2 <!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01//EN"
3 "http://www.w3.org/TR/html4/strict.dtd">
6 <meta http-equiv="Content-Type" content="text/html; charset=utf-8">
7 <link rel="stylesheet" href="llvm.css" type="text/css">
8 <title>LLVM 2.3 Release Notes</title>
12 <div class="doc_title">LLVM 2.3 Release Notes</div>
15 <li><a href="#intro">Introduction</a></li>
16 <li><a href="#changes">Major Changes and Sub-project Status</a></li>
17 <li><a href="#whatsnew">What's New?</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>
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 <!-- Done through Week-of-Mon-20080324.txt -->
30 <!-- *********************************************************************** -->
31 <div class="doc_section">
32 <a name="intro">Introduction</a>
34 <!-- *********************************************************************** -->
36 <div class="doc_text">
38 <p>This document contains the release notes for the LLVM compiler
39 infrastructure, release 2.3. Here we describe the status of LLVM, including
40 major improvements from the previous release and any known problems. All LLVM
41 releases may be downloaded from the <a href="http://llvm.org/releases/">LLVM
42 releases web site</a>.</p>
44 <p>For more information about LLVM, including information about the latest
45 release, please check out the <a href="http://llvm.org/">main LLVM
46 web site</a>. If you have questions or comments, the <a
47 href="http://mail.cs.uiuc.edu/mailman/listinfo/llvmdev">LLVM developer's mailing
48 list</a> is a good place to send them.</p>
50 <p>Note that if you are reading this file from a Subversion checkout or the
51 main LLVM web page, this document applies to the <i>next</i> release, not the
52 current one. To see the release notes for a specific releases, please see the
53 <a href="http://llvm.org/releases/">releases page</a>.</p>
57 <!-- *********************************************************************** -->
58 <div class="doc_section">
59 <a name="changes">Major Changes and Sub-project Status</a>
61 <!-- *********************************************************************** -->
63 <div class="doc_text">
65 <p>This is the fourteenth public release of the LLVM Compiler Infrastructure.
66 It includes a large number of features and refinements from LLVM 2.2.</p>
70 <!-- Unfinished features in 2.3:
76 <!--=========================================================================-->
77 <div class="doc_subsection">
78 <a name="changes">Major Changes in LLVM 2.3</a>
81 <div class="doc_text">
83 <p>LLVM 2.3 no longer supports llvm-gcc 4.0, it has been replaced with
86 <p>LLVM 2.3 no longer includes the <tt>llvm-upgrade</tt> tool. It was useful
87 for upgrading LLVM 1.9 files to LLVM 2.x syntax, but you can always use a
88 previous LLVM release to do this. One nice impact of this is that the LLVM
89 regression test suite no longer depends on llvm-upgrade, which makes it run
92 <p>The <tt>llvm2cpp</tt> tool has been folded into llc, use
93 <tt>llc -march=cpp</tt> instead of <tt>llvm2cpp</tt>.</p>
95 <p>LLVM API Changes:</p>
98 <li>Several core LLVM IR classes have migrated to use the
99 '<tt>FOOCLASS::Create(...)</tt>' pattern instead of '<tt>new
100 FOOCLASS(...)</tt>' (e.g. where FOOCLASS=<tt>BasicBlock</tt>). We hope to
101 standardize on <tt>FOOCLASS::Create</tt> for all IR classes in the future,
102 but not all of them have been moved over yet.</li>
103 <li>LLVM 2.3 renames the LLVMBuilder and LLVMFoldingBuilder classes to
104 <a href="http://llvm.org/doxygen/classllvm_1_1IRBuilder.html">IRBuilder</a>.
