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 2.9 Release Notes</title>
12 <h1 class="doc_title">LLVM 2.9 Release Notes</h1>
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 2.9</a></li>
21 <li><a href="#whatsnew">What's New in LLVM 2.9?</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>
31 <h1 style="color:red">These are in-progress notes for the upcoming LLVM 2.9
34 <a href="http://llvm.org/releases/2.8/docs/ReleaseNotes.html">LLVM 2.8
35 Release Notes</a>.</h1>
37 <!-- *********************************************************************** -->
39 <a name="intro">Introduction</a>
41 <!-- *********************************************************************** -->
43 <div class="doc_text">
45 <p>This document contains the release notes for the LLVM Compiler
46 Infrastructure, release 2.9. Here we describe the status of LLVM, including
47 major improvements from the previous release and significant known problems.
48 All LLVM releases may be downloaded from the <a
49 href="http://llvm.org/releases/">LLVM releases web site</a>.</p>
51 <p>For more information about LLVM, including information about the latest
52 release, please check out the <a href="http://llvm.org/">main LLVM
53 web site</a>. If you have questions or comments, the <a
54 href="http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev">LLVM Developer's
55 Mailing List</a> is a good place to send them.</p>
57 <p>Note that if you are reading this file from a Subversion checkout or the
58 main LLVM web page, this document applies to the <i>next</i> release, not the
59 current one. To see the release notes for a specific release, please see the
60 <a href="http://llvm.org/releases/">releases page</a>.</p>
64 <!-- Features that need text if they're finished for 3.1:
68 loop dependence analysis
69 CorrelatedValuePropagation
70 lib/Transforms/IPO/MergeFunctions.cpp => consider for 3.1.
73 <!-- *********************************************************************** -->
75 <a name="subproj">Sub-project Status Update</a>
77 <!-- *********************************************************************** -->
79 <div class="doc_text">
81 The LLVM 2.9 distribution currently consists of code from the core LLVM
82 repository (which roughly includes the LLVM optimizers, code generators
83 and supporting tools), the Clang repository and the llvm-gcc repository. In
84 addition to this code, the LLVM Project includes other sub-projects that are in
85 development. Here we include updates on these subprojects.
91 <!--=========================================================================-->
93 <a name="clang">Clang: C/C++/Objective-C Frontend Toolkit</a>
96 <div class="doc_text">
98 <p><a href="http://clang.llvm.org/">Clang</a> is an LLVM front end for the C,
99 C++, and Objective-C languages. Clang aims to provide a better user experience
100 through expressive diagnostics, a high level of conformance to language
101 standards, fast compilation, and low memory use. Like LLVM, Clang provides a
102 modular, library-based architecture that makes it suitable for creating or
103 integrating with other development tools. Clang is considered a
104 production-quality compiler for C, Objective-C, C++ and Objective-C++ on x86
105 (32- and 64-bit), and for darwin/arm targets.</p>
107 <p>In the LLVM 2.9 time-frame, the Clang team has made many improvements in C,
108 C++ and Objective-C support. C++ support is now generally rock solid, has
109 been exercised on a broad variety of code, and has several new <a
110 href="http://clang.llvm.org/cxx_status.html#cxx0x">C++'0x features</a>
111 implemented (such as rvalue references and variadic templates). LLVM 2.9 has
112 also brought in a large range of bug fixes and minor features (e.g. __label__
113 support), and is much more compatible with the Linux Kernel.</p>
115 <p>If Clang rejects your code that is built with another compiler, please take a
116 look at the <a href="http://clang.llvm.org/compatibility.html">language
117 compatibility</a> guide to make sure the issue isn't intentional or a known
125 <!--=========================================================================-->
127 <a name="dragonegg">DragonEgg: GCC front-ends, LLVM back-end</a>
130 <div class="doc_text">
132 <a href="http://dragonegg.llvm.org/">DragonEgg</a> is a
133 <a href="http://gcc.gnu.org/wiki/plugins">gcc plugin</a> that replaces GCC's
134 optimizers and code generators with LLVM's.
