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>Codegen progress/state [DANIEL]</p>
105 <!--=========================================================================-->
106 <div class="doc_subsection">
107 <a name="clangsa">Clang Static Analyzer</a>
110 <div class="doc_text">
112 <p>The Clang project also includes an early stage static source code analysis
113 tool for <a href="http://clang.llvm.org/StaticAnalysis.html">automatically
114 finding bugs</a> in C and Objective-C programs. The tool performs a growing set
115 of checks to find bugs that occur on a specific path within a program. Examples
116 of bugs the tool finds include logic errors such as null dereferences,
117 violations of various API rules, dead code, and potential memory leaks in
118 Objective-C programs. Since its inception, public feedback on the tool has been
119 extremely positive, and conservative estimates put the number of real bugs it
120 has found in industrial-quality software on the order of thousands.</p>
122 <p>The tool also provides a simple web GUI to inspect potential bugs found by
123 the tool. While still early in development, the GUI illustrates some of the key
124 features of Clang: accurate source location information, which is used by the
125 GUI to highlight specific code expressions that relate to a bug (including those
126 that span multiple lines) and built-in knowledge of macros, which is used to
127 perform inline expansion of macros within the GUI itself.</p>
129 <p>The set of checks performed by the static analyzer is gradually expanding,
131 future plans for the tool include full source-level inter-procedural analysis
132 and deeper checks such as buffer overrun detection. There are many opportunities
133 to extend and enhance the static analyzer, and anyone interested in working on
134 this project is encouraged to get involved!</p>
138 <!--=========================================================================-->
139 <div class="doc_subsection">
140 <a name="vmkit">VMKit: JVM/CLI Virtual Machine Implementation</a>
143 <div class="doc_text">
145 The <a href="http://vmkit.llvm.org/">VMKit project</a> is an implementation of
146 a JVM and a CLI Virtual Machines (Microsoft .NET is an
147 implementation of the CLI) using the Just-In-Time compiler of LLVM.</p>
149 <p>Following LLVM 2.4, VMKit has its first release 0.24 that you can find on
150 the <a href="http://llvm.org/releases/">release page</a>. The release includes
151 bug fixes, cleanup and new features. The major changes include:</p>
155 <li><p>Support for generics in the .Net virtual machine. This was implemented
156 by Tilmann Scheller during his Google Summer of Code project.</li></p>
157 <li><p>Initial support for the Mono class libraries.</p></li>
158 <li><p>Support for MacOSX/x86, following LLVM's support for exceptions in
160 <li><p>A new vmkit driver: a program to run java or .net applications. The
161 driver supports llvm command line arguments including the new "-fast" option.
163 <li><p>A new memory allocation scheme in the JVM that makes unloading a
164 class loader very fast.</p></li>
165 <li><p>VMKit now follows the LLVM Makefile machinery.</p></li>
172 <!-- *********************************************************************** -->
173 <div class="doc_section">
174 <a name="whatsnew">What's New in LLVM?</a>
176 <!-- *********************************************************************** -->
178 <div class="doc_text">
180 <p>This release includes a huge number of bug fixes, performance tweaks and
181 minor improvements. Some of the major improvements and new features are listed
186 <!--=========================================================================-->
187 <div class="doc_subsection">
188 <a name="majorfeatures">Major New Features</a>
191 <div class="doc_text">
193 <p>LLVM 2.4 includes several major new capabilities:</p>
196 <li><p>The most visible end-user change in LLVM 2.4 is that it includes many
