Gitiles
Code Review Sign In
gerrit-public.fairphone.software / fp2-dev / platform / external / llvm / e83ae23878a3f11e00169d747b17addd090bd657 / . / docs / ReleaseNotes-2.6.html
blob: 5a5a279778397da2a1d739dc3a807546deae07d6 [file] [log] [blame]
Duncan Sandse2e4a892009-06-24 08:38:48 +0000[diff] [blame]1<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01//EN"
2 "http://www.w3.org/TR/html4/strict.dtd">
3<html>
4<head>
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 2.6 Release Notes</title>
8</head>
9<body>
10
11<div class="doc_title">LLVM 2.6 Release Notes</div>
12
13<ol>
14 <li><a href="#intro">Introduction</a></li>
15 <li><a href="#subproj">Sub-project Status Update</a></li>
16 <li><a href="#externalproj">External Projects Using LLVM 2.6</a></li>
17 <li><a href="#whatsnew">What's New in LLVM 2.6?</a></li>
18 <li><a href="GettingStarted.html">Installation Instructions</a></li>
19 <li><a href="#portability">Portability and Supported Platforms</a></li>
20 <li><a href="#knownproblems">Known Problems</a></li>
21 <li><a href="#additionalinfo">Additional Information</a></li>
22</ol>
23
24<div class="doc_author">
25 <p>Written by the <a href="http://llvm.org">LLVM Team</a></p>
26</div>
27
28<!-- *********************************************************************** -->
29<div class="doc_section">
30 <a name="intro">Introduction</a>
31</div>
32<!-- *********************************************************************** -->
33
34<div class="doc_text">
35
36<p>This document contains the release notes for the LLVM Compiler
37Infrastructure, release 2.6. Here we describe the status of LLVM, including
38major improvements from the previous release and significant known problems.
39All LLVM releases may be downloaded from the <a
40href="http://llvm.org/releases/">LLVM releases web site</a>.</p>
41
42<p>For more information about LLVM, including information about the latest
43release, please check out the <a href="http://llvm.org/">main LLVM
44web site</a>. If you have questions or comments, the <a
45href="http://mail.cs.uiuc.edu/mailman/listinfo/llvmdev">LLVM Developer's Mailing
46List</a> is a good place to send them.</p>
47
48<p>Note that if you are reading this file from a Subversion checkout or the
49main LLVM web page, this document applies to the <i>next</i> release, not the
50current one. To see the release notes for a specific release, please see the
51<a href="http://llvm.org/releases/">releases page</a>.</p>
52
53</div>
Chris Lattner80ed2552009-10-08 07:01:46 +0000[diff] [blame]54
Chris Lattnerc758fec2009-10-05 02:12:39 +0000[diff] [blame]55
Chris Lattner7b0a6812009-10-08 06:27:53 +0000[diff] [blame]56<!--
57Almost dead code.
Chris Lattner80ed2552009-10-08 07:01:46 +0000[diff] [blame]58 include/llvm/Analysis/LiveValues.h => Dan
59 lib/Transforms/IPO/MergeFunctions.cpp => consider for 2.8.
60 llvm/Analysis/PointerTracking.h => Edwin wants this, consider for 2.8.
Chris Lattner7b0a6812009-10-08 06:27:53 +0000[diff] [blame]61-->
Chris Lattnerc758fec2009-10-05 02:12:39 +0000[diff] [blame]62
Chris Lattnerc758fec2009-10-05 02:12:39 +0000[diff] [blame]63
64<!-- Unfinished features in 2.6:
Chris Lattner80ed2552009-10-08 07:01:46 +0000[diff] [blame]65 gcc plugin.
Chris Lattnerc758fec2009-10-05 02:12:39 +0000[diff] [blame]66 strong phi elim
67 variable debug info for optimized code
Duncan Sandse2e4a892009-06-24 08:38:48 +0000[diff] [blame]68 postalloc scheduler: anti dependence breaking, hazard recognizer?
Chris Lattnerc758fec2009-10-05 02:12:39 +0000[diff] [blame]69 metadata
70 loop dependence analysis
Chris Lattner80ed2552009-10-08 07:01:46 +0000[diff] [blame]71 ELF Writer? How stable?
72 <li>PostRA scheduler improvements, ARM adoption (David Goodwin).</li>
Chris Lattner45178502009-10-09 06:24:25 +0000[diff] [blame]73 2.7 supports the GDB 7.0 jit interfaces for debug info.
Duncan Sandse2e4a892009-06-24 08:38:48 +0000[diff] [blame]74 -->
75
76 <!-- for announcement email:
Chris Lattnerc758fec2009-10-05 02:12:39 +0000[diff] [blame]77 Logo web page.
78 llvm devmtg
79 compiler_rt
Chris Lattner7b0a6812009-10-08 06:27:53 +0000[diff] [blame]80 klee web page at klee.llvm.org
81 Many new papers added to /pubs/
82 Mention gcc plugin.
83
Duncan Sandse2e4a892009-06-24 08:38:48 +0000[diff] [blame]84 -->
85
86<!-- *********************************************************************** -->
87<div class="doc_section">
88 <a name="subproj">Sub-project Status Update</a>
89</div>
90<!-- *********************************************************************** -->
91
92<div class="doc_text">
93<p>
94The LLVM 2.6 distribution currently consists of code from the core LLVM
95repository &mdash;which roughly includes the LLVM optimizers, code generators
96and supporting tools &mdash; and the llvm-gcc repository. In addition to this
97code, the LLVM Project includes other sub-projects that are in development. The
98two which are the most actively developed are the <a href="#clang">Clang
99Project</a> and the <a href="#vmkit">VMKit Project</a>.
100</p>
101
102</div>
103
104
105<!--=========================================================================-->
106<div class="doc_subsection">
107<a name="clang">Clang: C/C++/Objective-C Frontend Toolkit</a>
108</div>
109
110<div class="doc_text">
111
Chris Lattner28d02742009-10-09 05:01:15 +0000[diff] [blame]112<p>The <a href="http://clang.llvm.org/">Clang project</a> is an effort to build
113a set of new 'LLVM native' front-end technologies for the C family of languages.
114LLVM 2.6 is the first release to officially include Clang, and it provides a
115production quality C and Objective-C compiler. If you are interested in fast
116compiles and good diagnostics, we encourage you to try it out.</p>
117
118<p>In addition to supporting these languages, C++ support is also <a
119href="http://clang.llvm.org/cxx_status.html">well under way</a>, and mainline
120Clang is able to parse the libstdc++ 4.2 headers and even codegen simple apps.
