Gitiles
Code Review Sign In
gerrit-public.fairphone.software / fp2-dev / platform / external / llvm / bb8107aa09b1f7b43ea073afead7f10e9ac88934 / . / docs / ReleaseNotes.html
blob: 31ef5144ba8b03e54f64f75edf1f5276f5691fd3 [file] [log] [blame]
Dan Gohmanf17a25c2007-07-18 16:29:46 +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">
Jeffrey Yasskin8ee1b0a2010-01-28 01:14:43 +0000[diff] [blame]7 <title>LLVM 2.7 Release Notes</title>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]8</head>
9<body>
10
Jeffrey Yasskin8ee1b0a2010-01-28 01:14:43 +0000[diff] [blame]11<div class="doc_title">LLVM 2.7 Release Notes</div>
Mikhail Glushenkov808cd3e2008-10-13 02:08:34 +0000[diff] [blame]12
Chris Lattnere5236102010-03-17 04:02:39 +0000[diff] [blame]13<img align=right src="http://llvm.org/img/DragonSmall.png"
Gabor Greif92d28ea2010-04-22 10:21:43 +0000[diff] [blame]14 width="136" height="136" alt="LLVM Dragon Logo">
Chris Lattnere5236102010-03-17 04:02:39 +0000[diff] [blame]15
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]16<ol>
17 <li><a href="#intro">Introduction</a></li>
Chris Lattner78a130b2008-10-13 18:01:01 +0000[diff] [blame]18 <li><a href="#subproj">Sub-project Status Update</a></li>
Jeffrey Yasskin8ee1b0a2010-01-28 01:14:43 +0000[diff] [blame]19 <li><a href="#externalproj">External Projects Using LLVM 2.7</a></li>
20 <li><a href="#whatsnew">What's New in LLVM 2.7?</a></li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]21 <li><a href="GettingStarted.html">Installation Instructions</a></li>
22 <li><a href="#portability">Portability and Supported Platforms</a></li>
Dan Gohman0de549c2008-10-14 16:23:02 +0000[diff] [blame]23 <li><a href="#knownproblems">Known Problems</a></li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]24 <li><a href="#additionalinfo">Additional Information</a></li>
25</ol>
26
27<div class="doc_author">
Dan Gohman0de549c2008-10-14 16:23:02 +0000[diff] [blame]28 <p>Written by the <a href="http://llvm.org">LLVM Team</a></p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]29</div>
30
Chris Lattner6ba1be32010-04-13 06:37:00 +0000[diff] [blame]31<!--
Chris Lattnerabfaf732010-04-22 05:41:35 +0000[diff] [blame]32<h1 style="color:red">These are in-progress notes for the upcoming LLVM 2.8
Jeffrey Yasskin8ee1b0a2010-01-28 01:14:43 +0000[diff] [blame]33release.<br>
34You may prefer the
Chris Lattnerabfaf732010-04-22 05:41:35 +0000[diff] [blame]35<a href="http://llvm.org/releases/2.6/docs/ReleaseNotes.html">LLVM 2.7
Chris Lattner6ba1be32010-04-13 06:37:00 +0000[diff] [blame]36Release Notes</a>.</h1>-->
Jeffrey Yasskin8ee1b0a2010-01-28 01:14:43 +0000[diff] [blame]37
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]38<!-- *********************************************************************** -->
39<div class="doc_section">
40 <a name="intro">Introduction</a>
41</div>
42<!-- *********************************************************************** -->
43
44<div class="doc_text">
45
Chris Lattner92eacf82008-10-13 17:57:36 +0000[diff] [blame]46<p>This document contains the release notes for the LLVM Compiler
Jeffrey Yasskin8ee1b0a2010-01-28 01:14:43 +0000[diff] [blame]47Infrastructure, release 2.7. Here we describe the status of LLVM, including
Chris Lattner92eacf82008-10-13 17:57:36 +0000[diff] [blame]48major improvements from the previous release and significant known problems.
Mikhail Glushenkovf61066a2009-03-01 18:09:47 +0000[diff] [blame]49All LLVM releases may be downloaded from the <a
Chris Lattner92eacf82008-10-13 17:57:36 +0000[diff] [blame]50href="http://llvm.org/releases/">LLVM releases web site</a>.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]51
52<p>For more information about LLVM, including information about the latest
53release, please check out the <a href="http://llvm.org/">main LLVM
54web site</a>. If you have questions or comments, the <a
Chris Lattner58b1d0f2010-03-17 04:41:49 +0000[diff] [blame]55href="http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev">LLVM Developer's
56Mailing List</a> is a good place to send them.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]57
Mikhail Glushenkov808cd3e2008-10-13 02:08:34 +0000[diff] [blame]58<p>Note that if you are reading this file from a Subversion checkout or the
Chris Lattnerb0d6c502007-09-21 03:54:09 +0000[diff] [blame]59main LLVM web page, this document applies to the <i>next</i> release, not the
Gabor Greif7c65c4f2008-10-14 11:00:32 +0000[diff] [blame]60current one. To see the release notes for a specific release, please see the
Chris Lattnerb0d6c502007-09-21 03:54:09 +0000[diff] [blame]61<a href="http://llvm.org/releases/">releases page</a>.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]62
63</div>
Chris Lattner0b6ea7a2009-10-13 17:48:04 +0000[diff] [blame]64
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]65
Chris Lattner0b6ea7a2009-10-13 17:48:04 +0000[diff] [blame]66<!--
67Almost dead code.
68 include/llvm/Analysis/LiveValues.h => Dan
69 lib/Transforms/IPO/MergeFunctions.cpp => consider for 2.8.
70 llvm/Analysis/PointerTracking.h => Edwin wants this, consider for 2.8.
Chris Lattner6ba1be32010-04-13 06:37:00 +0000[diff] [blame]71 ABCD, GEPSplitterPass
Chris Lattner2bd5e652010-01-16 21:25:13 +0000[diff] [blame]72 MSIL backend?
Chris Lattner2b656342010-03-17 06:41:58 +0000[diff] [blame]73 lib/Transforms/Utils/SSI.cpp -> ABCD depends on it.
Chris Lattner0b6ea7a2009-10-13 17:48:04 +0000[diff] [blame]74-->
75
76
Jeffrey Yasskin8ee1b0a2010-01-28 01:14:43 +0000[diff] [blame]77<!-- Features that need text if they're finished for 2.7:
Chris Lattner8f33c262010-04-22 06:28:20 +0000[diff] [blame]78 combiner-aa?
Chris Lattner0b6ea7a2009-10-13 17:48:04 +0000[diff] [blame]79 strong phi elim
Chris Lattner6fd928d2010-04-21 05:17:40 +0000[diff] [blame]80 llvm.dbg.value: variable debug info for optimized code
Chris Lattner0b6ea7a2009-10-13 17:48:04 +0000[diff] [blame]81 loop dependence analysis
Chris Lattner8fda4952008-02-10 07:46:44 +0000[diff] [blame]82 -->
Mikhail Glushenkov808cd3e2008-10-13 02:08:34 +0000[diff] [blame]83
Chris Lattner4f076402008-10-12 19:47:48 +0000[diff] [blame]84 <!-- for announcement email:
Chris Lattner0b6ea7a2009-10-13 17:48:04 +0000[diff] [blame]85 Logo web page.
86 llvm devmtg
87 compiler_rt
88 KLEE web page at klee.llvm.org
89 Many new papers added to /pubs/
Chris Lattner2b656342010-03-17 06:41:58 +0000[diff] [blame]90 Mention gcc plugin.
Chris Lattnera53f4972009-02-25 06:34:50 +0000[diff] [blame]91 -->
Chris Lattner8fda4952008-02-10 07:46:44 +0000[diff] [blame]92
Chris Lattner78a130b2008-10-13 18:01:01 +0000[diff] [blame]93<!-- *********************************************************************** -->
94<div class="doc_section">
95 <a name="subproj">Sub-project Status Update</a>
Chris Lattnerf65224f2008-06-08 21:34:41 +0000[diff] [blame]96</div>
Chris Lattner78a130b2008-10-13 18:01:01 +0000[diff] [blame]97<!-- *********************************************************************** -->
Chris Lattnerf65224f2008-06-08 21:34:41 +0000[diff] [blame]98
99<div class="doc_text">
Chris Lattnerf6a0d382008-06-08 21:58:17 +0000[diff] [blame]100<p>
Jeffrey Yasskin8ee1b0a2010-01-28 01:14:43 +0000[diff] [blame]101The LLVM 2.7 distribution currently consists of code from the core LLVM
Chris Lattner0b6ea7a2009-10-13 17:48:04 +0000[diff] [blame]102repository (which roughly includes the LLVM optimizers, code generators
103and supporting tools), the Clang repository and the llvm-gcc repository. In
104addition to this code, the LLVM Project includes other sub-projects that are in
105development. Here we include updates on these subprojects.
