Gitiles
Code Review Sign In
gerrit-public.fairphone.software / fp2-dev / platform / external / llvm / b8e23ab91f339f5ea355ade92c78a1e8ec406c71 / . / docs / FAQ.html
blob: d94c368d0c7b0f3532010f5a4680bb2e971dd504 [file] [log] [blame]
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01//EN"
2 "http://www.w3.org/TR/html4/strict.dtd">
3<html>
4<head>
5 <title>LLVM: Frequently Asked Questions</title>
6 <style type="text/css">
7 @import url("llvm.css");
8 .question { font-weight: bold }
9 .answer { margin-left: 2em }
10 </style>
11</head>
12<body>
13
14<div class="doc_title">
15 LLVM: Frequently Asked Questions
16</div>
17
18<ol>
19 <li><a href="#license">License</a>
20 <ol>
21 <li>Why are the LLVM source code and the front-end distributed under different
22 licenses?</li>
23 <li>Does the University of Illinois Open Source License really qualify as an
24 "open source" license?</li>
25 <li>Can I modify LLVM source code and redistribute the modified source?</li>
26 <li>Can I modify LLVM source code and redistribute binaries or other tools
27 based on it, without redistributing the source?</li>
28 </ol></li>
29
30 <li><a href="#source">Source code</a>
31 <ol>
32 <li>In what language is LLVM written?</li>
33 <li>How portable is the LLVM source code?</li>
34 </ol></li>
35
36 <li><a href="#build">Build Problems</a>
37 <ol>
38 <li>When I run configure, it finds the wrong C compiler.</li>
39 <li>The <tt>configure</tt> script finds the right C compiler, but it uses the
40 LLVM linker from a previous build. What do I do?</li>
41 <li>When creating a dynamic library, I get a strange GLIBC error.</li>
42 <li>I've updated my source tree from Subversion, and now my build is trying
43 to use a file/directory that doesn't exist.</li>
44 <li>I've modified a Makefile in my source tree, but my build tree keeps using
45 the old version. What do I do?</li>
46 <li>I've upgraded to a new version of LLVM, and I get strange build
47 errors.</li>
48 <li>I've built LLVM and am testing it, but the tests freeze.</li>
49 <li>Why do test results differ when I perform different types of builds?</li>
50 <li>Compiling LLVM with GCC 3.3.2 fails, what should I do?</li>
51 <li>When I use the test suite, all of the C Backend tests fail. What is
52 wrong?</li>
53 <li>After Subversion update, rebuilding gives the error "No rule to make
54 target".</li>
55 <li><a href="#llvmc">The <tt>llvmc</tt> program gives me errors/doesn't
Bill Wendling4e05864c2007-09-22 09:54:47 +0000[diff] [blame]56 work.</a></li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]57 </ol></li>
58
59 <li><a href="#felangs">Source Languages</a>
60 <ol>
61 <li><a href="#langs">What source languages are supported?</a></li>
Gordon Henriksendb0558b2008-02-22 21:55:51 +0000[diff] [blame]62 <li><a href="#langirgen">I'd like to write a self-hosting LLVM compiler. How
63 should I interface with the LLVM middle-end optimizers and back-end code
Dan Gohman4dfac702008-11-24 17:18:39 +0000[diff] [blame]64 generators?</a></li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]65 <li><a href="#langhlsupp">What support is there for higher level source
66 language constructs for building a compiler?</a></li>
67 <li><a href="GetElementPtr.html">I don't understand the GetElementPtr
68 instruction. Help!</a></li>
69 </ol>
70
71 <li><a href="#cfe">Using the GCC Front End</a>
72 <ol>
73 <li>
74 When I compile software that uses a configure script, the configure script
75 thinks my system has all of the header files and libraries it is testing
76 for. How do I get configure to work correctly?
77 </li>
78
79 <li>
80 When I compile code using the LLVM GCC front end, it complains that it
81 cannot find libcrtend.a.
