1================================
2Frequently Asked Questions (FAQ)
3================================
4
5.. contents::
6   :local:
7
8
9License
10=======
11
12Does the University of Illinois Open Source License really qualify as an "open source" license?
13-----------------------------------------------------------------------------------------------
14Yes, the license is `certified
15<http://www.opensource.org/licenses/UoI-NCSA.php>`_ by the Open Source
16Initiative (OSI).
17
18
19Can I modify LLVM source code and redistribute the modified source?
20-------------------------------------------------------------------
21Yes.  The modified source distribution must retain the copyright notice and
22follow the three bulletted conditions listed in the `LLVM license
23<http://llvm.org/svn/llvm-project/llvm/trunk/LICENSE.TXT>`_.
24
25
26Can I modify the LLVM source code and redistribute binaries or other tools based on it, without redistributing the source?
27--------------------------------------------------------------------------------------------------------------------------
28Yes. This is why we distribute LLVM under a less restrictive license than GPL,
29as explained in the first question above.
30
31
32Source Code
33===========
34
35In what language is LLVM written?
36---------------------------------
37All of the LLVM tools and libraries are written in C++ with extensive use of
38the STL.
39
40
41How portable is the LLVM source code?
42-------------------------------------
43The LLVM source code should be portable to most modern Unix-like operating
44systems.  Most of the code is written in standard C++ with operating system
45services abstracted to a support library.  The tools required to build and
46test LLVM have been ported to a plethora of platforms.
47
48Some porting problems may exist in the following areas:
49
50* The autoconf/makefile build system relies heavily on UNIX shell tools,
51  like the Bourne Shell and sed.  Porting to systems without these tools
52  (MacOS 9, Plan 9) will require more effort.
53
54What API do I use to store a value to one of the virtual registers in LLVM IR's SSA representation?
55---------------------------------------------------------------------------------------------------
56
57In short: you can't. It's actually kind of a silly question once you grok
58what's going on. Basically, in code like:
59
60.. code-block:: llvm
61
62    %result = add i32 %foo, %bar
63
64, ``%result`` is just a name given to the ``Value`` of the ``add``
65instruction. In other words, ``%result`` *is* the add instruction. The
66"assignment" doesn't explicitly "store" anything to any "virtual register";
67the "``=``" is more like the mathematical sense of equality.
68
69Longer explanation: In order to generate a textual representation of the
70IR, some kind of name has to be given to each instruction so that other
71instructions can textually reference it. However, the isomorphic in-memory
72representation that you manipulate from C++ has no such restriction since
73instructions can simply keep pointers to any other ``Value``'s that they
74reference. In fact, the names of dummy numbered temporaries like ``%1`` are
75not explicitly represented in the in-memory representation at all (see
76``Value::getName()``).
77
78Build Problems
79==============
80
81When I run configure, it finds the wrong C compiler.
82----------------------------------------------------
83The ``configure`` script attempts to locate first ``gcc`` and then ``cc``,
84unless it finds compiler paths set in ``CC`` and ``CXX`` for the C and C++
85compiler, respectively.
86
87If ``configure`` finds the wrong compiler, either adjust your ``PATH``
88environment variable or set ``CC`` and ``CXX`` explicitly.
89
90
91The ``configure`` script finds the right C compiler, but it uses the LLVM tools from a previous build.  What do I do?
92---------------------------------------------------------------------------------------------------------------------
93The ``configure`` script uses the ``PATH`` to find executables, so if it's
94grabbing the wrong linker/assembler/etc, there are two ways to fix it:
95
96#. Adjust your ``PATH`` environment variable so that the correct program
97   appears first in the ``PATH``.  This may work, but may not be convenient
98   when you want them *first* in your path for other work.
99
100#. Run ``configure`` with an alternative ``PATH`` that is correct. In a
101   Bourne compatible shell, the syntax would be:
102
103.. code-block:: console
104
105   % PATH=[the path without the bad program] $LLVM_SRC_DIR/configure ...
106
107This is still somewhat inconvenient, but it allows ``configure`` to do its
108work without having to adjust your ``PATH`` permanently.
109
110
111When creating a dynamic library, I get a strange GLIBC error.