Bill Wendlingdde41b82009-03-02 04:28:57 +0000[diff] [blame]106</p>
Chris Lattner78a130b2008-10-13 18:01:01 +0000[diff] [blame]107
Chris Lattnerf6a0d382008-06-08 21:58:17 +0000[diff] [blame]108</div>
109
Chris Lattner78a130b2008-10-13 18:01:01 +0000[diff] [blame]110
Chris Lattnerf6a0d382008-06-08 21:58:17 +0000[diff] [blame]111<!--=========================================================================-->
Chris Lattner78a130b2008-10-13 18:01:01 +0000[diff] [blame]112<div class="doc_subsection">
Chris Lattnera3c683b2008-10-13 18:11:54 +0000[diff] [blame]113<a name="clang">Clang: C/C++/Objective-C Frontend Toolkit</a>
Chris Lattnerf6a0d382008-06-08 21:58:17 +0000[diff] [blame]114</div>
115
116<div class="doc_text">
117
Douglas Gregor3235b4f2010-04-22 20:42:40 +0000[diff] [blame]118<p><a href="http://clang.llvm.org/">Clang</a> is an LLVM front end for the C, C++, and Objective-C languages. Clang aims to provide a better user experience through expressive diagnostics, a high level of conformance to language standards, fast compilation, and low memory use. Like LLVM, Clang provides a modular, library-based architecture that makes it suitable for creating or integrating with other development tools. Clang is considered a production-quality compiler for C and Objective-C on x86 (32- and 64-bit).</p>
Chris Lattner0b6ea7a2009-10-13 17:48:04 +0000[diff] [blame]119
Jeffrey Yasskin8ee1b0a2010-01-28 01:14:43 +0000[diff] [blame]120<p>In the LLVM 2.7 time-frame, the Clang team has made many improvements:</p>
Bill Wendling385b0d32008-10-27 09:27:33 +0000[diff] [blame]121
Daniel Dunbarf3e35782008-10-14 23:25:09 +0000[diff] [blame]122<ul>
Douglas Gregor3235b4f2010-04-22 20:42:40 +0000[diff] [blame]123
124<li>C++ Support: Clang is now capable of self-hosting! While still alpha-quality, Clang's C++ support has matured enough to build LLVM and Clang, and C++ is now enabled by default. See the <a href="http://clang.llvm.org/cxx_compatibility.html">Clang C++ compatibility page</a> for common C++ migration issues.</li>
125
David Chisnall25635df2010-04-25 19:13:33 +0000[diff] [blame]126<li>Objective-C: Clang now includes experimental support for an updated Objective-C ABI on non-Darwin platforms. This includes support for non-fragile instance variables and accelerated proxies, as well as greater potential for future optimisations. The new ABI is used when compiling with the -fobjc-nonfragile-abi and -fgnu-runtime options. Code compiled with these options may be mixed with code compiled with GCC or clang using the old GNU ABI, but requires the libobjc2 runtime from the GNUstep project.</li>
127
Douglas Gregor3235b4f2010-04-22 20:42:40 +0000[diff] [blame]128<li>New warnings: Clang contains a number of new warnings, including control-flow warnings (unreachable code, missing return statements in a non-<code>void</code> function, etc.), sign-comparison warnings, and improved format-string warnings.</li>
Daniel Dunbar29f1e722010-03-25 16:09:18 +0000[diff] [blame]129
Daniel Dunbar29f1e722010-03-25 16:09:18 +0000[diff] [blame]130<li>CIndex API and Python bindings: Clang now includes a C API as part of the
Gabor Greif8abc1962010-04-25 21:27:54 +0000[diff] [blame]131CIndex library. Although we may make some changes to the API in the future, it
Daniel Dunbar29f1e722010-03-25 16:09:18 +0000[diff] [blame]132is intended to be stable and has been designed for use by external projects. See
133the Clang
134doxygen <a href="http://clang.llvm.org/doxygen/group__CINDEX.html">CIndex</a>
Wesley Peck43e93692010-04-22 14:19:00 +0000[diff] [blame]135documentation for more details. The CIndex API also includes a preliminary
Daniel Dunbar29f1e722010-03-25 16:09:18 +0000[diff] [blame]136set of Python bindings.</li>
137
138<li>ARM Support: Clang now has ABI support for both the Darwin and Linux ARM
139ABIs. Coupled with many improvements to the LLVM ARM backend, Clang is now
Gabor Greif8abc1962010-04-25 21:27:54 +0000[diff] [blame]140suitable for use as a beta quality ARM compiler.</li>
David Chisnall25635df2010-04-25 19:13:33 +0000[diff] [blame]141
Bill Wendlinga8fb81d2009-03-02 04:28:18 +0000[diff] [blame]142</ul>
Chris Lattnera3c683b2008-10-13 18:11:54 +0000[diff] [blame]143</div>
144
145<!--=========================================================================-->
146<div class="doc_subsection">
147<a name="clangsa">Clang Static Analyzer</a>
148</div>
149
150<div class="doc_text">
151
Chris Lattnerfa63e262010-04-22 06:38:11 +0000[diff] [blame]152<p>The <a href="http://clang-analyzer.llvm.org/">Clang Static Analyzer</a>
153 project is an effort to use static source code analysis techniques to
154 automatically find bugs in C and Objective-C programs (and hopefully <a
155 href="http://clang-analyzer.llvm.org/dev_cxx.html">C++ in the
156 future</a>!). The tool is very good at finding bugs that occur on specific
157 paths through code, such as on error conditions.</p>
Chris Lattnerc2d84672008-10-14 00:52:49 +0000[diff] [blame]158
Chris Lattnerfa63e262010-04-22 06:38:11 +0000[diff] [blame]159<p>In the LLVM 2.7 time-frame, the analyzer core has made several major and
160 minor improvements, including better support for tracking the fields of
161 structures, initial support (not enabled by default yet) for doing
162 interprocedural (cross-function) analysis, and new checks have been added.
163</p>
Chris Lattnerf6a0d382008-06-08 21:58:17 +0000[diff] [blame]164
165</div>
166
Chris Lattner78a130b2008-10-13 18:01:01 +0000[diff] [blame]167<!--=========================================================================-->
168<div class="doc_subsection">
Nicolas Geoffray2b139382008-10-14 19:23:04 +0000[diff] [blame]169<a name="vmkit">VMKit: JVM/CLI Virtual Machine Implementation</a>
Chris Lattner78a130b2008-10-13 18:01:01 +0000[diff] [blame]170</div>
171
172<div class="doc_text">
173<p>
Nicolas Geoffray2b139382008-10-14 19:23:04 +0000[diff] [blame]174The <a href="http://vmkit.llvm.org/">VMKit project</a> is an implementation of
Chris Lattner0b6ea7a2009-10-13 17:48:04 +0000[diff] [blame]175a JVM and a CLI Virtual Machine (Microsoft .NET is an
176implementation of the CLI) using LLVM for static and just-in-time
177compilation.</p>
Chris Lattner78a130b2008-10-13 18:01:01 +0000[diff] [blame]178
Chris Lattner0b6ea7a2009-10-13 17:48:04 +0000[diff] [blame]179<p>
Chris Lattnerd5b62522010-03-29 18:34:13 +0000[diff] [blame]180With the release of LLVM 2.7, VMKit has shifted to a great framework for writing
181virtual machines. VMKit now offers precise and efficient garbage collection with
182multi-threading support, thanks to the MMTk memory management toolkit, as well
183as just in time and ahead of time compilation with LLVM. The major changes in
184VMKit 0.27 are:</p>
Chris Lattner78a130b2008-10-13 18:01:01 +0000[diff] [blame]185
Nicolas Geoffray2b139382008-10-14 19:23:04 +0000[diff] [blame]186<ul>
187
Chris Lattnerd5b62522010-03-29 18:34:13 +0000[diff] [blame]188<li>Garbage collection: VMKit now uses the MMTk toolkit for garbage collectors.
189 The first collector to be ported is the MarkSweep collector, which is precise,
190 and drastically improves the performance of VMKit.</li>
191<li>Line number information in the JVM: by using the debug metadata of LLVM, the
192 JVM now supports precise line number information, useful when printing a stack
193 trace.</li>
194<li>Interface calls in the JVM: we implemented a variant of the Interface Method
195 Table technique for interface calls in the JVM.
196</li>
Nicolas Geoffray2b139382008-10-14 19:23:04 +0000[diff] [blame]197
198</ul>
Chris Lattner78a130b2008-10-13 18:01:01 +0000[diff] [blame]199</div>
200
Chris Lattner0b6ea7a2009-10-13 17:48:04 +0000[diff] [blame]201
202<!--=========================================================================-->
203<div class="doc_subsection">
204<a name="compiler-rt">compiler-rt: Compiler Runtime Library</a>
205</div>
206
207<div class="doc_text">
208<p>
209The new LLVM <a href="http://compiler-rt.llvm.org/">compiler-rt project</a>
210is a simple library that provides an implementation of the low-level
211target-specific hooks required by code generation and other runtime components.
212For example, when compiling for a 32-bit target, converting a double to a 64-bit
213unsigned integer is compiled into a runtime call to the "__fixunsdfdi"
214function. The compiler-rt library provides highly optimized implementations of
215this and other low-level routines (some are 3x faster than the equivalent
216libgcc routines).</p>
217
218<p>
219All of the code in the compiler-rt project is available under the standard LLVM
Chris Lattner6fd928d2010-04-21 05:17:40 +0000[diff] [blame]220License, a "BSD-style" license. New in LLVM 2.7: compiler_rt now
221supports ARM targets.</p>
Chris Lattner0b6ea7a2009-10-13 17:48:04 +0000[diff] [blame]222
223</div>
224
225<!--=========================================================================-->
226<div class="doc_subsection">
Duncan Sands2d9c3082010-04-02 09:23:15 +0000[diff] [blame]227<a name="dragonegg">DragonEgg: llvm-gcc ported to gcc-4.5</a>
Chris Lattner0b6ea7a2009-10-13 17:48:04 +0000[diff] [blame]228</div>
229
230<div class="doc_text">
231<p>
Duncan Sands2d9c3082010-04-02 09:23:15 +0000[diff] [blame]232<a href="http://dragonegg.llvm.org/">DragonEgg</a> is a port of llvm-gcc to
233gcc-4.5. Unlike llvm-gcc, which makes many intrusive changes to the underlying
234gcc-4.2 code, dragonegg in theory does not require any gcc-4.5 modifications
235whatsoever (currently one small patch is needed). This is thanks to the new
236<a href="http://gcc.gnu.org/wiki/plugins">gcc plugin architecture</a>, which
237makes it possible to modify the behaviour of gcc at runtime by loading a plugin,
238which is nothing more than a dynamic library which conforms to the gcc plugin
239interface. DragonEgg is a gcc plugin that causes the LLVM optimizers to be run
240instead of the gcc optimizers, and the LLVM code generators instead of the gcc
241code generators, just like llvm-gcc. To use it, you add
242"-fplugin=path/dragonegg.so" to the gcc-4.5 command line, and gcc-4.5 magically
243becomes llvm-gcc-4.5!
244</p>
245
246<p>
247DragonEgg is still a work in progress. Currently C works very well, while C++,
248Ada and Fortran work fairly well. All other languages either don't work at all,
249or only work poorly. For the moment only the x86-32 and x86-64 targets are
Duncan Sands566f40b2010-04-20 19:40:58 +0000[diff] [blame]250supported, and only on linux and darwin (darwin needs an additional gcc patch).
Duncan Sands2d9c3082010-04-02 09:23:15 +0000[diff] [blame]251</p>
252
253<p>
Duncan Sandsa9cc3892010-04-21 13:51:48 +0000[diff] [blame]254DragonEgg is a new project which is seeing its first release with llvm-2.7.
Chris Lattner0b6ea7a2009-10-13 17:48:04 +0000[diff] [blame]255</p>
256
257</div>
258
259
260<!--=========================================================================-->
261<div class="doc_subsection">
262<a name="mc">llvm-mc: Machine Code Toolkit</a>
263</div>
264
265<div class="doc_text">
266<p>
Chris Lattnerfa63e262010-04-22 06:38:11 +0000[diff] [blame]267The LLVM Machine Code (aka MC) sub-project of LLVM was created to solve a number
268of problems in the realm of assembly, disassembly, object file format handling,
269and a number of other related areas that CPU instruction-set level tools work
270in. It is a sub-project of LLVM which provides it with a number of advantages
271over other compilers that do not have tightly integrated assembly-level tools.
272For a gentle introduction, please see the <a
273href="http://blog.llvm.org/2010/04/intro-to-llvm-mc-project.html">Intro to the
274LLVM MC Project Blog Post</a>.
