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 2.8 Release Notes</title>
11 <div class="doc_title">LLVM 2.8 Release Notes</div>
13 <img align=right src="http://llvm.org/img/DragonSmall.png"
14 width="136" height="136" alt="LLVM Dragon Logo">
17 <li><a href="#intro">Introduction</a></li>
18 <li><a href="#subproj">Sub-project Status Update</a></li>
19 <li><a href="#externalproj">External Projects Using LLVM 2.8</a></li>
20 <li><a href="#whatsnew">What's New in LLVM 2.8?</a></li>
21 <li><a href="GettingStarted.html">Installation Instructions</a></li>
22 <li><a href="#knownproblems">Known Problems</a></li>
23 <li><a href="#additionalinfo">Additional Information</a></li>
26 <div class="doc_author">
27 <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 2.8
34 <a href="http://llvm.org/releases/2.7/docs/ReleaseNotes.html">LLVM 2.7
35 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 2.8. 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.
75 <!-- Features that need text if they're finished for 2.9:
78 loop dependence analysis
80 CorrelatedValuePropagation
83 <!-- Announcement, lldb, libc++ -->
86 MachineCSE tuned and on by default.
87 llvm.dbg.value: variable debug info for optimized code
88 MC Assembler backend is now real, does relaxation and is bitwise identical
89 with darwin assembler in huge majority of all cases.
90 new GHC calling convention
91 New half float intrinsics LangRef.html#int_fp16
92 Rewrote tblgen's type inference for backends to be more consistent and
93 diagnose more target bugs. This also allows limited support for writing
94 patterns for instructions that return multiple results, e.g. a virtual
95 register and a flag result. Stuff that used 'parallel' before should use
97 New ARM/Thumb disassembler support in MC.
98 New SSEDomainFix pass:
99 On Nehalem and newer CPUs there is a 2 cycle latency penalty on using a
100 register in a different domain than where it was defined. Some instructions
101 have equvivalents for different domains, like por/orps/orpd. The
102 SSEDomainFix pass tries to minimize the number of domain crossings by
103 changing between equvivalent opcodes where possible.
104 Support for the Intel AES instructions in the assembler.
105 memcpy, memmove, and memset now take address space qualified pointers + volatile.
106 per-instruction debug info metadata is much faster and uses less space (new DebugLoc class).
107 -ffunction-sections and -fdata-sections are supported on ELF targets.
108 Now iterate function passes when a cgsccpassmanager detects a devirtualization
109 -momit-leaf-frame-pointer now supported.
110 New -regalloc=fast, =local got removed
111 New -regalloc=default option that chooses a register allocator based on the -O optimization level.
112 New "trap values" concept: http://llvm.org/docs/LangRef.html#trapvalues
113 Improved trip count analysis for <= and >= loops, and uses sign overflow info.
115 X86 backend attempts to promote 16-bit integer operations to 32-bits to avoid
116 0x66 prefixes, which are slow on some microarchitectures and bloat the code
118 X87 fp stackifier is global!
119 LTO debug info support?
120 NEON: Better performance for QQQQ (4-consecutive Q register) instructions. New reg sequence abstraction?
121 New support for X86 "thiscall" calling convention (x86_thiscallcc in IR).
122 ARM: Better scheduling (list-hybrid, hybrid?)
123 New SubRegIndex tblgen class for targets -> jakob
124 ARM: Tail call support.
125 AVX support in the MC assembler. Full compiler support not done yet.
126 Atomics now get legalized when not natively supported (jim g)
127 ARM: General performance work and tuning.
131 <!-- *********************************************************************** -->
132 <div class="doc_section">
133 <a name="subproj">Sub-project Status Update</a>
135 <!-- *********************************************************************** -->
137 <div class="doc_text">
139 The LLVM 2.8 distribution currently consists of code from the core LLVM
140 repository (which roughly includes the LLVM optimizers, code generators
141 and supporting tools), the Clang repository and the llvm-gcc repository. In
142 addition to this code, the LLVM Project includes other sub-projects that are in
143 development. Here we include updates on these subprojects.