135 Currently it requires a patched version of gcc-4.5.
136 The plugin can target the x86-32 and x86-64 processor families and has been
137 used successfully on the Darwin, FreeBSD and Linux platforms.
138 The Ada, C, C++ and Fortran languages work well.
139 The plugin is capable of compiling plenty of Obj-C, Obj-C++ and Java but it is
140 not known whether the compiled code actually works or not!
144 The 2.9 release has the following notable changes:
146 <li>The plugin is much more stable when compiling Fortran.</li>
147 <li>Inline assembly where an asm output is tied to an input of a different size
148 is now supported in many more cases.</li>
149 <li>Basic support for the __float128 type was added. It is now possible to
150 generate LLVM IR from programs using __float128 but code generation does not
152 <li>Compiling Java programs no longer systematically crashes the plugin.</li>
157 <!--=========================================================================-->
159 <a name="compiler-rt">compiler-rt: Compiler Runtime Library</a>
162 <div class="doc_text">
164 The new LLVM <a href="http://compiler-rt.llvm.org/">compiler-rt project</a>
165 is a simple library that provides an implementation of the low-level
166 target-specific hooks required by code generation and other runtime components.
167 For example, when compiling for a 32-bit target, converting a double to a 64-bit
168 unsigned integer is compiled into a runtime call to the "__fixunsdfdi"
169 function. The compiler-rt library provides highly optimized implementations of
170 this and other low-level routines (some are 3x faster than the equivalent
171 libgcc routines).</p>
173 <p>In the LLVM 2.9 timeframe, compiler_rt has had several minor changes for
174 better ARM support, and a fairly major license change. All of the code in the
175 compiler-rt project is now <a href="DeveloperPolicy.html#license">dual
176 licensed</a> under MIT and UIUC license, which allows you to use compiler-rt
177 in applications without the binary copyright reproduction clause. If you
178 prefer the LLVM/UIUC license, you are free to continue using it under that
183 <!--=========================================================================-->
185 <a name="lldb">LLDB: Low Level Debugger</a>
188 <div class="doc_text">
190 <a href="http://lldb.llvm.org/">LLDB</a> is a brand new member of the LLVM
191 umbrella of projects. LLDB is a next generation, high-performance debugger. It
192 is built as a set of reusable components which highly leverage existing
193 libraries in the larger LLVM Project, such as the Clang expression parser, the
194 LLVM disassembler and the LLVM JIT.</p>
197 LLDB is has advanced by leaps and bounds in the 2.9 timeframe. It is
198 dramatically more stable and useful, and includes both a new <a
199 href="http://lldb.llvm.org/tutorial.html">tutorial</a> and a <a
200 href="http://lldb.llvm.org/lldb-gdb.html">side-by-side comparison with
205 <!--=========================================================================-->
207 <a name="libc++">libc++: C++ Standard Library</a>
210 <div class="doc_text">
212 <a href="http://libcxx.llvm.org/">libc++</a> is another new member of the LLVM
213 family. It is an implementation of the C++ standard library, written from the
214 ground up to specifically target the forthcoming C++'0X standard and focus on
215 delivering great performance.</p>
218 In the LLVM 2.9 timeframe, libc++ has had numerous bugs fixed, and is now being
219 co-developed with Clang's C++'0x mode.</p>
222 Like compiler_rt, libc++ is now <a href="DeveloperPolicy.html#license">dual
223 licensed</a> under the MIT and UIUC license, allowing it to be used more
231 <!--=========================================================================-->
234 <a name="klee">KLEE: A Symbolic Execution Virtual Machine</a>
237 <div class="doc_text">
239 <a href="http://klee.llvm.org/">KLEE</a> is a symbolic execution framework for
240 programs in LLVM bitcode form. KLEE tries to symbolically evaluate "all" paths
241 through the application and records state transitions that lead to fault
242 states. This allows it to construct testcases that lead to faults and can even
243 be used to verify some algorithms.