Chris Lattner0b6ea7a2009-10-13 17:48:04 +0000[diff] [blame]275</p>
276
Chris Lattnerfa63e262010-04-22 06:38:11 +0000[diff] [blame]277<p>2.7 includes major parts of the work required by the new MC Project. A few
278 targets have been refactored to support it, and work is underway to support a
Gabor Greif9e25ea72010-04-25 21:30:22 +0000[diff] [blame]279 native assembler in LLVM. This work is not complete in LLVM 2.7, but it has
Chris Lattnerfa63e262010-04-22 06:38:11 +0000[diff] [blame]280 made substantially more progress on LLVM mainline.</p>
281
282<p>One minor example of what MC can do is to transcode an AT&amp;T syntax
Gabor Greifadba54d2010-04-22 10:25:23 +0000[diff] [blame]283 X86 .s file into intel syntax. You can do this with something like:</p>
Chris Lattnerfa63e262010-04-22 06:38:11 +0000[diff] [blame]284<pre>
285 llvm-mc foo.s -output-asm-variant=1 -o foo-intel.s
286</pre>
Chris Lattner8a57e892010-04-21 06:23:40 +0000[diff] [blame]287
Chris Lattner0b6ea7a2009-10-13 17:48:04 +0000[diff] [blame]288</div>
289
290
Chris Lattnerb2cc5022009-02-26 22:33:38 +0000[diff] [blame]291<!-- *********************************************************************** -->
292<div class="doc_section">
Jeffrey Yasskin8ee1b0a2010-01-28 01:14:43 +0000[diff] [blame]293 <a name="externalproj">External Open Source Projects Using LLVM 2.7</a>
Chris Lattnerb2cc5022009-02-26 22:33:38 +0000[diff] [blame]294</div>
295<!-- *********************************************************************** -->
296
Chris Lattner0b6ea7a2009-10-13 17:48:04 +0000[diff] [blame]297<div class="doc_text">
298
299<p>An exciting aspect of LLVM is that it is used as an enabling technology for
300 a lot of other language and tools projects. This section lists some of the
Jeffrey Yasskin8ee1b0a2010-01-28 01:14:43 +0000[diff] [blame]301 projects that have already been updated to work with LLVM 2.7.</p>
Chris Lattner0b6ea7a2009-10-13 17:48:04 +0000[diff] [blame]302</div>
303
Chris Lattnerb2cc5022009-02-26 22:33:38 +0000[diff] [blame]304<!--=========================================================================-->
305<div class="doc_subsection">
306<a name="pure">Pure</a>
307</div>
308
309<div class="doc_text">
310<p>
Chris Lattner47202c32009-03-02 02:37:32 +0000[diff] [blame]311<a href="http://pure-lang.googlecode.com/">Pure</a>
312is an algebraic/functional programming language based on term rewriting.
Chris Lattnerb2cc5022009-02-26 22:33:38 +0000[diff] [blame]313Programs are collections of equations which are used to evaluate expressions in
314a symbolic fashion. Pure offers dynamic typing, eager and lazy evaluation,
Mikhail Glushenkovf61066a2009-03-01 18:09:47 +0000[diff] [blame]315lexical closures, a hygienic macro system (also based on term rewriting),
Chris Lattnerb2cc5022009-02-26 22:33:38 +0000[diff] [blame]316built-in list and matrix support (including list and matrix comprehensions) and
317an easy-to-use C interface. The interpreter uses LLVM as a backend to
318 JIT-compile Pure programs to fast native code.</p>
319
Chris Lattner1f8dc512010-03-17 17:25:49 +0000[diff] [blame]320<p>Pure versions 0.43 and later have been tested and are known to work with
321LLVM 2.7 (and continue to work with older LLVM releases >= 2.5).</p>
322
Chris Lattnerb2cc5022009-02-26 22:33:38 +0000[diff] [blame]323</div>
324
Chris Lattner1ad599a2009-02-28 18:58:01 +0000[diff] [blame]325<!--=========================================================================-->
326<div class="doc_subsection">
327<a name="RoadsendPHP">Roadsend PHP</a>
328</div>
329
330<div class="doc_text">
Chris Lattner0b6ea7a2009-10-13 17:48:04 +0000[diff] [blame]331<p>
332<a href="http://code.roadsend.com/rphp">Roadsend PHP</a> (rphp) is an open
Chris Lattner9e08de12009-03-02 19:07:24 +0000[diff] [blame]333source implementation of the PHP programming
Chris Lattner0b6ea7a2009-10-13 17:48:04 +0000[diff] [blame]334language that uses LLVM for its optimizer, JIT and static compiler. This is a
Chris Lattnere5236102010-03-17 04:02:39 +0000[diff] [blame]335reimplementation of an earlier project that is now based on LLVM.
Chris Lattner1f8dc512010-03-17 17:25:49 +0000[diff] [blame]336</p>
Chris Lattner1ad599a2009-02-28 18:58:01 +0000[diff] [blame]337</div>
Chris Lattnerb2cc5022009-02-26 22:33:38 +0000[diff] [blame]338
Chris Lattner0b6ea7a2009-10-13 17:48:04 +0000[diff] [blame]339<!--=========================================================================-->
340<div class="doc_subsection">
341<a name="UnladenSwallow">Unladen Swallow</a>
342</div>
343
344<div class="doc_text">
345<p>
346<a href="http://code.google.com/p/unladen-swallow/">Unladen Swallow</a> is a
347branch of <a href="http://python.org/">Python</a> intended to be fully
348compatible and significantly faster. It uses LLVM's optimization passes and JIT
Chris Lattnerd7db82c2010-03-18 06:52:15 +0000[diff] [blame]349compiler.
Chris Lattnere5236102010-03-17 04:02:39 +0000[diff] [blame]350</p>
Chris Lattner0b6ea7a2009-10-13 17:48:04 +0000[diff] [blame]351</div>
352
353<!--=========================================================================-->
354<div class="doc_subsection">
Chris Lattner49894862010-03-29 17:50:39 +0000[diff] [blame]355<a name="tce">TTA-based Codesign Environment (TCE)</a>
356</div>
357
358<div class="doc_text">
359<p>
360<a href="http://tce.cs.tut.fi/">TCE</a> is a toolset for designing
361application-specific processors (ASP) based on the Transport triggered
362architecture (TTA). The toolset provides a complete co-design flow from C/C++
363programs down to synthesizable VHDL and parallel program binaries. Processor
364customization points include the register files, function units, supported
365operations, and the interconnection network.</p>
366
367<p>TCE uses llvm-gcc/Clang and LLVM for C/C++ language support, target
368independent optimizations and also for parts of code generation. It generates
369new LLVM-based code generators "on the fly" for the designed TTA processors and
370loads them in to the compiler backend as runtime libraries to avoid per-target
371recompilation of larger parts of the compiler chain.</p>
372
373</div>
Chris Lattnerf6a0d382008-06-08 21:58:17 +0000[diff] [blame]374
John Criswelle75f04f2010-04-06 14:52:14 +0000[diff] [blame]375<!--=========================================================================-->
376<div class="doc_subsection">
377<a name="safecode">SAFECode Compiler</a>
378</div>
379
380<div class="doc_text">
381<p>
382<a href="http://safecode.cs.illinois.edu">SAFECode</a> is a memory safe C
383compiler built using LLVM. It takes standard, unannotated C code, analyzes the
384code to ensure that memory accesses and array indexing operations are safe, and
385instruments the code with run-time checks when safety cannot be proven
386statically.
387</p>
388</div>
389
Chris Lattner73372af2010-04-22 17:28:36 +0000[diff] [blame]390<!--=========================================================================-->
391<div class="doc_subsection">
392<a name="icedtea">IcedTea Java Virtual Machine Implementation</a>
393</div>
John Criswelle75f04f2010-04-06 14:52:14 +0000[diff] [blame]394
Chris Lattner73372af2010-04-22 17:28:36 +0000[diff] [blame]395<div class="doc_text">
396<p>
397<a href="http://icedtea.classpath.org/wiki/Main_Page">IcedTea</a> provides a
398harness to build OpenJDK using only free software build tools and to provide
399replacements for the not-yet free parts of OpenJDK. One of the extensions that
400IcedTea provides is a new JIT compiler named <a
401href="http://icedtea.classpath.org/wiki/ZeroSharkFaq">Shark</a> which uses LLVM
402to provide native code generation without introducing processor-dependent
403code.
404</p>
405<p>Icedtea6 1.8 and later have been tested and are known to work with
406LLVM 2.7 (and continue to work with older LLVM releases >= 2.6 as well).
407</p>
408</div>
409
410<!--=========================================================================-->
411<div class="doc_subsection">
412<a name="llvm-lua">LLVM-Lua</a>
413</div>
414
415<div class="doc_text">
416<p>
417<a href="http://code.google.com/p/llvm-lua/">LLVM-Lua</a> uses LLVM
418 to add JIT and static compiling support to the Lua VM. Lua
419bytecode is analyzed to remove type checks, then LLVM is used to compile the
420bytecode down to machine code.
421</p>
422<p>LLVM-Lua 1.2.0 have been tested and is known to work with LLVM 2.7.
423</p>
424</div>
Chris Lattner7a12c992010-04-22 21:34:16 +0000[diff] [blame]425
426<!--=========================================================================-->
427<div class="doc_subsection">
428<a name="MacRuby">MacRuby</a>
429</div>
430
431<div class="doc_text">
432<p>
433<a href="http://macruby.org">MacRuby</a> is an implementation of Ruby based on
434core Mac OS technologies, sponsored by Apple Inc. It uses LLVM at runtime for
435optimization passes, JIT compilation and exception handling. It also allows
436static (ahead-of-time) compilation of Ruby code straight to machine code.
437</p>
438<p>The upcoming MacRuby 0.6 release works with LLVM 2.7.
439</p>
440</div>
441
442
Chris Lattnerf6a0d382008-06-08 21:58:17 +0000[diff] [blame]443<!-- *********************************************************************** -->
444<div class="doc_section">
Jeffrey Yasskin8ee1b0a2010-01-28 01:14:43 +0000[diff] [blame]445 <a name="whatsnew">What's New in LLVM 2.7?</a>
Chris Lattnerf6a0d382008-06-08 21:58:17 +0000[diff] [blame]446</div>
447<!-- *********************************************************************** -->
448
Chris Lattner5cce7a52008-06-08 22:59:35 +0000[diff] [blame]449<div class="doc_text">
450
Chris Lattner0b6ea7a2009-10-13 17:48:04 +0000[diff] [blame]451<p>This release includes a huge number of bug fixes, performance tweaks and
Chris Lattner92eacf82008-10-13 17:57:36 +0000[diff] [blame]452minor improvements. Some of the major improvements and new features are listed
453in this section.
