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 2.5 Release Notes</title>
11 <div class="doc_title">LLVM 2.5 Release Notes</div>
14 <li><a href="#intro">Introduction</a></li>
15 <li><a href="#subproj">Sub-project Status Update</a></li>
16 <li><a href="#externalproj">External Projects Using LLVM 2.5</a></li>
17 <li><a href="#whatsnew">What's New in LLVM 2.5?</a></li>
18 <li><a href="GettingStarted.html">Installation Instructions</a></li>
19 <li><a href="#portability">Portability and Supported Platforms</a></li>
20 <li><a href="#knownproblems">Known Problems</a></li>
21 <li><a href="#additionalinfo">Additional Information</a></li>
24 <div class="doc_author">
25 <p>Written by the <a href="http://llvm.org">LLVM Team</a></p>
28 <!-- *********************************************************************** -->
29 <div class="doc_section">
30 <a name="intro">Introduction</a>
32 <!-- *********************************************************************** -->
34 <div class="doc_text">
36 <p>This document contains the release notes for the LLVM Compiler
37 Infrastructure, release 2.5. Here we describe the status of LLVM, including
38 major improvements from the previous release and significant known problems.
39 All LLVM releases may be downloaded from the <a
40 href="http://llvm.org/releases/">LLVM releases web site</a>.</p>
42 <p>For more information about LLVM, including information about the latest
43 release, please check out the <a href="http://llvm.org/">main LLVM
44 web site</a>. If you have questions or comments, the <a
45 href="http://mail.cs.uiuc.edu/mailman/listinfo/llvmdev">LLVM Developer's Mailing
46 List</a> is a good place to send them.</p>
48 <p>Note that if you are reading this file from a Subversion checkout or the
49 main LLVM web page, this document applies to the <i>next</i> release, not the
50 current one. To see the release notes for a specific release, please see the
51 <a href="http://llvm.org/releases/">releases page</a>.</p>
55 <!-- Unfinished features in 2.5:
58 target-specific intrinsics
60 pre-alloc splitter, strong phi elim
61 llc -enable-value-prop, propagation of value info (sign/zero ext info) from
63 debug info for optimized code
65 postalloc scheduler: anti dependence breaking, hazard recognizer?
67 initial support for debug line numbers when optimization enabled, not useful in
68 2.5 but will be for 2.6.
72 <!-- for announcement email:
75 <!-- *********************************************************************** -->
76 <div class="doc_section">
77 <a name="subproj">Sub-project Status Update</a>
79 <!-- *********************************************************************** -->
81 <div class="doc_text">
83 The LLVM 2.5 distribution currently consists of code from the core LLVM
84 repository —which roughly includes the LLVM optimizers, code generators
85 and supporting tools — and the llvm-gcc repository. In addition to this
86 code, the LLVM Project includes other sub-projects that are in development. The