82 </li>
83
84 <li>
85 How can I disable all optimizations when compiling code using the LLVM GCC front end?
86 </li>
87
Dan Gohman4dfac702008-11-24 17:18:39 +0000[diff] [blame]88 <li><a href="#translatecxx">Can I use LLVM to convert C++ code to C code?</a></li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]89
Dan Gohmanb8e23ab2009-02-10 17:26:53 +0000[diff] [blame^]90 <li><a href="#platformindependent">Can I compile C or C++ code to platform-independent LLVM bitcode?</a></li>
91
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]92 </ol>
93 </li>
94
95 <li><a href="#cfe_code">Questions about code generated by the GCC front-end</a>
96 <ol>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]97 <li><a href="#iosinit">What is this <tt>llvm.global_ctors</tt> and
98 <tt>_GLOBAL__I__tmp_webcompile...</tt> stuff that happens when I
99 #include &lt;iostream&gt;?</a></li>
100 <li><a href="#codedce">Where did all of my code go??</a></li>
101 <li><a href="#undef">What is this "<tt>undef</tt>" thing that shows up in my code?</a></li>
102 </ol>
103 </li>
104</ol>
105
106<div class="doc_author">
107 <p>Written by <a href="http://llvm.org">The LLVM Team</a></p>
108</div>
109
110
111<!-- *********************************************************************** -->
112<div class="doc_section">
113 <a name="license">License</a>
114</div>
115<!-- *********************************************************************** -->
116
117<div class="question">
118<p>Why are the LLVM source code and the front-end distributed under different
119licenses?</p>
120</div>
121
122<div class="answer">
123<p>The C/C++ front-ends are based on GCC and must be distributed under the GPL.
124Our aim is to distribute LLVM source code under a <em>much less restrictive</em>
125license, in particular one that does not compel users who distribute tools based
126on modifying the source to redistribute the modified source code as well.</p>
127</div>
128
129<div class="question">
130<p>Does the University of Illinois Open Source License really qualify as an
131"open source" license?</p>
132</div>
133
134<div class="answer">
135<p>Yes, the license is <a
136href="http://www.opensource.org/licenses/UoI-NCSA.php">certified</a> by the Open
137Source Initiative (OSI).</p>
138</div>
139
140<div class="question">
141<p>Can I modify LLVM source code and redistribute the modified source?</p>
142</div>
143
144<div class="answer">
145<p>Yes. The modified source distribution must retain the copyright notice and
146follow the three bulletted conditions listed in the <a
Misha Brukman5f135db2008-12-17 18:06:53 +0000[diff] [blame]147href="http://llvm.org/svn/llvm-project/llvm/trunk/LICENSE.TXT">LLVM license</a>.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]148</div>
149
150<div class="question">
151<p>Can I modify LLVM source code and redistribute binaries or other tools based
152on it, without redistributing the source?</p>
153</div>
154
155<div class="answer">
156<p>Yes, this is why we distribute LLVM under a less restrictive license than
157GPL, as explained in the first question above.</p>
158</div>
159
160<!-- *********************************************************************** -->
161<div class="doc_section">
162 <a name="source">Source Code</a>
163</div>
164<!-- *********************************************************************** -->
165
166<div class="question">
167<p>In what language is LLVM written?</p>
168</div>
169
170<div class="answer">
171<p>All of the LLVM tools and libraries are written in C++ with extensive use of
172the STL.</p>
173</div>
174
175<div class="question">
176<p>How portable is the LLVM source code?</p>
177</div>
178
179<div class="answer">
180<p>The LLVM source code should be portable to most modern UNIX-like operating
181systems. Most of the code is written in standard C++ with operating system
182services abstracted to a support library. The tools required to build and test
183LLVM have been ported to a plethora of platforms.</p>
184
185<p>Some porting problems may exist in the following areas:</p>
186
187<ul>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]188 <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 +0000[diff] [blame]189 compile as well on unsupported platforms.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]190
191 <li>The LLVM build system relies heavily on UNIX shell tools, like the Bourne