106 <li>MRegisterInfo was renamed to
107 <a href="http://llvm.org/doxygen/classllvm_1_1TargetRegisterInfo.html">
108 TargetRegisterInfo</a>.</li>
109 <li>The MappedFile class is gone, please use
110 <a href="http://llvm.org/doxygen/classllvm_1_1MemoryBuffer.html">
111 MemoryBuffer</a> instead.</li>
112 <li>The '<tt>-enable-eh</tt>' flag to llc has been removed. Now code should
113 encode whether it is safe to omit unwind information for a function by
114 tagging the Function object with the '<tt>nounwind</tt>' attribute.</li>
115 <li>The ConstantFP::get method that uses APFloat now takes one argument
116 instead of two. The type argument has been removed, and the type is
117 now inferred from the size of the given APFloat value.</li>
122 <!--=========================================================================-->
123 <div class="doc_subsection">
124 <a name="otherprojects">Other LLVM Sub-Projects</a>
127 <div class="doc_text">
129 The core LLVM 2.3 distribution currently consists of code from the core LLVM
130 repository (which roughly contains the LLVM optimizer, code generators and
131 supporting tools) and the llvm-gcc repository. In addition to this code, the
132 LLVM Project includes other sub-projects that are in development. The two which
133 are the most actively developed are the new <a href="#vmkit">vmkit Project</a>
134 and the <a href="#clang">Clang Project</a>.
138 <!--=========================================================================-->
139 <div class="doc_subsubsection">
140 <a name="vmkit">vmkit</a>
143 <div class="doc_text">
145 The "vmkit" project is a new addition to the LLVM family. It is an
146 implementation of 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>The JVM, called JnJVM, executes real-world applications such as Apache
150 projects (e.g. Felix and Tomcat) and the SpecJVM98 benchmark. It uses the GNU
151 Classpath project for the base classes. The CLI implementation, called N3, is
152 its in early stages but can execute simple applications and the "pnetmark"
153 benchmark. It uses the pnetlib project as its core library.</p>
155 <p>The 'vmkit' VMs compare in performance with industrial and top open-source
156 VMs on scientific applications. Besides the JIT, the VMs use many features of
157 the LLVM framework, including the standard set of optimizations, atomic
158 operations, custom function provider and memory manager for JITed methods, and
159 specific virtual machine optimizations. vmkit is not an official part of LLVM
160 2.3 release. It is publicly available under the LLVM license and can be
164 <div class="doc_code">
165 <pre>svn co http://llvm.org/svn/llvm-project/vmkit/trunk vmkit</pre>
170 <!--=========================================================================-->
171 <div class="doc_subsubsection">
172 <a name="clang">Clang</a>
175 <div class="doc_text">
177 <p>The <a href="http://clang.llvm.org/">Clang project</a> is an effort to build
178 a set of new 'LLVM native' front-end technologies for the LLVM optimizer
179 and code generator. Clang is continuing to make major strides forward in all
180 areas. Its C and Objective-C parsing support is very solid, and the code
181 generation support is far enough along to build many C applications. While not
182 yet production quality, it is progressing very nicely. In addition, C++
183 front-end work has started to make significant progress.</p>
185 <p>At this point, Clang is most useful if you are interested in source-to-source
186 transformations (such as refactoring) and other source-level tools for C and
187 Objective-C. Clang now also includes tools for turning C code into pretty HTML,
188 and includes a new <a href="http://clang.llvm.org/StaticAnalysis.html">static
189 analysis tool</a> in development. This tool focuses on automatically finding
190 bugs in C and Objective-C code.</p>
195 <!-- *********************************************************************** -->
196 <div class="doc_section">
197 <a name="whatsnew">What's New?</a>
199 <!-- *********************************************************************** -->
201 <div class="doc_text">
203 <p>LLVM 2.3 includes a huge number of bug fixes, performance tweaks and minor
204 improvements. Some of the major improvements and new features are listed in
209 <!--=========================================================================-->
210 <div class="doc_subsection">
211 <a name="majorfeatures">Major New Features</a>
214 <div class="doc_text">
216 <p>LLVM 2.3 includes several major new capabilities:</p>
219 <li><p>The biggest change in LLVM 2.3 is Multiple Return Value (MRV) support.
