Gitiles
Code Review Sign In
gerrit-public.fairphone.software / fp2-dev / platform / external / llvm / 8767bdca785849ff566f409d4d3801f274686e3a / . / docs / ReleaseNotes-2.6.html
blob: c2c6c1eba34623f2a43e3c1dbd8bc6f75e5094c6 [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>
Duncan Sandsccedda42009-10-10 19:16:25 +0000[diff] [blame]149in C and Objective-C programs. The tool performs checks to find
150bugs that occur on a specific path within a program.
151The set of checks performed by the static analyzer continues to expand, and
Duncan Sands54fbb412009-06-24 08:38:48 +0000[diff] [blame]152future plans for the tool include full source-level inter-procedural analysis
153and deeper checks such as buffer overrun detection. There are many opportunities
154to extend and enhance the static analyzer, and anyone interested in working on
155this project is encouraged to get involved!</p>
156
157</div>
158
159<!--=========================================================================-->
160<div class="doc_subsection">
161<a name="vmkit">VMKit: JVM/CLI Virtual Machine Implementation</a>
162</div>
163
164<div class="doc_text">
165<p>
166The <a href="http://vmkit.llvm.org/">VMKit project</a> is an implementation of
Nicolas Geoffray12c0e562009-10-09 10:13:08 +0000[diff] [blame]167a JVM and a CLI Virtual Machine (Microsoft .NET is an
Nicolas Geoffray0dcee362009-10-09 10:17:14 +0000[diff] [blame]168implementation of the CLI) using LLVM for static and just-in-time
169compilation.</p>
Duncan Sands54fbb412009-06-24 08:38:48 +0000[diff] [blame]170
Nicolas Geoffray0dcee362009-10-09 10:17:14 +0000[diff] [blame]171<p>
172VMKit version 0.26 builds with LLVM 2.6 and you can find it on its
Duncan Sandsdc5bd2b2009-10-10 19:30:16 +0000[diff] [blame]173<a href="http://vmkit.llvm.org/releases/">web page</a>. The release includes
Duncan Sands54fbb412009-06-24 08:38:48 +0000[diff] [blame]174bug fixes, cleanup and new features. The major changes are:</p>
175
176<ul>
177
Nicolas Geoffray0dcee362009-10-09 10:17:14 +0000[diff] [blame]178<li>A new llcj tool to generate shared libraries or executables of Java
179 files.</li>
Nicolas Geoffray13eff6a2009-10-09 13:17:57 +0000[diff] [blame]180<li>Cooperative garbage collection. </li>
Nicolas Geoffray12c0e562009-10-09 10:13:08 +0000[diff] [blame]181<li>Fast subtype checking (paper from Click et al [JGI'02]). </li>
Duncan Sandsdc5bd2b2009-10-10 19:30:16 +0000[diff] [blame]182<li>Implementation of a two-word header for Java objects instead of the original
Nicolas Geoffray0dcee362009-10-09 10:17:14 +0000[diff] [blame]183 three-word header. </li>
184<li>Better Java specification-compliance: division by zero checks, stack
185 overflow checks, finalization and references support. </li>
Duncan Sands54fbb412009-06-24 08:38:48 +0000[diff] [blame]186
187</ul>
188</div>
189
Chris Lattner6e6d33c2009-10-09 05:55:04 +0000[diff] [blame]190
191<!--=========================================================================-->
192<div class="doc_subsection">
193<a name="compiler-rt">compiler-rt: Compiler Runtime Library</a>
194</div>
195
196<div class="doc_text">
197<p>
198The new LLVM <a href="http://compiler-rt.llvm.org/">compiler-rt project</a>
199is a simple library that provides an implementation of the low-level
200target-specific hooks required by code generation and other runtime components.
201For example, when compiling for a 32-bit target, converting a double to a 64-bit
Duncan Sandsd4f53352009-10-10 20:06:04 +0000[diff] [blame]202unsigned integer is compiled into a runtime call to the "__fixunsdfdi"
Chris Lattner4d91d4d2009-10-10 18:33:13 +0000[diff] [blame]203function. The compiler-rt library provides highly optimized implementations of
204this and other low-level routines (some are 3x faster than the equivalent
205libgcc routines).</p>
Chris Lattner6e6d33c2009-10-09 05:55:04 +0000[diff] [blame]206
207<p>
208All of the code in the compiler-rt project is available under the standard LLVM
209License, a "BSD-style" license.</p>
210
211</div>
212
213<!--=========================================================================-->
214<div class="doc_subsection">
Chris Lattner51ead6d2009-10-10 19:00:55 +0000[diff] [blame]215<a name="klee">KLEE: Symbolic Execution and Automatic Test Case Generator</a>
Chris Lattner6e6d33c2009-10-09 05:55:04 +0000[diff] [blame]216</div>
217
218<div class="doc_text">
219<p>
Chris Lattner51ead6d2009-10-10 19:00:55 +0000[diff] [blame]220The new LLVM <a href="http://klee.llvm.org/">KLEE project</a> is a symbolic
221execution framework for programs in LLVM bitcode form. KLEE tries to
Chris Lattner6e6d33c2009-10-09 05:55:04 +0000[diff] [blame]222symbolically evaluate "all" paths through the application and records state
223transitions that lead to fault states. This allows it to construct testcases
224that lead to faults and can even be used to verify algorithms. For more
225details, please see the <a
226href="http://llvm.org/pubs/2008-12-OSDI-KLEE.html">OSDI 2008 paper</a> about
Chris Lattner51ead6d2009-10-10 19:00:55 +0000[diff] [blame]227KLEE.</p>
Chris Lattner6e6d33c2009-10-09 05:55:04 +0000[diff] [blame]228
229</div>
230
231<!--=========================================================================-->
232<div class="doc_subsection">
233<a name="dragonegg">Dragon Egg: An LLVM backend plugin for GCC</a>
234</div>
235
236<div class="doc_text">
237<p>
238<b>Duncan needs to write me</b>.
239</p>
240
241</div>
242
243
Chris Lattner169c7012009-10-08 07:01:46 +0000[diff] [blame]244<!--=========================================================================-->
245<div class="doc_subsection">
246<a name="mc">llvm-mc: Machine Code Toolkit</a>
247</div>
248
249<div class="doc_text">
250<p>
Chris Lattner0d6011062009-10-09 06:24:25 +0000[diff] [blame]251The LLVM Machine Code (MC) Toolkit project is a (very early) effort to build
252better tools for dealing with machine code, object file formats, etc. The idea
253is to be able to generate most of the target specific details of assemblers and
254disassemblers from existing LLVM target .td files (with suitable enhancements),
255and to build infrastructure for reading and writing common object file formats.
256One of the first deliverables is to build a full assembler and integrate it into
257the compiler, which is predicted to substantially reduce compile time in some
258scenarios.
259</p>
Chris Lattner169c7012009-10-08 07:01:46 +0000[diff] [blame]260
Chris Lattner0d6011062009-10-09 06:24:25 +0000[diff] [blame]261<p>In the LLVM 2.6 timeframe, the MC framework has grown to the point where it
262can reliably parse and pretty print (with some encoding information) a
263darwin/x86 .s file successfully, and has the very early phases of a Mach-O
264assembler in progress. Beyond the MC framework itself, major refactoring of the
265LLVM code generator has started. The idea is to make the code generator reason
266about the code it is producing in a much more semantic way, rather than a
267textual way. For example, the code generator now uses MCSection objects to
268represent section assignments, instead of text strings that print to .section
269directives.</p>
270
271<p>MC is an early and ongoing project that will hopefully continue to lead to
272many improvements in the code generator and build infrastructure useful for many
273other situations.