Bill Wendling4e05864c2007-09-22 09:54:47 +0000[diff] [blame]192 Shell and sed. Porting to systems without these tools (MacOS 9, Plan 9)
193 will require more effort.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]194</ul>
195
196</div>
197
198<!-- *********************************************************************** -->
199<div class="doc_section">
200 <a name="build">Build Problems</a>
201</div>
202<!-- *********************************************************************** -->
203
204<div class="question">
205<p>When I run configure, it finds the wrong C compiler.</p>
206</div>
207
208<div class="answer">
209
210<p>The <tt>configure</tt> script attempts to locate first <tt>gcc</tt> and then
211<tt>cc</tt>, unless it finds compiler paths set in <tt>CC</tt> and <tt>CXX</tt>
212for the C and C++ compiler, respectively.</p>
213
214<p>If <tt>configure</tt> finds the wrong compiler, either adjust your
215<tt>PATH</tt> environment variable or set <tt>CC</tt> and <tt>CXX</tt>
216explicitly.</p>
217
218</div>
219
220<div class="question">
221<p>The <tt>configure</tt> script finds the right C compiler, but it uses the
222LLVM linker from a previous build. What do I do?</p>
223</div>
224
225<div class="answer">
226<p>The <tt>configure</tt> script uses the <tt>PATH</tt> to find executables, so
227if it's grabbing the wrong linker/assembler/etc, there are two ways to fix
228it:</p>
229
230<ol>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]231 <li><p>Adjust your <tt>PATH</tt> environment variable so that the correct
Bill Wendling4e05864c2007-09-22 09:54:47 +0000[diff] [blame]232 program appears first in the <tt>PATH</tt>. This may work, but may not be
233 convenient when you want them <i>first</i> in your path for other
234 work.</p></li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]235
236 <li><p>Run <tt>configure</tt> with an alternative <tt>PATH</tt> that is
Bill Wendling4e05864c2007-09-22 09:54:47 +0000[diff] [blame]237 correct. In a Borne compatible shell, the syntax would be:</p>
238
Misha Brukman4fbcf682008-12-17 18:11:40 +0000[diff] [blame]239<pre class="doc_code">
Bill Wendling4e05864c2007-09-22 09:54:47 +0000[diff] [blame]240% PATH=[the path without the bad program] ./configure ...
241</pre>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]242
243 <p>This is still somewhat inconvenient, but it allows <tt>configure</tt>
Bill Wendling4e05864c2007-09-22 09:54:47 +0000[diff] [blame]244 to do its work without having to adjust your <tt>PATH</tt>
245 permanently.</p></li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]246</ol>
247
248</div>
249
250<div class="question">
251<p>When creating a dynamic library, I get a strange GLIBC error.</p>
252</div>
253
254<div class="answer">
255<p>Under some operating systems (i.e. Linux), libtool does not work correctly if
256GCC was compiled with the --disable-shared option. To work around this, install
257your own version of GCC that has shared libraries enabled by default.</p>
258</div>
259
260<div class="question">
Bill Wendling4e05864c2007-09-22 09:54:47 +0000[diff] [blame]261<p>I've updated my source tree from Subversion, and now my build is trying to
262use a file/directory that doesn't exist.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]263</div>
264
265<div class="answer">
266<p>You need to re-run configure in your object directory. When new Makefiles
267are added to the source tree, they have to be copied over to the object tree in
268order to be used by the build.</p>
269</div>
270
271<div class="question">
272<p>I've modified a Makefile in my source tree, but my build tree keeps using the
273old version. What do I do?</p>
274</div>
275
276<div class="answer">
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]277<p>If the Makefile already exists in your object tree, you
278can just run the following command in the top level directory of your object
279tree:</p>
280
Misha Brukman4fbcf682008-12-17 18:11:40 +0000[diff] [blame]281<pre class="doc_code">
282% ./config.status &lt;relative path to Makefile&gt;
283</pre>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]284
285<p>If the Makefile is new, you will have to modify the configure script to copy
286it over.</p>
287
288</div>
289
290<div class="question">
291<p>I've upgraded to a new version of LLVM, and I get strange build errors.</p>
292</div>
293
294<div class="answer">
295
296<p>Sometimes, changes to the LLVM source code alters how the build system works.