220 MRVs allow LLVM IR to directly represent functions that return multiple
221 values without having to pass them "by reference" in the LLVM IR. This
222 allows a front-end to generate more efficient code, as MRVs are generally
223 returned in registers if a target supports them. See the <a
224 href="LangRef.html#i_getresult">LLVM IR Reference</a> for more details.</p>
226 <p>MRVs are fully supported in the LLVM IR, but are not yet fully supported in
227 on all targets. However, it is generally safe to return up to 2 values from
228 a function: most targets should be able to handle at least that. MRV
229 support is a critical requirement for X86-64 ABI support, as X86-64 requires
230 the ability to return multiple registers from functions, and we use MRVs to
231 accomplish this in a direct way.</p></li>
233 <li><p>LLVM 2.3 includes a complete reimplementation of the "<tt>llvmc</tt>"
234 tool. It is designed to overcome several problems with the original
235 <tt>llvmc</tt> and to provide a superset of the features of the
236 '<tt>gcc</tt>' driver.</p>
238 <p>The main features of <tt>llvmc2</tt> are:
240 <li>Extended handling of command line options and smart rules for
241 dispatching them to different tools.</li>
242 <li>Flexible (and extensible) rules for defining different tools.</li>
243 <li>The different intermediate steps performed by tools are represented
244 as edges in the abstract graph.</li>
245 <li>The 'language' for driver behavior definition is tablegen and thus
246 it's relatively easy to add new features.</li>
247 <li>The definition of driver is transformed into set of C++ classes, thus
248 no runtime interpretation is needed.</li>
252 <li><p>LLVM 2.3 includes a completely rewritten interface for <a
253 href="LinkTimeOptimization.html">Link Time Optimization</a>. This interface
254 is written in C, which allows for easier integration with C code bases, and
255 incorporates improvements we learned about from the first incarnation of the
258 <li><p>The <a href="tutorial/LangImpl1.html">Kaleidoscope tutorial</a> now
259 includes a "port" of the tutorial that <a
260 href="tutorial/OCamlLangImpl1.html">uses the Ocaml bindings</a> to implement
261 the Kaleidoscope language.</p></li>
268 <!--=========================================================================-->
269 <div class="doc_subsection">
270 <a name="llvm-gcc">llvm-gcc 4.2 Improvements</a>
273 <div class="doc_text">
275 <p>LLVM 2.3 fully supports the llvm-gcc 4.2 front-end, and includes support
276 for the C, C++, Objective-C, Ada, and Fortran front-ends.</p>
280 <li>llvm-gcc 4.2 includes numerous fixes to better support the Objective-C
281 front-end. Objective-C now works very well on Mac OS/X.</li>
283 <li>Fortran <tt>EQUIVALENCE</tt>s are now supported by the gfortran
286 <li>llvm-gcc 4.2 includes many other fixes which improve conformance with the
287 relevant parts of the GCC testsuite.</li>
294 <!--=========================================================================-->
295 <div class="doc_subsection">
296 <a name="coreimprovements">LLVM Core Improvements</a>
299 <div class="doc_text">
300 <p>New features include:
304 <li>LLVM IR now directly represents "common" linkage, instead of representing it
305 as a form of weak linkage.</li>
307 <li>LLVM IR now has support for atomic operations, and this functionality can be
308 accessed through the llvm-gcc "<tt>__sync_synchronize</tt>",
309 "<tt>__sync_val_compare_and_swap</tt>", and related builtins. Support for
310 atomics are available in the Alpha, X86, X86-64, and PowerPC backends.</li>
312 <li>The C and Ocaml bindings have extended to cover pass managers, several
313 transformation passes, iteration over the LLVM IR, target data, and parameter
314 attribute lists.</li>
319 <!--=========================================================================-->
320 <div class="doc_subsection">
321 <a name="optimizer">Optimizer Improvements</a>
324 <div class="doc_text">
326 <p>In addition to a huge array of bug fixes and minor performance tweaks, the
327 LLVM 2.3 optimizers support a few major enhancements:</p>
331 <li><p>Loop index set splitting on by default.
