Gitiles
Code Review Sign In
gerrit-public.fairphone.software / fp2-dev / platform / external / llvm / 51ead6da49d101c8507fa07b7d1b35dc563abeed / . / docs / ReleaseNotes-2.6.html
blob: 5a209fcc632ab3d2345a61396343eed599d35dd2 [file] [log] [blame]
Duncan Sands54fbb412009-06-24 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 Lattner169c7012009-10-08 07:01:46 +0000[diff] [blame]54
Chris Lattnerb2d43472009-10-05 02:12:39 +0000[diff] [blame]55
Chris Lattner6aaf6902009-10-08 06:27:53 +0000[diff] [blame]56<!--
57Almost dead code.
Chris Lattner169c7012009-10-08 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 Lattner6aaf6902009-10-08 06:27:53 +0000[diff] [blame]61-->
Chris Lattnerb2d43472009-10-05 02:12:39 +0000[diff] [blame]62
Chris Lattnerb2d43472009-10-05 02:12:39 +0000[diff] [blame]63
64<!-- Unfinished features in 2.6:
Chris Lattner169c7012009-10-08 07:01:46 +0000[diff] [blame]65 gcc plugin.
Chris Lattnerb2d43472009-10-05 02:12:39 +0000[diff] [blame]66 strong phi elim
67 variable debug info for optimized code
Duncan Sands54fbb412009-06-24 08:38:48 +0000[diff] [blame]68 postalloc scheduler: anti dependence breaking, hazard recognizer?
Chris Lattnerb2d43472009-10-05 02:12:39 +0000[diff] [blame]69 metadata
70 loop dependence analysis
Chris Lattner169c7012009-10-08 07:01:46 +0000[diff] [blame]71 ELF Writer? How stable?
72 <li>PostRA scheduler improvements, ARM adoption (David Goodwin).</li>
Chris Lattner0d6011062009-10-09 06:24:25 +0000[diff] [blame]73 2.7 supports the GDB 7.0 jit interfaces for debug info.
Duncan Sands54fbb412009-06-24 08:38:48 +0000[diff] [blame]74 -->
75
76 <!-- for announcement email:
Chris Lattnerb2d43472009-10-05 02:12:39 +0000[diff] [blame]77 Logo web page.
78 llvm devmtg
79 compiler_rt
Chris Lattner51ead6d2009-10-10 19:00:55 +0000[diff] [blame^]80 KLEE web page at klee.llvm.org
Chris Lattner6aaf6902009-10-08 06:27:53 +0000[diff] [blame]81 Many new papers added to /pubs/
82 Mention gcc plugin.
83
Duncan Sands54fbb412009-06-24 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
Chris Lattner4d91d4d2009-10-10 18:33:13 +0000[diff] [blame]95repository (which roughly includes the LLVM optimizers, code generators
96and supporting tools), the Clang repository and the llvm-gcc repository. In
97addition to this code, the LLVM Project includes other sub-projects that are in
98development. Here we include updates on these subprojects.
Duncan Sands54fbb412009-06-24 08:38:48 +0000[diff] [blame]99</p>
100
101</div>
102
103
104<!--=========================================================================-->
105<div class="doc_subsection">
106<a name="clang">Clang: C/C++/Objective-C Frontend Toolkit</a>
107</div>
108
109<div class="doc_text">
110
Chris Lattner401ec6d2009-10-09 05:01:15 +0000[diff] [blame]111<p>The <a href="http://clang.llvm.org/">Clang project</a> is an effort to build
112a set of new 'LLVM native' front-end technologies for the C family of languages.
113LLVM 2.6 is the first release to officially include Clang, and it provides a
Chris Lattner4d91d4d2009-10-10 18:33:13 +0000[diff] [blame]114production quality C and Objective-C compiler. If you are interested in <a
115href="http://clang.llvm.org/performance.html">fast compiles</a> and
116<a href="http://clang.llvm.org/diagnostics.html">good diagnostics</a>, we
117encourage you to try it out. Clang currently compiles typical Objective-C code
1183x faster than GCC and compiles C code about 30% faster than GCC at -O0 -g
119(which is when the most pressure is on the frontend).</p>
Chris Lattner401ec6d2009-10-09 05:01:15 +0000[diff] [blame]120
121<p>In addition to supporting these languages, C++ support is also <a
122href="http://clang.llvm.org/cxx_status.html">well under way</a>, and mainline
123Clang is able to parse the libstdc++ 4.2 headers and even codegen simple apps.
124If you are interested in Clang C++ support or any other Clang feature, we
125strongly encourage you to get involved on the <a
Duncan Sands54fbb412009-06-24 08:38:48 +0000[diff] [blame]126href="http://lists.cs.uiuc.edu/mailman/listinfo/cfe-dev">Clang front-end mailing
127list</a>.</p>
128
129<p>In the LLVM 2.6 time-frame, the Clang team has made many improvements:</p>
130
131<ul>
Chris Lattner401ec6d2009-10-09 05:01:15 +0000[diff] [blame]132<li>C and Objective-C support are now considered production quality.</li>
Chris Lattner4d91d4d2009-10-10 18:33:13 +0000[diff] [blame]133<li>AuroraUX, FreeBSD, and OpenBSD are now supported.</li>
Chris Lattner401ec6d2009-10-09 05:01:15 +0000[diff] [blame]134<li>Most of Objective-C 2.0 is now supported with the GNU runtime.</li>
Duncan Sands54fbb412009-06-24 08:38:48 +0000[diff] [blame]135<li>Many many bugs are fixed and many features have been added.</li>
136</ul>
137</div>
138
139<!--=========================================================================-->
140<div class="doc_subsection">
141<a name="clangsa">Clang Static Analyzer</a>
142</div>
143
144<div class="doc_text">
145
Chris Lattner169c7012009-10-08 07:01:46 +0000[diff] [blame]146<p><b>UPDATE!</b> Previously announced in the 2.4 and 2.5 LLVM releases, the Clang project also
Duncan Sands54fbb412009-06-24 08:38:48 +0000[diff] [blame]147includes an early stage static source code analysis tool for <a
148href="http://clang.llvm.org/StaticAnalysis.html">automatically finding bugs</a>
149in C and Objective-C programs. The tool performs a growing set of checks to find
150bugs that occur on a specific path within a program.</p>
151
152<p>In the LLVM 2.6 time-frame there have been many significant improvements to
153XYZ.</p>
154
155<p>The set of checks performed by the static analyzer continues to expand, and
156future plans for the tool include full source-level inter-procedural analysis
157and deeper checks such as buffer overrun detection. There are many opportunities
158to extend and enhance the static analyzer, and anyone interested in working on
159this project is encouraged to get involved!</p>
160
161</div>
162
163<!--=========================================================================-->
164<div class="doc_subsection">
165<a name="vmkit">VMKit: JVM/CLI Virtual Machine Implementation</a>
166</div>
167
168<div class="doc_text">
169<p>
170The <a href="http://vmkit.llvm.org/">VMKit project</a> is an implementation of
Nicolas Geoffray12c0e562009-10-09 10:13:08 +0000[diff] [blame]171a JVM and a CLI Virtual Machine (Microsoft .NET is an
Nicolas Geoffray0dcee362009-10-09 10:17:14 +0000[diff] [blame]172implementation of the CLI) using LLVM for static and just-in-time
173compilation.</p>
Duncan Sands54fbb412009-06-24 08:38:48 +0000[diff] [blame]174
Nicolas Geoffray0dcee362009-10-09 10:17:14 +0000[diff] [blame]175<p>
176VMKit version 0.26 builds with LLVM 2.6 and you can find it on its
Duncan Sands54fbb412009-06-24 08:38:48 +0000[diff] [blame]177<a href="http://vmkit.llvm.org/releases/">webpage</a>. The release includes
178bug fixes, cleanup and new features. The major changes are:</p>
179
180<ul>
181
Nicolas Geoffray0dcee362009-10-09 10:17:14 +0000[diff] [blame]182<li>A new llcj tool to generate shared libraries or executables of Java
183 files.</li>
Nicolas Geoffray13eff6a2009-10-09 13:17:57 +0000[diff] [blame]184<li>Cooperative garbage collection. </li>
Nicolas Geoffray12c0e562009-10-09 10:13:08 +0000[diff] [blame]185<li>Fast subtype checking (paper from Click et al [JGI'02]). </li>
Nicolas Geoffray0dcee362009-10-09 10:17:14 +0000[diff] [blame]186<li>Implementation of a two-word header for Java objects instead of the orginal
187 three-word header. </li>
188<li>Better Java specification-compliance: division by zero checks, stack
189 overflow checks, finalization and references support. </li>
Duncan Sands54fbb412009-06-24 08:38:48 +0000[diff] [blame]190
191</ul>
192</div>
193
Chris Lattner6e6d33c2009-10-09 05:55:04 +0000[diff] [blame]194
195<!--=========================================================================-->
196<div class="doc_subsection">
197<a name="compiler-rt">compiler-rt: Compiler Runtime Library</a>
198</div>
199
200<div class="doc_text">
201<p>
202The new LLVM <a href="http://compiler-rt.llvm.org/">compiler-rt project</a>
203is a simple library that provides an implementation of the low-level
204target-specific hooks required by code generation and other runtime components.