Chris Lattner5cce7a52008-06-08 22:59:35 +0000[diff] [blame]454</p>
Chris Lattner0b6ea7a2009-10-13 17:48:04 +0000[diff] [blame]455
Chris Lattner8f33c262010-04-22 06:28:20 +0000[diff] [blame]456</div>
457
458<!--=========================================================================-->
459<div class="doc_subsection">
460<a name="orgchanges">LLVM Community Changes</a>
461</div>
462
463<div class="doc_text">
464
Chris Lattner6fd928d2010-04-21 05:17:40 +0000[diff] [blame]465<p>In addition to changes to the code, between LLVM 2.6 and 2.7, a number of
466organization changes have happened:
467</p>
468
469<ul>
Chris Lattnerabfaf732010-04-22 05:41:35 +0000[diff] [blame]470<li>LLVM has a new <a href="http://llvm.org/Logo.html">official logo</a>!</li>
Chris Lattner6fd928d2010-04-21 05:17:40 +0000[diff] [blame]471
Chris Lattner6fd928d2010-04-21 05:17:40 +0000[diff] [blame]472<li>Ted Kremenek and Doug Gregor have stepped forward as <a
473 href="http://llvm.org/docs/DeveloperPolicy.html#owners">Code Owners</a> of the
Chris Lattner8f33c262010-04-22 06:28:20 +0000[diff] [blame]474 Clang static analyzer and the Clang frontend, respectively.</li>
Chris Lattner8a57e892010-04-21 06:23:40 +0000[diff] [blame]475
476<li>LLVM now has an <a href="http://blog.llvm.org">official Blog</a> at
477 <a href="http://blog.llvm.org">http://blog.llvm.org</a>. This is a great way
478 to learn about new LLVM-related features as they are implemented. Several
479 features in this release are already explained on the blog.</li>
480
Chris Lattner6fd928d2010-04-21 05:17:40 +0000[diff] [blame]481<li>The LLVM web pages are now checked into the SVN server, in the "www",
482 "www-pubs" and "www-releases" SVN modules. Previously they were hidden in a
Chris Lattnerabfaf732010-04-22 05:41:35 +0000[diff] [blame]483 largely inaccessible old CVS server.</li>
Chris Lattner8a57e892010-04-21 06:23:40 +0000[diff] [blame]484
485<li><a href="http://llvm.org">llvm.org</a> is now hosted on a new (and much
486 faster) server. It is still graciously hosted at the University of Illinois
487 of Urbana Champaign.</li>
Chris Lattner6fd928d2010-04-21 05:17:40 +0000[diff] [blame]488</ul>
Chris Lattner5cce7a52008-06-08 22:59:35 +0000[diff] [blame]489</div>
490
Chris Lattnerf6a0d382008-06-08 21:58:17 +0000[diff] [blame]491<!--=========================================================================-->
Chris Lattnerf65224f2008-06-08 21:34:41 +0000[diff] [blame]492<div class="doc_subsection">
Chris Lattnere0798082008-02-06 18:00:06 +0000[diff] [blame]493<a name="majorfeatures">Major New Features</a>
494</div>
495
496<div class="doc_text">
497
Jeffrey Yasskin8ee1b0a2010-01-28 01:14:43 +0000[diff] [blame]498<p>LLVM 2.7 includes several major new capabilities:</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]499
500<ul>
Chris Lattnerabfaf732010-04-22 05:41:35 +0000[diff] [blame]501<li>2.7 includes initial support for the <a
Chris Lattner8a57e892010-04-21 06:23:40 +0000[diff] [blame]502 href="http://en.wikipedia.org/wiki/MicroBlaze">MicroBlaze</a> target.
503 MicroBlaze is a soft processor core designed for Xilinx FPGAs.</li>
Chris Lattner6fd928d2010-04-21 05:17:40 +0000[diff] [blame]504
Chris Lattnerabfaf732010-04-22 05:41:35 +0000[diff] [blame]505<li>2.7 includes a new LLVM IR "extensible metadata" feature. This feature
Chris Lattner8a57e892010-04-21 06:23:40 +0000[diff] [blame]506 supports many different use cases, including allowing front-end authors to
507 encode source level information into LLVM IR, which is consumed by later
508 language-specific passes. This is a great way to do high-level optimizations
509 like devirtualization, type-based alias analysis, etc. See the <a
510 href="http://blog.llvm.org/2010/04/extensible-metadata-in-llvm-ir.html">
511 Extensible Metadata Blog Post</a> for more information.</li>
512
Chris Lattnerabfaf732010-04-22 05:41:35 +0000[diff] [blame]513<li>2.7 encodes <a href="SourceLevelDebugging.html">debug information</a>
Chris Lattner8a57e892010-04-21 06:23:40 +0000[diff] [blame]514in a completely new way, built on extensible metadata. The new implementation
515is much more memory efficient and paves the way for improvements to optimized
516code debugging experience.</li>
Chris Lattner6fd928d2010-04-21 05:17:40 +0000[diff] [blame]517
Chris Lattnerabfaf732010-04-22 05:41:35 +0000[diff] [blame]518<li>2.7 now directly supports taking the address of a label and doing an
Chris Lattner8a57e892010-04-21 06:23:40 +0000[diff] [blame]519 indirect branch through a pointer. This is particularly useful for
520 interpreter loops, and is used to implement the GCC "address of label"
521 extension. For more information, see the <a
522href="http://blog.llvm.org/2010/01/address-of-label-and-indirect-branches.html">
523Address of Label and Indirect Branches in LLVM IR Blog Post</a>.
Chris Lattner6fd928d2010-04-21 05:17:40 +0000[diff] [blame]524
Chris Lattnerabfaf732010-04-22 05:41:35 +0000[diff] [blame]525<li>2.7 is the first release to start supporting APIs for assembling and
Chris Lattner8a57e892010-04-21 06:23:40 +0000[diff] [blame]526 disassembling target machine code. These APIs are useful for a variety of
527 low level clients, and are surfaced in the new "enhanced disassembly" API.
528 For more information see the <a
529 href="http://blog.llvm.org/2010/01/x86-disassembler.html">The X86
530 Disassembler Blog Post</a> for more information.</li>
Chris Lattner6fd928d2010-04-21 05:17:40 +0000[diff] [blame]531
Chris Lattnerfa63e262010-04-22 06:38:11 +0000[diff] [blame]532<li>2.7 includes major parts of the work required by the new MC Project,
533 see the <a href="#mc">MC update above</a> for more information.</li>
Chris Lattner6fd928d2010-04-21 05:17:40 +0000[diff] [blame]534
Chris Lattner5ba16862008-02-10 08:18:42 +0000[diff] [blame]535</ul>
Chris Lattner2b656342010-03-17 06:41:58 +0000[diff] [blame]536
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]537</div>
538
Chris Lattnerb0d6c502007-09-21 03:54:09 +0000[diff] [blame]539<!--=========================================================================-->
540<div class="doc_subsection">
Chris Lattner47202c32009-03-02 02:37:32 +0000[diff] [blame]541<a name="coreimprovements">LLVM IR and Core Improvements</a>
Chris Lattnerc3649452008-02-10 08:17:19 +0000[diff] [blame]542</div>
543
544<div class="doc_text">
Chris Lattner0b6ea7a2009-10-13 17:48:04 +0000[diff] [blame]545<p>LLVM IR has several new features for better support of new targets and that
546expose new optimization opportunities:</p>
Chris Lattnerc3649452008-02-10 08:17:19 +0000[diff] [blame]547
Chris Lattnerc5a4e312008-06-05 06:25:56 +0000[diff] [blame]548<ul>
Chris Lattnerabfaf732010-04-22 05:41:35 +0000[diff] [blame]549<li>LLVM IR now supports a 16-bit "half float" data type through <a
550 href="LangRef.html#int_fp16">two new intrinsics</a> and APFloat support.</li>
551<li>LLVM IR supports two new <a href="LangRef.html#fnattrs">function
Chris Lattner8a57e892010-04-21 06:23:40 +0000[diff] [blame]552 attributes</a>: inlinehint and alignstack(n). The former is a hint to the
Chris Lattnerabfaf732010-04-22 05:41:35 +0000[diff] [blame]553 optimizer that a function was declared 'inline' and thus the inliner should
554 weight it higher when considering inlining it. The later
Chris Lattner8a57e892010-04-21 06:23:40 +0000[diff] [blame]555 indicates to the code generator that the function diverges from the platform
556 ABI on stack alignment.</li>
557<li>The new <a href="LangRef.html#int_objectsize">llvm.objectsize</a> intrinsic
558 allows the optimizer to infer the sizes of memory objects in some cases.
Chris Lattnerabfaf732010-04-22 05:41:35 +0000[diff] [blame]559 This intrinsic is used to implement the GCC <tt>__builtin_object_size</tt>
Chris Lattner8a57e892010-04-21 06:23:40 +0000[diff] [blame]560 extension.</li>
Chris Lattner8f33c262010-04-22 06:28:20 +0000[diff] [blame]561<li>LLVM IR now supports marking load and store instructions with <a
562 href="LangRef.html#i_load">"non-temporal" hints</a> (building on the new
563 metadata feature). This hint encourages the code
564 generator to generate non-temporal accesses when possible, which are useful
565 for code that is carefully managing cache behavior. Currently, only the
566 X86 backend provides target support for this feature.</li>
567
Chris Lattner8a57e892010-04-21 06:23:40 +0000[diff] [blame]568<li>LLVM 2.7 has pre-alpha support for <a
Gabor Greifdd39cd12010-04-22 10:11:24 +0000[diff] [blame]569 href="LangRef.html#t_union">unions in LLVM IR</a>.
