082 expect small issues to happen. Also, llvm-gcc cannot build the mingw64
1083 runtime currently due to lack of support for the 'u' inline assembly
1084 constraint and for X87 floating point inline assembly.</li>
1085 <li>The X86-64 backend does not yet support the LLVM IR instruction
1086 <tt>va_arg</tt>. Currently, front-ends support variadic
1087 argument constructs on X86-64 by lowering them manually.</li>
1092 <!-- ======================================================================= -->
1093 <div class="doc_subsection">
1094 <a name="ppc-be">Known problems with the PowerPC back-end</a>
1097 <div class="doc_text">
1100 <li>The Linux PPC32/ABI support needs testing for the interpreter and static
1101 compilation, and lacks support for debug information.</li>
1106 <!-- ======================================================================= -->
1107 <div class="doc_subsection">
1108 <a name="arm-be">Known problems with the ARM back-end</a>
1111 <div class="doc_text">
1114 <li>Thumb mode works only on ARMv6 or higher processors. On sub-ARMv6
1115 processors, thumb programs can crash or produce wrong
1116 results (<a href="http://llvm.org/PR1388">PR1388</a>).</li>
1117 <li>Compilation for ARM Linux OABI (old ABI) is supported but not fully tested.
1123 <!-- ======================================================================= -->
1124 <div class="doc_subsection">
1125 <a name="sparc-be">Known problems with the SPARC back-end</a>
1128 <div class="doc_text">
1131 <li>The SPARC backend only supports the 32-bit SPARC ABI (-m32); it does not
1132 support the 64-bit SPARC ABI (-m64).</li>
1137 <!-- ======================================================================= -->
1138 <div class="doc_subsection">
1139 <a name="mips-be">Known problems with the MIPS back-end</a>
1142 <div class="doc_text">
1145 <li>64-bit MIPS targets are not supported yet.</li>
1150 <!-- ======================================================================= -->
1151 <div class="doc_subsection">
1152 <a name="alpha-be">Known problems with the Alpha back-end</a>
1155 <div class="doc_text">
1159 <li>On 21164s, some rare FP arithmetic sequences which may trap do not have the
1160 appropriate nops inserted to ensure restartability.</li>
1165 <!-- ======================================================================= -->
1166 <div class="doc_subsection">
1167 <a name="c-be">Known problems with the C back-end</a>
1170 <div class="doc_text">
1172 <p>The C backend has numerous problems and is not being actively maintained.
1173 Depending on it for anything serious is not advised.</p>
1176 <li><a href="http://llvm.org/PR802">The C backend has only basic support for
1177 inline assembly code</a>.</li>
1178 <li><a href="http://llvm.org/PR1658">The C backend violates the ABI of common
1179 C++ programs</a>, preventing intermixing between C++ compiled by the CBE and
1180 C++ code compiled with <tt>llc</tt> or native compilers.</li>
1181 <li>The C backend does not support all exception handling constructs.</li>
1182 <li>The C backend does not support arbitrary precision integers.</li>
1188 <!-- ======================================================================= -->
1189 <div class="doc_subsection">
1190 <a name="llvm-gcc">Known problems with the llvm-gcc front-end</a>
1193 <div class="doc_text">
1195 <p>llvm-gcc is generally very stable for the C family of languages. The only
1196 major language feature of GCC not supported by llvm-gcc is the
1197 <tt>__builtin_apply</tt> family of builtins. However, some extensions
1198 are only supported on some targets. For example, trampolines are only
1199 supported on some targets (these are used when you take the address of a
1200 nested function).</p>
1202 <p>Fortran support generally works, but there are still several unresolved bugs
1203 in <a href="http://llvm.org/bugs/">Bugzilla</a>. Please see the
1204 tools/gfortran component for details. Note that llvm-gcc is missing major
1205 Fortran performance work in the frontend and library that went into GCC after
1206 4.2. If you are interested in Fortran, we recommend that you consider using
1207 <a href="#dragonegg">dragonegg</a> instead.</p>
1209 <p>The llvm-gcc 4.2 Ada compiler has basic functionality, but is no longer being
1210 actively maintained. If you are interested in Ada, we recommend that you
1211 consider using <a href="#dragonegg">dragonegg</a> instead.</p>
1214 <!-- *********************************************************************** -->
1215 <div class="doc_section">
1216 <a name="additionalinfo">Additional Information</a>
1218 <!-- *********************************************************************** -->
1220 <div class="doc_text">
1222 <p>A wide variety of additional information is available on the <a
1223 href="http://llvm.org">LLVM web page</a>, in particular in the <a
1224 href="http://llvm.org/docs/">documentation</a> section. The web page also
1225 contains versions of the API documentation which is up-to-date with the
1226 Subversion version of the source code.
1227 You can access versions of these documents specific to this release by going
1228 into the "<tt>llvm/doc/</tt>" directory in the LLVM tree.</p>
1230 <p>If you have any questions or comments about LLVM, please feel free to contact
1231 us via the <a href="http://llvm.org/docs/#maillist"> mailing
1236 <!-- *********************************************************************** -->
1240 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
1241 src="http://jigsaw.w3.org/css-validator/images/vcss-blue" alt="Valid CSS"></a>
1242 <a href="http://validator.w3.org/check/referer"><img
1243 src="http://www.w3.org/Icons/valid-html401-blue" alt="Valid HTML 4.01"></a>
1245 <a href="http://llvm.org/">LLVM Compiler Infrastructure</a><br>
1246 Last modified: $Date$