112-------------------------------------------------------------
113Under some operating systems (i.e. Linux), libtool does not work correctly if
114GCC was compiled with the ``--disable-shared option``.  To work around this,
115install your own version of GCC that has shared libraries enabled by default.
116
117
118I've updated my source tree from Subversion, and now my build is trying to use a file/directory that doesn't exist.
119-------------------------------------------------------------------------------------------------------------------
120You need to re-run configure in your object directory.  When new Makefiles
121are added to the source tree, they have to be copied over to the object tree
122in order to be used by the build.
123
124
125I've modified a Makefile in my source tree, but my build tree keeps using the old version.  What do I do?
126---------------------------------------------------------------------------------------------------------
127If the Makefile already exists in your object tree, you can just run the
128following command in the top level directory of your object tree:
129
130.. code-block:: console
131
132   % ./config.status <relative path to Makefile>;
133
134If the Makefile is new, you will have to modify the configure script to copy
135it over.
136
137
138I've upgraded to a new version of LLVM, and I get strange build errors.
139-----------------------------------------------------------------------
140Sometimes, changes to the LLVM source code alters how the build system works.
141Changes in ``libtool``, ``autoconf``, or header file dependencies are
142especially prone to this sort of problem.
143
144The best thing to try is to remove the old files and re-build.  In most cases,
145this takes care of the problem.  To do this, just type ``make clean`` and then
146``make`` in the directory that fails to build.
147
148
149I've built LLVM and am testing it, but the tests freeze.
150--------------------------------------------------------
151This is most likely occurring because you built a profile or release
152(optimized) build of LLVM and have not specified the same information on the
153``gmake`` command line.
154
155For example, if you built LLVM with the command:
156
157.. code-block:: console
158
159   % gmake ENABLE_PROFILING=1
160
161...then you must run the tests with the following commands:
162
163.. code-block:: console
164
165   % cd llvm/test
166   % gmake ENABLE_PROFILING=1
167
168Why do test results differ when I perform different types of builds?
169--------------------------------------------------------------------
170The LLVM test suite is dependent upon several features of the LLVM tools and
171libraries.
172
173First, the debugging assertions in code are not enabled in optimized or
174profiling builds.  Hence, tests that used to fail may pass.
175
176Second, some tests may rely upon debugging options or behavior that is only
177available in the debug build.  These tests will fail in an optimized or
178profile build.
179
180
181Compiling LLVM with GCC 3.3.2 fails, what should I do?
182------------------------------------------------------
183This is `a bug in GCC <http://gcc.gnu.org/bugzilla/show_bug.cgi?id=13392>`_,
184and affects projects other than LLVM.  Try upgrading or downgrading your GCC.
185
186
187After Subversion update, rebuilding gives the error "No rule to make target".
188-----------------------------------------------------------------------------
189If the error is of the form:
190
191.. code-block:: console
192
193   gmake[2]: *** No rule to make target `/path/to/somefile',
194                 needed by `/path/to/another/file.d'.
195   Stop.
196
197This may occur anytime files are moved within the Subversion repository or
198removed entirely.  In this case, the best solution is to erase all ``.d``
199files, which list dependencies for source files, and rebuild:
200
201.. code-block:: console
202
203   % cd $LLVM_OBJ_DIR
204   % rm -f `find . -name \*\.d`
205   % gmake
206
207In other cases, it may be necessary to run ``make clean`` before rebuilding.
208
209
210Source Languages
211================
212
213What source languages are supported?
214------------------------------------
215LLVM currently has full support for C and C++ source languages. These are
216available through both `Clang <http://clang.llvm.org/>`_ and `DragonEgg
217<http://dragonegg.llvm.org/>`_.
218
219The PyPy developers are working on integrating LLVM into the PyPy backend so
220that PyPy language can translate to LLVM.
221
222
223I'd like to write a self-hosting LLVM compiler. How should I interface with the LLVM middle-end optimizers and back-end code generators?