197 optimizations and changes to make -O0 compile times much faster. You should see
198 improvements on the order of 30% (or more) faster than LLVM 2.3. There are many
199 pieces to this change, described in more detail below. The speedups and new
200 components can also be used for JIT compilers that want fast compilation as
203 <li><p>The biggest change to the LLVM IR is that Multiple Return Values (which
204 were introduced in LLVM 2.3) have been generalized to full support for "First
205 Class Aggregate" values in LLVM 2.4. This means that LLVM IR supports using
206 structs and arrays as values in a function. This capability is mostly useful
207 for front-end authors, who prefer to treat things like complex numbers, simple
208 tuples, dope vectors, etc as Value*'s instead of as a tuple of Value*'s or as
209 memory values. Bitcode files from LLVM 2.3 will automatically migrate to the
210 general representation.</p></li>
212 <li><p>LLVM 2.4 also includes an initial port for the PIC16 microprocessor. This
213 is the LLVM target that only has support for 8 bit registers, and a number of
214 other crazy constraints. While the port is still in early development stages,
215 it shows some interesting things you can do with LLVM.</p></li>
222 <!--=========================================================================-->
223 <div class="doc_subsection">
224 <a name="llvm-gcc">llvm-gcc 4.2 Improvements</a>
227 <div class="doc_text">
229 <p>LLVM fully supports the llvm-gcc 4.2 front-end, which marries the GCC
230 front-ends and driver with the LLVM optimizer and code generator. It currently
231 includes support for the C, C++, Objective-C, Ada, and Fortran front-ends.</p>
234 <li>LLVM 2.4 supports the full set of atomic <tt>__sync_*</tt> builtins. LLVM
235 2.3 only supported those used by OpenMP, but 2.4 supports them all. While
236 llvm-gcc supports all of these builtins, note that not all targets do. X86
237 support them all in both 32-bit and 64-bit mode and PowerPC supports them all
238 except for the 64-bit operations when in 32-bit mode.</li>
240 <li>llvm-gcc now supports an <tt>-flimited-precision</tt> option, which tells
241 the compiler that it is ok to use low-precision approximations of certain libm
242 functions (like tan, log, etc). This allows you to get high performance if you
243 only need (say) 14-bits of precision.</li>
245 <li>llvm-gcc now supports a C language extension known as "<a
246 href="http://lists.cs.uiuc.edu/pipermail/cfe-dev/2008-August/002670.html">Blocks
247 </a>". This feature is similar to nested functions and closures, but does not
248 require stack trampolines (with most ABIs) and supports returning closures
249 from functions that define them. Note that actually <em>using</em> Blocks
250 requires a small runtime that is not included with llvm-gcc.</li>
252 <li>llvm-gcc now supports a new <tt>-flto</tt> option. On systems that support
253 transparent Link Time Optimization (currently Darwin systems with Xcode 3.1 and
254 later) this allows the use of LTO with other optimization levels like -Os.
255 Previously, LTO could only be used with -O4, which implied optimizations in
256 -O3 that can increase code size.</li>
262 <!--=========================================================================-->
263 <div class="doc_subsection">
264 <a name="coreimprovements">LLVM Core Improvements</a>
267 <div class="doc_text">
268 <p>New features include:
272 <li>A major change to the <tt>Use</tt> class landed, which shrank it by 25%. Since
273 this is a pervasive part of the LLVM, it ended up reducing the memory use of
274 LLVM IR in general by 15% for most programs.</li>
276 <li>Values with no names are now pretty printed by <tt>llvm-dis</tt> more
277 nicely. They now print as "<tt>%3 = add i32 %A, 4</tt>" instead of
278 "<tt>add i32 %A, 4 ; <i32>:3</tt>", which makes it much easier to read.