149 <!--=========================================================================-->
150 <div class="doc_subsection">
151 <a name="clang">Clang: C/C++/Objective-C Frontend Toolkit</a>
154 <div class="doc_text">
156 <p><a href="http://clang.llvm.org/">Clang</a> is an LLVM front end for the C,
157 C++, and Objective-C languages. Clang aims to provide a better user experience
158 through expressive diagnostics, a high level of conformance to language
159 standards, fast compilation, and low memory use. Like LLVM, Clang provides a
160 modular, library-based architecture that makes it suitable for creating or
161 integrating with other development tools. Clang is considered a
162 production-quality compiler for C, Objective-C, C++ and Objective-C++ on x86
163 (32- and 64-bit), and for darwin-arm targets.</p>
165 <p>In the LLVM 2.8 time-frame, the Clang team has made many improvements:</p>
168 <li>Surely these guys have done something</li>
169 <li>X86-64 abi improvements? Did they make it in?</li>
173 <!--=========================================================================-->
174 <div class="doc_subsection">
175 <a name="clangsa">Clang Static Analyzer</a>
178 <div class="doc_text">
180 <p>The <a href="http://clang-analyzer.llvm.org/">Clang Static Analyzer</a>
181 project is an effort to use static source code analysis techniques to
182 automatically find bugs in C and Objective-C programs (and hopefully <a
183 href="http://clang-analyzer.llvm.org/dev_cxx.html">C++ in the
184 future</a>!). The tool is very good at finding bugs that occur on specific
185 paths through code, such as on error conditions.</p>
187 <p>The LLVM 2.8 release fixes a number of bugs and slightly improves precision
188 over 2.7, but there are no major new features in the release.
193 <!--=========================================================================-->
194 <div class="doc_subsection">
195 <a name="vmkit">VMKit: JVM/CLI Virtual Machine Implementation</a>
198 <div class="doc_text">
200 The <a href="http://vmkit.llvm.org/">VMKit project</a> is an implementation of
201 a JVM and a CLI Virtual Machine (Microsoft .NET is an
202 implementation of the CLI) using LLVM for static and just-in-time
205 <p>With the release of LLVM 2.8, ...</p>
210 <!--=========================================================================-->
211 <div class="doc_subsection">
212 <a name="compiler-rt">compiler-rt: Compiler Runtime Library</a>
215 <div class="doc_text">
217 The new LLVM <a href="http://compiler-rt.llvm.org/">compiler-rt project</a>
218 is a simple library that provides an implementation of the low-level
219 target-specific hooks required by code generation and other runtime components.
220 For example, when compiling for a 32-bit target, converting a double to a 64-bit
221 unsigned integer is compiled into a runtime call to the "__fixunsdfdi"
222 function. The compiler-rt library provides highly optimized implementations of
223 this and other low-level routines (some are 3x faster than the equivalent
224 libgcc routines).</p>
227 All of the code in the compiler-rt project is available under the standard LLVM
228 License, a "BSD-style" license. New in LLVM 2.8:
235 <!--=========================================================================-->
236 <div class="doc_subsection">
237 <a name="dragonegg">DragonEgg: llvm-gcc ported to gcc-4.5</a>
240 <div class="doc_text">
242 <a href="http://dragonegg.llvm.org/">DragonEgg</a> is a port of llvm-gcc to
243 gcc-4.5. Unlike llvm-gcc, dragonegg in theory does not require any gcc-4.5
244 modifications whatsoever (currently one small patch is needed) thanks to the
245 new <a href="http://gcc.gnu.org/wiki/plugins">gcc plugin architecture</a>.
246 DragonEgg is a gcc plugin that makes gcc-4.5 use the LLVM optimizers and code
247 generators instead of gcc's, just like with llvm-gcc.
251 DragonEgg is still a work in progress, but it is able to compile a lot of code,
252 for example all of gcc, LLVM and clang. Currently Ada, C, C++ and Fortran work
253 well, while all other languages either don't work at all or only work poorly.
254 For the moment only the x86-32 and x86-64 targets are supported, and only on
255 linux and darwin (darwin may need additional gcc patches).