224----------------------------------------------------------------------------------------------------------------------------------------
225Your compiler front-end will communicate with LLVM by creating a module in the
226LLVM intermediate representation (IR) format. Assuming you want to write your
227language's compiler in the language itself (rather than C++), there are 3
228major ways to tackle generating LLVM IR from a front-end:
229
2301. **Call into the LLVM libraries code using your language's FFI (foreign
231   function interface).**
232
233  * *for:* best tracks changes to the LLVM IR, .ll syntax, and .bc format
234
235  * *for:* enables running LLVM optimization passes without a emit/parse
236    overhead
237
238  * *for:* adapts well to a JIT context
239
240  * *against:* lots of ugly glue code to write
241
2422. **Emit LLVM assembly from your compiler's native language.**
243
244  * *for:* very straightforward to get started
245
246  * *against:* the .ll parser is slower than the bitcode reader when
247    interfacing to the middle end
248
249  * *against:* it may be harder to track changes to the IR
250
2513. **Emit LLVM bitcode from your compiler's native language.**
252
253  * *for:* can use the more-efficient bitcode reader when interfacing to the
254    middle end
255
256  * *against:* you'll have to re-engineer the LLVM IR object model and bitcode
257    writer in your language
258
259  * *against:* it may be harder to track changes to the IR
260
261If you go with the first option, the C bindings in include/llvm-c should help
262a lot, since most languages have strong support for interfacing with C. The
263most common hurdle with calling C from managed code is interfacing with the
264garbage collector. The C interface was designed to require very little memory
265management, and so is straightforward in this regard.
266
267What support is there for a higher level source language constructs for building a compiler?
268--------------------------------------------------------------------------------------------
269Currently, there isn't much. LLVM supports an intermediate representation
270which is useful for code representation but will not support the high level
271(abstract syntax tree) representation needed by most compilers. There are no
272facilities for lexical nor semantic analysis.
273
274
275I don't understand the ``GetElementPtr`` instruction. Help!
276-----------------------------------------------------------
277See `The Often Misunderstood GEP Instruction <GetElementPtr.html>`_.
278
279
280Using the C and C++ Front Ends
281==============================
282
283Can I compile C or C++ code to platform-independent LLVM bitcode?
284-----------------------------------------------------------------
285No. C and C++ are inherently platform-dependent languages. The most obvious
286example of this is the preprocessor. A very common way that C code is made
287portable is by using the preprocessor to include platform-specific code. In
288practice, information about other platforms is lost after preprocessing, so
289the result is inherently dependent on the platform that the preprocessing was
290targeting.
291
292Another example is ``sizeof``. It's common for ``sizeof(long)`` to vary
293between platforms. In most C front-ends, ``sizeof`` is expanded to a
294constant immediately, thus hard-wiring a platform-specific detail.
295
296Also, since many platforms define their ABIs in terms of C, and since LLVM is
297lower-level than C, front-ends currently must emit platform-specific IR in
298order to have the result conform to the platform ABI.
299
300
301Questions about code generated by the demo page
302===============================================
303
304What is this ``llvm.global_ctors`` and ``_GLOBAL__I_a...`` stuff that happens when I ``#include <iostream>``?
305-------------------------------------------------------------------------------------------------------------
306If you ``#include`` the ``<iostream>`` header into a C++ translation unit,
307the file will probably use the ``std::cin``/``std::cout``/... global objects.
308However, C++ does not guarantee an order of initialization between static
309objects in different translation units, so if a static ctor/dtor in your .cpp
310file used ``std::cout``, for example, the object would not necessarily be
311automatically initialized before your use.
312
313To make ``std::cout`` and friends work correctly in these scenarios, the STL
314that we use declares a static object that gets created in every translation
315unit that includes ``<iostream>``.  This object has a static constructor
316and destructor that initializes and destroys the global iostream objects
317before they could possibly be used in the file.  The code that you see in the
318``.ll`` file corresponds to the constructor and destructor registration code.
319
320If you would like to make it easier to *understand* the LLVM code generated
321by the compiler in the demo page, consider using ``printf()`` instead of
322``iostream``\s to print values.
323
324
325Where did all of my code go??
326-----------------------------
327If you are using the LLVM demo page, you may often wonder what happened to
328all of the code that you typed in.  Remember that the demo script is running
329the code through the LLVM optimizers, so if your code doesn't actually do
330anything useful, it might all be deleted.
331
332To prevent this, make sure that the code is actually needed.  For example, if
333you are computing some expression, return the value from the function instead
334of leaving it in a local variable.  If you really want to constrain the
335optimizer, you can read from and assign to ``volatile`` global variables.