281 <li>LLVM 2.4 includes some changes for better vector support. First, the shift
282 operations (<tt>shl</tt>, <tt>ashr</tt>, <tt>lshr</tt>) now all support vectors
283 and do an element-by-element shift (shifts of the whole vector can be
284 accomplished by bitcasting the vector to <1 x i128> for example). Second,
285 there is initial support in development for vector comparisons with the
286 <a href="LangRef.html#i_fcmp">fcmp</a>/<a href="LangRef.html#i_icmp">icmp</a>
287 instructions. These instructions compare two vectors and return a vector of
288 i1's for each result. Note that there is very little codegen support available
289 for any of these IR features though.</li>
291 <li>A new <tt>DebugInfoBuilder</tt> class is available, which makes it much
292 easier for front-ends to create debug info descriptors, similar to the way that
293 <tt>IRBuilder</tt> makes it easier to create LLVM IR.</li>
295 <li>The <tt>IRBuilder</tt> class is now parameterized by a class responsible
296 for constant folding. The default <tt>ConstantFolder</tt> class does target independent
297 constant folding. The <tt>NoFolder</tt> class does no constant folding at all, which is
298 useful when learning how LLVM works. The <tt>TargetFolder</tt> class folds the most,
299 doing target dependent constant folding.</li>
301 <li>LLVM now supports "function attributes", which allows us to separate return
302 value attributes from function attributes. LLVM now supports attributes on a
303 function itself, a return value, and its parameters. New supported function
304 attributes include noinline/alwaysinline and the "opt-size" flag which says the
305 function should be optimized for code size.</li>
307 <li>LLVM IR now directly represents "common" linkage, instead of
308 representing it as a form of weak linkage.</li>
314 <!--=========================================================================-->
315 <div class="doc_subsection">
316 <a name="optimizer">Optimizer Improvements</a>
319 <div class="doc_text">
321 <p>In addition to a huge array of bug fixes and minor performance tweaks, this
322 release includes a few major enhancements and additions to the optimizers:</p>
326 <li>The Global Value Numbering (GVN) pass now does local Partial Redundancy
327 Elimination (PRE) to eliminate some partially redundant expressions in cases
328 where doing so won't grow code size.</li>
330 <li>LLVM 2.4 includes a new loop deletion pass (which removes output-free
331 provably-finite loops) and a rewritten Aggressive Dead Code Elimination (ADCE)
332 pass that no longer uses control dependence information. These changes speed up
333 the optimizer and also prevent it from deleting output-free infinite
336 <li>The new AddReadAttrs pass works out which functions are read-only or
337 read-none (these correspond to 'pure' and 'const' in GCC) and marks them
338 with the appropriate attribute.</li>
340 <li>LLVM 2.4 now includes a new SparsePropagation framework, which makes it
341 trivial to build lattice-based dataflow solvers that operate over LLVM IR. Using
342 this interface means that you just define objects to represent your lattice
343 values and the transfer functions that operate on them. It handles the
344 mechanics of worklist processing, liveness tracking, handling PHI nodes,
347 <li>The Loop Strength Reduction and induction variable optimization passes have
348 several improvements to avoid inserting MAX expressions, to optimize simple
349 floating point induction variables and to analyze trip counts of more
352 <li>Various helper functions (ComputeMaskedBits, ComputeNumSignBits, etc) were
353 pulled out of the Instruction Combining pass and put into a new
354 <tt>ValueTracking.h</tt> header, where they can be reused by other passes.</li>
356 <li>The tail duplication pass has been removed from the standard optimizer
357 sequence used by llvm-gcc. This pass still exists, but the benefits it once
358 provided are now achieved by other passes.</li>
364 <!--=========================================================================-->
365 <div class="doc_subsection">
366 <a name="codegen">Code Generator Improvements</a>
369 <div class="doc_text">
371 <p>We have put a significant amount of work into the code generator infrastructure,
372 which allows us to implement more aggressive algorithms and make it run
376 <li>The target-independent code generator supports (and the X86 backend
377 currently implements) a new interface for "fast" instruction selection. This
378 interface is optimized to produce code as quickly as possible, sacrificing
379 code quality to do it. This is used by default at -O0 or when using
380 "llc -fast" on X86. It is straight-forward to add support for
381 other targets if faster -O0 compilation is desired.</li>
383 <li>In addition to the new 'fast' instruction selection path, many existing
384 pieces of the code generator have been optimized in significant ways.
