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 2.4 Release Notes</title>
11 <div class="doc_title">LLVM 2.4 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="#whatsnew">What's New in LLVM?</a></li>
17 <li><a href="GettingStarted.html">Installation Instructions</a></li>
18 <li><a href="#portability">Portability and Supported Platforms</a></li>
19 <li><a href="#knownproblems">Known Problems</a></li>
20 <li><a href="#additionalinfo">Additional Information</a></li>
23 <div class="doc_author">
24 <p>Written by the <a href="http://llvm.org">LLVM Team</a></p>
27 <!-- *********************************************************************** -->
28 <div class="doc_section">
29 <a name="intro">Introduction</a>
31 <!-- *********************************************************************** -->
33 <div class="doc_text">
35 <p>This document contains the release notes for the LLVM Compiler
36 Infrastructure, release 2.4. Here we describe the status of LLVM, including
37 major improvements from the previous release and significant known problems.
38 All LLVM releases may be downloaded from the <a
39 href="http://llvm.org/releases/">LLVM releases web site</a>.</p>
41 <p>For more information about LLVM, including information about the latest
42 release, please check out the <a href="http://llvm.org/">main LLVM
43 web site</a>. If you have questions or comments, the <a
44 href="http://mail.cs.uiuc.edu/mailman/listinfo/llvmdev">LLVM Developer's Mailing
45 List</a> is a good place to send them.</p>
47 <p>Note that if you are reading this file from a Subversion checkout or the
48 main LLVM web page, this document applies to the <i>next</i> release, not the
49 current one. To see the release notes for a specific release, please see the
50 <a href="http://llvm.org/releases/">releases page</a>.</p>
54 <!-- Unfinished features in 2.4:
58 llc -enable-value-prop, propagation of value info (sign/zero ext info) from
62 <!-- for announcement email:
67 <!-- *********************************************************************** -->
68 <div class="doc_section">
69 <a name="subproj">Sub-project Status Update</a>
71 <!-- *********************************************************************** -->
73 <div class="doc_text">
75 The LLVM 2.4 distribution currently consists of code from the core LLVM
76 repository (which roughly includes the LLVM optimizers, code generators and
77 supporting tools) and the llvm-gcc repository. In addition to this code, the
78 LLVM Project includes other sub-projects that are in development. The two which
79 are the most actively developed are the <a href="#clang">Clang Project</a> and
80 the <a href="#vmkit">VMKit Project</a>.
86 <!--=========================================================================-->
87 <div class="doc_subsection">
88 <a name="clang">Clang: C/C++/Objective-C Frontend Toolkit</a>
91 <div class="doc_text">
93 <p>The <a href="http://clang.llvm.org/">Clang project</a> is an effort to build
94 a set of new 'LLVM native' front-end technologies for the LLVM optimizer
95 and code generator. Clang is continuing to make major strides forward in all
96 areas. Its C and Objective-C parsing support is very solid, and the code
97 generation support is far enough along to build many C applications. While not
98 yet production quality, it is progressing very nicely. In addition, C++
99 front-end work has started to make significant progress.</p>
101 <p>Clang, in conjunction with the <tt>ccc</tt> driver, is now usable as a
102 replacement for gcc for building some small- to medium-sized C applications.