297Changes in libtool, autoconf, or header file dependencies are especially prone
298to this sort of problem.</p>
299
300<p>The best thing to try is to remove the old files and re-build. In most
301cases, this takes care of the problem. To do this, just type <tt>make
302clean</tt> and then <tt>make</tt> in the directory that fails to build.</p>
303
304</div>
305
306<div class="question">
307<p>I've built LLVM and am testing it, but the tests freeze.</p>
308</div>
309
310<div class="answer">
311
312<p>This is most likely occurring because you built a profile or release
313(optimized) build of LLVM and have not specified the same information on the
314<tt>gmake</tt> command line.</p>
315
316<p>For example, if you built LLVM with the command:</p>
317
Misha Brukman4fbcf682008-12-17 18:11:40 +0000[diff] [blame]318<pre class="doc_code">
319% gmake ENABLE_PROFILING=1
320</pre>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]321
322<p>...then you must run the tests with the following commands:</p>
323
Misha Brukman4fbcf682008-12-17 18:11:40 +0000[diff] [blame]324<pre class="doc_code">
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]325% cd llvm/test
326% gmake ENABLE_PROFILING=1
327</pre>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]328
329</div>
330
331<div class="question">
332<p>Why do test results differ when I perform different types of builds?</p>
333</div>
334
335<div class="answer">
336
337<p>The LLVM test suite is dependent upon several features of the LLVM tools and
338libraries.</p>
339
340<p>First, the debugging assertions in code are not enabled in optimized or
341profiling builds. Hence, tests that used to fail may pass.</p>
342
343<p>Second, some tests may rely upon debugging options or behavior that is only
344available in the debug build. These tests will fail in an optimized or profile
345build.</p>
346
347</div>
348
349<div class="question">
350<p>Compiling LLVM with GCC 3.3.2 fails, what should I do?</p>
351</div>
352
353<div class="answer">
354<p>This is <a href="http://gcc.gnu.org/PR?13392">a bug in GCC</a>, and
Bill Wendling4e05864c2007-09-22 09:54:47 +0000[diff] [blame]355affects projects other than LLVM. Try upgrading or downgrading your GCC.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]356</div>
357
358<div class="question">
Bill Wendling4e05864c2007-09-22 09:54:47 +0000[diff] [blame]359<p>After Subversion update, rebuilding gives the error "No rule to make
360target".</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]361</div>
362
363<div class="answer">
364<p>If the error is of the form:</p>
365
Misha Brukman4fbcf682008-12-17 18:11:40 +0000[diff] [blame]366<pre class="doc_code">
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]367gmake[2]: *** No rule to make target `/path/to/somefile', needed by
368`/path/to/another/file.d'.<br>
369Stop.
370</pre>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]371
372<p>This may occur anytime files are moved within the Subversion repository or
373removed entirely. In this case, the best solution is to erase all
374<tt>.d</tt> files, which list dependencies for source files, and rebuild:</p>
375
Misha Brukman4fbcf682008-12-17 18:11:40 +0000[diff] [blame]376<pre class="doc_code">
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]377% cd $LLVM_OBJ_DIR
378% rm -f `find . -name \*\.d`
379% gmake
380</pre>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]381
382<p>In other cases, it may be necessary to run <tt>make clean</tt> before
383rebuilding.</p>
384</div>
385
Bill Wendling4e05864c2007-09-22 09:54:47 +0000[diff] [blame]386<div class="question"><p><a name="llvmc">
387The <tt>llvmc</tt> program gives me errors/doesn't work.</a></p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]388</div>
389
390<div class="answer">
Bill Wendling4e05864c2007-09-22 09:54:47 +0000[diff] [blame]391<p><tt>llvmc</tt> is experimental and isn't really supported. We suggest
392using <tt>llvm-gcc</tt> instead.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]393</div>
394
395<!-- *********************************************************************** -->
396<div class="doc_section"><a name="felangs">Source Languages</a></div>
397
398<div class="question"><p>
399 <a name="langs">What source languages are supported?</a></p>
400</div>
401<div class="answer">
402 <p>LLVM currently has full support for C and C++ source languages. These are
403 available through a special version of GCC that LLVM calls the
404 <a href="#cfe">C Front End</a></p>
405 <p>There is an incomplete version of a Java front end available in the
406 <tt>java</tt> module. There is no documentation on this yet so
407 you'll need to download the code, compile it, and try it.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]408 <p>The PyPy developers are working on integrating LLVM into the PyPy backend
409 so that PyPy language can translate to LLVM.</p>
410</div>
Gordon Henriksen2c2e1482008-02-22 20:58:29 +0000[diff] [blame]411
412<div class="question"><p><a name="langirgen">
Gordon Henriksendb0558b2008-02-22 21:55:51 +0000[diff] [blame]413 I'd like to write a self-hosting LLVM compiler. How should I interface with
414 the LLVM middle-end optimizers and back-end code generators?