250 <!-- *********************************************************************** -->
252 <a name="externalproj">External Open Source Projects Using LLVM 2.9</a>
254 <!-- *********************************************************************** -->
256 <div class="doc_text">
258 <p>An exciting aspect of LLVM is that it is used as an enabling technology for
259 a lot of other language and tools projects. This section lists some of the
260 projects that have already been updated to work with LLVM 2.9.</p>
264 <!--=========================================================================-->
265 <h2>Crack Programming Language</h2>
267 <div class="doc_text">
269 <a href="http://code.google.com/p/crack-language/">Crack</a> aims to provide the
270 ease of development of a scripting language with the performance of a compiled
271 language. The language derives concepts from C++, Java and Python, incorporating
272 object-oriented programming, operator overloading and strong typing.</p>
276 <!--=========================================================================-->
277 <h2>TTA-based Codesign Environment (TCE)</h2>
279 <div class="doc_text">
280 <p>TCE is a toolset for designing application-specific processors (ASP) based on
281 the Transport triggered architecture (TTA). The toolset provides a complete
282 co-design flow from C/C++ programs down to synthesizable VHDL and parallel
283 program binaries. Processor customization points include the register files,
284 function units, supported operations, and the interconnection network.</p>
286 <p>TCE uses Clang and LLVM for C/C++ language support, target independent
287 optimizations and also for parts of code generation. It generates new LLVM-based
288 code generators "on the fly" for the designed TTA processors and loads them in
289 to the compiler backend as runtime libraries to avoid per-target recompilation
290 of larger parts of the compiler chain.</p>
295 <!--=========================================================================-->
298 <div class="doc_text">
299 <p><a href="http://gitorious.org/pinavm/pages/Home">PinaVM</a> is an open
300 source, <a href="http://www.systemc.org/">SystemC</a> front-end. Unlike many
301 other front-ends, PinaVM actually executes the elaboration of the
302 program analyzed using LLVM's JIT infrastructure. It later enriches the
303 bitcode with SystemC-specific information.</p>
306 <!--=========================================================================-->
309 <div class="doc_text">
310 <p><a href="http://pure-lang.googlecode.com/">Pure</a> is an
312 programming language based on term rewriting. Programs are collections
313 of equations which are used to evaluate expressions in a symbolic
314 fashion. The interpreter uses LLVM as a backend to JIT-compile Pure
315 programs to fast native code. Pure offers dynamic typing, eager and lazy
316 evaluation, lexical closures, a hygienic macro system (also based on
317 term rewriting), built-in list and matrix support (including list and
318 matrix comprehensions) and an easy-to-use interface to C and other
319 programming languages (including the ability to load LLVM bitcode
320 modules, and inline C, C++, Fortran and Faust code in Pure programs if
321 the corresponding LLVM-enabled compilers are installed).</p>
323 <p>Pure version 0.47 has been tested and is known to work with LLVM 2.9
324 (and continues to work with older LLVM releases >= 2.5).</p>
327 <!--=========================================================================-->
328 <h2 id="icedtea">IcedTea Java Virtual Machine Implementation</h2>
330 <div class="doc_text">
332 <a href="http://icedtea.classpath.org/wiki/Main_Page">IcedTea</a> provides a
333 harness to build OpenJDK using only free software build tools and to provide
334 replacements for the not-yet free parts of OpenJDK. One of the extensions that
335 IcedTea provides is a new JIT compiler named <a
336 href="http://icedtea.classpath.org/wiki/ZeroSharkFaq">Shark</a> which uses LLVM
337 to provide native code generation without introducing processor-dependent
341 <p> OpenJDK 7 b112, IcedTea6 1.9 and IcedTea7 1.13 and later have been tested
342 and are known to work with LLVM 2.9 (and continue to work with older LLVM
343 releases >= 2.6 as well).</p>