103 Additionally, Clang now has code generation support for Objective-C on Mac OS X
104 platform. Major highlights include:</p>
107 <li> Clang/ccc pass almost all of the LLVM test suite on Mac OS X and Linux
108 on the 32-bit x86 architecture. This includes significant C
109 applications such as <a href="http://www.sqlite.org">sqlite3</a>,
110 <a href="http://www.lua.org">lua</a>, and
111 <a href="http://www.clamav.net">Clam AntiVirus</a>. </li>
113 <li> Clang can build the majority of Objective-C examples shipped with the
114 Mac OS X Developer Tools. </li>
117 <p>Clang code generation still needs considerable testing and development,
118 however. Some areas under active development include:</p>
121 <li> Improved support for C and Objective-C features, for example
122 variable-length arrays, va_arg, exception handling (Obj-C), and garbage
123 collection (Obj-C). </li>
124 <li> ABI compatibility, especially for platforms other than 32-bit
130 <!--=========================================================================-->
131 <div class="doc_subsection">
132 <a name="clangsa">Clang Static Analyzer</a>
135 <div class="doc_text">
137 <p>The Clang project also includes an early stage static source code analysis
138 tool for <a href="http://clang.llvm.org/StaticAnalysis.html">automatically
139 finding bugs</a> in C and Objective-C programs. The tool performs a growing set
140 of checks to find bugs that occur on a specific path within a program. Examples
141 of bugs the tool finds include logic errors such as null dereferences,
142 violations of various API rules, dead code, and potential memory leaks in
143 Objective-C programs. Since its inception, public feedback on the tool has been
144 extremely positive, and conservative estimates put the number of real bugs it
145 has found in industrial-quality software on the order of thousands.</p>
147 <p>The tool also provides a simple web GUI to inspect potential bugs found by
148 the tool. While still early in development, the GUI illustrates some of the key
149 features of Clang: accurate source location information, which is used by the
150 GUI to highlight specific code expressions that relate to a bug (including those
151 that span multiple lines); and built-in knowledge of macros, which is used to
152 perform inline expansion of macros within the GUI itself.</p>
154 <p>The set of checks performed by the static analyzer is gradually expanding,
155 and future plans for the tool include full source-level inter-procedural
156 analysis and deeper checks such as buffer overrun detection. There are many
157 opportunities to extend and enhance the static analyzer, and anyone interested
158 in working on this project is encouraged to get involved!</p>
162 <!--=========================================================================-->
163 <div class="doc_subsection">
164 <a name="vmkit">VMKit: JVM/CLI Virtual Machine Implementation</a>
167 <div class="doc_text">
169 The <a href="http://vmkit.llvm.org/">VMKit project</a> is an implementation of
170 a JVM and a CLI Virtual Machines (Microsoft .NET is an
171 implementation of the CLI) using the Just-In-Time compiler of LLVM.</p>
173 <p>Following LLVM 2.4, VMKit has its first release 0.24 that you can find on its
174 <a href="http://vmkit.llvm.org/releases/">webpage</a>. The release includes
175 bug fixes, cleanup and new features. The major changes are:</p>
179 <li> Support for generics in the .Net virtual machine.</li>
180 <li> Initial support for the Mono class libraries. </li>
181 <li> Support for MacOSX/x86, following LLVM's support for exceptions in
182 JIT on MacOSX/x86. </li>
183 <li> A new vmkit driver: a program to run java or .net applications. The driver
184 supports llvm command line arguments including the new "-fast" option. </li>
185 <li> A new memory allocation scheme in the JVM that makes unloading a
186 class loader very fast. </li>
187 <li> VMKit now follows the LLVM Makefile machinery. </li>
193 <!-- *********************************************************************** -->
194 <div class="doc_section">
195 <a name="whatsnew">What's New in LLVM?</a>
197 <!-- *********************************************************************** -->
199 <div class="doc_text">
201 <p>This release includes a huge number of bug fixes, performance tweaks, and
202 minor improvements. Some of the major improvements and new features are listed
207 <!--=========================================================================-->
208 <div class="doc_subsection">