Gordon Henriksen2c2e1482008-02-22 20:58:29 +0000[diff] [blame]415</a></p></div>
416<div class="answer">
417 <p>Your compiler front-end will communicate with LLVM by creating a module in
Gordon Henriksendb0558b2008-02-22 21:55:51 +0000[diff] [blame]418 the LLVM intermediate representation (IR) format. Assuming you want to
419 write your language's compiler in the language itself (rather than C++),
420 there are 3 major ways to tackle generating LLVM IR from a front-end:</p>
Gordon Henriksen2c2e1482008-02-22 20:58:29 +0000[diff] [blame]421 <ul>
422 <li>
423 <strong>Call into the LLVM libraries code using your language's FFI
424 (foreign function interface).</strong>
425 <ul>
426 <li><em>for:</em> best tracks changes to the LLVM IR, .ll syntax,
427 and .bc format</li>
428 <li><em>for:</em> enables running LLVM optimization passes without a
429 emit/parse overhead</li>
430 <li><em>for:</em> adapts well to a JIT context</li>
431 <li><em>against:</em> lots of ugly glue code to write</li>
432 </ul>
433 </li>
434 <li>
435 <strong>Emit LLVM assembly from your compiler's native language.</strong>
436 <ul>
437 <li><em>for:</em> very straightforward to get started</li>
438 <li><em>against:</em> the .ll parser is slower than the bitcode reader
439 when interfacing to the middle end</li>
440 <li><em>against:</em> you'll have to re-engineer the LLVM IR object
441 model and asm writer in your language</li>
442 <li><em>against:</em> it may be harder to track changes to the IR</li>
443 </ul>
444 </li>
445 <li>
446 <strong>Emit LLVM bitcode from your compiler's native language.</strong>
447 <ul>
448 <li><em>for:</em> can use the more-efficient bitcode reader when
449 interfacing to the middle end</li>
450 <li><em>against:</em> you'll have to re-engineer the LLVM IR object
451 model and bitcode writer in your language</li>
452 <li><em>against:</em> it may be harder to track changes to the IR</li>
453 </ul>
454 </li>
455 </ul>
456 <p>If you go with the first option, the C bindings in include/llvm-c should
457 help a lot, since most languages have strong support for interfacing with
458 C. The most common hurdle with calling C from managed code is interfacing
459 with the garbage collector. The C interface was designed to require very
460 little memory management, and so is straightforward in this regard.</p>
461</div>
462
Bill Wendling4e05864c2007-09-22 09:54:47 +0000[diff] [blame]463<div class="question"><p><a name="langhlsupp">
464 What support is there for a higher level source language constructs for
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]465 building a compiler?</a></p>
466</div>
467<div class="answer">
468 <p>Currently, there isn't much. LLVM supports an intermediate representation
469 which is useful for code representation but will not support the high level
470 (abstract syntax tree) representation needed by most compilers. There are no
471 facilities for lexical nor semantic analysis. There is, however, a <i>mostly
472 implemented</i> configuration-driven
473 <a href="CompilerDriver.html">compiler driver</a> which simplifies the task
474 of running optimizations, linking, and executable generation.</p>
475</div>
476
Dan Gohman4dfac702008-11-24 17:18:39 +0000[diff] [blame]477<div class="question"><p><a name="getelementptr">
Bill Wendling4e05864c2007-09-22 09:54:47 +0000[diff] [blame]478 I don't understand the GetElementPtr instruction. Help!</a></p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]479</div>
480<div class="answer">
481 <p>See <a href="GetElementPtr.html">The Often Misunderstood GEP
Bill Wendling4e05864c2007-09-22 09:54:47 +0000[diff] [blame]482 Instruction</a>.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]483</div>
484
485<!-- *********************************************************************** -->
486<div class="doc_section">
487 <a name="cfe">Using the GCC Front End</a>
488</div>
489
490<div class="question">
491<p>
492When I compile software that uses a configure script, the configure script
493thinks my system has all of the header files and libraries it is testing for.
