blob: 8f89d0c31ad2c55fe414915ea6c005af9c9e3a6b [file] [log] [blame]
Dan Gohmanf17a25c2007-07-18 16:29:46 +00001<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01//EN"
2 "http://www.w3.org/TR/html4/strict.dtd">
3<html>
4<head>
Bill Wendlingc0e81142009-04-07 18:40:56 +00005 <meta http-equiv="Content-Type" content="text/html; charset=utf-8">
Dan Gohmanf17a25c2007-07-18 16:29:46 +00006 <title>LLVM: Frequently Asked Questions</title>
7 <style type="text/css">
8 @import url("llvm.css");
9 .question { font-weight: bold }
10 .answer { margin-left: 2em }
11 </style>
12</head>
13<body>
14
15<div class="doc_title">
16 LLVM: Frequently Asked Questions
17</div>
18
19<ol>
20 <li><a href="#license">License</a>
21 <ol>
Bill Wendlingc0e81142009-04-07 18:40:56 +000022 <li>Why are the LLVM source code and the front-end distributed under
23 different licenses?</li>
24
25 <li>Does the University of Illinois Open Source License really qualify as an
26 "open source" license?</li>
27
28 <li>Can I modify LLVM source code and redistribute the modified source?</li>
29
30 <li>Can I modify LLVM source code and redistribute binaries or other tools
31 based on it, without redistributing the source?</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +000032 </ol></li>
33
34 <li><a href="#source">Source code</a>
35 <ol>
Bill Wendlingc0e81142009-04-07 18:40:56 +000036 <li>In what language is LLVM written?</li>
37
38 <li>How portable is the LLVM source code?</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +000039 </ol></li>
40
41 <li><a href="#build">Build Problems</a>
42 <ol>
Bill Wendlingc0e81142009-04-07 18:40:56 +000043 <li>When I run configure, it finds the wrong C compiler.</li>
44
45 <li>The <tt>configure</tt> script finds the right C compiler, but it uses
46 the LLVM linker from a previous build. What do I do?</li>
47
48 <li>When creating a dynamic library, I get a strange GLIBC error.</li>
49
50 <li>I've updated my source tree from Subversion, and now my build is trying
51 to use a file/directory that doesn't exist.</li>
52
53 <li>I've modified a Makefile in my source tree, but my build tree keeps
54 using the old version. What do I do?</li>
55
56 <li>I've upgraded to a new version of LLVM, and I get strange build
57 errors.</li>
58
59 <li>I've built LLVM and am testing it, but the tests freeze.</li>
60
61 <li>Why do test results differ when I perform different types of
62 builds?</li>
63
64 <li>Compiling LLVM with GCC 3.3.2 fails, what should I do?</li>
65
66 <li>Compiling LLVM with GCC succeeds, but the resulting tools do not work,
67 what can be wrong?</li>
68
69 <li>When I use the test suite, all of the C Backend tests fail. What is
70 wrong?</li>
71
72 <li>After Subversion update, rebuilding gives the error "No rule to make
73 target".</li>
74
75 <li><a href="#llvmc">The <tt>llvmc</tt> program gives me errors/doesn't
76 work.</a></li>
Bill Wendling28436212009-04-07 18:51:13 +000077
78 <li><a href="#srcdir-objdir">When I compile LLVM-GCC with srcdir == objdir,
79 it fails. Why?</a></li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +000080 </ol></li>
81
82 <li><a href="#felangs">Source Languages</a>
83 <ol>
84 <li><a href="#langs">What source languages are supported?</a></li>
Bill Wendlingc0e81142009-04-07 18:40:56 +000085
Gordon Henriksendb0558b2008-02-22 21:55:51 +000086 <li><a href="#langirgen">I'd like to write a self-hosting LLVM compiler. How
Bill Wendlingc0e81142009-04-07 18:40:56 +000087 should I interface with the LLVM middle-end optimizers and back-end code
88 generators?</a></li>
89
Dan Gohmanf17a25c2007-07-18 16:29:46 +000090 <li><a href="#langhlsupp">What support is there for higher level source
Bill Wendlingc0e81142009-04-07 18:40:56 +000091 language constructs for building a compiler?</a></li>
92
Dan Gohmanf17a25c2007-07-18 16:29:46 +000093 <li><a href="GetElementPtr.html">I don't understand the GetElementPtr
94 instruction. Help!</a></li>
95 </ol>
96
97 <li><a href="#cfe">Using the GCC Front End</a>
98 <ol>
Bill Wendlingc0e81142009-04-07 18:40:56 +000099 <li>When I compile software that uses a configure script, the configure
100 script thinks my system has all of the header files and libraries it is
101 testing for. How do I get configure to work correctly?</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000102
Bill Wendlingc0e81142009-04-07 18:40:56 +0000103 <li>When I compile code using the LLVM GCC front end, it complains that it
104 cannot find libcrtend.a?</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000105
Bill Wendlingc0e81142009-04-07 18:40:56 +0000106 <li>How can I disable all optimizations when compiling code using the LLVM
107 GCC front end?</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000108
Bill Wendlingc0e81142009-04-07 18:40:56 +0000109 <li><a href="#translatecxx">Can I use LLVM to convert C++ code to C
110 code?</a></li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000111
Bill Wendlingc0e81142009-04-07 18:40:56 +0000112 <li><a href="#platformindependent">Can I compile C or C++ code to