259 The 2.8 release has the following notable changes:
261 <li>The plugin loads faster due to exporting fewer symbols.</li>
262 <li>Additional vector operations such as addps256 are now supported.</li>
263 <li>Ada global variables with no initial value are no longer zero initialized,
264 resulting in better optimization.</li>
265 <li>The '-fplugin-arg-dragonegg-enable-gcc-optzns' flag now runs all gcc
266 optimizers, rather than just a handful.</li>
267 <li>Fortran programs using common variables now link correctly.</li>
268 <li>GNU OMP constructs no longer crash the compiler.</li>
275 <!--=========================================================================-->
276 <div class="doc_subsection">
277 <a name="lldb">LLDB: Low Level Debugger</a>
280 <div class="doc_text">
282 <a href="http://lldb.llvm.org/">LLDB</a> is</p>
293 <!--=========================================================================-->
294 <div class="doc_subsection">
295 <a name="libc++">libc++: C++ Standard Library</a>
298 <div class="doc_text">
300 <a href="http://libc++.llvm.org/">libc++</a> is</p>
312 <!-- *********************************************************************** -->
313 <div class="doc_section">
314 <a name="externalproj">External Open Source Projects Using LLVM 2.8</a>
316 <!-- *********************************************************************** -->
318 <div class="doc_text">
320 <p>An exciting aspect of LLVM is that it is used as an enabling technology for
321 a lot of other language and tools projects. This section lists some of the
322 projects that have already been updated to work with LLVM 2.8.</p>
325 <!--=========================================================================-->
326 <div class="doc_subsection">
327 <a name="tce">TTA-based Codesign Environment (TCE)</a>
330 <div class="doc_text">
332 <a href="http://tce.cs.tut.fi/">TCE</a> is a toolset for designing
333 application-specific processors (ASP) based on the Transport triggered
334 architecture (TTA). The toolset provides a complete co-design flow from C/C++
335 programs down to synthesizable VHDL and parallel program binaries. Processor
336 customization points include the register files, function units, supported
337 operations, and the interconnection network.</p>
339 <p>TCE uses llvm-gcc/Clang and LLVM for C/C++ language support, target
340 independent optimizations and also for parts of code generation. It generates
341 new LLVM-based code generators "on the fly" for the designed TTA processors and
342 loads them in to the compiler backend as runtime libraries to avoid per-target
343 recompilation of larger parts of the compiler chain.</p>
347 <!--=========================================================================-->
348 <div class="doc_subsection">
349 <a name="Horizon">Horizon Bytecode Compiler</a>
352 <div class="doc_text">
354 <a href="http://www.quokforge.org/projects/horizon">Horizon</a> is a bytecode
355 language and compiler written on top of LLVM, intended for producing
356 single-address-space managed code operating systems that
357 run faster than the equivalent multiple-address-space C systems.
358 More in-depth blurb is available on <a
359 href="http://www.quokforge.org/projects/horizon/wiki/Wiki">the wiki</a>.</p>
363 <!--=========================================================================-->
364 <div class="doc_subsection">
365 <a name="clamav">Clam AntiVirus</a>
368 <div class="doc_text">
370 <a href=http://www.clamav.net>Clam AntiVirus</a> is an open source (GPL)
371 anti-virus toolkit for UNIX, designed especially for e-mail scanning on mail
372 gateways. Since version 0.96 it has <a
373 href="http://vrt-sourcefire.blogspot.com/2010/09/introduction-to-clamavs-low-level.html">bytecode
