zpu: wip on DAG to DAG
[llvm/zpu.git] / docs / ProgrammersManual.html
blobe8d81a25061daa2db8d873767113226f2c1b5b2e
1 <!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01//EN"
2 "http://www.w3.org/TR/html4/strict.dtd">
3 <html>
4 <head>
5 <meta http-equiv="Content-type" content="text/html;charset=UTF-8">
6 <title>LLVM Programmer's Manual</title>
7 <link rel="stylesheet" href="llvm.css" type="text/css">
8 </head>
9 <body>
11 <div class="doc_title">
12 LLVM Programmer's Manual
13 </div>
15 <ol>
16 <li><a href="#introduction">Introduction</a></li>
17 <li><a href="#general">General Information</a>
18 <ul>
19 <li><a href="#stl">The C++ Standard Template Library</a></li>
20 <!--
21 <li>The <tt>-time-passes</tt> option</li>
22 <li>How to use the LLVM Makefile system</li>
23 <li>How to write a regression test</li>
25 -->
26 </ul>
27 </li>
28 <li><a href="#apis">Important and useful LLVM APIs</a>
29 <ul>
30 <li><a href="#isa">The <tt>isa&lt;&gt;</tt>, <tt>cast&lt;&gt;</tt>
31 and <tt>dyn_cast&lt;&gt;</tt> templates</a> </li>
32 <li><a href="#string_apis">Passing strings (the <tt>StringRef</tt>
33 and <tt>Twine</tt> classes)</a>
34 <ul>
35 <li><a href="#StringRef">The <tt>StringRef</tt> class</a> </li>
36 <li><a href="#Twine">The <tt>Twine</tt> class</a> </li>
37 </ul>
38 </li>
39 <li><a href="#DEBUG">The <tt>DEBUG()</tt> macro and <tt>-debug</tt>
40 option</a>
41 <ul>
42 <li><a href="#DEBUG_TYPE">Fine grained debug info with <tt>DEBUG_TYPE</tt>
43 and the <tt>-debug-only</tt> option</a> </li>
44 </ul>
45 </li>
46 <li><a href="#Statistic">The <tt>Statistic</tt> class &amp; <tt>-stats</tt>
47 option</a></li>
48 <!--
49 <li>The <tt>InstVisitor</tt> template
50 <li>The general graph API
51 -->
52 <li><a href="#ViewGraph">Viewing graphs while debugging code</a></li>
53 </ul>
54 </li>
55 <li><a href="#datastructure">Picking the Right Data Structure for a Task</a>
56 <ul>
57 <li><a href="#ds_sequential">Sequential Containers (std::vector, std::list, etc)</a>
58 <ul>
59 <li><a href="#dss_fixedarrays">Fixed Size Arrays</a></li>
60 <li><a href="#dss_heaparrays">Heap Allocated Arrays</a></li>
61 <li><a href="#dss_smallvector">"llvm/ADT/SmallVector.h"</a></li>
62 <li><a href="#dss_vector">&lt;vector&gt;</a></li>
63 <li><a href="#dss_deque">&lt;deque&gt;</a></li>
64 <li><a href="#dss_list">&lt;list&gt;</a></li>
65 <li><a href="#dss_ilist">llvm/ADT/ilist.h</a></li>
66 <li><a href="#dss_other">Other Sequential Container Options</a></li>
67 </ul></li>
68 <li><a href="#ds_set">Set-Like Containers (std::set, SmallSet, SetVector, etc)</a>
69 <ul>
70 <li><a href="#dss_sortedvectorset">A sorted 'vector'</a></li>
71 <li><a href="#dss_smallset">"llvm/ADT/SmallSet.h"</a></li>
72 <li><a href="#dss_smallptrset">"llvm/ADT/SmallPtrSet.h"</a></li>
73 <li><a href="#dss_denseset">"llvm/ADT/DenseSet.h"</a></li>
74 <li><a href="#dss_FoldingSet">"llvm/ADT/FoldingSet.h"</a></li>
75 <li><a href="#dss_set">&lt;set&gt;</a></li>
76 <li><a href="#dss_setvector">"llvm/ADT/SetVector.h"</a></li>
77 <li><a href="#dss_uniquevector">"llvm/ADT/UniqueVector.h"</a></li>
78 <li><a href="#dss_otherset">Other Set-Like ContainerOptions</a></li>
79 </ul></li>
80 <li><a href="#ds_map">Map-Like Containers (std::map, DenseMap, etc)</a>
81 <ul>
82 <li><a href="#dss_sortedvectormap">A sorted 'vector'</a></li>
83 <li><a href="#dss_stringmap">"llvm/ADT/StringMap.h"</a></li>
84 <li><a href="#dss_indexedmap">"llvm/ADT/IndexedMap.h"</a></li>
85 <li><a href="#dss_densemap">"llvm/ADT/DenseMap.h"</a></li>
86 <li><a href="#dss_valuemap">"llvm/ADT/ValueMap.h"</a></li>
87 <li><a href="#dss_map">&lt;map&gt;</a></li>
88 <li><a href="#dss_othermap">Other Map-Like Container Options</a></li>
89 </ul></li>
90 <li><a href="#ds_string">String-like containers</a>
91 <!--<ul>
92 todo
93 </ul>--></li>
94 <li><a href="#ds_bit">BitVector-like containers</a>
95 <ul>
96 <li><a href="#dss_bitvector">A dense bitvector</a></li>
97 <li><a href="#dss_smallbitvector">A "small" dense bitvector</a></li>
98 <li><a href="#dss_sparsebitvector">A sparse bitvector</a></li>
99 </ul></li>
100 </ul>
101 </li>
102 <li><a href="#common">Helpful Hints for Common Operations</a>
103 <ul>
104 <li><a href="#inspection">Basic Inspection and Traversal Routines</a>
105 <ul>
106 <li><a href="#iterate_function">Iterating over the <tt>BasicBlock</tt>s
107 in a <tt>Function</tt></a> </li>
108 <li><a href="#iterate_basicblock">Iterating over the <tt>Instruction</tt>s
109 in a <tt>BasicBlock</tt></a> </li>
110 <li><a href="#iterate_institer">Iterating over the <tt>Instruction</tt>s
111 in a <tt>Function</tt></a> </li>
112 <li><a href="#iterate_convert">Turning an iterator into a
113 class pointer</a> </li>
114 <li><a href="#iterate_complex">Finding call sites: a more
115 complex example</a> </li>
116 <li><a href="#calls_and_invokes">Treating calls and invokes
117 the same way</a> </li>
118 <li><a href="#iterate_chains">Iterating over def-use &amp;
119 use-def chains</a> </li>
120 <li><a href="#iterate_preds">Iterating over predecessors &amp;
121 successors of blocks</a></li>
122 </ul>
123 </li>
124 <li><a href="#simplechanges">Making simple changes</a>
125 <ul>
126 <li><a href="#schanges_creating">Creating and inserting new
127 <tt>Instruction</tt>s</a> </li>
128 <li><a href="#schanges_deleting">Deleting <tt>Instruction</tt>s</a> </li>
129 <li><a href="#schanges_replacing">Replacing an <tt>Instruction</tt>
130 with another <tt>Value</tt></a> </li>
131 <li><a href="#schanges_deletingGV">Deleting <tt>GlobalVariable</tt>s</a> </li>
132 </ul>
133 </li>
134 <li><a href="#create_types">How to Create Types</a></li>
135 <!--
136 <li>Working with the Control Flow Graph
137 <ul>
138 <li>Accessing predecessors and successors of a <tt>BasicBlock</tt>
139 <li>
140 <li>
141 </ul>
142 -->
143 </ul>
144 </li>
146 <li><a href="#threading">Threads and LLVM</a>
147 <ul>
148 <li><a href="#startmultithreaded">Entering and Exiting Multithreaded Mode
149 </a></li>
150 <li><a href="#shutdown">Ending execution with <tt>llvm_shutdown()</tt></a></li>
151 <li><a href="#managedstatic">Lazy initialization with <tt>ManagedStatic</tt></a></li>
152 <li><a href="#llvmcontext">Achieving Isolation with <tt>LLVMContext</tt></a></li>
153 <li><a href="#jitthreading">Threads and the JIT</a></li>
154 </ul>
155 </li>
157 <li><a href="#advanced">Advanced Topics</a>
158 <ul>
159 <li><a href="#TypeResolve">LLVM Type Resolution</a>
160 <ul>
161 <li><a href="#BuildRecType">Basic Recursive Type Construction</a></li>
162 <li><a href="#refineAbstractTypeTo">The <tt>refineAbstractTypeTo</tt> method</a></li>
163 <li><a href="#PATypeHolder">The PATypeHolder Class</a></li>
164 <li><a href="#AbstractTypeUser">The AbstractTypeUser Class</a></li>
165 </ul></li>
167 <li><a href="#SymbolTable">The <tt>ValueSymbolTable</tt> and <tt>TypeSymbolTable</tt> classes</a></li>
168 <li><a href="#UserLayout">The <tt>User</tt> and owned <tt>Use</tt> classes' memory layout</a></li>
169 </ul></li>
171 <li><a href="#coreclasses">The Core LLVM Class Hierarchy Reference</a>
172 <ul>
173 <li><a href="#Type">The <tt>Type</tt> class</a> </li>
174 <li><a href="#Module">The <tt>Module</tt> class</a></li>
175 <li><a href="#Value">The <tt>Value</tt> class</a>
176 <ul>
177 <li><a href="#User">The <tt>User</tt> class</a>
178 <ul>
179 <li><a href="#Instruction">The <tt>Instruction</tt> class</a></li>
180 <li><a href="#Constant">The <tt>Constant</tt> class</a>
181 <ul>
182 <li><a href="#GlobalValue">The <tt>GlobalValue</tt> class</a>
183 <ul>
184 <li><a href="#Function">The <tt>Function</tt> class</a></li>
185 <li><a href="#GlobalVariable">The <tt>GlobalVariable</tt> class</a></li>
186 </ul>
187 </li>
188 </ul>
189 </li>
190 </ul>
191 </li>
192 <li><a href="#BasicBlock">The <tt>BasicBlock</tt> class</a></li>
193 <li><a href="#Argument">The <tt>Argument</tt> class</a></li>
194 </ul>
195 </li>
196 </ul>
197 </li>
198 </ol>
200 <div class="doc_author">
201 <p>Written by <a href="mailto:sabre@nondot.org">Chris Lattner</a>,
202 <a href="mailto:dhurjati@cs.uiuc.edu">Dinakar Dhurjati</a>,
203 <a href="mailto:ggreif@gmail.com">Gabor Greif</a>,
204 <a href="mailto:jstanley@cs.uiuc.edu">Joel Stanley</a>,
205 <a href="mailto:rspencer@x10sys.com">Reid Spencer</a> and
206 <a href="mailto:owen@apple.com">Owen Anderson</a></p>
207 </div>
209 <!-- *********************************************************************** -->
210 <div class="doc_section">
211 <a name="introduction">Introduction </a>
212 </div>
213 <!-- *********************************************************************** -->
215 <div class="doc_text">
217 <p>This document is meant to highlight some of the important classes and
218 interfaces available in the LLVM source-base. This manual is not
219 intended to explain what LLVM is, how it works, and what LLVM code looks
220 like. It assumes that you know the basics of LLVM and are interested
221 in writing transformations or otherwise analyzing or manipulating the
222 code.</p>
224 <p>This document should get you oriented so that you can find your
225 way in the continuously growing source code that makes up the LLVM
226 infrastructure. Note that this manual is not intended to serve as a
227 replacement for reading the source code, so if you think there should be
228 a method in one of these classes to do something, but it's not listed,
229 check the source. Links to the <a href="/doxygen/">doxygen</a> sources
230 are provided to make this as easy as possible.</p>
232 <p>The first section of this document describes general information that is
233 useful to know when working in the LLVM infrastructure, and the second describes
234 the Core LLVM classes. In the future this manual will be extended with
235 information describing how to use extension libraries, such as dominator
236 information, CFG traversal routines, and useful utilities like the <tt><a
237 href="/doxygen/InstVisitor_8h-source.html">InstVisitor</a></tt> template.</p>
239 </div>
241 <!-- *********************************************************************** -->
242 <div class="doc_section">
243 <a name="general">General Information</a>
244 </div>
245 <!-- *********************************************************************** -->
247 <div class="doc_text">
249 <p>This section contains general information that is useful if you are working
250 in the LLVM source-base, but that isn't specific to any particular API.</p>
252 </div>
254 <!-- ======================================================================= -->
255 <div class="doc_subsection">
256 <a name="stl">The C++ Standard Template Library</a>
257 </div>
259 <div class="doc_text">
261 <p>LLVM makes heavy use of the C++ Standard Template Library (STL),
262 perhaps much more than you are used to, or have seen before. Because of
263 this, you might want to do a little background reading in the
264 techniques used and capabilities of the library. There are many good
265 pages that discuss the STL, and several books on the subject that you
266 can get, so it will not be discussed in this document.</p>
268 <p>Here are some useful links:</p>
270 <ol>
272 <li><a href="http://www.dinkumware.com/manuals/#Standard C++ Library">Dinkumware
273 C++ Library reference</a> - an excellent reference for the STL and other parts
274 of the standard C++ library.</li>
276 <li><a href="http://www.tempest-sw.com/cpp/">C++ In a Nutshell</a> - This is an
277 O'Reilly book in the making. It has a decent Standard Library
278 Reference that rivals Dinkumware's, and is unfortunately no longer free since the
279 book has been published.</li>
281 <li><a href="http://www.parashift.com/c++-faq-lite/">C++ Frequently Asked
282 Questions</a></li>
284 <li><a href="http://www.sgi.com/tech/stl/">SGI's STL Programmer's Guide</a> -
285 Contains a useful <a
286 href="http://www.sgi.com/tech/stl/stl_introduction.html">Introduction to the
287 STL</a>.</li>
289 <li><a href="http://www.research.att.com/%7Ebs/C++.html">Bjarne Stroustrup's C++
290 Page</a></li>
292 <li><a href="http://64.78.49.204/">
293 Bruce Eckel's Thinking in C++, 2nd ed. Volume 2 Revision 4.0 (even better, get
294 the book).</a></li>
296 </ol>
298 <p>You are also encouraged to take a look at the <a
299 href="CodingStandards.html">LLVM Coding Standards</a> guide which focuses on how
300 to write maintainable code more than where to put your curly braces.</p>
302 </div>
304 <!-- ======================================================================= -->
305 <div class="doc_subsection">
306 <a name="stl">Other useful references</a>
307 </div>
309 <div class="doc_text">
311 <ol>
312 <li><a href="http://www.fortran-2000.com/ArnaudRecipes/sharedlib.html">Using
313 static and shared libraries across platforms</a></li>
314 </ol>
316 </div>
318 <!-- *********************************************************************** -->
319 <div class="doc_section">
320 <a name="apis">Important and useful LLVM APIs</a>
321 </div>
322 <!-- *********************************************************************** -->
324 <div class="doc_text">
326 <p>Here we highlight some LLVM APIs that are generally useful and good to
327 know about when writing transformations.</p>
329 </div>
331 <!-- ======================================================================= -->
332 <div class="doc_subsection">
333 <a name="isa">The <tt>isa&lt;&gt;</tt>, <tt>cast&lt;&gt;</tt> and
334 <tt>dyn_cast&lt;&gt;</tt> templates</a>
335 </div>
337 <div class="doc_text">
339 <p>The LLVM source-base makes extensive use of a custom form of RTTI.
340 These templates have many similarities to the C++ <tt>dynamic_cast&lt;&gt;</tt>
341 operator, but they don't have some drawbacks (primarily stemming from
342 the fact that <tt>dynamic_cast&lt;&gt;</tt> only works on classes that
343 have a v-table). Because they are used so often, you must know what they
344 do and how they work. All of these templates are defined in the <a
345 href="/doxygen/Casting_8h-source.html"><tt>llvm/Support/Casting.h</tt></a>
346 file (note that you very rarely have to include this file directly).</p>
348 <dl>
349 <dt><tt>isa&lt;&gt;</tt>: </dt>
351 <dd><p>The <tt>isa&lt;&gt;</tt> operator works exactly like the Java
352 "<tt>instanceof</tt>" operator. It returns true or false depending on whether
353 a reference or pointer points to an instance of the specified class. This can
354 be very useful for constraint checking of various sorts (example below).</p>
355 </dd>
357 <dt><tt>cast&lt;&gt;</tt>: </dt>
359 <dd><p>The <tt>cast&lt;&gt;</tt> operator is a "checked cast" operation. It
360 converts a pointer or reference from a base class to a derived class, causing
361 an assertion failure if it is not really an instance of the right type. This
362 should be used in cases where you have some information that makes you believe
363 that something is of the right type. An example of the <tt>isa&lt;&gt;</tt>
364 and <tt>cast&lt;&gt;</tt> template is:</p>
366 <div class="doc_code">
367 <pre>
368 static bool isLoopInvariant(const <a href="#Value">Value</a> *V, const Loop *L) {
369 if (isa&lt;<a href="#Constant">Constant</a>&gt;(V) || isa&lt;<a href="#Argument">Argument</a>&gt;(V) || isa&lt;<a href="#GlobalValue">GlobalValue</a>&gt;(V))
370 return true;
372 // <i>Otherwise, it must be an instruction...</i>
373 return !L-&gt;contains(cast&lt;<a href="#Instruction">Instruction</a>&gt;(V)-&gt;getParent());
375 </pre>
376 </div>
378 <p>Note that you should <b>not</b> use an <tt>isa&lt;&gt;</tt> test followed
379 by a <tt>cast&lt;&gt;</tt>, for that use the <tt>dyn_cast&lt;&gt;</tt>
380 operator.</p>
382 </dd>
384 <dt><tt>dyn_cast&lt;&gt;</tt>:</dt>
386 <dd><p>The <tt>dyn_cast&lt;&gt;</tt> operator is a "checking cast" operation.
387 It checks to see if the operand is of the specified type, and if so, returns a
388 pointer to it (this operator does not work with references). If the operand is
389 not of the correct type, a null pointer is returned. Thus, this works very
390 much like the <tt>dynamic_cast&lt;&gt;</tt> operator in C++, and should be
391 used in the same circumstances. Typically, the <tt>dyn_cast&lt;&gt;</tt>
392 operator is used in an <tt>if</tt> statement or some other flow control
393 statement like this:</p>
395 <div class="doc_code">
396 <pre>
397 if (<a href="#AllocationInst">AllocationInst</a> *AI = dyn_cast&lt;<a href="#AllocationInst">AllocationInst</a>&gt;(Val)) {
398 // <i>...</i>
400 </pre>
401 </div>
403 <p>This form of the <tt>if</tt> statement effectively combines together a call
404 to <tt>isa&lt;&gt;</tt> and a call to <tt>cast&lt;&gt;</tt> into one
405 statement, which is very convenient.</p>
407 <p>Note that the <tt>dyn_cast&lt;&gt;</tt> operator, like C++'s
408 <tt>dynamic_cast&lt;&gt;</tt> or Java's <tt>instanceof</tt> operator, can be
409 abused. In particular, you should not use big chained <tt>if/then/else</tt>
410 blocks to check for lots of different variants of classes. If you find
411 yourself wanting to do this, it is much cleaner and more efficient to use the
412 <tt>InstVisitor</tt> class to dispatch over the instruction type directly.</p>
414 </dd>
416 <dt><tt>cast_or_null&lt;&gt;</tt>: </dt>
418 <dd><p>The <tt>cast_or_null&lt;&gt;</tt> operator works just like the
419 <tt>cast&lt;&gt;</tt> operator, except that it allows for a null pointer as an
420 argument (which it then propagates). This can sometimes be useful, allowing
421 you to combine several null checks into one.</p></dd>
423 <dt><tt>dyn_cast_or_null&lt;&gt;</tt>: </dt>
425 <dd><p>The <tt>dyn_cast_or_null&lt;&gt;</tt> operator works just like the
426 <tt>dyn_cast&lt;&gt;</tt> operator, except that it allows for a null pointer
427 as an argument (which it then propagates). This can sometimes be useful,
428 allowing you to combine several null checks into one.</p></dd>
430 </dl>
432 <p>These five templates can be used with any classes, whether they have a
433 v-table or not. To add support for these templates, you simply need to add
434 <tt>classof</tt> static methods to the class you are interested casting
435 to. Describing this is currently outside the scope of this document, but there
436 are lots of examples in the LLVM source base.</p>
438 </div>
441 <!-- ======================================================================= -->
442 <div class="doc_subsection">
443 <a name="string_apis">Passing strings (the <tt>StringRef</tt>
444 and <tt>Twine</tt> classes)</a>
445 </div>
447 <div class="doc_text">
449 <p>Although LLVM generally does not do much string manipulation, we do have
450 several important APIs which take strings. Two important examples are the
451 Value class -- which has names for instructions, functions, etc. -- and the
452 StringMap class which is used extensively in LLVM and Clang.</p>
454 <p>These are generic classes, and they need to be able to accept strings which
455 may have embedded null characters. Therefore, they cannot simply take
456 a <tt>const char *</tt>, and taking a <tt>const std::string&amp;</tt> requires
457 clients to perform a heap allocation which is usually unnecessary. Instead,
458 many LLVM APIs use a <tt>StringRef</tt> or a <tt>const Twine&amp;</tt> for
459 passing strings efficiently.</p>
461 </div>
463 <!-- _______________________________________________________________________ -->
464 <div class="doc_subsubsection">
465 <a name="StringRef">The <tt>StringRef</tt> class</a>
466 </div>
468 <div class="doc_text">
470 <p>The <tt>StringRef</tt> data type represents a reference to a constant string
471 (a character array and a length) and supports the common operations available
472 on <tt>std:string</tt>, but does not require heap allocation.</p>
474 <p>It can be implicitly constructed using a C style null-terminated string,
475 an <tt>std::string</tt>, or explicitly with a character pointer and length.