385 SelectionDAG's are now pool allocated and use better algorithms in many
386 places, the ".s" file printers now use <tt>raw_ostream</tt> to emit text much faster,
387 etc. The end result of these improvements is that the compiler also takes
388 substantially less time to generate code that is just as good (and often
389 better) than before.</li>
391 <li>Each target has been split to separate the ".s" file printing logic from the
392 rest of the target. This enables JIT compilers that don't link in the
393 (somewhat large) code and data tables used for printing a ".s" file.</li>
395 <li>The code generator now includes a "stack slot coloring" pass, which packs
396 together individual spilled values into common stack slots. This reduces
397 the size of stack frames with many spills, which tends to increase L1 cache
400 <li>Various pieces of the register allocator (e.g. the coalescer and two-address
401 operation elimination pass) now know how to rematerialize trivial operations
402 to avoid copies and include several other optimizations.</li>
404 <li>The <a href="CodeGenerator.html#selectiondag_process">graphs</a> produced by
405 the <tt>llc -view-*-dags</tt> options are now significantly prettier and
408 <li>LLVM 2.4 includes a new register allocator based on Partitioned Boolean
409 Quadratic Programming (PBQP). This register allocator is still in
410 development, but is very simple and clean.</li>
417 <!--=========================================================================-->
418 <div class="doc_subsection">
419 <a name="targetspecific">Target Specific Improvements</a>
422 <div class="doc_text">
423 <p>New target-specific features include:
427 <li>Exception handling is supported by default on Linux/x86-64.</li>
428 <li>Position Independent Code (PIC) is now supported on Linux/x86-64.</li>
429 <li>@llvm.frameaddress now supports getting the frame address of stack frames
430 > 0 on x86/x86-64.</li>
431 <li>MIPS floating point support? [BRUNO]</li>
432 <li>The PowerPC backend now supports trampolines.</li>
438 <!--=========================================================================-->
439 <div class="doc_subsection">
440 <a name="otherimprovements">Other Improvements</a>
443 <div class="doc_text">
444 <p>New features include:
448 <li><tt>llvmc2</tt> (the generic compiler driver) gained plugin
449 support. It is now easier to experiment with <tt>llvmc2</tt> and
450 build your own tools based on it.</li>
452 <li>LLVM 2.4 includes a number of new generic algorithms and data structures,
453 include a scoped hash table, 'immutable' data structures, a simple
454 free-list manager, and a <tt>raw_ostream</tt> class.
455 The <tt>raw_ostream</tt> class and
456 <tt>format</tt> allow for efficient file output, and various pieces of LLVM
457 have switched over to use it. The eventual goal is to eliminate
458 std::ostream in favor of it.</li>
464 <!--=========================================================================-->
465 <div class="doc_subsection">
466 <a name="changes">Major Changes and Removed Features</a>
469 <div class="doc_text">
471 <p>If you're already an LLVM user or developer with out-of-tree changes based
472 on LLVM 2.3, this section lists some "gotchas" that you may run into upgrading
473 from the previous release.</p>
477 <li>The LLVM IR generated by llvm-gcc no longer names all instructions. This
478 makes it run faster, but may be more confusing to some people. If you
479 prefer to have names, the '<tt>opt -instnamer</tt>' pass will add names to
480 all instructions.</li>
482 <li>The LoadVN and GCSE passes have been removed from the tree. They are
483 obsolete and have been replaced with the GVN and MemoryDependence passes.
488 <p>In addition, many APIs have changed in this release. Some of the major LLVM
493 <li>Now, function attributes and return value attributes are managed
494 sepearately. Interface exported by <tt>ParameterAttributes.h</tt> header is now
495 experted by <tt>Attributes.h</tt> header. The new attributes interface changes are:
497 <li><tt>getParamAttrs</tt> method is now replaced by
498 <tt>getParamAttributes</tt>, <tt>getRetAttributes</tt> and
499 <tt>getFnAttributes</tt> methods.</li>
500 <li> Return value attributes are stored at index 0. Function attributes are
501 stored at index ~0U. Parameter attributes are stored at index that matches
502 parameter number.</li>
503 <li> <tt>ParamAttr</tt> namespace is now renamed as <tt>Attribute</tt>.</li>
504 <li> The name of the class</tt> that manages reference count of opaque
505 attributes is changed from <tt>PAListPtr</tt> to <tt>AttrListPtr</tt>.</li>
506 <li> <tt>ParamAttrsWithIndex</tt> is now renamed as <tt>AttributeWithIndex</tt>.