374 signatures</a> that allow writing detections for complex malware. It
375 uses LLVM's JIT to speed up the execution of bytecode on
376 X86,X86-64,PPC32/64, falling back to its own interpreter otherwise.
377 The git version was updated to work with LLVM 2.8
381 href="http://git.clamav.net/gitweb?p=clamav-bytecode-compiler.git;a=blob_plain;f=docs/user/clambc-user.pdf">
382 ClamAV bytecode compiler</a> uses Clang and LLVM to compile a C-like
383 language, insert runtime checks, and generate ClamAV bytecode.</p>
387 <!--=========================================================================-->
388 <div class="doc_subsection">
389 <a name="pure">Pure</a>
392 <div class="doc_text">
394 <a href="http://pure-lang.googlecode.com/">Pure</a>
395 is an algebraic/functional
396 programming language based on term rewriting. Programs are collections
397 of equations which are used to evaluate expressions in a symbolic
398 fashion. Pure offers dynamic typing, eager and lazy evaluation, lexical
399 closures, a hygienic macro system (also based on term rewriting),
400 built-in list and matrix support (including list and matrix
401 comprehensions) and an easy-to-use C interface. The interpreter uses
402 LLVM as a backend to JIT-compile Pure programs to fast native code.</p>
404 <p>Pure versions 0.44 and later have been tested and are known to work with
405 LLVM 2.8 (and continue to work with older LLVM releases >= 2.5).</p>
409 <!--=========================================================================-->
410 <div class="doc_subsection">
411 <a name="GHC">Glasgow Haskell Compiler (GHC)</a>
414 <div class="doc_text">
416 <a href="http://www.haskell.org/ghc/">GHC</a> is an open source,
417 state-of-the-art programming suite for
418 Haskell, a standard lazy functional programming language. It includes
419 an optimizing static compiler generating good code for a variety of
420 platforms, together with an interactive system for convenient, quick
423 <p>In addition to the existing C and native code generators, GHC 7.0 now
425 href="http://hackage.haskell.org/trac/ghc/wiki/Commentary/Compiler/Backends/LLVM">LLVM
426 code generator</a>. GHC supports LLVM 2.7 and later.</p>
430 <!--=========================================================================-->
431 <div class="doc_subsection">
432 <a name="Clay">Clay Programming Language</a>
435 <div class="doc_text">
437 <a href="http://tachyon.in/clay/">Clay</a> is a new systems programming
438 language that is specifically designed for generic programming. It makes
439 generic programming very concise thanks to whole program type propagation. It
440 uses LLVM as its backend.</p>
444 <!-- *********************************************************************** -->
445 <div class="doc_section">
446 <a name="whatsnew">What's New in LLVM 2.8?</a>
448 <!-- *********************************************************************** -->
450 <div class="doc_text">
452 <p>This release includes a huge number of bug fixes, performance tweaks and
453 minor improvements. Some of the major improvements and new features are listed
459 <!--=========================================================================-->
460 <div class="doc_subsection">
461 <a name="orgchanges">LLVM Community Changes</a>
464 <div class="doc_text">
466 <p>In addition to changes to the code, between LLVM 2.7 and 2.8, a number of
467 organization changes have happened:
471 <li>libc++ and lldb are new</li>
472 <li>Debugging optimized code support.</li>
476 <!--=========================================================================-->
477 <div class="doc_subsection">
478 <a name="majorfeatures">Major New Features</a>
481 <div class="doc_text">
483 <p>LLVM 2.8 includes several major new capabilities:</p>
486 <li>atomic lowering pass.</li>
487 <li>RegionInfo pass: opt -regions analyze" or "opt -view-regions".
488 <!-- Tobias Grosser --></li>
489 <li>ARMGlobalMerge: <!-- Anton --> </li>
495 <!--=========================================================================-->
496 <div class="doc_subsection">
497 <a name="coreimprovements">LLVM IR and Core Improvements</a>
500 <div class="doc_text">
501 <p>LLVM IR has several new features for better support of new targets and that
502 expose new optimization opportunities:</p>
506 <li>LLVM 2.8 changes the internal order of operands in <a
507 href="http://llvm.org/doxygen/classllvm_1_1InvokeInst.html"><tt>InvokeInst</tt></a>
508 and <a href="http://llvm.org/doxygen/classllvm_1_1CallInst.html"><tt>CallInst</tt></a>.
509 To be portable across releases, resort to <tt>CallSite</tt> and the
510 high-level accessors, such as <tt>getCalledValue</tt> and <tt>setUnwindDest</tt>.