346 <!--=========================================================================-->
347 <h2>Glasgow Haskell Compiler (GHC)</h2>
349 <div class="doc_text">
350 <p>GHC is an open source, state-of-the-art programming suite for Haskell,
351 a standard lazy functional programming language. It includes an
352 optimizing static compiler generating good code for a variety of
353 platforms, together with an interactive system for convenient, quick
356 <p>In addition to the existing C and native code generators, GHC 7.0 now
357 supports an LLVM code generator. GHC supports LLVM 2.7 and later.</p>
360 <!--=========================================================================-->
361 <h2>Polly - Polyhedral optimizations for LLVM</h2>
363 <div class="doc_text">
364 <p>Polly is a project that aims to provide advanced memory access optimizations
365 to better take advantage of SIMD units, cache hierarchies, multiple cores or
366 even vector accelerators for LLVM. Built around an abstract mathematical
367 description based on Z-polyhedra, it provides the infrastructure to develop
368 advanced optimizations in LLVM and to connect complex external optimizers. In
369 its first year of existence Polly already provides an exact value-based
370 dependency analysis as well as basic SIMD and OpenMP code generation support.
371 Furthermore, Polly can use PoCC(Pluto) an advanced optimizer for data-locality
376 <!-- *********************************************************************** -->
378 <a name="whatsnew">What's New in LLVM 2.9?</a>
380 <!-- *********************************************************************** -->
382 <div class="doc_text">
384 <p>This release includes a huge number of bug fixes, performance tweaks and
385 minor improvements. Some of the major improvements and new features are listed
391 <!--=========================================================================-->
393 <a name="majorfeatures">Major New Features</a>
396 <div class="doc_text">
398 <p>LLVM 2.9 includes several major new capabilities:</p>
403 TBAA: On by default in clang. Disable it with -fno-strict-aliasing.
404 Could be more aggressive for structs.
407 <li>New Nvidia PTX backend, not generally useful in 2.9 though.</li>
410 Much better debug info generated, particularly in optimized code situations.
414 inline asm multiple alternative constraint support.
418 New naming rules in coding standards: CodingStandards.html#ll_naming
425 <!--=========================================================================-->
427 <a name="coreimprovements">LLVM IR and Core Improvements</a>
430 <div class="doc_text">
431 <p>LLVM IR has several new features for better support of new targets and that
432 expose new optimization opportunities:</p>
435 <li>udiv, ashr, lshr, shl now have exact and nuw/nsw bits:
436 PR8862 / LangRef.html</li>
438 unnamed_addr + PR8927
440 new 'hotpatch' attribute: LangRef.html#fnattrs
446 <!--=========================================================================-->
448 <a name="optimizer">Optimizer Improvements</a>
451 <div class="doc_text">
453 <p>In addition to a large array of minor performance tweaks and bug fixes, this
454 release includes a few major enhancements and additions to the optimizers:</p>
457 <li>LTO has been improved to use MC for parsing inline asm and now
458 can build large programs like Firefox 4 on both OS X and Linux.</li>
461 LoopIdiom: memset/memcpy formation and memset_pattern on darwin. Build with
462 -ffreestanding or -fno-builtin if your memcpy is being compiled into infinite
468 LoopInstSimplify pass.
470 New <a href="WritingAnLLVMPass.html#RegionPass">RegionPass</a> infrastructure
471 for region-based optimizations.
473 Can optimize printf to iprintf when no floating point is used, for embedded
474 targets with smaller iprintf implementation.
476 Speedups to various mid-level passes:
477 GVN is much faster on functions with deep dominator trees / lots of BBs.
478 DomTree and DominatorFrontier are much faster to compute, and preserved by
479 more passes (so they are computed less often)
480 SRoA is also much faster and doesn't use DominanceFrontier.
482 DSE is more aggressive with stores of different types: e.g. a large store
483 following a small one to the same address.
