Gitiles
Code Review Sign In
gerrit-public.fairphone.software / fp2-dev / platform / external / llvm / 43b65e99061b5306bb4616f903788126952a8a0d / . / docs / ReleaseNotes.html
blob: c8aa59ffa4ec3c184da864c3d11ab1eb62da7e6c [file] [log] [blame]
Anton Korobeynikov0d176912008-06-08 10:23:46 +0000[diff] [blame]1
Misha Brukman500bc302003-11-22 00:38:41 +0000[diff] [blame]2<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01//EN"
3 "http://www.w3.org/TR/html4/strict.dtd">
4<html>
5<head>
Reid Spencer6454ed32004-11-18 18:38:58 +0000[diff] [blame]6 <meta http-equiv="Content-Type" content="text/html; charset=utf-8">
Misha Brukman500bc302003-11-22 00:38:41 +0000[diff] [blame]7 <link rel="stylesheet" href="llvm.css" type="text/css">
Chris Lattner2b659ef2008-02-12 06:29:45 +0000[diff] [blame]8 <title>LLVM 2.3 Release Notes</title>
Misha Brukman500bc302003-11-22 00:38:41 +0000[diff] [blame]9</head>
10<body>
Chris Lattner79c3fe12003-10-02 04:57:28 +0000[diff] [blame]11
Chris Lattner2b659ef2008-02-12 06:29:45 +0000[diff] [blame]12<div class="doc_title">LLVM 2.3 Release Notes</div>
Chris Lattner79c3fe12003-10-02 04:57:28 +0000[diff] [blame]13
14<ol>
Misha Brukman500bc302003-11-22 00:38:41 +0000[diff] [blame]15 <li><a href="#intro">Introduction</a></li>
Chris Lattner8348b472008-06-08 21:58:17 +0000[diff] [blame]16 <li><a href="#changes">Major Changes and Sub-project Status</a></li>
Misha Brukman500bc302003-11-22 00:38:41 +0000[diff] [blame]17 <li><a href="#whatsnew">What's New?</a></li>
Chris Lattner4b538b92004-04-30 22:17:12 +0000[diff] [blame]18 <li><a href="GettingStarted.html">Installation Instructions</a></li>
Misha Brukman500bc302003-11-22 00:38:41 +0000[diff] [blame]19 <li><a href="#portability">Portability and Supported Platforms</a></li>
Chris Lattner79c3fe12003-10-02 04:57:28 +0000[diff] [blame]20 <li><a href="#knownproblems">Known Problems</a>
Misha Brukman500bc302003-11-22 00:38:41 +0000[diff] [blame]21 <li><a href="#additionalinfo">Additional Information</a></li>
Chris Lattner79c3fe12003-10-02 04:57:28 +0000[diff] [blame]22</ol>
23
Chris Lattner7911ce22004-05-23 21:07:27 +0000[diff] [blame]24<div class="doc_author">
Chris Lattnerc463b272005-10-29 07:07:09 +0000[diff] [blame]25 <p>Written by the <a href="http://llvm.org">LLVM Team</a><p>
Misha Brukman500bc302003-11-22 00:38:41 +0000[diff] [blame]26</div>
Chris Lattner79c3fe12003-10-02 04:57:28 +0000[diff] [blame]27
Chris Lattner77d29b12008-06-05 08:02:49 +0000[diff] [blame]28<!-- Done through Week-of-Mon-20080324.txt -->
29
Chris Lattner79c3fe12003-10-02 04:57:28 +0000[diff] [blame]30<!-- *********************************************************************** -->
Misha Brukman500bc302003-11-22 00:38:41 +0000[diff] [blame]31<div class="doc_section">
32 <a name="intro">Introduction</a>
33</div>
Chris Lattner79c3fe12003-10-02 04:57:28 +0000[diff] [blame]34<!-- *********************************************************************** -->
35
Misha Brukman500bc302003-11-22 00:38:41 +0000[diff] [blame]36<div class="doc_text">
37
38<p>This document contains the release notes for the LLVM compiler
Chris Lattner2b659ef2008-02-12 06:29:45 +0000[diff] [blame]39infrastructure, release 2.3. Here we describe the status of LLVM, including
Chris Lattner57a460e2007-05-23 04:39:32 +0000[diff] [blame]40major improvements from the previous release and any known problems. All LLVM
Chris Lattnercaf06342007-05-09 04:58:11 +0000[diff] [blame]41releases may be downloaded from the <a href="http://llvm.org/releases/">LLVM
Chris Lattner57a460e2007-05-23 04:39:32 +0000[diff] [blame]42releases web site</a>.</p>
Chris Lattner19092612003-10-02 16:38:05 +0000[diff] [blame]43
Chris Lattner7506b1d2004-12-07 08:04:13 +0000[diff] [blame]44<p>For more information about LLVM, including information about the latest
Chris Lattnerc463b272005-10-29 07:07:09 +0000[diff] [blame]45release, please check out the <a href="http://llvm.org/">main LLVM
Chris Lattner47ad72c2003-10-07 21:38:31 +0000[diff] [blame]46web site</a>. If you have questions or comments, the <a
47href="http://mail.cs.uiuc.edu/mailman/listinfo/llvmdev">LLVM developer's mailing
Misha Brukman500bc302003-11-22 00:38:41 +0000[diff] [blame]48list</a> is a good place to send them.</p>
Chris Lattner79c3fe12003-10-02 04:57:28 +0000[diff] [blame]49
Reid Spencer669ed452007-07-09 08:04:31 +0000[diff] [blame]50<p>Note that if you are reading this file from a Subversion checkout or the
Chris Lattner84977642007-09-21 03:54:09 +0000[diff] [blame]51main LLVM web page, this document applies to the <i>next</i> release, not the
52current one. To see the release notes for a specific releases, please see the
53<a href="http://llvm.org/releases/">releases page</a>.</p>
Misha Brukman500bc302003-11-22 00:38:41 +0000[diff] [blame]54
55</div>
Chris Lattner79c3fe12003-10-02 04:57:28 +0000[diff] [blame]56
57<!-- *********************************************************************** -->
Misha Brukman500bc302003-11-22 00:38:41 +0000[diff] [blame]58<div class="doc_section">
Chris Lattner8348b472008-06-08 21:58:17 +0000[diff] [blame]59 <a name="changes">Major Changes and Sub-project Status</a>
Misha Brukman500bc302003-11-22 00:38:41 +0000[diff] [blame]60</div>
Chris Lattner79c3fe12003-10-02 04:57:28 +0000[diff] [blame]61<!-- *********************************************************************** -->
62
Misha Brukman500bc302003-11-22 00:38:41 +0000[diff] [blame]63<div class="doc_text">
64
Chris Lattner2b659ef2008-02-12 06:29:45 +0000[diff] [blame]65<p>This is the fourteenth public release of the LLVM Compiler Infrastructure.