513 You can no longer pass use_iterators directly to cast<> (and similar), because
514 these routines tend to perform costly dereference operations more than once. You
515 have to dereference the iterators yourself and pass them in.
518 llvm.memcpy.*, llvm.memset.*, llvm.memmove.* (and possibly other?) intrinsics
519 take an extra parameter now (i1 isVolatile), totaling 5 parameters.
520 If you were creating these intrinsic calls and prototypes yourself (as opposed
521 to using Intrinsic::getDeclaration), you can use UpgradeIntrinsicFunction/UpgradeIntrinsicCall
522 to be portable accross releases.
523 Note that you cannot use Intrinsic::getDeclaration() in a backwards compatible
524 way (needs 2/3 types now, in 2.7 it needed just 1).
527 SetCurrentDebugLocation takes a DebugLoc now instead of a MDNode.
528 Change your code to use
529 SetCurrentDebugLocation(DebugLoc::getFromDILocation(...)).
532 VISIBILITY_HIDDEN is gone.
535 The <tt>RegisterPass</tt> and <tt>RegisterAnalysisGroup</tt> templates are
536 considered deprecated, but continue to function in LLVM 2.8. Clients are
537 strongly advised to use the upcoming <tt>INITIALIZE_PASS()</tt> and
538 <tt>INITIALIZE_AG_PASS()</tt> macros instead.
540 SMDiagnostic takes different parameters now. //FIXME: how to upgrade?
543 The constructor for the Triple class no longer tries to understand odd triple
544 specifications. Frontends should ensure that they only pass valid triples to
545 LLVM. The Triple::normalize utility method has been added to help front-ends
546 deal with funky triples.
548 Some APIs got renamed:
550 <li>llvm_report_error -> report_fatal_error</li>
551 <li>llvm_install_error_handler -> install_fatal_error_handler</li>
552 <li>llvm::DwarfExceptionHandling -> llvm::JITExceptionHandling</li>
559 <!--=========================================================================-->
560 <div class="doc_subsection">
561 <a name="optimizer">Optimizer Improvements</a>
564 <div class="doc_text">
566 <p>In addition to a large array of minor performance tweaks and bug fixes, this
567 release includes a few major enhancements and additions to the optimizers:</p>
578 <!--=========================================================================-->
579 <div class="doc_subsection">
580 <a name="executionengine">Interpreter and JIT Improvements</a>
583 <div class="doc_text">
592 <!--=========================================================================-->
593 <div class="doc_subsection">
594 <a name="mc">MC Level Improvements</a>
597 <div class="doc_text">
601 The LLVM Machine Code (aka MC) sub-project of LLVM was created to solve a number
602 of problems in the realm of assembly, disassembly, object file format handling,
603 and a number of other related areas that CPU instruction-set level tools work
604 in. It is a sub-project of LLVM which provides it with a number of advantages
605 over other compilers that do not have tightly integrated assembly-level tools.
606 For a gentle introduction, please see the <a
607 href="http://blog.llvm.org/2010/04/intro-to-llvm-mc-project.html">Intro to the
608 LLVM MC Project Blog Post</a>.
611 <p>2.8 status here. Basic correctness, some obscure missing instructions on
612 mainline, on by default in clang.
613 Entire compiler backend converted to use mcstreamer.
619 <!--=========================================================================-->
620 <div class="doc_subsection">
621 <a name="codegen">Target Independent Code Generator Improvements</a>
624 <div class="doc_text">
626 <p>We have put a significant amount of work into the code generator
627 infrastructure, which allows us to implement more aggressive algorithms and make
631 <li>MachO writer works.</li>
635 <!--=========================================================================-->
636 <div class="doc_subsection">
637 <a name="x86">X86-32 and X86-64 Target Improvements</a>
640 <div class="doc_text">
641 <p>New features of the X86 target include:
645 <li>The X86 backend now supports holding X87 floating point stack values
646 in registers across basic blocks, dramatically improving performance of code
647 that uses long double, and when targetting CPUs that don't support SSE.</li>
653 <!--=========================================================================-->
654 <div class="doc_subsection">
655 <a name="ARM">ARM Target Improvements</a>
658 <div class="doc_text">
659 <p>New features of the ARM target include:
665 All of the NEON load and store intrinsics (llvm.arm.neon.vld* and
666 llvm.arm.neon.vst*) take an extra parameter to specify the alignment in bytes
667 of the memory being accessed.