336
337
338What is this "``undef``" thing that shows up in my code?
339--------------------------------------------------------
340``undef`` is the LLVM way of representing a value that is not defined.  You
341can get these if you do not initialize a variable before you use it.  For
342example, the C function:
343
344.. code-block:: c
345
346   int X() { int i; return i; }
347
348Is compiled to "``ret i32 undef``" because "``i``" never has a value specified
349for it.
350
351
352Why does instcombine + simplifycfg turn a call to a function with a mismatched calling convention into "unreachable"? Why not make the verifier reject it?
353----------------------------------------------------------------------------------------------------------------------------------------------------------
354This is a common problem run into by authors of front-ends that are using
355custom calling conventions: you need to make sure to set the right calling
356convention on both the function and on each call to the function.  For
357example, this code:
358
359.. code-block:: llvm
360
361   define fastcc void @foo() {
362       ret void
363   }
364   define void @bar() {
365       call void @foo()
366       ret void
367   }
368
369Is optimized to:
370
371.. code-block:: llvm
372
373   define fastcc void @foo() {
374       ret void
375   }
376   define void @bar() {
377       unreachable
378   }
379
380... with "``opt -instcombine -simplifycfg``".  This often bites people because
381"all their code disappears".  Setting the calling convention on the caller and
382callee is required for indirect calls to work, so people often ask why not
383make the verifier reject this sort of thing.
384
385The answer is that this code has undefined behavior, but it is not illegal.
386If we made it illegal, then every transformation that could potentially create
387this would have to ensure that it doesn't, and there is valid code that can
388create this sort of construct (in dead code).  The sorts of things that can
389cause this to happen are fairly contrived, but we still need to accept them.
390Here's an example:
391
392.. code-block:: llvm
393
394   define fastcc void @foo() {
395       ret void
396   }
397   define internal void @bar(void()* %FP, i1 %cond) {
398       br i1 %cond, label %T, label %F
399   T:
400       call void %FP()
401       ret void
402   F:
403       call fastcc void %FP()
404       ret void
405   }
406   define void @test() {
407       %X = or i1 false, false
408       call void @bar(void()* @foo, i1 %X)
409       ret void
410   }
411
412In this example, "test" always passes ``@foo``/``false`` into ``bar``, which
413ensures that it is dynamically called with the right calling conv (thus, the
414code is perfectly well defined).  If you run this through the inliner, you
415get this (the explicit "or" is there so that the inliner doesn't dead code
416eliminate a bunch of stuff):
417
418.. code-block:: llvm
419
420   define fastcc void @foo() {
421       ret void
422   }
423   define void @test() {
424       %X = or i1 false, false
425       br i1 %X, label %T.i, label %F.i
426   T.i:
427       call void @foo()
428       br label %bar.exit
429   F.i:
430       call fastcc void @foo()
431       br label %bar.exit
432   bar.exit:
433       ret void
434   }
435
436Here you can see that the inlining pass made an undefined call to ``@foo``
437with the wrong calling convention.  We really don't want to make the inliner
438have to know about this sort of thing, so it needs to be valid code.  In this
439case, dead code elimination can trivially remove the undefined code.  However,
440if ``%X`` was an input argument to ``@test``, the inliner would produce this:
441
442.. code-block:: llvm
443
444   define fastcc void @foo() {
445       ret void
446   }
447
448   define void @test(i1 %X) {
449       br i1 %X, label %T.i, label %F.i
450   T.i:
451       call void @foo()
452       br label %bar.exit
453   F.i:
454       call fastcc void @foo()
455       br label %bar.exit
456   bar.exit:
457       ret void
458   }
459
460The interesting thing about this is that ``%X`` *must* be false for the
461code to be well-defined, but no amount of dead code elimination will be able
462to delete the broken call as unreachable.  However, since
463``instcombine``/``simplifycfg`` turns the undefined call into unreachable, we
464end up with a branch on a condition that goes to unreachable: a branch to
465unreachable can never happen, so "``-inline -instcombine -simplifycfg``" is
466able to produce:
467
468.. code-block:: llvm
469
470   define fastcc void @foo() {
471      ret void
472   }
473   define void @test(i1 %X) {
474   F.i:
475      call fastcc void @foo()
476      ret void
477   }
478