205For example, when compiling for a 32-bit target, converting a double to a 64-bit
206unsigned integer is compiling into a runtime call to the "__fixunsdfdi"
Chris Lattner4d91d4d2009-10-10 18:33:13 +0000[diff] [blame]207function. The compiler-rt library provides highly optimized implementations of
208this and other low-level routines (some are 3x faster than the equivalent
209libgcc routines).</p>
Chris Lattner6e6d33c2009-10-09 05:55:04 +0000[diff] [blame]210
211<p>
212All of the code in the compiler-rt project is available under the standard LLVM
213License, a "BSD-style" license.</p>
214
215</div>
216
217<!--=========================================================================-->
218<div class="doc_subsection">
Chris Lattner51ead6d2009-10-10 19:00:55 +0000[diff] [blame^]219<a name="klee">KLEE: Symbolic Execution and Automatic Test Case Generator</a>
Chris Lattner6e6d33c2009-10-09 05:55:04 +0000[diff] [blame]220</div>
221
222<div class="doc_text">
223<p>
Chris Lattner51ead6d2009-10-10 19:00:55 +0000[diff] [blame^]224The new LLVM <a href="http://klee.llvm.org/">KLEE project</a> is a symbolic
225execution framework for programs in LLVM bitcode form. KLEE tries to
Chris Lattner6e6d33c2009-10-09 05:55:04 +0000[diff] [blame]226symbolically evaluate "all" paths through the application and records state
227transitions that lead to fault states. This allows it to construct testcases
228that lead to faults and can even be used to verify algorithms. For more
229details, please see the <a
230href="http://llvm.org/pubs/2008-12-OSDI-KLEE.html">OSDI 2008 paper</a> about
Chris Lattner51ead6d2009-10-10 19:00:55 +0000[diff] [blame^]231KLEE.</p>
Chris Lattner6e6d33c2009-10-09 05:55:04 +0000[diff] [blame]232
233</div>
234
235<!--=========================================================================-->
236<div class="doc_subsection">
237<a name="dragonegg">Dragon Egg: An LLVM backend plugin for GCC</a>
238</div>
239
240<div class="doc_text">
241<p>
242<b>Duncan needs to write me</b>.
243</p>
244
245</div>
246
247
Chris Lattner169c7012009-10-08 07:01:46 +0000[diff] [blame]248<!--=========================================================================-->
249<div class="doc_subsection">
250<a name="mc">llvm-mc: Machine Code Toolkit</a>
251</div>
252
253<div class="doc_text">
254<p>
Chris Lattner0d6011062009-10-09 06:24:25 +0000[diff] [blame]255The LLVM Machine Code (MC) Toolkit project is a (very early) effort to build
256better tools for dealing with machine code, object file formats, etc. The idea
257is to be able to generate most of the target specific details of assemblers and
258disassemblers from existing LLVM target .td files (with suitable enhancements),
259and to build infrastructure for reading and writing common object file formats.
260One of the first deliverables is to build a full assembler and integrate it into
261the compiler, which is predicted to substantially reduce compile time in some
262scenarios.
263</p>
Chris Lattner169c7012009-10-08 07:01:46 +0000[diff] [blame]264
Chris Lattner0d6011062009-10-09 06:24:25 +0000[diff] [blame]265<p>In the LLVM 2.6 timeframe, the MC framework has grown to the point where it
266can reliably parse and pretty print (with some encoding information) a
267darwin/x86 .s file successfully, and has the very early phases of a Mach-O
268assembler in progress. Beyond the MC framework itself, major refactoring of the
269LLVM code generator has started. The idea is to make the code generator reason
270about the code it is producing in a much more semantic way, rather than a
271textual way. For example, the code generator now uses MCSection objects to
272represent section assignments, instead of text strings that print to .section
273directives.</p>
274
275<p>MC is an early and ongoing project that will hopefully continue to lead to
276many improvements in the code generator and build infrastructure useful for many
277other situations.
Chris Lattner169c7012009-10-08 07:01:46 +0000[diff] [blame]278</p>
279
280</div>
281
282
Duncan Sands54fbb412009-06-24 08:38:48 +0000[diff] [blame]283<!-- *********************************************************************** -->
284<div class="doc_section">
285 <a name="externalproj">External Projects Using LLVM 2.6</a>
286</div>
287<!-- *********************************************************************** -->
288
Chris Lattner4d91d4d2009-10-10 18:33:13 +0000[diff] [blame]289<div class="doc_text">
290
291<p>An exciting aspect of LLVM is that it is used as an enabling technology for
292 a lot of other language and tools projects. This section lists some of the
293 projects that have already been updated to work with LLVM 2.6.</p>
294</div>
295
296
Chris Lattner169c7012009-10-08 07:01:46 +0000[diff] [blame]297<!--=========================================================================-->
298<div class="doc_subsection">
299<a name="Rubinius">Rubinius</a>
300</div>
301
302<div class="doc_text">
303<p><a href="http://github.com/evanphx/rubinius">Rubinius</a> is an environment
304for running Ruby code which strives to write as much of the core class
305implementation in Ruby as possible. Combined with a bytecode interpreting VM, it
306uses LLVM to optimize and compile ruby code down to machine code. Techniques
307such as type feedback, method inlining, and uncommon traps are all used to
308remove dynamism from ruby execution and increase performance.</p>
Chris Lattner59a30272009-10-08 16:01:33 +0000[diff] [blame]309
310<p>Since LLVM 2.5, Rubinius has made several major leaps forward, implementing
311a counter based JIT, type feedback, and speculative method inlining.
312</p>
313
Chris Lattner169c7012009-10-08 07:01:46 +0000[diff] [blame]314</div>
Chris Lattner2375deb2009-09-30 06:27:22 +0000[diff] [blame]315
316<!--=========================================================================-->
317<div class="doc_subsection">
318<a name="macruby">MacRuby</a>
319</div>
320
321<div class="doc_text">
322
323<p>
324<a href="http://macruby.org">MacRuby</a> is an implementation of Ruby on top of
325core Mac OS X technologies, such as the Objective-C common runtime and garbage
326collector, and the CoreFoundation framework. It is principally developed by
327Apple and aims at enabling the creation of full-fledged Mac OS X applications.
328</p>
329
330<p>
331MacRuby uses LLVM for optimization passes, JIT and AOT compilation of Ruby
332expressions. It also uses zero-cost DWARF exceptions to implement Ruby exception
333handling.</p>
334
335</div>
336
337
Duncan Sands54fbb412009-06-24 08:38:48 +0000[diff] [blame]338<!--=========================================================================-->
339<div class="doc_subsection">
340<a name="pure">Pure</a>
341</div>
342
343<div class="doc_text">
344<p>
345<a href="http://pure-lang.googlecode.com/">Pure</a>
346is an algebraic/functional programming language based on term rewriting.