476 For example, the <tt>StringRef</tt> find function is declared as:</p>
478 <pre class="doc_code">
479 iterator find(StringRef Key);
480 </pre>
482 <p>and clients can call it using any one of:</p>
484 <pre class="doc_code">
485 Map.find("foo"); <i>// Lookup "foo"</i>
486 Map.find(std::string("bar")); <i>// Lookup "bar"</i>
487 Map.find(StringRef("\0baz", 4)); <i>// Lookup "\0baz"</i>
488 </pre>
490 <p>Similarly, APIs which need to return a string may return a <tt>StringRef</tt>
491 instance, which can be used directly or converted to an <tt>std::string</tt>
492 using the <tt>str</tt> member function. See
493 "<tt><a href="/doxygen/classllvm_1_1StringRef_8h-source.html">llvm/ADT/StringRef.h</a></tt>"
494 for more information.</p>
496 <p>You should rarely use the <tt>StringRef</tt> class directly, because it contains
497 pointers to external memory it is not generally safe to store an instance of the
498 class (unless you know that the external storage will not be freed). StringRef is
499 small and pervasive enough in LLVM that it should always be passed by value.</p>
501 </div>
503 <!-- _______________________________________________________________________ -->
504 <div class="doc_subsubsection">
505 <a name="Twine">The <tt>Twine</tt> class</a>
506 </div>
508 <div class="doc_text">
510 <p>The <tt>Twine</tt> class is an efficient way for APIs to accept concatenated
511 strings. For example, a common LLVM paradigm is to name one instruction based on
512 the name of another instruction with a suffix, for example:</p>
514 <div class="doc_code">
515 <pre>
516 New = CmpInst::Create(<i>...</i>, SO->getName() + ".cmp");
517 </pre>
518 </div>
520 <p>The <tt>Twine</tt> class is effectively a
521 lightweight <a href="http://en.wikipedia.org/wiki/Rope_(computer_science)">rope</a>
522 which points to temporary (stack allocated) objects. Twines can be implicitly
523 constructed as the result of the plus operator applied to strings (i.e., a C
524 strings, an <tt>std::string</tt>, or a <tt>StringRef</tt>). The twine delays the
525 actual concatenation of strings until it is actually required, at which point
526 it can be efficiently rendered directly into a character array. This avoids
527 unnecessary heap allocation involved in constructing the temporary results of
528 string concatenation. See
529 "<tt><a href="/doxygen/classllvm_1_1Twine_8h-source.html">llvm/ADT/Twine.h</a></tt>"
530 for more information.</p>
532 <p>As with a <tt>StringRef</tt>, <tt>Twine</tt> objects point to external memory
533 and should almost never be stored or mentioned directly. They are intended
534 solely for use when defining a function which should be able to efficiently
535 accept concatenated strings.</p>
537 </div>
540 <!-- ======================================================================= -->
541 <div class="doc_subsection">
542 <a name="DEBUG">The <tt>DEBUG()</tt> macro and <tt>-debug</tt> option</a>
543 </div>
545 <div class="doc_text">
547 <p>Often when working on your pass you will put a bunch of debugging printouts
548 and other code into your pass. After you get it working, you want to remove
549 it, but you may need it again in the future (to work out new bugs that you run
550 across).</p>
552 <p> Naturally, because of this, you don't want to delete the debug printouts,
553 but you don't want them to always be noisy. A standard compromise is to comment
554 them out, allowing you to enable them if you need them in the future.</p>
556 <p>The "<tt><a href="/doxygen/Debug_8h-source.html">llvm/Support/Debug.h</a></tt>"
557 file provides a macro named <tt>DEBUG()</tt> that is a much nicer solution to
558 this problem. Basically, you can put arbitrary code into the argument of the
559 <tt>DEBUG</tt> macro, and it is only executed if '<tt>opt</tt>' (or any other
560 tool) is run with the '<tt>-debug</tt>' command line argument:</p>
562 <div class="doc_code">
563 <pre>
564 DEBUG(errs() &lt;&lt; "I am here!\n");
565 </pre>
566 </div>
568 <p>Then you can run your pass like this:</p>
570 <div class="doc_code">
571 <pre>
572 $ opt &lt; a.bc &gt; /dev/null -mypass
573 <i>&lt;no output&gt;</i>
574 $ opt &lt; a.bc &gt; /dev/null -mypass -debug
575 I am here!
576 </pre>
577 </div>
579 <p>Using the <tt>DEBUG()</tt> macro instead of a home-brewed solution allows you
580 to not have to create "yet another" command line option for the debug output for
581 your pass. Note that <tt>DEBUG()</tt> macros are disabled for optimized builds,
582 so they do not cause a performance impact at all (for the same reason, they
583 should also not contain side-effects!).</p>
585 <p>One additional nice thing about the <tt>DEBUG()</tt> macro is that you can
586 enable or disable it directly in gdb. Just use "<tt>set DebugFlag=0</tt>" or
587 "<tt>set DebugFlag=1</tt>" from the gdb if the program is running. If the
588 program hasn't been started yet, you can always just run it with
589 <tt>-debug</tt>.</p>
591 </div>
593 <!-- _______________________________________________________________________ -->
594 <div class="doc_subsubsection">
595 <a name="DEBUG_TYPE">Fine grained debug info with <tt>DEBUG_TYPE</tt> and
596 the <tt>-debug-only</tt> option</a>
597 </div>
599 <div class="doc_text">
601 <p>Sometimes you may find yourself in a situation where enabling <tt>-debug</tt>
602 just turns on <b>too much</b> information (such as when working on the code
603 generator). If you want to enable debug information with more fine-grained
604 control, you define the <tt>DEBUG_TYPE</tt> macro and the <tt>-debug</tt> only
605 option as follows:</p>
607 <div class="doc_code">
608 <pre>
609 #undef DEBUG_TYPE
610 DEBUG(errs() &lt;&lt; "No debug type\n");
611 #define DEBUG_TYPE "foo"
612 DEBUG(errs() &lt;&lt; "'foo' debug type\n");
613 #undef DEBUG_TYPE
614 #define DEBUG_TYPE "bar"
615 DEBUG(errs() &lt;&lt; "'bar' debug type\n"));
616 #undef DEBUG_TYPE
617 #define DEBUG_TYPE ""
618 DEBUG(errs() &lt;&lt; "No debug type (2)\n");
619 </pre>
620 </div>
622 <p>Then you can run your pass like this:</p>
624 <div class="doc_code">
625 <pre>
626 $ opt &lt; a.bc &gt; /dev/null -mypass
627 <i>&lt;no output&gt;</i>
628 $ opt &lt; a.bc &gt; /dev/null -mypass -debug
629 No debug type
630 'foo' debug type
631 'bar' debug type
632 No debug type (2)
633 $ opt &lt; a.bc &gt; /dev/null -mypass -debug-only=foo
634 'foo' debug type
635 $ opt &lt; a.bc &gt; /dev/null -mypass -debug-only=bar
636 'bar' debug type
637 </pre>
638 </div>
640 <p>Of course, in practice, you should only set <tt>DEBUG_TYPE</tt> at the top of
641 a file, to specify the debug type for the entire module (if you do this before
642 you <tt>#include "llvm/Support/Debug.h"</tt>, you don't have to insert the ugly
643 <tt>#undef</tt>'s). Also, you should use names more meaningful than "foo" and
644 "bar", because there is no system in place to ensure that names do not
645 conflict. If two different modules use the same string, they will all be turned
646 on when the name is specified. This allows, for example, all debug information
647 for instruction scheduling to be enabled with <tt>-debug-type=InstrSched</tt>,
648 even if the source lives in multiple files.</p>
650 <p>The <tt>DEBUG_WITH_TYPE</tt> macro is also available for situations where you
651 would like to set <tt>DEBUG_TYPE</tt>, but only for one specific <tt>DEBUG</tt>
652 statement. It takes an additional first parameter, which is the type to use. For
653 example, the preceding example could be written as:</p>
656 <div class="doc_code">
657 <pre>
658 DEBUG_WITH_TYPE("", errs() &lt;&lt; "No debug type\n");
659 DEBUG_WITH_TYPE("foo", errs() &lt;&lt; "'foo' debug type\n");
660 DEBUG_WITH_TYPE("bar", errs() &lt;&lt; "'bar' debug type\n"));
661 DEBUG_WITH_TYPE("", errs() &lt;&lt; "No debug type (2)\n");
662 </pre>
663 </div>
665 </div>
667 <!-- ======================================================================= -->
668 <div class="doc_subsection">
669 <a name="Statistic">The <tt>Statistic</tt> class &amp; <tt>-stats</tt>
670 option</a>
671 </div>
673 <div class="doc_text">
675 <p>The "<tt><a
676 href="/doxygen/Statistic_8h-source.html">llvm/ADT/Statistic.h</a></tt>" file
677 provides a class named <tt>Statistic</tt> that is used as a unified way to
678 keep track of what the LLVM compiler is doing and how effective various
679 optimizations are. It is useful to see what optimizations are contributing to
680 making a particular program run faster.</p>
682 <p>Often you may run your pass on some big program, and you're interested to see
683 how many times it makes a certain transformation. Although you can do this with
684 hand inspection, or some ad-hoc method, this is a real pain and not very useful
685 for big programs. Using the <tt>Statistic</tt> class makes it very easy to
686 keep track of this information, and the calculated information is presented in a
687 uniform manner with the rest of the passes being executed.</p>
689 <p>There are many examples of <tt>Statistic</tt> uses, but the basics of using
690 it are as follows:</p>
692 <ol>
693 <li><p>Define your statistic like this:</p>
695 <div class="doc_code">
696 <pre>
697 #define <a href="#DEBUG_TYPE">DEBUG_TYPE</a> "mypassname" <i>// This goes before any #includes.</i>
698 STATISTIC(NumXForms, "The # of times I did stuff");
699 </pre>
700 </div>
702 <p>The <tt>STATISTIC</tt> macro defines a static variable, whose name is
703 specified by the first argument. The pass name is taken from the DEBUG_TYPE
704 macro, and the description is taken from the second argument. The variable
705 defined ("NumXForms" in this case) acts like an unsigned integer.</p></li>
707 <li><p>Whenever you make a transformation, bump the counter:</p>
709 <div class="doc_code">
710 <pre>
711 ++NumXForms; // <i>I did stuff!</i>
712 </pre>
713 </div>
715 </li>
716 </ol>
718 <p>That's all you have to do. To get '<tt>opt</tt>' to print out the
719 statistics gathered, use the '<tt>-stats</tt>' option:</p>
721 <div class="doc_code">
722 <pre>
723 $ opt -stats -mypassname &lt; program.bc &gt; /dev/null
724 <i>... statistics output ...</i>
725 </pre>
726 </div>
728 <p> When running <tt>opt</tt> on a C file from the SPEC benchmark
729 suite, it gives a report that looks like this:</p>
731 <div class="doc_code">
732 <pre>
733 7646 bitcodewriter - Number of normal instructions
734 725 bitcodewriter - Number of oversized instructions
735 129996 bitcodewriter - Number of bitcode bytes written
736 2817 raise - Number of insts DCEd or constprop'd
737 3213 raise - Number of cast-of-self removed
738 5046 raise - Number of expression trees converted
739 75 raise - Number of other getelementptr's formed
740 138 raise - Number of load/store peepholes
741 42 deadtypeelim - Number of unused typenames removed from symtab
742 392 funcresolve - Number of varargs functions resolved
743 27 globaldce - Number of global variables removed
744 2 adce - Number of basic blocks removed
745 134 cee - Number of branches revectored
746 49 cee - Number of setcc instruction eliminated
747 532 gcse - Number of loads removed
748 2919 gcse - Number of instructions removed
749 86 indvars - Number of canonical indvars added
750 87 indvars - Number of aux indvars removed
751 25 instcombine - Number of dead inst eliminate
752 434 instcombine - Number of insts combined
753 248 licm - Number of load insts hoisted
754 1298 licm - Number of insts hoisted to a loop pre-header
755 3 licm - Number of insts hoisted to multiple loop preds (bad, no loop pre-header)
756 75 mem2reg - Number of alloca's promoted
757 1444 cfgsimplify - Number of blocks simplified
758 </pre>
759 </div>
761 <p>Obviously, with so many optimizations, having a unified framework for this
762 stuff is very nice. Making your pass fit well into the framework makes it more
763 maintainable and useful.</p>
765 </div>
767 <!-- ======================================================================= -->
768 <div class="doc_subsection">
769 <a name="ViewGraph">Viewing graphs while debugging code</a>
770 </div>
772 <div class="doc_text">
774 <p>Several of the important data structures in LLVM are graphs: for example
775 CFGs made out of LLVM <a href="#BasicBlock">BasicBlock</a>s, CFGs made out of
776 LLVM <a href="CodeGenerator.html#machinebasicblock">MachineBasicBlock</a>s, and
777 <a href="CodeGenerator.html#selectiondag_intro">Instruction Selection
778 DAGs</a>. In many cases, while debugging various parts of the compiler, it is
779 nice to instantly visualize these graphs.</p>
781 <p>LLVM provides several callbacks that are available in a debug build to do
782 exactly that. If you call the <tt>Function::viewCFG()</tt> method, for example,
783 the current LLVM tool will pop up a window containing the CFG for the function
784 where each basic block is a node in the graph, and each node contains the
785 instructions in the block. Similarly, there also exists
786 <tt>Function::viewCFGOnly()</tt> (does not include the instructions), the
787 <tt>MachineFunction::viewCFG()</tt> and <tt>MachineFunction::viewCFGOnly()</tt>,
788 and the <tt>SelectionDAG::viewGraph()</tt> methods. Within GDB, for example,
789 you can usually use something like <tt>call DAG.viewGraph()</tt> to pop
790 up a window. Alternatively, you can sprinkle calls to these functions in your
791 code in places you want to debug.</p>
793 <p>Getting this to work requires a small amount of configuration. On Unix
794 systems with X11, install the <a href="http://www.graphviz.org">graphviz</a>
795 toolkit, and make sure 'dot' and 'gv' are in your path. If you are running on
796 Mac OS/X, download and install the Mac OS/X <a
797 href="http://www.pixelglow.com/graphviz/">Graphviz program</a>, and add
798 <tt>/Applications/Graphviz.app/Contents/MacOS/</tt> (or wherever you install
799 it) to your path. Once in your system and path are set up, rerun the LLVM
800 configure script and rebuild LLVM to enable this functionality.</p>
802 <p><tt>SelectionDAG</tt> has been extended to make it easier to locate
803 <i>interesting</i> nodes in large complex graphs. From gdb, if you
804 <tt>call DAG.setGraphColor(<i>node</i>, "<i>color</i>")</tt>, then the
805 next <tt>call DAG.viewGraph()</tt> would highlight the node in the
806 specified color (choices of colors can be found at <a
807 href="http://www.graphviz.org/doc/info/colors.html">colors</a>.) More
808 complex node attributes can be provided with <tt>call
809 DAG.setGraphAttrs(<i>node</i>, "<i>attributes</i>")</tt> (choices can be
810 found at <a href="http://www.graphviz.org/doc/info/attrs.html">Graph
811 Attributes</a>.) If you want to restart and clear all the current graph
812 attributes, then you can <tt>call DAG.clearGraphAttrs()</tt>. </p>
814 </div>
816 <!-- *********************************************************************** -->
817 <div class="doc_section">
818 <a name="datastructure">Picking the Right Data Structure for a Task</a>
819 </div>
820 <!-- *********************************************************************** -->
822 <div class="doc_text">
824 <p>LLVM has a plethora of data structures in the <tt>llvm/ADT/</tt> directory,
825 and we commonly use STL data structures. This section describes the trade-offs
826 you should consider when you pick one.</p>
829 The first step is a choose your own adventure: do you want a sequential
830 container, a set-like container, or a map-like container? The most important
831 thing when choosing a container is the algorithmic properties of how you plan to
832 access the container. Based on that, you should use:</p>
834 <ul>
835 <li>a <a href="#ds_map">map-like</a> container if you need efficient look-up
836 of an value based on another value. Map-like containers also support
837 efficient queries for containment (whether a key is in the map). Map-like
838 containers generally do not support efficient reverse mapping (values to
839 keys). If you need that, use two maps. Some map-like containers also
840 support efficient iteration through the keys in sorted order. Map-like
841 containers are the most expensive sort, only use them if you need one of
842 these capabilities.</li>
844 <li>a <a href="#ds_set">set-like</a> container if you need to put a bunch of
845 stuff into a container that automatically eliminates duplicates. Some
846 set-like containers support efficient iteration through the elements in
847 sorted order. Set-like containers are more expensive than sequential
848 containers.
849 </li>
851 <li>a <a href="#ds_sequential">sequential</a> container provides
852 the most efficient way to add elements and keeps track of the order they are
853 added to the collection. They permit duplicates and support efficient
854 iteration, but do not support efficient look-up based on a key.
855 </li>
857 <li>a <a href="#ds_string">string</a> container is a specialized sequential
858 container or reference structure that is used for character or byte
859 arrays.</li>
861 <li>a <a href="#ds_bit">bit</a> container provides an efficient way to store and
862 perform set operations on sets of numeric id's, while automatically
863 eliminating duplicates. Bit containers require a maximum of 1 bit for each
864 identifier you want to store.
865 </li>
866 </ul>
869 Once the proper category of container is determined, you can fine tune the
870 memory use, constant factors, and cache behaviors of access by intelligently
871 picking a member of the category. Note that constant factors and cache behavior
872 can be a big deal. If you have a vector that usually only contains a few
873 elements (but could contain many), for example, it's much better to use
874 <a href="#dss_smallvector">SmallVector</a> than <a href="#dss_vector">vector</a>
875 . Doing so avoids (relatively) expensive malloc/free calls, which dwarf the
876 cost of adding the elements to the container. </p>
878 </div>
880 <!-- ======================================================================= -->
881 <div class="doc_subsection">
882 <a name="ds_sequential">Sequential Containers (std::vector, std::list, etc)</a>
883 </div>
885 <div class="doc_text">
886 There are a variety of sequential containers available for you, based on your
887 needs. Pick the first in this section that will do what you want.
888 </div>
890 <!-- _______________________________________________________________________ -->
891 <div class="doc_subsubsection">
892 <a name="dss_fixedarrays">Fixed Size Arrays</a>
893 </div>
895 <div class="doc_text">
896 <p>Fixed size arrays are very simple and very fast. They are good if you know
897 exactly how many elements you have, or you have a (low) upper bound on how many
898 you have.</p>
899 </div>
901 <!-- _______________________________________________________________________ -->
902 <div class="doc_subsubsection">
903 <a name="dss_heaparrays">Heap Allocated Arrays</a>
904 </div>
906 <div class="doc_text">
907 <p>Heap allocated arrays (new[] + delete[]) are also simple. They are good if
908 the number of elements is variable, if you know how many elements you will need
909 before the array is allocated, and if the array is usually large (if not,
910 consider a <a href="#dss_smallvector">SmallVector</a>). The cost of a heap
911 allocated array is the cost of the new/delete (aka malloc/free). Also note that
912 if you are allocating an array of a type with a constructor, the constructor and
913 destructors will be run for every element in the array (re-sizable vectors only
914 construct those elements actually used).</p>
915 </div>
917 <!-- _______________________________________________________________________ -->
918 <div class="doc_subsubsection">
919 <a name="dss_smallvector">"llvm/ADT/SmallVector.h"</a>
920 </div>
922 <div class="doc_text">
923 <p><tt>SmallVector&lt;Type, N&gt;</tt> is a simple class that looks and smells
924 just like <tt>vector&lt;Type&gt;</tt>:
925 it supports efficient iteration, lays out elements in memory order (so you can
926 do pointer arithmetic between elements), supports efficient push_back/pop_back
927 operations, supports efficient random access to its elements, etc.</p>
929 <p>The advantage of SmallVector is that it allocates space for
930 some number of elements (N) <b>in the object itself</b>. Because of this, if
931 the SmallVector is dynamically smaller than N, no malloc is performed. This can
932 be a big win in cases where the malloc/free call is far more expensive than the
933 code that fiddles around with the elements.</p>
935 <p>This is good for vectors that are "usually small" (e.g. the number of
936 predecessors/successors of a block is usually less than 8). On the other hand,
937 this makes the size of the SmallVector itself large, so you don't want to
938 allocate lots of them (doing so will waste a lot of space). As such,
939 SmallVectors are most useful when on the stack.</p>
941 <p>SmallVector also provides a nice portable and efficient replacement for
942 <tt>alloca</tt>.</p>
944 </div>
946 <!-- _______________________________________________________________________ -->
947 <div class="doc_subsubsection">
948 <a name="dss_vector">&lt;vector&gt;</a>
949 </div>
951 <div class="doc_text">
953 std::vector is well loved and respected. It is useful when SmallVector isn't:
954 when the size of the vector is often large (thus the small optimization will
955 rarely be a benefit) or if you will be allocating many instances of the vector
956 itself (which would waste space for elements that aren't in the container).
957 vector is also useful when interfacing with code that expects vectors :).