332 This transformation hoists conditions from loop bodies and reduces a loop's
333 iteration space to improve performance. For example,</p>
335 <div class="doc_code">
337 for (i = LB; i < UB; ++i)
343 <p>is transformed into:</p>
345 <p><div class="doc_code">
348 for (i = LB; i < NUB; ++i)
355 <li>LLVM now includes a new <tt>memcpy</tt> optimization pass which removes
356 dead <tt>memcpy</tt> calls, unneeded copies of aggregates, and performs
357 return slot optimization. The LLVM optimizer now notices long sequences of
358 consecutive stores and merges them into <tt>memcpy</tt>'s where profitable.</li>
360 <li>Alignment detection for vector memory references and for <tt>memcpy</tt> and
361 <tt>memset</tt> is now more aggressive.</li>
363 <li>The Aggressive Dead Code Elimination (ADCE) optimization has been rewritten
364 to make it both faster and safer in the presence of code containing infinite
365 loops. Some of its prior functionality has been factored out into the loop
366 deletion pass, which <em>is</em> safe for infinite loops. The new ADCE pass is
367 no longer based on control dependence, making it run faster.</li>
369 <li>The 'SimplifyLibCalls' pass, which optimizes calls to libc and libm
370 functions for C-based languages, has been rewritten to be a FunctionPass
371 instead a ModulePass. This allows it to be run more often and to be
372 included at -O1 in llvm-gcc. It was also extended to include more
373 optimizations and several corner case bugs were fixed.</li>
375 <li>LLVM now includes a simple 'Jump Threading' pass, which attempts to simplify
376 conditional branches using information about predecessor blocks, simplifying
377 the control flow graph. This pass is pretty basic at this point, but
378 catches some important cases and provides a foundation to build on.</li>
380 <li>Several corner case bugs which could lead to deleting volatile memory
381 accesses have been fixed.</li>
383 <li>Several optimizations have been sped up, leading to faster code generation
384 with the same code quality.</li>
390 <!--=========================================================================-->
391 <div class="doc_subsection">
392 <a name="codegen">Code Generator Improvements</a>
395 <div class="doc_text">
397 <p>We put a significant amount of work into the code generator infrastructure,
398 which allows us to implement more aggressive algorithms and make it run
402 <li>The code generator now has support for carrying information about memory
403 references throughout the entire code generation process, via the
404 <a href="http://llvm.org/doxygen/classllvm_1_1MachineMemOperand.html">
405 MachineMemOperand</a> class. In the future this will be used to improve
406 both pre-pass and post-pass scheduling, and to improve compiler-debugging
409 <li>The target-independent code generator infrastructure now uses LLVM's
410 <a href="http://llvm.org/doxygen/classllvm_1_1APInt.html">APInt</a>
411 class to handle integer values, which allows it to support integer types
412 larger than 64 bits (for example i128). Note that support for such types is
413 also dependent on target-specific support. Use of APInt is also a step
414 toward support for non-power-of-2 integer sizes.</li>
416 <li>LLVM 2.3 includes several compile time speedups for code with large basic
417 blocks, particularly in the instruction selection phase, register
418 allocation, scheduling, and tail merging/jump threading.</li>
420 <li>LLVM 2.3 includes several improvements which make llc's
421 <tt>--view-sunit-dags</tt> visualization of scheduling dependency graphs
422 easier to understand.</li>
424 <li>The code generator allows targets to write patterns that generate subreg
425 references directly in .td files now.</li>
427 <li><tt>memcpy</tt> lowering in the backend is more aggressive, particularly for
428 <tt>memcpy</tt> calls introduced by the code generator when handling
429 pass-by-value structure argument copies.</li>
431 <li>Inline assembly with multiple register results now returns those results
432 directly in the appropriate registers, rather than going through memory.
433 Inline assembly that uses constraints like "ir" with immediates now use the
434 'i' form when possible instead of always loading the value in a register.