209 <a name="majorfeatures">Major New Features</a>
212 <div class="doc_text">
214 <p>LLVM 2.4 includes several major new capabilities:</p>
217 <li><p>The most visible end-user change in LLVM 2.4 is that it includes many
218 optimizations and changes to make -O0 compile times much faster. You should see
219 improvements in speed on the order of 30% (or more) than in LLVM 2.3. There are
220 many pieces to this change described in more detail below. The speedups and new
221 components can also be used for JIT compilers that want fast
222 compilation.</p></li>
224 <li><p>The biggest change to the LLVM IR is that Multiple Return Values (which
225 were introduced in LLVM 2.3) have been generalized to full support for "First
226 Class Aggregate" values in LLVM 2.4. This means that LLVM IR supports using
227 structs and arrays as values in a function. This capability is mostly useful
228 for front-end authors, who prefer to treat things like complex numbers, simple
229 tuples, dope vectors, etc., as Value*'s instead of as a tuple of Value*'s or as
230 memory values. Bitcode files from LLVM 2.3 will automatically migrate to the
231 general representation.</p></li>
233 <li><p>LLVM 2.4 also includes an initial port for the PIC16 microprocessor. This
234 target only has support for 8 bit registers, and a number of other crazy
235 constraints. While the port is still in early development stages, it shows some
236 interesting things you can do with LLVM.</p></li>
243 <!--=========================================================================-->
244 <div class="doc_subsection">
245 <a name="llvm-gcc">llvm-gcc 4.2 Improvements</a>
248 <div class="doc_text">
250 <p>LLVM fully supports the llvm-gcc 4.2 front-end, which marries the GCC
251 front-ends and driver with the LLVM optimizer and code generator. It currently
252 includes support for the C, C++, Objective-C, Ada, and Fortran front-ends.</p>
255 <li>LLVM 2.4 supports the full set of atomic <tt>__sync_*</tt> builtins. LLVM
256 2.3 only supported those used by OpenMP, but 2.4 supports them all. Note that
257 while llvm-gcc supports all of these builtins, not all targets do. X86 support
258 them all in both 32-bit and 64-bit mode and PowerPC supports them all except for
259 the 64-bit operations when in 32-bit mode.</li>
261 <li>llvm-gcc now supports an <tt>-flimited-precision</tt> option, which tells
262 the compiler that it is okay to use low-precision approximations of certain libm
263 functions (like <tt>exp</tt>, <tt>log</tt>, etc). This allows you to get high
264 performance if you only need (say) 12-bits of precision.</li>
266 <li>llvm-gcc now supports a C language extension known as "<a
267 href="http://lists.cs.uiuc.edu/pipermail/cfe-dev/2008-August/002670.html">Blocks</a>".
268 This feature is similar to nested functions and closures, but does not
269 require stack trampolines (with most ABIs), and supports returning closures
270 from functions that define them. Note that actually <em>using</em> Blocks
271 requires a small runtime that is not included with llvm-gcc.</li>
273 <li>llvm-gcc now supports a new <tt>-flto</tt> option. On systems that support
274 transparent Link Time Optimization (currently Darwin systems with Xcode 3.1 and
275 later) this allows the use of LTO with other optimization levels like -Os.
276 Previously, LTO could only be used with -O4, which implied optimizations in
277 -O3 that can increase code size.</li>
283 <!--=========================================================================-->
284 <div class="doc_subsection">
285 <a name="coreimprovements">LLVM Core Improvements</a>
288 <div class="doc_text">
289 <p>New features include:</p>
292 <li>A major change to the <tt>Use</tt> class landed, which shrank it by 25%. Since
293 this is a pervasive part of the LLVM, it ended up reducing the memory use of
294 LLVM IR in general by 15% for most programs.</li>
296 <li>Values with no names are now pretty printed by <tt>llvm-dis</tt> more
297 nicely. They now print as "<tt>%3 = add i32 %A, 4</tt>" instead of
298 "<tt>add i32 %A, 4 ; <i32>:3</tt>", which makes it much easier to read.
301 <li>LLVM 2.4 includes some changes for better vector support. First, the shift
302 operations (<tt>shl</tt>, <tt>ashr</tt>, and <tt>lshr</tt>) now all support
303 vectors and do an element-by-element shift (shifts of the whole vector can be
304 accomplished by bitcasting the vector to <tt><1 x i128></tt>, for example). Second,
305 there is initial support in development for vector comparisons with the