Chris Lattner169c7012009-10-08 07:01:46 +0000[diff] [blame]274</p>
275
276</div>
277
278
Duncan Sands54fbb412009-06-24 08:38:48 +0000[diff] [blame]279<!-- *********************************************************************** -->
280<div class="doc_section">
281 <a name="externalproj">External Projects Using LLVM 2.6</a>
282</div>
283<!-- *********************************************************************** -->
284
Chris Lattner4d91d4d2009-10-10 18:33:13 +0000[diff] [blame]285<div class="doc_text">
286
287<p>An exciting aspect of LLVM is that it is used as an enabling technology for
288 a lot of other language and tools projects. This section lists some of the
289 projects that have already been updated to work with LLVM 2.6.</p>
290</div>
291
292
Chris Lattner169c7012009-10-08 07:01:46 +0000[diff] [blame]293<!--=========================================================================-->
294<div class="doc_subsection">
295<a name="Rubinius">Rubinius</a>
296</div>
297
298<div class="doc_text">
299<p><a href="http://github.com/evanphx/rubinius">Rubinius</a> is an environment
300for running Ruby code which strives to write as much of the core class
301implementation in Ruby as possible. Combined with a bytecode interpreting VM, it
302uses LLVM to optimize and compile ruby code down to machine code. Techniques
303such as type feedback, method inlining, and uncommon traps are all used to
304remove dynamism from ruby execution and increase performance.</p>
Chris Lattner59a30272009-10-08 16:01:33 +0000[diff] [blame]305
306<p>Since LLVM 2.5, Rubinius has made several major leaps forward, implementing
307a counter based JIT, type feedback, and speculative method inlining.
308</p>
309
Chris Lattner169c7012009-10-08 07:01:46 +0000[diff] [blame]310</div>
Chris Lattner2375deb2009-09-30 06:27:22 +0000[diff] [blame]311
312<!--=========================================================================-->
313<div class="doc_subsection">
314<a name="macruby">MacRuby</a>
315</div>
316
317<div class="doc_text">
318
319<p>
320<a href="http://macruby.org">MacRuby</a> is an implementation of Ruby on top of
321core Mac OS X technologies, such as the Objective-C common runtime and garbage
322collector, and the CoreFoundation framework. It is principally developed by
323Apple and aims at enabling the creation of full-fledged Mac OS X applications.
324</p>
325
326<p>
327MacRuby uses LLVM for optimization passes, JIT and AOT compilation of Ruby
328expressions. It also uses zero-cost DWARF exceptions to implement Ruby exception
329handling.</p>
330
331</div>
332
333
Duncan Sands54fbb412009-06-24 08:38:48 +0000[diff] [blame]334<!--=========================================================================-->
335<div class="doc_subsection">
336<a name="pure">Pure</a>
337</div>
338
339<div class="doc_text">
340<p>
341<a href="http://pure-lang.googlecode.com/">Pure</a>
342is an algebraic/functional programming language based on term rewriting.
343Programs are collections of equations which are used to evaluate expressions in
344a symbolic fashion. Pure offers dynamic typing, eager and lazy evaluation,
345lexical closures, a hygienic macro system (also based on term rewriting),
346built-in list and matrix support (including list and matrix comprehensions) and
347an easy-to-use C interface. The interpreter uses LLVM as a backend to
348 JIT-compile Pure programs to fast native code.</p>
349
Chris Lattner59a30272009-10-08 16:01:33 +0000[diff] [blame]350<p>Pure versions 0.31 and later have been tested and are known to work with
351LLVM 2.6 (and continue to work with older LLVM releases >= 2.3 as well).
Duncan Sands54fbb412009-06-24 08:38:48 +0000[diff] [blame]352</p>
353</div>
354
355
356<!--=========================================================================-->
357<div class="doc_subsection">
358<a name="ldc">LLVM D Compiler</a>
359</div>
360
361<div class="doc_text">
362<p>
363<a href="http://www.dsource.org/projects/ldc">LDC</a> is an implementation of
364the D Programming Language using the LLVM optimizer and code generator.
365The LDC project works great with the LLVM 2.6 release. General improvements in
366this
367cycle have included new inline asm constraint handling, better debug info
Duncan Sandsdc5bd2b2009-10-10 19:30:16 +0000[diff] [blame]368support, general bug fixes, and better x86-64 support. This has allowed
Duncan Sands54fbb412009-06-24 08:38:48 +0000[diff] [blame]369some major improvements in LDC, getting us much closer to being as
370fully featured as the original DMD compiler from DigitalMars.
371</p>
372</div>
373
374<!--=========================================================================-->
375<div class="doc_subsection">
376<a name="RoadsendPHP">Roadsend PHP</a>
377</div>
378
379<div class="doc_text">
Chris Lattner169c7012009-10-08 07:01:46 +0000[diff] [blame]380<p>
Chris Lattner169c7012009-10-08 07:01:46 +0000[diff] [blame]381<a href="http://code.roadsend.com/rphp">Roadsend PHP</a> (rphp) is an open
Duncan Sands54fbb412009-06-24 08:38:48 +0000[diff] [blame]382source implementation of the PHP programming
383language that uses LLVM for its optimizer, JIT, and static compiler. This is a
384reimplementation of an earlier project that is now based on LLVM.</p>
385</div>
386
Jeffrey Yasskin3a07d2f2009-06-24 21:09:13 +0000[diff] [blame]387<!--=========================================================================-->
388<div class="doc_subsection">
Chris Lattner6e6d33c2009-10-09 05:55:04 +0000[diff] [blame]389<a name="UnladenSwallow">Unladen Swallow</a>
Jeffrey Yasskin3a07d2f2009-06-24 21:09:13 +0000[diff] [blame]390</div>
391
392<div class="doc_text">
Chris Lattner169c7012009-10-08 07:01:46 +0000[diff] [blame]393<p>
Chris Lattner169c7012009-10-08 07:01:46 +0000[diff] [blame]394<a href="http://code.google.com/p/unladen-swallow/">Unladen Swallow</a> is a
Jeffrey Yasskin3a07d2f2009-06-24 21:09:13 +0000[diff] [blame]395branch of <a href="http://python.org/">Python</a> intended to be fully
396compatible and significantly faster. It uses LLVM's optimization passes and JIT
397compiler.</p>
398</div>
399
Chris Lattner6e6d33c2009-10-09 05:55:04 +0000[diff] [blame]400<!--=========================================================================-->
401<div class="doc_subsection">
402<a name="llvm-lua">llvm-lua</a>
403</div>
404
405<div class="doc_text">
406<p>
407<a href="http://code.google.com/p/llvm-lua/">LLVM-Lua</a> uses LLVM to add JIT
Duncan Sandsd4f53352009-10-10 20:06:04 +0000[diff] [blame]408and static compiling support to the Lua VM. Lua bytecode is analyzed to
409remove type checks, then LLVM is used to compile the bytecode down to machine
Chris Lattner6e6d33c2009-10-09 05:55:04 +0000[diff] [blame]410code.</p>
411</div>
412
Jeffrey Yasskin123b3922009-06-24 21:26:42 +0000[diff] [blame]413
Duncan Sands54fbb412009-06-24 08:38:48 +0000[diff] [blame]414
415<!-- *********************************************************************** -->
416<div class="doc_section">
417 <a name="whatsnew">What's New in LLVM 2.6?</a>
418</div>
419<!-- *********************************************************************** -->
420
421<div class="doc_text">
422
423<p>This release includes a huge number of bug fixes, performance tweaks, and
424minor improvements. Some of the major improvements and new features are listed
425in this section.
426</p>
Chris Lattner6aaf6902009-10-08 06:27:53 +0000[diff] [blame]427
Duncan Sands54fbb412009-06-24 08:38:48 +0000[diff] [blame]428</div>
429
430<!--=========================================================================-->
431<div class="doc_subsection">
432<a name="majorfeatures">Major New Features</a>
433</div>
434
435<div class="doc_text">
436
437<p>LLVM 2.6 includes several major new capabilities:</p>
438
439<ul>
Chris Lattner51ead6d2009-10-10 19:00:55 +0000[diff] [blame]440<li>New <a href="#compiler-rt">compiler-rt</a>, <A href="#klee">KLEE</a>,
Chris Lattner0d6011062009-10-09 06:24:25 +0000[diff] [blame]441 and <a href="#mc">machine code toolkit</a> sub-projects.</li>
442<li>Debug information now includes line numbers when optimizations are enabled.