435 This saves an instruction and reduces register use.</li>
437 <li>Added support for PIC/GOT style <a
438 href="CodeGenerator.html#tailcallopt">tail calls</a> on X86/32 and initial
439 support for tail calls on PowerPC 32 (it may also work on PowerPC 64 but is
440 not thoroughly tested).</li>
446 <!--=========================================================================-->
447 <div class="doc_subsection">
448 <a name="x86specific">X86/X86-64 Specific Improvements</a>
451 <div class="doc_text">
452 <p>New target-specific features include:
456 <li>llvm-gcc's X86-64 ABI conformance is far improved, particularly in the
457 area of passing and returning structures by value. llvm-gcc compiled code
458 now interoperates very well on X86-64 systems with other compilers.</li>
460 <li>Support for Win64 was added. This includes code generation itself, JIT
461 support, and necessary changes to llvm-gcc.</li>
463 <li>The LLVM X86 backend now supports the support SSE 4.1 instruction set, and
464 the llvm-gcc 4.2 front-end supports the SSE 4.1 compiler builtins. Various
465 generic vector operations (insert/extract/shuffle) are much more efficient
466 when SSE 4.1 is enabled. The JIT automatically takes advantage of these
467 instructions, but llvm-gcc must be explicitly told to use them, e.g. with
468 <tt>-march=penryn</tt>.</li>
470 <li>The X86 backend now does a number of optimizations that aim to avoid
471 converting numbers back and forth from SSE registers to the X87 floating
472 point stack. This is important because most X86 ABIs require return values
473 to be on the X87 Floating Point stack, but most CPUs prefer computation in
476 <li>The X86 backend supports stack realignment, which is particularly useful for
477 vector code on OS's without 16-byte aligned stacks, such as Linux and
480 <li>The X86 backend now supports the "sseregparm" options in GCC, which allow
481 functions to be tagged as passing floating point values in SSE
484 <li>Trampolines (taking the address of a nested function) now work on
487 <li><tt>__builtin_prefetch</tt> is now compiled into the appropriate prefetch
488 instructions instead of being ignored.</li>
490 <li>128-bit integers are now supported on X86-64 targets. This can be used
491 through <tt>__attribute__((TImode))</tt> in llvm-gcc.</li>
493 <li>The register allocator can now rematerialize PIC-base computations, which is
494 an important optimization for register use.</li>
496 <li>The "t" and "f" inline assembly constraints for the X87 floating point stack
497 now work. However, the "u" constraint is still not fully supported.</li>
503 <!--=========================================================================-->
504 <div class="doc_subsection">
505 <a name="targetspecific">Other Target Specific Improvements</a>
508 <div class="doc_text">
509 <p>New target-specific features include:
513 <li>The LLVM C backend now supports vector code.</li>
514 <li>The Cell SPU backend includes a number of improvements. It generates better
515 code and its stability/completeness is improving.</li>
522 <!--=========================================================================-->
523 <div class="doc_subsection">
524 <a name="otherimprovements">Other Improvements</a>
527 <div class="doc_text">
528 <p>New features include:
532 <li>LLVM now builds with GCC 4.3.</li>
533 <li>Bugpoint now supports running custom scripts (with the <tt>-run-custom</tt>
534 option) to determine how to execute the command and whether it is making
535 forward process.</li>
540 <!-- *********************************************************************** -->
541 <div class="doc_section">
542 <a name="portability">Portability and Supported Platforms</a>
544 <!-- *********************************************************************** -->
546 <div class="doc_text">
548 <p>LLVM is known to work on the following platforms:</p>
551 <li>Intel and AMD machines (IA32) running Red Hat Linux, Fedora Core and FreeBSD
552 (and probably other unix-like systems).</li>
553 <li>PowerPC and X86-based Mac OS X systems, running 10.3 and above in 32-bit and
555 <li>Intel and AMD machines running on Win32 using MinGW libraries (native).</li>
556 <li>Intel and AMD machines running on Win32 with the Cygwin libraries (limited
557 support is available for native builds with Visual C++).</li>
558 <li>Sun UltraSPARC workstations running Solaris 10.</li>
559 <li>Alpha-based machines running Debian GNU/Linux.</li>
560 <li>Itanium-based (IA64) machines running Linux and HP-UX.</li>
563 <p>The core LLVM infrastructure uses GNU autoconf to adapt itself
564 to the machine and operating system on which it is built. However, minor
565 porting may be required to get LLVM to work on new platforms. We welcome your
566 portability patches and reports of successful builds or error messages.</p>
570 <!-- *********************************************************************** -->
571 <div class="doc_section">
572 <a name="knownproblems">Known Problems</a>
574 <!-- *********************************************************************** -->
576 <div class="doc_text">
578 <p>This section contains all known problems with the LLVM system, listed by
579 component. As new problems are discovered, they will be added to these
580 sections. If you run into a problem, please check the <a
581 href="http://llvm.org/bugs/">LLVM bug database</a> and submit a bug if
582 there isn't already one.</p>
586 <!-- ======================================================================= -->
587 <div class="doc_subsection">