306 <tt><a href="LangRef.html#i_fcmp">fcmp</a>/<a href="LangRef.html#i_icmp">icmp</a></tt>
307 instructions. These instructions compare two vectors and return a vector of
308 <tt>i1</tt>'s for each result. Note that there is very little codegen support
309 available for any of these IR features though.</li>
311 <li>A new <tt>DebugInfoBuilder</tt> class is available, which makes it much
312 easier for front-ends to create debug info descriptors, similar to the way that
313 <tt>IRBuilder</tt> makes it easier to create LLVM IR.</li>
315 <li>The <tt>IRBuilder</tt> class is now parameterized by a class responsible
316 for constant folding. The default <tt>ConstantFolder</tt> class does target independent
317 constant folding. The <tt>NoFolder</tt> class does no constant folding at all, which is
318 useful when learning how LLVM works. The <tt>TargetFolder</tt> class folds the most,
319 doing target dependent constant folding.</li>
321 <li>LLVM now supports "function attributes", which allow us to separate return
322 value attributes from function attributes. LLVM now supports attributes on a
323 function itself, a return value, and its parameters. New supported function
324 attributes include <tt>noinline/alwaysinline</tt> and the <tt>opt-size</tt> flag,
325 which says the function should be optimized for code size.</li>
327 <li>LLVM IR now directly represents "common" linkage, instead of
328 representing it as a form of weak linkage.</li>
334 <!--=========================================================================-->
335 <div class="doc_subsection">
336 <a name="optimizer">Optimizer Improvements</a>
339 <div class="doc_text">
341 <p>In addition to a huge array of bug fixes and minor performance tweaks, this
342 release includes a few major enhancements and additions to the optimizers:</p>
346 <li>The Global Value Numbering (GVN) pass now does local Partial Redundancy
347 Elimination (PRE) to eliminate some partially redundant expressions in cases
348 where doing so won't grow code size.</li>
350 <li>LLVM 2.4 includes a new loop deletion pass (which removes output-free
351 provably-finite loops) and a rewritten Aggressive Dead Code Elimination (ADCE)
352 pass that no longer uses control dependence information. These changes speed up
353 the optimizer and also prevent it from deleting output-free infinite
356 <li>The new AddReadAttrs pass works out which functions are read-only or
357 read-none (these correspond to 'pure' and 'const' in GCC) and marks them
358 with the appropriate attribute.</li>
360 <li>LLVM 2.4 now includes a new SparsePropagation framework, which makes it
361 trivial to build lattice-based dataflow solvers that operate over LLVM IR. Using
362 this interface means that you just define objects to represent your lattice
363 values and the transfer functions that operate on them. It handles the
364 mechanics of worklist processing, liveness tracking, handling PHI nodes,
367 <li>The Loop Strength Reduction and induction variable optimization passes have
368 several improvements to avoid inserting MAX expressions, to optimize simple
369 floating point induction variables and to analyze trip counts of more
372 <li>Various helper functions (ComputeMaskedBits, ComputeNumSignBits, etc) were
373 pulled out of the Instruction Combining pass and put into a new
374 <tt>ValueTracking.h</tt> header, where they can be reused by other passes.</li>
376 <li>The tail duplication pass has been removed from the standard optimizer
377 sequence used by llvm-gcc. This pass still exists, but the benefits it once
378 provided are now achieved by other passes.</li>
384 <!--=========================================================================-->
385 <div class="doc_subsection">
386 <a name="codegen">Code Generator Improvements</a>
389 <div class="doc_text">
391 <p>We have put a significant amount of work into the code generator infrastructure,
392 which allows us to implement more aggressive algorithms and make it run
396 <li>The target-independent code generator supports (and the X86 backend
397 currently implements) a new interface for "fast" instruction selection. This
398 interface is optimized to produce code as quickly as possible, sacrificing
399 code quality to do it. This is used by default at -O0 or when using
400 "llc -fast" on X86. It is straight-forward to add support for
401 other targets if faster -O0 compilation is desired.</li>
403 <li>In addition to the new 'fast' instruction selection path, many existing
404 pieces of the code generator have been optimized in significant ways.