958 </p>
960 <p>One worthwhile note about std::vector: avoid code like this:</p>
962 <div class="doc_code">
963 <pre>
964 for ( ... ) {
965 std::vector&lt;foo&gt; V;
966 use V;
968 </pre>
969 </div>
971 <p>Instead, write this as:</p>
973 <div class="doc_code">
974 <pre>
975 std::vector&lt;foo&gt; V;
976 for ( ... ) {
977 use V;
978 V.clear();
980 </pre>
981 </div>
983 <p>Doing so will save (at least) one heap allocation and free per iteration of
984 the loop.</p>
986 </div>
988 <!-- _______________________________________________________________________ -->
989 <div class="doc_subsubsection">
990 <a name="dss_deque">&lt;deque&gt;</a>
991 </div>
993 <div class="doc_text">
994 <p>std::deque is, in some senses, a generalized version of std::vector. Like
995 std::vector, it provides constant time random access and other similar
996 properties, but it also provides efficient access to the front of the list. It
997 does not guarantee continuity of elements within memory.</p>
999 <p>In exchange for this extra flexibility, std::deque has significantly higher
1000 constant factor costs than std::vector. If possible, use std::vector or
1001 something cheaper.</p>
1002 </div>
1004 <!-- _______________________________________________________________________ -->
1005 <div class="doc_subsubsection">
1006 <a name="dss_list">&lt;list&gt;</a>
1007 </div>
1009 <div class="doc_text">
1010 <p>std::list is an extremely inefficient class that is rarely useful.
1011 It performs a heap allocation for every element inserted into it, thus having an
1012 extremely high constant factor, particularly for small data types. std::list
1013 also only supports bidirectional iteration, not random access iteration.</p>
1015 <p>In exchange for this high cost, std::list supports efficient access to both
1016 ends of the list (like std::deque, but unlike std::vector or SmallVector). In
1017 addition, the iterator invalidation characteristics of std::list are stronger
1018 than that of a vector class: inserting or removing an element into the list does
1019 not invalidate iterator or pointers to other elements in the list.</p>
1020 </div>
1022 <!-- _______________________________________________________________________ -->
1023 <div class="doc_subsubsection">
1024 <a name="dss_ilist">llvm/ADT/ilist.h</a>
1025 </div>
1027 <div class="doc_text">
1028 <p><tt>ilist&lt;T&gt;</tt> implements an 'intrusive' doubly-linked list. It is
1029 intrusive, because it requires the element to store and provide access to the
1030 prev/next pointers for the list.</p>
1032 <p><tt>ilist</tt> has the same drawbacks as <tt>std::list</tt>, and additionally
1033 requires an <tt>ilist_traits</tt> implementation for the element type, but it
1034 provides some novel characteristics. In particular, it can efficiently store
1035 polymorphic objects, the traits class is informed when an element is inserted or
1036 removed from the list, and <tt>ilist</tt>s are guaranteed to support a
1037 constant-time splice operation.</p>
1039 <p>These properties are exactly what we want for things like
1040 <tt>Instruction</tt>s and basic blocks, which is why these are implemented with
1041 <tt>ilist</tt>s.</p>
1043 Related classes of interest are explained in the following subsections:
1044 <ul>
1045 <li><a href="#dss_ilist_traits">ilist_traits</a></li>
1046 <li><a href="#dss_iplist">iplist</a></li>
1047 <li><a href="#dss_ilist_node">llvm/ADT/ilist_node.h</a></li>
1048 <li><a href="#dss_ilist_sentinel">Sentinels</a></li>
1049 </ul>
1050 </div>
1052 <!-- _______________________________________________________________________ -->
1053 <div class="doc_subsubsection">
1054 <a name="dss_ilist_traits">ilist_traits</a>
1055 </div>
1057 <div class="doc_text">
1058 <p><tt>ilist_traits&lt;T&gt;</tt> is <tt>ilist&lt;T&gt;</tt>'s customization
1059 mechanism. <tt>iplist&lt;T&gt;</tt> (and consequently <tt>ilist&lt;T&gt;</tt>)
1060 publicly derive from this traits class.</p>
1061 </div>
1063 <!-- _______________________________________________________________________ -->
1064 <div class="doc_subsubsection">
1065 <a name="dss_iplist">iplist</a>
1066 </div>
1068 <div class="doc_text">
1069 <p><tt>iplist&lt;T&gt;</tt> is <tt>ilist&lt;T&gt;</tt>'s base and as such
1070 supports a slightly narrower interface. Notably, inserters from
1071 <tt>T&amp;</tt> are absent.</p>
1073 <p><tt>ilist_traits&lt;T&gt;</tt> is a public base of this class and can be
1074 used for a wide variety of customizations.</p>
1075 </div>
1077 <!-- _______________________________________________________________________ -->
1078 <div class="doc_subsubsection">
1079 <a name="dss_ilist_node">llvm/ADT/ilist_node.h</a>
1080 </div>
1082 <div class="doc_text">
1083 <p><tt>ilist_node&lt;T&gt;</tt> implements a the forward and backward links
1084 that are expected by the <tt>ilist&lt;T&gt;</tt> (and analogous containers)
1085 in the default manner.</p>
1087 <p><tt>ilist_node&lt;T&gt;</tt>s are meant to be embedded in the node type
1088 <tt>T</tt>, usually <tt>T</tt> publicly derives from
1089 <tt>ilist_node&lt;T&gt;</tt>.</p>
1090 </div>
1092 <!-- _______________________________________________________________________ -->
1093 <div class="doc_subsubsection">
1094 <a name="dss_ilist_sentinel">Sentinels</a>
1095 </div>
1097 <div class="doc_text">
1098 <p><tt>ilist</tt>s have another specialty that must be considered. To be a good
1099 citizen in the C++ ecosystem, it needs to support the standard container
1100 operations, such as <tt>begin</tt> and <tt>end</tt> iterators, etc. Also, the
1101 <tt>operator--</tt> must work correctly on the <tt>end</tt> iterator in the
1102 case of non-empty <tt>ilist</tt>s.</p>
1104 <p>The only sensible solution to this problem is to allocate a so-called
1105 <i>sentinel</i> along with the intrusive list, which serves as the <tt>end</tt>
1106 iterator, providing the back-link to the last element. However conforming to the
1107 C++ convention it is illegal to <tt>operator++</tt> beyond the sentinel and it
1108 also must not be dereferenced.</p>
1110 <p>These constraints allow for some implementation freedom to the <tt>ilist</tt>
1111 how to allocate and store the sentinel. The corresponding policy is dictated
1112 by <tt>ilist_traits&lt;T&gt;</tt>. By default a <tt>T</tt> gets heap-allocated
1113 whenever the need for a sentinel arises.</p>
1115 <p>While the default policy is sufficient in most cases, it may break down when
1116 <tt>T</tt> does not provide a default constructor. Also, in the case of many
1117 instances of <tt>ilist</tt>s, the memory overhead of the associated sentinels
1118 is wasted. To alleviate the situation with numerous and voluminous
1119 <tt>T</tt>-sentinels, sometimes a trick is employed, leading to <i>ghostly
1120 sentinels</i>.</p>
1122 <p>Ghostly sentinels are obtained by specially-crafted <tt>ilist_traits&lt;T&gt;</tt>
1123 which superpose the sentinel with the <tt>ilist</tt> instance in memory. Pointer
1124 arithmetic is used to obtain the sentinel, which is relative to the
1125 <tt>ilist</tt>'s <tt>this</tt> pointer. The <tt>ilist</tt> is augmented by an
1126 extra pointer, which serves as the back-link of the sentinel. This is the only
1127 field in the ghostly sentinel which can be legally accessed.</p>
1128 </div>
1130 <!-- _______________________________________________________________________ -->
1131 <div class="doc_subsubsection">
1132 <a name="dss_other">Other Sequential Container options</a>
1133 </div>
1135 <div class="doc_text">
1136 <p>Other STL containers are available, such as std::string.</p>
1138 <p>There are also various STL adapter classes such as std::queue,
1139 std::priority_queue, std::stack, etc. These provide simplified access to an
1140 underlying container but don't affect the cost of the container itself.</p>
1142 </div>
1145 <!-- ======================================================================= -->
1146 <div class="doc_subsection">
1147 <a name="ds_set">Set-Like Containers (std::set, SmallSet, SetVector, etc)</a>
1148 </div>
1150 <div class="doc_text">
1152 <p>Set-like containers are useful when you need to canonicalize multiple values
1153 into a single representation. There are several different choices for how to do
1154 this, providing various trade-offs.</p>
1156 </div>
1159 <!-- _______________________________________________________________________ -->
1160 <div class="doc_subsubsection">
1161 <a name="dss_sortedvectorset">A sorted 'vector'</a>
1162 </div>
1164 <div class="doc_text">
1166 <p>If you intend to insert a lot of elements, then do a lot of queries, a
1167 great approach is to use a vector (or other sequential container) with
1168 std::sort+std::unique to remove duplicates. This approach works really well if
1169 your usage pattern has these two distinct phases (insert then query), and can be
1170 coupled with a good choice of <a href="#ds_sequential">sequential container</a>.
1171 </p>
1174 This combination provides the several nice properties: the result data is
1175 contiguous in memory (good for cache locality), has few allocations, is easy to
1176 address (iterators in the final vector are just indices or pointers), and can be
1177 efficiently queried with a standard binary or radix search.</p>
1179 </div>
1181 <!-- _______________________________________________________________________ -->
1182 <div class="doc_subsubsection">
1183 <a name="dss_smallset">"llvm/ADT/SmallSet.h"</a>
1184 </div>
1186 <div class="doc_text">
1188 <p>If you have a set-like data structure that is usually small and whose elements
1189 are reasonably small, a <tt>SmallSet&lt;Type, N&gt;</tt> is a good choice. This set
1190 has space for N elements in place (thus, if the set is dynamically smaller than
1191 N, no malloc traffic is required) and accesses them with a simple linear search.
1192 When the set grows beyond 'N' elements, it allocates a more expensive representation that
1193 guarantees efficient access (for most types, it falls back to std::set, but for
1194 pointers it uses something far better, <a
1195 href="#dss_smallptrset">SmallPtrSet</a>).</p>
1197 <p>The magic of this class is that it handles small sets extremely efficiently,
1198 but gracefully handles extremely large sets without loss of efficiency. The
1199 drawback is that the interface is quite small: it supports insertion, queries
1200 and erasing, but does not support iteration.</p>
1202 </div>
1204 <!-- _______________________________________________________________________ -->
1205 <div class="doc_subsubsection">
1206 <a name="dss_smallptrset">"llvm/ADT/SmallPtrSet.h"</a>
1207 </div>
1209 <div class="doc_text">
1211 <p>SmallPtrSet has all the advantages of <tt>SmallSet</tt> (and a <tt>SmallSet</tt> of pointers is
1212 transparently implemented with a <tt>SmallPtrSet</tt>), but also supports iterators. If
1213 more than 'N' insertions are performed, a single quadratically
1214 probed hash table is allocated and grows as needed, providing extremely
1215 efficient access (constant time insertion/deleting/queries with low constant
1216 factors) and is very stingy with malloc traffic.</p>
1218 <p>Note that, unlike <tt>std::set</tt>, the iterators of <tt>SmallPtrSet</tt> are invalidated
1219 whenever an insertion occurs. Also, the values visited by the iterators are not
1220 visited in sorted order.</p>
1222 </div>
1224 <!-- _______________________________________________________________________ -->
1225 <div class="doc_subsubsection">
1226 <a name="dss_denseset">"llvm/ADT/DenseSet.h"</a>
1227 </div>
1229 <div class="doc_text">
1232 DenseSet is a simple quadratically probed hash table. It excels at supporting
1233 small values: it uses a single allocation to hold all of the pairs that
1234 are currently inserted in the set. DenseSet is a great way to unique small
1235 values that are not simple pointers (use <a
1236 href="#dss_smallptrset">SmallPtrSet</a> for pointers). Note that DenseSet has
1237 the same requirements for the value type that <a
1238 href="#dss_densemap">DenseMap</a> has.
1239 </p>
1241 </div>
1243 <!-- _______________________________________________________________________ -->
1244 <div class="doc_subsubsection">
1245 <a name="dss_FoldingSet">"llvm/ADT/FoldingSet.h"</a>
1246 </div>
1248 <div class="doc_text">
1251 FoldingSet is an aggregate class that is really good at uniquing
1252 expensive-to-create or polymorphic objects. It is a combination of a chained
1253 hash table with intrusive links (uniqued objects are required to inherit from
1254 FoldingSetNode) that uses <a href="#dss_smallvector">SmallVector</a> as part of
1255 its ID process.</p>
1257 <p>Consider a case where you want to implement a "getOrCreateFoo" method for
1258 a complex object (for example, a node in the code generator). The client has a
1259 description of *what* it wants to generate (it knows the opcode and all the
1260 operands), but we don't want to 'new' a node, then try inserting it into a set
1261 only to find out it already exists, at which point we would have to delete it
1262 and return the node that already exists.
1263 </p>
1265 <p>To support this style of client, FoldingSet perform a query with a
1266 FoldingSetNodeID (which wraps SmallVector) that can be used to describe the
1267 element that we want to query for. The query either returns the element
1268 matching the ID or it returns an opaque ID that indicates where insertion should
1269 take place. Construction of the ID usually does not require heap traffic.</p>
1271 <p>Because FoldingSet uses intrusive links, it can support polymorphic objects
1272 in the set (for example, you can have SDNode instances mixed with LoadSDNodes).
1273 Because the elements are individually allocated, pointers to the elements are
1274 stable: inserting or removing elements does not invalidate any pointers to other
1275 elements.
1276 </p>
1278 </div>
1280 <!-- _______________________________________________________________________ -->
1281 <div class="doc_subsubsection">
1282 <a name="dss_set">&lt;set&gt;</a>
1283 </div>
1285 <div class="doc_text">
1287 <p><tt>std::set</tt> is a reasonable all-around set class, which is decent at
1288 many things but great at nothing. std::set allocates memory for each element
1289 inserted (thus it is very malloc intensive) and typically stores three pointers
1290 per element in the set (thus adding a large amount of per-element space
1291 overhead). It offers guaranteed log(n) performance, which is not particularly
1292 fast from a complexity standpoint (particularly if the elements of the set are
1293 expensive to compare, like strings), and has extremely high constant factors for
1294 lookup, insertion and removal.</p>
1296 <p>The advantages of std::set are that its iterators are stable (deleting or
1297 inserting an element from the set does not affect iterators or pointers to other
1298 elements) and that iteration over the set is guaranteed to be in sorted order.
1299 If the elements in the set are large, then the relative overhead of the pointers
1300 and malloc traffic is not a big deal, but if the elements of the set are small,
1301 std::set is almost never a good choice.</p>
1303 </div>
1305 <!-- _______________________________________________________________________ -->
1306 <div class="doc_subsubsection">
1307 <a name="dss_setvector">"llvm/ADT/SetVector.h"</a>
1308 </div>
1310 <div class="doc_text">
1311 <p>LLVM's SetVector&lt;Type&gt; is an adapter class that combines your choice of
1312 a set-like container along with a <a href="#ds_sequential">Sequential
1313 Container</a>. The important property
1314 that this provides is efficient insertion with uniquing (duplicate elements are
1315 ignored) with iteration support. It implements this by inserting elements into
1316 both a set-like container and the sequential container, using the set-like
1317 container for uniquing and the sequential container for iteration.
1318 </p>
1320 <p>The difference between SetVector and other sets is that the order of
1321 iteration is guaranteed to match the order of insertion into the SetVector.
1322 This property is really important for things like sets of pointers. Because
1323 pointer values are non-deterministic (e.g. vary across runs of the program on
1324 different machines), iterating over the pointers in the set will
1325 not be in a well-defined order.</p>
1328 The drawback of SetVector is that it requires twice as much space as a normal
1329 set and has the sum of constant factors from the set-like container and the
1330 sequential container that it uses. Use it *only* if you need to iterate over
1331 the elements in a deterministic order. SetVector is also expensive to delete
1332 elements out of (linear time), unless you use it's "pop_back" method, which is
1333 faster.
1334 </p>
1336 <p>SetVector is an adapter class that defaults to using std::vector and std::set
1337 for the underlying containers, so it is quite expensive. However,
1338 <tt>"llvm/ADT/SetVector.h"</tt> also provides a SmallSetVector class, which
1339 defaults to using a SmallVector and SmallSet of a specified size. If you use
1340 this, and if your sets are dynamically smaller than N, you will save a lot of
1341 heap traffic.</p>
1343 </div>
1345 <!-- _______________________________________________________________________ -->
1346 <div class="doc_subsubsection">
1347 <a name="dss_uniquevector">"llvm/ADT/UniqueVector.h"</a>
1348 </div>
1350 <div class="doc_text">
1353 UniqueVector is similar to <a href="#dss_setvector">SetVector</a>, but it
1354 retains a unique ID for each element inserted into the set. It internally
1355 contains a map and a vector, and it assigns a unique ID for each value inserted
1356 into the set.</p>
1358 <p>UniqueVector is very expensive: its cost is the sum of the cost of
1359 maintaining both the map and vector, it has high complexity, high constant
1360 factors, and produces a lot of malloc traffic. It should be avoided.</p>
1362 </div>
1365 <!-- _______________________________________________________________________ -->
1366 <div class="doc_subsubsection">
1367 <a name="dss_otherset">Other Set-Like Container Options</a>
1368 </div>
1370 <div class="doc_text">
1373 The STL provides several other options, such as std::multiset and the various
1374 "hash_set" like containers (whether from C++ TR1 or from the SGI library). We
1375 never use hash_set and unordered_set because they are generally very expensive
1376 (each insertion requires a malloc) and very non-portable.
1377 </p>
1379 <p>std::multiset is useful if you're not interested in elimination of
1380 duplicates, but has all the drawbacks of std::set. A sorted vector (where you
1381 don't delete duplicate entries) or some other approach is almost always
1382 better.</p>
1384 </div>
1386 <!-- ======================================================================= -->
1387 <div class="doc_subsection">
1388 <a name="ds_map">Map-Like Containers (std::map, DenseMap, etc)</a>
1389 </div>
1391 <div class="doc_text">
1392 Map-like containers are useful when you want to associate data to a key. As
1393 usual, there are a lot of different ways to do this. :)
1394 </div>
1396 <!-- _______________________________________________________________________ -->
1397 <div class="doc_subsubsection">
1398 <a name="dss_sortedvectormap">A sorted 'vector'</a>
1399 </div>
1401 <div class="doc_text">
1404 If your usage pattern follows a strict insert-then-query approach, you can
1405 trivially use the same approach as <a href="#dss_sortedvectorset">sorted vectors
1406 for set-like containers</a>. The only difference is that your query function
1407 (which uses std::lower_bound to get efficient log(n) lookup) should only compare
1408 the key, not both the key and value. This yields the same advantages as sorted
1409 vectors for sets.
1410 </p>
1411 </div>
1413 <!-- _______________________________________________________________________ -->
1414 <div class="doc_subsubsection">
1415 <a name="dss_stringmap">"llvm/ADT/StringMap.h"</a>
1416 </div>
1418 <div class="doc_text">
1421 Strings are commonly used as keys in maps, and they are difficult to support
1422 efficiently: they are variable length, inefficient to hash and compare when
1423 long, expensive to copy, etc. StringMap is a specialized container designed to
1424 cope with these issues. It supports mapping an arbitrary range of bytes to an
1425 arbitrary other object.</p>
1427 <p>The StringMap implementation uses a quadratically-probed hash table, where
1428 the buckets store a pointer to the heap allocated entries (and some other
1429 stuff). The entries in the map must be heap allocated because the strings are
1430 variable length. The string data (key) and the element object (value) are
1431 stored in the same allocation with the string data immediately after the element
1432 object. This container guarantees the "<tt>(char*)(&amp;Value+1)</tt>" points
1433 to the key string for a value.</p>
1435 <p>The StringMap is very fast for several reasons: quadratic probing is very
1436 cache efficient for lookups, the hash value of strings in buckets is not
1437 recomputed when looking up an element, StringMap rarely has to touch the
1438 memory for unrelated objects when looking up a value (even when hash collisions
1439 happen), hash table growth does not recompute the hash values for strings
1440 already in the table, and each pair in the map is store in a single allocation
1441 (the string data is stored in the same allocation as the Value of a pair).</p>
1443 <p>StringMap also provides query methods that take byte ranges, so it only ever
1444 copies a string if a value is inserted into the table.</p>
1445 </div>
1447 <!-- _______________________________________________________________________ -->
1448 <div class="doc_subsubsection">
1449 <a name="dss_indexedmap">"llvm/ADT/IndexedMap.h"</a>
1450 </div>
1452 <div class="doc_text">
1454 IndexedMap is a specialized container for mapping small dense integers (or
1455 values that can be mapped to small dense integers) to some other type. It is
1456 internally implemented as a vector with a mapping function that maps the keys to
1457 the dense integer range.
1458 </p>
1461 This is useful for cases like virtual registers in the LLVM code generator: they
1462 have a dense mapping that is offset by a compile-time constant (the first
1463 virtual register ID).</p>
1465 </div>
1467 <!-- _______________________________________________________________________ -->
1468 <div class="doc_subsubsection">
1469 <a name="dss_densemap">"llvm/ADT/DenseMap.h"</a>
1470 </div>
1472 <div class="doc_text">
1475 DenseMap is a simple quadratically probed hash table. It excels at supporting
1476 small keys and values: it uses a single allocation to hold all of the pairs that
1477 are currently inserted in the map. DenseMap is a great way to map pointers to
1478 pointers, or map other small types to each other.