121If you are interested in Clang C++ support or any other Clang feature, we
122strongly encourage you to get involved on the <a
Duncan Sandse2e4a892009-06-24 08:38:48 +0000[diff] [blame]123href="http://lists.cs.uiuc.edu/mailman/listinfo/cfe-dev">Clang front-end mailing
124list</a>.</p>
125
126<p>In the LLVM 2.6 time-frame, the Clang team has made many improvements:</p>
127
128<ul>
Chris Lattner28d02742009-10-09 05:01:15 +0000[diff] [blame]129<li>C and Objective-C support are now considered production quality.</li>
Chris Lattnerc32a5322009-09-30 06:27:22 +0000[diff] [blame]130<li>AuroraUX / FreeBSD &amp; OpenBSD Toolchain support.</li>
Chris Lattner28d02742009-10-09 05:01:15 +0000[diff] [blame]131<li>Most of Objective-C 2.0 is now supported with the GNU runtime.</li>
Duncan Sandse2e4a892009-06-24 08:38:48 +0000[diff] [blame]132<li>Many many bugs are fixed and many features have been added.</li>
133</ul>
134</div>
135
136<!--=========================================================================-->
137<div class="doc_subsection">
138<a name="clangsa">Clang Static Analyzer</a>
139</div>
140
141<div class="doc_text">
142
Chris Lattner80ed2552009-10-08 07:01:46 +0000[diff] [blame]143<p><b>UPDATE!</b> Previously announced in the 2.4 and 2.5 LLVM releases, the Clang project also
Duncan Sandse2e4a892009-06-24 08:38:48 +0000[diff] [blame]144includes an early stage static source code analysis tool for <a
145href="http://clang.llvm.org/StaticAnalysis.html">automatically finding bugs</a>
146in C and Objective-C programs. The tool performs a growing set of checks to find
147bugs that occur on a specific path within a program.</p>
148
149<p>In the LLVM 2.6 time-frame there have been many significant improvements to
150XYZ.</p>
151
152<p>The set of checks performed by the static analyzer continues to expand, and
153future plans for the tool include full source-level inter-procedural analysis
154and deeper checks such as buffer overrun detection. There are many opportunities
155to extend and enhance the static analyzer, and anyone interested in working on
156this project is encouraged to get involved!</p>
157
158</div>
159
160<!--=========================================================================-->
161<div class="doc_subsection">
162<a name="vmkit">VMKit: JVM/CLI Virtual Machine Implementation</a>
163</div>
164
165<div class="doc_text">
166<p>
167The <a href="http://vmkit.llvm.org/">VMKit project</a> is an implementation of
Nicolas Geoffraye83ae232009-10-09 10:13:08 +0000[diff] [blame^]168a JVM and a CLI Virtual Machine (Microsoft .NET is an
169implementation of the CLI) using LLVM for static and just-in-time compilation.</p>
Duncan Sandse2e4a892009-06-24 08:38:48 +0000[diff] [blame]170
Nicolas Geoffraye83ae232009-10-09 10:13:08 +0000[diff] [blame^]171<p>VMKit version 0.26 builds with LLVM 2.6 and you can find it on its
Duncan Sandse2e4a892009-06-24 08:38:48 +0000[diff] [blame]172<a href="http://vmkit.llvm.org/releases/">webpage</a>. The release includes
173bug fixes, cleanup and new features. The major changes are:</p>
174
175<ul>
176
Nicolas Geoffraye83ae232009-10-09 10:13:08 +0000[diff] [blame^]177<li>A new llcj tool to generate shared libraries or executables of Java files.</li>
178<li>A rewriting of the JVM C++ codebase to call LLVM garbage collection (GC) intrinsics when it manipulates GC objects. </li>
179<li>Cooperative GC and precise stack scanning using the GC framework of LLVM. </li>
180<li>Fast subtype checking (paper from Click et al [JGI'02]). </li>
181<li>Implementation of a two-word header for Java objects instead of the orginal three-word header. </li>
182<li>Better Java specification-compliance: division by zero checks, stack overflow checks, finalization and references support. </li>
Duncan Sandse2e4a892009-06-24 08:38:48 +0000[diff] [blame]183
184</ul>
185</div>
186
Chris Lattner6ee62f82009-10-09 05:55:04 +0000[diff] [blame]187
188<!--=========================================================================-->
189<div class="doc_subsection">
190<a name="compiler-rt">compiler-rt: Compiler Runtime Library</a>
191</div>
192
193<div class="doc_text">
194<p>
195The new LLVM <a href="http://compiler-rt.llvm.org/">compiler-rt project</a>
196is a simple library that provides an implementation of the low-level
197target-specific hooks required by code generation and other runtime components.
198For example, when compiling for a 32-bit target, converting a double to a 64-bit
199unsigned integer is compiling into a runtime call to the "__fixunsdfdi"
200function. The compiler-rt library provides optimized implementations of this and
201other low-level routines.</p>
202
203<p>
204All of the code in the compiler-rt project is available under the standard LLVM
205License, a "BSD-style" license.</p>
206
207</div>
208
209<!--=========================================================================-->
210<div class="doc_subsection">
211<a name="klee">klee: Symbolic Execution and Automatic Test Case Generator</a>
212</div>
213
214<div class="doc_text">
215<p>
216The new LLVM <a href="http://klee.llvm.org/">klee project</a> is a symbolic
217execution framework for programs in LLVM bitcode form. Klee tries to
218symbolically evaluate "all" paths through the application and records state
219transitions that lead to fault states. This allows it to construct testcases
220that lead to faults and can even be used to verify algorithms. For more
221details, please see the <a
222href="http://llvm.org/pubs/2008-12-OSDI-KLEE.html">OSDI 2008 paper</a> about
223Klee.</p>
224
225</div>
226
227<!--=========================================================================-->
228<div class="doc_subsection">
229<a name="dragonegg">Dragon Egg: An LLVM backend plugin for GCC</a>
230</div>
231
232<div class="doc_text">
233<p>
234<b>Duncan needs to write me</b>.
235</p>
236
237</div>
238
239
Chris Lattner80ed2552009-10-08 07:01:46 +0000[diff] [blame]240<!--=========================================================================-->
241<div class="doc_subsection">
242<a name="mc">llvm-mc: Machine Code Toolkit</a>
243</div>
244
245<div class="doc_text">
246<p>
Chris Lattner45178502009-10-09 06:24:25 +0000[diff] [blame]247The LLVM Machine Code (MC) Toolkit project is a (very early) effort to build
248better tools for dealing with machine code, object file formats, etc. The idea
249is to be able to generate most of the target specific details of assemblers and
250disassemblers from existing LLVM target .td files (with suitable enhancements),
251and to build infrastructure for reading and writing common object file formats.
252One of the first deliverables is to build a full assembler and integrate it into
253the compiler, which is predicted to substantially reduce compile time in some
254scenarios.
255</p>
Chris Lattner80ed2552009-10-08 07:01:46 +0000[diff] [blame]256
Chris Lattner45178502009-10-09 06:24:25 +0000[diff] [blame]257<p>In the LLVM 2.6 timeframe, the MC framework has grown to the point where it
258can reliably parse and pretty print (with some encoding information) a
259darwin/x86 .s file successfully, and has the very early phases of a Mach-O
260assembler in progress. Beyond the MC framework itself, major refactoring of the
261LLVM code generator has started. The idea is to make the code generator reason
262about the code it is producing in a much more semantic way, rather than a
263textual way. For example, the code generator now uses MCSection objects to
264represent section assignments, instead of text strings that print to .section
265directives.</p>
266
267<p>MC is an early and ongoing project that will hopefully continue to lead to
268many improvements in the code generator and build infrastructure useful for many
269other situations.
Chris Lattner80ed2552009-10-08 07:01:46 +0000[diff] [blame]270</p>
271
272</div>
273
274
Duncan Sandse2e4a892009-06-24 08:38:48 +0000[diff] [blame]275<!-- *********************************************************************** -->
276<div class="doc_section">
277 <a name="externalproj">External Projects Using LLVM 2.6</a>
278</div>
279<!-- *********************************************************************** -->
280
Chris Lattner80ed2552009-10-08 07:01:46 +0000[diff] [blame]281<!--=========================================================================-->
282<div class="doc_subsection">
283<a name="Rubinius">Rubinius</a>
284</div>
285
286<div class="doc_text">
287<p><a href="http://github.com/evanphx/rubinius">Rubinius</a> is an environment
288for running Ruby code which strives to write as much of the core class
289implementation in Ruby as possible. Combined with a bytecode interpreting VM, it
290uses LLVM to optimize and compile ruby code down to machine code. Techniques
291such as type feedback, method inlining, and uncommon traps are all used to
292remove dynamism from ruby execution and increase performance.</p>
Chris Lattnerd4a537b2009-10-08 16:01:33 +0000[diff] [blame]293
294<p>Since LLVM 2.5, Rubinius has made several major leaps forward, implementing
295a counter based JIT, type feedback, and speculative method inlining.