Chris Lattner8f33c262010-04-22 06:28:20 +0000[diff] [blame]570 Unfortunately, this support is not really usable in 2.7, so if you're
Chris Lattner8a57e892010-04-21 06:23:40 +0000[diff] [blame]571 interested in pushing it forward, please help contribute to LLVM mainline.</li>
Gabor Greif92d28ea2010-04-22 10:21:43 +0000[diff] [blame]572
Chris Lattnerc3649452008-02-10 08:17:19 +0000[diff] [blame]573</ul>
Mikhail Glushenkov808cd3e2008-10-13 02:08:34 +0000[diff] [blame]574
Chris Lattnerc3649452008-02-10 08:17:19 +0000[diff] [blame]575</div>
576
577<!--=========================================================================-->
578<div class="doc_subsection">
Chris Lattner8fda4952008-02-10 07:46:44 +0000[diff] [blame]579<a name="optimizer">Optimizer Improvements</a>
580</div>
581
582<div class="doc_text">
583
Chris Lattner0b6ea7a2009-10-13 17:48:04 +0000[diff] [blame]584<p>In addition to a large array of minor performance tweaks and bug fixes, this
Chris Lattnerea9c2242008-10-13 21:50:36 +0000[diff] [blame]585release includes a few major enhancements and additions to the optimizers:</p>
Chris Lattner8fda4952008-02-10 07:46:44 +0000[diff] [blame]586
587<ul>
588
Chris Lattnerabfaf732010-04-22 05:41:35 +0000[diff] [blame]589<li>The inliner reuses now merges arrays stack objects in different callees when
590 inlining multiple call sites into one function. This reduces the stack size
591 of the resultant function.</li>
592<li>The -basicaa alias analysis pass (which is the default) has been improved to
593 be less dependent on "type safe" pointers. It can now look through bitcasts
594 and other constructs more aggressively, allowing better load/store
595 optimization.</li>
596<li>The load elimination optimization in the GVN Pass [<a
597href="http://blog.llvm.org/2009/12/introduction-to-load-elimination-in-gvn.html">intro
598 blog post</a>] has been substantially improved to be more aggressive about
599 partial redundancy elimination and do more aggressive phi translation. Please
600 see the <a
601 href="http://blog.llvm.org/2009/12/advanced-topics-in-redundant-load.html">
602 Advanced Topics in Redundant Load Elimination with a Focus on PHI Translation
603 Blog Post</a> for more details.</li>
604<li>The module <a href="LangRef.html#datalayout">target data string</a> now
Wesley Peck63c86992010-04-22 13:28:34 +0000[diff] [blame]605 includes a notion of 'native' integer data types for the target. This
606 helps mid-level optimizations avoid promoting complex sequences of
607 operations to data types that are not natively supported (e.g. converting
608 i32 operations to i64 on 32-bit chips).</li>
Chris Lattnerabfaf732010-04-22 05:41:35 +0000[diff] [blame]609<li>The mid-level optimizer is now conservative when operating on a module with
610 no target data. Previously, it would default to SparcV9 settings, which is
611 not what most people expected.</li>
Chris Lattner556aac32010-04-21 06:42:24 +0000[diff] [blame]612<li>Jump threading is now much more aggressive at simplifying correlated
Chris Lattnerabfaf732010-04-22 05:41:35 +0000[diff] [blame]613 conditionals and threading blocks with otherwise complex logic. It has
614 subsumed the old "Conditional Propagation" pass, and -condprop has been
615 removed from LLVM 2.7.</li>
616<li>The -instcombine pass has been refactored from being one huge file to being
617 a library of its own. Internally, it uses a customized IRBuilder to clean
618 it up and simplify it.</li>
Chris Lattner6fd928d2010-04-21 05:17:40 +0000[diff] [blame]619
Chris Lattnerabfaf732010-04-22 05:41:35 +0000[diff] [blame]620<li>The optimal edge profiling pass is reliable and much more complete than in
621 2.6. It can be used with the llvm-prof tool but isn't wired up to the
622 llvm-gcc and clang command line options yet.</li>
Chris Lattner47202c32009-03-02 02:37:32 +0000[diff] [blame]623
Chris Lattnerabfaf732010-04-22 05:41:35 +0000[diff] [blame]624<li>A new experimental alias analysis implementation, -scev-aa, has been added.
625 It uses LLVM's Scalar Evolution implementation to do symbolic analysis of
626 pointer offset expressions to disambiguate pointers. It can catch a few
627 cases that basicaa cannot, particularly in complex loop nests.</li>
Chris Lattner556aac32010-04-21 06:42:24 +0000[diff] [blame]628
Chris Lattner8f33c262010-04-22 06:28:20 +0000[diff] [blame]629<li>The default pass ordering has been tweaked for improved optimization
Chris Lattnerabfaf732010-04-22 05:41:35 +0000[diff] [blame]630 effectiveness.</li>
Chris Lattner556aac32010-04-21 06:42:24 +0000[diff] [blame]631
Chris Lattner0b6ea7a2009-10-13 17:48:04 +0000[diff] [blame]632</ul>
Chris Lattner47202c32009-03-02 02:37:32 +0000[diff] [blame]633
Chris Lattner0b6ea7a2009-10-13 17:48:04 +0000[diff] [blame]634</div>
Chris Lattner47202c32009-03-02 02:37:32 +0000[diff] [blame]635
Chris Lattner47202c32009-03-02 02:37:32 +0000[diff] [blame]636
Chris Lattner0b6ea7a2009-10-13 17:48:04 +0000[diff] [blame]637<!--=========================================================================-->
638<div class="doc_subsection">
639<a name="executionengine">Interpreter and JIT Improvements</a>
640</div>
Chris Lattnerea9c2242008-10-13 21:50:36 +0000[diff] [blame]641
Chris Lattner0b6ea7a2009-10-13 17:48:04 +0000[diff] [blame]642<div class="doc_text">
643
644<ul>
Wesley Peck0d7f61e2010-04-22 13:36:27 +0000[diff] [blame]645<li>The JIT now supports generating debug information and is compatible with
646the new GDB 7.0 (and later) interfaces for registering dynamically generated
647debug info.</li>
Chris Lattner6fd928d2010-04-21 05:17:40 +0000[diff] [blame]648
Chris Lattner8f33c262010-04-22 06:28:20 +0000[diff] [blame]649<li>The JIT now <a href="http://llvm.org/PR5184">defaults
Jeffrey Yasskin40715142010-01-29 19:10:38 +0000[diff] [blame]650to compiling eagerly</a> to avoid a race condition in the lazy JIT.
651Clients that still want the lazy JIT can switch it on by calling
652<tt>ExecutionEngine::DisableLazyCompilation(false)</tt>.</li>
Chris Lattner6fd928d2010-04-21 05:17:40 +0000[diff] [blame]653
Jeffrey Yasskinc0577082010-02-11 01:07:39 +0000[diff] [blame]654<li>It is now possible to create more than one JIT instance in the same process.
655These JITs can generate machine code in parallel,
656although <a href="http://llvm.org/docs/ProgrammersManual.html#jitthreading">you
657still have to obey the other threading restrictions</a>.</li>
Chris Lattner6fd928d2010-04-21 05:17:40 +0000[diff] [blame]658
Chris Lattner0eef6e42008-06-08 02:45:07 +0000[diff] [blame]659</ul>
660
661</div>
662
663<!--=========================================================================-->
664<div class="doc_subsection">
Chris Lattner5f393542009-03-02 03:24:11 +0000[diff] [blame]665<a name="codegen">Target Independent Code Generator Improvements</a>
Chris Lattner0eef6e42008-06-08 02:45:07 +0000[diff] [blame]666</div>
667
668<div class="doc_text">
669
Mikhail Glushenkovf61066a2009-03-01 18:09:47 +0000[diff] [blame]670<p>We have put a significant amount of work into the code generator
671infrastructure, which allows us to implement more aggressive algorithms and make
672it run faster:</p>
Chris Lattner0eef6e42008-06-08 02:45:07 +0000[diff] [blame]673
674<ul>
Chris Lattner8f33c262010-04-22 06:28:20 +0000[diff] [blame]675<li>The 'llc -asm-verbose' option (which is now the default) has been enhanced
676 to emit many useful comments to .s files indicating information about spill
677 slots and loop nest structure. This should make it much easier to read and
678 understand assembly files. This is wired up in llvm-gcc and clang to
679 the <tt>-fverbose-asm</tt> option.</li>
680
Dan Gohman93087ff2010-04-22 20:50:43 +0000[diff] [blame]681<li>New LSR with "full strength reduction" mode, which can reduce address
682 register pressure in loops where address generation is important.</li>
Chris Lattner8f33c262010-04-22 06:28:20 +0000[diff] [blame]683
684<li>A new codegen level Common Subexpression Elimination pass (MachineCSE)
685 is available and enabled by default. It catches redundancies exposed by
686 lowering.</li>
687<li>A new pre-register-allocation tail duplication pass is available and enabled
688 by default, it can substantially improve branch prediction quality in some
689 cases.</li>
690<li>A new sign and zero extension optimization pass (OptimizeExtsPass)
691 is available and enabled by default. This pass can takes advantage
692 architecture features like x86-64 implicit zero extension behavior and
693 sub-registers.</li>
694<li>The code generator now supports a mode where it attempts to preserve the
695 order of instructions in the input code. This is important for source that
696 is hand scheduled and extremely sensitive to scheduling. It is compatible
697 with the GCC <tt>-fno-schedule-insns</tt> option.</li>
698<li>The target-independent code generator now supports generating code with
699 arbitrary numbers of result values. Returning more values than was
700 previously supported is handled by returning through a hidden pointer. In
701 2.7, only the X86 and XCore targets have adopted support for this
702 though.</li>
703<li>The code generator now supports generating code that follows the
704 <a href="LangRef.html#callingconv">Glasgow Haskell Compiler Calling
705 Convention</a> and ABI.</li>
706<li>The "<a href="CodeGenerator.html#selectiondag_select">DAG instruction
707 selection</a>" phase of the code generator has been largely rewritten for
708 2.7. Previously, tblgen spit out tons of C++ code which was compiled and
709 linked into the target to do the pattern matching, now it emits a much
710 smaller table which is read by the target-independent code. The primary
711 advantages of this approach is that the size and compile time of various
712 targets is much improved. The X86 code generator shrunk by 1.5MB of code,
713 for example.</li>
714<li>Almost the entire code generator has switched to emitting code through the
715 MC interfaces instead of printing textually to the .s file. This led to a
716 number of cleanups and speedups. In 2.7, debug an exception handling
717 information does not go through MC yet.</li>
Chris Lattnerb0d6c502007-09-21 03:54:09 +0000[diff] [blame]718</ul>
Chris Lattnerb0d6c502007-09-21 03:54:09 +0000[diff] [blame]719</div>
720
Chris Lattner47202c32009-03-02 02:37:32 +0000[diff] [blame]721<!--=========================================================================-->
722<div class="doc_subsection">
723<a name="x86">X86-32 and X86-64 Target Improvements</a>
724</div>
725
726<div class="doc_text">
727<p>New features of the X86 target include:
728</p>
729
730<ul>
Chris Lattner556aac32010-04-21 06:42:24 +0000[diff] [blame]731<li>The X86 backend now optimizes tails calls much more aggressively for