443 This allows statistical sampling tools like oprofile and Shark to map
444 samples back to source lines.</li>
445<li>LLVM now includes new experimental backends to support the MSP430, SystemZ,
Chris Lattner6e6d33c2009-10-09 05:55:04 +0000[diff] [blame]446 and BlackFin architectures.</li>
447<li>LLVM supports a new <a href="GoldPlugin.html">Gold Linker Plugin</a> which
448 enables support for <a href="LinkTimeOptimization.html">transparent
449 link-time optimization</a> on ELF targets when used with the Gold binutils
450 linker.</li>
Chris Lattner4d91d4d2009-10-10 18:33:13 +0000[diff] [blame]451<li>LLVM now supports doing optimization and code generation on multiple
452 threads. Please see the <a href="ProgrammersManual.html#threading">LLVM
453 Programmer's Manual</a> for more information.</li>
Chris Lattner896d2582009-10-09 06:36:25 +0000[diff] [blame]454<li>LLVM now has experimental support for <a
455 href="http://nondot.org/~sabre/LLVMNotes/EmbeddedMetadata.txt">embedded
456 metadata</a> in LLVM IR, though the implementation is not guaranteed to be
457 final and the .bc file format may change in future releases. Debug info
Chris Lattner4d91d4d2009-10-10 18:33:13 +0000[diff] [blame]458 does not yet use this format in LLVM 2.6.</li>
Duncan Sands54fbb412009-06-24 08:38:48 +0000[diff] [blame]459</ul>
460
461</div>
462
Duncan Sands54fbb412009-06-24 08:38:48 +0000[diff] [blame]463<!--=========================================================================-->
464<div class="doc_subsection">
465<a name="coreimprovements">LLVM IR and Core Improvements</a>
466</div>
467
468<div class="doc_text">
Chris Lattner4d91d4d2009-10-10 18:33:13 +0000[diff] [blame]469<p>LLVM IR has several new features for better support of new targets and that
470expose new optimization opportunities:</p>
Duncan Sands54fbb412009-06-24 08:38:48 +0000[diff] [blame]471
472<ul>
Chris Lattner4d91d4d2009-10-10 18:33:13 +0000[diff] [blame]473<li>The <a href="LangRef.html#i_add">add</a>, <a
474 href="LangRef.html#i_sub">sub</a>, and <a href="LangRef.html#i_mul">mul</a>
Duncan Sandsd4f53352009-10-10 20:06:04 +0000[diff] [blame]475 instructions have been split into integer and floating point versions (like
Chris Lattner4d91d4d2009-10-10 18:33:13 +0000[diff] [blame]476 divide and remainder), introducing new <a
477 href="LangRef.html#i_fadd">fadd</a>, <a href="LangRef.html#i_fsub">fsub</a>,
478 and <a href="LangRef.html#i_fmul">fmul</a> instructions.</li>
479<li>The <a href="LangRef.html#i_add">add</a>, <a
480 href="LangRef.html#i_sub">sub</a>, and <a href="LangRef.html#i_mul">mul</a>
481 instructions now support optional "nsw" and "nuw" bits which indicate that
482 the operation is guaranteed to not overflow (in the signed or
483 unsigned case, respectively). This gives the optimizer more information and
Duncan Sandsd4f53352009-10-10 20:06:04 +0000[diff] [blame]484 can be used for things like C signed integer values, which are undefined on
Chris Lattner4d91d4d2009-10-10 18:33:13 +0000[diff] [blame]485 overflow.</li>
486<li>The <a href="LangRef.html#i_sdiv">sdiv</a> instruction now supports an
487 optional "exact" flag which indicates that the result of the division is
Duncan Sandsd4f53352009-10-10 20:06:04 +0000[diff] [blame]488 guaranteed to have a remainder of zero. This is useful for optimizing pointer
Chris Lattner4d91d4d2009-10-10 18:33:13 +0000[diff] [blame]489 subtraction in C.</li>
490<li>The <a href="LangRef.html#i_getelementptr">getelementptr</a> instruction now
491 supports arbitrary integer index values for array/pointer indices. This
Benjamin Kramer08f0add2009-10-10 19:45:13 +0000[diff] [blame]492 allows for better code generation on 16-bit targets like PIC16.</li>
Chris Lattner4d91d4d2009-10-10 18:33:13 +0000[diff] [blame]493<li>The <a href="LangRef.html#i_getelementptr">getelementptr</a> instruction now
494 supports an "inbounds" optimization hint that tells the optimizer that the
495 pointer is guaranteed to be within its allocated object.</li>
496<li>LLVM now support a series of new linkage types for global values which allow
497 for better optimization and new capabilities:
498 <ul>
499 <li><a href="LangRef.html#linkage_linkonce">linkonce_odr</a> and
500 <a href="LangRef.html#linkage_weak">weak_odr</a> have the same linkage
501 semantics as the non-"odr" linkage types. The difference is that these
502 linkage types indicate that all definitions of the specified function
503 are guaranteed to have the same semantics. This allows inlining
504 templates functions in C++ but not inlining weak functions in C,
505 which previously both got the same linkage type.</li>
506 <li><a href="LangRef.html#linkage_available_externally">available_externally
507 </a> is a new linkage type that gives the optimizer visibility into the
508 definition of a function (allowing inlining and side effect analysis)
509 but that does not cause code to be generated. This allows better
510 optimization of "GNU inline" functions, extern templates, etc.</li>
511 <li><a href="LangRef.html#linkage_linker_private">linker_private</a> is a
512 new linkage type (which is only useful on Mac OS X) that is used for
513 some metadata generation and other obscure things.</li>
514 </ul></li>
515<li>Finally, target-specific intrinsics can now return multiple values, which
516 is useful for modeling target operations with multiple results.</li>
Duncan Sands54fbb412009-06-24 08:38:48 +0000[diff] [blame]517</ul>
518
519</div>
520
521<!--=========================================================================-->
522<div class="doc_subsection">
523<a name="optimizer">Optimizer Improvements</a>
524</div>
525
526<div class="doc_text">
527
Chris Lattner51ead6d2009-10-10 19:00:55 +0000[diff] [blame]528<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]529release includes a few major enhancements and additions to the optimizers:</p>
530
531<ul>
532
Chris Lattner51ead6d2009-10-10 19:00:55 +0000[diff] [blame]533<li>The <a href="Passes.html#scalarrepl">Scalar Replacement of Aggregates</a>
534 pass has many improvements that allow it to better promote vector unions,
Duncan Sandsd4f53352009-10-10 20:06:04 +0000[diff] [blame]535 variables which are memset, and much more strange code that can happen to
536 do bitfield accesses to register operations. An interesting change is that
Chris Lattner51ead6d2009-10-10 19:00:55 +0000[diff] [blame]537 it now produces "unusual" integer sizes (like i1704) in some cases and lets
538 other optimizers clean things up.</li>
539<li>The <a href="Passes.html#loop-reduce">Loop Strength Reduction</a> pass now
540 promotes small integer induction variables to 64-bit on 64-bit targets,
Duncan Sandsd4f53352009-10-10 20:06:04 +0000[diff] [blame]541 which provides a major performance boost for much numerical code. It also
Chris Lattner51ead6d2009-10-10 19:00:55 +0000[diff] [blame]542 promotes shorts to int on 32-bit hosts, etc. LSR now also analyzes pointer
543 expressions (e.g. getelementptrs), as well as integers.</li>
544<li>The <a href="Passes.html#gvn">GVN</a> pass now eliminates partial
545 redundancies of loads in simple cases.</li>
546<li>The <a href="Passes.html#inline">Inliner</a> now reuses stack space when
Duncan Sandsdc5bd2b2009-10-10 19:30:16 +0000[diff] [blame]547 inlining similar arrays from multiple callees into one caller.</li>
Chris Lattner51ead6d2009-10-10 19:00:55 +0000[diff] [blame]548<li>LLVM includes a new experimental Static Single Information (SSI)
549 construction pass.</li>
Chris Lattner6aaf6902009-10-08 06:27:53 +0000[diff] [blame]550</li>
Duncan Sands54fbb412009-06-24 08:38:48 +0000[diff] [blame]551
552</ul>
553
554</div>
555
Chris Lattnerfdf24e72009-10-10 21:40:13 +0000[diff] [blame]556
557<!--=========================================================================-->
558<div class="doc_subsection">
559<a name="executionengine">Interpreter and JIT Improvements</a>