113 platform-independent LLVM bitcode?</a></li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000114 </ol>
115 </li>
116
117 <li><a href="#cfe_code">Questions about code generated by the GCC front-end</a>
118 <ol>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000119 <li><a href="#iosinit">What is this <tt>llvm.global_ctors</tt> and
120 <tt>_GLOBAL__I__tmp_webcompile...</tt> stuff that happens when I
121 #include &lt;iostream&gt;?</a></li>
Bill Wendlingc0e81142009-04-07 18:40:56 +0000122
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000123 <li><a href="#codedce">Where did all of my code go??</a></li>
Bill Wendlingc0e81142009-04-07 18:40:56 +0000124
125 <li><a href="#undef">What is this "<tt>undef</tt>" thing that shows up in
126 my code?</a></li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000127 </ol>
128 </li>
129</ol>
130
131<div class="doc_author">
132 <p>Written by <a href="http://llvm.org">The LLVM Team</a></p>
133</div>
134
135
136<!-- *********************************************************************** -->
137<div class="doc_section">
138 <a name="license">License</a>
139</div>
140<!-- *********************************************************************** -->
141
142<div class="question">
143<p>Why are the LLVM source code and the front-end distributed under different
Bill Wendlingc0e81142009-04-07 18:40:56 +0000144 licenses?</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000145</div>
146
147<div class="answer">
148<p>The C/C++ front-ends are based on GCC and must be distributed under the GPL.
Bill Wendlingc0e81142009-04-07 18:40:56 +0000149 Our aim is to distribute LLVM source code under a <em>much less
150 restrictive</em> license, in particular one that does not compel users who
151 distribute tools based on modifying the source to redistribute the modified
152 source code as well.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000153</div>
154
155<div class="question">
156<p>Does the University of Illinois Open Source License really qualify as an
Bill Wendlingc0e81142009-04-07 18:40:56 +0000157 "open source" license?</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000158</div>
159
160<div class="answer">
Bill Wendlingc0e81142009-04-07 18:40:56 +0000161<p>Yes, the license
162 is <a href="http://www.opensource.org/licenses/UoI-NCSA.php">certified</a> by
163 the Open Source Initiative (OSI).</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000164</div>
165
166<div class="question">
167<p>Can I modify LLVM source code and redistribute the modified source?</p>
168</div>
169
170<div class="answer">
171<p>Yes. The modified source distribution must retain the copyright notice and
Bill Wendlingc0e81142009-04-07 18:40:56 +0000172 follow the three bulletted conditions listed in
173 the <a href="http://llvm.org/svn/llvm-project/llvm/trunk/LICENSE.TXT">LLVM
174 license</a>.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000175</div>
176
177<div class="question">
178<p>Can I modify LLVM source code and redistribute binaries or other tools based
Bill Wendlingc0e81142009-04-07 18:40:56 +0000179 on it, without redistributing the source?</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000180</div>
181
182<div class="answer">
Bill Wendlingc0e81142009-04-07 18:40:56 +0000183<p>Yes. This is why we distribute LLVM under a less restrictive license than
184 GPL, as explained in the first question above.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000185</div>
186
187<!-- *********************************************************************** -->
188<div class="doc_section">
189 <a name="source">Source Code</a>
190</div>
191<!-- *********************************************************************** -->
192
193<div class="question">
194<p>In what language is LLVM written?</p>
195</div>
196
197<div class="answer">
198<p>All of the LLVM tools and libraries are written in C++ with extensive use of
Bill Wendlingc0e81142009-04-07 18:40:56 +0000199 the STL.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000200</div>
201
202<div class="question">
203<p>How portable is the LLVM source code?</p>
204</div>
205
206<div class="answer">
207<p>The LLVM source code should be portable to most modern UNIX-like operating
208systems. Most of the code is written in standard C++ with operating system
209services abstracted to a support library. The tools required to build and test
210LLVM have been ported to a plethora of platforms.</p>
211
212<p>Some porting problems may exist in the following areas:</p>
213
214<ul>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000215 <li>The GCC front end code is not as portable as the LLVM suite, so it may not
Bill Wendling4e05864c2007-09-22 09:54:47 +0000216 compile as well on unsupported platforms.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000217
218 <li>The LLVM build system relies heavily on UNIX shell tools, like the Bourne
Bill Wendling4e05864c2007-09-22 09:54:47 +0000219 Shell and sed. Porting to systems without these tools (MacOS 9, Plan 9)
220 will require more effort.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000221</ul>
222