511 <li>The <tt>DbgStopPointInst</tt> methods <tt>getDirectory</tt> and
512 <tt>getFileName</tt> now return <tt>Value*</tt> instead of strings. These can be
513 converted to strings using <tt>llvm::GetConstantStringInfo</tt> defined via
514 "<tt>llvm/Analysis/ValueTracking.h</tt>".</li>
516 <li>The APIs to create various instructions have changed from lower case
517 "create" methods to upper case "Create" methods (e.g.
518 <tt>BinaryOperator::create</tt>). LLVM 2.4 includes both cases, but the
519 lower case ones are removed in mainline, please migrate.</li>
521 <li>Various header files like "<tt>llvm/ADT/iterator</tt>" were given a ".h" suffix.
522 Change your code to #include "<tt>llvm/ADT/iterator.h</tt>" instead.</li>
524 <li>In the code generator, many <tt>MachineOperand</tt> predicates were renamed to be
525 shorter (e.g. <tt>isFrameIndex()</tt> -> <tt>isFI()</tt>),
526 <tt>SDOperand</tt> was renamed to <tt>SDValue</tt> (and the "<tt>Val</tt>"
527 member was changed to be the <tt>getNode()</tt> accessor), and the
528 <tt>MVT::ValueType</tt> enum has been replaced with an "<tt>MVT</tt>"
529 struct. The <tt>getSignExtended</tt> and <tt>getValue</tt> methods in the
530 ConstantSDNode class were renamed to <tt>getSExtValue</tt> and
531 <tt>getZExtValue</tt> respectively, to be more consistent with
532 the <tt>ConstantInt</tt> class.</li>
539 <!-- *********************************************************************** -->
540 <div class="doc_section">
541 <a name="portability">Portability and Supported Platforms</a>
543 <!-- *********************************************************************** -->
545 <div class="doc_text">
547 <p>LLVM is known to work on the following platforms:</p>
550 <li>Intel and AMD machines (IA32) running Red Hat Linux, Fedora Core and FreeBSD
551 (and probably other unix-like systems).</li>
552 <li>PowerPC and X86-based Mac OS X systems, running 10.3 and above in 32-bit and
554 <li>Intel and AMD machines running on Win32 using MinGW libraries (native).</li>
555 <li>Intel and AMD machines running on Win32 with the Cygwin libraries (limited
556 support is available for native builds with Visual C++).</li>
557 <li>Sun UltraSPARC workstations running Solaris 10.</li>
558 <li>Alpha-based machines running Debian GNU/Linux.</li>
559 <li>Itanium-based (IA64) machines running Linux and HP-UX.</li>
562 <p>The core LLVM infrastructure uses GNU autoconf to adapt itself
563 to the machine and operating system on which it is built. However, minor
564 porting may be required to get LLVM to work on new platforms. We welcome your
565 portability patches and reports of successful builds or error messages.</p>
569 <!-- *********************************************************************** -->
570 <div class="doc_section">
571 <a name="knownproblems">Known Problems</a>
573 <!-- *********************************************************************** -->
575 <div class="doc_text">
577 <p>This section contains all known problems with the LLVM system, listed by
578 component. As new problems are discovered, they will be added to these
579 sections. If you run into a problem, please check the <a
580 href="http://llvm.org/bugs/">LLVM bug database</a> and submit a bug if
581 there isn't already one.</p>
585 <!-- ======================================================================= -->
586 <div class="doc_subsection">
587 <a name="experimental">Experimental features included with this release</a>
590 <div class="doc_text">
592 <p>The following components of this LLVM release are either untested, known to
593 be broken or unreliable, or are in early development. These components should
594 not be relied on, and bugs should not be filed against them, but they may be
595 useful to some people. In particular, if you would like to work on one of these
596 components, please contact us on the <a
597 href="http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev">LLVMdev list</a>.</p>
600 <li>The MSIL, IA64, Alpha, SPU, MIPS, and PIC16 backends are experimental.</li>
601 <li>The llc "<tt>-filetype=asm</tt>" (the default) is the only supported
602 value for this option.</li>
607 <!-- ======================================================================= -->
608 <div class="doc_subsection">
609 <a name="x86-be">Known problems with the X86 back-end</a>
612 <div class="doc_text">
615 <li>The X86 backend does not yet support