Chris Lattner791f77b2008-06-05 06:25:56 +0000[diff] [blame]66It includes a large number of features and refinements from LLVM 2.2.</p>
Chris Lattnera5099962003-12-06 20:22:41 +0000[diff] [blame]67
Chris Lattner80453c92004-05-24 04:50:25 +0000[diff] [blame]68</div>
69
Chris Lattner2b659ef2008-02-12 06:29:45 +0000[diff] [blame]70<!-- Unfinished features in 2.3:
Chris Lattneracce85d2008-02-10 07:46:44 +0000[diff] [blame]71 Machine LICM
72 Machine Sinking
73 LegalizeDAGTypes
74 -->
75
Chris Lattnera5099962003-12-06 20:22:41 +0000[diff] [blame]76<!--=========================================================================-->
Chris Lattner8a458762005-05-15 05:44:51 +0000[diff] [blame]77<div class="doc_subsection">
Chris Lattner77d29b12008-06-05 08:02:49 +0000[diff] [blame]78<a name="changes">Major Changes in LLVM 2.3</a>
Chris Lattnerb4c68cd2008-02-10 07:04:35 +0000[diff] [blame]79</div>
80
81<div class="doc_text">
82
Chris Lattnerea34f642008-06-08 21:34:41 +0000[diff] [blame]83<p>LLVM 2.3 no longer supports llvm-gcc 4.0, it has been replaced with
84 llvm-gcc 4.2.</p>
Chris Lattner252b83d2008-02-06 18:00:06 +0000[diff] [blame]85
Chris Lattnerea34f642008-06-08 21:34:41 +0000[diff] [blame]86<p>LLVM 2.3 no longer includes the <tt>llvm-upgrade</tt> tool. It was useful
87 for upgrading LLVM 1.9 files to LLVM 2.x syntax, but you can always use a
88 previous LLVM release to do this. One nice impact of this is that the LLVM
89 regression test suite no longer depends on llvm-upgrade, which makes it run
90 faster.</p>
91
92<p>The <tt>llvm2cpp</tt> tool has been folded into llc, use
Chris Lattnere6e1b352008-06-08 21:19:07 +0000[diff] [blame]93 <tt>llc -march=cpp</tt> instead of <tt>llvm2cpp</tt>.</p>
94
Chris Lattner0b832202008-06-08 02:45:07 +0000[diff] [blame]95<p>LLVM API Changes:</p>
Chris Lattner77d29b12008-06-05 08:02:49 +0000[diff] [blame]96
Chris Lattner0b832202008-06-08 02:45:07 +0000[diff] [blame]97<ul>
Chris Lattnerea34f642008-06-08 21:34:41 +0000[diff] [blame]98<li>Several core LLVM IR classes have migrated to use the
99 '<tt>FOOCLASS::Create(...)</tt>' pattern instead of '<tt>new
100 FOOCLASS(...)</tt>' (e.g. where FOOCLASS=<tt>BasicBlock</tt>). We hope to
101 standardize on <tt>FOOCLASS::Create</tt> for all IR classes in the future,
102 but not all of them have been moved over yet.</li>
Chris Lattner0b832202008-06-08 02:45:07 +0000[diff] [blame]103<li>LLVM 2.3 renames the LLVMBuilder and LLVMFoldingBuilder classes to
Dan Gohmane3280782008-06-08 22:57:59 +0000[diff] [blame]104 <a href="http://llvm.org/doxygen/classllvm_1_1IRBuilder.html">IRBuilder</a>.
105 </li>
106<li>MRegisterInfo was renamed to
107 <a href="http://llvm.org/doxygen/classllvm_1_1TargetRegisterInfo.html">
108 TargetRegisterInfo</a>.</li>
109<li>The MappedFile class is gone, please use
110 <a href="http://llvm.org/doxygen/classllvm_1_1MemoryBuffer.html">
111 MemoryBuffer</a> instead.</li>
Chris Lattnerea34f642008-06-08 21:34:41 +0000[diff] [blame]112<li>The '<tt>-enable-eh</tt>' flag to llc has been removed. Now code should
113 encode whether it is safe to omit unwind information for a function by
114 tagging the Function object with the '<tt>nounwind</tt>' attribute.</li>
Chris Lattner0b832202008-06-08 02:45:07 +0000[diff] [blame]115
116</ul>
Chris Lattner252b83d2008-02-06 18:00:06 +0000[diff] [blame]117</div>
118
119<!--=========================================================================-->
120<div class="doc_subsection">
Chris Lattnerea34f642008-06-08 21:34:41 +0000[diff] [blame]121<a name="otherprojects">Other LLVM Sub-Projects</a>
122</div>
123
124<div class="doc_text">
Chris Lattner8348b472008-06-08 21:58:17 +0000[diff] [blame]125<p>
126The core LLVM 2.3 distribution currently consists of code from the core LLVM
127repository (which roughly contains the LLVM optimizer, code generators and
128supporting tools) and the llvm-gcc repository. In addition to this code, the
129LLVM Project includes other sub-projects that are in development. The two which
130are the most actively developed are the new <a href="#vmkit">vmkit Project</a>
131and the <a href="#clang">Clang Project</a>.
132</p>
133</div>
134
135<!--=========================================================================-->
136<div class="doc_subsubsection">
137<a name="vmkit">vmkit</a>
138</div>
139
140<div class="doc_text">
141<p>
142The "vmkit" project is a new addition to the LLVM family. It is an
143implementation of a JVM and a CLI Virtual Machines (Microsoft .NET is an
144implementation of the CLI) using the Just-In-Time compiler of LLVM.</p>
145
146<p>The JVM, called JnJVM, executes real-world applications such as Apache
147projects (e.g. Felix and Tomcat) and the SpecJVM98 benchmark. It uses the GNU
148Classpath project for the base classes. The CLI implementation, called N3, is
149its in early stages but can execute simple applications and the "pnetmark"
150benchmark. It uses the pnetlib project as its core library.</p>
151
152<p>The 'vmkit' VMs compare in performance with industrial and top open-source
153VMs on scientific applications. Besides the JIT, the VMs use many features of
154the LLVM framework, including the standard set of optimizations, atomic
155operations, custom function provider and memory manager for JITed methods, and
156specific virtual machine optimizations. vmkit is not an official part of LLVM
1572.3 release. It is publicly available under the LLVM license and can be
158downloaded from:
159</p>
Chris Lattnerea34f642008-06-08 21:34:41 +0000[diff] [blame]160
161<p>
Chris Lattner8348b472008-06-08 21:58:17 +0000[diff] [blame]162<tt>svn co http://llvm.org/svn/llvm-project/vmkit/trunk vmkit</tt>
Chris Lattnerea34f642008-06-08 21:34:41 +0000[diff] [blame]163</p>
164
165</div>
166
167<!--=========================================================================-->
Chris Lattner8348b472008-06-08 21:58:17 +0000[diff] [blame]168<div class="doc_subsubsection">
169<a name="clang">Clang</a>
170</div>
171
172<div class="doc_text">
173
174<p>The <a href="http://clang.llvm.org/">Clang project</a> is an effort to build
175a set of new 'LLVM native' front-end technologies for the LLVM optimizer
176and code generator. Clang is continuing to make major strides forward in all
177areas. Its C and Objective-C parsing support is very solid, and the code
178generation support is far enough along to build many C applications. While not
179yet production quality, it is progressing very nicely. In addition, C++
180front-end work has started to make significant progress.</p>
181
182<p>At this point, Clang is most useful if you are interested in source-to-source
183transformations (such as refactoring) and other source-level tools for C and
184Objective-C. Clang now also includes tools for turning C code into pretty HTML,
185and includes a new <a href="http://clang.llvm.org/StaticAnalysis.html">static
186analysis tool</a> in development. This tool is automatically focused on finding
187bugs in C and Objective-C code.</p>
188
189</div>
190
191
192<!-- *********************************************************************** -->
193<div class="doc_section">
194 <a name="whatsnew">What's New?</a>
195</div>
196<!-- *********************************************************************** -->
197
Chris Lattnerf8e0b4e2008-06-08 22:59:35 +0000[diff] [blame]198<div class="doc_text">
199
200<p>LLVM 2.3 includes a huge number of bug fixes, performance tweaks and minor
201improvements. Some of the major improvements and new features are listed in
202this section.
203</p>
204</div>
205
Chris Lattner8348b472008-06-08 21:58:17 +0000[diff] [blame]206<!--=========================================================================-->
Chris Lattnerea34f642008-06-08 21:34:41 +0000[diff] [blame]207<div class="doc_subsection">
Chris Lattner252b83d2008-02-06 18:00:06 +0000[diff] [blame]208<a name="majorfeatures">Major New Features</a>
209</div>
210
211<div class="doc_text">
212
Chris Lattner2b659ef2008-02-12 06:29:45 +0000[diff] [blame]213<p>LLVM 2.3 includes several major new capabilities:</p>
Chris Lattnerf3e5bc62007-05-14 06:56:09 +0000[diff] [blame]214
215<ul>
Chris Lattnerea34f642008-06-08 21:34:41 +0000[diff] [blame]216<li><p>The biggest change in LLVM 2.3 is Multiple Return Value (MRV) support.