223</div>
224
225<!-- *********************************************************************** -->
226<div class="doc_section">
227 <a name="build">Build Problems</a>
228</div>
229<!-- *********************************************************************** -->
230
231<div class="question">
232<p>When I run configure, it finds the wrong C compiler.</p>
233</div>
234
235<div class="answer">
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000236<p>The <tt>configure</tt> script attempts to locate first <tt>gcc</tt> and then
Bill Wendlingc0e81142009-04-07 18:40:56 +0000237 <tt>cc</tt>, unless it finds compiler paths set in <tt>CC</tt>
238 and <tt>CXX</tt> for the C and C++ compiler, respectively.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000239
240<p>If <tt>configure</tt> finds the wrong compiler, either adjust your
Bill Wendlingc0e81142009-04-07 18:40:56 +0000241 <tt>PATH</tt> environment variable or set <tt>CC</tt> and <tt>CXX</tt>
242 explicitly.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000243
244</div>
245
246<div class="question">
247<p>The <tt>configure</tt> script finds the right C compiler, but it uses the
Bill Wendlingc0e81142009-04-07 18:40:56 +0000248 LLVM linker from a previous build. What do I do?</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000249</div>
250
251<div class="answer">
252<p>The <tt>configure</tt> script uses the <tt>PATH</tt> to find executables, so
Bill Wendlingc0e81142009-04-07 18:40:56 +0000253 if it's grabbing the wrong linker/assembler/etc, there are two ways to fix
254 it:</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000255
256<ol>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000257 <li><p>Adjust your <tt>PATH</tt> environment variable so that the correct
Bill Wendling4e05864c2007-09-22 09:54:47 +0000258 program appears first in the <tt>PATH</tt>. This may work, but may not be
259 convenient when you want them <i>first</i> in your path for other
260 work.</p></li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000261
262 <li><p>Run <tt>configure</tt> with an alternative <tt>PATH</tt> that is
Bill Wendling4e05864c2007-09-22 09:54:47 +0000263 correct. In a Borne compatible shell, the syntax would be:</p>
264
Misha Brukman4fbcf682008-12-17 18:11:40 +0000265<pre class="doc_code">
Bill Wendling4e05864c2007-09-22 09:54:47 +0000266% PATH=[the path without the bad program] ./configure ...
267</pre>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000268
269 <p>This is still somewhat inconvenient, but it allows <tt>configure</tt>
Bill Wendling4e05864c2007-09-22 09:54:47 +0000270 to do its work without having to adjust your <tt>PATH</tt>
271 permanently.</p></li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000272</ol>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000273</div>
274
275<div class="question">
276<p>When creating a dynamic library, I get a strange GLIBC error.</p>
277</div>
278
279<div class="answer">
280<p>Under some operating systems (i.e. Linux), libtool does not work correctly if
Bill Wendlingc0e81142009-04-07 18:40:56 +0000281 GCC was compiled with the --disable-shared option. To work around this,
282 install your own version of GCC that has shared libraries enabled by
283 default.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000284</div>
285
286<div class="question">
Bill Wendlingc0e81142009-04-07 18:40:56 +0000287<p>I've updated my source tree from Subversion, and now my build is trying to
288 use a file/directory that doesn't exist.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000289</div>
290
291<div class="answer">
292<p>You need to re-run configure in your object directory. When new Makefiles
Bill Wendlingc0e81142009-04-07 18:40:56 +0000293 are added to the source tree, they have to be copied over to the object tree
294 in order to be used by the build.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000295</div>
296
297<div class="question">
298<p>I've modified a Makefile in my source tree, but my build tree keeps using the
Bill Wendlingc0e81142009-04-07 18:40:56 +0000299 old version. What do I do?</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000300</div>
301
302<div class="answer">
Bill Wendlingc0e81142009-04-07 18:40:56 +0000303<p>If the Makefile already exists in your object tree, you can just run the
304 following command in the top level directory of your object tree:</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000305
Misha Brukman4fbcf682008-12-17 18:11:40 +0000306<pre class="doc_code">
307% ./config.status &lt;relative path to Makefile&gt;
308</pre>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000309
310<p>If the Makefile is new, you will have to modify the configure script to copy
Bill Wendlingc0e81142009-04-07 18:40:56 +0000311 it over.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000312</div>
313
314<div class="question">
315<p>I've upgraded to a new version of LLVM, and I get strange build errors.</p>
316</div>
317
318<div class="answer">
319
320<p>Sometimes, changes to the LLVM source code alters how the build system works.