347Programs are collections of equations which are used to evaluate expressions in
348a symbolic fashion. Pure offers dynamic typing, eager and lazy evaluation,
349lexical closures, a hygienic macro system (also based on term rewriting),
350built-in list and matrix support (including list and matrix comprehensions) and
351an easy-to-use C interface. The interpreter uses LLVM as a backend to
352 JIT-compile Pure programs to fast native code.</p>
353
Chris Lattner59a30272009-10-08 16:01:33 +0000[diff] [blame]354<p>Pure versions 0.31 and later have been tested and are known to work with
355LLVM 2.6 (and continue to work with older LLVM releases >= 2.3 as well).
Duncan Sands54fbb412009-06-24 08:38:48 +0000[diff] [blame]356</p>
357</div>
358
359
360<!--=========================================================================-->
361<div class="doc_subsection">
362<a name="ldc">LLVM D Compiler</a>
363</div>
364
365<div class="doc_text">
366<p>
367<a href="http://www.dsource.org/projects/ldc">LDC</a> is an implementation of
368the D Programming Language using the LLVM optimizer and code generator.
369The LDC project works great with the LLVM 2.6 release. General improvements in
370this
371cycle have included new inline asm constraint handling, better debug info
372support, general bugfixes, and better x86-64 support. This has allowed
373some major improvements in LDC, getting us much closer to being as
374fully featured as the original DMD compiler from DigitalMars.
375</p>
376</div>
377
378<!--=========================================================================-->
379<div class="doc_subsection">
380<a name="RoadsendPHP">Roadsend PHP</a>
381</div>
382
383<div class="doc_text">
Chris Lattner169c7012009-10-08 07:01:46 +0000[diff] [blame]384<p>
Chris Lattner169c7012009-10-08 07:01:46 +0000[diff] [blame]385<a href="http://code.roadsend.com/rphp">Roadsend PHP</a> (rphp) is an open
Duncan Sands54fbb412009-06-24 08:38:48 +0000[diff] [blame]386source implementation of the PHP programming
387language that uses LLVM for its optimizer, JIT, and static compiler. This is a
388reimplementation of an earlier project that is now based on LLVM.</p>
389</div>
390
Jeffrey Yasskin3a07d2f2009-06-24 21:09:13 +0000[diff] [blame]391<!--=========================================================================-->
392<div class="doc_subsection">
Chris Lattner6e6d33c2009-10-09 05:55:04 +0000[diff] [blame]393<a name="UnladenSwallow">Unladen Swallow</a>
Jeffrey Yasskin3a07d2f2009-06-24 21:09:13 +0000[diff] [blame]394</div>
395
396<div class="doc_text">
Chris Lattner169c7012009-10-08 07:01:46 +0000[diff] [blame]397<p>
Chris Lattner169c7012009-10-08 07:01:46 +0000[diff] [blame]398<a href="http://code.google.com/p/unladen-swallow/">Unladen Swallow</a> is a
Jeffrey Yasskin3a07d2f2009-06-24 21:09:13 +0000[diff] [blame]399branch of <a href="http://python.org/">Python</a> intended to be fully
400compatible and significantly faster. It uses LLVM's optimization passes and JIT
401compiler.</p>
402</div>
403
Chris Lattner6e6d33c2009-10-09 05:55:04 +0000[diff] [blame]404<!--=========================================================================-->
405<div class="doc_subsection">
406<a name="llvm-lua">llvm-lua</a>
407</div>
408
409<div class="doc_text">
410<p>
411<a href="http://code.google.com/p/llvm-lua/">LLVM-Lua</a> uses LLVM to add JIT
412&amp; static compiling support to the Lua VM. Lua bytecode is analyzed to
413remove type checks, then LLVM is used to compile those bytecodes down to machine
414code.</p>
415</div>
416
Jeffrey Yasskin123b3922009-06-24 21:26:42 +0000[diff] [blame]417
Duncan Sands54fbb412009-06-24 08:38:48 +0000[diff] [blame]418
419<!-- *********************************************************************** -->
420<div class="doc_section">
421 <a name="whatsnew">What's New in LLVM 2.6?</a>
422</div>
423<!-- *********************************************************************** -->
424
425<div class="doc_text">
426
427<p>This release includes a huge number of bug fixes, performance tweaks, and
428minor improvements. Some of the major improvements and new features are listed
429in this section.
430</p>
Chris Lattner6aaf6902009-10-08 06:27:53 +0000[diff] [blame]431
Duncan Sands54fbb412009-06-24 08:38:48 +0000[diff] [blame]432</div>
433
434<!--=========================================================================-->
435<div class="doc_subsection">
436<a name="majorfeatures">Major New Features</a>
437</div>
438
439<div class="doc_text">
440
441<p>LLVM 2.6 includes several major new capabilities:</p>
442
443<ul>
Chris Lattner51ead6d2009-10-10 19:00:55 +0000[diff] [blame^]444<li>New <a href="#compiler-rt">compiler-rt</a>, <A href="#klee">KLEE</a>,
Chris Lattner0d6011062009-10-09 06:24:25 +0000[diff] [blame]445 and <a href="#mc">machine code toolkit</a> sub-projects.</li>
446<li>Debug information now includes line numbers when optimizations are enabled.
447 This allows statistical sampling tools like oprofile and Shark to map
448 samples back to source lines.</li>
449<li>LLVM now includes new experimental backends to support the MSP430, SystemZ,
Chris Lattner6e6d33c2009-10-09 05:55:04 +0000[diff] [blame]450 and BlackFin architectures.</li>
451<li>LLVM supports a new <a href="GoldPlugin.html">Gold Linker Plugin</a> which
452 enables support for <a href="LinkTimeOptimization.html">transparent
453 link-time optimization</a> on ELF targets when used with the Gold binutils
454 linker.</li>
Chris Lattner4d91d4d2009-10-10 18:33:13 +0000[diff] [blame]455<li>LLVM now supports doing optimization and code generation on multiple
456 threads. Please see the <a href="ProgrammersManual.html#threading">LLVM
457 Programmer's Manual</a> for more information.</li>
Chris Lattner896d2582009-10-09 06:36:25 +0000[diff] [blame]458<li>LLVM now has experimental support for <a
459 href="http://nondot.org/~sabre/LLVMNotes/EmbeddedMetadata.txt">embedded
460 metadata</a> in LLVM IR, though the implementation is not guaranteed to be
461 final and the .bc file format may change in future releases. Debug info
Chris Lattner4d91d4d2009-10-10 18:33:13 +0000[diff] [blame]462 does not yet use this format in LLVM 2.6.</li>
Duncan Sands54fbb412009-06-24 08:38:48 +0000[diff] [blame]463</ul>
464
465</div>
466
Duncan Sands54fbb412009-06-24 08:38:48 +0000[diff] [blame]467<!--=========================================================================-->
468<div class="doc_subsection">
469<a name="coreimprovements">LLVM IR and Core Improvements</a>
470</div>
471
472<div class="doc_text">
Chris Lattner4d91d4d2009-10-10 18:33:13 +0000[diff] [blame]473<p>LLVM IR has several new features for better support of new targets and that
474expose new optimization opportunities:</p>
Duncan Sands54fbb412009-06-24 08:38:48 +0000[diff] [blame]475
476<ul>
Chris Lattner4d91d4d2009-10-10 18:33:13 +0000[diff] [blame]477<li>The <a href="LangRef.html#i_add">add</a>, <a
478 href="LangRef.html#i_sub">sub</a>, and <a href="LangRef.html#i_mul">mul</a>
479 instructions have been split into integer and floating point version (like
480 divide and remainder), introducing new <a
481 href="LangRef.html#i_fadd">fadd</a>, <a href="LangRef.html#i_fsub">fsub</a>,
482 and <a href="LangRef.html#i_fmul">fmul</a> instructions.</li>
483<li>The <a href="LangRef.html#i_add">add</a>, <a
484 href="LangRef.html#i_sub">sub</a>, and <a href="LangRef.html#i_mul">mul</a>
485 instructions now support optional "nsw" and "nuw" bits which indicate that
486 the operation is guaranteed to not overflow (in the signed or
487 unsigned case, respectively). This gives the optimizer more information and
488 can be used for things C signed integer values, which are undefined on
489 overflow.</li>
490<li>The <a href="LangRef.html#i_sdiv">sdiv</a> instruction now supports an
491 optional "exact" flag which indicates that the result of the division is
492 guaranteed to have a remainder of zero. This is useful to optimize pointer
493 subtraction in C.</li>
494<li>The <a href="LangRef.html#i_getelementptr">getelementptr</a> instruction now
495 supports arbitrary integer index values for array/pointer indices. This
496 allows for better better code generation on 16-bit targets like PIC16.</li>
497<li>The <a href="LangRef.html#i_getelementptr">getelementptr</a> instruction now
498 supports an "inbounds" optimization hint that tells the optimizer that the
499 pointer is guaranteed to be within its allocated object.</li>
500<li>LLVM now support a series of new linkage types for global values which allow