217 MRVs allow LLVM IR to directly represent functions that return multiple
218 values without having to pass them "by reference" in the LLVM IR. This
219 allows a front-end to generate more efficient code, as MRVs are generally
220 returned in registers if a target supports them. See the <a
221 href="LangRef.html#i_getresult">LLVM IR Reference</a> for more details.</p>
222
223 <p>MRVs are fully supported in the LLVM IR, but are not yet fully supported in
224 on all targets. However, it is generally safe to return up to 2 values from
225 a function: most targets should be able to handle at least that. MRV
226 support is a critical requirement for X86-64 ABI support, as X86-64 requires
227 the ability to return multiple registers from functions, and we use MRVs to
228 accomplish this in a direct way.</p></li>
Chris Lattnerbc5786b2008-06-05 06:57:39 +0000[diff] [blame]229
Chris Lattnerea34f642008-06-08 21:34:41 +0000[diff] [blame]230<li><p>LLVM 2.3 includes a complete reimplementation of the "<tt>llvmc</tt>"
231 tool. It is designed to overcome several problems with the original
232 <tt>llvmc</tt> and to provide a superset of the features of the
233 '<tt>gcc</tt>' driver.</p>
Chris Lattnerbc5786b2008-06-05 06:57:39 +0000[diff] [blame]234
Chris Lattnerea34f642008-06-08 21:34:41 +0000[diff] [blame]235<p>The main features of <tt>llvmc2</tt> are:
236 <ul>
237 <li>Extended handling of command line options and smart rules for
238 dispatching them to different tools.</li>
239 <li>Flexible (and extensible) rules for defining different tools.</li>
240 <li>The different intermediate steps performed by tools are represented
241 as edges in the abstract graph.</li>
242 <li>The 'language' for driver behavior definition is tablegen and thus
243 it's relatively easy to add new features.</li>
244 <li>The definition of driver is transformed into set of C++ classes, thus
245 no runtime interpretation is needed.</li>
246 </ul></p>
247 </li>
Chris Lattnerbc5786b2008-06-05 06:57:39 +0000[diff] [blame]248
Chris Lattnerea34f642008-06-08 21:34:41 +0000[diff] [blame]249<li><p>LLVM 2.3 includes a completely rewritten interface for <a
250 href="LinkTimeOptimization.html">Link Time Optimization</a>. This interface
251 is written in C, which allows for easier integration with C code bases, and
252 incorporates improvements we learned about from the first incarnation of the
253 interface.</p></li>
Chris Lattnerbc5786b2008-06-05 06:57:39 +0000[diff] [blame]254
Chris Lattnerea34f642008-06-08 21:34:41 +0000[diff] [blame]255<li><p>The <a href="tutorial/LangImpl1.html">Kaleidoscope tutorial</a> now
256 includes a "port" of the tutorial that <a
257 href="tutorial/OCamlLangImpl1.html">uses the Ocaml bindings</a> to implement
258 the Kaleidoscope language.</p></li>
Chris Lattnerbc5786b2008-06-05 06:57:39 +0000[diff] [blame]259
Chris Lattner8170c102008-02-10 08:18:42 +0000[diff] [blame]260</ul>
261
Chris Lattnerf3e5bc62007-05-14 06:56:09 +0000[diff] [blame]262</div>
263
Chris Lattner791f77b2008-06-05 06:25:56 +0000[diff] [blame]264
265<!--=========================================================================-->
266<div class="doc_subsection">
Chris Lattnerea34f642008-06-08 21:34:41 +0000[diff] [blame]267<a name="llvm-gcc">llvm-gcc 4.2 Improvements</a>
Chris Lattner791f77b2008-06-05 06:25:56 +0000[diff] [blame]268</div>
269
270<div class="doc_text">
271
Chris Lattnerf8e0b4e2008-06-08 22:59:35 +0000[diff] [blame]272<p>LLVM 2.3 fully supports the llvm-gcc 4.2 front-end, and includes support
273for the C, C++, Objective-C, Ada, and Fortran front-ends.</p>
Chris Lattner791f77b2008-06-05 06:25:56 +0000[diff] [blame]274
Chris Lattnerf8e0b4e2008-06-08 22:59:35 +0000[diff] [blame]275<p>
276<ul>
277<li>llvm-gcc 4.2 includes numerous fixes to better support the Objective-C
278front-end. Objective-C now works very well on Mac OS/X.</li>
Chris Lattner791f77b2008-06-05 06:25:56 +0000[diff] [blame]279
Chris Lattnerf8e0b4e2008-06-08 22:59:35 +0000[diff] [blame]280<li>Fortran EQUIVALENCEs are now supported by the gfortran front-end.</li>
Chris Lattner8348b472008-06-08 21:58:17 +0000[diff] [blame]281
Chris Lattnerf8e0b4e2008-06-08 22:59:35 +0000[diff] [blame]282<li>llvm-gcc 4.2 includes many other fixes which improve conformance with the
283relevant parts of the GCC testsuite.</li>
284
285</ul></p>
Chris Lattner791f77b2008-06-05 06:25:56 +0000[diff] [blame]286
Chris Lattner791f77b2008-06-05 06:25:56 +0000[diff] [blame]287</div>
288
289
Chris Lattner84977642007-09-21 03:54:09 +0000[diff] [blame]290<!--=========================================================================-->
291<div class="doc_subsection">
Chris Lattnerf304ffc2008-02-10 08:17:19 +0000[diff] [blame]292<a name="coreimprovements">LLVM Core Improvements</a>
293</div>
294
295<div class="doc_text">
296<p>New features include:
297</p>
298
Chris Lattner791f77b2008-06-05 06:25:56 +0000[diff] [blame]299<ul>
Chris Lattnerf8e0b4e2008-06-08 22:59:35 +0000[diff] [blame]300<li>LLVM IR now directly represents "common" linkage, instead of representing it
301as a form of weak linkage.</li>
302
303<li>LLVM IR now has support for atomic operations, and this functionality can
304be accessed through the llvm-gcc "__sync_synchronize",
305"__sync_val_compare_and_swap", and related builtins. Support for atomics are
306available in the Alpha, X86, X86-64, and PowerPC backends.</li>
307
308<li>The C and Ocaml bindings have extended to cover pass managers, several
309transformation passes, iteration over the LLVM IR, target data, and parameter
310attribute lists.</li>
Chris Lattnerf304ffc2008-02-10 08:17:19 +0000[diff] [blame]311</ul>
312
313</div>
314
315<!--=========================================================================-->
316<div class="doc_subsection">
Chris Lattneracce85d2008-02-10 07:46:44 +0000[diff] [blame]317<a name="optimizer">Optimizer Improvements</a>
318</div>
319
320<div class="doc_text">
321
Chris Lattnerf304ffc2008-02-10 08:17:19 +0000[diff] [blame]322<p>In addition to a huge array of bug fixes and minor performance tweaks, the
Chris Lattner2b659ef2008-02-12 06:29:45 +0000[diff] [blame]323LLVM 2.3 optimizers support a few major enhancements:</p>
Chris Lattneracce85d2008-02-10 07:46:44 +0000[diff] [blame]324
325<ul>
326
Chris Lattnerf8e0b4e2008-06-08 22:59:35 +0000[diff] [blame]327<li><p>Loop index set splitting on by default.