588 <a name="experimental">Experimental features included with this release</a>
591 <div class="doc_text">
593 <p>The following components of this LLVM release are either untested, known to
594 be broken or unreliable, or are in early development. These components should
595 not be relied on, and bugs should not be filed against them, but they may be
596 useful to some people. In particular, if you would like to work on one of these
597 components, please contact us on the <a
598 href="http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev">LLVMdev list</a>.</p>
601 <li>The MSIL, IA64, Alpha, SPU, and MIPS backends are experimental.</li>
602 <li>The llc "<tt>-filetype=asm</tt>" (the default) is the only supported
603 value for this option.</li>
608 <!-- ======================================================================= -->
609 <div class="doc_subsection">
610 <a name="x86-be">Known problems with the X86 back-end</a>
613 <div class="doc_text">
616 <li>The X86 backend does not yet support
617 all <a href="http://llvm.org/PR879">inline assembly that uses the X86
618 floating point stack</a>. It supports the 'f' and 't' constraints, but not
620 <li>The X86 backend generates inefficient floating point code when configured
621 to generate code for systems that don't have SSE2.</li>
622 <li>Win64 code generation wasn't widely tested. Everything should work, but we
623 expect small issues to happen. Also, llvm-gcc cannot build mingw64 runtime
625 to <a href="http://llvm.org/PR2255">several</a>
626 <a href="http://llvm.org/PR2257">bugs</a> due to lack of support for the
627 'u' inline assembly constraint and X87 floating point inline assembly.</li>
628 <li>The X86-64 backend does not yet support position-independent code (PIC)
629 generation on Linux targets.</li>
630 <li>The X86-64 backend does not yet support the LLVM IR instruction
631 <tt>va_arg</tt>. Currently, the llvm-gcc front-end supports variadic
632 argument constructs on X86-64 by lowering them manually.</li>
637 <!-- ======================================================================= -->
638 <div class="doc_subsection">
639 <a name="ppc-be">Known problems with the PowerPC back-end</a>
642 <div class="doc_text">
645 <li>The Linux PPC32/ABI support needs testing for the interpreter and static
646 compilation, and lacks support for debug information.</li>
651 <!-- ======================================================================= -->
652 <div class="doc_subsection">
653 <a name="arm-be">Known problems with the ARM back-end</a>
656 <div class="doc_text">
659 <li>Thumb mode works only on ARMv6 or higher processors. On sub-ARMv6
660 processors, thumb programs can crash or produce wrong
661 results (<a href="http://llvm.org/PR1388">PR1388</a>).</li>
662 <li>Compilation for ARM Linux OABI (old ABI) is supported, but not fully tested.
664 <li>There is a bug in QEMU-ARM (<= 0.9.0) which causes it to incorrectly
666 programs compiled with LLVM. Please use more recent versions of QEMU.</li>
671 <!-- ======================================================================= -->
672 <div class="doc_subsection">
673 <a name="sparc-be">Known problems with the SPARC back-end</a>
676 <div class="doc_text">
679 <li>The SPARC backend only supports the 32-bit SPARC ABI (-m32), it does not
680 support the 64-bit SPARC ABI (-m64).</li>
685 <!-- ======================================================================= -->
686 <div class="doc_subsection">
687 <a name="alpha-be">Known problems with the Alpha back-end</a>
690 <div class="doc_text">
694 <li>On 21164s, some rare FP arithmetic sequences which may trap do not have the
695 appropriate nops inserted to ensure restartability.</li>
700 <!-- ======================================================================= -->
701 <div class="doc_subsection">
702 <a name="ia64-be">Known problems with the IA64 back-end</a>
705 <div class="doc_text">
708 <li>The Itanium backend is highly experimental, and has a number of known
709 issues. We are looking for a maintainer for the Itanium backend. If you
710 are interested, please contact the llvmdev mailing list.</li>
715 <!-- ======================================================================= -->
716 <div class="doc_subsection">
717 <a name="c-be">Known problems with the C back-end</a>
720 <div class="doc_text">
723 <li><a href="http://llvm.org/PR802">The C backend has only basic support for
724 inline assembly code</a>.</li>
725 <li><a href="http://llvm.org/PR1658">The C backend violates the ABI of common
726 C++ programs</a>, preventing intermixing between C++ compiled by the CBE and
727 C++ code compiled with llc or native compilers.</li>
728 <li>The C backend does not support all exception handling constructs.</li>
734 <!-- ======================================================================= -->
735 <div class="doc_subsection">
736 <a name="c-fe">Known problems with the llvm-gcc C front-end</a>
739 <div class="doc_text">
741 <p>llvm-gcc does not currently support <a href="http://llvm.org/PR869">Link-Time
742 Optimization</a> on most platforms "out-of-the-box". Please inquire on the
743 llvmdev mailing list if you are interested.</p>
745 <p>The only major language feature of GCC not supported by llvm-gcc is
746 the <tt>__builtin_apply</tt> family of builtins. However, some extensions
747 are only supported on some targets. For example, trampolines are only
748 supported on some targets (these are used when you take the address of a
749 nested function).</p>
751 <p>If you run into GCC extensions which are not supported, please let us know.