87 two which are the most actively developed are the <a href="#clang">Clang
88 Project</a> and the <a href="#vmkit">VMKit Project</a>.
94 <!--=========================================================================-->
95 <div class="doc_subsection">
96 <a name="clang">Clang: C/C++/Objective-C Frontend Toolkit</a>
99 <div class="doc_text">
101 <p>The <a href="http://clang.llvm.org/">Clang project</a> is an effort to build
102 a set of new 'LLVM native' front-end technologies for the LLVM optimizer and
103 code generator. While Clang is not included in the LLVM 2.5 release, it is
104 continuing to make major strides forward in all areas. Its C and Objective-C
105 parsing and code generation support is now very solid. For example, it is
106 capable of successfully building many real-world applications for X86-32
108 including <a href="http://wiki.freebsd.org/BuildingFreeBSDWithClang">the FreeBSD
109 kernel</a>. C++ is also
110 making <a href="http://clang.llvm.org/cxx_status.html">incredible progress</a>,
111 and work on templates has recently started.</p>
113 <p>While Clang is not yet production quality, it is progressing very nicely and
114 is quite usable for building many C and Objective-C applications. If you are
115 interested in fast compiles and good diagnostics, we encourage you to try it out
116 by <a href="http://clang.llvm.org/get_started.html">building from mainline</a>
117 and reporting any issues you hit to the <a
118 href="http://lists.cs.uiuc.edu/mailman/listinfo/cfe-dev">Clang front-end mailing
121 <p>In the LLVM 2.5 time-frame, the Clang team has made many improvements:</p>
124 <li>Clang now has a new driver, which is focused on providing a GCC-compatible
126 <li>The X86-64 ABI is now supported.</li>
127 <li>Precompiled header support is now implemented.</li>
128 <li>Objective-C support is significantly improved beyond LLVM 2.4, supporting
129 many features, such as Objective-C Garbage Collection.</li>
130 <li>Many many bugs are fixed and many features have been added.</li>
134 <!--=========================================================================-->
135 <div class="doc_subsection">
136 <a name="clangsa">Clang Static Analyzer</a>
139 <div class="doc_text">
141 <p>Previously announced in the last LLVM release, the Clang project also
142 includes an early stage static source code analysis tool for <a
143 href="http://clang.llvm.org/StaticAnalysis.html">automatically finding bugs</a>
144 in C and Objective-C programs. The tool performs a growing set of checks to find
145 bugs that occur on a specific path within a program.</p>
147 <p>In the LLVM 2.5 time-frame there have been many significant improvements to
148 the analyzer's core path simulation engine and machinery for generating
149 path-based bug reports to end-users. Particularly noteworthy improvements
150 include experimental support for full field-sensitivity and reasoning about heap
151 objects as well as an improved value-constraints subengine that does a much
152 better job of reasoning about inequality relationships (e.g., <tt>x > 2</tt>)
153 between variables and constants.
155 <p>The set of checks performed by the static analyzer continue to expand, and
156 future plans for the tool include full source-level inter-procedural analysis
157 and deeper checks such as buffer overrun detection. There are many opportunities
158 to extend and enhance the static analyzer, and anyone interested in working on
159 this project is encouraged to get involved!</p>
163 <!--=========================================================================-->
164 <div class="doc_subsection">
165 <a name="vmkit">VMKit: JVM/CLI Virtual Machine Implementation</a>
168 <div class="doc_text">
170 The <a href="http://vmkit.llvm.org/">VMKit project</a> is an implementation of
171 a JVM and a CLI Virtual Machines (Microsoft .NET is an
172 implementation of the CLI) using the Just-In-Time compiler of LLVM.</p>
174 <p>Following LLVM 2.5, VMKit has its first release that you can find on its
175 <a href="http://vmkit.llvm.org/releases/">webpage</a>. The release includes
176 bug fixes, cleanup and new features. The major changes are:</p>
180 <li>Ahead of Time compiler: compiles .class files to llvm .bc. VMKit uses this
181 functionality to native compile the standard classes (eg java.lang.String).
182 Users can compile AOT .class files into dynamic libraries and run them with the
185 <li>New exception model: the dwarf exception model is very slow for
186 exception-intensive applications, so the JVM has had a new implementation of
187 exceptions which check at each function call if an exception happened. There is
188 a low performance penalty on applications without exceptions, but it is a big
189 gain for exception-intensive applications. For example the jack benchmark in
190 Spec JVM98 is 6x faster (performance gain of 83%).</li>
192 <li>New support for OSX/X64, Linux/X64 (with the Boehm GC), Linux/ppc32.</li>
197 <!-- *********************************************************************** -->
198 <div class="doc_section">
199 <a name="externalproj">External Projects Using LLVM 2.5</a>
201 <!-- *********************************************************************** -->
203 <!--=========================================================================-->
204 <div class="doc_subsection">
205 <a name="pure">Pure</a>
208 <div class="doc_text">
210 <a href="http://pure-lang.googlecode.com/">Pure</a>
211 is an algebraic/functional programming language based on term rewriting.
212 Programs are collections of equations which are used to evaluate expressions in
213 a symbolic fashion. Pure offers dynamic typing, eager and lazy evaluation,
214 lexical closures, a hygienic macro system (also based on term rewriting),
215 built-in list and matrix support (including list and matrix comprehensions) and
216 an easy-to-use C interface. The interpreter uses LLVM as a backend to
217 JIT-compile Pure programs to fast native code.</p>
219 <p>In addition to the usual algebraic data structures, Pure also has
220 MATLAB-style matrices in order to support numeric computations and signal
221 processing in an efficient way. Pure is mainly aimed at mathematical
222 applications right now, but it has been designed as a general purpose language.