328This transformation hoists conditions from loop bodies and reduces a loop's
329iteration space to improve performance. For example,</p>
330
Devang Patel83a15d72008-06-05 21:44:00 +0000[diff] [blame]331<pre>
Chris Lattnerc4e6b372008-06-08 20:25:30 +0000[diff] [blame]332for (i = LB; i &lt; UB; ++i)
333 if (i &lt;= NV)
Devang Patel83a15d72008-06-05 21:44:00 +0000[diff] [blame]334 LOOP_BODY
335</pre>
Chris Lattnerf8e0b4e2008-06-08 22:59:35 +0000[diff] [blame]336
337<p>is transformed into:</p>
338
Devang Patel83a15d72008-06-05 21:44:00 +0000[diff] [blame]339<pre>
340NUB = min(NV+1, UB)
Chris Lattnerc4e6b372008-06-08 20:25:30 +0000[diff] [blame]341for (i = LB; i &lt; NUB; ++i)
Devang Patel83a15d72008-06-05 21:44:00 +0000[diff] [blame]342 LOOP_BODY
343</pre>
344</li>
Chris Lattner0b832202008-06-08 02:45:07 +0000[diff] [blame]345
Chris Lattnerf8e0b4e2008-06-08 22:59:35 +0000[diff] [blame]346<li>LLVM now includes a new <tt>memcpy</tt> optimization pass which removes
Duncan Sands27aff872008-06-08 20:18:35 +0000[diff] [blame]347dead <tt>memcpy</tt> calls, unneeded copies of aggregates, and performs
Chris Lattner0b832202008-06-08 02:45:07 +0000[diff] [blame]348return slot optimization. The LLVM optimizer now notices long sequences of
Duncan Sands27aff872008-06-08 20:18:35 +0000[diff] [blame]349consecutive stores and merges them into <tt>memcpy</tt>'s where profitable.</li>
Chris Lattner0b832202008-06-08 02:45:07 +0000[diff] [blame]350
Gabor Greif96a89c72008-06-05 18:39:01 +0000[diff] [blame]351<li>Alignment detection for vector memory references and for <tt>memcpy</tt> and
352<tt>memset</tt> is now more aggressive.</li>
Owen Anderson30073ff2008-06-06 16:23:15 +0000[diff] [blame]353
Chris Lattnerf8e0b4e2008-06-08 22:59:35 +0000[diff] [blame]354<li>The Aggressive Dead Code Elimination (ADCE) optimization has been rewritten
Chris Lattner0b832202008-06-08 02:45:07 +0000[diff] [blame]355to make it both faster and safer in the presence of code containing infinite
356loops. Some of its prior functionality has been factored out into the loop
Chris Lattnerf8e0b4e2008-06-08 22:59:35 +0000[diff] [blame]357deletion pass, which <em>is</em> safe for infinite loops. The new ADCE pass is
358no longer based on control dependence, making it run faster.</li>
Chris Lattner0b832202008-06-08 02:45:07 +0000[diff] [blame]359
Chris Lattner0b832202008-06-08 02:45:07 +0000[diff] [blame]360<li>The 'SimplifyLibCalls' pass, which optimizes calls to libc and libm
361 functions for C-based languages, has been rewritten to be a FunctionPass
362 instead a ModulePass. This allows it to be run more often and to be
363 included at -O1 in llvm-gcc. It was also extended to include more
Duncan Sands27aff872008-06-08 20:18:35 +0000[diff] [blame]364 optimizations and several corner case bugs were fixed.</li>
Chris Lattner0b832202008-06-08 02:45:07 +0000[diff] [blame]365
Duncan Sands47eff2b2008-06-08 19:38:43 +0000[diff] [blame]366<li>LLVM now includes a simple 'Jump Threading' pass, which attempts to simplify
Duncan Sands27aff872008-06-08 20:18:35 +0000[diff] [blame]367 conditional branches using information about predecessor blocks, simplifying
Chris Lattnere6e1b352008-06-08 21:19:07 +0000[diff] [blame]368 the control flow graph. This pass is pretty basic at this point, but
369 catches some important cases and provides a foundation to build on.</li>
370
371<li>Several corner case bugs which could lead to deleting volatile memory
372 accesses have been fixed.</li>
Chris Lattnerf8e0b4e2008-06-08 22:59:35 +0000[diff] [blame]373
374<li>Several optimizations have been sped up, leading to faster code generation
375 with the same code quality.</li>
376
Chris Lattner0b832202008-06-08 02:45:07 +0000[diff] [blame]377</ul>
378
379</div>
380
381<!--=========================================================================-->
382<div class="doc_subsection">
383<a name="codegen">Code Generator Improvements</a>
384</div>
385
386<div class="doc_text">
387
388<p>We put a significant amount of work into the code generator infrastructure,
389which allows us to implement more aggressive algorithms and make it run
390faster:</p>
391
392<ul>
Dan Gohmanfd0d22e2008-06-08 22:49:27 +0000[diff] [blame]393<li>The code generator now has support for carrying information about memory
394 references throughout the entire code generation process, via the
395 <a href="http://llvm.org/doxygen/classllvm_1_1MachineMemOperand.html">
396 MachineMemOperand</a> class. In the future this will be used to improve
397 both pre-pass and post-pass scheduling, and to improve compiler-debugging
398 output.</li>
Chris Lattner0b832202008-06-08 02:45:07 +0000[diff] [blame]399
Dan Gohmane3280782008-06-08 22:57:59 +0000[diff] [blame]400<li>The target-independent code generator infrastructure now uses LLVM's
401 <a href="http://llvm.org/doxygen/classllvm_1_1APInt.html">APInt</a>
Chris Lattner0b832202008-06-08 02:45:07 +0000[diff] [blame]402 class to handle integer values, which allows it to support integer types
Chris Lattner43b65e92008-06-08 23:06:47 +0000[diff] [blame^]403 larger than 64 bits (for example i128). Note that support for such types is
404 also dependent on target-specific support. Use of APInt is also a step
405 toward support for non-power-of-2 integer sizes.</li>
Chris Lattner0b832202008-06-08 02:45:07 +0000[diff] [blame]406
Chris Lattnere6e1b352008-06-08 21:19:07 +0000[diff] [blame]407<li>LLVM 2.3 includes several compile time speedups for code with large basic
Dan Gohmaneedf6fe2008-06-08 22:44:43 +0000[diff] [blame]408 blocks, particularly in the instruction selection phase, register
409 allocation, scheduling, and tail merging/jump threading.</li>
Chris Lattner0b832202008-06-08 02:45:07 +0000[diff] [blame]410
Chris Lattner43b65e92008-06-08 23:06:47 +0000[diff] [blame^]411<li>LLVM 2.3 includes several improvements which make llc's
412 <tt>--view-sunit-dags</tt> visualization of scheduling dependency graphs
413 easier to understand.</li>
Chris Lattner0b832202008-06-08 02:45:07 +0000[diff] [blame]414
415<li>The code generator allows targets to write patterns that generate subreg
416 references directly in .td files now.</li>
417
418<li><tt>memcpy</tt> lowering in the backend is more aggressive, particularly for
419 <tt>memcpy</tt> calls introduced by the code generator when handling
420 pass-by-value structure argument copies.</li>
Chris Lattnere6e1b352008-06-08 21:19:07 +0000[diff] [blame]421
422<li>Inline assembly with multiple register results now returns those results
423 directly in the appropriate registers, rather than going through memory.
424 Inline assembly that uses constraints like "ir" with immediates now use the
425 'i' form when possible instead of always loading the value in a register.