Bill Wendlingc0e81142009-04-07 18:40:56 +0000321 Changes in libtool, autoconf, or header file dependencies are especially
322 prone to this sort of problem.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000323
324<p>The best thing to try is to remove the old files and re-build. In most
Bill Wendlingc0e81142009-04-07 18:40:56 +0000325 cases, this takes care of the problem. To do this, just type <tt>make
326 clean</tt> and then <tt>make</tt> in the directory that fails to build.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000327</div>
328
329<div class="question">
330<p>I've built LLVM and am testing it, but the tests freeze.</p>
331</div>
332
333<div class="answer">
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000334<p>This is most likely occurring because you built a profile or release
Bill Wendlingc0e81142009-04-07 18:40:56 +0000335 (optimized) build of LLVM and have not specified the same information on the
336 <tt>gmake</tt> command line.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000337
338<p>For example, if you built LLVM with the command:</p>
339
Misha Brukman4fbcf682008-12-17 18:11:40 +0000340<pre class="doc_code">
341% gmake ENABLE_PROFILING=1
342</pre>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000343
344<p>...then you must run the tests with the following commands:</p>
345
Misha Brukman4fbcf682008-12-17 18:11:40 +0000346<pre class="doc_code">
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000347% cd llvm/test
348% gmake ENABLE_PROFILING=1
349</pre>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000350</div>
351
352<div class="question">
353<p>Why do test results differ when I perform different types of builds?</p>
354</div>
355
356<div class="answer">
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000357<p>The LLVM test suite is dependent upon several features of the LLVM tools and
Bill Wendlingc0e81142009-04-07 18:40:56 +0000358 libraries.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000359
360<p>First, the debugging assertions in code are not enabled in optimized or
Bill Wendlingc0e81142009-04-07 18:40:56 +0000361 profiling builds. Hence, tests that used to fail may pass.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000362
363<p>Second, some tests may rely upon debugging options or behavior that is only
Bill Wendlingc0e81142009-04-07 18:40:56 +0000364 available in the debug build. These tests will fail in an optimized or
365 profile build.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000366</div>
367
368<div class="question">
369<p>Compiling LLVM with GCC 3.3.2 fails, what should I do?</p>
370</div>
371
372<div class="answer">
Bill Wendlingc0e81142009-04-07 18:40:56 +0000373<p>This is <a href="http://gcc.gnu.org/bugzilla/show_bug.cgi?id=13392">a bug in
374 GCC</a>, and affects projects other than LLVM. Try upgrading or downgrading
375 your GCC.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000376</div>
377
378<div class="question">
Bill Wendlingc0e81142009-04-07 18:40:56 +0000379<p>Compiling LLVM with GCC succeeds, but the resulting tools do not work, what
380 can be wrong?</p>
Gabor Greif5b6141c2009-03-02 19:08:05 +0000381</div>
382
383<div class="answer">
Bill Wendlingc0e81142009-04-07 18:40:56 +0000384<p>Several versions of GCC have shown a weakness in miscompiling the LLVM
385 codebase. Please consult your compiler version (<tt>gcc --version</tt>) to
386 find out whether it is <a href="GettingStarted.html#brokengcc">broken</a>.
387 If so, your only option is to upgrade GCC to a known good version.</p>
Gabor Greif5b6141c2009-03-02 19:08:05 +0000388</div>
389
390<div class="question">
Bill Wendling4e05864c2007-09-22 09:54:47 +0000391<p>After Subversion update, rebuilding gives the error "No rule to make
Bill Wendlingc0e81142009-04-07 18:40:56 +0000392 target".</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000393</div>
394
395<div class="answer">
396<p>If the error is of the form:</p>
397
Misha Brukman4fbcf682008-12-17 18:11:40 +0000398<pre class="doc_code">
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000399gmake[2]: *** No rule to make target `/path/to/somefile', needed by
400`/path/to/another/file.d'.<br>
401Stop.