494How do I get configure to work correctly?
495</p>
496</div>
497
498<div class="answer">
499<p>
500The configure script is getting things wrong because the LLVM linker allows
501symbols to be undefined at link time (so that they can be resolved during JIT
502or translation to the C back end). That is why configure thinks your system
503"has everything."
504</p>
505<p>
506To work around this, perform the following steps:
507</p>
508<ol>
509 <li>Make sure the CC and CXX environment variables contains the full path to
510 the LLVM GCC front end.</li>
511
512 <li>Make sure that the regular C compiler is first in your PATH. </li>
513
514 <li>Add the string "-Wl,-native" to your CFLAGS environment variable.</li>
515</ol>
516
517<p>
518This will allow the <tt>llvm-ld</tt> linker to create a native code executable
519instead of shell script that runs the JIT. Creating native code requires
520standard linkage, which in turn will allow the configure script to find out if
521code is not linking on your system because the feature isn't available on your
522system.</p>
523</div>
524
525<div class="question">
526<p>
527When I compile code using the LLVM GCC front end, it complains that it cannot
528find libcrtend.a.
529</p>
530</div>
531
532<div class="answer">
533<p>
534The only way this can happen is if you haven't installed the runtime library. To
535correct this, do:</p>
536
Misha Brukman4fbcf682008-12-17 18:11:40 +0000[diff] [blame]537<pre class="doc_code">
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]538% cd llvm/runtime
539% make clean ; make install-bytecode
540</pre>
541</div>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]542
543<div class="question">
544<p>
545How can I disable all optimizations when compiling code using the LLVM GCC front end?
546</p>
547</div>
548
549<div class="answer">
550<p>
551Passing "-Wa,-disable-opt -Wl,-disable-opt" will disable *all* cleanup and
552optimizations done at the llvm level, leaving you with the truly horrible
553code that you desire.
554</p>
555</div>
556
557
558<div class="question">
559<p>
Dan Gohman4dfac702008-11-24 17:18:39 +0000[diff] [blame]560<a name="translatecxx">Can I use LLVM to convert C++ code to C code?</a>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]561</p>
562</div>
563
564<div class="answer">
565<p>Yes, you can use LLVM to convert code from any language LLVM supports to C.