1479 </p>
1482 There are several aspects of DenseMap that you should be aware of, however. The
1483 iterators in a densemap are invalidated whenever an insertion occurs, unlike
1484 map. Also, because DenseMap allocates space for a large number of key/value
1485 pairs (it starts with 64 by default), it will waste a lot of space if your keys
1486 or values are large. Finally, you must implement a partial specialization of
1487 DenseMapInfo for the key that you want, if it isn't already supported. This
1488 is required to tell DenseMap about two special marker values (which can never be
1489 inserted into the map) that it needs internally.</p>
1491 </div>
1493 <!-- _______________________________________________________________________ -->
1494 <div class="doc_subsubsection">
1495 <a name="dss_valuemap">"llvm/ADT/ValueMap.h"</a>
1496 </div>
1498 <div class="doc_text">
1501 ValueMap is a wrapper around a <a href="#dss_densemap">DenseMap</a> mapping
1502 Value*s (or subclasses) to another type. When a Value is deleted or RAUW'ed,
1503 ValueMap will update itself so the new version of the key is mapped to the same
1504 value, just as if the key were a WeakVH. You can configure exactly how this
1505 happens, and what else happens on these two events, by passing
1506 a <code>Config</code> parameter to the ValueMap template.</p>
1508 </div>
1510 <!-- _______________________________________________________________________ -->
1511 <div class="doc_subsubsection">
1512 <a name="dss_map">&lt;map&gt;</a>
1513 </div>
1515 <div class="doc_text">
1518 std::map has similar characteristics to <a href="#dss_set">std::set</a>: it uses
1519 a single allocation per pair inserted into the map, it offers log(n) lookup with
1520 an extremely large constant factor, imposes a space penalty of 3 pointers per
1521 pair in the map, etc.</p>
1523 <p>std::map is most useful when your keys or values are very large, if you need
1524 to iterate over the collection in sorted order, or if you need stable iterators
1525 into the map (i.e. they don't get invalidated if an insertion or deletion of
1526 another element takes place).</p>
1528 </div>
1530 <!-- _______________________________________________________________________ -->
1531 <div class="doc_subsubsection">
1532 <a name="dss_othermap">Other Map-Like Container Options</a>
1533 </div>
1535 <div class="doc_text">
1538 The STL provides several other options, such as std::multimap and the various
1539 "hash_map" like containers (whether from C++ TR1 or from the SGI library). We
1540 never use hash_set and unordered_set because they are generally very expensive
1541 (each insertion requires a malloc) and very non-portable.</p>
1543 <p>std::multimap is useful if you want to map a key to multiple values, but has
1544 all the drawbacks of std::map. A sorted vector or some other approach is almost
1545 always better.</p>
1547 </div>
1549 <!-- ======================================================================= -->
1550 <div class="doc_subsection">
1551 <a name="ds_string">String-like containers</a>
1552 </div>
1554 <div class="doc_text">
1557 TODO: const char* vs stringref vs smallstring vs std::string. Describe twine,
1558 xref to #string_apis.
1559 </p>
1561 </div>
1563 <!-- ======================================================================= -->
1564 <div class="doc_subsection">
1565 <a name="ds_bit">Bit storage containers (BitVector, SparseBitVector)</a>
1566 </div>
1568 <div class="doc_text">
1569 <p>Unlike the other containers, there are only two bit storage containers, and
1570 choosing when to use each is relatively straightforward.</p>
1572 <p>One additional option is
1573 <tt>std::vector&lt;bool&gt;</tt>: we discourage its use for two reasons 1) the
1574 implementation in many common compilers (e.g. commonly available versions of
1575 GCC) is extremely inefficient and 2) the C++ standards committee is likely to
1576 deprecate this container and/or change it significantly somehow. In any case,
1577 please don't use it.</p>
1578 </div>
1580 <!-- _______________________________________________________________________ -->
1581 <div class="doc_subsubsection">
1582 <a name="dss_bitvector">BitVector</a>
1583 </div>
1585 <div class="doc_text">
1586 <p> The BitVector container provides a dynamic size set of bits for manipulation.
1587 It supports individual bit setting/testing, as well as set operations. The set
1588 operations take time O(size of bitvector), but operations are performed one word
1589 at a time, instead of one bit at a time. This makes the BitVector very fast for
1590 set operations compared to other containers. Use the BitVector when you expect
1591 the number of set bits to be high (IE a dense set).
1592 </p>
1593 </div>
1595 <!-- _______________________________________________________________________ -->
1596 <div class="doc_subsubsection">
1597 <a name="dss_smallbitvector">SmallBitVector</a>
1598 </div>
1600 <div class="doc_text">
1601 <p> The SmallBitVector container provides the same interface as BitVector, but
1602 it is optimized for the case where only a small number of bits, less than
1603 25 or so, are needed. It also transparently supports larger bit counts, but
1604 slightly less efficiently than a plain BitVector, so SmallBitVector should
1605 only be used when larger counts are rare.
1606 </p>
1609 At this time, SmallBitVector does not support set operations (and, or, xor),
1610 and its operator[] does not provide an assignable lvalue.
1611 </p>
1612 </div>
1614 <!-- _______________________________________________________________________ -->
1615 <div class="doc_subsubsection">
1616 <a name="dss_sparsebitvector">SparseBitVector</a>
1617 </div>
1619 <div class="doc_text">
1620 <p> The SparseBitVector container is much like BitVector, with one major
1621 difference: Only the bits that are set, are stored. This makes the
1622 SparseBitVector much more space efficient than BitVector when the set is sparse,
1623 as well as making set operations O(number of set bits) instead of O(size of
1624 universe). The downside to the SparseBitVector is that setting and testing of random bits is O(N), and on large SparseBitVectors, this can be slower than BitVector. In our implementation, setting or testing bits in sorted order
1625 (either forwards or reverse) is O(1) worst case. Testing and setting bits within 128 bits (depends on size) of the current bit is also O(1). As a general statement, testing/setting bits in a SparseBitVector is O(distance away from last set bit).
1626 </p>
1627 </div>
1629 <!-- *********************************************************************** -->
1630 <div class="doc_section">
1631 <a name="common">Helpful Hints for Common Operations</a>
1632 </div>
1633 <!-- *********************************************************************** -->
1635 <div class="doc_text">
1637 <p>This section describes how to perform some very simple transformations of
1638 LLVM code. This is meant to give examples of common idioms used, showing the
1639 practical side of LLVM transformations. <p> Because this is a "how-to" section,
1640 you should also read about the main classes that you will be working with. The
1641 <a href="#coreclasses">Core LLVM Class Hierarchy Reference</a> contains details
1642 and descriptions of the main classes that you should know about.</p>
1644 </div>
1646 <!-- NOTE: this section should be heavy on example code -->
1647 <!-- ======================================================================= -->
1648 <div class="doc_subsection">
1649 <a name="inspection">Basic Inspection and Traversal Routines</a>
1650 </div>
1652 <div class="doc_text">
1654 <p>The LLVM compiler infrastructure have many different data structures that may
1655 be traversed. Following the example of the C++ standard template library, the
1656 techniques used to traverse these various data structures are all basically the
1657 same. For a enumerable sequence of values, the <tt>XXXbegin()</tt> function (or
1658 method) returns an iterator to the start of the sequence, the <tt>XXXend()</tt>
1659 function returns an iterator pointing to one past the last valid element of the
1660 sequence, and there is some <tt>XXXiterator</tt> data type that is common
1661 between the two operations.</p>
1663 <p>Because the pattern for iteration is common across many different aspects of
1664 the program representation, the standard template library algorithms may be used
1665 on them, and it is easier to remember how to iterate. First we show a few common
1666 examples of the data structures that need to be traversed. Other data
1667 structures are traversed in very similar ways.</p>
1669 </div>
1671 <!-- _______________________________________________________________________ -->
1672 <div class="doc_subsubsection">
1673 <a name="iterate_function">Iterating over the </a><a
1674 href="#BasicBlock"><tt>BasicBlock</tt></a>s in a <a
1675 href="#Function"><tt>Function</tt></a>
1676 </div>
1678 <div class="doc_text">
1680 <p>It's quite common to have a <tt>Function</tt> instance that you'd like to
1681 transform in some way; in particular, you'd like to manipulate its
1682 <tt>BasicBlock</tt>s. To facilitate this, you'll need to iterate over all of
1683 the <tt>BasicBlock</tt>s that constitute the <tt>Function</tt>. The following is
1684 an example that prints the name of a <tt>BasicBlock</tt> and the number of
1685 <tt>Instruction</tt>s it contains:</p>
1687 <div class="doc_code">
1688 <pre>
1689 // <i>func is a pointer to a Function instance</i>
1690 for (Function::iterator i = func-&gt;begin(), e = func-&gt;end(); i != e; ++i)
1691 // <i>Print out the name of the basic block if it has one, and then the</i>
1692 // <i>number of instructions that it contains</i>
1693 errs() &lt;&lt; "Basic block (name=" &lt;&lt; i-&gt;getName() &lt;&lt; ") has "
1694 &lt;&lt; i-&gt;size() &lt;&lt; " instructions.\n";
1695 </pre>
1696 </div>
1698 <p>Note that i can be used as if it were a pointer for the purposes of
1699 invoking member functions of the <tt>Instruction</tt> class. This is
1700 because the indirection operator is overloaded for the iterator
1701 classes. In the above code, the expression <tt>i-&gt;size()</tt> is
1702 exactly equivalent to <tt>(*i).size()</tt> just like you'd expect.</p>
1704 </div>
1706 <!-- _______________________________________________________________________ -->
1707 <div class="doc_subsubsection">
1708 <a name="iterate_basicblock">Iterating over the </a><a
1709 href="#Instruction"><tt>Instruction</tt></a>s in a <a
1710 href="#BasicBlock"><tt>BasicBlock</tt></a>
1711 </div>
1713 <div class="doc_text">
1715 <p>Just like when dealing with <tt>BasicBlock</tt>s in <tt>Function</tt>s, it's
1716 easy to iterate over the individual instructions that make up
1717 <tt>BasicBlock</tt>s. Here's a code snippet that prints out each instruction in
1718 a <tt>BasicBlock</tt>:</p>
1720 <div class="doc_code">
1721 <pre>
1722 // <i>blk is a pointer to a BasicBlock instance</i>
1723 for (BasicBlock::iterator i = blk-&gt;begin(), e = blk-&gt;end(); i != e; ++i)
1724 // <i>The next statement works since operator&lt;&lt;(ostream&amp;,...)</i>
1725 // <i>is overloaded for Instruction&amp;</i>
1726 errs() &lt;&lt; *i &lt;&lt; "\n";
1727 </pre>
1728 </div>
1730 <p>However, this isn't really the best way to print out the contents of a
1731 <tt>BasicBlock</tt>! Since the ostream operators are overloaded for virtually
1732 anything you'll care about, you could have just invoked the print routine on the
1733 basic block itself: <tt>errs() &lt;&lt; *blk &lt;&lt; "\n";</tt>.</p>
1735 </div>
1737 <!-- _______________________________________________________________________ -->
1738 <div class="doc_subsubsection">
1739 <a name="iterate_institer">Iterating over the </a><a
1740 href="#Instruction"><tt>Instruction</tt></a>s in a <a
1741 href="#Function"><tt>Function</tt></a>
1742 </div>
1744 <div class="doc_text">
1746 <p>If you're finding that you commonly iterate over a <tt>Function</tt>'s
1747 <tt>BasicBlock</tt>s and then that <tt>BasicBlock</tt>'s <tt>Instruction</tt>s,
1748 <tt>InstIterator</tt> should be used instead. You'll need to include <a
1749 href="/doxygen/InstIterator_8h-source.html"><tt>llvm/Support/InstIterator.h</tt></a>,
1750 and then instantiate <tt>InstIterator</tt>s explicitly in your code. Here's a
1751 small example that shows how to dump all instructions in a function to the standard error stream:<p>
1753 <div class="doc_code">
1754 <pre>
1755 #include "<a href="/doxygen/InstIterator_8h-source.html">llvm/Support/InstIterator.h</a>"
1757 // <i>F is a pointer to a Function instance</i>
1758 for (inst_iterator I = inst_begin(F), E = inst_end(F); I != E; ++I)
1759 errs() &lt;&lt; *I &lt;&lt; "\n";
1760 </pre>
1761 </div>
1763 <p>Easy, isn't it? You can also use <tt>InstIterator</tt>s to fill a
1764 work list with its initial contents. For example, if you wanted to
1765 initialize a work list to contain all instructions in a <tt>Function</tt>
1766 F, all you would need to do is something like:</p>
1768 <div class="doc_code">
1769 <pre>
1770 std::set&lt;Instruction*&gt; worklist;
1771 // or better yet, SmallPtrSet&lt;Instruction*, 64&gt; worklist;
1773 for (inst_iterator I = inst_begin(F), E = inst_end(F); I != E; ++I)
1774 worklist.insert(&amp;*I);
1775 </pre>
1776 </div>
1778 <p>The STL set <tt>worklist</tt> would now contain all instructions in the
1779 <tt>Function</tt> pointed to by F.</p>
1781 </div>
1783 <!-- _______________________________________________________________________ -->
1784 <div class="doc_subsubsection">
1785 <a name="iterate_convert">Turning an iterator into a class pointer (and
1786 vice-versa)</a>
1787 </div>
1789 <div class="doc_text">
1791 <p>Sometimes, it'll be useful to grab a reference (or pointer) to a class
1792 instance when all you've got at hand is an iterator. Well, extracting
1793 a reference or a pointer from an iterator is very straight-forward.
1794 Assuming that <tt>i</tt> is a <tt>BasicBlock::iterator</tt> and <tt>j</tt>
1795 is a <tt>BasicBlock::const_iterator</tt>:</p>
1797 <div class="doc_code">
1798 <pre>
1799 Instruction&amp; inst = *i; // <i>Grab reference to instruction reference</i>
1800 Instruction* pinst = &amp;*i; // <i>Grab pointer to instruction reference</i>
1801 const Instruction&amp; inst = *j;
1802 </pre>
1803 </div>
1805 <p>However, the iterators you'll be working with in the LLVM framework are
1806 special: they will automatically convert to a ptr-to-instance type whenever they
1807 need to. Instead of dereferencing the iterator and then taking the address of
1808 the result, you can simply assign the iterator to the proper pointer type and
1809 you get the dereference and address-of operation as a result of the assignment
1810 (behind the scenes, this is a result of overloading casting mechanisms). Thus
1811 the last line of the last example,</p>
1813 <div class="doc_code">
1814 <pre>
1815 Instruction *pinst = &amp;*i;
1816 </pre>
1817 </div>
1819 <p>is semantically equivalent to</p>
1821 <div class="doc_code">
1822 <pre>
1823 Instruction *pinst = i;
1824 </pre>
1825 </div>
1827 <p>It's also possible to turn a class pointer into the corresponding iterator,
1828 and this is a constant time operation (very efficient). The following code
1829 snippet illustrates use of the conversion constructors provided by LLVM
1830 iterators. By using these, you can explicitly grab the iterator of something
1831 without actually obtaining it via iteration over some structure:</p>
1833 <div class="doc_code">
1834 <pre>
1835 void printNextInstruction(Instruction* inst) {
1836 BasicBlock::iterator it(inst);
1837 ++it; // <i>After this line, it refers to the instruction after *inst</i>
1838 if (it != inst-&gt;getParent()-&gt;end()) errs() &lt;&lt; *it &lt;&lt; "\n";
1840 </pre>
1841 </div>
1843 <p>Unfortunately, these implicit conversions come at a cost; they prevent
1844 these iterators from conforming to standard iterator conventions, and thus
1845 from being usable with standard algorithms and containers. For example, they
1846 prevent the following code, where <tt>B</tt> is a <tt>BasicBlock</tt>,
1847 from compiling:</p>
1849 <div class="doc_code">
1850 <pre>
1851 llvm::SmallVector&lt;llvm::Instruction *, 16&gt;(B-&gt;begin(), B-&gt;end());
1852 </pre>
1853 </div>
1855 <p>Because of this, these implicit conversions may be removed some day,
1856 and <tt>operator*</tt> changed to return a pointer instead of a reference.</p>
1858 </div>
1860 <!--_______________________________________________________________________-->
1861 <div class="doc_subsubsection">
1862 <a name="iterate_complex">Finding call sites: a slightly more complex
1863 example</a>
1864 </div>
1866 <div class="doc_text">
1868 <p>Say that you're writing a FunctionPass and would like to count all the
1869 locations in the entire module (that is, across every <tt>Function</tt>) where a
1870 certain function (i.e., some <tt>Function</tt>*) is already in scope. As you'll
1871 learn later, you may want to use an <tt>InstVisitor</tt> to accomplish this in a
1872 much more straight-forward manner, but this example will allow us to explore how
1873 you'd do it if you didn't have <tt>InstVisitor</tt> around. In pseudo-code, this
1874 is what we want to do:</p>
1876 <div class="doc_code">
1877 <pre>
1878 initialize callCounter to zero
1879 for each Function f in the Module
1880 for each BasicBlock b in f
1881 for each Instruction i in b
1882 if (i is a CallInst and calls the given function)
1883 increment callCounter
1884 </pre>
1885 </div>
1887 <p>And the actual code is (remember, because we're writing a
1888 <tt>FunctionPass</tt>, our <tt>FunctionPass</tt>-derived class simply has to
1889 override the <tt>runOnFunction</tt> method):</p>
1891 <div class="doc_code">
1892 <pre>
1893 Function* targetFunc = ...;
1895 class OurFunctionPass : public FunctionPass {
1896 public:
1897 OurFunctionPass(): callCounter(0) { }
1899 virtual runOnFunction(Function&amp; F) {
1900 for (Function::iterator b = F.begin(), be = F.end(); b != be; ++b) {
1901 for (BasicBlock::iterator i = b-&gt;begin(), ie = b-&gt;end(); i != ie; ++i) {
1902 if (<a href="#CallInst">CallInst</a>* callInst = <a href="#isa">dyn_cast</a>&lt;<a
1903 href="#CallInst">CallInst</a>&gt;(&amp;*i)) {
1904 // <i>We know we've encountered a call instruction, so we</i>
1905 // <i>need to determine if it's a call to the</i>
1906 // <i>function pointed to by m_func or not.</i>
1907 if (callInst-&gt;getCalledFunction() == targetFunc)
1908 ++callCounter;
1914 private:
1915 unsigned callCounter;
1917 </pre>
1918 </div>
1920 </div>
1922 <!--_______________________________________________________________________-->
1923 <div class="doc_subsubsection">
1924 <a name="calls_and_invokes">Treating calls and invokes the same way</a>
1925 </div>
1927 <div class="doc_text">
1929 <p>You may have noticed that the previous example was a bit oversimplified in
1930 that it did not deal with call sites generated by 'invoke' instructions. In
1931 this, and in other situations, you may find that you want to treat
1932 <tt>CallInst</tt>s and <tt>InvokeInst</tt>s the same way, even though their
1933 most-specific common base class is <tt>Instruction</tt>, which includes lots of
1934 less closely-related things. For these cases, LLVM provides a handy wrapper
1935 class called <a
1936 href="http://llvm.org/doxygen/classllvm_1_1CallSite.html"><tt>CallSite</tt></a>.
1937 It is essentially a wrapper around an <tt>Instruction</tt> pointer, with some
1938 methods that provide functionality common to <tt>CallInst</tt>s and
1939 <tt>InvokeInst</tt>s.</p>
1941 <p>This class has "value semantics": it should be passed by value, not by
1942 reference and it should not be dynamically allocated or deallocated using
1943 <tt>operator new</tt> or <tt>operator delete</tt>. It is efficiently copyable,
1944 assignable and constructable, with costs equivalents to that of a bare pointer.
1945 If you look at its definition, it has only a single pointer member.</p>
1947 </div>
1949 <!--_______________________________________________________________________-->
1950 <div class="doc_subsubsection">
1951 <a name="iterate_chains">Iterating over def-use &amp; use-def chains</a>
1952 </div>
1954 <div class="doc_text">
1956 <p>Frequently, we might have an instance of the <a
1957 href="/doxygen/classllvm_1_1Value.html">Value Class</a> and we want to
1958 determine which <tt>User</tt>s use the <tt>Value</tt>. The list of all
1959 <tt>User</tt>s of a particular <tt>Value</tt> is called a <i>def-use</i> chain.