501 for better optimization and new capabilities:
502 <ul>
503 <li><a href="LangRef.html#linkage_linkonce">linkonce_odr</a> and
504 <a href="LangRef.html#linkage_weak">weak_odr</a> have the same linkage
505 semantics as the non-"odr" linkage types. The difference is that these
506 linkage types indicate that all definitions of the specified function
507 are guaranteed to have the same semantics. This allows inlining
508 templates functions in C++ but not inlining weak functions in C,
509 which previously both got the same linkage type.</li>
510 <li><a href="LangRef.html#linkage_available_externally">available_externally
511 </a> is a new linkage type that gives the optimizer visibility into the
512 definition of a function (allowing inlining and side effect analysis)
513 but that does not cause code to be generated. This allows better
514 optimization of "GNU inline" functions, extern templates, etc.</li>
515 <li><a href="LangRef.html#linkage_linker_private">linker_private</a> is a
516 new linkage type (which is only useful on Mac OS X) that is used for
517 some metadata generation and other obscure things.</li>
518 </ul></li>
519<li>Finally, target-specific intrinsics can now return multiple values, which
520 is useful for modeling target operations with multiple results.</li>
Duncan Sands54fbb412009-06-24 08:38:48 +0000[diff] [blame]521</ul>
522
523</div>
524
525<!--=========================================================================-->
526<div class="doc_subsection">
527<a name="optimizer">Optimizer Improvements</a>
528</div>
529
530<div class="doc_text">
531
Chris Lattner51ead6d2009-10-10 19:00:55 +0000[diff] [blame^]532<p>In addition to a large array of minor performance tweaks and bug fixes, this
Duncan Sands54fbb412009-06-24 08:38:48 +0000[diff] [blame]533release includes a few major enhancements and additions to the optimizers:</p>
534
535<ul>
536
Chris Lattner51ead6d2009-10-10 19:00:55 +0000[diff] [blame^]537<li>The <a href="Passes.html#scalarrepl">Scalar Replacement of Aggregates</a>
538 pass has many improvements that allow it to better promote vector unions,
539 variables which are memset, and much more strange code that can happen do
540 to bitfield accesses to register operations. An interesting change is that
541 it now produces "unusual" integer sizes (like i1704) in some cases and lets
542 other optimizers clean things up.</li>
543<li>The <a href="Passes.html#loop-reduce">Loop Strength Reduction</a> pass now
544 promotes small integer induction variables to 64-bit on 64-bit targets,
545 which provides a major performance boost many for numerical code. It also
546 promotes shorts to int on 32-bit hosts, etc. LSR now also analyzes pointer
547 expressions (e.g. getelementptrs), as well as integers.</li>
548<li>The <a href="Passes.html#gvn">GVN</a> pass now eliminates partial
549 redundancies of loads in simple cases.</li>
550<li>The <a href="Passes.html#inline">Inliner</a> now reuses stack space when
551 inlining similiar arrays from multiple callees into one caller.</li>
552<li>LLVM includes a new experimental Static Single Information (SSI)
553 construction pass.</li>
Chris Lattner6aaf6902009-10-08 06:27:53 +0000[diff] [blame]554</li>
Duncan Sands54fbb412009-06-24 08:38:48 +0000[diff] [blame]555
556</ul>
557
558</div>
559
560<!--=========================================================================-->
561<div class="doc_subsection">
562<a name="codegen">Target Independent Code Generator Improvements</a>
563</div>
564
565<div class="doc_text">
566
567<p>We have put a significant amount of work into the code generator
568infrastructure, which allows us to implement more aggressive algorithms and make
569it run faster:</p>
570
571<ul>
572
Chris Lattner51ead6d2009-10-10 19:00:55 +0000[diff] [blame^]573<li>The <tt>llc -asm-verbose</tt> option (exposed from llvm-gcc and clang as
574 <tt>-fverbose-asm</tt>) now adds a lot of useful information in comments to
575 the generated .s file. This information includes location information (if
576 built with <tt>-g</tt>) and loop nest information.</li>
577<li>The code generator now supports a new MachineVerifier pass which is useful
578 for finding bugs in targets and ccodegen passes.</li>
579<li>The Machine LICM is now enabled by default. It hoists instructions out of
580 loops (such as constant pool loads, loads from readonly stubs, vector
581 constant synthesization code, etc) and is currently configured to only do so
582 when the hoisted operation can be rematerialized.</li>
583<li>The Machine Sinking pass is now enabled by default. This pass moves
584 side-effect free operations down the CFG so that they are executed on fewer
585 paths through a function.</li>
Chris Lattner6aaf6902009-10-08 06:27:53 +0000[diff] [blame]586<li>Tblgen now supports multiclass inheritance and a number of new string and
587 list operations like !(subst), !(foreach), !car, !cdr, !null, !if, !cast.
588 These make the .td files more expressive and allow more aggressive factoring
589 of duplication across instruction patterns.</li>
Chris Lattner51ead6d2009-10-10 19:00:55 +0000[diff] [blame^]590<li>Target-specific intrinsics can now be added without having to hack VMCore to
591 add them. This makes it easier to maintain out-of-tree targets.</li>
Chris Lattner6aaf6902009-10-08 06:27:53 +0000[diff] [blame]592<li>Regalloc improvements for commuting, various spiller peephole optimizations, cross-class coalescing.</li>
593<li><tt>llc -enable-value-prop</tt>, propagation of value info (sign/zero ext info) from one MBB to another</li>
594<li>Regalloc hints for allocation stuff: Evan r73381/r73671. Finished/enabled?</li>
595<li>Stack slot coloring for register spills (denser stack frames)</li>
596<li>SelectionDAGS: New BuildVectorSDNode (r65296), and ISD::VECTOR_SHUFFLE (r69952 / PR2957)</li>
Chris Lattner51ead6d2009-10-10 19:00:55 +0000[diff] [blame^]597<li>The Prolog/Epilog Insertion Pass now has experimental support for performing
598 the "shrink wrapping" optimization, which moves spills and reloads around in
599 the CFG to avoid doing saves on paths that don't need them.</li>
600<li>LLVM includes new experimental support for writing ELF .o files directly
601 from the compiler. It works well for many simple C testcases, but doesn't
602 support exception handling, debug info, inline assembly, etc.</li>
Chris Lattner6e6d33c2009-10-09 05:55:04 +0000[diff] [blame]603<li>Targets can now specify register allocation hints through
Chris Lattner51ead6d2009-10-10 19:00:55 +0000[diff] [blame^]604 MachineRegisterInfo:: setRegAllocationHint. A regalloc hint consists of hint
605 type and physical register number. A hint type of zero specifies a register
606 allocation preference. Other hint type values are target specific which are
607 resolved by TargetRegisterInfo::ResolveRegAllocHint. An example of which is
608 the ARM target can uses register hint to request that the register allocator
609 provide an even / odd register pair to two virtual registers.</li>
Duncan Sands54fbb412009-06-24 08:38:48 +0000[diff] [blame]610</ul>
611</div>
612
613<!--=========================================================================-->
614<div class="doc_subsection">
615<a name="x86">X86-32 and X86-64 Target Improvements</a>
616</div>
617
618<div class="doc_text">
619<p>New features of the X86 target include:
620</p>
621
622<ul>
623
Chris Lattner6aaf6902009-10-08 06:27:53 +0000[diff] [blame]624<li>Preliminary support for addrspace 256 -> GS, 257 -> FS, known problems: CodeGenerator.html#x86_memory</li>
625<li>Support for softfloat modes, typically used by OS kernels.</li>
626
627<li>X86-64: better modeling of implicit zero extensions, eliminates a lot of redundant zexts</li>
628<li>X86-64 TLS support for local exec and initial exec.</li>
629<li>Better modeling of H registerts as subregs.</li>
630<li>Vector icmp/fcmp now work with SSE codegen.</li>
631<li>SSE 4.2 support.</li>
632<li>all global variable reference logic is now in ClassifyGlobalReference.</li>
633</li>
Duncan Sands54fbb412009-06-24 08:38:48 +0000[diff] [blame]634</ul>
635
636</div>
637
638<!--=========================================================================-->
639<div class="doc_subsection">
640<a name="pic16">PIC16 Target Improvements</a>
641</div>
642
643<div class="doc_text">
644<p>New features of the PIC16 target include:
645</p>
646
647<ul>
Chris Lattner6aaf6902009-10-08 06:27:53 +0000[diff] [blame]648<li>Support for floating-point, indirect function calls, and
649 passing/returning aggregate types to functions.