566Note that the generated C code will be very low level (all loops are lowered
567to gotos, etc) and not very pretty (comments are stripped, original source
568formatting is totally lost, variables are renamed, expressions are regrouped),
djga0237ec2009-01-25 16:04:50 +0000[diff] [blame]569so this may not be what you're looking for. Also, there are several
570limitations noted below.<p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]571
572<p>Use commands like this:</p>
573
574<ol>
Bill Wendling4e05864c2007-09-22 09:54:47 +0000[diff] [blame]575 <li><p>Compile your program as normal with llvm-g++:</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]576
Misha Brukman4fbcf682008-12-17 18:11:40 +0000[diff] [blame]577<pre class="doc_code">
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]578% llvm-g++ x.cpp -o program
579</pre>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]580
Bill Wendling4e05864c2007-09-22 09:54:47 +0000[diff] [blame]581 <p>or:</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]582
Misha Brukman4fbcf682008-12-17 18:11:40 +0000[diff] [blame]583<pre class="doc_code">
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]584% llvm-g++ a.cpp -c
585% llvm-g++ b.cpp -c
586% llvm-g++ a.o b.o -o program
587</pre>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]588
Bill Wendling4e05864c2007-09-22 09:54:47 +0000[diff] [blame]589 <p>With llvm-gcc3, this will generate program and program.bc. The .bc
590 file is the LLVM version of the program all linked together.</p></li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]591
Bill Wendling4e05864c2007-09-22 09:54:47 +0000[diff] [blame]592 <li><p>Convert the LLVM code to C code, using the LLC tool with the C
593 backend:</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]594
Misha Brukman4fbcf682008-12-17 18:11:40 +0000[diff] [blame]595<pre class="doc_code">
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]596% llc -march=c program.bc -o program.c
Misha Brukman4fbcf682008-12-17 18:11:40 +0000[diff] [blame]597</pre></li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]598
Bill Wendling4e05864c2007-09-22 09:54:47 +0000[diff] [blame]599<li><p>Finally, compile the C file:</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]600
Misha Brukman4fbcf682008-12-17 18:11:40 +0000[diff] [blame]601<pre class="doc_code">
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]602% cc x.c
Misha Brukman4fbcf682008-12-17 18:11:40 +0000[diff] [blame]603</pre></li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]604
605</ol>
606
djga0237ec2009-01-25 16:04:50 +0000[diff] [blame]607<p>Using LLVM does not eliminate the need for C++ library support.
608If you use the llvm-g++ front-end, the generated code will depend on
609g++'s C++ support libraries in the same way that code generated from
610g++ would. If you use another C++ front-end, the generated code will
611depend on whatever library that front-end would normally require.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]612
djga0237ec2009-01-25 16:04:50 +0000[diff] [blame]613<p>If you are working on a platform that does not provide any C++
614libraries, you may be able to manually compile libstdc++ to LLVM
Bill Wendling4e05864c2007-09-22 09:54:47 +0000[diff] [blame]615bitcode, statically link it into your program, then use the commands above to
djga0237ec2009-01-25 16:04:50 +0000[diff] [blame]616convert the whole result into C code. Alternatively, you might compile the
Bill Wendling4e05864c2007-09-22 09:54:47 +0000[diff] [blame]617libraries and your application into two different chunks of C code and link
618them.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]619
djga0237ec2009-01-25 16:04:50 +0000[diff] [blame]620<p>Note that, by default, the C back end does not support exception handling. If
621you want/need it for a certain program, you can enable it by passing
622"-enable-correct-eh-support" to the llc program. The resultant code will use
623setjmp/longjmp to implement exception support that is relatively slow, and
624not C++-ABI-conforming on most platforms, but otherwise correct.</p>
625
626<p>Also, there are a number of other limitations of the C backend that
627cause it to produce code that does not fully conform to the C++ ABI on
628most platforms. Some of the C++ programs in LLVM's test suite are known
629to fail when compiled with the C back end because of ABI incompatiblities
630with standard C++ libraries.</p>
631
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]632</div>
633
Dan Gohmanb8e23ab2009-02-10 17:26:53 +0000[diff] [blame^]634<div class="question">
635<p>
636<a name="platformindependent">Can I compile C or C++ code to platform-independent LLVM bitcode?</a>
637</p>
638</div>
639
640<div class="answer">
641
642<p>No. C and C++ are inherently platform-dependent languages. The most
643obvious example of this is the preprocessor. A very common way that C code