1960 For example, let's say we have a <tt>Function*</tt> named <tt>F</tt> to a
1961 particular function <tt>foo</tt>. Finding all of the instructions that
1962 <i>use</i> <tt>foo</tt> is as simple as iterating over the <i>def-use</i> chain
1963 of <tt>F</tt>:</p>
1965 <div class="doc_code">
1966 <pre>
1967 Function *F = ...;
1969 for (Value::use_iterator i = F-&gt;use_begin(), e = F-&gt;use_end(); i != e; ++i)
1970 if (Instruction *Inst = dyn_cast&lt;Instruction&gt;(*i)) {
1971 errs() &lt;&lt; "F is used in instruction:\n";
1972 errs() &lt;&lt; *Inst &lt;&lt; "\n";
1974 </pre>
1975 </div>
1977 <p>Note that dereferencing a <tt>Value::use_iterator</tt> is not a very cheap
1978 operation. Instead of performing <tt>*i</tt> above several times, consider
1979 doing it only once in the loop body and reusing its result.</p>
1981 <p>Alternatively, it's common to have an instance of the <a
1982 href="/doxygen/classllvm_1_1User.html">User Class</a> and need to know what
1983 <tt>Value</tt>s are used by it. The list of all <tt>Value</tt>s used by a
1984 <tt>User</tt> is known as a <i>use-def</i> chain. Instances of class
1985 <tt>Instruction</tt> are common <tt>User</tt>s, so we might want to iterate over
1986 all of the values that a particular instruction uses (that is, the operands of
1987 the particular <tt>Instruction</tt>):</p>
1989 <div class="doc_code">
1990 <pre>
1991 Instruction *pi = ...;
1993 for (User::op_iterator i = pi-&gt;op_begin(), e = pi-&gt;op_end(); i != e; ++i) {
1994 Value *v = *i;
1995 // <i>...</i>
1997 </pre>
1998 </div>
2000 <p>Declaring objects as <tt>const</tt> is an important tool of enforcing
2001 mutation free algorithms (such as analyses, etc.). For this purpose above
2002 iterators come in constant flavors as <tt>Value::const_use_iterator</tt>
2003 and <tt>Value::const_op_iterator</tt>. They automatically arise when
2004 calling <tt>use/op_begin()</tt> on <tt>const Value*</tt>s or
2005 <tt>const User*</tt>s respectively. Upon dereferencing, they return
2006 <tt>const Use*</tt>s. Otherwise the above patterns remain unchanged.</p>
2008 </div>
2010 <!--_______________________________________________________________________-->
2011 <div class="doc_subsubsection">
2012 <a name="iterate_preds">Iterating over predecessors &amp;
2013 successors of blocks</a>
2014 </div>
2016 <div class="doc_text">
2018 <p>Iterating over the predecessors and successors of a block is quite easy
2019 with the routines defined in <tt>"llvm/Support/CFG.h"</tt>. Just use code like
2020 this to iterate over all predecessors of BB:</p>
2022 <div class="doc_code">
2023 <pre>
2024 #include "llvm/Support/CFG.h"
2025 BasicBlock *BB = ...;
2027 for (pred_iterator PI = pred_begin(BB), E = pred_end(BB); PI != E; ++PI) {
2028 BasicBlock *Pred = *PI;
2029 // <i>...</i>
2031 </pre>
2032 </div>
2034 <p>Similarly, to iterate over successors use
2035 succ_iterator/succ_begin/succ_end.</p>
2037 </div>
2040 <!-- ======================================================================= -->
2041 <div class="doc_subsection">
2042 <a name="simplechanges">Making simple changes</a>
2043 </div>
2045 <div class="doc_text">
2047 <p>There are some primitive transformation operations present in the LLVM
2048 infrastructure that are worth knowing about. When performing
2049 transformations, it's fairly common to manipulate the contents of basic
2050 blocks. This section describes some of the common methods for doing so
2051 and gives example code.</p>
2053 </div>
2055 <!--_______________________________________________________________________-->
2056 <div class="doc_subsubsection">
2057 <a name="schanges_creating">Creating and inserting new
2058 <tt>Instruction</tt>s</a>
2059 </div>
2061 <div class="doc_text">
2063 <p><i>Instantiating Instructions</i></p>
2065 <p>Creation of <tt>Instruction</tt>s is straight-forward: simply call the
2066 constructor for the kind of instruction to instantiate and provide the necessary
2067 parameters. For example, an <tt>AllocaInst</tt> only <i>requires</i> a
2068 (const-ptr-to) <tt>Type</tt>. Thus:</p>
2070 <div class="doc_code">
2071 <pre>
2072 AllocaInst* ai = new AllocaInst(Type::Int32Ty);
2073 </pre>
2074 </div>
2076 <p>will create an <tt>AllocaInst</tt> instance that represents the allocation of
2077 one integer in the current stack frame, at run time. Each <tt>Instruction</tt>
2078 subclass is likely to have varying default parameters which change the semantics
2079 of the instruction, so refer to the <a
2080 href="/doxygen/classllvm_1_1Instruction.html">doxygen documentation for the subclass of
2081 Instruction</a> that you're interested in instantiating.</p>
2083 <p><i>Naming values</i></p>
2085 <p>It is very useful to name the values of instructions when you're able to, as
2086 this facilitates the debugging of your transformations. If you end up looking
2087 at generated LLVM machine code, you definitely want to have logical names
2088 associated with the results of instructions! By supplying a value for the
2089 <tt>Name</tt> (default) parameter of the <tt>Instruction</tt> constructor, you
2090 associate a logical name with the result of the instruction's execution at
2091 run time. For example, say that I'm writing a transformation that dynamically
2092 allocates space for an integer on the stack, and that integer is going to be
2093 used as some kind of index by some other code. To accomplish this, I place an
2094 <tt>AllocaInst</tt> at the first point in the first <tt>BasicBlock</tt> of some
2095 <tt>Function</tt>, and I'm intending to use it within the same
2096 <tt>Function</tt>. I might do:</p>
2098 <div class="doc_code">
2099 <pre>
2100 AllocaInst* pa = new AllocaInst(Type::Int32Ty, 0, "indexLoc");
2101 </pre>
2102 </div>
2104 <p>where <tt>indexLoc</tt> is now the logical name of the instruction's
2105 execution value, which is a pointer to an integer on the run time stack.</p>
2107 <p><i>Inserting instructions</i></p>
2109 <p>There are essentially two ways to insert an <tt>Instruction</tt>
2110 into an existing sequence of instructions that form a <tt>BasicBlock</tt>:</p>
2112 <ul>
2113 <li>Insertion into an explicit instruction list
2115 <p>Given a <tt>BasicBlock* pb</tt>, an <tt>Instruction* pi</tt> within that
2116 <tt>BasicBlock</tt>, and a newly-created instruction we wish to insert
2117 before <tt>*pi</tt>, we do the following: </p>
2119 <div class="doc_code">
2120 <pre>
2121 BasicBlock *pb = ...;
2122 Instruction *pi = ...;
2123 Instruction *newInst = new Instruction(...);
2125 pb-&gt;getInstList().insert(pi, newInst); // <i>Inserts newInst before pi in pb</i>
2126 </pre>
2127 </div>
2129 <p>Appending to the end of a <tt>BasicBlock</tt> is so common that
2130 the <tt>Instruction</tt> class and <tt>Instruction</tt>-derived
2131 classes provide constructors which take a pointer to a
2132 <tt>BasicBlock</tt> to be appended to. For example code that
2133 looked like: </p>
2135 <div class="doc_code">
2136 <pre>
2137 BasicBlock *pb = ...;
2138 Instruction *newInst = new Instruction(...);
2140 pb-&gt;getInstList().push_back(newInst); // <i>Appends newInst to pb</i>
2141 </pre>
2142 </div>
2144 <p>becomes: </p>
2146 <div class="doc_code">
2147 <pre>
2148 BasicBlock *pb = ...;
2149 Instruction *newInst = new Instruction(..., pb);
2150 </pre>
2151 </div>
2153 <p>which is much cleaner, especially if you are creating
2154 long instruction streams.</p></li>
2156 <li>Insertion into an implicit instruction list
2158 <p><tt>Instruction</tt> instances that are already in <tt>BasicBlock</tt>s
2159 are implicitly associated with an existing instruction list: the instruction
2160 list of the enclosing basic block. Thus, we could have accomplished the same
2161 thing as the above code without being given a <tt>BasicBlock</tt> by doing:
2162 </p>
2164 <div class="doc_code">
2165 <pre>
2166 Instruction *pi = ...;
2167 Instruction *newInst = new Instruction(...);
2169 pi-&gt;getParent()-&gt;getInstList().insert(pi, newInst);
2170 </pre>
2171 </div>
2173 <p>In fact, this sequence of steps occurs so frequently that the
2174 <tt>Instruction</tt> class and <tt>Instruction</tt>-derived classes provide
2175 constructors which take (as a default parameter) a pointer to an
2176 <tt>Instruction</tt> which the newly-created <tt>Instruction</tt> should
2177 precede. That is, <tt>Instruction</tt> constructors are capable of
2178 inserting the newly-created instance into the <tt>BasicBlock</tt> of a
2179 provided instruction, immediately before that instruction. Using an
2180 <tt>Instruction</tt> constructor with a <tt>insertBefore</tt> (default)
2181 parameter, the above code becomes:</p>
2183 <div class="doc_code">
2184 <pre>
2185 Instruction* pi = ...;
2186 Instruction* newInst = new Instruction(..., pi);
2187 </pre>
2188 </div>
2190 <p>which is much cleaner, especially if you're creating a lot of
2191 instructions and adding them to <tt>BasicBlock</tt>s.</p></li>
2192 </ul>
2194 </div>
2196 <!--_______________________________________________________________________-->
2197 <div class="doc_subsubsection">
2198 <a name="schanges_deleting">Deleting <tt>Instruction</tt>s</a>
2199 </div>
2201 <div class="doc_text">
2203 <p>Deleting an instruction from an existing sequence of instructions that form a
2204 <a href="#BasicBlock"><tt>BasicBlock</tt></a> is very straight-forward. First,
2205 you must have a pointer to the instruction that you wish to delete. Second, you
2206 need to obtain the pointer to that instruction's basic block. You use the
2207 pointer to the basic block to get its list of instructions and then use the
2208 erase function to remove your instruction. For example:</p>
2210 <div class="doc_code">
2211 <pre>
2212 <a href="#Instruction">Instruction</a> *I = .. ;
2213 I-&gt;eraseFromParent();
2214 </pre>
2215 </div>
2217 </div>
2219 <!--_______________________________________________________________________-->
2220 <div class="doc_subsubsection">
2221 <a name="schanges_replacing">Replacing an <tt>Instruction</tt> with another
2222 <tt>Value</tt></a>
2223 </div>
2225 <div class="doc_text">
2227 <p><i>Replacing individual instructions</i></p>
2229 <p>Including "<a href="/doxygen/BasicBlockUtils_8h-source.html">llvm/Transforms/Utils/BasicBlockUtils.h</a>"
2230 permits use of two very useful replace functions: <tt>ReplaceInstWithValue</tt>
2231 and <tt>ReplaceInstWithInst</tt>.</p>
2233 <h4><a name="schanges_deleting">Deleting <tt>Instruction</tt>s</a></h4>
2235 <ul>
2236 <li><tt>ReplaceInstWithValue</tt>
2238 <p>This function replaces all uses of a given instruction with a value,
2239 and then removes the original instruction. The following example
2240 illustrates the replacement of the result of a particular
2241 <tt>AllocaInst</tt> that allocates memory for a single integer with a null
2242 pointer to an integer.</p>
2244 <div class="doc_code">
2245 <pre>
2246 AllocaInst* instToReplace = ...;
2247 BasicBlock::iterator ii(instToReplace);
2249 ReplaceInstWithValue(instToReplace-&gt;getParent()-&gt;getInstList(), ii,
2250 Constant::getNullValue(PointerType::getUnqual(Type::Int32Ty)));
2251 </pre></div></li>
2253 <li><tt>ReplaceInstWithInst</tt>
2255 <p>This function replaces a particular instruction with another
2256 instruction, inserting the new instruction into the basic block at the
2257 location where the old instruction was, and replacing any uses of the old
2258 instruction with the new instruction. The following example illustrates
2259 the replacement of one <tt>AllocaInst</tt> with another.</p>
2261 <div class="doc_code">
2262 <pre>
2263 AllocaInst* instToReplace = ...;
2264 BasicBlock::iterator ii(instToReplace);
2266 ReplaceInstWithInst(instToReplace-&gt;getParent()-&gt;getInstList(), ii,
2267 new AllocaInst(Type::Int32Ty, 0, "ptrToReplacedInt"));
2268 </pre></div></li>
2269 </ul>
2271 <p><i>Replacing multiple uses of <tt>User</tt>s and <tt>Value</tt>s</i></p>
2273 <p>You can use <tt>Value::replaceAllUsesWith</tt> and
2274 <tt>User::replaceUsesOfWith</tt> to change more than one use at a time. See the
2275 doxygen documentation for the <a href="/doxygen/classllvm_1_1Value.html">Value Class</a>
2276 and <a href="/doxygen/classllvm_1_1User.html">User Class</a>, respectively, for more
2277 information.</p>
2279 <!-- Value::replaceAllUsesWith User::replaceUsesOfWith Point out:
2280 include/llvm/Transforms/Utils/ especially BasicBlockUtils.h with:
2281 ReplaceInstWithValue, ReplaceInstWithInst -->
2283 </div>
2285 <!--_______________________________________________________________________-->
2286 <div class="doc_subsubsection">
2287 <a name="schanges_deletingGV">Deleting <tt>GlobalVariable</tt>s</a>
2288 </div>
2290 <div class="doc_text">
2292 <p>Deleting a global variable from a module is just as easy as deleting an
2293 Instruction. First, you must have a pointer to the global variable that you wish
2294 to delete. You use this pointer to erase it from its parent, the module.
2295 For example:</p>
2297 <div class="doc_code">
2298 <pre>
2299 <a href="#GlobalVariable">GlobalVariable</a> *GV = .. ;
2301 GV-&gt;eraseFromParent();
2302 </pre>
2303 </div>
2305 </div>
2307 <!-- ======================================================================= -->
2308 <div class="doc_subsection">
2309 <a name="create_types">How to Create Types</a>
2310 </div>
2312 <div class="doc_text">
2314 <p>In generating IR, you may need some complex types. If you know these types
2315 statically, you can use <tt>TypeBuilder&lt;...&gt;::get()</tt>, defined
2316 in <tt>llvm/Support/TypeBuilder.h</tt>, to retrieve them. <tt>TypeBuilder</tt>
2317 has two forms depending on whether you're building types for cross-compilation
2318 or native library use. <tt>TypeBuilder&lt;T, true&gt;</tt> requires
2319 that <tt>T</tt> be independent of the host environment, meaning that it's built
2320 out of types from
2321 the <a href="/doxygen/namespacellvm_1_1types.html"><tt>llvm::types</tt></a>
2322 namespace and pointers, functions, arrays, etc. built of
2323 those. <tt>TypeBuilder&lt;T, false&gt;</tt> additionally allows native C types
2324 whose size may depend on the host compiler. For example,</p>
2326 <div class="doc_code">
2327 <pre>
2328 FunctionType *ft = TypeBuilder&lt;types::i&lt;8&gt;(types::i&lt;32&gt;*), true&gt;::get();
2329 </pre>
2330 </div>
2332 <p>is easier to read and write than the equivalent</p>
2334 <div class="doc_code">
2335 <pre>
2336 std::vector&lt;const Type*&gt; params;
2337 params.push_back(PointerType::getUnqual(Type::Int32Ty));
2338 FunctionType *ft = FunctionType::get(Type::Int8Ty, params, false);
2339 </pre>
2340 </div>
2342 <p>See the <a href="/doxygen/TypeBuilder_8h-source.html#l00001">class
2343 comment</a> for more details.</p>
2345 </div>
2347 <!-- *********************************************************************** -->
2348 <div class="doc_section">
2349 <a name="threading">Threads and LLVM</a>
2350 </div>
2351 <!-- *********************************************************************** -->
2353 <div class="doc_text">
2355 This section describes the interaction of the LLVM APIs with multithreading,
2356 both on the part of client applications, and in the JIT, in the hosted
2357 application.
2358 </p>
2361 Note that LLVM's support for multithreading is still relatively young. Up
2362 through version 2.5, the execution of threaded hosted applications was
2363 supported, but not threaded client access to the APIs. While this use case is
2364 now supported, clients <em>must</em> adhere to the guidelines specified below to
2365 ensure proper operation in multithreaded mode.
2366 </p>
2369 Note that, on Unix-like platforms, LLVM requires the presence of GCC's atomic
2370 intrinsics in order to support threaded operation. If you need a
2371 multhreading-capable LLVM on a platform without a suitably modern system
2372 compiler, consider compiling LLVM and LLVM-GCC in single-threaded mode, and
2373 using the resultant compiler to build a copy of LLVM with multithreading
2374 support.
2375 </p>
2376 </div>
2378 <!-- ======================================================================= -->
2379 <div class="doc_subsection">
2380 <a name="startmultithreaded">Entering and Exiting Multithreaded Mode</a>
2381 </div>
2383 <div class="doc_text">
2386 In order to properly protect its internal data structures while avoiding
2387 excessive locking overhead in the single-threaded case, the LLVM must intialize
2388 certain data structures necessary to provide guards around its internals. To do
2389 so, the client program must invoke <tt>llvm_start_multithreaded()</tt> before
2390 making any concurrent LLVM API calls. To subsequently tear down these
2391 structures, use the <tt>llvm_stop_multithreaded()</tt> call. You can also use
2392 the <tt>llvm_is_multithreaded()</tt> call to check the status of multithreaded
2393 mode.
2394 </p>
2397 Note that both of these calls must be made <em>in isolation</em>. That is to
2398 say that no other LLVM API calls may be executing at any time during the
2399 execution of <tt>llvm_start_multithreaded()</tt> or <tt>llvm_stop_multithreaded
2400 </tt>. It's is the client's responsibility to enforce this isolation.
2401 </p>
2404 The return value of <tt>llvm_start_multithreaded()</tt> indicates the success or
2405 failure of the initialization. Failure typically indicates that your copy of
2406 LLVM was built without multithreading support, typically because GCC atomic
2407 intrinsics were not found in your system compiler. In this case, the LLVM API
2408 will not be safe for concurrent calls. However, it <em>will</em> be safe for
2409 hosting threaded applications in the JIT, though <a href="#jitthreading">care
2410 must be taken</a> to ensure that side exits and the like do not accidentally
2411 result in concurrent LLVM API calls.
2412 </p>
2413 </div>
2415 <!-- ======================================================================= -->
2416 <div class="doc_subsection">
2417 <a name="shutdown">Ending Execution with <tt>llvm_shutdown()</tt></a>
2418 </div>
2420 <div class="doc_text">
2422 When you are done using the LLVM APIs, you should call <tt>llvm_shutdown()</tt>
2423 to deallocate memory used for internal structures. This will also invoke
2424 <tt>llvm_stop_multithreaded()</tt> if LLVM is operating in multithreaded mode.
2425 As such, <tt>llvm_shutdown()</tt> requires the same isolation guarantees as
2426 <tt>llvm_stop_multithreaded()</tt>.
2427 </p>
2430 Note that, if you use scope-based shutdown, you can use the
2431 <tt>llvm_shutdown_obj</tt> class, which calls <tt>llvm_shutdown()</tt> in its
2432 destructor.
2433 </div>
2435 <!-- ======================================================================= -->
2436 <div class="doc_subsection">
2437 <a name="managedstatic">Lazy Initialization with <tt>ManagedStatic</tt></a>
2438 </div>
2440 <div class="doc_text">
2442 <tt>ManagedStatic</tt> is a utility class in LLVM used to implement static
2443 initialization of static resources, such as the global type tables. Before the
2444 invocation of <tt>llvm_shutdown()</tt>, it implements a simple lazy
2445 initialization scheme. Once <tt>llvm_start_multithreaded()</tt> returns,
2446 however, it uses double-checked locking to implement thread-safe lazy
2447 initialization.
2448 </p>
2451 Note that, because no other threads are allowed to issue LLVM API calls before
2452 <tt>llvm_start_multithreaded()</tt> returns, it is possible to have
2453 <tt>ManagedStatic</tt>s of <tt>llvm::sys::Mutex</tt>s.
2454 </p>
2457 The <tt>llvm_acquire_global_lock()</tt> and <tt>llvm_release_global_lock</tt>
2458 APIs provide access to the global lock used to implement the double-checked
2459 locking for lazy initialization. These should only be used internally to LLVM,
2460 and only if you know what you're doing!
2461 </p>
2462 </div>
2464 <!-- ======================================================================= -->
2465 <div class="doc_subsection">
2466 <a name="llvmcontext">Achieving Isolation with <tt>LLVMContext</tt></a>
2467 </div>
2469 <div class="doc_text">
2471 <tt>LLVMContext</tt> is an opaque class in the LLVM API which clients can use
2472 to operate multiple, isolated instances of LLVM concurrently within the same
2473 address space. For instance, in a hypothetical compile-server, the compilation
2474 of an individual translation unit is conceptually independent from all the
2475 others, and it would be desirable to be able to compile incoming translation
2476 units concurrently on independent server threads. Fortunately,
2477 <tt>LLVMContext</tt> exists to enable just this kind of scenario!
2478 </p>
2481 Conceptually, <tt>LLVMContext</tt> provides isolation. Every LLVM entity
2482 (<tt>Module</tt>s, <tt>Value</tt>s, <tt>Type</tt>s, <tt>Constant</tt>s, etc.)
2483 in LLVM's in-memory IR belongs to an <tt>LLVMContext</tt>. Entities in
2484 different contexts <em>cannot</em> interact with each other: <tt>Module</tt>s in
2485 different contexts cannot be linked together, <tt>Function</tt>s cannot be added
2486 to <tt>Module</tt>s in different contexts, etc. What this means is that is is
2487 safe to compile on multiple threads simultaneously, as long as no two threads
2488 operate on entities within the same context.
2489 </p>
2492 In practice, very few places in the API require the explicit specification of a
2493 <tt>LLVMContext</tt>, other than the <tt>Type</tt> creation/lookup APIs.
2494 Because every <tt>Type</tt> carries a reference to its owning context, most
2495 other entities can determine what context they belong to by looking at their
2496 own <tt>Type</tt>. If you are adding new entities to LLVM IR, please try to
2497 maintain this interface design.
2498 </p>
2501 For clients that do <em>not</em> require the benefits of isolation, LLVM
2502 provides a convenience API <tt>getGlobalContext()</tt>. This returns a global,
2503 lazily initialized <tt>LLVMContext</tt> that may be used in situations where
2504 isolation is not a concern.
2505 </p>
2506 </div>
2508 <!-- ======================================================================= -->
2509 <div class="doc_subsection">
2510 <a name="jitthreading">Threads and the JIT</a>
2511 </div>
2513 <div class="doc_text">
2515 LLVM's "eager" JIT compiler is safe to use in threaded programs. Multiple
2516 threads can call <tt>ExecutionEngine::getPointerToFunction()</tt> or
2517 <tt>ExecutionEngine::runFunction()</tt> concurrently, and multiple threads can
2518 run code output by the JIT concurrently. The user must still ensure that only
2519 one thread accesses IR in a given <tt>LLVMContext</tt> while another thread
2520 might be modifying it. One way to do that is to always hold the JIT lock while
2521 accessing IR outside the JIT (the JIT <em>modifies</em> the IR by adding
2522 <tt>CallbackVH</tt>s). Another way is to only
2523 call <tt>getPointerToFunction()</tt> from the <tt>LLVMContext</tt>'s thread.