670 The llvm.arm.neon.vaba intrinsic (vector absolute difference and
671 accumulate) has been removed. This operation is now represented using
672 the llvm.arm.neon.vabd intrinsic (vector absolute difference) followed by a
676 The llvm.arm.neon.vabdl and llvm.arm.neon.vabal intrinsics (lengthening
677 vector absolute difference with and without accumlation) have been removed.
678 They are represented using the llvm.arm.neon.vabd intrinsic (vector absolute
679 difference) followed by a vector zero-extend operation, and for vabal,
683 The llvm.arm.neon.vmovn intrinsic has been removed. Calls of this intrinsic
684 are now replaced by vector truncate operations.
687 The llvm.arm.neon.vmovls and llvm.arm.neon.vmovlu intrinsics have been
688 removed. They are now represented as vector sign-extend (vmovls) and
689 zero-extend (vmovlu) operations.
692 The llvm.arm.neon.vaddl*, llvm.arm.neon.vaddw*, llvm.arm.neon.vsubl*, and
693 llvm.arm.neon.vsubw* intrinsics (lengthening vector add and subtract) have
694 been removed. They are replaced by vector add and vector subtract operations
695 where one (vaddw, vsubw) or both (vaddl, vsubl) of the operands are either
696 sign-extended or zero-extended.
699 The llvm.arm.neon.vmulls, llvm.arm.neon.vmullu, llvm.arm.neon.vmlal*, and
700 llvm.arm.neon.vmlsl* intrinsics (lengthening vector multiply with and without
701 accumulation and subtraction) have been removed. These operations are now
702 represented as vector multiplications where the operands are either
703 sign-extended or zero-extended, followed by a vector add for vmlal or a
704 vector subtract for vmlsl. Note that the polynomial vector multiply
705 intrinsic, llvm.arm.neon.vmullp, remains unchanged.
711 <!--=========================================================================-->
712 <div class="doc_subsection">
713 <a name="newapis">New Useful APIs</a>
716 <div class="doc_text">
718 <p>This release includes a number of new APIs that are used internally, which
719 may also be useful for external clients.
729 <!--=========================================================================-->
730 <div class="doc_subsection">
731 <a name="otherimprovements">Other Improvements and New Features</a>
734 <div class="doc_text">
735 <p>Other miscellaneous features include:</p>
744 <!--=========================================================================-->
745 <div class="doc_subsection">
746 <a name="changes">Major Changes and Removed Features</a>
749 <div class="doc_text">
751 <p>If you're already an LLVM user or developer with out-of-tree changes based
752 on LLVM 2.7, this section lists some "gotchas" that you may run into upgrading
753 from the previous release.</p>
756 <li>.ll file doesn't produce #uses comments anymore, to get them, run a .bc file
757 through "llvm-dis --show-annotations".</li>
758 <li>MSIL Backend removed.</li>
759 <li>ABCD and SSI passes removed.</li>
760 <li>'Union' LLVM IR feature removed.</li>
763 <p>In addition, many APIs have changed in this release. Some of the major LLVM
772 <!-- *********************************************************************** -->
773 <div class="doc_section">
774 <a name="knownproblems">Known Problems</a>
776 <!-- *********************************************************************** -->
778 <div class="doc_text">
780 <p>This section contains significant known problems with the LLVM system,
781 listed by component. If you run into a problem, please check the <a
782 href="http://llvm.org/bugs/">LLVM bug database</a> and submit a bug if
783 there isn't already one.</p>
787 <!-- ======================================================================= -->
788 <div class="doc_subsection">
789 <a name="experimental">Experimental features included with this release</a>