732 functions that use the standard C calling convention.</li>
733<li>The X86 backend now models scalar SSE registers as subregs of the SSE vector
734 registers, making the code generator more aggressive in cases where scalars
735 and vector types are mixed.</li>
Chris Lattner5f393542009-03-02 03:24:11 +0000[diff] [blame]736
Chris Lattner47202c32009-03-02 02:37:32 +0000[diff] [blame]737</ul>
738
Chris Lattner47202c32009-03-02 02:37:32 +0000[diff] [blame]739</div>
Chris Lattnerb0d6c502007-09-21 03:54:09 +0000[diff] [blame]740
741<!--=========================================================================-->
742<div class="doc_subsection">
Chris Lattner0b6ea7a2009-10-13 17:48:04 +0000[diff] [blame]743<a name="ARM">ARM Target Improvements</a>
Chris Lattner7ccc40d2009-03-01 02:30:21 +0000[diff] [blame]744</div>
745
746<div class="doc_text">
Chris Lattner0b6ea7a2009-10-13 17:48:04 +0000[diff] [blame]747<p>New features of the ARM target include:
748</p>
Chris Lattner7ccc40d2009-03-01 02:30:21 +0000[diff] [blame]749
750<ul>
Chris Lattner7ccc40d2009-03-01 02:30:21 +0000[diff] [blame]751
Chris Lattner8a57e892010-04-21 06:23:40 +0000[diff] [blame]752<li>The ARM backend now generates instructions in unified assembly syntax.</li>
753
754<li>llvm-gcc now has complete support for the ARM v7 NEON instruction set. This
755 support differs slightly from the GCC implementation. Please see the
756 <a
757href="http://blog.llvm.org/2010/04/arm-advanced-simd-neon-intrinsics-and.html">
758 ARM Advanced SIMD (NEON) Intrinsics and Types in LLVM Blog Post</a> for
759 helpful information if migrating code from GCC to LLVM-GCC.</li>
760
Wesley Peck0e9beac2010-04-22 13:43:14 +0000[diff] [blame]761<li>The ARM and Thumb code generators now use register scavenging for stack
Jim Grosbachc5595772010-04-22 18:28:43 +0000[diff] [blame]762 object address materialization. This allows the use of R3 as a general
763 purpose register in Thumb1 code, as it was previous reserved for use in
764 stack address materialization. Secondly, sequential uses of the same
765 value will now re-use the materialized constant.</li>
766
Wesley Peck0e9beac2010-04-22 13:43:14 +0000[diff] [blame]767<li>The ARM backend now has good support for ARMv4 targets and has been tested
Chris Lattner8a57e892010-04-21 06:23:40 +0000[diff] [blame]768 on StrongARM hardware. Previously, LLVM only supported ARMv4T and
769 newer chips.</li>
Jim Grosbach95254a42010-04-22 18:33:31 +0000[diff] [blame]770
771<li>Atomic builtins are now supported for ARMv6 and ARMv7 (__sync_synchronize,
772 __sync_fetch_and_add, etc.).</li>
773
Chris Lattner0b6ea7a2009-10-13 17:48:04 +0000[diff] [blame]774</ul>
Chris Lattner7ccc40d2009-03-01 02:30:21 +0000[diff] [blame]775
Chris Lattner7ccc40d2009-03-01 02:30:21 +0000[diff] [blame]776
Chris Lattner0b6ea7a2009-10-13 17:48:04 +0000[diff] [blame]777</div>
Chris Lattner7ccc40d2009-03-01 02:30:21 +0000[diff] [blame]778
Chris Lattner0b6ea7a2009-10-13 17:48:04 +0000[diff] [blame]779<!--=========================================================================-->
780<div class="doc_subsection">
Chris Lattner0b6ea7a2009-10-13 17:48:04 +0000[diff] [blame]781<a name="newapis">New Useful APIs</a>
782</div>
783
784<div class="doc_text">
785
786<p>This release includes a number of new APIs that are used internally, which
787 may also be useful for external clients.
788</p>
789
790<ul>
Chris Lattnerabfaf732010-04-22 05:41:35 +0000[diff] [blame]791<li>The optimizer uses the new CodeMetrics class to measure the size of code.
Chris Lattner9d966ed2010-04-22 17:39:38 +0000[diff] [blame]792 Various passes (like the inliner, loop unswitcher, etc) all use this to make
793 more accurate estimates of the code size impact of various
Chris Lattnerabfaf732010-04-22 05:41:35 +0000[diff] [blame]794 optimizations.</li>
795<li>A new <a href="http://llvm.org/doxygen/InstructionSimplify_8h-source.html">
Wesley Peck7b6a2ec2010-04-22 13:50:46 +0000[diff] [blame]796 llvm/Analysis/InstructionSimplify.h</a> interface is available for doing
Chris Lattnerabfaf732010-04-22 05:41:35 +0000[diff] [blame]797 symbolic simplification of instructions (e.g. <tt>a+0</tt> -&gt; <tt>a</tt>)
798 without requiring the instruction to exist. This centralizes a lot of
799 ad-hoc symbolic manipulation code scattered in various passes.</li>
800<li>The optimizer now uses a new <a
801 href="http://llvm.org/doxygen/SSAUpdater_8h-source.html">SSAUpdater</a>
802 class which efficiently supports
803 doing unstructured SSA update operations. This centralized a bunch of code
Wesley Peck7b6a2ec2010-04-22 13:50:46 +0000[diff] [blame]804 scattered throughout various passes (e.g. jump threading, lcssa,
805 loop rotate, etc) for doing this sort of thing. The code generator has a
806 similar <a href="http://llvm.org/doxygen/MachineSSAUpdater_8h-source.html">
Chris Lattnerabfaf732010-04-22 05:41:35 +0000[diff] [blame]807 MachineSSAUpdater</a> class.</li>
808<li>The <a href="http://llvm.org/doxygen/Regex_8h-source.html">
809 llvm/Support/Regex.h</a> header exposes a platform independent regular
Chris Lattner8a57e892010-04-21 06:23:40 +0000[diff] [blame]810 expression API. Building on this, the <a
811 href="TestingGuide.html#FileCheck">FileCheck</a> utility now supports
812 regular exressions.</li>
813<li>raw_ostream now supports a circular "debug stream" accessed with "dbgs()".
814 By default, this stream works the same way as "errs()", but if you pass
815 <tt>-debug-buffer-size=1000</tt> to opt, the debug stream is capped to a
816 fixed sized circular buffer and the output is printed at the end of the
817 program's execution. This is helpful if you have a long lived compiler
818 process and you're interested in seeing snapshots in time.</li>
Chris Lattner0b6ea7a2009-10-13 17:48:04 +0000[diff] [blame]819</ul>
820
821
822</div>
823
824<!--=========================================================================-->
825<div class="doc_subsection">
826<a name="otherimprovements">Other Improvements and New Features</a>
827</div>
828
829<div class="doc_text">
830<p>Other miscellaneous features include:</p>
831
832<ul>
Chris Lattner8a57e892010-04-21 06:23:40 +0000[diff] [blame]833<li>You can now build LLVM as a big dynamic library (e.g. "libllvm2.7.so"). To
834 get this, configure LLVM with the --enable-shared option.</li>
835
Wesley Pecka85292d2010-04-22 13:53:54 +0000[diff] [blame]836<li>LLVM command line tools now overwrite their output by default. Previously,
837 they would only do this with -f. This makes them more convenient to use, and
Chris Lattner6fd928d2010-04-21 05:17:40 +0000[diff] [blame]838 behave more like standard unix tools.</li>
839
840<li>The opt and llc tools now autodetect whether their input is a .ll or .bc
841 file, and automatically do the right thing. This means you don't need to
842 explicitly use the llvm-as tool for most things.</li>
Chris Lattner7ccc40d2009-03-01 02:30:21 +0000[diff] [blame]843</ul>
844
845</div>
846
Chris Lattnere45f18b2008-06-05 08:02:49 +0000[diff] [blame]847
848<!--=========================================================================-->
849<div class="doc_subsection">
Chris Lattner92eacf82008-10-13 17:57:36 +0000[diff] [blame]850<a name="changes">Major Changes and Removed Features</a>
851</div>
852
853<div class="doc_text">
854
Chris Lattnercc64a322008-10-13 22:06:31 +0000[diff] [blame]855<p>If you're already an LLVM user or developer with out-of-tree changes based
Jeffrey Yasskin8ee1b0a2010-01-28 01:14:43 +0000[diff] [blame]856on LLVM 2.6, this section lists some "gotchas" that you may run into upgrading
Chris Lattnercc64a322008-10-13 22:06:31 +0000[diff] [blame]857from the previous release.</p>
Chris Lattner92eacf82008-10-13 17:57:36 +0000[diff] [blame]858
859<ul>
Chris Lattner6fd928d2010-04-21 05:17:40 +0000[diff] [blame]860
861<li>
862The Andersen's alias analysis ("anders-aa") pass, the Predicate Simplifier
863("predsimplify") pass, the LoopVR pass, the GVNPRE pass, and the random sampling
864profiling ("rsprofiling") passes have all been removed. They were not being
865actively maintained and had substantial problems. If you are interested in
866these components, you are welcome to ressurect them from SVN, fix the
867correctness problems, and resubmit them to mainline.</li>
868
869<li>LLVM now defaults to building most libraries with RTTI turned off, providing
870a code size reduction. Packagers who are interested in building LLVM to support
871plugins that require RTTI information should build with "make REQUIRE_RTTI=1"
872and should read the new <a href="Packaging.html">Advice on Packaging LLVM</a>
873document.</li>
874
Jeffrey Yasskin260af332010-02-09 23:03:44 +0000[diff] [blame]875<li>The LLVM interpreter now defaults to <em>not</em> using <tt>libffi</tt> even
876if you have it installed. This makes it more likely that an LLVM built on one
877system will work when copied to a similar system. To use <tt>libffi</tt>,
Chris Lattner8f33c262010-04-22 06:28:20 +0000[diff] [blame]878configure with <tt>--enable-libffi</tt>.</li>
Chris Lattnercc64a322008-10-13 22:06:31 +0000[diff] [blame]879
Chris Lattner6fd928d2010-04-21 05:17:40 +0000[diff] [blame]880<li>Debug information uses a completely different representation, an LLVM 2.6
881.bc file should work with LLVM 2.7, but debug info won't come forward.</li>
882
883<li>The LLVM 2.6 (and earlier) "malloc" and "free" instructions got removed,
884 along with LowerAllocations pass. Now you should just use a call to the
885 malloc and free functions in libc. These calls are optimized as well as
886 the old instructions were.</li>
887</ul>
Chris Lattnercc64a322008-10-13 22:06:31 +0000[diff] [blame]888
889<p>In addition, many APIs have changed in this release. Some of the major LLVM
890API changes are:</p>
891
892<ul>
Chris Lattner6fd928d2010-04-21 05:17:40 +0000[diff] [blame]893<li>Just about everything has been converted to use raw_ostream instead of
894 std::ostream.</li>
895<li>llvm/ADT/iterator.h has been removed, just use &lt;iterator&gt;
896 instead.</li>
897<li>The Streams.h file and "DOUT" got removed, use "DEBUG(errs() &lt;&lt; ...);"
898 instead.</li>
Chris Lattner43f7a1d2010-04-22 17:25:00 +0000[diff] [blame]899<li>The TargetAsmInfo interface was renamed to MCAsmInfo.</li>
Jeffrey Yasskin2ca16212010-01-28 01:41:20 +0000[diff] [blame]900<li><tt>ModuleProvider</tt> has been <a
Gabor Greif92d28ea2010-04-22 10:21:43 +0000[diff] [blame]901href="http://llvm.org/viewvc/llvm-project?view=rev&amp;revision=94686">removed</a>
Jeffrey Yasskin2ca16212010-01-28 01:41:20 +0000[diff] [blame]902and its methods moved to <tt>Module</tt> and <tt>GlobalValue</tt>.