650<li>The code generator is able to generate debug info into output COFF files.
651<li>Support for placing an object into a specific section or at a specific
652 address in memory.</li>
Duncan Sands54fbb412009-06-24 08:38:48 +0000[diff] [blame]653</ul>
654
655<p>Things not yet supported:</p>
656
657<ul>
Duncan Sands54fbb412009-06-24 08:38:48 +0000[diff] [blame]658<li>Variable arguments.</li>
Duncan Sands54fbb412009-06-24 08:38:48 +0000[diff] [blame]659<li>Interrupts/programs.</li>
Duncan Sands54fbb412009-06-24 08:38:48 +0000[diff] [blame]660</ul>
661
662</div>
663
Bob Wilson80333842009-08-12 21:19:49 +0000[diff] [blame]664<!--=========================================================================-->
665<div class="doc_subsection">
666<a name="ARM">ARM Target Improvements</a>
667</div>
668
669<div class="doc_text">
670<p>New features of the ARM target include:
671</p>
672
673<ul>
674
675<li>Preliminary support for processors, such as the Cortex-A8 and Cortex-A9,
Sandeep Patel16eac042009-08-20 15:01:16 +0000[diff] [blame]676that implement version v7-A of the ARM architecture. The ARM backend now
677supports both the Thumb2 and Advanced SIMD (Neon) instruction sets. The
678AAPCS-VFP "hard float" calling conventions are also supported with the
679<tt>-float-abi=hard</tt> flag. These features are still somewhat experimental
680and subject to change. The Neon intrinsics, in particular, may change in future
681releases of LLVM.
Bob Wilson80333842009-08-12 21:19:49 +0000[diff] [blame]682</li>
Chris Lattner6aaf6902009-10-08 06:27:53 +0000[diff] [blame]683
684 ARM AAPCS-VFP hard float ABI is supported.
685 ARM calling convention code is now tblgen generated instead of manual.
686 ARM: NEON support. neonfp for doing single precision fp with neon instead of VFP.
687
Bob Wilson80333842009-08-12 21:19:49 +0000[diff] [blame]688</ul>
689
690</div>
691
Duncan Sands54fbb412009-06-24 08:38:48 +0000[diff] [blame]692<!--=========================================================================-->
693<div class="doc_subsection">
Chris Lattner6aaf6902009-10-08 06:27:53 +0000[diff] [blame]694<a name="OtherTarget">Other Target Specific Improvements</a>
Duncan Sands54fbb412009-06-24 08:38:48 +0000[diff] [blame]695</div>
696
697<div class="doc_text">
Chris Lattner6aaf6902009-10-08 06:27:53 +0000[diff] [blame]698<p>New features of other targets include:
699</p>
Duncan Sands54fbb412009-06-24 08:38:48 +0000[diff] [blame]700
701<ul>
Chris Lattner6aaf6902009-10-08 06:27:53 +0000[diff] [blame]702<li>Mips now supports O32 Calling Convention.</li>
Chris Lattner896d2582009-10-09 06:36:25 +0000[diff] [blame]703<li>Many improvements to the 32-bit PowerPC SVR4 ABI (used on powerpc-linux)
704 support, lots of bugs fixed.</li>
705<li>Added support for the 64-bit PowerPC SVR4 ABI (used on powerpc64-linux).
706 Needs more testing.</li>
Chris Lattner6aaf6902009-10-08 06:27:53 +0000[diff] [blame]707</ul>
708
709</div>
710
711<!--=========================================================================-->
712<div class="doc_subsection">
Chris Lattner0d6011062009-10-09 06:24:25 +0000[diff] [blame]713<a name="executionengine">Interpreter and JIT Improvements</a>
714</div>
715
716<div class="doc_text">
717
718<ul>
719<li>The JIT now supports generating more than 16M of code.</li>
720<li>When configured with --with-oprofile, the JIT can now inform oprofile about
721 JIT'd code, allowing oprofile to get line number and function name
722 information for JIT'd functions.</li>
723<li>When "libffi" is available, the LLVM interpreter now uses it, which supports
724 calling almost arbitrary external (natively compiled) functions.</li>
725<li>Clients of the JIT can now register a 'JITEventListener' object to receive
726 callbacks when the JIT emits or frees machine code. The OProfile support
727 uses this mechanism.</li>
728</ul>
729
730</div>
731
732
733<!--=========================================================================-->
734<div class="doc_subsection">
Chris Lattner6aaf6902009-10-08 06:27:53 +0000[diff] [blame]735<a name="newapis">New Useful APIs</a>
736</div>
737
738<div class="doc_text">
739
740<ul>
741<li>New EngineBuilder class for creating JITs: r76276</li>
742 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.
743 StringRef class, Twine class.
744 New WeakVH and AssertingVH and CallbackVH classes.
745 New llvm/ADT/Triple class.
746 llvm_report_error() error handling API (llvm/Support/ErrorHandling.h)
747 New llvm/System/Atomic.h, llvm/System/RWMutex.h for portable atomic ops, rw locks.
748 New SourceMgr, SMLoc classes for simple parsers with caret diagnostics and #include support, (used by
749 tablegen, llvm-mc, the .ll parser, FileCheck, etc)
750
Chris Lattner0d6011062009-10-09 06:24:25 +0000[diff] [blame]751</ul>
Chris Lattner6aaf6902009-10-08 06:27:53 +0000[diff] [blame]752
753
754</div>
755
756<!--=========================================================================-->
757<div class="doc_subsection">
758<a name="otherimprovements">Other Improvements and New Features</a>
759</div>
760
761<div class="doc_text">
762<p>Other miscellaneous features include:</p>
763
764<ul>
Chris Lattner0d6011062009-10-09 06:24:25 +0000[diff] [blame]765<li>LLVM now includes a new internal '<a
766 href="http://llvm.org/cmds/FileCheck.html">FileCheck</a>' tool which allows
767 writing much more accurate regression tests that run faster. Please see the
768 <a href="TestingGuide.html#FileCheck">FileCheck section of the Testing
769 Guide</a> for more information.</li>
Chris Lattner6e6d33c2009-10-09 05:55:04 +0000[diff] [blame]770<li>LLVM profile information support has been significantly improved to produce
771correct use counts, and has support for edge profiling with reduced runtime
772overhead. Combined, the generated profile information is both more correct and
773imposes about half as much overhead (2.6. from 12% to 6% overhead on SPEC
774CPU2000).</li>
Chris Lattner169c7012009-10-08 07:01:46 +0000[diff] [blame]775<li>Many extensions to the C APIs.</li>
Chris Lattner0d6011062009-10-09 06:24:25 +0000[diff] [blame]776<li>LLVM 2.6 includes a brand new experimental LLVM bindings to the Ada2005
777programming language.</li>
Chris Lattner169c7012009-10-08 07:01:46 +0000[diff] [blame]778
779<li>LLVMC:
780
781* Dynamic plugins now work on Windows.