426 This saves an instruction and reduces register use.</li>
427
428<li>Added support for PIC/GOT style tail calls on x86/32 and initial support
429 for tail calls on PowerPC 32 (it may also work on ppc64 but not
430 thoroughly tested).</li>
Chris Lattner84977642007-09-21 03:54:09 +0000[diff] [blame]431</ul>
432
433</div>
434
435
436<!--=========================================================================-->
437<div class="doc_subsection">
Chris Lattner77d29b12008-06-05 08:02:49 +0000[diff] [blame]438<a name="x86specific">X86/X86-64 Specific Improvements</a>
Chris Lattner84977642007-09-21 03:54:09 +0000[diff] [blame]439</div>
Chris Lattner60893e52007-05-18 06:33:02 +0000[diff] [blame]440
Chris Lattner738bd302006-04-18 06:32:08 +0000[diff] [blame]441<div class="doc_text">
Chris Lattnerf304ffc2008-02-10 08:17:19 +0000[diff] [blame]442<p>New target-specific features include:
Chris Lattnerf3e5bc62007-05-14 06:56:09 +0000[diff] [blame]443</p>
Chris Lattner26299222006-11-18 07:51:14 +0000[diff] [blame]444
445<ul>
Chris Lattner791f77b2008-06-05 06:25:56 +0000[diff] [blame]446<li>llvm-gcc's X86-64 ABI conformance is far improved, particularly in the
Gabor Greif96a89c72008-06-05 18:39:01 +0000[diff] [blame]447 area of passing and returning structures by value. llvm-gcc compiled code
Chris Lattner791f77b2008-06-05 06:25:56 +0000[diff] [blame]448 now interoperates very well on X86-64 systems with other compilers.</li>
449
Duncan Sands47eff2b2008-06-08 19:38:43 +0000[diff] [blame]450<li>Support for Win64 was added. This includes code generation itself, JIT
Chris Lattner43b65e92008-06-08 23:06:47 +0000[diff] [blame^]451 support, and necessary changes to llvm-gcc.</li>
Anton Korobeynikova6094be2008-06-08 10:24:13 +0000[diff] [blame]452
Chris Lattner791f77b2008-06-05 06:25:56 +0000[diff] [blame]453<li>The LLVM X86 backend now supports the support SSE 4.1 instruction set, and
Chris Lattner0b832202008-06-08 02:45:07 +0000[diff] [blame]454 the llvm-gcc 4.2 front-end supports the SSE 4.1 compiler builtins. Various
455 generic vector operations (insert/extract/shuffle) are much more efficient
456 when SSE 4.1 is enabled. The JIT automatically takes advantage of these
457 instructions, but llvm-gcc must be explicitly told to use them, e.g. with
458 <tt>-march=penryn</tt>.</li>
Anton Korobeynikova6094be2008-06-08 10:24:13 +0000[diff] [blame]459
Chris Lattner791f77b2008-06-05 06:25:56 +0000[diff] [blame]460<li>The X86 backend now does a number of optimizations that aim to avoid
461 converting numbers back and forth from SSE registers to the X87 floating
Chris Lattner43b65e92008-06-08 23:06:47 +0000[diff] [blame^]462 point stack. This is important because most X86 ABIs require return values
463 to be on the X87 Floating Point stack, but most CPUs prefer computation in
464 the SSE units.</li>
Anton Korobeynikova6094be2008-06-08 10:24:13 +0000[diff] [blame]465
Chris Lattnerbc5786b2008-06-05 06:57:39 +0000[diff] [blame]466<li>The X86 backend supports stack realignment, which is particularly useful for
Chris Lattner43b65e92008-06-08 23:06:47 +0000[diff] [blame^]467 vector code on OS's without 16-byte aligned stacks, such as Linux and
468 Windows.</li>
Chris Lattner791f77b2008-06-05 06:25:56 +0000[diff] [blame]469
470<li>The X86 backend now supports the "sseregparm" options in GCC, which allow
471 functions to be tagged as passing floating point values in SSE
472 registers.</li>
473
474<li>Trampolines (taking the address of a nested function) now work on
Duncan Sands27aff872008-06-08 20:18:35 +0000[diff] [blame]475 Linux/X86-64.</li>
Anton Korobeynikova6094be2008-06-08 10:24:13 +0000[diff] [blame]476
Chris Lattner77d29b12008-06-05 08:02:49 +0000[diff] [blame]477<li><tt>__builtin_prefetch</tt> is now compiled into the appropriate prefetch
478 instructions instead of being ignored.</li>
Chris Lattner791f77b2008-06-05 06:25:56 +0000[diff] [blame]479
Chris Lattner43b65e92008-06-08 23:06:47 +0000[diff] [blame^]480<li>128-bit integers are now supported on X86-64 targets. This can be used
481 through <tt>__attribute__((TImode))</tt> in llvm-gcc.</li>
Dan Gohman4e16d402008-06-05 16:15:39 +0000[diff] [blame]482
Chris Lattner43b65e92008-06-08 23:06:47 +0000[diff] [blame^]483<li>The register allocator can now rematerialize PIC-base computations, which is
484 an important optimization for register use.</li>
Chris Lattner0b832202008-06-08 02:45:07 +0000[diff] [blame]485
Chris Lattnere6e1b352008-06-08 21:19:07 +0000[diff] [blame]486<li>The "t" and "f" inline assembly constraints for the X87 floating point stack
487 now work. However, the "u" constraint is still not fully supported.</li>
488
Chris Lattnerbfb17ab2007-05-17 21:41:31 +0000[diff] [blame]489</ul>
Chris Lattner84977642007-09-21 03:54:09 +0000[diff] [blame]490
Chris Lattnerbfb17ab2007-05-17 21:41:31 +0000[diff] [blame]491</div>
Chris Lattnerc5d658a2006-03-03 00:34:26 +0000[diff] [blame]492
Chris Lattner84977642007-09-21 03:54:09 +0000[diff] [blame]493<!--=========================================================================-->
494<div class="doc_subsection">
Chris Lattner77d29b12008-06-05 08:02:49 +0000[diff] [blame]495<a name="targetspecific">Other Target Specific Improvements</a>
496</div>
497
498<div class="doc_text">
499<p>New target-specific features include:
500</p>
501
502<ul>
Gabor Greif96a89c72008-06-05 18:39:01 +0000[diff] [blame]503<li>The LLVM C backend now supports vector code.</li>
Chris Lattner43b65e92008-06-08 23:06:47 +0000[diff] [blame^]504<li>The Cell SPU backend includes a number of improvements. It generates better
505 code and its stability/completeness is improving.</li>
Chris Lattner77d29b12008-06-05 08:02:49 +0000[diff] [blame]506</ul>
507
508</div>
509
510
511
512<!--=========================================================================-->
513<div class="doc_subsection">
Chris Lattner84977642007-09-21 03:54:09 +0000[diff] [blame]514<a name="otherimprovements">Other Improvements</a>
515</div>
516
517<div class="doc_text">
518<p>New features include:
519</p>
520
521<ul>
Chris Lattnerbc5786b2008-06-05 06:57:39 +0000[diff] [blame]522<li>LLVM now builds with GCC 4.3.</li>
Chris Lattnere6e1b352008-06-08 21:19:07 +0000[diff] [blame]523<li>Bugpoint now supports running custom scripts (with the <tt>-run-custom</tt>
524 option) to determine how to execute the command and whether it is making
525 forward process.</li>
Chris Lattner84977642007-09-21 03:54:09 +0000[diff] [blame]526</ul>
527
528</div>
Chris Lattnerc5d658a2006-03-03 00:34:26 +0000[diff] [blame]529
Chris Lattner19092612003-10-02 16:38:05 +0000[diff] [blame]530<!-- *********************************************************************** -->
Misha Brukman500bc302003-11-22 00:38:41 +0000[diff] [blame]531<div class="doc_section">
532 <a name="portability">Portability and Supported Platforms</a>
533</div>
Chris Lattner19092612003-10-02 16:38:05 +0000[diff] [blame]534<!-- *********************************************************************** -->
535
Misha Brukman500bc302003-11-22 00:38:41 +0000[diff] [blame]536<div class="doc_text">
537
John Criswell0b5b5e92004-12-08 20:35:47 +0000[diff] [blame]538<p>LLVM is known to work on the following platforms:</p>
Chris Lattner4654bdb2004-06-01 18:22:41 +0000[diff] [blame]539
540<ul>
Gabor Greif96a89c72008-06-05 18:39:01 +0000[diff] [blame]541<li>Intel and AMD machines (IA32) running Red Hat Linux, Fedora Core and FreeBSD
Reid Spencer00812e22005-05-17 02:47:27 +0000[diff] [blame]542 (and probably other unix-like systems).</li>