223 The dynamic interpreter environment and the C interface make it possible to use
224 it as a kind of functional scripting language for many application areas.
229 <!--=========================================================================-->
230 <div class="doc_subsection">
231 <a name="ldc">LLVM D Compiler</a>
234 <div class="doc_text">
236 <a href="http://www.dsource.org/projects/ldc">LDC</a> is an implementation of
237 the D Programming Language using the LLVM optimizer and code generator.
238 LDC project works great with the LLVM 2.5 release. General improvmenets in this
239 cycle have included new inline asm constraint handling, better debug info
240 support, general bugfixes, and better x86-64 support. This has allowed
241 some major improvements in LDC, getting us much closer to being as
242 fully featured as the original DMD compiler from DigitalMars.
246 <!--=========================================================================-->
247 <div class="doc_subsection">
248 <a name="RoadsendPHP">Roadsend PHP</a>
251 <div class="doc_text">
252 <p><a href="http://code.roadsend.com/rphp">Roadsend PHP</a> (rphp) is an open
253 source compiler for the PHP programming language that uses LLVM for its
254 optimizer, JIT, and static compiler. This is a reimplementation of an earlier
255 project that is now based on the LLVM.</p>
259 <!-- *********************************************************************** -->
260 <div class="doc_section">
261 <a name="whatsnew">What's New in LLVM 2.5?</a>
263 <!-- *********************************************************************** -->
265 <div class="doc_text">
267 <p>This release includes a huge number of bug fixes, performance tweaks, and
268 minor improvements. Some of the major improvements and new features are listed
273 <!--=========================================================================-->
274 <div class="doc_subsection">
275 <a name="majorfeatures">Major New Features</a>
278 <div class="doc_text">
280 <p>LLVM 2.5 includes several major new capabilities:</p>
283 <li>LLVM 2.5 includes a brand new <a
284 href="http://en.wikipedia.org/wiki/XCore">XCore</a> backend.</li>
286 <li>llvm-gcc now generally supports the GFortan front-end, and the precompiled
287 release binaries now support Fortran, even on Mac OS/X.</li>
289 <li>CMake is now used by the <a href="GettingStartedVS.html">LLVM build process
290 on Windows</a>. It automatically generates Visual Studio project files (and
291 more) from a set of simple text files. This makes it much easier to
292 maintain. In time, we'd like to standardize on CMake for everything.</li>
294 <li>LLVM 2.5 now uses (and includes) Google Test for unit testing.</li>
296 <li>The LLVM native code generator now supports arbitrary precision integers.
297 Types like <tt>i33</tt> have long been valid in the LLVM IR, but were previously
298 only supported by the interpreter. Note that the C backend still does not
301 <li>LLVM 2.5 no longer uses 'bison', so it is easier to build on Windows.</li>
307 <!--=========================================================================-->
308 <div class="doc_subsection">
309 <a name="llvm-gcc">llvm-gcc 4.2 Improvements</a>
312 <div class="doc_text">
314 <p>LLVM fully supports the llvm-gcc 4.2 front-end, which marries the GCC
315 front-ends and driver with the LLVM optimizer and code generator. It currently
316 includes support for the C, C++, Objective-C, Ada, and Fortran front-ends.</p>
319 <li>In this release, the GCC inliner is completely disabled. Previously the GCC
320 inliner was used to handle always-inline functions and other cases. This caused
321 problems with code size growth, and it is completely disabled in this
324 <li>llvm-gcc (and LLVM in general) now support code generation for stack
325 canaries, which is an effective form of <a
326 href="http://en.wikipedia.org/wiki/Stack-smashing_protection">buffer overflow
327 protection</a>. llvm-gcc supports this with the <tt>-fstack-protector</tt>
328 command line option (just like GCC). In LLVM IR, you can request code
329 generation for stack canaries with function attributes.