616 all <a href="http://llvm.org/PR879">inline assembly that uses the X86
617 floating point stack</a>. It supports the 'f' and 't' constraints, but not
619 <li>The X86 backend generates inefficient floating point code when configured
620 to generate code for systems that don't have SSE2.</li>
621 <li>Win64 code generation wasn't widely tested. Everything should work, but we
622 expect small issues to happen. Also, llvm-gcc cannot build mingw64 runtime
624 to <a href="http://llvm.org/PR2255">several</a>
625 <a href="http://llvm.org/PR2257">bugs</a> due to lack of support for the
626 'u' inline assembly constraint and X87 floating point inline assembly.</li>
627 <li>The X86-64 backend does not yet support the LLVM IR instruction
628 <tt>va_arg</tt>. Currently, the llvm-gcc front-end supports variadic
629 argument constructs on X86-64 by lowering them manually.</li>
634 <!-- ======================================================================= -->
635 <div class="doc_subsection">
636 <a name="ppc-be">Known problems with the PowerPC back-end</a>
639 <div class="doc_text">
642 <li>The Linux PPC32/ABI support needs testing for the interpreter and static
643 compilation, and lacks support for debug information.</li>
648 <!-- ======================================================================= -->
649 <div class="doc_subsection">
650 <a name="arm-be">Known problems with the ARM back-end</a>
653 <div class="doc_text">
656 <li>Thumb mode works only on ARMv6 or higher processors. On sub-ARMv6
657 processors, thumb programs can crash or produce wrong
658 results (<a href="http://llvm.org/PR1388">PR1388</a>).</li>
659 <li>Compilation for ARM Linux OABI (old ABI) is supported, but not fully tested.
661 <li>There is a bug in QEMU-ARM (<= 0.9.0) which causes it to incorrectly
663 programs compiled with LLVM. Please use more recent versions of QEMU.</li>
668 <!-- ======================================================================= -->
669 <div class="doc_subsection">
670 <a name="sparc-be">Known problems with the SPARC back-end</a>
673 <div class="doc_text">
676 <li>The SPARC backend only supports the 32-bit SPARC ABI (-m32), it does not
677 support the 64-bit SPARC ABI (-m64).</li>
682 <!-- ======================================================================= -->
683 <div class="doc_subsection">
684 <a name="alpha-be">Known problems with the Alpha back-end</a>
687 <div class="doc_text">
691 <li>On 21164s, some rare FP arithmetic sequences which may trap do not have the
692 appropriate nops inserted to ensure restartability.</li>
697 <!-- ======================================================================= -->
698 <div class="doc_subsection">
699 <a name="ia64-be">Known problems with the IA64 back-end</a>
702 <div class="doc_text">
705 <li>The Itanium backend is highly experimental, and has a number of known
706 issues. We are looking for a maintainer for the Itanium backend. If you
707 are interested, please contact the llvmdev mailing list.</li>
712 <!-- ======================================================================= -->
713 <div class="doc_subsection">
714 <a name="c-be">Known problems with the C back-end</a>
717 <div class="doc_text">
720 <li><a href="http://llvm.org/PR802">The C backend has only basic support for
721 inline assembly code</a>.</li>
722 <li><a href="http://llvm.org/PR1658">The C backend violates the ABI of common
723 C++ programs</a>, preventing intermixing between C++ compiled by the CBE and
724 C++ code compiled with llc or native compilers.</li>
725 <li>The C backend does not support all exception handling constructs.</li>
731 <!-- ======================================================================= -->
732 <div class="doc_subsection">
733 <a name="c-fe">Known problems with the llvm-gcc C front-end</a>
736 <div class="doc_text">
738 <p>llvm-gcc does not currently support <a href="http://llvm.org/PR869">Link-Time
739 Optimization</a> on most platforms "out-of-the-box". Please inquire on the
740 llvmdev mailing list if you are interested.</p>
742 <p>The only major language feature of GCC not supported by llvm-gcc is
743 the <tt>__builtin_apply</tt> family of builtins. However, some extensions
744 are only supported on some targets. For example, trampolines are only
745 supported on some targets (these are used when you take the address of a
746 nested function).</p>
748 <p>If you run into GCC extensions which are not supported, please let us know.