560</div>
561
562<div class="doc_text">
563
564<ul>
565<li>LLVM has a new "EngineBuilder" class which makes it more obvious how to
566 set up and configure an ExecutionEngine (a JIT or interpreter).</li>
567<li>The JIT now supports generating more than 16M of code.</li>
568<li>When configured with --with-oprofile, the JIT can now inform oprofile about
569 JIT'd code, allowing oprofile to get line number and function name
570 information for JIT'd functions.</li>
571<li>When "libffi" is available, the LLVM interpreter now uses it, which supports
572 calling almost arbitrary external (natively compiled) functions.</li>
573<li>Clients of the JIT can now register a 'JITEventListener' object to receive
574 callbacks when the JIT emits or frees machine code. The OProfile support
575 uses this mechanism.</li>
576</ul>
577
578</div>
579
Duncan Sands54fbb412009-06-24 08:38:48 +0000[diff] [blame]580<!--=========================================================================-->
581<div class="doc_subsection">
582<a name="codegen">Target Independent Code Generator Improvements</a>
583</div>
584
585<div class="doc_text">
586
587<p>We have put a significant amount of work into the code generator
588infrastructure, which allows us to implement more aggressive algorithms and make
589it run faster:</p>
590
591<ul>
592
Chris Lattner51ead6d2009-10-10 19:00:55 +0000[diff] [blame]593<li>The <tt>llc -asm-verbose</tt> option (exposed from llvm-gcc and clang as
594 <tt>-fverbose-asm</tt>) now adds a lot of useful information in comments to
595 the generated .s file. This information includes location information (if
596 built with <tt>-g</tt>) and loop nest information.</li>
597<li>The code generator now supports a new MachineVerifier pass which is useful
Duncan Sandsdc5bd2b2009-10-10 19:30:16 +0000[diff] [blame]598 for finding bugs in targets and codegen passes.</li>
Chris Lattner51ead6d2009-10-10 19:00:55 +0000[diff] [blame]599<li>The Machine LICM is now enabled by default. It hoists instructions out of
Duncan Sandsdc5bd2b2009-10-10 19:30:16 +0000[diff] [blame]600 loops (such as constant pool loads, loads from read-only stubs, vector
Chris Lattner51ead6d2009-10-10 19:00:55 +0000[diff] [blame]601 constant synthesization code, etc) and is currently configured to only do so
602 when the hoisted operation can be rematerialized.</li>
603<li>The Machine Sinking pass is now enabled by default. This pass moves
604 side-effect free operations down the CFG so that they are executed on fewer
605 paths through a function.</li>
Chris Lattnerfdf24e72009-10-10 21:40:13 +0000[diff] [blame]606<li>The code generator now performs "stack slot coloring" of register spills,
Chris Lattner10abb752009-10-10 19:26:21 +0000[diff] [blame]607 which allows spill slots to be reused. This leads to smaller stack frames
608 in cases where there are lots of register spills.</li>
609<li>The register allocator has many improvements to take better advantage of
610 commutable operations, various spiller peephole optimizations, and can now
611 coalesce cross-register-class copies.</li>
Chris Lattner6aaf6902009-10-08 06:27:53 +0000[diff] [blame]612<li>Tblgen now supports multiclass inheritance and a number of new string and
613 list operations like !(subst), !(foreach), !car, !cdr, !null, !if, !cast.
614 These make the .td files more expressive and allow more aggressive factoring
615 of duplication across instruction patterns.</li>
Chris Lattner51ead6d2009-10-10 19:00:55 +0000[diff] [blame]616<li>Target-specific intrinsics can now be added without having to hack VMCore to
617 add them. This makes it easier to maintain out-of-tree targets.</li>
Chris Lattner10abb752009-10-10 19:26:21 +0000[diff] [blame]618<li>The instruction selector is better at propagating information about values
619 (such as whether they are sign/zero extended etc) across basic block
620 boundaries.</li>
Chris Lattnerfdf24e72009-10-10 21:40:13 +0000[diff] [blame]621<li>The SelectionDAG datastructure has new nodes for representing buildvector
622 and <a href="http://llvm.org/PR2957">vector shuffle</a> operations. This
623 makes operations and pattern matching more efficient and easier to get
624 right.</li>
Chris Lattner51ead6d2009-10-10 19:00:55 +0000[diff] [blame]625<li>The Prolog/Epilog Insertion Pass now has experimental support for performing
626 the "shrink wrapping" optimization, which moves spills and reloads around in
627 the CFG to avoid doing saves on paths that don't need them.</li>
628<li>LLVM includes new experimental support for writing ELF .o files directly
629 from the compiler. It works well for many simple C testcases, but doesn't
630 support exception handling, debug info, inline assembly, etc.</li>
Chris Lattner6e6d33c2009-10-09 05:55:04 +0000[diff] [blame]631<li>Targets can now specify register allocation hints through
Chris Lattner51ead6d2009-10-10 19:00:55 +0000[diff] [blame]632 MachineRegisterInfo:: setRegAllocationHint. A regalloc hint consists of hint
633 type and physical register number. A hint type of zero specifies a register
634 allocation preference. Other hint type values are target specific which are
Duncan Sandsd4f53352009-10-10 20:06:04 +0000[diff] [blame]635 resolved by TargetRegisterInfo::ResolveRegAllocHint. An example is the ARM
636 target which uses register hints to request that the register allocator
Chris Lattner51ead6d2009-10-10 19:00:55 +0000[diff] [blame]637 provide an even / odd register pair to two virtual registers.</li>
Duncan Sands54fbb412009-06-24 08:38:48 +0000[diff] [blame]638</ul>
639</div>
640
641<!--=========================================================================-->
642<div class="doc_subsection">
643<a name="x86">X86-32 and X86-64 Target Improvements</a>
644</div>
645
646<div class="doc_text">
647<p>New features of the X86 target include:
648</p>
649
650<ul>
651
Chris Lattnerfdf24e72009-10-10 21:40:13 +0000[diff] [blame]652<li>SSE 4.2 builtins are now supported.</li>
653<li>GCC-compatible soft float modes are now supported, which are typically used
654 by OS kernels.</li>
655<li>X86-64 now models implicit zero extensions better, which allows the code
656 generator to remove a lot of redundant zexts. It also models the 8-bit "H"
657 registers as sugregs, which allows they to be used in some tricky
658 situations.</li>
659<li>X86-64 now supports the "local exec" and "initial exec" thread local storage
660 model.</li>
661<li>The vector forms of the <a href="LangRef.html#i_icmp">icmp</a> and <a
662 href="LangRef.html#i_fcmp">fcmp</a> instructions now select to efficient
663 SSE operations.</li>
664<li>The X86 backend has preliminary support for <a
665 href="CodeGenerator.html#x86_memory">mapping address spaces to segment
666 register references</a>. This allows you to write GS or FS relative memory
667 accesses directly in LLVM IR for cases where you know exactly what you're
668 doing (such as in an OS kernel). There are some known problems with this
669 support, but it works in simple cases.</li>
670<li>The X86 code generator has been refactored to move all global variable
671 reference logic to one place
672 (<tt>X86Subtarget::ClassifyGlobalReference</tt>) which
673 makes it easier to reason about.</li>
Chris Lattner6aaf6902009-10-08 06:27:53 +0000[diff] [blame]674</li>
Duncan Sands54fbb412009-06-24 08:38:48 +0000[diff] [blame]675</ul>
676
677</div>
678
679<!--=========================================================================-->
680<div class="doc_subsection">
681<a name="pic16">PIC16 Target Improvements</a>
682</div>
683
684<div class="doc_text">
685<p>New features of the PIC16 target include:
686</p>
687
688<ul>
Chris Lattner6aaf6902009-10-08 06:27:53 +0000[diff] [blame]689<li>Support for floating-point, indirect function calls, and
690 passing/returning aggregate types to functions.
691<li>The code generator is able to generate debug info into output COFF files.