2524 </p>
2526 <p>When the JIT is configured to compile lazily (using
2527 <tt>ExecutionEngine::DisableLazyCompilation(false)</tt>), there is currently a
2528 <a href="http://llvm.org/bugs/show_bug.cgi?id=5184">race condition</a> in
2529 updating call sites after a function is lazily-jitted. It's still possible to
2530 use the lazy JIT in a threaded program if you ensure that only one thread at a
2531 time can call any particular lazy stub and that the JIT lock guards any IR
2532 access, but we suggest using only the eager JIT in threaded programs.
2533 </p>
2534 </div>
2536 <!-- *********************************************************************** -->
2537 <div class="doc_section">
2538 <a name="advanced">Advanced Topics</a>
2539 </div>
2540 <!-- *********************************************************************** -->
2542 <div class="doc_text">
2544 This section describes some of the advanced or obscure API's that most clients
2545 do not need to be aware of. These API's tend manage the inner workings of the
2546 LLVM system, and only need to be accessed in unusual circumstances.
2547 </p>
2548 </div>
2550 <!-- ======================================================================= -->
2551 <div class="doc_subsection">
2552 <a name="TypeResolve">LLVM Type Resolution</a>
2553 </div>
2555 <div class="doc_text">
2558 The LLVM type system has a very simple goal: allow clients to compare types for
2559 structural equality with a simple pointer comparison (aka a shallow compare).
2560 This goal makes clients much simpler and faster, and is used throughout the LLVM
2561 system.
2562 </p>
2565 Unfortunately achieving this goal is not a simple matter. In particular,
2566 recursive types and late resolution of opaque types makes the situation very
2567 difficult to handle. Fortunately, for the most part, our implementation makes
2568 most clients able to be completely unaware of the nasty internal details. The
2569 primary case where clients are exposed to the inner workings of it are when
2570 building a recursive type. In addition to this case, the LLVM bitcode reader,
2571 assembly parser, and linker also have to be aware of the inner workings of this
2572 system.
2573 </p>
2576 For our purposes below, we need three concepts. First, an "Opaque Type" is
2577 exactly as defined in the <a href="LangRef.html#t_opaque">language
2578 reference</a>. Second an "Abstract Type" is any type which includes an
2579 opaque type as part of its type graph (for example "<tt>{ opaque, i32 }</tt>").
2580 Third, a concrete type is a type that is not an abstract type (e.g. "<tt>{ i32,
2581 float }</tt>").
2582 </p>
2584 </div>
2586 <!-- ______________________________________________________________________ -->
2587 <div class="doc_subsubsection">
2588 <a name="BuildRecType">Basic Recursive Type Construction</a>
2589 </div>
2591 <div class="doc_text">
2594 Because the most common question is "how do I build a recursive type with LLVM",
2595 we answer it now and explain it as we go. Here we include enough to cause this
2596 to be emitted to an output .ll file:
2597 </p>
2599 <div class="doc_code">
2600 <pre>
2601 %mylist = type { %mylist*, i32 }
2602 </pre>
2603 </div>
2606 To build this, use the following LLVM APIs:
2607 </p>
2609 <div class="doc_code">
2610 <pre>
2611 // <i>Create the initial outer struct</i>
2612 <a href="#PATypeHolder">PATypeHolder</a> StructTy = OpaqueType::get();
2613 std::vector&lt;const Type*&gt; Elts;
2614 Elts.push_back(PointerType::getUnqual(StructTy));
2615 Elts.push_back(Type::Int32Ty);
2616 StructType *NewSTy = StructType::get(Elts);
2618 // <i>At this point, NewSTy = "{ opaque*, i32 }". Tell VMCore that</i>
2619 // <i>the struct and the opaque type are actually the same.</i>
2620 cast&lt;OpaqueType&gt;(StructTy.get())-&gt;<a href="#refineAbstractTypeTo">refineAbstractTypeTo</a>(NewSTy);
2622 // <i>NewSTy is potentially invalidated, but StructTy (a <a href="#PATypeHolder">PATypeHolder</a>) is</i>
2623 // <i>kept up-to-date</i>
2624 NewSTy = cast&lt;StructType&gt;(StructTy.get());
2626 // <i>Add a name for the type to the module symbol table (optional)</i>
2627 MyModule-&gt;addTypeName("mylist", NewSTy);
2628 </pre>
2629 </div>
2632 This code shows the basic approach used to build recursive types: build a
2633 non-recursive type using 'opaque', then use type unification to close the cycle.
2634 The type unification step is performed by the <tt><a
2635 href="#refineAbstractTypeTo">refineAbstractTypeTo</a></tt> method, which is
2636 described next. After that, we describe the <a
2637 href="#PATypeHolder">PATypeHolder class</a>.
2638 </p>
2640 </div>
2642 <!-- ______________________________________________________________________ -->
2643 <div class="doc_subsubsection">
2644 <a name="refineAbstractTypeTo">The <tt>refineAbstractTypeTo</tt> method</a>
2645 </div>
2647 <div class="doc_text">
2649 The <tt>refineAbstractTypeTo</tt> method starts the type unification process.
2650 While this method is actually a member of the DerivedType class, it is most
2651 often used on OpaqueType instances. Type unification is actually a recursive
2652 process. After unification, types can become structurally isomorphic to
2653 existing types, and all duplicates are deleted (to preserve pointer equality).
2654 </p>
2657 In the example above, the OpaqueType object is definitely deleted.
2658 Additionally, if there is an "{ \2*, i32}" type already created in the system,
2659 the pointer and struct type created are <b>also</b> deleted. Obviously whenever
2660 a type is deleted, any "Type*" pointers in the program are invalidated. As
2661 such, it is safest to avoid having <i>any</i> "Type*" pointers to abstract types
2662 live across a call to <tt>refineAbstractTypeTo</tt> (note that non-abstract
2663 types can never move or be deleted). To deal with this, the <a
2664 href="#PATypeHolder">PATypeHolder</a> class is used to maintain a stable
2665 reference to a possibly refined type, and the <a
2666 href="#AbstractTypeUser">AbstractTypeUser</a> class is used to update more
2667 complex datastructures.
2668 </p>
2670 </div>
2672 <!-- ______________________________________________________________________ -->
2673 <div class="doc_subsubsection">
2674 <a name="PATypeHolder">The PATypeHolder Class</a>
2675 </div>
2677 <div class="doc_text">
2679 PATypeHolder is a form of a "smart pointer" for Type objects. When VMCore
2680 happily goes about nuking types that become isomorphic to existing types, it
2681 automatically updates all PATypeHolder objects to point to the new type. In the
2682 example above, this allows the code to maintain a pointer to the resultant
2683 resolved recursive type, even though the Type*'s are potentially invalidated.
2684 </p>
2687 PATypeHolder is an extremely light-weight object that uses a lazy union-find
2688 implementation to update pointers. For example the pointer from a Value to its
2689 Type is maintained by PATypeHolder objects.
2690 </p>
2692 </div>
2694 <!-- ______________________________________________________________________ -->
2695 <div class="doc_subsubsection">
2696 <a name="AbstractTypeUser">The AbstractTypeUser Class</a>
2697 </div>
2699 <div class="doc_text">
2702 Some data structures need more to perform more complex updates when types get
2703 resolved. To support this, a class can derive from the AbstractTypeUser class.
2704 This class
2705 allows it to get callbacks when certain types are resolved. To register to get
2706 callbacks for a particular type, the DerivedType::{add/remove}AbstractTypeUser
2707 methods can be called on a type. Note that these methods only work for <i>
2708 abstract</i> types. Concrete types (those that do not include any opaque
2709 objects) can never be refined.
2710 </p>
2711 </div>
2714 <!-- ======================================================================= -->
2715 <div class="doc_subsection">
2716 <a name="SymbolTable">The <tt>ValueSymbolTable</tt> and
2717 <tt>TypeSymbolTable</tt> classes</a>
2718 </div>
2720 <div class="doc_text">
2721 <p>The <tt><a href="http://llvm.org/doxygen/classllvm_1_1ValueSymbolTable.html">
2722 ValueSymbolTable</a></tt> class provides a symbol table that the <a
2723 href="#Function"><tt>Function</tt></a> and <a href="#Module">
2724 <tt>Module</tt></a> classes use for naming value definitions. The symbol table
2725 can provide a name for any <a href="#Value"><tt>Value</tt></a>.
2726 The <tt><a href="http://llvm.org/doxygen/classllvm_1_1TypeSymbolTable.html">
2727 TypeSymbolTable</a></tt> class is used by the <tt>Module</tt> class to store
2728 names for types.</p>
2730 <p>Note that the <tt>SymbolTable</tt> class should not be directly accessed
2731 by most clients. It should only be used when iteration over the symbol table
2732 names themselves are required, which is very special purpose. Note that not
2733 all LLVM
2734 <tt><a href="#Value">Value</a></tt>s have names, and those without names (i.e. they have
2735 an empty name) do not exist in the symbol table.
2736 </p>
2738 <p>These symbol tables support iteration over the values/types in the symbol
2739 table with <tt>begin/end/iterator</tt> and supports querying to see if a
2740 specific name is in the symbol table (with <tt>lookup</tt>). The
2741 <tt>ValueSymbolTable</tt> class exposes no public mutator methods, instead,
2742 simply call <tt>setName</tt> on a value, which will autoinsert it into the
2743 appropriate symbol table. For types, use the Module::addTypeName method to
2744 insert entries into the symbol table.</p>
2746 </div>
2750 <!-- ======================================================================= -->
2751 <div class="doc_subsection">
2752 <a name="UserLayout">The <tt>User</tt> and owned <tt>Use</tt> classes' memory layout</a>
2753 </div>
2755 <div class="doc_text">
2756 <p>The <tt><a href="http://llvm.org/doxygen/classllvm_1_1User.html">
2757 User</a></tt> class provides a basis for expressing the ownership of <tt>User</tt>
2758 towards other <tt><a href="http://llvm.org/doxygen/classllvm_1_1Value.html">
2759 Value</a></tt>s. The <tt><a href="http://llvm.org/doxygen/classllvm_1_1Use.html">
2760 Use</a></tt> helper class is employed to do the bookkeeping and to facilitate <i>O(1)</i>
2761 addition and removal.</p>
2763 <!-- ______________________________________________________________________ -->
2764 <div class="doc_subsubsection">
2765 <a name="Use2User">Interaction and relationship between <tt>User</tt> and <tt>Use</tt> objects</a>
2766 </div>
2768 <div class="doc_text">
2770 A subclass of <tt>User</tt> can choose between incorporating its <tt>Use</tt> objects
2771 or refer to them out-of-line by means of a pointer. A mixed variant
2772 (some <tt>Use</tt>s inline others hung off) is impractical and breaks the invariant
2773 that the <tt>Use</tt> objects belonging to the same <tt>User</tt> form a contiguous array.
2774 </p>
2775 </div>
2778 We have 2 different layouts in the <tt>User</tt> (sub)classes:
2779 <ul>
2780 <li><p>Layout a)
2781 The <tt>Use</tt> object(s) are inside (resp. at fixed offset) of the <tt>User</tt>
2782 object and there are a fixed number of them.</p>
2784 <li><p>Layout b)
2785 The <tt>Use</tt> object(s) are referenced by a pointer to an
2786 array from the <tt>User</tt> object and there may be a variable
2787 number of them.</p>
2788 </ul>
2790 As of v2.4 each layout still possesses a direct pointer to the
2791 start of the array of <tt>Use</tt>s. Though not mandatory for layout a),
2792 we stick to this redundancy for the sake of simplicity.
2793 The <tt>User</tt> object also stores the number of <tt>Use</tt> objects it
2794 has. (Theoretically this information can also be calculated
2795 given the scheme presented below.)</p>
2797 Special forms of allocation operators (<tt>operator new</tt>)
2798 enforce the following memory layouts:</p>
2800 <ul>
2801 <li><p>Layout a) is modelled by prepending the <tt>User</tt> object by the <tt>Use[]</tt> array.</p>
2803 <pre>
2804 ...---.---.---.---.-------...
2805 | P | P | P | P | User
2806 '''---'---'---'---'-------'''
2807 </pre>
2809 <li><p>Layout b) is modelled by pointing at the <tt>Use[]</tt> array.</p>
2810 <pre>
2811 .-------...
2812 | User
2813 '-------'''
2816 .---.---.---.---...
2817 | P | P | P | P |
2818 '---'---'---'---'''
2819 </pre>
2820 </ul>
2821 <i>(In the above figures '<tt>P</tt>' stands for the <tt>Use**</tt> that
2822 is stored in each <tt>Use</tt> object in the member <tt>Use::Prev</tt>)</i>
2824 <!-- ______________________________________________________________________ -->
2825 <div class="doc_subsubsection">
2826 <a name="Waymarking">The waymarking algorithm</a>
2827 </div>
2829 <div class="doc_text">
2831 Since the <tt>Use</tt> objects are deprived of the direct (back)pointer to
2832 their <tt>User</tt> objects, there must be a fast and exact method to
2833 recover it. This is accomplished by the following scheme:</p>
2834 </div>
2836 A bit-encoding in the 2 LSBits (least significant bits) of the <tt>Use::Prev</tt> allows to find the
2837 start of the <tt>User</tt> object:
2838 <ul>
2839 <li><tt>00</tt> &mdash;&gt; binary digit 0</li>
2840 <li><tt>01</tt> &mdash;&gt; binary digit 1</li>
2841 <li><tt>10</tt> &mdash;&gt; stop and calculate (<tt>s</tt>)</li>
2842 <li><tt>11</tt> &mdash;&gt; full stop (<tt>S</tt>)</li>
2843 </ul>
2845 Given a <tt>Use*</tt>, all we have to do is to walk till we get
2846 a stop and we either have a <tt>User</tt> immediately behind or
2847 we have to walk to the next stop picking up digits
2848 and calculating the offset:</p>
2849 <pre>
2850 .---.---.---.---.---.---.---.---.---.---.---.---.---.---.---.---.----------------
2851 | 1 | s | 1 | 0 | 1 | 0 | s | 1 | 1 | 0 | s | 1 | 1 | s | 1 | S | User (or User*)
2852 '---'---'---'---'---'---'---'---'---'---'---'---'---'---'---'---'----------------
2853 |+15 |+10 |+6 |+3 |+1
2854 | | | | |__>
2855 | | | |__________>
2856 | | |______________________>
2857 | |______________________________________>
2858 |__________________________________________________________>
2859 </pre>
2861 Only the significant number of bits need to be stored between the
2862 stops, so that the <i>worst case is 20 memory accesses</i> when there are
2863 1000 <tt>Use</tt> objects associated with a <tt>User</tt>.</p>
2865 <!-- ______________________________________________________________________ -->
2866 <div class="doc_subsubsection">
2867 <a name="ReferenceImpl">Reference implementation</a>
2868 </div>
2870 <div class="doc_text">
2872 The following literate Haskell fragment demonstrates the concept:</p>
2873 </div>
2875 <div class="doc_code">
2876 <pre>
2877 > import Test.QuickCheck
2879 > digits :: Int -> [Char] -> [Char]
2880 > digits 0 acc = '0' : acc
2881 > digits 1 acc = '1' : acc
2882 > digits n acc = digits (n `div` 2) $ digits (n `mod` 2) acc
2884 > dist :: Int -> [Char] -> [Char]
2885 > dist 0 [] = ['S']
2886 > dist 0 acc = acc
2887 > dist 1 acc = let r = dist 0 acc in 's' : digits (length r) r
2888 > dist n acc = dist (n - 1) $ dist 1 acc
2890 > takeLast n ss = reverse $ take n $ reverse ss
2892 > test = takeLast 40 $ dist 20 []
2894 </pre>
2895 </div>
2897 Printing &lt;test&gt; gives: <tt>"1s100000s11010s10100s1111s1010s110s11s1S"</tt></p>
2899 The reverse algorithm computes the length of the string just by examining
2900 a certain prefix:</p>
2902 <div class="doc_code">
2903 <pre>
2904 > pref :: [Char] -> Int
2905 > pref "S" = 1
2906 > pref ('s':'1':rest) = decode 2 1 rest
2907 > pref (_:rest) = 1 + pref rest
2909 > decode walk acc ('0':rest) = decode (walk + 1) (acc * 2) rest
2910 > decode walk acc ('1':rest) = decode (walk + 1) (acc * 2 + 1) rest
2911 > decode walk acc _ = walk + acc
2913 </pre>
2914 </div>
2916 Now, as expected, printing &lt;pref test&gt; gives <tt>40</tt>.</p>
2918 We can <i>quickCheck</i> this with following property:</p>
2920 <div class="doc_code">
2921 <pre>
2922 > testcase = dist 2000 []
2923 > testcaseLength = length testcase
2925 > identityProp n = n > 0 && n <= testcaseLength ==> length arr == pref arr
2926 > where arr = takeLast n testcase
2928 </pre>
2929 </div>
2931 As expected &lt;quickCheck identityProp&gt; gives:</p>
2933 <pre>
2934 *Main> quickCheck identityProp
2935 OK, passed 100 tests.
2936 </pre>
2938 Let's be a bit more exhaustive:</p>
2940 <div class="doc_code">
2941 <pre>
2943 > deepCheck p = check (defaultConfig { configMaxTest = 500 }) p
2945 </pre>
2946 </div>
2948 And here is the result of &lt;deepCheck identityProp&gt;:</p>
2950 <pre>
2951 *Main> deepCheck identityProp
2952 OK, passed 500 tests.
2953 </pre>
2955 <!-- ______________________________________________________________________ -->
2956 <div class="doc_subsubsection">
2957 <a name="Tagging">Tagging considerations</a>
2958 </div>
2961 To maintain the invariant that the 2 LSBits of each <tt>Use**</tt> in <tt>Use</tt>
2962 never change after being set up, setters of <tt>Use::Prev</tt> must re-tag the
2963 new <tt>Use**</tt> on every modification. Accordingly getters must strip the
2964 tag bits.</p>
2966 For layout b) instead of the <tt>User</tt> we find a pointer (<tt>User*</tt> with LSBit set).
2967 Following this pointer brings us to the <tt>User</tt>. A portable trick ensures
2968 that the first bytes of <tt>User</tt> (if interpreted as a pointer) never has
2969 the LSBit set. (Portability is relying on the fact that all known compilers place the
2970 <tt>vptr</tt> in the first word of the instances.)</p>
2972 </div>
2974 <!-- *********************************************************************** -->
2975 <div class="doc_section">
2976 <a name="coreclasses">The Core LLVM Class Hierarchy Reference </a>
2977 </div>
2978 <!-- *********************************************************************** -->
2980 <div class="doc_text">
2981 <p><tt>#include "<a href="/doxygen/Type_8h-source.html">llvm/Type.h</a>"</tt>
2982 <br>doxygen info: <a href="/doxygen/classllvm_1_1Type.html">Type Class</a></p>
2984 <p>The Core LLVM classes are the primary means of representing the program
2985 being inspected or transformed. The core LLVM classes are defined in
2986 header files in the <tt>include/llvm/</tt> directory, and implemented in
2987 the <tt>lib/VMCore</tt> directory.</p>
2989 </div>
2991 <!-- ======================================================================= -->
2992 <div class="doc_subsection">
2993 <a name="Type">The <tt>Type</tt> class and Derived Types</a>
2994 </div>
2996 <div class="doc_text">
2998 <p><tt>Type</tt> is a superclass of all type classes. Every <tt>Value</tt> has
2999 a <tt>Type</tt>. <tt>Type</tt> cannot be instantiated directly but only
3000 through its subclasses. Certain primitive types (<tt>VoidType</tt>,
3001 <tt>LabelType</tt>, <tt>FloatType</tt> and <tt>DoubleType</tt>) have hidden
3002 subclasses. They are hidden because they offer no useful functionality beyond
3003 what the <tt>Type</tt> class offers except to distinguish themselves from
3004 other subclasses of <tt>Type</tt>.</p>
3005 <p>All other types are subclasses of <tt>DerivedType</tt>. Types can be
3006 named, but this is not a requirement. There exists exactly
3007 one instance of a given shape at any one time. This allows type equality to
3008 be performed with address equality of the Type Instance. That is, given two
3009 <tt>Type*</tt> values, the types are identical if the pointers are identical.
3010 </p>
3011 </div>
3013 <!-- _______________________________________________________________________ -->
3014 <div class="doc_subsubsection">
3015 <a name="m_Type">Important Public Methods</a>
3016 </div>
3018 <div class="doc_text">
3020 <ul>
3021 <li><tt>bool isIntegerTy() const</tt>: Returns true for any integer type.</li>
3023 <li><tt>bool isFloatingPointTy()</tt>: Return true if this is one of the five
3024 floating point types.</li>
3026 <li><tt>bool isAbstract()</tt>: Return true if the type is abstract (contains
3027 an OpaqueType anywhere in its definition).</li>
3029 <li><tt>bool isSized()</tt>: Return true if the type has known size. Things
3030 that don't have a size are abstract types, labels and void.</li>
3032 </ul>
3033 </div>
3035 <!-- _______________________________________________________________________ -->
3036 <div class="doc_subsubsection">
3037 <a name="derivedtypes">Important Derived Types</a>
3038 </div>
3039 <div class="doc_text">
3040 <dl>
3041 <dt><tt>IntegerType</tt></dt>
3042 <dd>Subclass of DerivedType that represents integer types of any bit width.
3043 Any bit width between <tt>IntegerType::MIN_INT_BITS</tt> (1) and
3044 <tt>IntegerType::MAX_INT_BITS</tt> (~8 million) can be represented.