296</p>
297
Chris Lattner80ed2552009-10-08 07:01:46 +0000[diff] [blame]298</div>
Chris Lattnerc32a5322009-09-30 06:27:22 +0000[diff] [blame]299
300<!--=========================================================================-->
301<div class="doc_subsection">
302<a name="macruby">MacRuby</a>
303</div>
304
305<div class="doc_text">
306
307<p>
308<a href="http://macruby.org">MacRuby</a> is an implementation of Ruby on top of
309core Mac OS X technologies, such as the Objective-C common runtime and garbage
310collector, and the CoreFoundation framework. It is principally developed by
311Apple and aims at enabling the creation of full-fledged Mac OS X applications.
312</p>
313
314<p>
315MacRuby uses LLVM for optimization passes, JIT and AOT compilation of Ruby
316expressions. It also uses zero-cost DWARF exceptions to implement Ruby exception
317handling.</p>
318
319</div>
320
321
Duncan Sandse2e4a892009-06-24 08:38:48 +0000[diff] [blame]322<!--=========================================================================-->
323<div class="doc_subsection">
324<a name="pure">Pure</a>
325</div>
326
327<div class="doc_text">
328<p>
329<a href="http://pure-lang.googlecode.com/">Pure</a>
330is an algebraic/functional programming language based on term rewriting.
331Programs are collections of equations which are used to evaluate expressions in
332a symbolic fashion. Pure offers dynamic typing, eager and lazy evaluation,
333lexical closures, a hygienic macro system (also based on term rewriting),
334built-in list and matrix support (including list and matrix comprehensions) and
335an easy-to-use C interface. The interpreter uses LLVM as a backend to
336 JIT-compile Pure programs to fast native code.</p>
337
Chris Lattnerd4a537b2009-10-08 16:01:33 +0000[diff] [blame]338<p>Pure versions 0.31 and later have been tested and are known to work with
339LLVM 2.6 (and continue to work with older LLVM releases >= 2.3 as well).
Duncan Sandse2e4a892009-06-24 08:38:48 +0000[diff] [blame]340</p>
341</div>
342
343
344<!--=========================================================================-->
345<div class="doc_subsection">
346<a name="ldc">LLVM D Compiler</a>
347</div>
348
349<div class="doc_text">
350<p>
351<a href="http://www.dsource.org/projects/ldc">LDC</a> is an implementation of
352the D Programming Language using the LLVM optimizer and code generator.
353The LDC project works great with the LLVM 2.6 release. General improvements in
354this
355cycle have included new inline asm constraint handling, better debug info
356support, general bugfixes, and better x86-64 support. This has allowed
357some major improvements in LDC, getting us much closer to being as
358fully featured as the original DMD compiler from DigitalMars.
359</p>
360</div>
361
362<!--=========================================================================-->
363<div class="doc_subsection">
364<a name="RoadsendPHP">Roadsend PHP</a>
365</div>
366
367<div class="doc_text">
Chris Lattner80ed2552009-10-08 07:01:46 +0000[diff] [blame]368<p>
Chris Lattner80ed2552009-10-08 07:01:46 +0000[diff] [blame]369<a href="http://code.roadsend.com/rphp">Roadsend PHP</a> (rphp) is an open
Duncan Sandse2e4a892009-06-24 08:38:48 +0000[diff] [blame]370source implementation of the PHP programming
371language that uses LLVM for its optimizer, JIT, and static compiler. This is a
372reimplementation of an earlier project that is now based on LLVM.</p>
373</div>
374
Jeffrey Yasskinba2aa782009-06-24 21:09:13 +0000[diff] [blame]375<!--=========================================================================-->
376<div class="doc_subsection">
Chris Lattner6ee62f82009-10-09 05:55:04 +0000[diff] [blame]377<a name="UnladenSwallow">Unladen Swallow</a>
Jeffrey Yasskinba2aa782009-06-24 21:09:13 +0000[diff] [blame]378</div>
379
380<div class="doc_text">
Chris Lattner80ed2552009-10-08 07:01:46 +0000[diff] [blame]381<p>
Chris Lattner80ed2552009-10-08 07:01:46 +0000[diff] [blame]382<a href="http://code.google.com/p/unladen-swallow/">Unladen Swallow</a> is a
Jeffrey Yasskinba2aa782009-06-24 21:09:13 +0000[diff] [blame]383branch of <a href="http://python.org/">Python</a> intended to be fully
384compatible and significantly faster. It uses LLVM's optimization passes and JIT
385compiler.</p>
386</div>
387
Chris Lattner6ee62f82009-10-09 05:55:04 +0000[diff] [blame]388<!--=========================================================================-->
389<div class="doc_subsection">
390<a name="llvm-lua">llvm-lua</a>
391</div>
392
393<div class="doc_text">
394<p>
395<a href="http://code.google.com/p/llvm-lua/">LLVM-Lua</a> uses LLVM to add JIT
396&amp; static compiling support to the Lua VM. Lua bytecode is analyzed to
397remove type checks, then LLVM is used to compile those bytecodes down to machine
398code.</p>
399</div>
400
Jeffrey Yasskin5e98cb72009-06-24 21:26:42 +0000[diff] [blame]401
Duncan Sandse2e4a892009-06-24 08:38:48 +0000[diff] [blame]402
403<!-- *********************************************************************** -->
404<div class="doc_section">
405 <a name="whatsnew">What's New in LLVM 2.6?</a>
406</div>
407<!-- *********************************************************************** -->
408
409<div class="doc_text">
410
411<p>This release includes a huge number of bug fixes, performance tweaks, and
412minor improvements. Some of the major improvements and new features are listed
413in this section.
414</p>
Chris Lattner7b0a6812009-10-08 06:27:53 +0000[diff] [blame]415
Duncan Sandse2e4a892009-06-24 08:38:48 +0000[diff] [blame]416</div>
417
418<!--=========================================================================-->
419<div class="doc_subsection">
420<a name="majorfeatures">Major New Features</a>
421</div>
422
423<div class="doc_text">
424
425<p>LLVM 2.6 includes several major new capabilities:</p>
426
427<ul>
Chris Lattner6ee62f82009-10-09 05:55:04 +0000[diff] [blame]428<li>New <a href="#compiler-rt">compiler-rt</a>, <A href="#klee">klee</a>,
Chris Lattner45178502009-10-09 06:24:25 +0000[diff] [blame]429 and <a href="#mc">machine code toolkit</a> sub-projects.</li>
430<li>Debug information now includes line numbers when optimizations are enabled.