692<li>Support for placing an object into a specific section or at a specific
693 address in memory.</li>
Duncan Sands54fbb412009-06-24 08:38:48 +0000[diff] [blame]694</ul>
695
696<p>Things not yet supported:</p>
697
698<ul>
Duncan Sands54fbb412009-06-24 08:38:48 +0000[diff] [blame]699<li>Variable arguments.</li>
Duncan Sands54fbb412009-06-24 08:38:48 +0000[diff] [blame]700<li>Interrupts/programs.</li>
Duncan Sands54fbb412009-06-24 08:38:48 +0000[diff] [blame]701</ul>
702
703</div>
704
Bob Wilson80333842009-08-12 21:19:49 +0000[diff] [blame]705<!--=========================================================================-->
706<div class="doc_subsection">
707<a name="ARM">ARM Target Improvements</a>
708</div>
709
710<div class="doc_text">
711<p>New features of the ARM target include:
712</p>
713
714<ul>
715
716<li>Preliminary support for processors, such as the Cortex-A8 and Cortex-A9,
Sandeep Patel16eac042009-08-20 15:01:16 +0000[diff] [blame]717that implement version v7-A of the ARM architecture. The ARM backend now
Chris Lattnerfdf24e72009-10-10 21:40:13 +0000[diff] [blame]718supports both the Thumb2 and Advanced SIMD (Neon) instruction sets.</li>
719
720<li>The AAPCS-VFP "hard float" calling conventions are also supported with the
721<tt>-float-abi=hard</tt> flag.</li>
722
723<li>The ARM calling convention code is now tblgen generated instead of C++
724 code.</li>
Bob Wilson80333842009-08-12 21:19:49 +0000[diff] [blame]725</li>
Chris Lattner6aaf6902009-10-08 06:27:53 +0000[diff] [blame]726
Chris Lattnerfdf24e72009-10-10 21:40:13 +0000[diff] [blame]727
728<p>These features are still somewhat experimental
729and subject to change. The Neon intrinsics, in particular, may change in future
730releases of LLVM.</p>
Chris Lattner6aaf6902009-10-08 06:27:53 +0000[diff] [blame]731
Bob Wilson80333842009-08-12 21:19:49 +0000[diff] [blame]732</ul>
733
734</div>
735
Duncan Sands54fbb412009-06-24 08:38:48 +0000[diff] [blame]736<!--=========================================================================-->
737<div class="doc_subsection">
Chris Lattner6aaf6902009-10-08 06:27:53 +0000[diff] [blame]738<a name="OtherTarget">Other Target Specific Improvements</a>
Duncan Sands54fbb412009-06-24 08:38:48 +0000[diff] [blame]739</div>
740
741<div class="doc_text">
Chris Lattner6aaf6902009-10-08 06:27:53 +0000[diff] [blame]742<p>New features of other targets include:
743</p>
Duncan Sands54fbb412009-06-24 08:38:48 +0000[diff] [blame]744
745<ul>
Chris Lattner6aaf6902009-10-08 06:27:53 +0000[diff] [blame]746<li>Mips now supports O32 Calling Convention.</li>
Chris Lattner896d2582009-10-09 06:36:25 +0000[diff] [blame]747<li>Many improvements to the 32-bit PowerPC SVR4 ABI (used on powerpc-linux)
748 support, lots of bugs fixed.</li>
749<li>Added support for the 64-bit PowerPC SVR4 ABI (used on powerpc64-linux).
750 Needs more testing.</li>
Chris Lattner6aaf6902009-10-08 06:27:53 +0000[diff] [blame]751</ul>
752
753</div>
754
755<!--=========================================================================-->
756<div class="doc_subsection">
757<a name="newapis">New Useful APIs</a>
758</div>
759
760<div class="doc_text">
761
Chris Lattner835f2532009-10-10 22:02:58 +0000[diff] [blame]762<p>This release includes a number of new APIs that are used internally, which
763 may also be useful for external clients.
764</p>
765
Chris Lattner6aaf6902009-10-08 06:27:53 +0000[diff] [blame]766<ul>
Chris Lattner835f2532009-10-10 22:02:58 +0000[diff] [blame]767<li>New <a href="http://llvm.org/doxygen/PrettyStackTrace_8h-source.html">
768 PrettyStackTrace classes</a> allows crashes of llvm tools (and applications
769 that integrate them) to provide more detailed indication of what the
770 compiler was doing at the time of the crash (e.g. running a pass).
771 At the top level for each LLVM tool, it includes the command line arguments.
772 </li>
773<li>New <a href="http://llvm.org/doxygen/StringRef_8h-source.html">StringRef</a>
774 and <a href="http://llvm.org/doxygen/Twine_8h-source.html">Twine</a> classes
775 make operations on character ranges and
776 string concatenation to be more efficient. StringRef is just a <tt>const
777 char*</tt> with a length, Twine is a light-weight rope.</li>
778<li>LLVM has new WeakVH, AssertingVH, and CallbackVH classes, which make it
779 easier to write LLVM IR transformations. WeakVH is automatically drops to
780 null when the referenced Value is deleted, and is updated across a
781 replaceAllUsesWith operation. AssertingVH aborts the program if the
782 referenced value is destroyed while it is being referenced. CallbackVH is
783 a customizable class for handling value references. See <a
784 href="http://llvm.org/doxygen/ValueHandle_8h-source.html">ValueHandle.h</a>
785 for more information.</li>
786<li>The new '<a href="http://llvm.org/doxygen/Triple_8h-source.html">Triple
787 </a>' class centralizes a lot of logic that reasons about target
788 triples.</li>
789<li>The new '<a href="http://llvm.org/doxygen/ErrorHandling.h_8h-source.html">
790 llvm_report_error()</a>' set of APIs allows tools to embed the LLVM
791 optimizer and backend and recover from previously unrecoverable errors.</li>
792<li>LLVM has new abstractions for <a
793 href="http://llvm.org/doxygen/Atomic_8h-source.html">atomic operations</a>
794 and <a href="http://llvm.org/doxygen/RWMutex_8h-source.html">reader/writer
795 locks</a>.</li>
796<li>LLVM has new <a href="http://llvm.org/doxygen/SourceMgr_8h-source.html">
797 SourceMgr and SMLoc classes</a> which implement caret diagnostics and basic
798 include stack processing for simple parsers. It is used by tablegen,
799 llvm-mc, the .ll parser and FileCheck.</li>
Chris Lattner0d6011062009-10-09 06:24:25 +0000[diff] [blame]800</ul>
Chris Lattner6aaf6902009-10-08 06:27:53 +0000[diff] [blame]801
802
803</div>
804
805<!--=========================================================================-->
806<div class="doc_subsection">
807<a name="otherimprovements">Other Improvements and New Features</a>
808</div>
809
810<div class="doc_text">
811<p>Other miscellaneous features include:</p>
812
813<ul>
Chris Lattner0d6011062009-10-09 06:24:25 +0000[diff] [blame]814<li>LLVM now includes a new internal '<a
815 href="http://llvm.org/cmds/FileCheck.html">FileCheck</a>' tool which allows
816 writing much more accurate regression tests that run faster. Please see the
817 <a href="TestingGuide.html#FileCheck">FileCheck section of the Testing
818 Guide</a> for more information.</li>
Chris Lattner6e6d33c2009-10-09 05:55:04 +0000[diff] [blame]819<li>LLVM profile information support has been significantly improved to produce
820correct use counts, and has support for edge profiling with reduced runtime
821overhead. Combined, the generated profile information is both more correct and
822imposes about half as much overhead (2.6. from 12% to 6% overhead on SPEC
823CPU2000).</li>
Chris Lattner835f2532009-10-10 22:02:58 +0000[diff] [blame]824<li>The C bindings (in the llvm/include/llvm-c directory) include many newly
825 supported APIs.</li>
Chris Lattner0d6011062009-10-09 06:24:25 +0000[diff] [blame]826<li>LLVM 2.6 includes a brand new experimental LLVM bindings to the Ada2005
Chris Lattner835f2532009-10-10 22:02:58 +0000[diff] [blame]827 programming language.</li>
Chris Lattner169c7012009-10-08 07:01:46 +0000[diff] [blame]828
Chris Lattner835f2532009-10-10 22:02:58 +0000[diff] [blame]829<li>The LLVMC driver has several new features:
830 <ul>
831 <li>Dynamic plugins now work on Windows.</li>
832 <li>New option property: init. Makes possible to provide default values for
833 options defined in plugins (interface to cl::init).</li>
834 <li>New example: Skeleton, shows how to create a standalone LLVMC-based
835 driver.</li>
836 <li>New example: mcc16, a driver for the PIC16 toolchain.</li>
837 </ul>
838</li>
Chris Lattner169c7012009-10-08 07:01:46 +0000[diff] [blame]839
Duncan Sands54fbb412009-06-24 08:38:48 +0000[diff] [blame]840</ul>
841
842</div>
843
844
845<!--=========================================================================-->
846<div class="doc_subsection">
847<a name="changes">Major Changes and Removed Features</a>
848</div>
849
850<div class="doc_text">
851
852<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]853on LLVM 2.5, this section lists some "gotchas" that you may run into upgrading
Duncan Sands54fbb412009-06-24 08:38:48 +0000[diff] [blame]854from the previous release.</p>
855
856<ul>
Chris Lattner8767bdc2009-10-10 22:13:38 +0000[diff] [blame^]857<li>The Itanium (IA64) backend has been removed. It was not actively supported
858 and had bitrotted.</li>
Duncan Sandsd4f53352009-10-10 20:06:04 +0000[diff] [blame]859<li>The BigBlock register allocator has been removed, it had also bitrotted.</li>
Chris Lattner6aaf6902009-10-08 06:27:53 +0000[diff] [blame]860<li>The C Backend (-march=c) is no longer considered part of the LLVM release
861criteria. We still want it to work, but no one is maintaining it and it lacks
862support for arbitrary precision integers and other important IR features.</li>
Duncan Sands54fbb412009-06-24 08:38:48 +0000[diff] [blame]863
Chris Lattner8767bdc2009-10-10 22:13:38 +0000[diff] [blame^]864<li>All LLVM tools now default to overwriting their output file, behaving more
865 like standard unix tools. Previously, this only happened with the '-f'
866 option.</li>
867<li>LLVM build now builds all libraries as .a files instead of some
Chris Lattner169c7012009-10-08 07:01:46 +0000[diff] [blame]868 libraries as relinked .o files. This requires some APIs like
Chris Lattner8767bdc2009-10-10 22:13:38 +0000[diff] [blame^]869 InitializeAllTargets.h.