486 We now optimize various idioms for overflow detection into check of the flag
487 register on various CPUs, e.g.:
488 unsigned long t = a+b;
498 <p>In addition to these features that are done in 2.8, there is preliminary
499 support in the release for Type Based Alias Analysis
500 Preliminary work on TBAA but not usable in 2.8.
501 New CorrelatedValuePropagation pass, not on by default in 2.8 yet.
506 <!--=========================================================================-->
508 <a name="mc">MC Level Improvements</a>
511 <div class="doc_text">
513 The LLVM Machine Code (aka MC) subsystem was created to solve a number
514 of problems in the realm of assembly, disassembly, object file format handling,
515 and a number of other related areas that CPU instruction-set level tools work
519 <li>MC is now used by default for ELF systems on x86 and
521 <li>MC supports and CodeGen uses the <tt>.loc</tt> directives for
522 producing line number debug info. This produces more compact line
524 <li>MC supports the <tt>.cfi_*</tt> directives for producing DWARF
525 frame information, but it is still not used by CodeGen by default.</li>
526 <li>COFF support?</li>
529 MC Assembler: X86 now generates much better diagnostics for common errors,
530 is much faster at matching instructions, is much more bug-compatible with
531 the GAS assembler, and is now generally useful for a broad range of X86
535 ELF MC support: on by default in clang. There are still known missing features
536 for human written assembly.
539 Some basic <a href="CodeGenerator.html#mc">internals documentation</a> for MC.
541 MC Assembler support for .file and .loc.
543 tblgen support for assembler aliases: <a
544 href="CodeGenerator.html#na_instparsing">MnemonicAlias and InstAlias</a>
546 Win32 PE-COFF support in the MC assembler has made a lot of progress in the 2.9
547 timeframe, but is still not generally useful. Please see
548 "http://llvm.org/bugs/showdependencytree.cgi?id=9100&hide_resolved=1" for open bugs?
551 lib/Object and llvm-objdump
552 Experimental format independent object file manipulation library.
553 * Supports PE/COFF and ELF.
554 * llvm-nm extended to work with object files. Exactly matches
555 binutils-nm for the files I've tested.
556 * llvm-objdump added with support for disassembly (no relocations displayed).
562 <p>For more information, please see the <a
563 href="http://blog.llvm.org/2010/04/intro-to-llvm-mc-project.html">Intro to the
564 LLVM MC Project Blog Post</a>.
569 <!--=========================================================================-->
571 <a name="codegen">Target Independent Code Generator Improvements</a>
574 <div class="doc_text">
576 <p>We have put a significant amount of work into the code generator
577 infrastructure, which allows us to implement more aggressive algorithms and make
583 <li>The pre-register-allocation (preRA) instruction scheduler models register
584 pressure much more accurately in some cases. This allows the adoption of more
585 aggressive scheduling heuristics.
588 LiveDebugVariables is a new pass that keeps track of debugging information for
589 user variables that are kept in registers in optimized builds.
592 Scheduler now models operand latency and pipeline forwarding.
594 Major regalloc rewrite, not on by default for 2.9 and not advised to use it.
595 * New basic register allocator that can be used as a safe fallback when
596 debugging. Enable with -regalloc=basic.
597 * New infrastructure for live range splitting. SplitKit can break a live
598 interval into smaller pieces while preserving SSA form, and SpillPlacement
599 can help find the best split points. This is a work in progress so the API
601 * The inline spiller has learned to clean up after live range splitting. It
602 can hoist spills out of loops, and it can eliminate redundant spills.