402</pre>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000403
Bill Wendlingc0e81142009-04-07 18:40:56 +0000404<p>This may occur anytime files are moved within the Subversion repository or
405 removed entirely. In this case, the best solution is to erase all
406 <tt>.d</tt> files, which list dependencies for source files, and rebuild:</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000407
Misha Brukman4fbcf682008-12-17 18:11:40 +0000408<pre class="doc_code">
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000409% cd $LLVM_OBJ_DIR
410% rm -f `find . -name \*\.d`
411% gmake
412</pre>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000413
414<p>In other cases, it may be necessary to run <tt>make clean</tt> before
Bill Wendlingc0e81142009-04-07 18:40:56 +0000415 rebuilding.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000416</div>
417
Bill Wendling28436212009-04-07 18:51:13 +0000418<div class="question">
419<p><a name="llvmc">The <tt>llvmc</tt> program gives me errors/doesn't
420 work.</a></p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000421</div>
422
423<div class="answer">
Bill Wendling4e05864c2007-09-22 09:54:47 +0000424<p><tt>llvmc</tt> is experimental and isn't really supported. We suggest
Bill Wendlingc0e81142009-04-07 18:40:56 +0000425 using <tt>llvm-gcc</tt> instead.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000426</div>
427
Bill Wendling28436212009-04-07 18:51:13 +0000428<div class="question">
Bill Wendling4bf848d2009-04-07 18:54:06 +0000429<p><a name="srcdir-objdir">When I compile LLVM-GCC with srcdir == objdir, it
Bill Wendling28436212009-04-07 18:51:13 +0000430 fails. Why?</a></p>
431</div>
432
433<div class="answer">
434<p>The <tt>GNUmakefile</tt> in the top-level directory of LLVM-GCC is a special
435 <tt>Makefile</tt> used by Apple to invoke the <tt>build_gcc</tt> script after
436 setting up a special environment. This has the unforunate side-effect that
437 trying to build LLVM-GCC with srcdir == objdir in a "non-Apple way" invokes
438 the <tt>GNUmakefile</tt> instead of <tt>Makefile</tt>. Because the
439 environment isn't set up correctly to do this, the build fails.</p>
440
441<p>People not building LLVM-GCC the "Apple way" need to build LLVM-GCC with
442 srcdir != objdir, or simply remove the GNUmakefile entirely.</p>
443
444<p>We regret the inconvenience.</p>
445</div>
446
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000447<!-- *********************************************************************** -->
448<div class="doc_section"><a name="felangs">Source Languages</a></div>
449
Bill Wendlingc0e81142009-04-07 18:40:56 +0000450<div class="question">
451<p><a name="langs">What source languages are supported?</a></p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000452</div>
Gordon Henriksen2c2e1482008-02-22 20:58:29 +0000453
Gordon Henriksen2c2e1482008-02-22 20:58:29 +0000454<div class="answer">
Bill Wendlingc0e81142009-04-07 18:40:56 +0000455<p>LLVM currently has full support for C and C++ source languages. These are
456 available through a special version of GCC that LLVM calls the
457 <a href="#cfe">C Front End</a></p>
458
459<p>There is an incomplete version of a Java front end available in the
460 <tt>java</tt> module. There is no documentation on this yet so you'll need to
461 download the code, compile it, and try it.</p>
462
463<p>The PyPy developers are working on integrating LLVM into the PyPy backend so
464 that PyPy language can translate to LLVM.</p>
Gordon Henriksen2c2e1482008-02-22 20:58:29 +0000465</div>
466
Bill Wendlingc0e81142009-04-07 18:40:56 +0000467<div class="question">
468<p><a name="langirgen">I'd like to write a self-hosting LLVM compiler. How
469 should I interface with the LLVM middle-end optimizers and back-end code
470 generators?</a></p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000471</div>
472
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000473<div class="answer">
Bill Wendlingc0e81142009-04-07 18:40:56 +0000474<p>Your compiler front-end will communicate with LLVM by creating a module in
475 the LLVM intermediate representation (IR) format. Assuming you want to write
476 your language's compiler in the language itself (rather than C++), there are
477 3 major ways to tackle generating LLVM IR from a front-end:</p>
478
479<ul>
480 <li><strong>Call into the LLVM libraries code using your language's FFI
481 (foreign function interface).</strong>
482
483 <ul>
484 <li><em>for:</em> best tracks changes to the LLVM IR, .ll syntax, and .bc
485 format</li>
486
487 <li><em>for:</em> enables running LLVM optimization passes without a
488 emit/parse overhead</li>
489
490 <li><em>for:</em> adapts well to a JIT context</li>
491
492 <li><em>against:</em> lots of ugly glue code to write</li>
493 </ul></li>
494
495 <li> <strong>Emit LLVM assembly from your compiler's native language.</strong>
496 <ul>
497 <li><em>for:</em> very straightforward to get started</li>
498
499 <li><em>against:</em> the .ll parser is slower than the bitcode reader
500 when interfacing to the middle end</li>
501
502 <li><em>against:</em> you'll have to re-engineer the LLVM IR object model
503 and asm writer in your language</li>
504
505 <li><em>against:</em> it may be harder to track changes to the IR</li>
506 </ul></li>
507
508 <li><strong>Emit LLVM bitcode from your compiler's native language.</strong>