870 </li>
871</ul>
Chris Lattner169c7012009-10-08 07:01:46 +0000[diff] [blame]872
Duncan Sands54fbb412009-06-24 08:38:48 +0000[diff] [blame]873
874<p>In addition, many APIs have changed in this release. Some of the major LLVM
875API changes are:</p>
876
877<ul>
Chris Lattner8767bdc2009-10-10 22:13:38 +0000[diff] [blame^]878<li>All uses of hash_set and hash_map have been removed from the LLVM tree and
879 the wrapper headers have been removed.</li>
880<li>The llvm/Streams.h and DOUT member of Debug.h have been removed. The
881 llvm::Ostream class has been completely removed and replaced with uses of
882 raw_ostream.</li>
Owen Andersonbe3fe4e2009-07-02 16:48:38 +0000[diff] [blame]883<li>LLVM's global uniquing tables for <tt>Type</tt>s and <tt>Constant</tt>s have
884 been privatized into members of an <tt>LLVMContext</tt>. A number of APIs
885 now take an <tt>LLVMContext</tt> as a parameter. To smooth the transition
886 for clients that will only ever use a single context, the new
887 <tt>getGlobalContext()</tt> API can be used to access a default global
888 context which can be passed in any and all cases where a context is
889 required.
Duncan Sands54fbb412009-06-24 08:38:48 +0000[diff] [blame]890<li>The <tt>getABITypeSize</tt> methods are now called <tt>getAllocSize</tt>.</li>
Dan Gohman79537c92009-07-07 20:05:15 +0000[diff] [blame]891<li>The <tt>Add</tt>, <tt>Sub</tt>, and <tt>Mul</tt> operators are no longer
892 overloaded for floating-point types. Floating-point addition, subtraction,
893 and multiplication are now represented with new operators <tt>FAdd</tt>,
894 <tt>FSub</tt>, and <tt>FMul</tt>. In the <tt>IRBuilder</tt> API,
895 <tt>CreateAdd</tt>, <tt>CreateSub</tt>, <tt>CreateMul</tt>, and
896 <tt>CreateNeg</tt> should only be used for integer arithmetic now;
897 <tt>CreateFAdd</tt>, <tt>CreateFSub</tt>, <tt>CreateFMul</tt>, and
898 <tt>CreateFNeg</tt> should now be used for floating-point arithmetic.</li>
Daniel Dunbara7d9e052009-07-12 20:41:27 +0000[diff] [blame]899<li>The DynamicLibrary class can no longer be constructed, its functionality has
900 moved to static member functions.</li>
Dan Gohmanb6b11702009-07-15 19:59:19 +0000[diff] [blame]901<li><tt>raw_fd_ostream</tt>'s constructor for opening a given filename now
902 takes an extra <tt>Force</tt> argument. If <tt>Force</tt> is set to
903 <tt>false</tt>, an error will be reported if a file with the given name
904 already exists. If <tt>Force</tt> is set to <tt>true</tt>, the file will
905 be silently truncated (which is the behavior before this flag was
906 added).</li>
Edwin Török07768e22009-07-21 20:27:10 +0000[diff] [blame]907<li><tt>SCEVHandle</tt> no longer exists, because reference counting is no
908longer done for <tt>SCEV*</tt> objects, instead <tt>const SCEV*</tt> should be
909used.</li>
Daniel Dunbarbbfb87a2009-07-25 05:26:53 +0000[diff] [blame]910
Daniel Dunbare3572ba2009-07-25 04:41:11 +0000[diff] [blame]911<li>Many APIs, notably <tt>llvm::Value</tt>, now use the <tt>StringRef</tt>
912and <tt>Twine</tt> classes instead of passing <tt>const char*</tt>
913or <tt>std::string</tt>, as described in
914the <a href="ProgrammersManual.html#string_apis">Programmer's Manual</a>. Most
Chris Lattner8767bdc2009-10-10 22:13:38 +0000[diff] [blame^]915clients should be unaffected by this transition, unless they are used to
916<tt>Value::getName()</tt> returning a string. Here are some tips on updating to
9172.6:
Daniel Dunbarbbfb87a2009-07-25 05:26:53 +0000[diff] [blame]918 <ul>
919 <li><tt>getNameStr()</tt> is still available, and matches the old
920 behavior. Replacing <tt>getName()</tt> calls with this is an safe option,
921 although more efficient alternatives are now possible.</li>
922
923 <li>If you were just relying on <tt>getName()</tt> being able to be sent to
924 a <tt>std::ostream</tt>, consider migrating
925 to <tt>llvm::raw_ostream</tt>.</li>
926
927 <li>If you were using <tt>getName().c_str()</tt> to get a <tt>const
928 char*</tt> pointer to the name, you can use <tt>getName().data()</tt>.
929 Note that this string (as before), may not be the entire name if the
Benjamin Kramer08f0add2009-10-10 19:45:13 +0000[diff] [blame]930 name contains embedded null characters.</li>
Daniel Dunbarbbfb87a2009-07-25 05:26:53 +0000[diff] [blame]931
932 <li>If you were using operator plus on the result of <tt>getName()</tt> and
933 treating the result as an <tt>std::string</tt>, you can either
Duncan Sandsd4f53352009-10-10 20:06:04 +0000[diff] [blame]934 use <tt>Twine::str</tt> to get the result as an <tt>std::string</tt>, or
Daniel Dunbarbbfb87a2009-07-25 05:26:53 +0000[diff] [blame]935 could move to a <tt>Twine</tt> based design.</li>
Daniel Dunbare03513b2009-07-25 23:55:21 +0000[diff] [blame]936
937 <li><tt>isName()</tt> should be replaced with comparison
Benjamin Kramer7b2136d2009-08-05 15:42:44 +0000[diff] [blame]938 against <tt>getName()</tt> (this is now efficient).