903Most clients can remove uses of <tt>ExistingModuleProvider</tt>,
904replace <tt>getBitcodeModuleProvider</tt> with
905<tt>getLazyBitcodeModule</tt>, and pass their <tt>Module</tt> to
906functions that used to accept <tt>ModuleProvider</tt>. Clients who
907wrote their own <tt>ModuleProvider</tt>s will need to derive from
908<tt>GVMaterializer</tt> instead and use
909<tt>Module::setMaterializer</tt> to attach it to a
910<tt>Module</tt>.</li>
911
912<li><tt>GhostLinkage</tt> has given up the ghost.
913<tt>GlobalValue</tt>s that have not yet been read from their backing
914storage have the same linkage they will have after being read in.
915Clients must replace calls to
916<tt>GlobalValue::hasNotBeenReadFromBitcode</tt> with
917<tt>GlobalValue::isMaterializable</tt>.</li>
Daniel Dunbar61f41422010-02-10 04:09:52 +0000[diff] [blame]918
Duncan Sandsdd29fc82010-02-17 17:20:17 +0000[diff] [blame]919<li>The <tt>isInteger</tt>, <tt>isIntOrIntVector</tt>, <tt>isFloatingPoint</tt>,
920<tt>isFPOrFPVector</tt> and <tt>isFPOrFPVector</tt> methods have been renamed
921<tt>isIntegerTy</tt>, <tt>isIntOrIntVectorTy</tt>, <tt>isFloatingPointTy</tt>,
922<tt>isFPOrFPVectorTy</tt> and <tt>isFPOrFPVectorTy</tt> respectively.</li>
Chris Lattner43f7a1d2010-04-22 17:25:00 +0000[diff] [blame]923
924<li><tt>llvm::Instruction::clone()</tt> no longer takes argument.</li>
925<li><tt>raw_fd_ostream</tt>'s constructor now takes a flag argument, not individual
926 booleans (see <tt>include/llvm/Support/raw_ostream.h</tt> for details).</li>
927<li>Some header files have been renamed:
928<ul>
929 <li><tt>llvm/Support/AIXDataTypesFix.h</tt> to
930 <tt>llvm/System/AIXDataTypesFix.h</tt></li>
931 <li><tt>llvm/Support/DataTypes.h</tt> to <tt>llvm/System/DataTypes.h</tt></li>
932 <li><tt>llvm/Transforms/Utils/InlineCost.h</tt> to
933 <tt>llvm/Analysis/InlineCost.h</tt></li>
934 <li><tt>llvm/Support/Mangler.h</tt> to <tt>llvm/Target/Mangler.h</tt></li>
935 <li><tt>llvm/Analysis/Passes.h</tt> to <tt>llvm/CodeGen/Passes.h</tt></li>
936</ul></li>
Devang Patel0e3da1a2008-10-14 20:03:43 +0000[diff] [blame]937</ul>
Chris Lattner92eacf82008-10-13 17:57:36 +0000[diff] [blame]938
Chris Lattner92eacf82008-10-13 17:57:36 +0000[diff] [blame]939</div>
940
941
942
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]943<!-- *********************************************************************** -->
944<div class="doc_section">
945 <a name="portability">Portability and Supported Platforms</a>
946</div>
947<!-- *********************************************************************** -->
948
949<div class="doc_text">
950
951<p>LLVM is known to work on the following platforms:</p>
952
953<ul>
Mikhail Glushenkovf61066a2009-03-01 18:09:47 +0000[diff] [blame]954<li>Intel and AMD machines (IA32, X86-64, AMD64, EMT-64) running Red Hat
Chris Lattner0b6ea7a2009-10-13 17:48:04 +0000[diff] [blame]955 Linux, Fedora Core, FreeBSD and AuroraUX (and probably other unix-like
956 systems).</li>
Chris Lattner8f33c262010-04-22 06:28:20 +0000[diff] [blame]957<li>PowerPC and X86-based Mac OS X systems, running 10.4 and above in 32-bit
Chris Lattner0b6ea7a2009-10-13 17:48:04 +0000[diff] [blame]958 and 64-bit modes.</li>
Chris Lattnere0798082008-02-06 18:00:06 +0000[diff] [blame]959<li>Intel and AMD machines running on Win32 using MinGW libraries (native).</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]960<li>Intel and AMD machines running on Win32 with the Cygwin libraries (limited
961 support is available for native builds with Visual C++).</li>
Chris Lattner9235ce52009-10-16 16:30:58 +0000[diff] [blame]962<li>Sun x86 and AMD64 machines running Solaris 10, OpenSolaris 0906.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]963<li>Alpha-based machines running Debian GNU/Linux.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]964</ul>
965
Chris Lattnerba948522008-06-05 06:57:39 +0000[diff] [blame]966<p>The core LLVM infrastructure uses GNU autoconf to adapt itself
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]967to the machine and operating system on which it is built. However, minor
968porting may be required to get LLVM to work on new platforms. We welcome your
969portability patches and reports of successful builds or error messages.</p>
970
971</div>
972
973<!-- *********************************************************************** -->
974<div class="doc_section">
975 <a name="knownproblems">Known Problems</a>
976</div>
977<!-- *********************************************************************** -->
978
979<div class="doc_text">
980
Mikhail Glushenkovf61066a2009-03-01 18:09:47 +0000[diff] [blame]981<p>This section contains significant known problems with the LLVM system,
Chris Lattnerb1ad47c2008-11-10 05:40:34 +0000[diff] [blame]982listed by component. If you run into a problem, please check the <a
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]983href="http://llvm.org/bugs/">LLVM bug database</a> and submit a bug if
984there isn't already one.</p>
985
Chris Lattner1f8dc512010-03-17 17:25:49 +0000[diff] [blame]986<ul>
Chris Lattner0b6ea7a2009-10-13 17:48:04 +0000[diff] [blame]987<li>LLVM will not correctly compile on Solaris and/or OpenSolaris
988using the stock GCC 3.x.x series 'out the box',
Chris Lattnerc09ffd22009-11-03 21:50:09 +0000[diff] [blame]989See: <a href="GettingStarted.html#brokengcc">Broken versions of GCC and other tools</a>.
Chris Lattner0b6ea7a2009-10-13 17:48:04 +0000[diff] [blame]990However, A <a href="http://pkg.auroraux.org/GCC">Modern GCC Build</a>
991for x86/x86-64 has been made available from the third party AuroraUX Project
992that has been meticulously tested for bootstrapping LLVM &amp; Clang.</li>
993</ul>
994
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]995</div>
996
997<!-- ======================================================================= -->
998<div class="doc_subsection">
999 <a name="experimental">Experimental features included with this release</a>
1000</div>
1001
1002<div class="doc_text">
1003
1004<p>The following components of this LLVM release are either untested, known to
1005be broken or unreliable, or are in early development. These components should
1006not be relied on, and bugs should not be filed against them, but they may be
1007useful to some people. In particular, if you would like to work on one of these
Chris Lattnerdaeb3232008-02-12 06:29:45 +0000[diff] [blame]1008components, please contact us on the <a
1009href="http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev">LLVMdev list</a>.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1010
1011<ul>
Wesley Peckbe6dc822010-03-18 14:31:30 +0000[diff] [blame]1012<li>The MSIL, Alpha, SPU, MIPS, PIC16, Blackfin, MSP430, SystemZ and MicroBlaze
1013 backends are experimental.</li>
Chris Lattner8f33c262010-04-22 06:28:20 +0000[diff] [blame]1014<li><tt>llc</tt> "<tt>-filetype=asm</tt>" (the default) is the only
Chris Lattner58b1d0f2010-03-17 04:41:49 +0000[diff] [blame]1015 supported value for this option. The MachO writer is experimental, and
1016 works much better in mainline SVN.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1017</ul>
1018
1019</div>
1020
1021<!-- ======================================================================= -->
1022<div class="doc_subsection">
1023 <a name="x86-be">Known problems with the X86 back-end</a>
1024</div>
1025
1026<div class="doc_text">
1027
1028<ul>
Anton Korobeynikov77d15e22008-06-08 10:24:13 +0000[diff] [blame]1029 <li>The X86 backend does not yet support
1030 all <a href="http://llvm.org/PR879">inline assembly that uses the X86
1031 floating point stack</a>. It supports the 'f' and 't' constraints, but not
1032 'u'.</li>
Chris Lattnerd95b5962008-06-08 21:19:07 +0000[diff] [blame]1033 <li>The X86 backend generates inefficient floating point code when configured
1034 to generate code for systems that don't have SSE2.</li>
Duncan Sands8b4f3772008-06-08 19:38:43 +0000[diff] [blame]1035 <li>Win64 code generation wasn't widely tested. Everything should work, but we
John Criswellaea8c752009-03-02 15:28:15 +0000[diff] [blame]1036 expect small issues to happen. Also, llvm-gcc cannot build the mingw64
Chris Lattner8f33c262010-04-22 06:28:20 +0000[diff] [blame]1037 runtime currently due to lack of support for the 'u' inline assembly
1038 constraint and for X87 floating point inline assembly.</li>
Dan Gohman3882ecb2008-06-08 23:05:11 +0000[diff] [blame]1039 <li>The X86-64 backend does not yet support the LLVM IR instruction
Chris Lattner8f33c262010-04-22 06:28:20 +0000[diff] [blame]1040 <tt>va_arg</tt>. Currently, front-ends support variadic
Dan Gohman3882ecb2008-06-08 23:05:11 +0000[diff] [blame]1041 argument constructs on X86-64 by lowering them manually.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1042</ul>
1043
1044</div>
1045
1046<!-- ======================================================================= -->
1047<div class="doc_subsection">
1048 <a name="ppc-be">Known problems with the PowerPC back-end</a>
1049</div>
1050
1051<div class="doc_text">
1052
1053<ul>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1054<li>The Linux PPC32/ABI support needs testing for the interpreter and static
1055compilation, and lacks support for debug information.</li>
1056</ul>
1057
1058</div>
1059
1060<!-- ======================================================================= -->
1061<div class="doc_subsection">
1062 <a name="arm-be">Known problems with the ARM back-end</a>
1063</div>
1064
1065<div class="doc_text">
1066
1067<ul>
1068<li>Thumb mode works only on ARMv6 or higher processors. On sub-ARMv6
Duncan Sands1827fa62007-09-26 15:59:54 +0000[diff] [blame]1069processors, thumb programs can crash or produce wrong
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1070results (<a href="http://llvm.org/PR1388">PR1388</a>).</li>