336 <!--=========================================================================-->
337 <div class="doc_subsection">
338 <a name="coreimprovements">LLVM IR and Core Improvements</a>
341 <div class="doc_text">
342 <p>LLVM IR has several new features that are used by our existing front-ends and
343 can be useful if you are writing a front-end for LLVM:</p>
346 <li>The <a href="LangRef.html#i_shufflevector">shufflevector</a> instruction
347 has been generalized to allow different shuffle mask width than its input
348 vectors. This allows you to use shufflevector to combine two
349 "<4 x float>" vectors into a "<8 x float>" for example.</li>
351 <li>LLVM IR now supports new intrinsics for computing and acting on <a
352 href="LangRef.html#int_overflow">overflow of integer operations</a>. This allows
353 efficient code generation for languages that must trap or throw an exception on
354 overflow. While these intrinsics work on all targets, they only generate
355 efficient code on X86 so far.</li>
357 <li>LLVM IR now supports a new <a href="LangRef.html#linkage">private
358 linkage</a> type to produce labels that are stripped by the assembler before it
359 produces a .o file (thus they are invisible to the linker).</li>
361 <li>LLVM IR supports two new attributes for better alias analysis. The <a
362 href="LangRef.html#paramattrs">noalias</a> attribute can now be used on the
363 return value of a function to indicate that it returns new memory (e.g.
364 'malloc', 'calloc', etc).</li>
366 <li>The new <a href="LangRef.html#paramattrs">nocapture</a> attribute can be
367 used on pointer arguments to functions that access through but do not return the
368 pointer in a data structure that out lives the call (e.g. 'strlen', 'memcpy',
369 and many others). The simplifylibcalls pass applies these attributes to
370 standard libc functions.</li>
372 <li>The parser for ".ll" files in lib/AsmParser is now completely rewritten as a
373 recursive descent parser. This parser produces better error messages (including
374 caret diagnostics) is less fragile (less likely to crash on strange things) does
375 not leak memory, is more efficient, and eliminates LLVM's last use of the
378 <li>Debug information representation and manipulation internals have been
379 consolidated to use a new set of classes in
380 <tt>llvm/Analysis/DebugInfo.h</tt> classes. These routines are more
381 efficient, robust, and extensible and replace the older mechanisms.
382 llvm-gcc, clang, and the code generator now use them to create and process
383 debug information.</li>
389 <!--=========================================================================-->
390 <div class="doc_subsection">
391 <a name="optimizer">Optimizer Improvements</a>
394 <div class="doc_text">
396 <p>In addition to a huge array of bug fixes and minor performance tweaks, this
397 release includes a few major enhancements and additions to the optimizers:</p>
401 <li>The loop optimizer now improves floating point induction variables in
402 several ways, including adding shadow induction variables to avoid
403 "integer <-> floating point" conversions in loops when safe.</li>
405 <li>The "-mem2reg" pass is now much faster on code with huge basic blocks.</li>
407 <li>The "-jump-threading" pass is more powerful: it is iterative
408 and handles threading based on values with fully and partially redundant
411 <li>The "-memdep" memory dependence analysis pass (used by GVN and memcpyopt) is
412 both faster and more aggressive.</li>
414 <li>The "-scalarrepl" scalar replacement of aggregates pass is more aggressive
415 about promoting unions to registers.</li>
421 <!--=========================================================================-->
422 <div class="doc_subsection">
423 <a name="codegen">Target Independent Code Generator Improvements</a>
426 <div class="doc_text">
428 <p>We have put a significant amount of work into the code generator
429 infrastructure, which allows us to implement more aggressive algorithms and make
433 <li>The <a href="WritingAnLLVMBackend.html">Writing an LLVM Compiler
434 Backend</a> document has been greatly expanded and is substantially more
437 <li>The SelectionDAG type legalization logic has been completely rewritten, is
438 now more powerful (it supports arbitrary precision integer types for example),
439 and more correct in several corner cases. The type legalizer converts
440 operations on types that are not natively supported by the target machine into
441 equivalent code sequences that only use natively supported types. The old type
442 legalizer is still available (for now) and will be used if
443 <tt>-disable-legalize-types</tt> is passed to the code generator.