644is made portable is by using the preprocessor to include platform-specific
645code. In practice, information about other platforms is lost after
646preprocessing, so the result is inherently dependent on the platform that
647the preprocessing was targetting.</p>
648
649<p>Another example is <tt>sizeof</tt>. It's common for <tt>sizeof(long)</tt>
650to vary between platforms. In most C front-ends, <tt>sizeof</tt> is expanded
651to a constant immediately, thus hardwaring a platform-specific detail.</p>
652
653<p>Also, since many platforms define their ABIs in terms of C, and since
654LLVM is lower-level than C, front-ends currently must emit platform-specific
655IR in order to have the result conform to the platform ABI.</p>
656
657</div>
658
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]659<!-- *********************************************************************** -->
660<div class="doc_section">
661 <a name="cfe_code">Questions about code generated by the GCC front-end</a>
662</div>
663
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]664<div class="question">
665<a name="iosinit"></a>
666<p> What is this <tt>llvm.global_ctors</tt> and
667<tt>_GLOBAL__I__tmp_webcompile...</tt> stuff that happens when I #include
668&lt;iostream&gt;?</p>
669</div>
670
671<div class="answer">
672
673<p>If you #include the &lt;iostream&gt; header into a C++ translation unit, the
674file will probably use the <tt>std::cin</tt>/<tt>std::cout</tt>/... global
675objects. However, C++ does not guarantee an order of initialization between
676static objects in different translation units, so if a static ctor/dtor in your
677.cpp file used <tt>std::cout</tt>, for example, the object would not necessarily
678be automatically initialized before your use.</p>
679
680<p>To make <tt>std::cout</tt> and friends work correctly in these scenarios, the
681STL that we use declares a static object that gets created in every translation
682unit that includes <tt>&lt;iostream&gt;</tt>. This object has a static
683constructor and destructor that initializes and destroys the global iostream
684objects before they could possibly be used in the file. The code that you see
685in the .ll file corresponds to the constructor and destructor registration code.
686</p>
687
688<p>If you would like to make it easier to <b>understand</b> the LLVM code
689generated by the compiler in the demo page, consider using <tt>printf()</tt>
690instead of <tt>iostream</tt>s to print values.</p>
691
692</div>
693
694<!--=========================================================================-->
695
696<div class="question"><p>
697<a name="codedce"></a>
698Where did all of my code go??
699</p></div>
700
701<div class="answer">
702<p>
703If you are using the LLVM demo page, you may often wonder what happened to all
704of the code that you typed in. Remember that the demo script is running the
705code through the LLVM optimizers, so if your code doesn't actually do anything
706useful, it might all be deleted.
707</p>
708
709<p>
710To prevent this, make sure that the code is actually needed. For example, if
711you are computing some expression, return the value from the function instead of
712leaving it in a local variable. If you really want to constrain the optimizer,
713you can read from and assign to <tt>volatile</tt> global variables.
714</p>
715</div>
716
717<!--=========================================================================-->
718
719<div class="question"><p>
720<a name="undef"></a>
721<p>What is this "<tt>undef</tt>" thing that shows up in my code?
722</p></div>
723
724<div class="answer">
725<p>
726<a href="LangRef.html#undef"><tt>undef</tt></a> is the LLVM way of representing
727a value that is not defined. You can get these if you do not initialize a
728variable before you use it. For example, the C function:</p>
729
Misha Brukman4fbcf682008-12-17 18:11:40 +0000[diff] [blame]730<pre class="doc_code">
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]731int X() { int i; return i; }
732</pre>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]733
734<p>Is compiled to "<tt>ret i32 undef</tt>" because "<tt>i</tt>" never has
735a value specified for it.</p>
736</div>
737
738<!-- *********************************************************************** -->
739
740<hr>
741<address>
742 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
Misha Brukman947321d2008-12-11 17:34:48 +0000[diff] [blame]743 src="http://jigsaw.w3.org/css-validator/images/vcss-blue" alt="Valid CSS"></a>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]744 <a href="http://validator.w3.org/check/referer"><img
Misha Brukman947321d2008-12-11 17:34:48 +0000[diff] [blame]745 src="http://www.w3.org/Icons/valid-html401-blue" alt="Valid HTML 4.01"></a>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]746
747 <a href="http://llvm.org">LLVM Compiler Infrastructure</a><br>
748 Last modified: $Date$
749</address>
750
751</body>
752</html>