603 Rematerialization works with live range splitting.
604 * New greedy register allocator using live range splitting. This will be the
605 default register allocator in the next LLVM release, but it is not turned on
613 <!--=========================================================================-->
615 <a name="x86">X86-32 and X86-64 Target Improvements</a>
618 <div class="doc_text">
619 <p>New features and major changes in the X86 target include:
624 X86: Reimplemented all of MMX to introduce a new LLVM IR x86_mmx type. Now
625 random types like <2 x i32> are not iseld to mmx without emms. The
626 -disable-mmx flag is gone now.
630 X86 support for FS/GS relative loads and stores using address space 256/257 are
635 X86: Much better codegen for several cases using adc/sbb instead of cmovs for
636 conditional increment and other idioms.
640 The X86 backend has adopted a new preRA scheduling
641 mode, "list-ilp", to shorten the height of instruction schedules
642 without inducing register spills.
645 MC assembler support for 3dNow! and 3DNowA instructions.
647 <li>Several bugs have been fixed for Windows x64 code generator.</li>
652 <!--=========================================================================-->
654 <a name="ARM">ARM Target Improvements</a>
657 <div class="doc_text">
658 <p>New features of the ARM target include:
662 <li>ARM Fast ISel</li>
663 <li>ARM: New code placement pass.</li>
664 <li>ARM: Improved code generation for Cortex-A8 and Cortex-A9 CPUs.</li>
665 <li>ARM: __builtin_prefetch turns into prefetch instructions.</li>
666 <li>Countless ARM microoptimizations.</li>
668 <li> The ARM backend preRA scheduler now models machine resources at cycle
669 granularity. This allows the scheduler to both accurately model
670 instruction latency and avoid overcommitting functional units.</li>
676 <!--=========================================================================-->
678 <a name="OtherTS">Other Target Specific Improvements</a>
681 <div class="doc_text">
683 PPC: Switched to MCInstPrinter, and MCCodeEmitter. Ready to implement support
684 for directly writing out mach-o object files, but noone seems interested.
686 MicroBlaze: major updates for aggressive delay slot filler, MC-based assembly
687 printing, assembly instruction parsing, ELF .o file emission, and MC
688 instruction disassembler.
690 SPARC: Many improvements, including using the Y registers for multiplications
691 and addition of a simple delay slot filler.
696 <!--=========================================================================-->
698 <a name="changes">Major Changes and Removed Features</a>
701 <div class="doc_text">
703 <p>If you're already an LLVM user or developer with out-of-tree changes based
704 on LLVM 2.8, this section lists some "gotchas" that you may run into upgrading
705 from the previous release.</p>
708 last release for llvm-gcc
710 - DIBuilder provides simpler interface for front ends like Clang to encode debug info in LLVM IR.
711 - This interface hides implementation details (e.g. DIDerivedType, existence of compile unit etc..) that any front end should not know about.
712 For example, DIFactory DebugFactory;
713 Ty = DebugFactory.CreateDerivedType(DW_TAG_volatile_type,
714 findRegion(TYPE_CONTEXT(type)),
716 getOrCreateFile(main_input_filename),
718 NodeSizeInBits(type),
719 NodeAlignInBits(type),
724 DbgTy = DBuilder.createQualifiedType(DW_TAG_volatile_type, MainTy);
725 DIFactory is gone now.
732 LoopIndexSplit pass was removed, unmaintained.
733 LiveValues, SimplifyHalfPowrLibCalls, and GEPSplitter were removed.
734 Removed the PartialSpecialization pass, it was unmaintained and buggy.
736 DIFactory removed, use DIBuilder instead.
738 Triple::normalize is new, llvm triples are always stored in normalized form internally.
740 Triple x86_64--mingw64 is obsoleted. Use x86_64--mingw32 instead.
742 PointerTracking has been removed from mainline, moved to ClamAV.
748 <!--=========================================================================-->
750 <a name="api_changes">Internal API Changes</a>
753 <div class="doc_text">
755 <p>In addition, many APIs have changed in this release. Some of the major
756 LLVM API changes are:</p>
760 include/llvm/System merged into include/llvm/Support.