Chris Lattner000c73b2008-02-06 06:30:34 +0000[diff] [blame]543<li>PowerPC and X86-based Mac OS X systems, running 10.3 and above in 32-bit and
Chris Lattner26299222006-11-18 07:51:14 +0000[diff] [blame]544 64-bit modes.</li>
Chris Lattner252b83d2008-02-06 18:00:06 +0000[diff] [blame]545<li>Intel and AMD machines running on Win32 using MinGW libraries (native).</li>
Chris Lattner57a460e2007-05-23 04:39:32 +0000[diff] [blame]546<li>Intel and AMD machines running on Win32 with the Cygwin libraries (limited
547 support is available for native builds with Visual C++).</li>
Gabor Greif96a89c72008-06-05 18:39:01 +0000[diff] [blame]548<li>Sun UltraSPARC workstations running Solaris 10.</li>
John Criswell9321fa82005-05-13 20:28:15 +0000[diff] [blame]549<li>Alpha-based machines running Debian GNU/Linux.</li>
Gabor Greif96a89c72008-06-05 18:39:01 +0000[diff] [blame]550<li>Itanium-based (IA64) machines running Linux and HP-UX.</li>
Chris Lattner4654bdb2004-06-01 18:22:41 +0000[diff] [blame]551</ul>
552
Chris Lattnerbc5786b2008-06-05 06:57:39 +0000[diff] [blame]553<p>The core LLVM infrastructure uses GNU autoconf to adapt itself
Brian Gaekeb0fd7612004-05-09 05:28:35 +0000[diff] [blame]554to the machine and operating system on which it is built. However, minor
555porting may be required to get LLVM to work on new platforms. We welcome your
556portability patches and reports of successful builds or error messages.</p>
Chris Lattner79c3fe12003-10-02 04:57:28 +0000[diff] [blame]557
Misha Brukman500bc302003-11-22 00:38:41 +0000[diff] [blame]558</div>
Chris Lattner79c3fe12003-10-02 04:57:28 +0000[diff] [blame]559
560<!-- *********************************************************************** -->
Misha Brukman500bc302003-11-22 00:38:41 +0000[diff] [blame]561<div class="doc_section">
562 <a name="knownproblems">Known Problems</a>
563</div>
Chris Lattner79c3fe12003-10-02 04:57:28 +0000[diff] [blame]564<!-- *********************************************************************** -->
565
Misha Brukman500bc302003-11-22 00:38:41 +0000[diff] [blame]566<div class="doc_text">
567
568<p>This section contains all known problems with the LLVM system, listed by
Chris Lattner79c3fe12003-10-02 04:57:28 +0000[diff] [blame]569component. As new problems are discovered, they will be added to these
Chris Lattner5eccca42003-12-12 21:22:16 +0000[diff] [blame]570sections. If you run into a problem, please check the <a
Chris Lattnerc463b272005-10-29 07:07:09 +0000[diff] [blame]571href="http://llvm.org/bugs/">LLVM bug database</a> and submit a bug if
Chris Lattner5eccca42003-12-12 21:22:16 +0000[diff] [blame]572there isn't already one.</p>
Chris Lattner79c3fe12003-10-02 04:57:28 +0000[diff] [blame]573
Misha Brukman500bc302003-11-22 00:38:41 +0000[diff] [blame]574</div>
Chris Lattner79c3fe12003-10-02 04:57:28 +0000[diff] [blame]575
Chris Lattnerf5ee1702004-03-14 02:03:02 +0000[diff] [blame]576<!-- ======================================================================= -->
577<div class="doc_subsection">
578 <a name="experimental">Experimental features included with this release</a>
579</div>
Chris Lattner79c3fe12003-10-02 04:57:28 +0000[diff] [blame]580
Chris Lattnerf5ee1702004-03-14 02:03:02 +0000[diff] [blame]581<div class="doc_text">
582
Misha Brukman6df9e2c2004-05-12 21:46:05 +0000[diff] [blame]583<p>The following components of this LLVM release are either untested, known to
584be broken or unreliable, or are in early development. These components should
585not be relied on, and bugs should not be filed against them, but they may be
586useful to some people. In particular, if you would like to work on one of these
Chris Lattner2b659ef2008-02-12 06:29:45 +0000[diff] [blame]587components, please contact us on the <a
588href="http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev">LLVMdev list</a>.</p>
Chris Lattnerf5ee1702004-03-14 02:03:02 +0000[diff] [blame]589
590<ul>
Chris Lattner2b659ef2008-02-12 06:29:45 +0000[diff] [blame]591<li>The MSIL, IA64, Alpha, SPU, and MIPS backends are experimental.</li>
Duncan Sands27aff872008-06-08 20:18:35 +0000[diff] [blame]592<li>The llc "<tt>-filetype=asm</tt>" (the default) is the only supported
Chris Lattner000c73b2008-02-06 06:30:34 +0000[diff] [blame]593 value for this option.</li>
Chris Lattnerf5ee1702004-03-14 02:03:02 +0000[diff] [blame]594</ul>
595
596</div>
Chris Lattner79c3fe12003-10-02 04:57:28 +0000[diff] [blame]597
Misha Brukman500bc302003-11-22 00:38:41 +0000[diff] [blame]598<!-- ======================================================================= -->
599<div class="doc_subsection">
Chris Lattner26299222006-11-18 07:51:14 +0000[diff] [blame]600 <a name="x86-be">Known problems with the X86 back-end</a>
John Criswellc0c186d2005-11-08 21:11:33 +0000[diff] [blame]601</div>
602
603<div class="doc_text">
604
605<ul>
Anton Korobeynikova6094be2008-06-08 10:24:13 +0000[diff] [blame]606 <li>The X86 backend does not yet support
607 all <a href="http://llvm.org/PR879">inline assembly that uses the X86
608 floating point stack</a>. It supports the 'f' and 't' constraints, but not
609 'u'.</li>
Chris Lattnere6e1b352008-06-08 21:19:07 +0000[diff] [blame]610 <li>The X86 backend generates inefficient floating point code when configured
611 to generate code for systems that don't have SSE2.</li>
Duncan Sands47eff2b2008-06-08 19:38:43 +0000[diff] [blame]612 <li>Win64 code generation wasn't widely tested. Everything should work, but we
Anton Korobeynikova6094be2008-06-08 10:24:13 +0000[diff] [blame]613 expect small issues to happen. Also, llvm-gcc cannot build mingw64 runtime
614 currently due
615 to <a href="http://llvm.org/PR2255">several</a>
616 <a href="http://llvm.org/PR2257">bugs</a> in FP stackifier
Dan Gohman8207ba92008-06-08 23:05:11 +0000[diff] [blame]617 <li>The X86-64 backend does not yet support position-independent code (PIC)
618 generation on Linux targets.</li>
619 <li>The X86-64 backend does not yet support the LLVM IR instruction
620 <tt>va_arg</tt>. Currently, the llvm-gcc front-end supports variadic
621 argument constructs on X86-64 by lowering them manually.</li>
Chris Lattner26299222006-11-18 07:51:14 +0000[diff] [blame]622</ul>
623
624</div>
625
626<!-- ======================================================================= -->
627<div class="doc_subsection">
628 <a name="ppc-be">Known problems with the PowerPC back-end</a>
629</div>
630
631<div class="doc_text">
632
633<ul>
Nicolas Geoffraye4285dc2007-05-15 09:21:28 +0000[diff] [blame]634<li>The Linux PPC32/ABI support needs testing for the interpreter and static
Chris Lattner57a460e2007-05-23 04:39:32 +0000[diff] [blame]635compilation, and lacks support for debug information.</li>
Chris Lattner26299222006-11-18 07:51:14 +0000[diff] [blame]636</ul>
637
638</div>
639
640<!-- ======================================================================= -->
641<div class="doc_subsection">
Chris Lattnerf3e5bc62007-05-14 06:56:09 +0000[diff] [blame]642 <a name="arm-be">Known problems with the ARM back-end</a>
643</div>
644
645<div class="doc_text">
646
647<ul>
Chris Lattner57a460e2007-05-23 04:39:32 +0000[diff] [blame]648<li>Thumb mode works only on ARMv6 or higher processors. On sub-ARMv6
Duncan Sandsc90d68b2007-09-26 15:59:54 +0000[diff] [blame]649processors, thumb programs can crash or produce wrong
Chris Lattner57a460e2007-05-23 04:39:32 +0000[diff] [blame]650results (<a href="http://llvm.org/PR1388">PR1388</a>).</li>
Chris Lattnerf3e5bc62007-05-14 06:56:09 +0000[diff] [blame]651<li>Compilation for ARM Linux OABI (old ABI) is supported, but not fully tested.