792 <div class="doc_text">
794 <p>The following components of this LLVM release are either untested, known to
795 be broken or unreliable, or are in early development. These components should
796 not be relied on, and bugs should not be filed against them, but they may be
797 useful to some people. In particular, if you would like to work on one of these
798 components, please contact us on the <a
799 href="http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev">LLVMdev list</a>.</p>
802 <li>The Alpha, SPU, MIPS, PIC16, Blackfin, MSP430, SystemZ and MicroBlaze
803 backends are experimental.</li>
804 <li><tt>llc</tt> "<tt>-filetype=asm</tt>" (the default) is the only
805 supported value for this option. XXX Update me</li>
810 <!-- ======================================================================= -->
811 <div class="doc_subsection">
812 <a name="x86-be">Known problems with the X86 back-end</a>
815 <div class="doc_text">
818 <li>The X86 backend does not yet support
819 all <a href="http://llvm.org/PR879">inline assembly that uses the X86
820 floating point stack</a>. It supports the 'f' and 't' constraints, but not
822 <li>Win64 code generation wasn't widely tested. Everything should work, but we
823 expect small issues to happen. Also, llvm-gcc cannot build the mingw64
824 runtime currently due to lack of support for the 'u' inline assembly
825 constraint and for X87 floating point inline assembly.</li>
826 <li>The X86-64 backend does not yet support the LLVM IR instruction
827 <tt>va_arg</tt>. Currently, front-ends support variadic
828 argument constructs on X86-64 by lowering them manually.</li>
833 <!-- ======================================================================= -->
834 <div class="doc_subsection">
835 <a name="ppc-be">Known problems with the PowerPC back-end</a>
838 <div class="doc_text">
841 <li>The Linux PPC32/ABI support needs testing for the interpreter and static
842 compilation, and lacks support for debug information.</li>
847 <!-- ======================================================================= -->
848 <div class="doc_subsection">
849 <a name="arm-be">Known problems with the ARM back-end</a>
852 <div class="doc_text">
855 <li>Thumb mode works only on ARMv6 or higher processors. On sub-ARMv6
856 processors, thumb programs can crash or produce wrong
857 results (<a href="http://llvm.org/PR1388">PR1388</a>).</li>
858 <li>Compilation for ARM Linux OABI (old ABI) is supported but not fully tested.
864 <!-- ======================================================================= -->
865 <div class="doc_subsection">
866 <a name="sparc-be">Known problems with the SPARC back-end</a>
869 <div class="doc_text">
872 <li>The SPARC backend only supports the 32-bit SPARC ABI (-m32); it does not
873 support the 64-bit SPARC ABI (-m64).</li>
878 <!-- ======================================================================= -->
879 <div class="doc_subsection">
880 <a name="mips-be">Known problems with the MIPS back-end</a>
883 <div class="doc_text">
886 <li>64-bit MIPS targets are not supported yet.</li>
891 <!-- ======================================================================= -->
892 <div class="doc_subsection">
893 <a name="alpha-be">Known problems with the Alpha back-end</a>
896 <div class="doc_text">
900 <li>On 21164s, some rare FP arithmetic sequences which may trap do not have the
901 appropriate nops inserted to ensure restartability.</li>
906 <!-- ======================================================================= -->
907 <div class="doc_subsection">
908 <a name="c-be">Known problems with the C back-end</a>
911 <div class="doc_text">
913 <p>The C backend has numerous problems and is not being actively maintained.