446 <li>The code generator now supports widening illegal vectors to larger legal
447 ones (for example, converting operations on <3 x float> to work on
448 <4 x float>) which is very important for common graphics
451 <li>The assembly printers for each target are now split out into their own
452 libraries that are separate from the main code generation logic. This reduces
453 code size of JIT compilers by not requiring them to be linked in.</li>
455 <li>The 'fast' instruction selection path (used at -O0 and for fast JIT
456 compilers) now supports accelerating codegen for code that uses exception
457 handling constructs.</li>
459 <li>The optional PBQP register allocator now supports register coalescing.</li>
463 <!--=========================================================================-->
464 <div class="doc_subsection">
465 <a name="x86">X86-32 and X86-64 Target Improvements</a>
468 <div class="doc_text">
469 <p>New features of the X86 target include:
473 <li>The "<a href="LangRef.html#int_returnaddress">llvm.returnaddress</a>"
474 intrinsic (which is used to implement "__builtin_return_address") now supports
475 non-zero stack depths on X86.</li>
477 <li>The X86 backend now supports code generation of vector shift operations
478 using SSE instructions.</li>
480 <li>X86-64 code generation now takes advantage of red zone (unless
481 -mno-red-zone option is specified).</li>
483 <li>The X86 backend now supports using address space #256 in LLVM IR as a way of
484 performing memory references off the GS segment register. This allows a
485 front-end to take advantage of very low-level programming techniques when
486 targetting X86 CPUs. See test/CodeGen/X86/movgs.ll for a simple example.</li>
488 <li>The X86 backend now supports a <tt>-disable-mmx</tt> command line option to
489 prevent use of MMX even on chips that support it. This is important for cases
490 where code does not contain the proper "llvm.x86.mmx.emms" intrinsics.</li>
492 <li>The X86 JIT now detects the new Intel "<a
493 href="http://en.wikipedia.org/wiki/Intel_Core_i7">Core i7</a>" and <a
494 href="http://en.wikipedia.org/wiki/Intel_Atom">Atom</a>" chips,
495 auto-configuring itself appropriately for the features of these chips.</li>
497 <li>The JIT now supports exception handling constructs on Linux/X86-64 and
499 <li>The JIT supports Thread Local Storage (TLS) on Linux/X86-32 but not yet on
506 <!--=========================================================================-->
507 <div class="doc_subsection">
508 <a name="pic16">PIC16 Target Improvements</a>
511 <div class="doc_text">
512 <p>New features of the PIC16 target include:
516 <li>Both direct and indirect load/stores work now.</li>
517 <li>Logical, bitwise and conditional operations now work for integer data
519 <li>Function calls involving basic types work now.</li>
520 <li>Support for integer arrays.</li>
521 <li>Compiler can now emit libcalls for operations not support by m/c insns.</li>
522 <li>Support for both data and rom address spaces.</li>
525 <p>Things not yet supported:</p>
528 <li>Floating point.</li>
529 <li>Passing/returning aggregate types to/from functions.</li>
530 <li>Variable arguments.</li>
531 <li>Indirect function calls.</li>
532 <li>Interrupts/programs.</li>
539 <!--=========================================================================-->
540 <div class="doc_subsection">
541 <a name="llvmc">Improvements in LLVMC</a>
544 <div class="doc_text">
545 <p>New features include:</p>
548 <li>Beginning with LLVM 2.5, <tt>llvmc2</tt> is known as
549 just <tt>llvmc</tt>. The old <tt>llvmc</tt> driver was removed.</li>
551 <li>The Clang plugin was substantially improved and is now enabled
552 by default. The command <tt>llvmc --clang</tt> can be now used as a
553 synonym to <tt>ccc</tt>.</li>
555 <li>There is now a <tt>--check-graph</tt> option which is supposed to catch
556 common errors like multiple default edges, mismatched output/input language
557 names and cycles. In general, these checks can't be done at compile-time
558 because of the need to support plugins.</li>
560 <li>Plugins are now more flexible and can refer to compilation graph nodes and
561 options defined in other plugins. To manage dependencies, a priority-sorting
562 mechanism was introduced. This change affects the TableGen file syntax; see the
563 documentation for details.</li>
565 <li>Hooks can now be provided with arguments. The syntax is "<tt>$CALL(MyHook,
566 'Arg1', 'Arg2', 'Arg #3')</tt>".</li>
568 <li>A new option type: multi-valued option, for options that take more than one
569 argument (for example, "<tt>-foo a b c</tt>").</li>
571 <li>New option properties: '<tt>one_or_more</tt>', '<tt>zero_or_more</tt>',
572 '<tt>hidden</tt>' and '<tt>really_hidden</tt>'.</li>
574 <li>The '<tt>case</tt>' expression gained an '<tt>error</tt>' action and
575 an '<tt>empty</tt>' test (equivalent to "<tt>(not (not_empty ...))</tt>").</li>
577 <li>Documentation now looks more consistent to the rest of the LLVM
578 docs. There is also a man page now.</li>
585 <!--=========================================================================-->
586 <div class="doc_subsection">