652</li>
Chris Lattnere6e1b352008-06-08 21:19:07 +0000[diff] [blame]653<li>There is a bug in QEMU-ARM (&lt;= 0.9.0) which causes it to incorrectly
654 execute
Chris Lattner57a460e2007-05-23 04:39:32 +0000[diff] [blame]655programs compiled with LLVM. Please use more recent versions of QEMU.</li>
Chris Lattnerf3e5bc62007-05-14 06:56:09 +0000[diff] [blame]656</ul>
657
658</div>
659
660<!-- ======================================================================= -->
661<div class="doc_subsection">
Chris Lattner26299222006-11-18 07:51:14 +0000[diff] [blame]662 <a name="sparc-be">Known problems with the SPARC back-end</a>
663</div>
664
665<div class="doc_text">
666
667<ul>
668<li>The SPARC backend only supports the 32-bit SPARC ABI (-m32), it does not
669 support the 64-bit SPARC ABI (-m64).</li>
670</ul>
671
672</div>
673
674<!-- ======================================================================= -->
675<div class="doc_subsection">
Chris Lattner26299222006-11-18 07:51:14 +0000[diff] [blame]676 <a name="alpha-be">Known problems with the Alpha back-end</a>
677</div>
678
679<div class="doc_text">
680
681<ul>
682
683<li>On 21164s, some rare FP arithmetic sequences which may trap do not have the
684appropriate nops inserted to ensure restartability.</li>
685
John Criswellc0c186d2005-11-08 21:11:33 +0000[diff] [blame]686</ul>
687</div>
688
Chris Lattner26299222006-11-18 07:51:14 +0000[diff] [blame]689<!-- ======================================================================= -->
690<div class="doc_subsection">
691 <a name="ia64-be">Known problems with the IA64 back-end</a>
692</div>
693
694<div class="doc_text">
695
696<ul>
697
698<li>C++ programs are likely to fail on IA64, as calls to <tt>setjmp</tt> are
699made where the argument is not 16-byte aligned, as required on IA64. (Strictly
700speaking this is not a bug in the IA64 back-end; it will also be encountered
701when building C++ programs using the C back-end.)</li>
702
703<li>The C++ front-end does not use <a href="http://llvm.org/PR406">IA64
704ABI compliant layout of v-tables</a>. In particular, it just stores function
705pointers instead of function descriptors in the vtable. This bug prevents
706mixing C++ code compiled with LLVM with C++ objects compiled by other C++
707compilers.</li>
708
709<li>There are a few ABI violations which will lead to problems when mixing LLVM
710output with code built with other compilers, particularly for floating-point
711programs.</li>
712
Duncan Sands47eff2b2008-06-08 19:38:43 +0000[diff] [blame]713<li>Defining vararg functions is not supported (but calling them is OK).</li>
Chris Lattner26299222006-11-18 07:51:14 +0000[diff] [blame]714
Chris Lattnercaf06342007-05-09 04:58:11 +0000[diff] [blame]715<li>The Itanium backend has bitrotted somewhat.</li>
Chris Lattner26299222006-11-18 07:51:14 +0000[diff] [blame]716</ul>
717
718</div>
719
720<!-- ======================================================================= -->
721<div class="doc_subsection">
Chris Lattnerf3e5bc62007-05-14 06:56:09 +0000[diff] [blame]722 <a name="c-be">Known problems with the C back-end</a>
Chris Lattner26299222006-11-18 07:51:14 +0000[diff] [blame]723</div>
724
725<div class="doc_text">
726
727<ul>
Chris Lattner5733b272008-06-05 06:35:40 +0000[diff] [blame]728<li><a href="http://llvm.org/PR802">The C backend has only basic support for
729 inline assembly code</a>.</li>
Chris Lattner725a0d82007-09-26 06:01:35 +0000[diff] [blame]730<li><a href="http://llvm.org/PR1658">The C backend violates the ABI of common
731 C++ programs</a>, preventing intermixing between C++ compiled by the CBE and
Duncan Sands27aff872008-06-08 20:18:35 +0000[diff] [blame]732 C++ code compiled with llc or native compilers.</li>
Duncan Sandsf74c0cc2008-02-10 13:40:55 +0000[diff] [blame]733<li>The C backend does not support all exception handling constructs.</li>
Chris Lattner26299222006-11-18 07:51:14 +0000[diff] [blame]734</ul>
735
736</div>
John Criswellc0c186d2005-11-08 21:11:33 +0000[diff] [blame]737
Misha Brukman500bc302003-11-22 00:38:41 +0000[diff] [blame]738
739<!-- ======================================================================= -->
740<div class="doc_subsection">
Chris Lattner5733b272008-06-05 06:35:40 +0000[diff] [blame]741 <a name="c-fe">Known problems with the llvm-gcc C front-end</a>
Misha Brukman500bc302003-11-22 00:38:41 +0000[diff] [blame]742</div>
Chris Lattner47588f92003-10-02 05:07:23 +0000[diff] [blame]743
Chris Lattner79c3fe12003-10-02 04:57:28 +0000[diff] [blame]744<!-- _______________________________________________________________________ -->
Chris Lattnerb40efb32003-12-14 05:03:43 +0000[diff] [blame]745<div class="doc_subsubsection">Bugs</div>
Chris Lattner79c3fe12003-10-02 04:57:28 +0000[diff] [blame]746
Misha Brukman500bc302003-11-22 00:38:41 +0000[diff] [blame]747<div class="doc_text">
Chris Lattnerc5d658a2006-03-03 00:34:26 +0000[diff] [blame]748
Chris Lattner252b83d2008-02-06 18:00:06 +0000[diff] [blame]749<p>llvm-gcc does not currently support <a href="http://llvm.org/PR869">Link-Time
750Optimization</a> on most platforms "out-of-the-box". Please inquire on the
Chris Lattnerf3e5bc62007-05-14 06:56:09 +0000[diff] [blame]751llvmdev mailing list if you are interested.</p>
Chris Lattnerd9ea0172006-08-08 17:27:28 +0000[diff] [blame]752
Chris Lattner5733b272008-06-05 06:35:40 +0000[diff] [blame]753<p>The only major language feature of GCC not supported by llvm-gcc is
754 the <tt>__builtin_apply</tt> family of builtins. However, some extensions
755 are only supported on some targets. For example, trampolines are only
Duncan Sands27aff872008-06-08 20:18:35 +0000[diff] [blame]756 supported on some targets (these are used when you take the address of a
757 nested function).</p>
Misha Brukman500bc302003-11-22 00:38:41 +0000[diff] [blame]758
Chris Lattner5733b272008-06-05 06:35:40 +0000[diff] [blame]759<p>If you run into GCC extensions which are not supported, please let us know.