405 SelectionDAG's are now pool allocated and use better algorithms in many
406 places, the ".s" file printers now use <tt>raw_ostream</tt> to emit text much faster,
407 etc. The end result of these improvements is that the compiler also takes
408 substantially less time to generate code that is just as good (and often
409 better) than before.</li>
411 <li>Each target has been split to separate the ".s" file printing logic from the
412 rest of the target. This enables JIT compilers that don't link in the
413 (somewhat large) code and data tables used for printing a ".s" file.</li>
415 <li>The code generator now includes a "stack slot coloring" pass, which packs
416 together individual spilled values into common stack slots. This reduces
417 the size of stack frames with many spills, which tends to increase L1 cache
420 <li>Various pieces of the register allocator (e.g. the coalescer and two-address
421 operation elimination pass) now know how to rematerialize trivial operations
422 to avoid copies and include several other optimizations.</li>
424 <li>The <a href="CodeGenerator.html#selectiondag_process">graphs</a> produced by
425 the <tt>llc -view-*-dags</tt> options are now significantly prettier and
428 <li>LLVM 2.4 includes a new register allocator based on Partitioned Boolean
429 Quadratic Programming (PBQP). This register allocator is still in
430 development, but is very simple and clean.</li>
437 <!--=========================================================================-->
438 <div class="doc_subsection">
439 <a name="targetspecific">Target Specific Improvements</a>
442 <div class="doc_text">
443 <p>New target-specific features include:
447 <li>Exception handling is supported by default on Linux/x86-64.</li>
448 <li>Position Independent Code (PIC) is now supported on Linux/x86-64.</li>
449 <li><tt>@llvm.frameaddress</tt> now supports getting the frame address of stack frames
450 > 0 on x86/x86-64.</li>
451 <li>MIPS has improved a lot since last release, the most important changes
452 are: Little endian support, floating point support, allegrex core and
453 intrinsics support. O32 ABI is improved but isn't complete. The EABI
454 was implemented and is fully supported. We also have support for small
455 sections and gp_rel relocation for its access, a threshold in bytes can be
456 specified through command line.</li>
457 <li>The PowerPC backend now supports trampolines.</li>
463 <!--=========================================================================-->
464 <div class="doc_subsection">
465 <a name="otherimprovements">Other Improvements</a>
468 <div class="doc_text">
469 <p>New features include:
473 <li><tt>llvmc2</tt> (the generic compiler driver) gained plugin
474 support. It is now easier to experiment with <tt>llvmc2</tt> and
475 build your own tools based on it.</li>
477 <li>LLVM 2.4 includes a number of new generic algorithms and data structures,
478 including a scoped hash table, 'immutable' data structures, a simple
479 free-list manager, and a <tt>raw_ostream</tt> class.
480 The <tt>raw_ostream</tt> class and
481 <tt>format</tt> allow for efficient file output, and various pieces of LLVM
482 have switched over to use it. The eventual goal is to eliminate
483 use of <tt>std::ostream</tt> in favor of it.</li>
485 <li>LLVM 2.4 includes an optional build system based on CMake. It
486 still is in its early stages but can be useful for Visual C++
487 users who can not use the Visual Studio IDE.</li>
493 <!--=========================================================================-->
494 <div class="doc_subsection">
495 <a name="changes">Major Changes and Removed Features</a>
498 <div class="doc_text">
500 <p>If you're already an LLVM user or developer with out-of-tree changes based
501 on LLVM 2.3, this section lists some "gotchas" that you may run into upgrading
502 from the previous release.</p>
506 <li>The LLVM IR generated by llvm-gcc no longer names all instructions. This
507 makes it run faster, but may be more confusing to some people. If you
508 prefer to have names, the '<tt>opt -instnamer</tt>' pass will add names to
509 all instructions.</li>
511 <li>The LoadVN and GCSE passes have been removed from the tree. They are
512 obsolete and have been replaced with the GVN and MemoryDependence passes.
517 <p>In addition, many APIs have changed in this release. Some of the major LLVM
522 <li>Now, function attributes and return value attributes are managed
523 separately. Interface exported by <tt>ParameterAttributes.h</tt> header is now
524 exported by <tt>Attributes.h</tt> header. The new attributes interface changes are:
526 <li><tt>getParamAttrs</tt> method is now replaced by
527 <tt>getParamAttributes</tt>, <tt>getRetAttributes</tt> and
528 <tt>getFnAttributes</tt> methods.</li>
529 <li> Return value attributes are stored at index 0. Function attributes are
530 stored at index ~0U. Parameter attributes are stored at index that matches
531 parameter number.</li>
532 <li> <tt>ParamAttr</tt> namespace is now renamed as <tt>Attribute</tt>.</li>
533 <li> The name of the class that manages reference count of opaque
534 attributes is changed from <tt>PAListPtr</tt> to <tt>AttrListPtr</tt>.</li>
535 <li> <tt>ParamAttrsWithIndex</tt> is now renamed as <tt>AttributeWithIndex</tt>.
540 <li>The <tt>DbgStopPointInst</tt> methods <tt>getDirectory</tt> and
541 <tt>getFileName</tt> now return <tt>Value*</tt> instead of strings. These can be
542 converted to strings using <tt>llvm::GetConstantStringInfo</tt> defined via
543 "<tt>llvm/Analysis/ValueTracking.h</tt>".</li>
545 <li>The APIs to create various instructions have changed from lower case
546 "create" methods to upper case "Create" methods (e.g.