587 <a name="changes">Major Changes and Removed Features</a>
590 <div class="doc_text">
592 <p>If you're already an LLVM user or developer with out-of-tree changes based
593 on LLVM 2.4, this section lists some "gotchas" that you may run into upgrading
594 from the previous release.</p>
598 <li>llvm-gcc defaults to -fno-math-errno on all X86 targets.</li>
602 <p>In addition, many APIs have changed in this release. Some of the major LLVM
617 <!-- *********************************************************************** -->
618 <div class="doc_section">
619 <a name="portability">Portability and Supported Platforms</a>
621 <!-- *********************************************************************** -->
623 <div class="doc_text">
625 <p>LLVM is known to work on the following platforms:</p>
628 <li>Intel and AMD machines (IA32, X86-64, AMD64, EMT-64) running Red Hat
629 Linux, Fedora Core and FreeBSD (and probably other unix-like systems).</li>
630 <li>PowerPC and X86-based Mac OS X systems, running 10.3 and above in 32-bit
631 and 64-bit modes.</li>
632 <li>Intel and AMD machines running on Win32 using MinGW libraries (native).</li>
633 <li>Intel and AMD machines running on Win32 with the Cygwin libraries (limited
634 support is available for native builds with Visual C++).</li>
635 <li>Sun UltraSPARC workstations running Solaris 10.</li>
636 <li>Alpha-based machines running Debian GNU/Linux.</li>
637 <li>Itanium-based (IA64) machines running Linux and HP-UX.</li>
640 <p>The core LLVM infrastructure uses GNU autoconf to adapt itself
641 to the machine and operating system on which it is built. However, minor
642 porting may be required to get LLVM to work on new platforms. We welcome your
643 portability patches and reports of successful builds or error messages.</p>
647 <!-- *********************************************************************** -->
648 <div class="doc_section">
649 <a name="knownproblems">Known Problems</a>
651 <!-- *********************************************************************** -->
653 <div class="doc_text">
655 <p>This section contains significant known problems with the LLVM system,
656 listed by component. If you run into a problem, please check the <a
657 href="http://llvm.org/bugs/">LLVM bug database</a> and submit a bug if
658 there isn't already one.</p>
662 <!-- ======================================================================= -->
663 <div class="doc_subsection">
664 <a name="experimental">Experimental features included with this release</a>
667 <div class="doc_text">
669 <p>The following components of this LLVM release are either untested, known to
670 be broken or unreliable, or are in early development. These components should
671 not be relied on, and bugs should not be filed against them, but they may be
672 useful to some people. In particular, if you would like to work on one of these
673 components, please contact us on the <a
674 href="http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev">LLVMdev list</a>.</p>
677 <li>The MSIL, IA64, Alpha, SPU, MIPS, and PIC16 backends are experimental.</li>
678 <li>The llc "<tt>-filetype=asm</tt>" (the default) is the only supported
679 value for this option.</li>
684 <!-- ======================================================================= -->
685 <div class="doc_subsection">
686 <a name="x86-be">Known problems with the X86 back-end</a>
689 <div class="doc_text">
692 <li>The X86 backend does not yet support
693 all <a href="http://llvm.org/PR879">inline assembly that uses the X86
694 floating point stack</a>. It supports the 'f' and 't' constraints, but not
696 <li>The X86 backend generates inefficient floating point code when configured
697 to generate code for systems that don't have SSE2.</li>
698 <li>Win64 code generation wasn't widely tested. Everything should work, but we
699 expect small issues to happen. Also, llvm-gcc cannot build mingw64 runtime
701 to <a href="http://llvm.org/PR2255">several</a>
702 <a href="http://llvm.org/PR2257">bugs</a> due to lack of support for the
703 'u' inline assembly constraint and X87 floating point inline assembly.</li>
704 <li>The X86-64 backend does not yet support the LLVM IR instruction
705 <tt>va_arg</tt>. Currently, the llvm-gcc and front-ends support variadic
706 argument constructs on X86-64 by lowering them manually.</li>
711 <!-- ======================================================================= -->
712 <div class="doc_subsection">
713 <a name="ppc-be">Known problems with the PowerPC back-end</a>
716 <div class="doc_text">
719 <li>The Linux PPC32/ABI support needs testing for the interpreter and static
720 compilation, and lacks support for debug information.</li>
725 <!-- ======================================================================= -->
726 <div class="doc_subsection">
727 <a name="arm-be">Known problems with the ARM back-end</a>
730 <div class="doc_text">
733 <li>Thumb mode works only on ARMv6 or higher processors. On sub-ARMv6
734 processors, thumb programs can crash or produce wrong
735 results (<a href="http://llvm.org/PR1388">PR1388</a>).</li>
736 <li>Compilation for ARM Linux OABI (old ABI) is supported, but not fully tested.