431 This allows statistical sampling tools like oprofile and Shark to map
432 samples back to source lines.</li>
433<li>LLVM now includes new experimental backends to support the MSP430, SystemZ,
Chris Lattner6ee62f82009-10-09 05:55:04 +0000[diff] [blame]434 and BlackFin architectures.</li>
435<li>LLVM supports a new <a href="GoldPlugin.html">Gold Linker Plugin</a> which
436 enables support for <a href="LinkTimeOptimization.html">transparent
437 link-time optimization</a> on ELF targets when used with the Gold binutils
438 linker.</li>
Chris Lattner45178502009-10-09 06:24:25 +0000[diff] [blame]439<li>LLVM now supports doing optimization and code generation on multiple threads
440 by allowing multiple "LLVMContext" objects to exist. Please see the <a
441 href="ProgrammersManual.html#threading">threading entry in the Programmer's
442 Manual</a> for more information.</li>
Chris Lattnera8f42142009-10-09 06:36:25 +0000[diff] [blame]443<li>LLVM now has experimental support for <a
444 href="http://nondot.org/~sabre/LLVMNotes/EmbeddedMetadata.txt">embedded
445 metadata</a> in LLVM IR, though the implementation is not guaranteed to be
446 final and the .bc file format may change in future releases. Debug info
447 does not yet use this format in LLVM 2.6.</p>
448
Duncan Sandse2e4a892009-06-24 08:38:48 +0000[diff] [blame]449</ul>
450
451</div>
452
Duncan Sandse2e4a892009-06-24 08:38:48 +0000[diff] [blame]453<!--=========================================================================-->
454<div class="doc_subsection">
455<a name="coreimprovements">LLVM IR and Core Improvements</a>
456</div>
457
458<div class="doc_text">
459<p>LLVM IR has several new features that are used by our existing front-ends and
460can be useful if you are writing a front-end for LLVM:</p>
461
462<ul>
Chris Lattner7b0a6812009-10-08 06:27:53 +0000[diff] [blame]463<li>Getelementpr instruction now allows any integer type for array/pointer indexes.</li>
464<li>Inbounds for GEP</li>
465<li>NSW/NUW/exact div</li>
466<li>LSR promotes int induction variables to 64-bit on 64-bit targets, major perf boost for numerical code.</li>
467<li>LSR now analyzes pointer expressions (e.g. getelementptrs), not just integers.</li>
468<li>new linkage types linkonce_odr, weak_odr, linker_private, and available_externally.</li>
469<li>New fadd, fsub, fmul instructions and classes. </li>
470<li>Target intrinsics can now return multiple results.</li>
Duncan Sandse2e4a892009-06-24 08:38:48 +0000[diff] [blame]471</ul>
472
473</div>
474
475<!--=========================================================================-->
476<div class="doc_subsection">
477<a name="optimizer">Optimizer Improvements</a>
478</div>
479
480<div class="doc_text">
481
482<p>In addition to a large array of bug fixes and minor performance tweaks, this
483release includes a few major enhancements and additions to the optimizers:</p>
484
485<ul>
486
Chris Lattner7b0a6812009-10-08 06:27:53 +0000[diff] [blame]487<li>SRoA improvements for vector unions, memset, arbitrary weird bitfield accesses etc. It now produces "strange" sized integers.</li>
488<li>Inliner reuse stack space when inlining arrays?</li>
489<li>Enabled GVN Load PRE.</li>
490<li>New Static Single Information (SSI) construction pass (not used by anything yet, experimental).</li>
491</li>
Duncan Sandse2e4a892009-06-24 08:38:48 +0000[diff] [blame]492
493</ul>
494
495</div>
496
497<!--=========================================================================-->
498<div class="doc_subsection">
499<a name="codegen">Target Independent Code Generator Improvements</a>
500</div>
501
502<div class="doc_text">
503
504<p>We have put a significant amount of work into the code generator
505infrastructure, which allows us to implement more aggressive algorithms and make
506it run faster:</p>
507
508<ul>
509
Chris Lattner7b0a6812009-10-08 06:27:53 +0000[diff] [blame]510<li> -asm-verbose now prints location info (with -g) and loop nest info.</li>
511<li>Tblgen now supports multiclass inheritance and a number of new string and
512 list operations like !(subst), !(foreach), !car, !cdr, !null, !if, !cast.
513 These make the .td files more expressive and allow more aggressive factoring
514 of duplication across instruction patterns.</li>
515<li>New MachineVerifier pass.</li>
516<li>Machine LICM, hoists things like constant pool loads, loads from readonly stubs, vector constant synthesization code, etc.</li>
517<li>Machine Sinking</li>
518<li>target-specific intrinsics (r63765)</li>
519<li>Regalloc improvements for commuting, various spiller peephole optimizations, cross-class coalescing.</li>
520<li><tt>llc -enable-value-prop</tt>, propagation of value info (sign/zero ext info) from one MBB to another</li>
521<li>Regalloc hints for allocation stuff: Evan r73381/r73671. Finished/enabled?</li>
522<li>Stack slot coloring for register spills (denser stack frames)</li>
523<li>SelectionDAGS: New BuildVectorSDNode (r65296), and ISD::VECTOR_SHUFFLE (r69952 / PR2957)</li>
Chris Lattner80ed2552009-10-08 07:01:46 +0000[diff] [blame]524<li>Experimental support for shrink wrapping support in PEI.</li>
Chris Lattner6ee62f82009-10-09 05:55:04 +0000[diff] [blame]525<li>Experimental support for writing ELF .o files directly from the compiler,
526 it works well for many simple C testcases, but doesn't support exception
527 handling, debug info, inline assembly, etc.</li>
528<li>Targets can now specify register allocation hints through
529 MachineRegisterInfo:: setRegAllocationHint. A regalloc hint consists 1) hint
530 type, 2) physical register number. A hint type of zero specifies a register
531 allocation preference. Other hint type values are target specific which are
532 resolved by TargetRegisterInfo::ResolveRegAllocHint. An example of which is
533 the ARM target can uses register hint to request that the register allocator
534 provide an even / odd register pair to two virtual registers. It is
535 important to note the register allocation hints are just hints. There is no
536 guarantee the register allocators will be able to satisfy the hints.</li>
537
Duncan Sandse2e4a892009-06-24 08:38:48 +0000[diff] [blame]538</ul>
539</div>
540
541<!--=========================================================================-->
542<div class="doc_subsection">
543<a name="x86">X86-32 and X86-64 Target Improvements</a>
544</div>
545
546<div class="doc_text">
547<p>New features of the X86 target include:
548</p>
549
550<ul>
551
Chris Lattner7b0a6812009-10-08 06:27:53 +0000[diff] [blame]552<li>Preliminary support for addrspace 256 -> GS, 257 -> FS, known problems: CodeGenerator.html#x86_memory</li>
553<li>Support for softfloat modes, typically used by OS kernels.</li>
554
555<li>X86-64: better modeling of implicit zero extensions, eliminates a lot of redundant zexts</li>
556<li>X86-64 TLS support for local exec and initial exec.</li>
557<li>Better modeling of H registerts as subregs.</li>
558<li>Vector icmp/fcmp now work with SSE codegen.</li>
559<li>SSE 4.2 support.</li>
560<li>all global variable reference logic is now in ClassifyGlobalReference.</li>
561</li>
Duncan Sandse2e4a892009-06-24 08:38:48 +0000[diff] [blame]562</ul>
563
564</div>
565
566<!--=========================================================================-->
567<div class="doc_subsection">
568<a name="pic16">PIC16 Target Improvements</a>
569</div>
570
571<div class="doc_text">
572<p>New features of the PIC16 target include:
573</p>
574
575<ul>
Chris Lattner7b0a6812009-10-08 06:27:53 +0000[diff] [blame]576<li>Support for floating-point, indirect function calls, and
577 passing/returning aggregate types to functions.
578<li>The code generator is able to generate debug info into output COFF files.