547 <tt>BinaryOperator::create</tt>). LLVM 2.4 includes both cases, but the
548 lower case ones are removed in mainline (2.5 and later), please migrate.</li>
550 <li>Various header files like "<tt>llvm/ADT/iterator</tt>" were given a ".h" suffix.
551 Change your code to #include "<tt>llvm/ADT/iterator.h</tt>" instead.</li>
553 <li>The <tt>getresult</tt> instruction has been removed and replaced with the
554 <tt>extractvalue</tt> instruction. This is part of support for first class
557 <li>In the code generator, many <tt>MachineOperand</tt> predicates were renamed to be
558 shorter (e.g. <tt>isFrameIndex()</tt> -> <tt>isFI()</tt>),
559 <tt>SDOperand</tt> was renamed to <tt>SDValue</tt> (and the "<tt>Val</tt>"
560 member was changed to be the <tt>getNode()</tt> accessor), and the
561 <tt>MVT::ValueType</tt> enum has been replaced with an "<tt>MVT</tt>"
562 struct. The <tt>getSignExtended</tt> and <tt>getValue</tt> methods in the
563 ConstantSDNode class were renamed to <tt>getSExtValue</tt> and
564 <tt>getZExtValue</tt> respectively, to be more consistent with
565 the <tt>ConstantInt</tt> class.</li>
572 <!-- *********************************************************************** -->
573 <div class="doc_section">
574 <a name="portability">Portability and Supported Platforms</a>
576 <!-- *********************************************************************** -->
578 <div class="doc_text">
580 <p>LLVM is known to work on the following platforms:</p>
583 <li>Intel and AMD machines (IA32, X86-64, AMD64, EMT-64) running Red Hat
584 Linux, Fedora Core and FreeBSD (and probably other unix-like systems).</li>
585 <li>PowerPC and X86-based Mac OS X systems, running 10.3 and above in 32-bit
586 and 64-bit modes.</li>
587 <li>Intel and AMD machines running on Win32 using MinGW libraries (native).</li>
588 <li>Intel and AMD machines running on Win32 with the Cygwin libraries (limited
589 support is available for native builds with Visual C++).</li>
590 <li>Sun UltraSPARC workstations running Solaris 10.</li>
591 <li>Alpha-based machines running Debian GNU/Linux.</li>
592 <li>Itanium-based (IA64) machines running Linux and HP-UX.</li>
595 <p>The core LLVM infrastructure uses GNU autoconf to adapt itself
596 to the machine and operating system on which it is built. However, minor
597 porting may be required to get LLVM to work on new platforms. We welcome your
598 portability patches and reports of successful builds or error messages.</p>
602 <!-- *********************************************************************** -->
603 <div class="doc_section">
604 <a name="knownproblems">Known Problems</a>
606 <!-- *********************************************************************** -->
608 <div class="doc_text">
610 <p>This section contains significant known problems with the LLVM system,
611 listed by component. If you run into a problem, please check the <a
612 href="http://llvm.org/bugs/">LLVM bug database</a> and submit a bug if
613 there isn't already one.</p>
617 <!-- ======================================================================= -->
618 <div class="doc_subsection">
619 <a name="experimental">Experimental features included with this release</a>
622 <div class="doc_text">
624 <p>The following components of this LLVM release are either untested, known to
625 be broken or unreliable, or are in early development. These components should
626 not be relied on, and bugs should not be filed against them, but they may be
627 useful to some people. In particular, if you would like to work on one of these
628 components, please contact us on the <a
629 href="http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev">LLVMdev list</a>.</p>
632 <li>The MSIL, IA64, Alpha, SPU, MIPS, and PIC16 backends are experimental.</li>
633 <li>The llc "<tt>-filetype=asm</tt>" (the default) is the only supported
634 value for this option.</li>
639 <!-- ======================================================================= -->
640 <div class="doc_subsection">
641 <a name="x86-be">Known problems with the X86 back-end</a>
644 <div class="doc_text">
647 <li>The X86 backend does not yet support
648 all <a href="http://llvm.org/PR879">inline assembly that uses the X86
649 floating point stack</a>. It supports the 'f' and 't' constraints, but not
651 <li>The X86 backend generates inefficient floating point code when configured
652 to generate code for systems that don't have SSE2.</li>