738 <li>There is a bug in QEMU-ARM (<= 0.9.0) which causes it to incorrectly
740 programs compiled with LLVM. Please use more recent versions of QEMU.</li>
745 <!-- ======================================================================= -->
746 <div class="doc_subsection">
747 <a name="sparc-be">Known problems with the SPARC back-end</a>
750 <div class="doc_text">
753 <li>The SPARC backend only supports the 32-bit SPARC ABI (-m32), it does not
754 support the 64-bit SPARC ABI (-m64).</li>
759 <!-- ======================================================================= -->
760 <div class="doc_subsection">
761 <a name="mips-be">Known problems with the MIPS back-end</a>
764 <div class="doc_text">
767 <li>The O32 ABI is not fully supported.</li>
768 <li>64-bit MIPS targets are not supported yet.</li>
773 <!-- ======================================================================= -->
774 <div class="doc_subsection">
775 <a name="alpha-be">Known problems with the Alpha back-end</a>
778 <div class="doc_text">
782 <li>On 21164s, some rare FP arithmetic sequences which may trap do not have the
783 appropriate nops inserted to ensure restartability.</li>
788 <!-- ======================================================================= -->
789 <div class="doc_subsection">
790 <a name="ia64-be">Known problems with the IA64 back-end</a>
793 <div class="doc_text">
796 <li>The Itanium backend is highly experimental, and has a number of known
797 issues. We are looking for a maintainer for the Itanium backend. If you
798 are interested, please contact the LLVMdev mailing list.</li>
803 <!-- ======================================================================= -->
804 <div class="doc_subsection">
805 <a name="c-be">Known problems with the C back-end</a>
808 <div class="doc_text">
811 <li><a href="http://llvm.org/PR802">The C backend has only basic support for
812 inline assembly code</a>.</li>
813 <li><a href="http://llvm.org/PR1658">The C backend violates the ABI of common
814 C++ programs</a>, preventing intermixing between C++ compiled by the CBE and
815 C++ code compiled with llc or native compilers.</li>
816 <li>The C backend does not support all exception handling constructs.</li>
817 <li>The C backend does not support arbitrary precision integers.</li>
823 <!-- ======================================================================= -->
824 <div class="doc_subsection">
825 <a name="c-fe">Known problems with the llvm-gcc C front-end</a>
828 <div class="doc_text">
830 <p>llvm-gcc does not currently support <a href="http://llvm.org/PR869">Link-Time
831 Optimization</a> on most platforms "out-of-the-box". Please inquire on the
832 LLVMdev mailing list if you are interested.</p>
834 <p>The only major language feature of GCC not supported by llvm-gcc is
835 the <tt>__builtin_apply</tt> family of builtins. However, some extensions
836 are only supported on some targets. For example, trampolines are only
837 supported on some targets (these are used when you take the address of a
838 nested function).</p>
840 <p>If you run into GCC extensions which are not supported, please let us know.