579<li>Support for placing an object into a specific section or at a specific
580 address in memory.</li>
Duncan Sandse2e4a892009-06-24 08:38:48 +0000[diff] [blame]581</ul>
582
583<p>Things not yet supported:</p>
584
585<ul>
Duncan Sandse2e4a892009-06-24 08:38:48 +0000[diff] [blame]586<li>Variable arguments.</li>
Duncan Sandse2e4a892009-06-24 08:38:48 +0000[diff] [blame]587<li>Interrupts/programs.</li>
Duncan Sandse2e4a892009-06-24 08:38:48 +0000[diff] [blame]588</ul>
589
590</div>
591
Bob Wilson755cbe02009-08-12 21:19:49 +0000[diff] [blame]592<!--=========================================================================-->
593<div class="doc_subsection">
594<a name="ARM">ARM Target Improvements</a>
595</div>
596
597<div class="doc_text">
598<p>New features of the ARM target include:
599</p>
600
601<ul>
602
603<li>Preliminary support for processors, such as the Cortex-A8 and Cortex-A9,
Sandeep Patel64e39612009-08-20 15:01:16 +0000[diff] [blame]604that implement version v7-A of the ARM architecture. The ARM backend now
605supports both the Thumb2 and Advanced SIMD (Neon) instruction sets. The
606AAPCS-VFP "hard float" calling conventions are also supported with the
607<tt>-float-abi=hard</tt> flag. These features are still somewhat experimental
608and subject to change. The Neon intrinsics, in particular, may change in future
609releases of LLVM.
Bob Wilson755cbe02009-08-12 21:19:49 +0000[diff] [blame]610</li>
Chris Lattner7b0a6812009-10-08 06:27:53 +0000[diff] [blame]611
612 ARM AAPCS-VFP hard float ABI is supported.
613 ARM calling convention code is now tblgen generated instead of manual.
614 ARM: NEON support. neonfp for doing single precision fp with neon instead of VFP.
615
Bob Wilson755cbe02009-08-12 21:19:49 +0000[diff] [blame]616</ul>
617
618</div>
619
Duncan Sandse2e4a892009-06-24 08:38:48 +0000[diff] [blame]620<!--=========================================================================-->
621<div class="doc_subsection">
Chris Lattner7b0a6812009-10-08 06:27:53 +0000[diff] [blame]622<a name="OtherTarget">Other Target Specific Improvements</a>
Duncan Sandse2e4a892009-06-24 08:38:48 +0000[diff] [blame]623</div>
624
625<div class="doc_text">
Chris Lattner7b0a6812009-10-08 06:27:53 +0000[diff] [blame]626<p>New features of other targets include:
627</p>
Duncan Sandse2e4a892009-06-24 08:38:48 +0000[diff] [blame]628
629<ul>
Chris Lattner7b0a6812009-10-08 06:27:53 +0000[diff] [blame]630<li>Mips now supports O32 Calling Convention.</li>
Chris Lattnera8f42142009-10-09 06:36:25 +0000[diff] [blame]631<li>Many improvements to the 32-bit PowerPC SVR4 ABI (used on powerpc-linux)
632 support, lots of bugs fixed.</li>
633<li>Added support for the 64-bit PowerPC SVR4 ABI (used on powerpc64-linux).
634 Needs more testing.</li>
Chris Lattner7b0a6812009-10-08 06:27:53 +0000[diff] [blame]635</ul>
636
637</div>
638
639<!--=========================================================================-->
640<div class="doc_subsection">
Chris Lattner45178502009-10-09 06:24:25 +0000[diff] [blame]641<a name="executionengine">Interpreter and JIT Improvements</a>
642</div>
643
644<div class="doc_text">
645
646<ul>
647<li>The JIT now supports generating more than 16M of code.</li>
648<li>When configured with --with-oprofile, the JIT can now inform oprofile about
649 JIT'd code, allowing oprofile to get line number and function name
650 information for JIT'd functions.</li>
651<li>When "libffi" is available, the LLVM interpreter now uses it, which supports
652 calling almost arbitrary external (natively compiled) functions.</li>
653<li>Clients of the JIT can now register a 'JITEventListener' object to receive
654 callbacks when the JIT emits or frees machine code. The OProfile support
655 uses this mechanism.</li>
656</ul>
657
658</div>
659
660
661<!--=========================================================================-->
662<div class="doc_subsection">
Chris Lattner7b0a6812009-10-08 06:27:53 +0000[diff] [blame]663<a name="newapis">New Useful APIs</a>
664</div>
665
666<div class="doc_text">
667
668<ul>
669<li>New EngineBuilder class for creating JITs: r76276</li>
670 New PrettyStackTrace, crashes of llvm tools should give some indication of what the compiler was doing at the time of the crash (e.g. running a pass), and print out command line arguments.
671 StringRef class, Twine class.
672 New WeakVH and AssertingVH and CallbackVH classes.
673 New llvm/ADT/Triple class.
674 llvm_report_error() error handling API (llvm/Support/ErrorHandling.h)
675 New llvm/System/Atomic.h, llvm/System/RWMutex.h for portable atomic ops, rw locks.
676 New SourceMgr, SMLoc classes for simple parsers with caret diagnostics and #include support, (used by
677 tablegen, llvm-mc, the .ll parser, FileCheck, etc)
678
Chris Lattner45178502009-10-09 06:24:25 +0000[diff] [blame]679</ul>
Chris Lattner7b0a6812009-10-08 06:27:53 +0000[diff] [blame]680
681
682</div>
683
684<!--=========================================================================-->
685<div class="doc_subsection">
686<a name="otherimprovements">Other Improvements and New Features</a>
687</div>
688
689<div class="doc_text">
690<p>Other miscellaneous features include:</p>
691
692<ul>
Chris Lattner45178502009-10-09 06:24:25 +0000[diff] [blame]693<li>LLVM now includes a new internal '<a
694 href="http://llvm.org/cmds/FileCheck.html">FileCheck</a>' tool which allows
695 writing much more accurate regression tests that run faster. Please see the
696 <a href="TestingGuide.html#FileCheck">FileCheck section of the Testing
697 Guide</a> for more information.</li>
Chris Lattner6ee62f82009-10-09 05:55:04 +0000[diff] [blame]698<li>LLVM profile information support has been significantly improved to produce
699correct use counts, and has support for edge profiling with reduced runtime
700overhead. Combined, the generated profile information is both more correct and
701imposes about half as much overhead (2.6. from 12% to 6% overhead on SPEC
702CPU2000).</li>
Chris Lattner80ed2552009-10-08 07:01:46 +0000[diff] [blame]703<li>Many extensions to the C APIs.</li>
Chris Lattner45178502009-10-09 06:24:25 +0000[diff] [blame]704<li>LLVM 2.6 includes a brand new experimental LLVM bindings to the Ada2005
705programming language.</li>
Chris Lattner80ed2552009-10-08 07:01:46 +0000[diff] [blame]706
707<li>LLVMC:
708
709* Dynamic plugins now work on Windows.
710* New option property: init. Makes possible to provide default values for
711 options defined in plugins (interface to cl::init).
712* New example: Skeleton, shows how to create a standalone LLVMC-based driver.