653 <li>Win64 code generation wasn't widely tested. Everything should work, but we
654 expect small issues to happen. Also, llvm-gcc cannot build mingw64 runtime
656 to <a href="http://llvm.org/PR2255">several</a>
657 <a href="http://llvm.org/PR2257">bugs</a> due to lack of support for the
658 'u' inline assembly constraint and X87 floating point inline assembly.</li>
659 <li>The X86-64 backend does not yet support the LLVM IR instruction
660 <tt>va_arg</tt>. Currently, the llvm-gcc front-end supports variadic
661 argument constructs on X86-64 by lowering them manually.</li>
666 <!-- ======================================================================= -->
667 <div class="doc_subsection">
668 <a name="ppc-be">Known problems with the PowerPC back-end</a>
671 <div class="doc_text">
674 <li>The Linux PPC32/ABI support needs testing for the interpreter and static
675 compilation, and lacks support for debug information.</li>
680 <!-- ======================================================================= -->
681 <div class="doc_subsection">
682 <a name="arm-be">Known problems with the ARM back-end</a>
685 <div class="doc_text">
688 <li>Thumb mode works only on ARMv6 or higher processors. On sub-ARMv6
689 processors, thumb programs can crash or produce wrong
690 results (<a href="http://llvm.org/PR1388">PR1388</a>).</li>
691 <li>Compilation for ARM Linux OABI (old ABI) is supported, but not fully tested.
693 <li>There is a bug in QEMU-ARM (<= 0.9.0) which causes it to incorrectly
695 programs compiled with LLVM. Please use more recent versions of QEMU.</li>
700 <!-- ======================================================================= -->
701 <div class="doc_subsection">
702 <a name="sparc-be">Known problems with the SPARC back-end</a>
705 <div class="doc_text">
708 <li>The SPARC backend only supports the 32-bit SPARC ABI (-m32), it does not
709 support the 64-bit SPARC ABI (-m64).</li>
714 <!-- ======================================================================= -->
715 <div class="doc_subsection">
716 <a name="mips-be">Known problems with the MIPS back-end</a>
719 <div class="doc_text">
722 <li>The O32 ABI is not fully supported.</li>
723 <li>64-bit MIPS targets are not supported yet.</li>
728 <!-- ======================================================================= -->
729 <div class="doc_subsection">
730 <a name="alpha-be">Known problems with the Alpha back-end</a>
733 <div class="doc_text">
737 <li>On 21164s, some rare FP arithmetic sequences which may trap do not have the
738 appropriate nops inserted to ensure restartability.</li>
743 <!-- ======================================================================= -->
744 <div class="doc_subsection">
745 <a name="ia64-be">Known problems with the IA64 back-end</a>
748 <div class="doc_text">
751 <li>The Itanium backend is highly experimental, and has a number of known
752 issues. We are looking for a maintainer for the Itanium backend. If you
753 are interested, please contact the LLVMdev mailing list.</li>
758 <!-- ======================================================================= -->
759 <div class="doc_subsection">
760 <a name="c-be">Known problems with the C back-end</a>
763 <div class="doc_text">
766 <li><a href="http://llvm.org/PR802">The C backend has only basic support for
767 inline assembly code</a>.</li>
768 <li><a href="http://llvm.org/PR1658">The C backend violates the ABI of common
769 C++ programs</a>, preventing intermixing between C++ compiled by the CBE and
770 C++ code compiled with llc or native compilers.</li>
771 <li>The C backend does not support all exception handling constructs.</li>
777 <!-- ======================================================================= -->
778 <div class="doc_subsection">
779 <a name="c-fe">Known problems with the llvm-gcc C front-end</a>
782 <div class="doc_text">
784 <p>llvm-gcc does not currently support <a href="http://llvm.org/PR869">Link-Time
785 Optimization</a> on most platforms "out-of-the-box". Please inquire on the
786 LLVMdev mailing list if you are interested.</p>
788 <p>The only major language feature of GCC not supported by llvm-gcc is
789 the <tt>__builtin_apply</tt> family of builtins. However, some extensions
790 are only supported on some targets. For example, trampolines are only
791 supported on some targets (these are used when you take the address of a
792 nested function).</p>
794 <p>If you run into GCC extensions which are not supported, please let us know.