509
510 <ul>
511 <li><em>for:</em> can use the more-efficient bitcode reader when
512 interfacing to the middle end</li>
513
514 <li><em>against:</em> you'll have to re-engineer the LLVM IR object
515 model and bitcode writer in your language</li>
516
517 <li><em>against:</em> it may be harder to track changes to the IR</li>
518 </ul></li>
519</ul>
520
521<p>If you go with the first option, the C bindings in include/llvm-c should help
522 a lot, since most languages have strong support for interfacing with C. The
523 most common hurdle with calling C from managed code is interfacing with the
524 garbage collector. The C interface was designed to require very little memory
525 management, and so is straightforward in this regard.</p>
526</div>
527
528<div class="question">
529<p><a name="langhlsupp">What support is there for a higher level source language
530 constructs for building a compiler?</a></p>
531</div>
532
533<div class="answer">
534<p>Currently, there isn't much. LLVM supports an intermediate representation
535 which is useful for code representation but will not support the high level
536 (abstract syntax tree) representation needed by most compilers. There are no
537 facilities for lexical nor semantic analysis. There is, however, a <i>mostly
538 implemented</i> configuration-driven
539 <a href="CompilerDriver.html">compiler driver</a> which simplifies the task
540 of running optimizations, linking, and executable generation.</p>
541</div>
542
543<div class="question">
544<p><a name="getelementptr">I don't understand the GetElementPtr
545 instruction. Help!</a></p>
546</div>
547
548<div class="answer">
549<p>See <a href="GetElementPtr.html">The Often Misunderstood GEP
Bill Wendling4e05864c2007-09-22 09:54:47 +0000550 Instruction</a>.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000551</div>
552
553<!-- *********************************************************************** -->
554<div class="doc_section">
555 <a name="cfe">Using the GCC Front End</a>
556</div>
557
558<div class="question">
Bill Wendlingc0e81142009-04-07 18:40:56 +0000559<p>When I compile software that uses a configure script, the configure script
560 thinks my system has all of the header files and libraries it is testing for.
561 How do I get configure to work correctly?</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000562</div>
563
564<div class="answer">
Bill Wendlingc0e81142009-04-07 18:40:56 +0000565<p>The configure script is getting things wrong because the LLVM linker allows
566 symbols to be undefined at link time (so that they can be resolved during JIT
567 or translation to the C back end). That is why configure thinks your system
568 "has everything."</p>
569
570<p>To work around this, perform the following steps:</p>
571
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000572<ol>
Bill Wendlingc0e81142009-04-07 18:40:56 +0000573 <li>Make sure the CC and CXX environment variables contains the full path to
574 the LLVM GCC front end.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000575
576 <li>Make sure that the regular C compiler is first in your PATH. </li>
577
578 <li>Add the string "-Wl,-native" to your CFLAGS environment variable.</li>
579</ol>
580
Bill Wendlingc0e81142009-04-07 18:40:56 +0000581<p>This will allow the <tt>llvm-ld</tt> linker to create a native code
582 executable instead of shell script that runs the JIT. Creating native code
583 requires standard linkage, which in turn will allow the configure script to
584 find out if code is not linking on your system because the feature isn't
585 available on your system.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000586</div>
587
588<div class="question">
Bill Wendlingc0e81142009-04-07 18:40:56 +0000589<p>When I compile code using the LLVM GCC front end, it complains that it cannot
590 find libcrtend.a.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000591</p>
592</div>
593
594<div class="answer">
Bill Wendlingc0e81142009-04-07 18:40:56 +0000595<p>The only way this can happen is if you haven't installed the runtime
596 library. To correct this, do:</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000597
Misha Brukman4fbcf682008-12-17 18:11:40 +0000598<pre class="doc_code">
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000599% cd llvm/runtime
600% make clean ; make install-bytecode
601</pre>
602</div>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000603
604<div class="question">
Bill Wendlingc0e81142009-04-07 18:40:56 +0000605<p>How can I disable all optimizations when compiling code using the LLVM GCC
606 front end?</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000607</div>
608
609<div class="answer">
Bill Wendlingc0e81142009-04-07 18:40:56 +0000610<p>Passing "-Wa,-disable-opt -Wl,-disable-opt" will disable *all* cleanup and
611 optimizations done at the llvm level, leaving you with the truly horrible
612 code that you desire.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000613</div>
614
615
616<div class="question">
Bill Wendlingc0e81142009-04-07 18:40:56 +0000617<p><a name="translatecxx">Can I use LLVM to convert C++ code to C code?</a></p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000618</div>
619
620<div class="answer">
621<p>Yes, you can use LLVM to convert code from any language LLVM supports to C.