782* New option property: init. Makes possible to provide default values for
783 options defined in plugins (interface to cl::init).
784* New example: Skeleton, shows how to create a standalone LLVMC-based driver.
785* New example: mcc16, a driver for the PIC16 toolchain.</li>
786
Duncan Sands54fbb412009-06-24 08:38:48 +0000[diff] [blame]787</ul>
788
789</div>
790
791
792<!--=========================================================================-->
793<div class="doc_subsection">
794<a name="changes">Major Changes and Removed Features</a>
795</div>
796
797<div class="doc_text">
798
799<p>If you're already an LLVM user or developer with out-of-tree changes based
Owen Andersonbe3fe4e2009-07-02 16:48:38 +0000[diff] [blame]800on LLVM 2.5, this section lists some "gotchas" that you may run into upgrading
Duncan Sands54fbb412009-06-24 08:38:48 +0000[diff] [blame]801from the previous release.</p>
802
803<ul>
804
Chris Lattner6aaf6902009-10-08 06:27:53 +0000[diff] [blame]805<li>The Itanium (IA64) backend has been removed. It was not supported and
806 bitrotted.</li>
807<li>The BigBlock register allocator has been removed, it also bitrotted.</li>
808<li>The C Backend (-march=c) is no longer considered part of the LLVM release
809criteria. We still want it to work, but no one is maintaining it and it lacks
810support for arbitrary precision integers and other important IR features.</li>
Duncan Sands54fbb412009-06-24 08:38:48 +0000[diff] [blame]811</ul>
812
Chris Lattner169c7012009-10-08 07:01:46 +0000[diff] [blame]813 LLVM build now builds all libraries as .a files instead of some
814 libraries as relinked .o files. This requires some APIs like
815 InitializeAllTargets.h. TargetRegistry!
816
817
818
Duncan Sands54fbb412009-06-24 08:38:48 +0000[diff] [blame]819
820<p>In addition, many APIs have changed in this release. Some of the major LLVM
821API changes are:</p>
822
Chris Lattner6aaf6902009-10-08 06:27:53 +0000[diff] [blame]823
824 API Cleanup:
825 no use of hash_set/hash_map, no more llvm::OStream
826 Use raw_ostream for everything, killed off llvm/Streams.h and DOUT
827
Duncan Sands54fbb412009-06-24 08:38:48 +0000[diff] [blame]828<ul>
Owen Andersonbe3fe4e2009-07-02 16:48:38 +0000[diff] [blame]829<li>LLVM's global uniquing tables for <tt>Type</tt>s and <tt>Constant</tt>s have
830 been privatized into members of an <tt>LLVMContext</tt>. A number of APIs
831 now take an <tt>LLVMContext</tt> as a parameter. To smooth the transition
832 for clients that will only ever use a single context, the new
833 <tt>getGlobalContext()</tt> API can be used to access a default global
834 context which can be passed in any and all cases where a context is
835 required.
Duncan Sands54fbb412009-06-24 08:38:48 +0000[diff] [blame]836<li>The <tt>getABITypeSize</tt> methods are now called <tt>getAllocSize</tt>.</li>
Dan Gohman79537c92009-07-07 20:05:15 +0000[diff] [blame]837<li>The <tt>Add</tt>, <tt>Sub</tt>, and <tt>Mul</tt> operators are no longer
838 overloaded for floating-point types. Floating-point addition, subtraction,
839 and multiplication are now represented with new operators <tt>FAdd</tt>,
840 <tt>FSub</tt>, and <tt>FMul</tt>. In the <tt>IRBuilder</tt> API,
841 <tt>CreateAdd</tt>, <tt>CreateSub</tt>, <tt>CreateMul</tt>, and
842 <tt>CreateNeg</tt> should only be used for integer arithmetic now;
843 <tt>CreateFAdd</tt>, <tt>CreateFSub</tt>, <tt>CreateFMul</tt>, and
844 <tt>CreateFNeg</tt> should now be used for floating-point arithmetic.</li>
Daniel Dunbara7d9e052009-07-12 20:41:27 +0000[diff] [blame]845<li>The DynamicLibrary class can no longer be constructed, its functionality has
846 moved to static member functions.</li>
Dan Gohmanb6b11702009-07-15 19:59:19 +0000[diff] [blame]847<li><tt>raw_fd_ostream</tt>'s constructor for opening a given filename now
848 takes an extra <tt>Force</tt> argument. If <tt>Force</tt> is set to
849 <tt>false</tt>, an error will be reported if a file with the given name
850 already exists. If <tt>Force</tt> is set to <tt>true</tt>, the file will
851 be silently truncated (which is the behavior before this flag was
852 added).</li>
Edwin Török07768e22009-07-21 20:27:10 +0000[diff] [blame]853<li><tt>SCEVHandle</tt> no longer exists, because reference counting is no
854longer done for <tt>SCEV*</tt> objects, instead <tt>const SCEV*</tt> should be
855used.</li>
Daniel Dunbarbbfb87a2009-07-25 05:26:53 +0000[diff] [blame]856
Daniel Dunbare3572ba2009-07-25 04:41:11 +0000[diff] [blame]857<li>Many APIs, notably <tt>llvm::Value</tt>, now use the <tt>StringRef</tt>
858and <tt>Twine</tt> classes instead of passing <tt>const char*</tt>
859or <tt>std::string</tt>, as described in
860the <a href="ProgrammersManual.html#string_apis">Programmer's Manual</a>. Most
Daniel Dunbarbbfb87a2009-07-25 05:26:53 +0000[diff] [blame]861clients 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:
862 <ul>
863 <li><tt>getNameStr()</tt> is still available, and matches the old
864 behavior. Replacing <tt>getName()</tt> calls with this is an safe option,
865 although more efficient alternatives are now possible.</li>
866
867 <li>If you were just relying on <tt>getName()</tt> being able to be sent to
868 a <tt>std::ostream</tt>, consider migrating
869 to <tt>llvm::raw_ostream</tt>.</li>
870
871 <li>If you were using <tt>getName().c_str()</tt> to get a <tt>const
872 char*</tt> pointer to the name, you can use <tt>getName().data()</tt>.
873 Note that this string (as before), may not be the entire name if the
874 name containts embedded null characters.</li>
875
876 <li>If you were using operator plus on the result of <tt>getName()</tt> and
877 treating the result as an <tt>std::string</tt>, you can either
878 uses <tt>Twine::str</tt> to get the result as an <tt>std::string</tt>, or
879 could move to a <tt>Twine</tt> based design.</li>
Daniel Dunbare03513b2009-07-25 23:55:21 +0000[diff] [blame]880
881 <li><tt>isName()</tt> should be replaced with comparison
Benjamin Kramer7b2136d2009-08-05 15:42:44 +0000[diff] [blame]882 against <tt>getName()</tt> (this is now efficient).
Daniel Dunbarbbfb87a2009-07-25 05:26:53 +0000[diff] [blame]883 </ul>
884</li>
885
Daniel Dunbar48224ee2009-07-26 02:12:58 +0000[diff] [blame]886<li>The registration interfaces for backend Targets has changed (what was
Daniel Dunbarc9a70092009-07-26 05:41:39 +0000[diff] [blame]887previously TargetMachineRegistry). For backend authors, see the <a href="WritingAnLLVMBackend.html#TargetRegistration">Writing An LLVM Backend</a> guide. For clients, the notable API changes are:
888 <ul>
889 <li><tt>TargetMachineRegistry</tt> has been renamed
890 to <tt>TargetRegistry</tt>.</li>
891
892 <li>Clients should move to using the <tt>TargetRegistry::lookupTarget()</tt>
893 function to find targets.</li>
894 </ul>
895</li>
Daniel Dunbarbbfb87a2009-07-25 05:26:53 +0000[diff] [blame]896
Edwin Török07768e22009-07-21 20:27:10 +0000[diff] [blame]897<li>llvm-dis now fails if output file exists, instead of dumping to stdout.