3045 <ul>
3046 <li><tt>static const IntegerType* get(unsigned NumBits)</tt>: get an integer
3047 type of a specific bit width.</li>
3048 <li><tt>unsigned getBitWidth() const</tt>: Get the bit width of an integer
3049 type.</li>
3050 </ul>
3051 </dd>
3052 <dt><tt>SequentialType</tt></dt>
3053 <dd>This is subclassed by ArrayType and PointerType
3054 <ul>
3055 <li><tt>const Type * getElementType() const</tt>: Returns the type of each
3056 of the elements in the sequential type. </li>
3057 </ul>
3058 </dd>
3059 <dt><tt>ArrayType</tt></dt>
3060 <dd>This is a subclass of SequentialType and defines the interface for array
3061 types.
3062 <ul>
3063 <li><tt>unsigned getNumElements() const</tt>: Returns the number of
3064 elements in the array. </li>
3065 </ul>
3066 </dd>
3067 <dt><tt>PointerType</tt></dt>
3068 <dd>Subclass of SequentialType for pointer types.</dd>
3069 <dt><tt>VectorType</tt></dt>
3070 <dd>Subclass of SequentialType for vector types. A
3071 vector type is similar to an ArrayType but is distinguished because it is
3072 a first class type whereas ArrayType is not. Vector types are used for
3073 vector operations and are usually small vectors of of an integer or floating
3074 point type.</dd>
3075 <dt><tt>StructType</tt></dt>
3076 <dd>Subclass of DerivedTypes for struct types.</dd>
3077 <dt><tt><a name="FunctionType">FunctionType</a></tt></dt>
3078 <dd>Subclass of DerivedTypes for function types.
3079 <ul>
3080 <li><tt>bool isVarArg() const</tt>: Returns true if it's a vararg
3081 function</li>
3082 <li><tt> const Type * getReturnType() const</tt>: Returns the
3083 return type of the function.</li>
3084 <li><tt>const Type * getParamType (unsigned i)</tt>: Returns
3085 the type of the ith parameter.</li>
3086 <li><tt> const unsigned getNumParams() const</tt>: Returns the
3087 number of formal parameters.</li>
3088 </ul>
3089 </dd>
3090 <dt><tt>OpaqueType</tt></dt>
3091 <dd>Sublcass of DerivedType for abstract types. This class
3092 defines no content and is used as a placeholder for some other type. Note
3093 that OpaqueType is used (temporarily) during type resolution for forward
3094 references of types. Once the referenced type is resolved, the OpaqueType
3095 is replaced with the actual type. OpaqueType can also be used for data
3096 abstraction. At link time opaque types can be resolved to actual types
3097 of the same name.</dd>
3098 </dl>
3099 </div>
3103 <!-- ======================================================================= -->
3104 <div class="doc_subsection">
3105 <a name="Module">The <tt>Module</tt> class</a>
3106 </div>
3108 <div class="doc_text">
3110 <p><tt>#include "<a
3111 href="/doxygen/Module_8h-source.html">llvm/Module.h</a>"</tt><br> doxygen info:
3112 <a href="/doxygen/classllvm_1_1Module.html">Module Class</a></p>
3114 <p>The <tt>Module</tt> class represents the top level structure present in LLVM
3115 programs. An LLVM module is effectively either a translation unit of the
3116 original program or a combination of several translation units merged by the
3117 linker. The <tt>Module</tt> class keeps track of a list of <a
3118 href="#Function"><tt>Function</tt></a>s, a list of <a
3119 href="#GlobalVariable"><tt>GlobalVariable</tt></a>s, and a <a
3120 href="#SymbolTable"><tt>SymbolTable</tt></a>. Additionally, it contains a few
3121 helpful member functions that try to make common operations easy.</p>
3123 </div>
3125 <!-- _______________________________________________________________________ -->
3126 <div class="doc_subsubsection">
3127 <a name="m_Module">Important Public Members of the <tt>Module</tt> class</a>
3128 </div>
3130 <div class="doc_text">
3132 <ul>
3133 <li><tt>Module::Module(std::string name = "")</tt></li>
3134 </ul>
3136 <p>Constructing a <a href="#Module">Module</a> is easy. You can optionally
3137 provide a name for it (probably based on the name of the translation unit).</p>
3139 <ul>
3140 <li><tt>Module::iterator</tt> - Typedef for function list iterator<br>
3141 <tt>Module::const_iterator</tt> - Typedef for const_iterator.<br>
3143 <tt>begin()</tt>, <tt>end()</tt>
3144 <tt>size()</tt>, <tt>empty()</tt>
3146 <p>These are forwarding methods that make it easy to access the contents of
3147 a <tt>Module</tt> object's <a href="#Function"><tt>Function</tt></a>
3148 list.</p></li>
3150 <li><tt>Module::FunctionListType &amp;getFunctionList()</tt>
3152 <p> Returns the list of <a href="#Function"><tt>Function</tt></a>s. This is
3153 necessary to use when you need to update the list or perform a complex
3154 action that doesn't have a forwarding method.</p>
3156 <p><!-- Global Variable --></p></li>
3157 </ul>
3159 <hr>
3161 <ul>
3162 <li><tt>Module::global_iterator</tt> - Typedef for global variable list iterator<br>
3164 <tt>Module::const_global_iterator</tt> - Typedef for const_iterator.<br>
3166 <tt>global_begin()</tt>, <tt>global_end()</tt>
3167 <tt>global_size()</tt>, <tt>global_empty()</tt>
3169 <p> These are forwarding methods that make it easy to access the contents of
3170 a <tt>Module</tt> object's <a
3171 href="#GlobalVariable"><tt>GlobalVariable</tt></a> list.</p></li>
3173 <li><tt>Module::GlobalListType &amp;getGlobalList()</tt>
3175 <p>Returns the list of <a
3176 href="#GlobalVariable"><tt>GlobalVariable</tt></a>s. This is necessary to
3177 use when you need to update the list or perform a complex action that
3178 doesn't have a forwarding method.</p>
3180 <p><!-- Symbol table stuff --> </p></li>
3181 </ul>
3183 <hr>
3185 <ul>
3186 <li><tt><a href="#SymbolTable">SymbolTable</a> *getSymbolTable()</tt>
3188 <p>Return a reference to the <a href="#SymbolTable"><tt>SymbolTable</tt></a>
3189 for this <tt>Module</tt>.</p>
3191 <p><!-- Convenience methods --></p></li>
3192 </ul>
3194 <hr>
3196 <ul>
3197 <li><tt><a href="#Function">Function</a> *getFunction(const std::string
3198 &amp;Name, const <a href="#FunctionType">FunctionType</a> *Ty)</tt>
3200 <p>Look up the specified function in the <tt>Module</tt> <a
3201 href="#SymbolTable"><tt>SymbolTable</tt></a>. If it does not exist, return
3202 <tt>null</tt>.</p></li>
3204 <li><tt><a href="#Function">Function</a> *getOrInsertFunction(const
3205 std::string &amp;Name, const <a href="#FunctionType">FunctionType</a> *T)</tt>
3207 <p>Look up the specified function in the <tt>Module</tt> <a
3208 href="#SymbolTable"><tt>SymbolTable</tt></a>. If it does not exist, add an
3209 external declaration for the function and return it.</p></li>
3211 <li><tt>std::string getTypeName(const <a href="#Type">Type</a> *Ty)</tt>
3213 <p>If there is at least one entry in the <a
3214 href="#SymbolTable"><tt>SymbolTable</tt></a> for the specified <a
3215 href="#Type"><tt>Type</tt></a>, return it. Otherwise return the empty
3216 string.</p></li>
3218 <li><tt>bool addTypeName(const std::string &amp;Name, const <a
3219 href="#Type">Type</a> *Ty)</tt>
3221 <p>Insert an entry in the <a href="#SymbolTable"><tt>SymbolTable</tt></a>
3222 mapping <tt>Name</tt> to <tt>Ty</tt>. If there is already an entry for this
3223 name, true is returned and the <a
3224 href="#SymbolTable"><tt>SymbolTable</tt></a> is not modified.</p></li>
3225 </ul>
3227 </div>
3230 <!-- ======================================================================= -->
3231 <div class="doc_subsection">
3232 <a name="Value">The <tt>Value</tt> class</a>
3233 </div>
3235 <div class="doc_text">
3237 <p><tt>#include "<a href="/doxygen/Value_8h-source.html">llvm/Value.h</a>"</tt>
3238 <br>
3239 doxygen info: <a href="/doxygen/classllvm_1_1Value.html">Value Class</a></p>
3241 <p>The <tt>Value</tt> class is the most important class in the LLVM Source
3242 base. It represents a typed value that may be used (among other things) as an
3243 operand to an instruction. There are many different types of <tt>Value</tt>s,
3244 such as <a href="#Constant"><tt>Constant</tt></a>s,<a
3245 href="#Argument"><tt>Argument</tt></a>s. Even <a
3246 href="#Instruction"><tt>Instruction</tt></a>s and <a
3247 href="#Function"><tt>Function</tt></a>s are <tt>Value</tt>s.</p>
3249 <p>A particular <tt>Value</tt> may be used many times in the LLVM representation
3250 for a program. For example, an incoming argument to a function (represented
3251 with an instance of the <a href="#Argument">Argument</a> class) is "used" by
3252 every instruction in the function that references the argument. To keep track
3253 of this relationship, the <tt>Value</tt> class keeps a list of all of the <a
3254 href="#User"><tt>User</tt></a>s that is using it (the <a
3255 href="#User"><tt>User</tt></a> class is a base class for all nodes in the LLVM
3256 graph that can refer to <tt>Value</tt>s). This use list is how LLVM represents
3257 def-use information in the program, and is accessible through the <tt>use_</tt>*
3258 methods, shown below.</p>
3260 <p>Because LLVM is a typed representation, every LLVM <tt>Value</tt> is typed,
3261 and this <a href="#Type">Type</a> is available through the <tt>getType()</tt>
3262 method. In addition, all LLVM values can be named. The "name" of the
3263 <tt>Value</tt> is a symbolic string printed in the LLVM code:</p>
3265 <div class="doc_code">
3266 <pre>
3267 %<b>foo</b> = add i32 1, 2
3268 </pre>
3269 </div>
3271 <p><a name="nameWarning">The name of this instruction is "foo".</a> <b>NOTE</b>
3272 that the name of any value may be missing (an empty string), so names should
3273 <b>ONLY</b> be used for debugging (making the source code easier to read,
3274 debugging printouts), they should not be used to keep track of values or map
3275 between them. For this purpose, use a <tt>std::map</tt> of pointers to the
3276 <tt>Value</tt> itself instead.</p>
3278 <p>One important aspect of LLVM is that there is no distinction between an SSA
3279 variable and the operation that produces it. Because of this, any reference to
3280 the value produced by an instruction (or the value available as an incoming
3281 argument, for example) is represented as a direct pointer to the instance of
3282 the class that
3283 represents this value. Although this may take some getting used to, it
3284 simplifies the representation and makes it easier to manipulate.</p>
3286 </div>
3288 <!-- _______________________________________________________________________ -->
3289 <div class="doc_subsubsection">
3290 <a name="m_Value">Important Public Members of the <tt>Value</tt> class</a>
3291 </div>
3293 <div class="doc_text">
3295 <ul>
3296 <li><tt>Value::use_iterator</tt> - Typedef for iterator over the
3297 use-list<br>
3298 <tt>Value::const_use_iterator</tt> - Typedef for const_iterator over
3299 the use-list<br>
3300 <tt>unsigned use_size()</tt> - Returns the number of users of the
3301 value.<br>
3302 <tt>bool use_empty()</tt> - Returns true if there are no users.<br>
3303 <tt>use_iterator use_begin()</tt> - Get an iterator to the start of
3304 the use-list.<br>
3305 <tt>use_iterator use_end()</tt> - Get an iterator to the end of the
3306 use-list.<br>
3307 <tt><a href="#User">User</a> *use_back()</tt> - Returns the last
3308 element in the list.
3309 <p> These methods are the interface to access the def-use
3310 information in LLVM. As with all other iterators in LLVM, the naming
3311 conventions follow the conventions defined by the <a href="#stl">STL</a>.</p>
3312 </li>
3313 <li><tt><a href="#Type">Type</a> *getType() const</tt>
3314 <p>This method returns the Type of the Value.</p>
3315 </li>
3316 <li><tt>bool hasName() const</tt><br>
3317 <tt>std::string getName() const</tt><br>
3318 <tt>void setName(const std::string &amp;Name)</tt>
3319 <p> This family of methods is used to access and assign a name to a <tt>Value</tt>,
3320 be aware of the <a href="#nameWarning">precaution above</a>.</p>
3321 </li>
3322 <li><tt>void replaceAllUsesWith(Value *V)</tt>
3324 <p>This method traverses the use list of a <tt>Value</tt> changing all <a
3325 href="#User"><tt>User</tt>s</a> of the current value to refer to
3326 "<tt>V</tt>" instead. For example, if you detect that an instruction always
3327 produces a constant value (for example through constant folding), you can
3328 replace all uses of the instruction with the constant like this:</p>
3330 <div class="doc_code">
3331 <pre>
3332 Inst-&gt;replaceAllUsesWith(ConstVal);
3333 </pre>
3334 </div>
3336 </ul>
3338 </div>
3340 <!-- ======================================================================= -->
3341 <div class="doc_subsection">
3342 <a name="User">The <tt>User</tt> class</a>
3343 </div>
3345 <div class="doc_text">
3348 <tt>#include "<a href="/doxygen/User_8h-source.html">llvm/User.h</a>"</tt><br>
3349 doxygen info: <a href="/doxygen/classllvm_1_1User.html">User Class</a><br>
3350 Superclass: <a href="#Value"><tt>Value</tt></a></p>
3352 <p>The <tt>User</tt> class is the common base class of all LLVM nodes that may
3353 refer to <a href="#Value"><tt>Value</tt></a>s. It exposes a list of "Operands"
3354 that are all of the <a href="#Value"><tt>Value</tt></a>s that the User is
3355 referring to. The <tt>User</tt> class itself is a subclass of
3356 <tt>Value</tt>.</p>
3358 <p>The operands of a <tt>User</tt> point directly to the LLVM <a
3359 href="#Value"><tt>Value</tt></a> that it refers to. Because LLVM uses Static
3360 Single Assignment (SSA) form, there can only be one definition referred to,
3361 allowing this direct connection. This connection provides the use-def
3362 information in LLVM.</p>
3364 </div>
3366 <!-- _______________________________________________________________________ -->
3367 <div class="doc_subsubsection">
3368 <a name="m_User">Important Public Members of the <tt>User</tt> class</a>
3369 </div>
3371 <div class="doc_text">
3373 <p>The <tt>User</tt> class exposes the operand list in two ways: through
3374 an index access interface and through an iterator based interface.</p>
3376 <ul>
3377 <li><tt>Value *getOperand(unsigned i)</tt><br>
3378 <tt>unsigned getNumOperands()</tt>
3379 <p> These two methods expose the operands of the <tt>User</tt> in a
3380 convenient form for direct access.</p></li>
3382 <li><tt>User::op_iterator</tt> - Typedef for iterator over the operand
3383 list<br>
3384 <tt>op_iterator op_begin()</tt> - Get an iterator to the start of
3385 the operand list.<br>
3386 <tt>op_iterator op_end()</tt> - Get an iterator to the end of the
3387 operand list.
3388 <p> Together, these methods make up the iterator based interface to
3389 the operands of a <tt>User</tt>.</p></li>
3390 </ul>
3392 </div>
3394 <!-- ======================================================================= -->
3395 <div class="doc_subsection">
3396 <a name="Instruction">The <tt>Instruction</tt> class</a>
3397 </div>
3399 <div class="doc_text">
3401 <p><tt>#include "</tt><tt><a
3402 href="/doxygen/Instruction_8h-source.html">llvm/Instruction.h</a>"</tt><br>
3403 doxygen info: <a href="/doxygen/classllvm_1_1Instruction.html">Instruction Class</a><br>
3404 Superclasses: <a href="#User"><tt>User</tt></a>, <a
3405 href="#Value"><tt>Value</tt></a></p>
3407 <p>The <tt>Instruction</tt> class is the common base class for all LLVM
3408 instructions. It provides only a few methods, but is a very commonly used
3409 class. The primary data tracked by the <tt>Instruction</tt> class itself is the
3410 opcode (instruction type) and the parent <a
3411 href="#BasicBlock"><tt>BasicBlock</tt></a> the <tt>Instruction</tt> is embedded
3412 into. To represent a specific type of instruction, one of many subclasses of
3413 <tt>Instruction</tt> are used.</p>
3415 <p> Because the <tt>Instruction</tt> class subclasses the <a
3416 href="#User"><tt>User</tt></a> class, its operands can be accessed in the same
3417 way as for other <a href="#User"><tt>User</tt></a>s (with the
3418 <tt>getOperand()</tt>/<tt>getNumOperands()</tt> and
3419 <tt>op_begin()</tt>/<tt>op_end()</tt> methods).</p> <p> An important file for
3420 the <tt>Instruction</tt> class is the <tt>llvm/Instruction.def</tt> file. This
3421 file contains some meta-data about the various different types of instructions
3422 in LLVM. It describes the enum values that are used as opcodes (for example
3423 <tt>Instruction::Add</tt> and <tt>Instruction::ICmp</tt>), as well as the
3424 concrete sub-classes of <tt>Instruction</tt> that implement the instruction (for
3425 example <tt><a href="#BinaryOperator">BinaryOperator</a></tt> and <tt><a
3426 href="#CmpInst">CmpInst</a></tt>). Unfortunately, the use of macros in
3427 this file confuses doxygen, so these enum values don't show up correctly in the
3428 <a href="/doxygen/classllvm_1_1Instruction.html">doxygen output</a>.</p>
3430 </div>
3432 <!-- _______________________________________________________________________ -->
3433 <div class="doc_subsubsection">
3434 <a name="s_Instruction">Important Subclasses of the <tt>Instruction</tt>
3435 class</a>
3436 </div>
3437 <div class="doc_text">
3438 <ul>
3439 <li><tt><a name="BinaryOperator">BinaryOperator</a></tt>
3440 <p>This subclasses represents all two operand instructions whose operands
3441 must be the same type, except for the comparison instructions.</p></li>
3442 <li><tt><a name="CastInst">CastInst</a></tt>
3443 <p>This subclass is the parent of the 12 casting instructions. It provides
3444 common operations on cast instructions.</p>
3445 <li><tt><a name="CmpInst">CmpInst</a></tt>
3446 <p>This subclass respresents the two comparison instructions,
3447 <a href="LangRef.html#i_icmp">ICmpInst</a> (integer opreands), and
3448 <a href="LangRef.html#i_fcmp">FCmpInst</a> (floating point operands).</p>
3449 <li><tt><a name="TerminatorInst">TerminatorInst</a></tt>
3450 <p>This subclass is the parent of all terminator instructions (those which
3451 can terminate a block).</p>
3452 </ul>
3453 </div>
3455 <!-- _______________________________________________________________________ -->
3456 <div class="doc_subsubsection">
3457 <a name="m_Instruction">Important Public Members of the <tt>Instruction</tt>
3458 class</a>
3459 </div>
3461 <div class="doc_text">
3463 <ul>
3464 <li><tt><a href="#BasicBlock">BasicBlock</a> *getParent()</tt>
3465 <p>Returns the <a href="#BasicBlock"><tt>BasicBlock</tt></a> that
3466 this <tt>Instruction</tt> is embedded into.</p></li>
3467 <li><tt>bool mayWriteToMemory()</tt>
3468 <p>Returns true if the instruction writes to memory, i.e. it is a
3469 <tt>call</tt>,<tt>free</tt>,<tt>invoke</tt>, or <tt>store</tt>.</p></li>
3470 <li><tt>unsigned getOpcode()</tt>
3471 <p>Returns the opcode for the <tt>Instruction</tt>.</p></li>
3472 <li><tt><a href="#Instruction">Instruction</a> *clone() const</tt>
3473 <p>Returns another instance of the specified instruction, identical
3474 in all ways to the original except that the instruction has no parent
3475 (ie it's not embedded into a <a href="#BasicBlock"><tt>BasicBlock</tt></a>),
3476 and it has no name</p></li>
3477 </ul>
3479 </div>
3481 <!-- ======================================================================= -->
3482 <div class="doc_subsection">
3483 <a name="Constant">The <tt>Constant</tt> class and subclasses</a>
3484 </div>
3486 <div class="doc_text">
3488 <p>Constant represents a base class for different types of constants. It
3489 is subclassed by ConstantInt, ConstantArray, etc. for representing
3490 the various types of Constants. <a href="#GlobalValue">GlobalValue</a> is also
3491 a subclass, which represents the address of a global variable or function.
3492 </p>
3494 </div>
3496 <!-- _______________________________________________________________________ -->
3497 <div class="doc_subsubsection">Important Subclasses of Constant </div>
3498 <div class="doc_text">
3499 <ul>
3500 <li>ConstantInt : This subclass of Constant represents an integer constant of
3501 any width.
3502 <ul>
3503 <li><tt>const APInt&amp; getValue() const</tt>: Returns the underlying
3504 value of this constant, an APInt value.</li>
3505 <li><tt>int64_t getSExtValue() const</tt>: Converts the underlying APInt
3506 value to an int64_t via sign extension. If the value (not the bit width)
3507 of the APInt is too large to fit in an int64_t, an assertion will result.
3508 For this reason, use of this method is discouraged.</li>
3509 <li><tt>uint64_t getZExtValue() const</tt>: Converts the underlying APInt
3510 value to a uint64_t via zero extension. IF the value (not the bit width)
3511 of the APInt is too large to fit in a uint64_t, an assertion will result.