713* New example: mcc16, a driver for the PIC16 toolchain.</li>
714
Duncan Sandse2e4a892009-06-24 08:38:48 +0000[diff] [blame]715</ul>
716
717</div>
718
719
720<!--=========================================================================-->
721<div class="doc_subsection">
722<a name="changes">Major Changes and Removed Features</a>
723</div>
724
725<div class="doc_text">
726
727<p>If you're already an LLVM user or developer with out-of-tree changes based
Owen Andersone27be3a2009-07-02 16:48:38 +0000[diff] [blame]728on LLVM 2.5, this section lists some "gotchas" that you may run into upgrading
Duncan Sandse2e4a892009-06-24 08:38:48 +0000[diff] [blame]729from the previous release.</p>
730
731<ul>
732
Chris Lattner7b0a6812009-10-08 06:27:53 +0000[diff] [blame]733<li>The Itanium (IA64) backend has been removed. It was not supported and
734 bitrotted.</li>
735<li>The BigBlock register allocator has been removed, it also bitrotted.</li>
736<li>The C Backend (-march=c) is no longer considered part of the LLVM release
737criteria. We still want it to work, but no one is maintaining it and it lacks
738support for arbitrary precision integers and other important IR features.</li>
Duncan Sandse2e4a892009-06-24 08:38:48 +0000[diff] [blame]739</ul>
740
Chris Lattner80ed2552009-10-08 07:01:46 +0000[diff] [blame]741 LLVM build now builds all libraries as .a files instead of some
742 libraries as relinked .o files. This requires some APIs like
743 InitializeAllTargets.h. TargetRegistry!
744
745
746
Duncan Sandse2e4a892009-06-24 08:38:48 +0000[diff] [blame]747
748<p>In addition, many APIs have changed in this release. Some of the major LLVM
749API changes are:</p>
750
Chris Lattner7b0a6812009-10-08 06:27:53 +0000[diff] [blame]751
752 API Cleanup:
753 no use of hash_set/hash_map, no more llvm::OStream
754 Use raw_ostream for everything, killed off llvm/Streams.h and DOUT
755
Duncan Sandse2e4a892009-06-24 08:38:48 +0000[diff] [blame]756<ul>
Owen Andersone27be3a2009-07-02 16:48:38 +0000[diff] [blame]757<li>LLVM's global uniquing tables for <tt>Type</tt>s and <tt>Constant</tt>s have
758 been privatized into members of an <tt>LLVMContext</tt>. A number of APIs
759 now take an <tt>LLVMContext</tt> as a parameter. To smooth the transition
760 for clients that will only ever use a single context, the new
761 <tt>getGlobalContext()</tt> API can be used to access a default global
762 context which can be passed in any and all cases where a context is
763 required.
Duncan Sandse2e4a892009-06-24 08:38:48 +0000[diff] [blame]764<li>The <tt>getABITypeSize</tt> methods are now called <tt>getAllocSize</tt>.</li>
Dan Gohmanf667e7e2009-07-07 20:05:15 +0000[diff] [blame]765<li>The <tt>Add</tt>, <tt>Sub</tt>, and <tt>Mul</tt> operators are no longer
766 overloaded for floating-point types. Floating-point addition, subtraction,
767 and multiplication are now represented with new operators <tt>FAdd</tt>,
768 <tt>FSub</tt>, and <tt>FMul</tt>. In the <tt>IRBuilder</tt> API,
769 <tt>CreateAdd</tt>, <tt>CreateSub</tt>, <tt>CreateMul</tt>, and
770 <tt>CreateNeg</tt> should only be used for integer arithmetic now;
771 <tt>CreateFAdd</tt>, <tt>CreateFSub</tt>, <tt>CreateFMul</tt>, and
772 <tt>CreateFNeg</tt> should now be used for floating-point arithmetic.</li>
Daniel Dunbar02e73132009-07-12 20:41:27 +0000[diff] [blame]773<li>The DynamicLibrary class can no longer be constructed, its functionality has
774 moved to static member functions.</li>
Dan Gohmane4d54d72009-07-15 19:59:19 +0000[diff] [blame]775<li><tt>raw_fd_ostream</tt>'s constructor for opening a given filename now
776 takes an extra <tt>Force</tt> argument. If <tt>Force</tt> is set to
777 <tt>false</tt>, an error will be reported if a file with the given name
778 already exists. If <tt>Force</tt> is set to <tt>true</tt>, the file will
779 be silently truncated (which is the behavior before this flag was
780 added).</li>
Torok Edwinf6fa8ae2009-07-21 20:27:10 +0000[diff] [blame]781<li><tt>SCEVHandle</tt> no longer exists, because reference counting is no
782longer done for <tt>SCEV*</tt> objects, instead <tt>const SCEV*</tt> should be
783used.</li>
Daniel Dunbar8b5ee822009-07-25 05:26:53 +0000[diff] [blame]784
Daniel Dunbar6e0d1cb2009-07-25 04:41:11 +0000[diff] [blame]785<li>Many APIs, notably <tt>llvm::Value</tt>, now use the <tt>StringRef</tt>
786and <tt>Twine</tt> classes instead of passing <tt>const char*</tt>
787or <tt>std::string</tt>, as described in
788the <a href="ProgrammersManual.html#string_apis">Programmer's Manual</a>. Most
Daniel Dunbar8b5ee822009-07-25 05:26:53 +0000[diff] [blame]789clients should be unaffected by this transition, unless they are used to <tt>Value::getName()</tt> returning a string. Here are some tips on updating to 2.6:
790 <ul>
791 <li><tt>getNameStr()</tt> is still available, and matches the old
792 behavior. Replacing <tt>getName()</tt> calls with this is an safe option,
793 although more efficient alternatives are now possible.</li>
794
795 <li>If you were just relying on <tt>getName()</tt> being able to be sent to
796 a <tt>std::ostream</tt>, consider migrating
797 to <tt>llvm::raw_ostream</tt>.</li>
798
799 <li>If you were using <tt>getName().c_str()</tt> to get a <tt>const
800 char*</tt> pointer to the name, you can use <tt>getName().data()</tt>.
801 Note that this string (as before), may not be the entire name if the
802 name containts embedded null characters.</li>
803
804 <li>If you were using operator plus on the result of <tt>getName()</tt> and
805 treating the result as an <tt>std::string</tt>, you can either
806 uses <tt>Twine::str</tt> to get the result as an <tt>std::string</tt>, or
807 could move to a <tt>Twine</tt> based design.</li>
Daniel Dunbar03d76512009-07-25 23:55:21 +0000[diff] [blame]808
809 <li><tt>isName()</tt> should be replaced with comparison
Benjamin Kramere15192b2009-08-05 15:42:44 +0000[diff] [blame]810 against <tt>getName()</tt> (this is now efficient).
Daniel Dunbar8b5ee822009-07-25 05:26:53 +0000[diff] [blame]811 </ul>
812</li>
813
Daniel Dunbara5881e32009-07-26 02:12:58 +0000[diff] [blame]814<li>The registration interfaces for backend Targets has changed (what was
Daniel Dunbard6b06b12009-07-26 05:41:39 +0000[diff] [blame]815previously TargetMachineRegistry). For backend authors, see the <a href="WritingAnLLVMBackend.html#TargetRegistration">Writing An LLVM Backend</a> guide. For clients, the notable API changes are:
816 <ul>
817 <li><tt>TargetMachineRegistry</tt> has been renamed
818 to <tt>TargetRegistry</tt>.</li>
819
820 <li>Clients should move to using the <tt>TargetRegistry::lookupTarget()</tt>
821 function to find targets.</li>
822 </ul>
823</li>
Daniel Dunbar8b5ee822009-07-25 05:26:53 +0000[diff] [blame]824
Torok Edwinf6fa8ae2009-07-21 20:27:10 +0000[diff] [blame]825<li>llvm-dis now fails if output file exists, instead of dumping to stdout.
Daniel Dunbar6e0d1cb2009-07-25 04:41:11 +0000[diff] [blame]826FIXME: describe any other tool changes due to the raw_fd_ostream change. FIXME:
827This is not an API change, maybe there should be a tool changes section?</li>
Torok Edwinf6fa8ae2009-07-21 20:27:10 +0000[diff] [blame]828<li>temporarely due to Context API change passes should call doInitialization()
829method of the pass they inherit from, otherwise Context is NULL.