763 APInt API changes, see PR5207.
765 MVT::Flag renamed to MVT::Glue
768 error_code + libsystem + PathV2 changes
769 The system_error header from C++0x was added.
770 * Use if (error_code ec = function()) to check for error conditions
771 from functions which return it.
772 * error_code::message returns a human readable description of the error.
774 PathV1 has been deprecated in favor of PathV2 (sorry I didn't finish
775 this before the release).
776 * No Path class, use a r-value convertible to a twine instead.
777 * Assumes all paths are UTF-8.
783 <!-- *********************************************************************** -->
785 <a name="knownproblems">Known Problems</a>
787 <!-- *********************************************************************** -->
789 <div class="doc_text">
791 <p>This section contains significant known problems with the LLVM system,
792 listed by component. If you run into a problem, please check the <a
793 href="http://llvm.org/bugs/">LLVM bug database</a> and submit a bug if
794 there isn't already one.</p>
798 <!-- ======================================================================= -->
800 <a name="experimental">Experimental features included with this release</a>
803 <div class="doc_text">
805 <p>The following components of this LLVM release are either untested, known to
806 be broken or unreliable, or are in early development. These components should
807 not be relied on, and bugs should not be filed against them, but they may be
808 useful to some people. In particular, if you would like to work on one of these
809 components, please contact us on the <a
810 href="http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev">LLVMdev list</a>.</p>
813 <li>The Alpha, Blackfin, CellSPU, MicroBlaze, MSP430, MIPS, PTX, SystemZ
814 and XCore backends are experimental.</li>
815 <li><tt>llc</tt> "<tt>-filetype=obj</tt>" is experimental on all targets
816 other than darwin-i386 and darwin-x86_64. FIXME: Not true on ELF anymore?</li>
822 <!-- ======================================================================= -->
824 <a name="x86-be">Known problems with the X86 back-end</a>
827 <div class="doc_text">
830 <li>The X86 backend does not yet support
831 all <a href="http://llvm.org/PR879">inline assembly that uses the X86
832 floating point stack</a>. It supports the 'f' and 't' constraints, but not
834 <li>The X86-64 backend does not yet support the LLVM IR instruction
835 <tt>va_arg</tt>. Currently, front-ends support variadic
836 argument constructs on X86-64 by lowering them manually.</li>
837 <li>Windows x64 (aka Win64) code generator has a few issues.
839 <li>llvm-gcc cannot build the mingw-w64 runtime currently
840 due to lack of support for the 'u' inline assembly
841 constraint and for X87 floating point inline assembly.</li>
842 <li>On mingw-w64, you will see unresolved symbol <tt>__chkstk</tt>
843 due to <a href="http://llvm.org/bugs/show_bug.cgi?id=8919">Bug 8919</a>.
844 It is fixed in <a href="http://lists.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20110321/118499.html">r128206</a>.</li>
845 <li>Miss-aligned MOVDQA might crash your program. It is due to
846 <a href="http://llvm.org/bugs/show_bug.cgi?id=9483">Bug 9483</a>,
847 lack of handling aligned internal globals.</li>
855 <!-- ======================================================================= -->
857 <a name="ppc-be">Known problems with the PowerPC back-end</a>
860 <div class="doc_text">
863 <li>The Linux PPC32/ABI support needs testing for the interpreter and static
864 compilation, and lacks support for debug information.</li>
869 <!-- ======================================================================= -->
871 <a name="arm-be">Known problems with the ARM back-end</a>
874 <div class="doc_text">
877 <li>Thumb mode works only on ARMv6 or higher processors. On sub-ARMv6
878 processors, thumb programs can crash or produce wrong
879 results (<a href="http://llvm.org/PR1388">PR1388</a>).</li>
880 <li>Compilation for ARM Linux OABI (old ABI) is supported but not fully tested.