Bill Wendlingc0e81142009-04-07 18:40:56 +0000622 Note that the generated C code will be very low level (all loops are lowered
623 to gotos, etc) and not very pretty (comments are stripped, original source
624 formatting is totally lost, variables are renamed, expressions are
625 regrouped), so this may not be what you're looking for. Also, there are
626 several limitations noted below.<p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000627
628<p>Use commands like this:</p>
629
630<ol>
Bill Wendling4e05864c2007-09-22 09:54:47 +0000631 <li><p>Compile your program as normal with llvm-g++:</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000632
Misha Brukman4fbcf682008-12-17 18:11:40 +0000633<pre class="doc_code">
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000634% llvm-g++ x.cpp -o program
635</pre>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000636
Bill Wendlingc0e81142009-04-07 18:40:56 +0000637 <p>or:</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000638
Misha Brukman4fbcf682008-12-17 18:11:40 +0000639<pre class="doc_code">
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000640% llvm-g++ a.cpp -c
641% llvm-g++ b.cpp -c
642% llvm-g++ a.o b.o -o program
643</pre>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000644
Bill Wendling4e05864c2007-09-22 09:54:47 +0000645 <p>With llvm-gcc3, this will generate program and program.bc. The .bc
646 file is the LLVM version of the program all linked together.</p></li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000647
Bill Wendling4e05864c2007-09-22 09:54:47 +0000648 <li><p>Convert the LLVM code to C code, using the LLC tool with the C
649 backend:</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000650
Misha Brukman4fbcf682008-12-17 18:11:40 +0000651<pre class="doc_code">
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000652% llc -march=c program.bc -o program.c
Misha Brukman4fbcf682008-12-17 18:11:40 +0000653</pre></li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000654
Bill Wendlingc0e81142009-04-07 18:40:56 +0000655 <li><p>Finally, compile the C file:</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000656
Misha Brukman4fbcf682008-12-17 18:11:40 +0000657<pre class="doc_code">
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000658% cc x.c
Misha Brukman4fbcf682008-12-17 18:11:40 +0000659</pre></li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000660
661</ol>
662
Bill Wendlingc0e81142009-04-07 18:40:56 +0000663<p>Using LLVM does not eliminate the need for C++ library support. If you use
664 the llvm-g++ front-end, the generated code will depend on g++'s C++ support
665 libraries in the same way that code generated from g++ would. If you use
666 another C++ front-end, the generated code will depend on whatever library
667 that front-end would normally require.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000668
Bill Wendlingc0e81142009-04-07 18:40:56 +0000669<p>If you are working on a platform that does not provide any C++ libraries, you
670 may be able to manually compile libstdc++ to LLVM bitcode, statically link it
671 into your program, then use the commands above to convert the whole result
672 into C code. Alternatively, you might compile the libraries and your
673 application into two different chunks of C code and link them.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000674
Bill Wendlingc0e81142009-04-07 18:40:56 +0000675<p>Note that, by default, the C back end does not support exception handling.
676 If you want/need it for a certain program, you can enable it by passing
677 "-enable-correct-eh-support" to the llc program. The resultant code will use
678 setjmp/longjmp to implement exception support that is relatively slow, and
679 not C++-ABI-conforming on most platforms, but otherwise correct.</p>
djga0237ec2009-01-25 16:04:50 +0000680
Bill Wendlingc0e81142009-04-07 18:40:56 +0000681<p>Also, there are a number of other limitations of the C backend that cause it
682 to produce code that does not fully conform to the C++ ABI on most
683 platforms. Some of the C++ programs in LLVM's test suite are known to fail
684 when compiled with the C back end because of ABI incompatiblities with
685 standard C++ libraries.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000686</div>
687
Dan Gohmanb8e23ab2009-02-10 17:26:53 +0000688<div class="question">
Bill Wendlingc0e81142009-04-07 18:40:56 +0000689<p><a name="platformindependent">Can I compile C or C++ code to
690 platform-independent LLVM bitcode?</a></p>
Dan Gohmanb8e23ab2009-02-10 17:26:53 +0000691</div>
692
693<div class="answer">
Bill Wendlingc0e81142009-04-07 18:40:56 +0000694<p>No. C and C++ are inherently platform-dependent languages. The most obvious
695 example of this is the preprocessor. A very common way that C code is made
696 portable is by using the preprocessor to include platform-specific code. In
697 practice, information about other platforms is lost after preprocessing, so
698 the result is inherently dependent on the platform that the preprocessing was
699 targetting.</p>
Dan Gohmanb8e23ab2009-02-10 17:26:53 +0000700
Bill Wendlingc0e81142009-04-07 18:40:56 +0000701<p>Another example is <tt>sizeof</tt>. It's common for <tt>sizeof(long)</tt> to
702 vary between platforms. In most C front-ends, <tt>sizeof</tt> is expanded to
703 a constant immediately, thus hardwaring a platform-specific detail.</p>
Dan Gohmanb8e23ab2009-02-10 17:26:53 +0000704
Bill Wendlingc0e81142009-04-07 18:40:56 +0000705<p>Also, since many platforms define their ABIs in terms of C, and since LLVM is