756 <!-- ======================================================================= -->
757 <div class="doc_subsection">
758 <a name="c++-fe">Known problems with the llvm-gcc C++ front-end</a>
761 <div class="doc_text">
763 <p>The C++ front-end is considered to be fully
764 tested and works for a number of non-trivial programs, including LLVM
765 itself, Qt, Mozilla, etc.</p>
768 <li>Exception handling works well on the X86 and PowerPC targets, including
769 X86-64 darwin. This works when linking to a libstdc++ compiled by GCC. It is
770 supported on X86-64 linux, but that is disabled by default in this release.</li>
776 <!-- ======================================================================= -->
777 <div class="doc_subsection">
778 <a name="ada-fe">Known problems with the llvm-gcc Ada front-end</a>
781 <div class="doc_text">
782 The llvm-gcc 4.2 Ada compiler works fairly well, however this is not a mature
783 technology and problems should be expected.
785 <li>The Ada front-end currently only builds on X86-32. This is mainly due
786 to lack of trampoline support (pointers to nested functions) on other platforms,
787 however it <a href="http://llvm.org/PR2006">also fails to build on X86-64</a>
788 which does support trampolines.</li>
789 <li>The Ada front-end <a href="http://llvm.org/PR2007">fails to bootstrap</a>.
790 Workaround: configure with --disable-bootstrap.</li>
791 <li>The c380004 and <a href="http://llvm.org/PR2010">c393010</a> ACATS tests
792 fail (c380004 also fails with gcc-4.2 mainline). When built at -O3, the
793 <a href="http://llvm.org/PR2421">cxg2021</a> ACATS test also fails.</li>
794 <li>Some gcc specific Ada tests continue to crash the compiler. The testsuite
795 reports most tests as having failed even though they pass.</li>
796 <li>The -E binder option (exception backtraces)
797 <a href="http://llvm.org/PR1982">does not work</a> and will result in programs
798 crashing if an exception is raised. Workaround: do not use -E.</li>
799 <li>Only discrete types <a href="http://llvm.org/PR1981">are allowed to start
800 or finish at a non-byte offset</a> in a record. Workaround: do not pack records
801 or use representation clauses that result in a field of a non-discrete type
802 starting or finishing in the middle of a byte.</li>
803 <li>The <tt>lli</tt> interpreter <a href="http://llvm.org/PR2009">considers
804 'main' as generated by the Ada binder to be invalid</a>.
805 Workaround: hand edit the file to use pointers for <tt>argv</tt> and
806 <tt>envp</tt> rather than integers.</li>
807 <li>The <tt>-fstack-check</tt> option <a href="http://llvm.org/PR2008">is
812 <!-- *********************************************************************** -->
813 <div class="doc_section">
814 <a name="additionalinfo">Additional Information</a>
816 <!-- *********************************************************************** -->
818 <div class="doc_text">
820 <p>A wide variety of additional information is available on the <a
821 href="http://llvm.org">LLVM web page</a>, in particular in the <a
822 href="http://llvm.org/docs/">documentation</a> section. The web page also
823 contains versions of the API documentation which is up-to-date with the
824 Subversion version of the source code.
825 You can access versions of these documents specific to this release by going
826 into the "<tt>llvm/doc/</tt>" directory in the LLVM tree.</p>
828 <p>If you have any questions or comments about LLVM, please feel free to contact
829 us via the <a href="http://llvm.org/docs/#maillist"> mailing
834 <!-- *********************************************************************** -->
838 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
839 src="http://jigsaw.w3.org/css-validator/images/vcss" alt="Valid CSS!"></a>
840 <a href="http://validator.w3.org/check/referer"><img
841 src="http://www.w3.org/Icons/valid-html401" alt="Valid HTML 4.01!"></a>
843 <a href="http://llvm.org/">LLVM Compiler Infrastructure</a><br>
844 Last modified: $Date$