799 <!-- ======================================================================= -->
800 <div class="doc_subsection">
801 <a name="c++-fe">Known problems with the llvm-gcc C++ front-end</a>
804 <div class="doc_text">
806 <p>The C++ front-end is considered to be fully
807 tested and works for a number of non-trivial programs, including LLVM
808 itself, Qt, Mozilla, etc.</p>
811 <li>Exception handling works well on the X86 and PowerPC targets. Currently
812 only Linux and Darwin targets are supported (both 32 and 64 bit).</li>
817 <!-- ======================================================================= -->
818 <div class="doc_subsection">
819 <a name="fortran-fe">Known problems with the llvm-gcc Fortran front-end</a>
822 <div class="doc_text">
824 <li>Fortran support generally works, but there are still several unresolved bugs
825 in Bugzilla. Please see the tools/gfortran component for details.</li>
827 <li>The Fortran front-end currently does not build on Darwin (without tweaks)
828 due to unresolved dependencies on the C front-end.</li>
832 <!-- ======================================================================= -->
833 <div class="doc_subsection">
834 <a name="ada-fe">Known problems with the llvm-gcc Ada front-end</a>
837 <div class="doc_text">
838 The llvm-gcc 4.2 Ada compiler works fairly well, however this is not a mature
839 technology and problems should be expected.
841 <li>The Ada front-end currently only builds on X86-32. This is mainly due
842 to lack of trampoline support (pointers to nested functions) on other platforms,
843 however it <a href="http://llvm.org/PR2006">also fails to build on X86-64</a>
844 which does support trampolines.</li>
845 <li>The Ada front-end <a href="http://llvm.org/PR2007">fails to bootstrap</a>.
846 Workaround: configure with --disable-bootstrap.</li>
847 <li>The c380004, <a href="http://llvm.org/PR2010">c393010</a>
848 and <a href="http://llvm.org/PR2421">cxg2021</a> ACATS tests fail
849 (c380004 also fails with gcc-4.2 mainline).</li>
850 <li>Some gcc specific Ada tests continue to crash the compiler.</li>
851 <li>The -E binder option (exception backtraces)
852 <a href="http://llvm.org/PR1982">does not work</a> and will result in programs
853 crashing if an exception is raised. Workaround: do not use -E.</li>
854 <li>Only discrete types <a href="http://llvm.org/PR1981">are allowed to start
855 or finish at a non-byte offset</a> in a record. Workaround: do not pack records
856 or use representation clauses that result in a field of a non-discrete type
857 starting or finishing in the middle of a byte.</li>
858 <li>The <tt>lli</tt> interpreter <a href="http://llvm.org/PR2009">considers
859 'main' as generated by the Ada binder to be invalid</a>.
860 Workaround: hand edit the file to use pointers for <tt>argv</tt> and
861 <tt>envp</tt> rather than integers.</li>
862 <li>The <tt>-fstack-check</tt> option <a href="http://llvm.org/PR2008">is
867 <!-- *********************************************************************** -->
868 <div class="doc_section">
869 <a name="additionalinfo">Additional Information</a>
871 <!-- *********************************************************************** -->
873 <div class="doc_text">
875 <p>A wide variety of additional information is available on the <a
876 href="http://llvm.org">LLVM web page</a>, in particular in the <a
877 href="http://llvm.org/docs/">documentation</a> section. The web page also
878 contains versions of the API documentation which is up-to-date with the
879 Subversion version of the source code.
880 You can access versions of these documents specific to this release by going
881 into the "<tt>llvm/doc/</tt>" directory in the LLVM tree.</p>
883 <p>If you have any questions or comments about LLVM, please feel free to contact
884 us via the <a href="http://llvm.org/docs/#maillist"> mailing
889 <!-- *********************************************************************** -->
893 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
894 src="http://jigsaw.w3.org/css-validator/images/vcss-blue" alt="Valid CSS"></a>
895 <a href="http://validator.w3.org/check/referer"><img
896 src="http://www.w3.org/Icons/valid-html401-blue" alt="Valid HTML 4.01"></a>
898 <a href="http://llvm.org/">LLVM Compiler Infrastructure</a><br>
899 Last modified: $Date$