Daniel Dunbarbbfb87a2009-07-25 05:26:53 +0000[diff] [blame]939 </ul>
940</li>
941
Daniel Dunbar48224ee2009-07-26 02:12:58 +0000[diff] [blame]942<li>The registration interfaces for backend Targets has changed (what was
Chris Lattner8767bdc2009-10-10 22:13:38 +0000[diff] [blame^]943previously TargetMachineRegistry). For backend authors, see the <a
944href="WritingAnLLVMBackend.html#TargetRegistration">Writing An LLVM Backend</a>
945guide. For clients, the notable API changes are:
Daniel Dunbarc9a70092009-07-26 05:41:39 +0000[diff] [blame]946 <ul>
947 <li><tt>TargetMachineRegistry</tt> has been renamed
948 to <tt>TargetRegistry</tt>.</li>
949
950 <li>Clients should move to using the <tt>TargetRegistry::lookupTarget()</tt>
951 function to find targets.</li>
952 </ul>
953</li>
Duncan Sands54fbb412009-06-24 08:38:48 +0000[diff] [blame]954</ul>
955
956</div>
957
958
959
960<!-- *********************************************************************** -->
961<div class="doc_section">
962 <a name="portability">Portability and Supported Platforms</a>
963</div>
964<!-- *********************************************************************** -->
965
966<div class="doc_text">
967
968<p>LLVM is known to work on the following platforms:</p>
969
970<ul>
971<li>Intel and AMD machines (IA32, X86-64, AMD64, EMT-64) running Red Hat
Chris Lattner8767bdc2009-10-10 22:13:38 +0000[diff] [blame^]972 Linux, Fedora Core, FreeBSD and AuroraUX (and probably other unix-like
973 systems).</li>
Duncan Sands54fbb412009-06-24 08:38:48 +0000[diff] [blame]974<li>PowerPC and X86-based Mac OS X systems, running 10.3 and above in 32-bit
Chris Lattner8767bdc2009-10-10 22:13:38 +0000[diff] [blame^]975 and 64-bit modes.</li>
Duncan Sands54fbb412009-06-24 08:38:48 +0000[diff] [blame]976<li>Intel and AMD machines running on Win32 using MinGW libraries (native).</li>
977<li>Intel and AMD machines running on Win32 with the Cygwin libraries (limited
978 support is available for native builds with Visual C++).</li>
979<li>Sun UltraSPARC workstations running Solaris 10.</li>
980<li>Alpha-based machines running Debian GNU/Linux.</li>
Duncan Sands54fbb412009-06-24 08:38:48 +0000[diff] [blame]981</ul>
982
983<p>The core LLVM infrastructure uses GNU autoconf to adapt itself
984to the machine and operating system on which it is built. However, minor
985porting may be required to get LLVM to work on new platforms. We welcome your
986portability patches and reports of successful builds or error messages.</p>
987
988</div>
989
990<!-- *********************************************************************** -->
991<div class="doc_section">
992 <a name="knownproblems">Known Problems</a>
993</div>
994<!-- *********************************************************************** -->
995
996<div class="doc_text">
997
998<p>This section contains significant known problems with the LLVM system,
999listed by component. If you run into a problem, please check the <a
1000href="http://llvm.org/bugs/">LLVM bug database</a> and submit a bug if
1001there isn't already one.</p>
1002
Chris Lattnera8f23072009-07-21 22:47:03 +0000[diff] [blame]1003<ul>
Chris Lattner4050d252009-07-21 23:17:26 +0000[diff] [blame]1004<li>LLVM will not correctly compile on Solaris and/or OpenSolaris
Chris Lattnera8f23072009-07-21 22:47:03 +0000[diff] [blame]1005using the stock GCC 3.x.x series 'out the box',
1006See: <a href="#brokengcc">Broken versions of GCC and other tools</a>.
1007However, A <a href="http://pkg.auroraux.org/GCC">Modern GCC Build</a>
1008for x86/x64 has been made available from the third party AuroraUX Project
Chris Lattner6aaf6902009-10-08 06:27:53 +0000[diff] [blame]1009that has been meticulously tested for bootstrapping LLVM &amp; Clang.</li>
Chris Lattnera8f23072009-07-21 22:47:03 +0000[diff] [blame]1010</ul>
1011
Duncan Sands54fbb412009-06-24 08:38:48 +0000[diff] [blame]1012</div>
1013
1014<!-- ======================================================================= -->
1015<div class="doc_subsection">
1016 <a name="experimental">Experimental features included with this release</a>
1017</div>
1018
1019<div class="doc_text">
1020
1021<p>The following components of this LLVM release are either untested, known to
1022be broken or unreliable, or are in early development. These components should
1023not be relied on, and bugs should not be filed against them, but they may be
1024useful to some people. In particular, if you would like to work on one of these
1025components, please contact us on the <a
1026href="http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev">LLVMdev list</a>.</p>
1027
1028<ul>
Chris Lattner8767bdc2009-10-10 22:13:38 +0000[diff] [blame^]1029<li>The MSIL, Alpha, SPU, MIPS, PIC16, Blackfin, MSP430 and SystemZ backends are
1030 experimental.</li>
Duncan Sands54fbb412009-06-24 08:38:48 +0000[diff] [blame]1031<li>The <tt>llc</tt> "<tt>-filetype=asm</tt>" (the default) is the only
Chris Lattner8767bdc2009-10-10 22:13:38 +0000[diff] [blame^]1032 supported value for this option. The ELF writer is experimental.</li>
1033<li>The implementation of Andersen's Alias Analysis has many known bugs.</li>
Duncan Sands54fbb412009-06-24 08:38:48 +0000[diff] [blame]1034</ul>
1035
1036</div>
1037
1038<!-- ======================================================================= -->
1039<div class="doc_subsection">
1040 <a name="x86-be">Known problems with the X86 back-end</a>
1041</div>
1042
1043<div class="doc_text">
1044
1045<ul>
1046 <li>The X86 backend does not yet support
1047 all <a href="http://llvm.org/PR879">inline assembly that uses the X86
1048 floating point stack</a>. It supports the 'f' and 't' constraints, but not
1049 'u'.</li>
1050 <li>The X86 backend generates inefficient floating point code when configured
1051 to generate code for systems that don't have SSE2.</li>
1052 <li>Win64 code generation wasn't widely tested. Everything should work, but we
1053 expect small issues to happen. Also, llvm-gcc cannot build the mingw64
1054 runtime currently due
1055 to <a href="http://llvm.org/PR2255">several</a>
1056 <a href="http://llvm.org/PR2257">bugs</a> and due to lack of support for
1057 the
1058 'u' inline assembly constraint and for X87 floating point inline assembly.</li>
1059 <li>The X86-64 backend does not yet support the LLVM IR instruction
1060 <tt>va_arg</tt>. Currently, the llvm-gcc and front-ends support variadic
1061 argument constructs on X86-64 by lowering them manually.</li>
1062</ul>
1063
1064</div>
1065
1066<!-- ======================================================================= -->
1067<div class="doc_subsection">
1068 <a name="ppc-be">Known problems with the PowerPC back-end</a>
1069</div>
1070
1071<div class="doc_text">
1072
1073<ul>
1074<li>The Linux PPC32/ABI support needs testing for the interpreter and static
1075compilation, and lacks support for debug information.</li>
1076</ul>
1077
1078</div>
1079
1080<!-- ======================================================================= -->
1081<div class="doc_subsection">
1082 <a name="arm-be">Known problems with the ARM back-end</a>
1083</div>
1084
1085<div class="doc_text">
1086
1087<ul>
Bob Wilson80333842009-08-12 21:19:49 +0000[diff] [blame]1088<li>Support for the Advanced SIMD (Neon) instruction set is still incomplete
1089and not well tested. Some features may not work at all, and the code quality
1090may be poor in some cases.</li>
Duncan Sands54fbb412009-06-24 08:38:48 +0000[diff] [blame]1091<li>Thumb mode works only on ARMv6 or higher processors. On sub-ARMv6
1092processors, thumb programs can crash or produce wrong
1093results (<a href="http://llvm.org/PR1388">PR1388</a>).</li>
1094<li>Compilation for ARM Linux OABI (old ABI) is supported but not fully tested.