914 Depending on it for anything serious is not advised.</p>
917 <li><a href="http://llvm.org/PR802">The C backend has only basic support for
918 inline assembly code</a>.</li>
919 <li><a href="http://llvm.org/PR1658">The C backend violates the ABI of common
920 C++ programs</a>, preventing intermixing between C++ compiled by the CBE and
921 C++ code compiled with <tt>llc</tt> or native compilers.</li>
922 <li>The C backend does not support all exception handling constructs.</li>
923 <li>The C backend does not support arbitrary precision integers.</li>
929 <!-- ======================================================================= -->
930 <div class="doc_subsection">
931 <a name="llvm-gcc">Known problems with the llvm-gcc front-end</a>
934 <div class="doc_text">
936 <p>llvm-gcc is generally very stable for the C family of languages. The only
937 major language feature of GCC not supported by llvm-gcc is the
938 <tt>__builtin_apply</tt> family of builtins. However, some extensions
939 are only supported on some targets. For example, trampolines are only
940 supported on some targets (these are used when you take the address of a
941 nested function).</p>
943 <p>Fortran support generally works, but there are still several unresolved bugs
944 in <a href="http://llvm.org/bugs/">Bugzilla</a>. Please see the
945 tools/gfortran component for details. Note that llvm-gcc is missing major
946 Fortran performance work in the frontend and library that went into GCC after
947 4.2. If you are interested in Fortran, we recommend that you consider using
948 <a href="#dragonegg">dragonegg</a> instead.</p>
950 <p>The llvm-gcc 4.2 Ada compiler has basic functionality. However, this is not a
951 mature technology, and problems should be expected. For example:</p>
953 <li>The Ada front-end currently only builds on X86-32. This is mainly due
954 to lack of trampoline support (pointers to nested functions) on other platforms.
955 However, it <a href="http://llvm.org/PR2006">also fails to build on X86-64</a>
956 which does support trampolines.</li>
957 <li>The Ada front-end <a href="http://llvm.org/PR2007">fails to bootstrap</a>.
958 This is due to lack of LLVM support for <tt>setjmp</tt>/<tt>longjmp</tt> style
959 exception handling, which is used internally by the compiler.
960 Workaround: configure with <tt>--disable-bootstrap</tt>.</li>
961 <li>The c380004, <a href="http://llvm.org/PR2010">c393010</a>
962 and <a href="http://llvm.org/PR2421">cxg2021</a> ACATS tests fail
963 (c380004 also fails with gcc-4.2 mainline).
964 If the compiler is built with checks disabled then <a href="http://llvm.org/PR2010">c393010</a>
965 causes the compiler to go into an infinite loop, using up all system memory.</li>
966 <li>Some GCC specific Ada tests continue to crash the compiler.</li>
967 <li>The <tt>-E</tt> binder option (exception backtraces)
968 <a href="http://llvm.org/PR1982">does not work</a> and will result in programs
969 crashing if an exception is raised. Workaround: do not use <tt>-E</tt>.</li>
970 <li>Only discrete types <a href="http://llvm.org/PR1981">are allowed to start
971 or finish at a non-byte offset</a> in a record. Workaround: do not pack records
972 or use representation clauses that result in a field of a non-discrete type
973 starting or finishing in the middle of a byte.</li>
974 <li>The <tt>lli</tt> interpreter <a href="http://llvm.org/PR2009">considers
975 'main' as generated by the Ada binder to be invalid</a>.
976 Workaround: hand edit the file to use pointers for <tt>argv</tt> and
977 <tt>envp</tt> rather than integers.</li>
978 <li>The <tt>-fstack-check</tt> option <a href="http://llvm.org/PR2008">is
983 <!-- *********************************************************************** -->
984 <div class="doc_section">
985 <a name="additionalinfo">Additional Information</a>
987 <!-- *********************************************************************** -->
989 <div class="doc_text">
991 <p>A wide variety of additional information is available on the <a
992 href="http://llvm.org">LLVM web page</a>, in particular in the <a
993 href="http://llvm.org/docs/">documentation</a> section. The web page also
994 contains versions of the API documentation which is up-to-date with the
995 Subversion version of the source code.
996 You can access versions of these documents specific to this release by going
997 into the "<tt>llvm/doc/</tt>" directory in the LLVM tree.</p>
999 <p>If you have any questions or comments about LLVM, please feel free to contact
1000 us via the <a href="http://llvm.org/docs/#maillist"> mailing
1005 <!-- *********************************************************************** -->
1009 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
1010 src="http://jigsaw.w3.org/css-validator/images/vcss-blue" alt="Valid CSS"></a>
1011 <a href="http://validator.w3.org/check/referer"><img
1012 src="http://www.w3.org/Icons/valid-html401-blue" alt="Valid HTML 4.01"></a>
1014 <a href="http://llvm.org/">LLVM Compiler Infrastructure</a><br>
1015 Last modified: $Date$