845 <!-- ======================================================================= -->
846 <div class="doc_subsection">
847 <a name="c++-fe">Known problems with the llvm-gcc C++ front-end</a>
850 <div class="doc_text">
852 <p>The C++ front-end is considered to be fully
853 tested and works for a number of non-trivial programs, including LLVM
854 itself, Qt, Mozilla, etc.</p>
857 <li>Exception handling works well on the X86 and PowerPC targets. Currently
858 only Linux and Darwin targets are supported (both 32 and 64 bit).</li>
863 <!-- ======================================================================= -->
864 <div class="doc_subsection">
865 <a name="fortran-fe">Known problems with the llvm-gcc Fortran front-end</a>
868 <div class="doc_text">
870 <li>Fortran support generally works, but there are still several unresolved bugs
871 in Bugzilla. Please see the tools/gfortran component for details.</li>
873 <li>The Fortran front-end currently does not build on Darwin (without tweaks)
874 due to unresolved dependencies on the C front-end.</li>
878 <!-- ======================================================================= -->
879 <div class="doc_subsection">
880 <a name="ada-fe">Known problems with the llvm-gcc Ada front-end</a>
883 <div class="doc_text">
884 The llvm-gcc 4.2 Ada compiler works fairly well, however this is not a mature
885 technology and problems should be expected.
887 <li>The Ada front-end currently only builds on X86-32. This is mainly due
888 to lack of trampoline support (pointers to nested functions) on other platforms,
889 however it <a href="http://llvm.org/PR2006">also fails to build on X86-64</a>
890 which does support trampolines.</li>
891 <li>The Ada front-end <a href="http://llvm.org/PR2007">fails to bootstrap</a>.
892 This is due to lack of LLVM support for <tt>setjmp</tt>/<tt>longjmp</tt> style
893 exception handling, which is used internally by the compiler.
894 Workaround: configure with --disable-bootstrap.</li>
895 <li>The c380004, <a href="http://llvm.org/PR2010">c393010</a>
896 and <a href="http://llvm.org/PR2421">cxg2021</a> ACATS tests fail
897 (c380004 also fails with gcc-4.2 mainline).
898 If the compiler is built with checks disabled then <a href="http://llvm.org/PR2010">c393010</a>
899 causes the compiler to go into an infinite loop, using up all system memory.</li>
900 <li>Some gcc specific Ada tests continue to crash the compiler.</li>
901 <li>The -E binder option (exception backtraces)
902 <a href="http://llvm.org/PR1982">does not work</a> and will result in programs
903 crashing if an exception is raised. Workaround: do not use -E.</li>
904 <li>Only discrete types <a href="http://llvm.org/PR1981">are allowed to start
905 or finish at a non-byte offset</a> in a record. Workaround: do not pack records
906 or use representation clauses that result in a field of a non-discrete type
907 starting or finishing in the middle of a byte.</li>
908 <li>The <tt>lli</tt> interpreter <a href="http://llvm.org/PR2009">considers
909 'main' as generated by the Ada binder to be invalid</a>.
910 Workaround: hand edit the file to use pointers for <tt>argv</tt> and
911 <tt>envp</tt> rather than integers.</li>
912 <li>The <tt>-fstack-check</tt> option <a href="http://llvm.org/PR2008">is
917 <!-- *********************************************************************** -->
918 <div class="doc_section">
919 <a name="additionalinfo">Additional Information</a>
921 <!-- *********************************************************************** -->
923 <div class="doc_text">
925 <p>A wide variety of additional information is available on the <a
926 href="http://llvm.org">LLVM web page</a>, in particular in the <a
927 href="http://llvm.org/docs/">documentation</a> section. The web page also
928 contains versions of the API documentation which is up-to-date with the
929 Subversion version of the source code.
930 You can access versions of these documents specific to this release by going
931 into the "<tt>llvm/doc/</tt>" directory in the LLVM tree.</p>
933 <p>If you have any questions or comments about LLVM, please feel free to contact
934 us via the <a href="http://llvm.org/docs/#maillist"> mailing
939 <!-- *********************************************************************** -->
943 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
944 src="http://jigsaw.w3.org/css-validator/images/vcss-blue" alt="Valid CSS"></a>
945 <a href="http://validator.w3.org/check/referer"><img
946 src="http://www.w3.org/Icons/valid-html401-blue" alt="Valid HTML 4.01"></a>
948 <a href="http://llvm.org/">LLVM Compiler Infrastructure</a><br>
949 Last modified: $Date$