830FIXME: remove this entry when this is no longer needed.<li>
Duncan Sandse2e4a892009-06-24 08:38:48 +0000[diff] [blame]831</ul>
832
833</div>
834
835
836
837<!-- *********************************************************************** -->
838<div class="doc_section">
839 <a name="portability">Portability and Supported Platforms</a>
840</div>
841<!-- *********************************************************************** -->
842
843<div class="doc_text">
844
845<p>LLVM is known to work on the following platforms:</p>
846
847<ul>
848<li>Intel and AMD machines (IA32, X86-64, AMD64, EMT-64) running Red Hat
Chris Lattner3e7b5ca2009-07-21 22:47:03 +0000[diff] [blame]849Linux, Fedora Core, FreeBSD and AuroraUX (and probably other unix-like systems).</li>
Duncan Sandse2e4a892009-06-24 08:38:48 +0000[diff] [blame]850<li>PowerPC and X86-based Mac OS X systems, running 10.3 and above in 32-bit
851and 64-bit modes.</li>
852<li>Intel and AMD machines running on Win32 using MinGW libraries (native).</li>
853<li>Intel and AMD machines running on Win32 with the Cygwin libraries (limited
854 support is available for native builds with Visual C++).</li>
855<li>Sun UltraSPARC workstations running Solaris 10.</li>
856<li>Alpha-based machines running Debian GNU/Linux.</li>
Duncan Sandse2e4a892009-06-24 08:38:48 +0000[diff] [blame]857</ul>
858
859<p>The core LLVM infrastructure uses GNU autoconf to adapt itself
860to the machine and operating system on which it is built. However, minor
861porting may be required to get LLVM to work on new platforms. We welcome your
862portability patches and reports of successful builds or error messages.</p>
863
864</div>
865
866<!-- *********************************************************************** -->
867<div class="doc_section">
868 <a name="knownproblems">Known Problems</a>
869</div>
870<!-- *********************************************************************** -->
871
872<div class="doc_text">
873
874<p>This section contains significant known problems with the LLVM system,
875listed by component. If you run into a problem, please check the <a
876href="http://llvm.org/bugs/">LLVM bug database</a> and submit a bug if
877there isn't already one.</p>
878
Chris Lattner3e7b5ca2009-07-21 22:47:03 +0000[diff] [blame]879<ul>
Chris Lattneraedb59a2009-07-21 23:17:26 +0000[diff] [blame]880<li>LLVM will not correctly compile on Solaris and/or OpenSolaris
Chris Lattner3e7b5ca2009-07-21 22:47:03 +0000[diff] [blame]881using the stock GCC 3.x.x series 'out the box',
882See: <a href="#brokengcc">Broken versions of GCC and other tools</a>.
883However, A <a href="http://pkg.auroraux.org/GCC">Modern GCC Build</a>
884for x86/x64 has been made available from the third party AuroraUX Project
Chris Lattner7b0a6812009-10-08 06:27:53 +0000[diff] [blame]885that has been meticulously tested for bootstrapping LLVM &amp; Clang.</li>
Chris Lattner3e7b5ca2009-07-21 22:47:03 +0000[diff] [blame]886</ul>
887
Duncan Sandse2e4a892009-06-24 08:38:48 +0000[diff] [blame]888</div>
889
890<!-- ======================================================================= -->
891<div class="doc_subsection">
892 <a name="experimental">Experimental features included with this release</a>
893</div>
894
895<div class="doc_text">
896
897<p>The following components of this LLVM release are either untested, known to
898be broken or unreliable, or are in early development. These components should
899not be relied on, and bugs should not be filed against them, but they may be
900useful to some people. In particular, if you would like to work on one of these
901components, please contact us on the <a
902href="http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev">LLVMdev list</a>.</p>
903
904<ul>
Dan Gohmand2cb3d22009-07-24 00:30:09 +0000[diff] [blame]905<li>The MSIL, Alpha, SPU, MIPS, and PIC16 backends are experimental.</li>
Duncan Sandse2e4a892009-06-24 08:38:48 +0000[diff] [blame]906<li>The <tt>llc</tt> "<tt>-filetype=asm</tt>" (the default) is the only
907 supported value for this option.</li>
908</ul>
909
910</div>
911
912<!-- ======================================================================= -->
913<div class="doc_subsection">
914 <a name="x86-be">Known problems with the X86 back-end</a>
915</div>
916
917<div class="doc_text">
918
919<ul>
920 <li>The X86 backend does not yet support
921 all <a href="http://llvm.org/PR879">inline assembly that uses the X86
922 floating point stack</a>. It supports the 'f' and 't' constraints, but not
923 'u'.</li>
924 <li>The X86 backend generates inefficient floating point code when configured
925 to generate code for systems that don't have SSE2.</li>
926 <li>Win64 code generation wasn't widely tested. Everything should work, but we
927 expect small issues to happen. Also, llvm-gcc cannot build the mingw64
928 runtime currently due
929 to <a href="http://llvm.org/PR2255">several</a>
930 <a href="http://llvm.org/PR2257">bugs</a> and due to lack of support for
931 the
932 'u' inline assembly constraint and for X87 floating point inline assembly.</li>
933 <li>The X86-64 backend does not yet support the LLVM IR instruction
934 <tt>va_arg</tt>. Currently, the llvm-gcc and front-ends support variadic
935 argument constructs on X86-64 by lowering them manually.</li>
936</ul>
937
938</div>
939
940<!-- ======================================================================= -->
941<div class="doc_subsection">
942 <a name="ppc-be">Known problems with the PowerPC back-end</a>
943</div>
944
945<div class="doc_text">
946
947<ul>
948<li>The Linux PPC32/ABI support needs testing for the interpreter and static
949compilation, and lacks support for debug information.</li>
950</ul>
951
952</div>
953
954<!-- ======================================================================= -->
955<div class="doc_subsection">
956 <a name="arm-be">Known problems with the ARM back-end</a>
957</div>
958
959<div class="doc_text">
960
961<ul>
Bob Wilson755cbe02009-08-12 21:19:49 +0000[diff] [blame]962<li>Support for the Advanced SIMD (Neon) instruction set is still incomplete
963and not well tested. Some features may not work at all, and the code quality
964may be poor in some cases.</li>
Duncan Sandse2e4a892009-06-24 08:38:48 +0000[diff] [blame]965<li>Thumb mode works only on ARMv6 or higher processors. On sub-ARMv6
966processors, thumb programs can crash or produce wrong
967results (<a href="http://llvm.org/PR1388">PR1388</a>).</li>
968<li>Compilation for ARM Linux OABI (old ABI) is supported but not fully tested.