886 <!-- ======================================================================= -->
888 <a name="sparc-be">Known problems with the SPARC back-end</a>
891 <div class="doc_text">
894 <li>The SPARC backend only supports the 32-bit SPARC ABI (-m32); it does not
895 support the 64-bit SPARC ABI (-m64).</li>
900 <!-- ======================================================================= -->
902 <a name="mips-be">Known problems with the MIPS back-end</a>
905 <div class="doc_text">
908 <li>64-bit MIPS targets are not supported yet.</li>
913 <!-- ======================================================================= -->
915 <a name="alpha-be">Known problems with the Alpha back-end</a>
918 <div class="doc_text">
922 <li>On 21164s, some rare FP arithmetic sequences which may trap do not have the
923 appropriate nops inserted to ensure restartability.</li>
928 <!-- ======================================================================= -->
930 <a name="c-be">Known problems with the C back-end</a>
933 <div class="doc_text">
935 <p>The C backend has numerous problems and is not being actively maintained.
936 Depending on it for anything serious is not advised.</p>
939 <li><a href="http://llvm.org/PR802">The C backend has only basic support for
940 inline assembly code</a>.</li>
941 <li><a href="http://llvm.org/PR1658">The C backend violates the ABI of common
942 C++ programs</a>, preventing intermixing between C++ compiled by the CBE and
943 C++ code compiled with <tt>llc</tt> or native compilers.</li>
944 <li>The C backend does not support all exception handling constructs.</li>
945 <li>The C backend does not support arbitrary precision integers.</li>
951 <!-- ======================================================================= -->
953 <a name="llvm-gcc">Known problems with the llvm-gcc front-end</a>
956 <div class="doc_text">
958 <p>llvm-gcc is generally very stable for the C family of languages. The only
959 major language feature of GCC not supported by llvm-gcc is the
960 <tt>__builtin_apply</tt> family of builtins. However, some extensions
961 are only supported on some targets. For example, trampolines are only
962 supported on some targets (these are used when you take the address of a
963 nested function).</p>
965 <p>Fortran support generally works, but there are still several unresolved bugs
966 in <a href="http://llvm.org/bugs/">Bugzilla</a>. Please see the
967 tools/gfortran component for details. Note that llvm-gcc is missing major
968 Fortran performance work in the frontend and library that went into GCC after
969 4.2. If you are interested in Fortran, we recommend that you consider using
970 <a href="#dragonegg">dragonegg</a> instead.</p>
972 <p>The llvm-gcc 4.2 Ada compiler has basic functionality, but is no longer being
973 actively maintained. If you are interested in Ada, we recommend that you
974 consider using <a href="#dragonegg">dragonegg</a> instead.</p>
977 <!-- *********************************************************************** -->
979 <a name="additionalinfo">Additional Information</a>
981 <!-- *********************************************************************** -->
983 <div class="doc_text">
985 <p>A wide variety of additional information is available on the <a
986 href="http://llvm.org">LLVM web page</a>, in particular in the <a
987 href="http://llvm.org/docs/">documentation</a> section. The web page also
988 contains versions of the API documentation which is up-to-date with the
989 Subversion version of the source code.
990 You can access versions of these documents specific to this release by going
991 into the "<tt>llvm/doc/</tt>" directory in the LLVM tree.</p>
993 <p>If you have any questions or comments about LLVM, please feel free to contact
994 us via the <a href="http://llvm.org/docs/#maillist"> mailing
999 <!-- *********************************************************************** -->
1003 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
1004 src="http://jigsaw.w3.org/css-validator/images/vcss-blue" alt="Valid CSS"></a>
1005 <a href="http://validator.w3.org/check/referer"><img
1006 src="http://www.w3.org/Icons/valid-html401-blue" alt="Valid HTML 4.01"></a>
1008 <a href="http://llvm.org/">LLVM Compiler Infrastructure</a><br>
1009 Last modified: $Date$