706 lower-level than C, front-ends currently must emit platform-specific IR in
707 order to have the result conform to the platform ABI.</p>
Dan Gohmanb8e23ab2009-02-10 17:26:53 +0000708</div>
709
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000710<!-- *********************************************************************** -->
711<div class="doc_section">
712 <a name="cfe_code">Questions about code generated by the GCC front-end</a>
713</div>
714
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000715<div class="question">
Bill Wendlingc0e81142009-04-07 18:40:56 +0000716<p><a name="iosinit">What is this <tt>llvm.global_ctors</tt> and
717 <tt>_GLOBAL__I__tmp_webcompile...</tt> stuff that happens when I <tt>#include
718 &lt;iostream&gt;</tt>?</a></p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000719</div>
720
721<div class="answer">
Bill Wendlingc0e81142009-04-07 18:40:56 +0000722<p>If you <tt>#include</tt> the <tt>&lt;iostream&gt;</tt> header into a C++
723 translation unit, the file will probably use
724 the <tt>std::cin</tt>/<tt>std::cout</tt>/... global objects. However, C++
725 does not guarantee an order of initialization between static objects in
726 different translation units, so if a static ctor/dtor in your .cpp file
727 used <tt>std::cout</tt>, for example, the object would not necessarily be
728 automatically initialized before your use.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000729
730<p>To make <tt>std::cout</tt> and friends work correctly in these scenarios, the
Bill Wendlingc0e81142009-04-07 18:40:56 +0000731 STL that we use declares a static object that gets created in every
732 translation unit that includes <tt>&lt;iostream&gt;</tt>. This object has a
733 static constructor and destructor that initializes and destroys the global
734 iostream objects before they could possibly be used in the file. The code
735 that you see in the .ll file corresponds to the constructor and destructor
736 registration code.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000737</p>
738
739<p>If you would like to make it easier to <b>understand</b> the LLVM code
Bill Wendlingc0e81142009-04-07 18:40:56 +0000740 generated by the compiler in the demo page, consider using <tt>printf()</tt>
741 instead of <tt>iostream</tt>s to print values.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000742</div>
743
744<!--=========================================================================-->
745
Bill Wendlingc0e81142009-04-07 18:40:56 +0000746<div class="question">
747<p><a name="codedce">Where did all of my code go??</a></p>
748</div>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000749
750<div class="answer">
Bill Wendlingc0e81142009-04-07 18:40:56 +0000751<p>If you are using the LLVM demo page, you may often wonder what happened to
752 all of the code that you typed in. Remember that the demo script is running
753 the code through the LLVM optimizers, so if your code doesn't actually do
754 anything useful, it might all be deleted.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000755
Bill Wendlingc0e81142009-04-07 18:40:56 +0000756<p>To prevent this, make sure that the code is actually needed. For example, if
757 you are computing some expression, return the value from the function instead
758 of leaving it in a local variable. If you really want to constrain the
759 optimizer, you can read from and assign to <tt>volatile</tt> global
760 variables.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000761</div>
762
763<!--=========================================================================-->
764
Bill Wendlingc0e81142009-04-07 18:40:56 +0000765<div class="question">
766<p><a name="undef">What is this "<tt>undef</tt>" thing that shows up in my
Bill Wendlinga5b79d72009-04-07 18:52:30 +0000767 code?</a></p>
Bill Wendlingc0e81142009-04-07 18:40:56 +0000768</div>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000769
770<div class="answer">
Bill Wendlingc0e81142009-04-07 18:40:56 +0000771<p><a href="LangRef.html#undef"><tt>undef</tt></a> is the LLVM way of
772 representing a value that is not defined. You can get these if you do not
773 initialize a variable before you use it. For example, the C function:</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000774
Misha Brukman4fbcf682008-12-17 18:11:40 +0000775<pre class="doc_code">
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000776int X() { int i; return i; }
777</pre>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000778
Bill Wendlingc0e81142009-04-07 18:40:56 +0000779<p>Is compiled to "<tt>ret i32 undef</tt>" because "<tt>i</tt>" never has a
780 value specified for it.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000781</div>
782
783<!-- *********************************************************************** -->
784
785<hr>
786<address>
787 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
Misha Brukman947321d2008-12-11 17:34:48 +0000788 src="http://jigsaw.w3.org/css-validator/images/vcss-blue" alt="Valid CSS"></a>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000789 <a href="http://validator.w3.org/check/referer"><img
Misha Brukman947321d2008-12-11 17:34:48 +0000790 src="http://www.w3.org/Icons/valid-html401-blue" alt="Valid HTML 4.01"></a>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000791
792 <a href="http://llvm.org">LLVM Compiler Infrastructure</a><br>
793 Last modified: $Date$
794</address>
795
796</body>
797</html>