Daniel Dunbare3572ba2009-07-25 04:41:11 +0000[diff] [blame]898FIXME: describe any other tool changes due to the raw_fd_ostream change. FIXME:
899This is not an API change, maybe there should be a tool changes section?</li>
Edwin Török07768e22009-07-21 20:27:10 +0000[diff] [blame]900<li>temporarely due to Context API change passes should call doInitialization()
901method of the pass they inherit from, otherwise Context is NULL.
902FIXME: remove this entry when this is no longer needed.<li>
Duncan Sands54fbb412009-06-24 08:38:48 +0000[diff] [blame]903</ul>
904
905</div>
906
907
908
909<!-- *********************************************************************** -->
910<div class="doc_section">
911 <a name="portability">Portability and Supported Platforms</a>
912</div>
913<!-- *********************************************************************** -->
914
915<div class="doc_text">
916
917<p>LLVM is known to work on the following platforms:</p>
918
919<ul>
920<li>Intel and AMD machines (IA32, X86-64, AMD64, EMT-64) running Red Hat
Chris Lattnera8f23072009-07-21 22:47:03 +0000[diff] [blame]921Linux, Fedora Core, FreeBSD and AuroraUX (and probably other unix-like systems).</li>
Duncan Sands54fbb412009-06-24 08:38:48 +0000[diff] [blame]922<li>PowerPC and X86-based Mac OS X systems, running 10.3 and above in 32-bit
923and 64-bit modes.</li>
924<li>Intel and AMD machines running on Win32 using MinGW libraries (native).</li>
925<li>Intel and AMD machines running on Win32 with the Cygwin libraries (limited
926 support is available for native builds with Visual C++).</li>
927<li>Sun UltraSPARC workstations running Solaris 10.</li>
928<li>Alpha-based machines running Debian GNU/Linux.</li>
Duncan Sands54fbb412009-06-24 08:38:48 +0000[diff] [blame]929</ul>
930
931<p>The core LLVM infrastructure uses GNU autoconf to adapt itself
932to the machine and operating system on which it is built. However, minor
933porting may be required to get LLVM to work on new platforms. We welcome your
934portability patches and reports of successful builds or error messages.</p>
935
936</div>
937
938<!-- *********************************************************************** -->
939<div class="doc_section">
940 <a name="knownproblems">Known Problems</a>
941</div>
942<!-- *********************************************************************** -->
943
944<div class="doc_text">
945
946<p>This section contains significant known problems with the LLVM system,
947listed by component. If you run into a problem, please check the <a
948href="http://llvm.org/bugs/">LLVM bug database</a> and submit a bug if
949there isn't already one.</p>
950
Chris Lattnera8f23072009-07-21 22:47:03 +0000[diff] [blame]951<ul>
Chris Lattner4050d252009-07-21 23:17:26 +0000[diff] [blame]952<li>LLVM will not correctly compile on Solaris and/or OpenSolaris
Chris Lattnera8f23072009-07-21 22:47:03 +0000[diff] [blame]953using the stock GCC 3.x.x series 'out the box',
954See: <a href="#brokengcc">Broken versions of GCC and other tools</a>.
955However, A <a href="http://pkg.auroraux.org/GCC">Modern GCC Build</a>
956for x86/x64 has been made available from the third party AuroraUX Project
Chris Lattner6aaf6902009-10-08 06:27:53 +0000[diff] [blame]957that has been meticulously tested for bootstrapping LLVM &amp; Clang.</li>
Chris Lattnera8f23072009-07-21 22:47:03 +0000[diff] [blame]958</ul>
959
Duncan Sands54fbb412009-06-24 08:38:48 +0000[diff] [blame]960</div>
961
962<!-- ======================================================================= -->
963<div class="doc_subsection">
964 <a name="experimental">Experimental features included with this release</a>
965</div>
966
967<div class="doc_text">
968
969<p>The following components of this LLVM release are either untested, known to
970be broken or unreliable, or are in early development. These components should
971not be relied on, and bugs should not be filed against them, but they may be
972useful to some people. In particular, if you would like to work on one of these
973components, please contact us on the <a
974href="http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev">LLVMdev list</a>.</p>
975
976<ul>
Dan Gohman2a5ddf32009-07-24 00:30:09 +0000[diff] [blame]977<li>The MSIL, Alpha, SPU, MIPS, and PIC16 backends are experimental.</li>
Duncan Sands54fbb412009-06-24 08:38:48 +0000[diff] [blame]978<li>The <tt>llc</tt> "<tt>-filetype=asm</tt>" (the default) is the only
979 supported value for this option.</li>
980</ul>
981
982</div>
983
984<!-- ======================================================================= -->
985<div class="doc_subsection">
986 <a name="x86-be">Known problems with the X86 back-end</a>
987</div>
988
989<div class="doc_text">
990
991<ul>
992 <li>The X86 backend does not yet support
993 all <a href="http://llvm.org/PR879">inline assembly that uses the X86
994 floating point stack</a>. It supports the 'f' and 't' constraints, but not
995 'u'.</li>
996 <li>The X86 backend generates inefficient floating point code when configured
997 to generate code for systems that don't have SSE2.</li>
998 <li>Win64 code generation wasn't widely tested. Everything should work, but we
999 expect small issues to happen. Also, llvm-gcc cannot build the mingw64
1000 runtime currently due
1001 to <a href="http://llvm.org/PR2255">several</a>
1002 <a href="http://llvm.org/PR2257">bugs</a> and due to lack of support for
1003 the
1004 'u' inline assembly constraint and for X87 floating point inline assembly.</li>
1005 <li>The X86-64 backend does not yet support the LLVM IR instruction
1006 <tt>va_arg</tt>. Currently, the llvm-gcc and front-ends support variadic
1007 argument constructs on X86-64 by lowering them manually.</li>
1008</ul>
1009
1010</div>
1011
1012<!-- ======================================================================= -->
1013<div class="doc_subsection">
1014 <a name="ppc-be">Known problems with the PowerPC back-end</a>
1015</div>
1016
1017<div class="doc_text">
1018
1019<ul>
1020<li>The Linux PPC32/ABI support needs testing for the interpreter and static
1021compilation, and lacks support for debug information.</li>
1022</ul>
1023
1024</div>
1025
1026<!-- ======================================================================= -->
1027<div class="doc_subsection">
1028 <a name="arm-be">Known problems with the ARM back-end</a>
1029</div>
1030
1031<div class="doc_text">
1032
1033<ul>
Bob Wilson80333842009-08-12 21:19:49 +0000[diff] [blame]1034<li>Support for the Advanced SIMD (Neon) instruction set is still incomplete
1035and not well tested. Some features may not work at all, and the code quality
1036may be poor in some cases.</li>
Duncan Sands54fbb412009-06-24 08:38:48 +0000[diff] [blame]1037<li>Thumb mode works only on ARMv6 or higher processors. On sub-ARMv6
1038processors, thumb programs can crash or produce wrong
1039results (<a href="http://llvm.org/PR1388">PR1388</a>).</li>
1040<li>Compilation for ARM Linux OABI (old ABI) is supported but not fully tested.