969</li>
Duncan Sandse2e4a892009-06-24 08:38:48 +0000[diff] [blame]970</ul>
971
972</div>
973
974<!-- ======================================================================= -->
975<div class="doc_subsection">
976 <a name="sparc-be">Known problems with the SPARC back-end</a>
977</div>
978
979<div class="doc_text">
980
981<ul>
982<li>The SPARC backend only supports the 32-bit SPARC ABI (-m32); it does not
983 support the 64-bit SPARC ABI (-m64).</li>
984</ul>
985
986</div>
987
988<!-- ======================================================================= -->
989<div class="doc_subsection">
990 <a name="mips-be">Known problems with the MIPS back-end</a>
991</div>
992
993<div class="doc_text">
994
995<ul>
996<li>The O32 ABI is not fully supported.</li>
997<li>64-bit MIPS targets are not supported yet.</li>
998</ul>
999
1000</div>
1001
1002<!-- ======================================================================= -->
1003<div class="doc_subsection">
1004 <a name="alpha-be">Known problems with the Alpha back-end</a>
1005</div>
1006
1007<div class="doc_text">
1008
1009<ul>
1010
1011<li>On 21164s, some rare FP arithmetic sequences which may trap do not have the
1012appropriate nops inserted to ensure restartability.</li>
1013
1014</ul>
1015</div>
1016
1017<!-- ======================================================================= -->
1018<div class="doc_subsection">
Duncan Sandse2e4a892009-06-24 08:38:48 +0000[diff] [blame]1019 <a name="c-be">Known problems with the C back-end</a>
1020</div>
1021
1022<div class="doc_text">
1023
1024<ul>
1025<li><a href="http://llvm.org/PR802">The C backend has only basic support for
1026 inline assembly code</a>.</li>
1027<li><a href="http://llvm.org/PR1658">The C backend violates the ABI of common
1028 C++ programs</a>, preventing intermixing between C++ compiled by the CBE and
1029 C++ code compiled with <tt>llc</tt> or native compilers.</li>
1030<li>The C backend does not support all exception handling constructs.</li>
1031<li>The C backend does not support arbitrary precision integers.</li>
1032</ul>
1033
1034</div>
1035
1036
1037<!-- ======================================================================= -->
1038<div class="doc_subsection">
1039 <a name="c-fe">Known problems with the llvm-gcc C front-end</a>
1040</div>
1041
1042<div class="doc_text">
1043
1044<p>llvm-gcc does not currently support <a href="http://llvm.org/PR869">Link-Time
1045Optimization</a> on most platforms "out-of-the-box". Please inquire on the
1046LLVMdev mailing list if you are interested.</p>
1047
1048<p>The only major language feature of GCC not supported by llvm-gcc is
1049 the <tt>__builtin_apply</tt> family of builtins. However, some extensions
1050 are only supported on some targets. For example, trampolines are only
1051 supported on some targets (these are used when you take the address of a
1052 nested function).</p>
1053
1054<p>If you run into GCC extensions which are not supported, please let us know.
1055</p>
1056
1057</div>
1058
1059<!-- ======================================================================= -->
1060<div class="doc_subsection">
1061 <a name="c++-fe">Known problems with the llvm-gcc C++ front-end</a>
1062</div>
1063
1064<div class="doc_text">
1065
1066<p>The C++ front-end is considered to be fully
1067tested and works for a number of non-trivial programs, including LLVM
1068itself, Qt, Mozilla, etc.</p>
1069
1070<ul>
1071<li>Exception handling works well on the X86 and PowerPC targets. Currently
1072 only Linux and Darwin targets are supported (both 32 and 64 bit).</li>
1073</ul>
1074
1075</div>
1076
1077<!-- ======================================================================= -->
1078<div class="doc_subsection">
1079 <a name="fortran-fe">Known problems with the llvm-gcc Fortran front-end</a>
1080</div>
1081
1082<div class="doc_text">
1083<ul>
1084<li>Fortran support generally works, but there are still several unresolved bugs
1085 in Bugzilla. Please see the tools/gfortran component for details.</li>
1086</ul>
1087</div>
1088
1089<!-- ======================================================================= -->
1090<div class="doc_subsection">
1091 <a name="ada-fe">Known problems with the llvm-gcc Ada front-end</a>
1092</div>
1093
1094<div class="doc_text">
1095The llvm-gcc 4.2 Ada compiler works fairly well; however, this is not a mature
1096technology, and problems should be expected.
1097<ul>
1098<li>The Ada front-end currently only builds on X86-32. This is mainly due
1099to lack of trampoline support (pointers to nested functions) on other platforms.
1100However, it <a href="http://llvm.org/PR2006">also fails to build on X86-64</a>
1101which does support trampolines.</li>
1102<li>The Ada front-end <a href="http://llvm.org/PR2007">fails to bootstrap</a>.
1103This is due to lack of LLVM support for <tt>setjmp</tt>/<tt>longjmp</tt> style
1104exception handling, which is used internally by the compiler.
1105Workaround: configure with --disable-bootstrap.</li>
1106<li>The c380004, <a href="http://llvm.org/PR2010">c393010</a>
1107and <a href="http://llvm.org/PR2421">cxg2021</a> ACATS tests fail
1108(c380004 also fails with gcc-4.2 mainline).
1109If the compiler is built with checks disabled then <a href="http://llvm.org/PR2010">c393010</a>
1110causes the compiler to go into an infinite loop, using up all system memory.</li>
1111<li>Some GCC specific Ada tests continue to crash the compiler.</li>
1112<li>The -E binder option (exception backtraces)
1113<a href="http://llvm.org/PR1982">does not work</a> and will result in programs
1114crashing if an exception is raised. Workaround: do not use -E.</li>
1115<li>Only discrete types <a href="http://llvm.org/PR1981">are allowed to start
1116or finish at a non-byte offset</a> in a record. Workaround: do not pack records
1117or use representation clauses that result in a field of a non-discrete type
1118starting or finishing in the middle of a byte.</li>
1119<li>The <tt>lli</tt> interpreter <a href="http://llvm.org/PR2009">considers
1120'main' as generated by the Ada binder to be invalid</a>.
1121Workaround: hand edit the file to use pointers for <tt>argv</tt> and
1122<tt>envp</tt> rather than integers.</li>
1123<li>The <tt>-fstack-check</tt> option <a href="http://llvm.org/PR2008">is
1124ignored</a>.</li>
1125</ul>
1126</div>
1127
Erick Tryzelaar17167be2009-09-28 04:42:55 +0000[diff] [blame]1128<!-- ======================================================================= -->
1129<div class="doc_subsection">
1130 <a name="ocaml-bindingse">Known problems with the O'Caml bindings</a>
1131</div>
1132
1133<div class="doc_text">
1134
1135<p>The Llvm.Linkage module is broken, and has incorrect values. Only
1136Llvm.Linkage.External, Llvm.Linkage.Available_externally, and
1137Llvm.Linkage.Link_once will be correct. If you need any of the other linkage
1138modes, you'll have to write an external C library in order to expose the
1139functionality. This has been fixed in the trunk.</p>
1140</div>
1141
Duncan Sandse2e4a892009-06-24 08:38:48 +0000[diff] [blame]1142<!-- *********************************************************************** -->
1143<div class="doc_section">
1144 <a name="additionalinfo">Additional Information</a>
1145</div>
1146<!-- *********************************************************************** -->
1147
1148<div class="doc_text">
1149
1150<p>A wide variety of additional information is available on the <a
1151href="http://llvm.org">LLVM web page</a>, in particular in the <a
1152href="http://llvm.org/docs/">documentation</a> section. The web page also
1153contains versions of the API documentation which is up-to-date with the
1154Subversion version of the source code.
1155You can access versions of these documents specific to this release by going
1156into the "<tt>llvm/doc/</tt>" directory in the LLVM tree.</p>
1157
1158<p>If you have any questions or comments about LLVM, please feel free to contact
1159us via the <a href="http://llvm.org/docs/#maillist"> mailing
1160lists</a>.</p>
1161
1162</div>
1163
1164<!-- *********************************************************************** -->
1165
1166<hr>
1167<address>
1168 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
1169 src="http://jigsaw.w3.org/css-validator/images/vcss-blue" alt="Valid CSS"></a>
1170 <a href="http://validator.w3.org/check/referer"><img
1171 src="http://www.w3.org/Icons/valid-html401-blue" alt="Valid HTML 4.01"></a>
1172
1173 <a href="http://llvm.org/">LLVM Compiler Infrastructure</a><br>
1174 Last modified: $Date$
1175</address>
1176
1177</body>
1178</html>