1095</li>
Duncan Sands54fbb412009-06-24 08:38:48 +0000[diff] [blame]1096</ul>
1097
1098</div>
1099
1100<!-- ======================================================================= -->
1101<div class="doc_subsection">
1102 <a name="sparc-be">Known problems with the SPARC back-end</a>
1103</div>
1104
1105<div class="doc_text">
1106
1107<ul>
1108<li>The SPARC backend only supports the 32-bit SPARC ABI (-m32); it does not
1109 support the 64-bit SPARC ABI (-m64).</li>
1110</ul>
1111
1112</div>
1113
1114<!-- ======================================================================= -->
1115<div class="doc_subsection">
1116 <a name="mips-be">Known problems with the MIPS back-end</a>
1117</div>
1118
1119<div class="doc_text">
1120
1121<ul>
Duncan Sands54fbb412009-06-24 08:38:48 +0000[diff] [blame]1122<li>64-bit MIPS targets are not supported yet.</li>
1123</ul>
1124
1125</div>
1126
1127<!-- ======================================================================= -->
1128<div class="doc_subsection">
1129 <a name="alpha-be">Known problems with the Alpha back-end</a>
1130</div>
1131
1132<div class="doc_text">
1133
1134<ul>
1135
1136<li>On 21164s, some rare FP arithmetic sequences which may trap do not have the
1137appropriate nops inserted to ensure restartability.</li>
1138
1139</ul>
1140</div>
1141
1142<!-- ======================================================================= -->
1143<div class="doc_subsection">
Duncan Sands54fbb412009-06-24 08:38:48 +0000[diff] [blame]1144 <a name="c-be">Known problems with the C back-end</a>
1145</div>
1146
1147<div class="doc_text">
1148
1149<ul>
1150<li><a href="http://llvm.org/PR802">The C backend has only basic support for
1151 inline assembly code</a>.</li>
1152<li><a href="http://llvm.org/PR1658">The C backend violates the ABI of common
1153 C++ programs</a>, preventing intermixing between C++ compiled by the CBE and
1154 C++ code compiled with <tt>llc</tt> or native compilers.</li>
1155<li>The C backend does not support all exception handling constructs.</li>
1156<li>The C backend does not support arbitrary precision integers.</li>
1157</ul>
1158
1159</div>
1160
1161
1162<!-- ======================================================================= -->
1163<div class="doc_subsection">
1164 <a name="c-fe">Known problems with the llvm-gcc C front-end</a>
1165</div>
1166
1167<div class="doc_text">
1168
Duncan Sands54fbb412009-06-24 08:38:48 +0000[diff] [blame]1169<p>The only major language feature of GCC not supported by llvm-gcc is
1170 the <tt>__builtin_apply</tt> family of builtins. However, some extensions
1171 are only supported on some targets. For example, trampolines are only
1172 supported on some targets (these are used when you take the address of a
1173 nested function).</p>
1174
1175<p>If you run into GCC extensions which are not supported, please let us know.
1176</p>
1177
1178</div>
1179
1180<!-- ======================================================================= -->
1181<div class="doc_subsection">
1182 <a name="c++-fe">Known problems with the llvm-gcc C++ front-end</a>
1183</div>
1184
1185<div class="doc_text">
1186
1187<p>The C++ front-end is considered to be fully
1188tested and works for a number of non-trivial programs, including LLVM
1189itself, Qt, Mozilla, etc.</p>
1190
1191<ul>
1192<li>Exception handling works well on the X86 and PowerPC targets. Currently
1193 only Linux and Darwin targets are supported (both 32 and 64 bit).</li>
1194</ul>
1195
1196</div>
1197
1198<!-- ======================================================================= -->
1199<div class="doc_subsection">
1200 <a name="fortran-fe">Known problems with the llvm-gcc Fortran front-end</a>
1201</div>
1202
1203<div class="doc_text">
1204<ul>
1205<li>Fortran support generally works, but there are still several unresolved bugs
1206 in Bugzilla. Please see the tools/gfortran component for details.</li>
1207</ul>
1208</div>
1209
1210<!-- ======================================================================= -->
1211<div class="doc_subsection">
1212 <a name="ada-fe">Known problems with the llvm-gcc Ada front-end</a>
1213</div>
1214
1215<div class="doc_text">
1216The llvm-gcc 4.2 Ada compiler works fairly well; however, this is not a mature
1217technology, and problems should be expected.
1218<ul>
1219<li>The Ada front-end currently only builds on X86-32. This is mainly due
1220to lack of trampoline support (pointers to nested functions) on other platforms.
1221However, it <a href="http://llvm.org/PR2006">also fails to build on X86-64</a>
1222which does support trampolines.</li>
1223<li>The Ada front-end <a href="http://llvm.org/PR2007">fails to bootstrap</a>.
1224This is due to lack of LLVM support for <tt>setjmp</tt>/<tt>longjmp</tt> style
1225exception handling, which is used internally by the compiler.
1226Workaround: configure with --disable-bootstrap.</li>
1227<li>The c380004, <a href="http://llvm.org/PR2010">c393010</a>
1228and <a href="http://llvm.org/PR2421">cxg2021</a> ACATS tests fail
1229(c380004 also fails with gcc-4.2 mainline).
1230If the compiler is built with checks disabled then <a href="http://llvm.org/PR2010">c393010</a>
1231causes the compiler to go into an infinite loop, using up all system memory.</li>
1232<li>Some GCC specific Ada tests continue to crash the compiler.</li>
1233<li>The -E binder option (exception backtraces)
1234<a href="http://llvm.org/PR1982">does not work</a> and will result in programs
1235crashing if an exception is raised. Workaround: do not use -E.</li>
1236<li>Only discrete types <a href="http://llvm.org/PR1981">are allowed to start
1237or finish at a non-byte offset</a> in a record. Workaround: do not pack records
1238or use representation clauses that result in a field of a non-discrete type
1239starting or finishing in the middle of a byte.</li>
1240<li>The <tt>lli</tt> interpreter <a href="http://llvm.org/PR2009">considers
1241'main' as generated by the Ada binder to be invalid</a>.
1242Workaround: hand edit the file to use pointers for <tt>argv</tt> and
1243<tt>envp</tt> rather than integers.</li>
1244<li>The <tt>-fstack-check</tt> option <a href="http://llvm.org/PR2008">is
1245ignored</a>.</li>
1246</ul>
1247</div>
1248
Erick Tryzelaarb4cf9752009-09-28 04:42:55 +0000[diff] [blame]1249<!-- ======================================================================= -->
1250<div class="doc_subsection">
1251 <a name="ocaml-bindingse">Known problems with the O'Caml bindings</a>
1252</div>
1253
1254<div class="doc_text">
1255
1256<p>The Llvm.Linkage module is broken, and has incorrect values. Only
1257Llvm.Linkage.External, Llvm.Linkage.Available_externally, and
1258Llvm.Linkage.Link_once will be correct. If you need any of the other linkage
1259modes, you'll have to write an external C library in order to expose the
1260functionality. This has been fixed in the trunk.</p>
1261</div>
1262
Duncan Sands54fbb412009-06-24 08:38:48 +0000[diff] [blame]1263<!-- *********************************************************************** -->
1264<div class="doc_section">
1265 <a name="additionalinfo">Additional Information</a>
1266</div>
1267<!-- *********************************************************************** -->
1268
1269<div class="doc_text">
1270
1271<p>A wide variety of additional information is available on the <a
1272href="http://llvm.org">LLVM web page</a>, in particular in the <a
1273href="http://llvm.org/docs/">documentation</a> section. The web page also
1274contains versions of the API documentation which is up-to-date with the
1275Subversion version of the source code.
1276You can access versions of these documents specific to this release by going
1277into the "<tt>llvm/doc/</tt>" directory in the LLVM tree.</p>
1278
1279<p>If you have any questions or comments about LLVM, please feel free to contact
1280us via the <a href="http://llvm.org/docs/#maillist"> mailing
1281lists</a>.</p>
1282
1283</div>
1284
1285<!-- *********************************************************************** -->
1286
1287<hr>
1288<address>
1289 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
1290 src="http://jigsaw.w3.org/css-validator/images/vcss-blue" alt="Valid CSS"></a>
1291 <a href="http://validator.w3.org/check/referer"><img
1292 src="http://www.w3.org/Icons/valid-html401-blue" alt="Valid HTML 4.01"></a>
1293
1294 <a href="http://llvm.org/">LLVM Compiler Infrastructure</a><br>
1295 Last modified: $Date$
1296</address>
1297
1298</body>
1299</html>