1041</li>
Duncan Sands54fbb412009-06-24 08:38:48 +0000[diff] [blame]1042</ul>
1043
1044</div>
1045
1046<!-- ======================================================================= -->
1047<div class="doc_subsection">
1048 <a name="sparc-be">Known problems with the SPARC back-end</a>
1049</div>
1050
1051<div class="doc_text">
1052
1053<ul>
1054<li>The SPARC backend only supports the 32-bit SPARC ABI (-m32); it does not
1055 support the 64-bit SPARC ABI (-m64).</li>
1056</ul>
1057
1058</div>
1059
1060<!-- ======================================================================= -->
1061<div class="doc_subsection">
1062 <a name="mips-be">Known problems with the MIPS back-end</a>
1063</div>
1064
1065<div class="doc_text">
1066
1067<ul>
1068<li>The O32 ABI is not fully supported.</li>
1069<li>64-bit MIPS targets are not supported yet.</li>
1070</ul>
1071
1072</div>
1073
1074<!-- ======================================================================= -->
1075<div class="doc_subsection">
1076 <a name="alpha-be">Known problems with the Alpha back-end</a>
1077</div>
1078
1079<div class="doc_text">
1080
1081<ul>
1082
1083<li>On 21164s, some rare FP arithmetic sequences which may trap do not have the
1084appropriate nops inserted to ensure restartability.</li>
1085
1086</ul>
1087</div>
1088
1089<!-- ======================================================================= -->
1090<div class="doc_subsection">
Duncan Sands54fbb412009-06-24 08:38:48 +0000[diff] [blame]1091 <a name="c-be">Known problems with the C back-end</a>
1092</div>
1093
1094<div class="doc_text">
1095
1096<ul>
1097<li><a href="http://llvm.org/PR802">The C backend has only basic support for
1098 inline assembly code</a>.</li>
1099<li><a href="http://llvm.org/PR1658">The C backend violates the ABI of common
1100 C++ programs</a>, preventing intermixing between C++ compiled by the CBE and
1101 C++ code compiled with <tt>llc</tt> or native compilers.</li>
1102<li>The C backend does not support all exception handling constructs.</li>
1103<li>The C backend does not support arbitrary precision integers.</li>
1104</ul>
1105
1106</div>
1107
1108
1109<!-- ======================================================================= -->
1110<div class="doc_subsection">
1111 <a name="c-fe">Known problems with the llvm-gcc C front-end</a>
1112</div>
1113
1114<div class="doc_text">
1115
1116<p>llvm-gcc does not currently support <a href="http://llvm.org/PR869">Link-Time
1117Optimization</a> on most platforms "out-of-the-box". Please inquire on the
1118LLVMdev mailing list if you are interested.</p>
1119
1120<p>The only major language feature of GCC not supported by llvm-gcc is
1121 the <tt>__builtin_apply</tt> family of builtins. However, some extensions
1122 are only supported on some targets. For example, trampolines are only
1123 supported on some targets (these are used when you take the address of a
1124 nested function).</p>
1125
1126<p>If you run into GCC extensions which are not supported, please let us know.
1127</p>
1128
1129</div>
1130
1131<!-- ======================================================================= -->
1132<div class="doc_subsection">
1133 <a name="c++-fe">Known problems with the llvm-gcc C++ front-end</a>
1134</div>
1135
1136<div class="doc_text">
1137
1138<p>The C++ front-end is considered to be fully
1139tested and works for a number of non-trivial programs, including LLVM
1140itself, Qt, Mozilla, etc.</p>
1141
1142<ul>
1143<li>Exception handling works well on the X86 and PowerPC targets. Currently
1144 only Linux and Darwin targets are supported (both 32 and 64 bit).</li>
1145</ul>
1146
1147</div>
1148
1149<!-- ======================================================================= -->
1150<div class="doc_subsection">
1151 <a name="fortran-fe">Known problems with the llvm-gcc Fortran front-end</a>
1152</div>
1153
1154<div class="doc_text">
1155<ul>
1156<li>Fortran support generally works, but there are still several unresolved bugs
1157 in Bugzilla. Please see the tools/gfortran component for details.</li>
1158</ul>
1159</div>
1160
1161<!-- ======================================================================= -->
1162<div class="doc_subsection">
1163 <a name="ada-fe">Known problems with the llvm-gcc Ada front-end</a>
1164</div>
1165
1166<div class="doc_text">
1167The llvm-gcc 4.2 Ada compiler works fairly well; however, this is not a mature
1168technology, and problems should be expected.
1169<ul>
1170<li>The Ada front-end currently only builds on X86-32. This is mainly due
1171to lack of trampoline support (pointers to nested functions) on other platforms.
1172However, it <a href="http://llvm.org/PR2006">also fails to build on X86-64</a>
1173which does support trampolines.</li>
1174<li>The Ada front-end <a href="http://llvm.org/PR2007">fails to bootstrap</a>.
1175This is due to lack of LLVM support for <tt>setjmp</tt>/<tt>longjmp</tt> style
1176exception handling, which is used internally by the compiler.
1177Workaround: configure with --disable-bootstrap.</li>
1178<li>The c380004, <a href="http://llvm.org/PR2010">c393010</a>
1179and <a href="http://llvm.org/PR2421">cxg2021</a> ACATS tests fail
1180(c380004 also fails with gcc-4.2 mainline).
1181If the compiler is built with checks disabled then <a href="http://llvm.org/PR2010">c393010</a>
1182causes the compiler to go into an infinite loop, using up all system memory.</li>
1183<li>Some GCC specific Ada tests continue to crash the compiler.</li>
1184<li>The -E binder option (exception backtraces)
1185<a href="http://llvm.org/PR1982">does not work</a> and will result in programs
1186crashing if an exception is raised. Workaround: do not use -E.</li>
1187<li>Only discrete types <a href="http://llvm.org/PR1981">are allowed to start
1188or finish at a non-byte offset</a> in a record. Workaround: do not pack records
1189or use representation clauses that result in a field of a non-discrete type
1190starting or finishing in the middle of a byte.</li>
1191<li>The <tt>lli</tt> interpreter <a href="http://llvm.org/PR2009">considers
1192'main' as generated by the Ada binder to be invalid</a>.
1193Workaround: hand edit the file to use pointers for <tt>argv</tt> and
1194<tt>envp</tt> rather than integers.</li>
1195<li>The <tt>-fstack-check</tt> option <a href="http://llvm.org/PR2008">is
1196ignored</a>.</li>
1197</ul>
1198</div>
1199
Erick Tryzelaarb4cf9752009-09-28 04:42:55 +0000[diff] [blame]1200<!-- ======================================================================= -->
1201<div class="doc_subsection">
1202 <a name="ocaml-bindingse">Known problems with the O'Caml bindings</a>
1203</div>
1204
1205<div class="doc_text">
1206
1207<p>The Llvm.Linkage module is broken, and has incorrect values. Only
1208Llvm.Linkage.External, Llvm.Linkage.Available_externally, and
1209Llvm.Linkage.Link_once will be correct. If you need any of the other linkage
1210modes, you'll have to write an external C library in order to expose the
1211functionality. This has been fixed in the trunk.</p>
1212</div>
1213
Duncan Sands54fbb412009-06-24 08:38:48 +0000[diff] [blame]1214<!-- *********************************************************************** -->
1215<div class="doc_section">
1216 <a name="additionalinfo">Additional Information</a>
1217</div>
1218<!-- *********************************************************************** -->
1219
1220<div class="doc_text">
1221
1222<p>A wide variety of additional information is available on the <a
1223href="http://llvm.org">LLVM web page</a>, in particular in the <a
1224href="http://llvm.org/docs/">documentation</a> section. The web page also
1225contains versions of the API documentation which is up-to-date with the
1226Subversion version of the source code.
1227You can access versions of these documents specific to this release by going
1228into the "<tt>llvm/doc/</tt>" directory in the LLVM tree.</p>
1229
1230<p>If you have any questions or comments about LLVM, please feel free to contact
1231us via the <a href="http://llvm.org/docs/#maillist"> mailing
1232lists</a>.</p>
1233
1234</div>
1235
1236<!-- *********************************************************************** -->
1237
1238<hr>
1239<address>
1240 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
1241 src="http://jigsaw.w3.org/css-validator/images/vcss-blue" alt="Valid CSS"></a>
1242 <a href="http://validator.w3.org/check/referer"><img
1243 src="http://www.w3.org/Icons/valid-html401-blue" alt="Valid HTML 4.01"></a>
1244
1245 <a href="http://llvm.org/">LLVM Compiler Infrastructure</a><br>
1246 Last modified: $Date$
1247</address>
1248
1249</body>
1250</html>