753 <!-- ======================================================================= -->
754 <div class="doc_subsection">
755 <a name="c++-fe">Known problems with the llvm-gcc C++ front-end</a>
758 <div class="doc_text">
760 <p>The C++ front-end is considered to be fully
761 tested and works for a number of non-trivial programs, including LLVM
762 itself, Qt, Mozilla, etc.</p>
765 <li>Exception handling works well on the X86 and PowerPC targets. Currently
766 only linux and darwin targets are supported (both 32 and 64 bit).</li>
772 <!-- ======================================================================= -->
773 <div class="doc_subsection">
774 <a name="ada-fe">Known problems with the llvm-gcc Ada front-end</a>
777 <div class="doc_text">
778 The llvm-gcc 4.2 Ada compiler works fairly well, however this is not a mature
779 technology and problems should be expected.
781 <li>The Ada front-end currently only builds on X86-32. This is mainly due
782 to lack of trampoline support (pointers to nested functions) on other platforms,
783 however it <a href="http://llvm.org/PR2006">also fails to build on X86-64</a>
784 which does support trampolines.</li>
785 <li>The Ada front-end <a href="http://llvm.org/PR2007">fails to bootstrap</a>.
786 Workaround: configure with --disable-bootstrap.</li>
787 <li>The c380004, <a href="http://llvm.org/PR2010">c393010</a>
788 and <a href="http://llvm.org/PR2421">cxg2021</a> ACATS tests fail
789 (c380004 also fails with gcc-4.2 mainline).</li>
790 <li>Some gcc specific Ada tests continue to crash the compiler.</li>
791 <li>The -E binder option (exception backtraces)
792 <a href="http://llvm.org/PR1982">does not work</a> and will result in programs
793 crashing if an exception is raised. Workaround: do not use -E.</li>
794 <li>Only discrete types <a href="http://llvm.org/PR1981">are allowed to start
795 or finish at a non-byte offset</a> in a record. Workaround: do not pack records
796 or use representation clauses that result in a field of a non-discrete type
797 starting or finishing in the middle of a byte.</li>
798 <li>The <tt>lli</tt> interpreter <a href="http://llvm.org/PR2009">considers
799 'main' as generated by the Ada binder to be invalid</a>.
800 Workaround: hand edit the file to use pointers for <tt>argv</tt> and
801 <tt>envp</tt> rather than integers.</li>
802 <li>The <tt>-fstack-check</tt> option <a href="http://llvm.org/PR2008">is
807 <!-- *********************************************************************** -->
808 <div class="doc_section">
809 <a name="additionalinfo">Additional Information</a>
811 <!-- *********************************************************************** -->
813 <div class="doc_text">
815 <p>A wide variety of additional information is available on the <a
816 href="http://llvm.org">LLVM web page</a>, in particular in the <a
817 href="http://llvm.org/docs/">documentation</a> section. The web page also
818 contains versions of the API documentation which is up-to-date with the
819 Subversion version of the source code.
820 You can access versions of these documents specific to this release by going
821 into the "<tt>llvm/doc/</tt>" directory in the LLVM tree.</p>
823 <p>If you have any questions or comments about LLVM, please feel free to contact
824 us via the <a href="http://llvm.org/docs/#maillist"> mailing
829 <!-- *********************************************************************** -->
833 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
834 src="http://jigsaw.w3.org/css-validator/images/vcss" alt="Valid CSS!"></a>
835 <a href="http://validator.w3.org/check/referer"><img
836 src="http://www.w3.org/Icons/valid-html401" alt="Valid HTML 4.01!"></a>
838 <a href="http://llvm.org/">LLVM Compiler Infrastructure</a><br>
839 Last modified: $Date$