760</p>
Chris Lattner79c3fe12003-10-02 04:57:28 +0000[diff] [blame]761
Misha Brukman500bc302003-11-22 00:38:41 +0000[diff] [blame]762</div>
Chris Lattner79c3fe12003-10-02 04:57:28 +0000[diff] [blame]763
Misha Brukman500bc302003-11-22 00:38:41 +0000[diff] [blame]764<!-- ======================================================================= -->
765<div class="doc_subsection">
Chris Lattner5733b272008-06-05 06:35:40 +0000[diff] [blame]766 <a name="c++-fe">Known problems with the llvm-gcc C++ front-end</a>
Misha Brukman500bc302003-11-22 00:38:41 +0000[diff] [blame]767</div>
Chris Lattner79c3fe12003-10-02 04:57:28 +0000[diff] [blame]768
Misha Brukman500bc302003-11-22 00:38:41 +0000[diff] [blame]769<div class="doc_text">
770
Chris Lattnerf3e5bc62007-05-14 06:56:09 +0000[diff] [blame]771<p>The C++ front-end is considered to be fully
Chris Lattner7506b1d2004-12-07 08:04:13 +0000[diff] [blame]772tested and works for a number of non-trivial programs, including LLVM
Chris Lattnerf3e5bc62007-05-14 06:56:09 +0000[diff] [blame]773itself, Qt, Mozilla, etc.</p>
Misha Brukman500bc302003-11-22 00:38:41 +0000[diff] [blame]774
Misha Brukman500bc302003-11-22 00:38:41 +0000[diff] [blame]775<ul>
Chris Lattner5733b272008-06-05 06:35:40 +0000[diff] [blame]776<li>Exception handling works well on the X86 and PowerPC targets, including
Duncan Sands27aff872008-06-08 20:18:35 +0000[diff] [blame]777X86-64 darwin. This works when linking to a libstdc++ compiled by GCC. It is
778supported on X86-64 linux, but that is disabled by default in this release.</li>
Misha Brukman500bc302003-11-22 00:38:41 +0000[diff] [blame]779</ul>
Chris Lattnerfcc54b32003-10-07 22:14:37 +0000[diff] [blame]780
Misha Brukman500bc302003-11-22 00:38:41 +0000[diff] [blame]781</div>
782
Misha Brukman500bc302003-11-22 00:38:41 +0000[diff] [blame]783
Chris Lattner2b659ef2008-02-12 06:29:45 +0000[diff] [blame]784<!-- ======================================================================= -->
785<div class="doc_subsection">
Chris Lattner5733b272008-06-05 06:35:40 +0000[diff] [blame]786 <a name="ada-fe">Known problems with the llvm-gcc Ada front-end</a>
Chris Lattner2b659ef2008-02-12 06:29:45 +0000[diff] [blame]787</div>
788
789<div class="doc_text">
790The llvm-gcc 4.2 Ada compiler works fairly well, however this is not a mature
791technology and problems should be expected.
792<ul>
Duncan Sands27aff872008-06-08 20:18:35 +0000[diff] [blame]793<li>The Ada front-end currently only builds on X86-32. This is mainly due
Chris Lattner2b659ef2008-02-12 06:29:45 +0000[diff] [blame]794to lack of trampoline support (pointers to nested functions) on other platforms,
Duncan Sands27aff872008-06-08 20:18:35 +0000[diff] [blame]795however it <a href="http://llvm.org/PR2006">also fails to build on X86-64</a>
Chris Lattner2b659ef2008-02-12 06:29:45 +0000[diff] [blame]796which does support trampolines.</li>
797<li>The Ada front-end <a href="http://llvm.org/PR2007">fails to bootstrap</a>.
798Workaround: configure with --disable-bootstrap.</li>
799<li>The c380004 and <a href="http://llvm.org/PR2010">c393010</a> ACATS tests
Duncan Sands27aff872008-06-08 20:18:35 +0000[diff] [blame]800fail (c380004 also fails with gcc-4.2 mainline). When built at -O3, the
801<a href="http://llvm.org/PR2421">cxg2021</a> ACATS test also fails.</li>
802<li>Some gcc specific Ada tests continue to crash the compiler. The testsuite
803reports most tests as having failed even though they pass.</li>
Chris Lattner2b659ef2008-02-12 06:29:45 +0000[diff] [blame]804<li>The -E binder option (exception backtraces)
805<a href="http://llvm.org/PR1982">does not work</a> and will result in programs
806crashing if an exception is raised. Workaround: do not use -E.</li>
807<li>Only discrete types <a href="http://llvm.org/PR1981">are allowed to start
808or finish at a non-byte offset</a> in a record. Workaround: do not pack records
809or use representation clauses that result in a field of a non-discrete type
810starting or finishing in the middle of a byte.</li>
Chris Lattnere6e1b352008-06-08 21:19:07 +0000[diff] [blame]811<li>The <tt>lli</tt> interpreter <a href="http://llvm.org/PR2009">considers
812'main' as generated by the Ada binder to be invalid</a>.
813Workaround: hand edit the file to use pointers for <tt>argv</tt> and
814<tt>envp</tt> rather than integers.</li>
815<li>The <tt>-fstack-check</tt> option <a href="http://llvm.org/PR2008">is
816ignored</a>.</li>
Chris Lattner2b659ef2008-02-12 06:29:45 +0000[diff] [blame]817</ul>
818</div>
819
Chris Lattner79c3fe12003-10-02 04:57:28 +0000[diff] [blame]820<!-- *********************************************************************** -->
Misha Brukman500bc302003-11-22 00:38:41 +0000[diff] [blame]821<div class="doc_section">
822 <a name="additionalinfo">Additional Information</a>
823</div>
Chris Lattner79c3fe12003-10-02 04:57:28 +0000[diff] [blame]824<!-- *********************************************************************** -->
825
Misha Brukman500bc302003-11-22 00:38:41 +0000[diff] [blame]826<div class="doc_text">
827
Chris Lattner416db102005-05-16 17:13:10 +0000[diff] [blame]828<p>A wide variety of additional information is available on the <a
Chris Lattnerb4b0ce72007-05-18 00:44:29 +0000[diff] [blame]829href="http://llvm.org">LLVM web page</a>, in particular in the <a
830href="http://llvm.org/docs/">documentation</a> section. The web page also
Reid Spencer669ed452007-07-09 08:04:31 +0000[diff] [blame]831contains versions of the API documentation which is up-to-date with the
832Subversion version of the source code.
Misha Brukman109d9e82005-03-30 19:14:24 +0000[diff] [blame]833You can access versions of these documents specific to this release by going
834into the "<tt>llvm/doc/</tt>" directory in the LLVM tree.</p>
Chris Lattner79c3fe12003-10-02 04:57:28 +0000[diff] [blame]835
Misha Brukman500bc302003-11-22 00:38:41 +0000[diff] [blame]836<p>If you have any questions or comments about LLVM, please feel free to contact
Chris Lattnerc463b272005-10-29 07:07:09 +0000[diff] [blame]837us via the <a href="http://llvm.org/docs/#maillist"> mailing
Chris Lattner5eccca42003-12-12 21:22:16 +0000[diff] [blame]838lists</a>.</p>
Chris Lattner79c3fe12003-10-02 04:57:28 +0000[diff] [blame]839
Misha Brukman500bc302003-11-22 00:38:41 +0000[diff] [blame]840</div>
Chris Lattner79c3fe12003-10-02 04:57:28 +0000[diff] [blame]841
842<!-- *********************************************************************** -->
Chris Lattner79c3fe12003-10-02 04:57:28 +0000[diff] [blame]843
Misha Brukman500bc302003-11-22 00:38:41 +0000[diff] [blame]844<hr>
Misha Brukman2061e892003-11-22 01:23:39 +0000[diff] [blame]845<address>
Misha Brukman38847d52003-12-21 22:53:21 +0000[diff] [blame]846 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
Misha Brukman2061e892003-11-22 01:23:39 +0000[diff] [blame]847 src="http://jigsaw.w3.org/css-validator/images/vcss" alt="Valid CSS!"></a>
848 <a href="http://validator.w3.org/check/referer"><img
Chris Lattner8170c102008-02-10 08:18:42 +0000[diff] [blame]849 src="http://www.w3.org/Icons/valid-html401" alt="Valid HTML 4.01!"></a>
Misha Brukman2061e892003-11-22 01:23:39 +0000[diff] [blame]850
Chris Lattnerb4b0ce72007-05-18 00:44:29 +0000[diff] [blame]851 <a href="http://llvm.org/">LLVM Compiler Infrastructure</a><br>
Misha Brukman500bc302003-11-22 00:38:41 +0000[diff] [blame]852 Last modified: $Date$
Misha Brukman2061e892003-11-22 01:23:39 +0000[diff] [blame]853</address>
Chris Lattner79c3fe12003-10-02 04:57:28 +0000[diff] [blame]854
Misha Brukman500bc302003-11-22 00:38:41 +0000[diff] [blame]855</body>
856</html>