John Criswellaea8c752009-03-02 15:28:15 +0000[diff] [blame]1071<li>Compilation for ARM Linux OABI (old ABI) is supported but not fully tested.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1072</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1073</ul>
1074
1075</div>
1076
1077<!-- ======================================================================= -->
1078<div class="doc_subsection">
1079 <a name="sparc-be">Known problems with the SPARC back-end</a>
1080</div>
1081
1082<div class="doc_text">
1083
1084<ul>
John Criswellaea8c752009-03-02 15:28:15 +0000[diff] [blame]1085<li>The SPARC backend only supports the 32-bit SPARC ABI (-m32); it does not
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1086 support the 64-bit SPARC ABI (-m64).</li>
1087</ul>
1088
1089</div>
1090
1091<!-- ======================================================================= -->
1092<div class="doc_subsection">
Bruno Cardoso Lopes62d45d72008-10-25 14:56:26 +0000[diff] [blame]1093 <a name="mips-be">Known problems with the MIPS back-end</a>
1094</div>
1095
1096<div class="doc_text">
1097
1098<ul>
Bruno Cardoso Lopes62d45d72008-10-25 14:56:26 +0000[diff] [blame]1099<li>64-bit MIPS targets are not supported yet.</li>
1100</ul>
1101
1102</div>
1103
1104<!-- ======================================================================= -->
1105<div class="doc_subsection">
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1106 <a name="alpha-be">Known problems with the Alpha back-end</a>
1107</div>
1108
1109<div class="doc_text">
1110
1111<ul>
1112
1113<li>On 21164s, some rare FP arithmetic sequences which may trap do not have the
1114appropriate nops inserted to ensure restartability.</li>
1115
1116</ul>
1117</div>
1118
1119<!-- ======================================================================= -->
1120<div class="doc_subsection">
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1121 <a name="c-be">Known problems with the C back-end</a>
1122</div>
1123
1124<div class="doc_text">
1125
1126<ul>
Chris Lattner499c07c2008-06-05 06:35:40 +0000[diff] [blame]1127<li><a href="http://llvm.org/PR802">The C backend has only basic support for
1128 inline assembly code</a>.</li>
Chris Lattnere0194462007-09-26 06:01:35 +0000[diff] [blame]1129<li><a href="http://llvm.org/PR1658">The C backend violates the ABI of common
1130 C++ programs</a>, preventing intermixing between C++ compiled by the CBE and
Gabor Greife33c5322009-03-02 12:02:51 +0000[diff] [blame]1131 C++ code compiled with <tt>llc</tt> or native compilers.</li>
Duncan Sandsbd0ad6e2008-02-10 13:40:55 +0000[diff] [blame]1132<li>The C backend does not support all exception handling constructs.</li>
Duncan Sandse6e264c2009-02-25 11:51:54 +0000[diff] [blame]1133<li>The C backend does not support arbitrary precision integers.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1134</ul>
1135
1136</div>
1137
1138
1139<!-- ======================================================================= -->
1140<div class="doc_subsection">
Chris Lattner58b1d0f2010-03-17 04:41:49 +0000[diff] [blame]1141 <a name="c-fe">Known problems with the llvm-gcc C and C++ front-end</a>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1142</div>
1143
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1144<div class="doc_text">
1145
Chris Lattner499c07c2008-06-05 06:35:40 +0000[diff] [blame]1146<p>The only major language feature of GCC not supported by llvm-gcc is
1147 the <tt>__builtin_apply</tt> family of builtins. However, some extensions
1148 are only supported on some targets. For example, trampolines are only
Duncan Sandsa73561c2008-06-08 20:18:35 +0000[diff] [blame]1149 supported on some targets (these are used when you take the address of a
1150 nested function).</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1151
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1152</div>
1153
1154<!-- ======================================================================= -->
1155<div class="doc_subsection">
Chris Lattner6c8a1b82008-10-30 03:58:13 +0000[diff] [blame]1156 <a name="fortran-fe">Known problems with the llvm-gcc Fortran front-end</a>
1157</div>
1158
1159<div class="doc_text">
Gabor Greif1fc8fbc2008-11-04 21:50:59 +0000[diff] [blame]1160<ul>
Chris Lattner6c8a1b82008-10-30 03:58:13 +0000[diff] [blame]1161<li>Fortran support generally works, but there are still several unresolved bugs
Chris Lattner0b6ea7a2009-10-13 17:48:04 +0000[diff] [blame]1162 in <a href="http://llvm.org/bugs/">Bugzilla</a>. Please see the
1163 tools/gfortran component for details.</li>
Gabor Greif1fc8fbc2008-11-04 21:50:59 +0000[diff] [blame]1164</ul>
Chris Lattner6c8a1b82008-10-30 03:58:13 +0000[diff] [blame]1165</div>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1166
Chris Lattnerdaeb3232008-02-12 06:29:45 +0000[diff] [blame]1167<!-- ======================================================================= -->
1168<div class="doc_subsection">
Chris Lattner499c07c2008-06-05 06:35:40 +0000[diff] [blame]1169 <a name="ada-fe">Known problems with the llvm-gcc Ada front-end</a>
Chris Lattnerdaeb3232008-02-12 06:29:45 +0000[diff] [blame]1170</div>
1171
1172<div class="doc_text">
John Criswellaea8c752009-03-02 15:28:15 +0000[diff] [blame]1173The llvm-gcc 4.2 Ada compiler works fairly well; however, this is not a mature
1174technology, and problems should be expected.
Chris Lattnerdaeb3232008-02-12 06:29:45 +0000[diff] [blame]1175<ul>
Duncan Sandsa73561c2008-06-08 20:18:35 +0000[diff] [blame]1176<li>The Ada front-end currently only builds on X86-32. This is mainly due
John Criswellaea8c752009-03-02 15:28:15 +0000[diff] [blame]1177to lack of trampoline support (pointers to nested functions) on other platforms.
1178However, it <a href="http://llvm.org/PR2006">also fails to build on X86-64</a>
Chris Lattnerdaeb3232008-02-12 06:29:45 +0000[diff] [blame]1179which does support trampolines.</li>
1180<li>The Ada front-end <a href="http://llvm.org/PR2007">fails to bootstrap</a>.
Duncan Sandsf73473e2009-02-25 11:59:06 +0000[diff] [blame]1181This is due to lack of LLVM support for <tt>setjmp</tt>/<tt>longjmp</tt> style
1182exception handling, which is used internally by the compiler.
Chris Lattner0b6ea7a2009-10-13 17:48:04 +0000[diff] [blame]1183Workaround: configure with <tt>--disable-bootstrap</tt>.</li>
Duncan Sands2dbc2e82008-10-13 17:27:23 +0000[diff] [blame]1184<li>The c380004, <a href="http://llvm.org/PR2010">c393010</a>
1185and <a href="http://llvm.org/PR2421">cxg2021</a> ACATS tests fail
Duncan Sandsf73473e2009-02-25 11:59:06 +0000[diff] [blame]1186(c380004 also fails with gcc-4.2 mainline).
1187If the compiler is built with checks disabled then <a href="http://llvm.org/PR2010">c393010</a>
1188causes the compiler to go into an infinite loop, using up all system memory.</li>
Duncan Sandsc2a29a02009-03-02 16:35:57 +0000[diff] [blame]1189<li>Some GCC specific Ada tests continue to crash the compiler.</li>
Chris Lattner0b6ea7a2009-10-13 17:48:04 +0000[diff] [blame]1190<li>The <tt>-E</tt> binder option (exception backtraces)
Chris Lattnerdaeb3232008-02-12 06:29:45 +0000[diff] [blame]1191<a href="http://llvm.org/PR1982">does not work</a> and will result in programs
Chris Lattner0b6ea7a2009-10-13 17:48:04 +0000[diff] [blame]1192crashing if an exception is raised. Workaround: do not use <tt>-E</tt>.</li>
Chris Lattnerdaeb3232008-02-12 06:29:45 +0000[diff] [blame]1193<li>Only discrete types <a href="http://llvm.org/PR1981">are allowed to start
1194or finish at a non-byte offset</a> in a record. Workaround: do not pack records
1195or use representation clauses that result in a field of a non-discrete type
1196starting or finishing in the middle of a byte.</li>
Chris Lattnerd95b5962008-06-08 21:19:07 +0000[diff] [blame]1197<li>The <tt>lli</tt> interpreter <a href="http://llvm.org/PR2009">considers
1198'main' as generated by the Ada binder to be invalid</a>.
1199Workaround: hand edit the file to use pointers for <tt>argv</tt> and
1200<tt>envp</tt> rather than integers.</li>
1201<li>The <tt>-fstack-check</tt> option <a href="http://llvm.org/PR2008">is
1202ignored</a>.</li>
Chris Lattnerdaeb3232008-02-12 06:29:45 +0000[diff] [blame]1203</ul>
1204</div>
1205
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1206<!-- *********************************************************************** -->
1207<div class="doc_section">
1208 <a name="additionalinfo">Additional Information</a>
1209</div>
1210<!-- *********************************************************************** -->
1211
1212<div class="doc_text">
1213
1214<p>A wide variety of additional information is available on the <a
1215href="http://llvm.org">LLVM web page</a>, in particular in the <a
1216href="http://llvm.org/docs/">documentation</a> section. The web page also
1217contains versions of the API documentation which is up-to-date with the
1218Subversion version of the source code.
1219You can access versions of these documents specific to this release by going
1220into the "<tt>llvm/doc/</tt>" directory in the LLVM tree.</p>
1221
1222<p>If you have any questions or comments about LLVM, please feel free to contact
1223us via the <a href="http://llvm.org/docs/#maillist"> mailing
1224lists</a>.</p>
1225
1226</div>
1227
1228<!-- *********************************************************************** -->
1229
1230<hr>
1231<address>
1232 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
Misha Brukman947321d2008-12-11 17:34:48 +0000[diff] [blame]1233 src="http://jigsaw.w3.org/css-validator/images/vcss-blue" alt="Valid CSS"></a>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1234 <a href="http://validator.w3.org/check/referer"><img
Misha Brukman947321d2008-12-11 17:34:48 +0000[diff] [blame]1235 src="http://www.w3.org/Icons/valid-html401-blue" alt="Valid HTML 4.01"></a>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1236
1237 <a href="http://llvm.org/">LLVM Compiler Infrastructure</a><br>
1238 Last modified: $Date$
1239</address>
1240
1241</body>
1242</html>