3512 For this reason, use of this method is discouraged.</li>
3513 <li><tt>static ConstantInt* get(const APInt&amp; Val)</tt>: Returns the
3514 ConstantInt object that represents the value provided by <tt>Val</tt>.
3515 The type is implied as the IntegerType that corresponds to the bit width
3516 of <tt>Val</tt>.</li>
3517 <li><tt>static ConstantInt* get(const Type *Ty, uint64_t Val)</tt>:
3518 Returns the ConstantInt object that represents the value provided by
3519 <tt>Val</tt> for integer type <tt>Ty</tt>.</li>
3520 </ul>
3521 </li>
3522 <li>ConstantFP : This class represents a floating point constant.
3523 <ul>
3524 <li><tt>double getValue() const</tt>: Returns the underlying value of
3525 this constant. </li>
3526 </ul>
3527 </li>
3528 <li>ConstantArray : This represents a constant array.
3529 <ul>
3530 <li><tt>const std::vector&lt;Use&gt; &amp;getValues() const</tt>: Returns
3531 a vector of component constants that makeup this array. </li>
3532 </ul>
3533 </li>
3534 <li>ConstantStruct : This represents a constant struct.
3535 <ul>
3536 <li><tt>const std::vector&lt;Use&gt; &amp;getValues() const</tt>: Returns
3537 a vector of component constants that makeup this array. </li>
3538 </ul>
3539 </li>
3540 <li>GlobalValue : This represents either a global variable or a function. In
3541 either case, the value is a constant fixed address (after linking).
3542 </li>
3543 </ul>
3544 </div>
3547 <!-- ======================================================================= -->
3548 <div class="doc_subsection">
3549 <a name="GlobalValue">The <tt>GlobalValue</tt> class</a>
3550 </div>
3552 <div class="doc_text">
3554 <p><tt>#include "<a
3555 href="/doxygen/GlobalValue_8h-source.html">llvm/GlobalValue.h</a>"</tt><br>
3556 doxygen info: <a href="/doxygen/classllvm_1_1GlobalValue.html">GlobalValue
3557 Class</a><br>
3558 Superclasses: <a href="#Constant"><tt>Constant</tt></a>,
3559 <a href="#User"><tt>User</tt></a>, <a href="#Value"><tt>Value</tt></a></p>
3561 <p>Global values (<a href="#GlobalVariable"><tt>GlobalVariable</tt></a>s or <a
3562 href="#Function"><tt>Function</tt></a>s) are the only LLVM values that are
3563 visible in the bodies of all <a href="#Function"><tt>Function</tt></a>s.
3564 Because they are visible at global scope, they are also subject to linking with
3565 other globals defined in different translation units. To control the linking
3566 process, <tt>GlobalValue</tt>s know their linkage rules. Specifically,
3567 <tt>GlobalValue</tt>s know whether they have internal or external linkage, as
3568 defined by the <tt>LinkageTypes</tt> enumeration.</p>
3570 <p>If a <tt>GlobalValue</tt> has internal linkage (equivalent to being
3571 <tt>static</tt> in C), it is not visible to code outside the current translation
3572 unit, and does not participate in linking. If it has external linkage, it is
3573 visible to external code, and does participate in linking. In addition to
3574 linkage information, <tt>GlobalValue</tt>s keep track of which <a
3575 href="#Module"><tt>Module</tt></a> they are currently part of.</p>
3577 <p>Because <tt>GlobalValue</tt>s are memory objects, they are always referred to
3578 by their <b>address</b>. As such, the <a href="#Type"><tt>Type</tt></a> of a
3579 global is always a pointer to its contents. It is important to remember this
3580 when using the <tt>GetElementPtrInst</tt> instruction because this pointer must
3581 be dereferenced first. For example, if you have a <tt>GlobalVariable</tt> (a
3582 subclass of <tt>GlobalValue)</tt> that is an array of 24 ints, type <tt>[24 x
3583 i32]</tt>, then the <tt>GlobalVariable</tt> is a pointer to that array. Although
3584 the address of the first element of this array and the value of the
3585 <tt>GlobalVariable</tt> are the same, they have different types. The
3586 <tt>GlobalVariable</tt>'s type is <tt>[24 x i32]</tt>. The first element's type
3587 is <tt>i32.</tt> Because of this, accessing a global value requires you to
3588 dereference the pointer with <tt>GetElementPtrInst</tt> first, then its elements
3589 can be accessed. This is explained in the <a href="LangRef.html#globalvars">LLVM
3590 Language Reference Manual</a>.</p>
3592 </div>
3594 <!-- _______________________________________________________________________ -->
3595 <div class="doc_subsubsection">
3596 <a name="m_GlobalValue">Important Public Members of the <tt>GlobalValue</tt>
3597 class</a>
3598 </div>
3600 <div class="doc_text">
3602 <ul>
3603 <li><tt>bool hasInternalLinkage() const</tt><br>
3604 <tt>bool hasExternalLinkage() const</tt><br>
3605 <tt>void setInternalLinkage(bool HasInternalLinkage)</tt>
3606 <p> These methods manipulate the linkage characteristics of the <tt>GlobalValue</tt>.</p>
3607 <p> </p>
3608 </li>
3609 <li><tt><a href="#Module">Module</a> *getParent()</tt>
3610 <p> This returns the <a href="#Module"><tt>Module</tt></a> that the
3611 GlobalValue is currently embedded into.</p></li>
3612 </ul>
3614 </div>
3616 <!-- ======================================================================= -->
3617 <div class="doc_subsection">
3618 <a name="Function">The <tt>Function</tt> class</a>
3619 </div>
3621 <div class="doc_text">
3623 <p><tt>#include "<a
3624 href="/doxygen/Function_8h-source.html">llvm/Function.h</a>"</tt><br> doxygen
3625 info: <a href="/doxygen/classllvm_1_1Function.html">Function Class</a><br>
3626 Superclasses: <a href="#GlobalValue"><tt>GlobalValue</tt></a>,
3627 <a href="#Constant"><tt>Constant</tt></a>,
3628 <a href="#User"><tt>User</tt></a>,
3629 <a href="#Value"><tt>Value</tt></a></p>
3631 <p>The <tt>Function</tt> class represents a single procedure in LLVM. It is
3632 actually one of the more complex classes in the LLVM hierarchy because it must
3633 keep track of a large amount of data. The <tt>Function</tt> class keeps track
3634 of a list of <a href="#BasicBlock"><tt>BasicBlock</tt></a>s, a list of formal
3635 <a href="#Argument"><tt>Argument</tt></a>s, and a
3636 <a href="#SymbolTable"><tt>SymbolTable</tt></a>.</p>
3638 <p>The list of <a href="#BasicBlock"><tt>BasicBlock</tt></a>s is the most
3639 commonly used part of <tt>Function</tt> objects. The list imposes an implicit
3640 ordering of the blocks in the function, which indicate how the code will be
3641 laid out by the backend. Additionally, the first <a
3642 href="#BasicBlock"><tt>BasicBlock</tt></a> is the implicit entry node for the
3643 <tt>Function</tt>. It is not legal in LLVM to explicitly branch to this initial
3644 block. There are no implicit exit nodes, and in fact there may be multiple exit
3645 nodes from a single <tt>Function</tt>. If the <a
3646 href="#BasicBlock"><tt>BasicBlock</tt></a> list is empty, this indicates that
3647 the <tt>Function</tt> is actually a function declaration: the actual body of the
3648 function hasn't been linked in yet.</p>
3650 <p>In addition to a list of <a href="#BasicBlock"><tt>BasicBlock</tt></a>s, the
3651 <tt>Function</tt> class also keeps track of the list of formal <a
3652 href="#Argument"><tt>Argument</tt></a>s that the function receives. This
3653 container manages the lifetime of the <a href="#Argument"><tt>Argument</tt></a>
3654 nodes, just like the <a href="#BasicBlock"><tt>BasicBlock</tt></a> list does for
3655 the <a href="#BasicBlock"><tt>BasicBlock</tt></a>s.</p>
3657 <p>The <a href="#SymbolTable"><tt>SymbolTable</tt></a> is a very rarely used
3658 LLVM feature that is only used when you have to look up a value by name. Aside
3659 from that, the <a href="#SymbolTable"><tt>SymbolTable</tt></a> is used
3660 internally to make sure that there are not conflicts between the names of <a
3661 href="#Instruction"><tt>Instruction</tt></a>s, <a
3662 href="#BasicBlock"><tt>BasicBlock</tt></a>s, or <a
3663 href="#Argument"><tt>Argument</tt></a>s in the function body.</p>
3665 <p>Note that <tt>Function</tt> is a <a href="#GlobalValue">GlobalValue</a>
3666 and therefore also a <a href="#Constant">Constant</a>. The value of the function
3667 is its address (after linking) which is guaranteed to be constant.</p>
3668 </div>
3670 <!-- _______________________________________________________________________ -->
3671 <div class="doc_subsubsection">
3672 <a name="m_Function">Important Public Members of the <tt>Function</tt>
3673 class</a>
3674 </div>
3676 <div class="doc_text">
3678 <ul>
3679 <li><tt>Function(const </tt><tt><a href="#FunctionType">FunctionType</a>
3680 *Ty, LinkageTypes Linkage, const std::string &amp;N = "", Module* Parent = 0)</tt>
3682 <p>Constructor used when you need to create new <tt>Function</tt>s to add
3683 the the program. The constructor must specify the type of the function to
3684 create and what type of linkage the function should have. The <a
3685 href="#FunctionType"><tt>FunctionType</tt></a> argument
3686 specifies the formal arguments and return value for the function. The same
3687 <a href="#FunctionType"><tt>FunctionType</tt></a> value can be used to
3688 create multiple functions. The <tt>Parent</tt> argument specifies the Module
3689 in which the function is defined. If this argument is provided, the function
3690 will automatically be inserted into that module's list of
3691 functions.</p></li>
3693 <li><tt>bool isDeclaration()</tt>
3695 <p>Return whether or not the <tt>Function</tt> has a body defined. If the
3696 function is "external", it does not have a body, and thus must be resolved
3697 by linking with a function defined in a different translation unit.</p></li>
3699 <li><tt>Function::iterator</tt> - Typedef for basic block list iterator<br>
3700 <tt>Function::const_iterator</tt> - Typedef for const_iterator.<br>
3702 <tt>begin()</tt>, <tt>end()</tt>
3703 <tt>size()</tt>, <tt>empty()</tt>
3705 <p>These are forwarding methods that make it easy to access the contents of
3706 a <tt>Function</tt> object's <a href="#BasicBlock"><tt>BasicBlock</tt></a>
3707 list.</p></li>
3709 <li><tt>Function::BasicBlockListType &amp;getBasicBlockList()</tt>
3711 <p>Returns the list of <a href="#BasicBlock"><tt>BasicBlock</tt></a>s. This
3712 is necessary to use when you need to update the list or perform a complex
3713 action that doesn't have a forwarding method.</p></li>
3715 <li><tt>Function::arg_iterator</tt> - Typedef for the argument list
3716 iterator<br>
3717 <tt>Function::const_arg_iterator</tt> - Typedef for const_iterator.<br>
3719 <tt>arg_begin()</tt>, <tt>arg_end()</tt>
3720 <tt>arg_size()</tt>, <tt>arg_empty()</tt>
3722 <p>These are forwarding methods that make it easy to access the contents of
3723 a <tt>Function</tt> object's <a href="#Argument"><tt>Argument</tt></a>
3724 list.</p></li>
3726 <li><tt>Function::ArgumentListType &amp;getArgumentList()</tt>
3728 <p>Returns the list of <a href="#Argument"><tt>Argument</tt></a>s. This is
3729 necessary to use when you need to update the list or perform a complex
3730 action that doesn't have a forwarding method.</p></li>
3732 <li><tt><a href="#BasicBlock">BasicBlock</a> &amp;getEntryBlock()</tt>
3734 <p>Returns the entry <a href="#BasicBlock"><tt>BasicBlock</tt></a> for the
3735 function. Because the entry block for the function is always the first
3736 block, this returns the first block of the <tt>Function</tt>.</p></li>
3738 <li><tt><a href="#Type">Type</a> *getReturnType()</tt><br>
3739 <tt><a href="#FunctionType">FunctionType</a> *getFunctionType()</tt>
3741 <p>This traverses the <a href="#Type"><tt>Type</tt></a> of the
3742 <tt>Function</tt> and returns the return type of the function, or the <a
3743 href="#FunctionType"><tt>FunctionType</tt></a> of the actual
3744 function.</p></li>
3746 <li><tt><a href="#SymbolTable">SymbolTable</a> *getSymbolTable()</tt>
3748 <p> Return a pointer to the <a href="#SymbolTable"><tt>SymbolTable</tt></a>
3749 for this <tt>Function</tt>.</p></li>
3750 </ul>
3752 </div>
3754 <!-- ======================================================================= -->
3755 <div class="doc_subsection">
3756 <a name="GlobalVariable">The <tt>GlobalVariable</tt> class</a>
3757 </div>
3759 <div class="doc_text">
3761 <p><tt>#include "<a
3762 href="/doxygen/GlobalVariable_8h-source.html">llvm/GlobalVariable.h</a>"</tt>
3763 <br>
3764 doxygen info: <a href="/doxygen/classllvm_1_1GlobalVariable.html">GlobalVariable
3765 Class</a><br>
3766 Superclasses: <a href="#GlobalValue"><tt>GlobalValue</tt></a>,
3767 <a href="#Constant"><tt>Constant</tt></a>,
3768 <a href="#User"><tt>User</tt></a>,
3769 <a href="#Value"><tt>Value</tt></a></p>
3771 <p>Global variables are represented with the (surprise surprise)
3772 <tt>GlobalVariable</tt> class. Like functions, <tt>GlobalVariable</tt>s are also
3773 subclasses of <a href="#GlobalValue"><tt>GlobalValue</tt></a>, and as such are
3774 always referenced by their address (global values must live in memory, so their
3775 "name" refers to their constant address). See
3776 <a href="#GlobalValue"><tt>GlobalValue</tt></a> for more on this. Global
3777 variables may have an initial value (which must be a
3778 <a href="#Constant"><tt>Constant</tt></a>), and if they have an initializer,
3779 they may be marked as "constant" themselves (indicating that their contents
3780 never change at runtime).</p>
3781 </div>
3783 <!-- _______________________________________________________________________ -->
3784 <div class="doc_subsubsection">
3785 <a name="m_GlobalVariable">Important Public Members of the
3786 <tt>GlobalVariable</tt> class</a>
3787 </div>
3789 <div class="doc_text">
3791 <ul>
3792 <li><tt>GlobalVariable(const </tt><tt><a href="#Type">Type</a> *Ty, bool
3793 isConstant, LinkageTypes&amp; Linkage, <a href="#Constant">Constant</a>
3794 *Initializer = 0, const std::string &amp;Name = "", Module* Parent = 0)</tt>
3796 <p>Create a new global variable of the specified type. If
3797 <tt>isConstant</tt> is true then the global variable will be marked as
3798 unchanging for the program. The Linkage parameter specifies the type of
3799 linkage (internal, external, weak, linkonce, appending) for the variable.
3800 If the linkage is InternalLinkage, WeakAnyLinkage, WeakODRLinkage,
3801 LinkOnceAnyLinkage or LinkOnceODRLinkage,&nbsp; then the resultant
3802 global variable will have internal linkage. AppendingLinkage concatenates
3803 together all instances (in different translation units) of the variable
3804 into a single variable but is only applicable to arrays. &nbsp;See
3805 the <a href="LangRef.html#modulestructure">LLVM Language Reference</a> for
3806 further details on linkage types. Optionally an initializer, a name, and the
3807 module to put the variable into may be specified for the global variable as
3808 well.</p></li>
3810 <li><tt>bool isConstant() const</tt>
3812 <p>Returns true if this is a global variable that is known not to
3813 be modified at runtime.</p></li>
3815 <li><tt>bool hasInitializer()</tt>
3817 <p>Returns true if this <tt>GlobalVariable</tt> has an intializer.</p></li>
3819 <li><tt><a href="#Constant">Constant</a> *getInitializer()</tt>
3821 <p>Returns the initial value for a <tt>GlobalVariable</tt>. It is not legal
3822 to call this method if there is no initializer.</p></li>
3823 </ul>
3825 </div>
3828 <!-- ======================================================================= -->
3829 <div class="doc_subsection">
3830 <a name="BasicBlock">The <tt>BasicBlock</tt> class</a>
3831 </div>
3833 <div class="doc_text">
3835 <p><tt>#include "<a
3836 href="/doxygen/BasicBlock_8h-source.html">llvm/BasicBlock.h</a>"</tt><br>
3837 doxygen info: <a href="/doxygen/classllvm_1_1BasicBlock.html">BasicBlock
3838 Class</a><br>
3839 Superclass: <a href="#Value"><tt>Value</tt></a></p>
3841 <p>This class represents a single entry multiple exit section of the code,
3842 commonly known as a basic block by the compiler community. The
3843 <tt>BasicBlock</tt> class maintains a list of <a
3844 href="#Instruction"><tt>Instruction</tt></a>s, which form the body of the block.
3845 Matching the language definition, the last element of this list of instructions
3846 is always a terminator instruction (a subclass of the <a
3847 href="#TerminatorInst"><tt>TerminatorInst</tt></a> class).</p>
3849 <p>In addition to tracking the list of instructions that make up the block, the
3850 <tt>BasicBlock</tt> class also keeps track of the <a
3851 href="#Function"><tt>Function</tt></a> that it is embedded into.</p>
3853 <p>Note that <tt>BasicBlock</tt>s themselves are <a
3854 href="#Value"><tt>Value</tt></a>s, because they are referenced by instructions
3855 like branches and can go in the switch tables. <tt>BasicBlock</tt>s have type
3856 <tt>label</tt>.</p>
3858 </div>
3860 <!-- _______________________________________________________________________ -->
3861 <div class="doc_subsubsection">
3862 <a name="m_BasicBlock">Important Public Members of the <tt>BasicBlock</tt>
3863 class</a>
3864 </div>
3866 <div class="doc_text">
3867 <ul>
3869 <li><tt>BasicBlock(const std::string &amp;Name = "", </tt><tt><a
3870 href="#Function">Function</a> *Parent = 0)</tt>
3872 <p>The <tt>BasicBlock</tt> constructor is used to create new basic blocks for
3873 insertion into a function. The constructor optionally takes a name for the new
3874 block, and a <a href="#Function"><tt>Function</tt></a> to insert it into. If
3875 the <tt>Parent</tt> parameter is specified, the new <tt>BasicBlock</tt> is
3876 automatically inserted at the end of the specified <a
3877 href="#Function"><tt>Function</tt></a>, if not specified, the BasicBlock must be
3878 manually inserted into the <a href="#Function"><tt>Function</tt></a>.</p></li>
3880 <li><tt>BasicBlock::iterator</tt> - Typedef for instruction list iterator<br>
3881 <tt>BasicBlock::const_iterator</tt> - Typedef for const_iterator.<br>
3882 <tt>begin()</tt>, <tt>end()</tt>, <tt>front()</tt>, <tt>back()</tt>,
3883 <tt>size()</tt>, <tt>empty()</tt>
3884 STL-style functions for accessing the instruction list.
3886 <p>These methods and typedefs are forwarding functions that have the same
3887 semantics as the standard library methods of the same names. These methods
3888 expose the underlying instruction list of a basic block in a way that is easy to
3889 manipulate. To get the full complement of container operations (including
3890 operations to update the list), you must use the <tt>getInstList()</tt>
3891 method.</p></li>
3893 <li><tt>BasicBlock::InstListType &amp;getInstList()</tt>
3895 <p>This method is used to get access to the underlying container that actually
3896 holds the Instructions. This method must be used when there isn't a forwarding
3897 function in the <tt>BasicBlock</tt> class for the operation that you would like
3898 to perform. Because there are no forwarding functions for "updating"
3899 operations, you need to use this if you want to update the contents of a
3900 <tt>BasicBlock</tt>.</p></li>
3902 <li><tt><a href="#Function">Function</a> *getParent()</tt>
3904 <p> Returns a pointer to <a href="#Function"><tt>Function</tt></a> the block is
3905 embedded into, or a null pointer if it is homeless.</p></li>
3907 <li><tt><a href="#TerminatorInst">TerminatorInst</a> *getTerminator()</tt>
3909 <p> Returns a pointer to the terminator instruction that appears at the end of
3910 the <tt>BasicBlock</tt>. If there is no terminator instruction, or if the last
3911 instruction in the block is not a terminator, then a null pointer is
3912 returned.</p></li>
3914 </ul>
3916 </div>
3919 <!-- ======================================================================= -->
3920 <div class="doc_subsection">
3921 <a name="Argument">The <tt>Argument</tt> class</a>
3922 </div>
3924 <div class="doc_text">
3926 <p>This subclass of Value defines the interface for incoming formal
3927 arguments to a function. A Function maintains a list of its formal
3928 arguments. An argument has a pointer to the parent Function.</p>
3930 </div>
3932 <!-- *********************************************************************** -->
3933 <hr>
3934 <address>
3935 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
3936 src="http://jigsaw.w3.org/css-validator/images/vcss-blue" alt="Valid CSS"></a>
3937 <a href="http://validator.w3.org/check/referer"><img
3938 src="http://www.w3.org/Icons/valid-html401" alt="Valid HTML 4.01 Strict"></a>
3940 <a href="mailto:dhurjati@cs.uiuc.edu">Dinakar Dhurjati</a> and
3941 <a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
3942 <a href="http://llvm.org">The LLVM Compiler Infrastructure</a><br>
3943 Last modified: $Date$
3944 </address>
3946 </body>
3947 </html>