1 <!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01//EN"
2 "http://www.w3.org/TR/html4/strict.dtd">
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">
11 <div class="doc_title">
12 LLVM Programmer's Manual
16 <li><a href="#introduction">Introduction</a></li>
17 <li><a href="#general">General Information</a>
19 <li><a href="#stl">The C++ Standard Template Library</a></li>
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>
28 <li><a href="#apis">Important and useful LLVM APIs</a>
30 <li><a href="#isa">The <tt>isa<></tt>, <tt>cast<></tt>
31 and <tt>dyn_cast<></tt> templates</a> </li>
32 <li><a href="#string_apis">Passing strings (the <tt>StringRef</tt>
33 and <tt>Twine</tt> classes)</a>
35 <li><a href="#StringRef">The <tt>StringRef</tt> class</a> </li>
36 <li><a href="#Twine">The <tt>Twine</tt> class</a> </li>
39 <li><a href="#DEBUG">The <tt>DEBUG()</tt> macro and <tt>-debug</tt>
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>
46 <li><a href="#Statistic">The <tt>Statistic</tt> class & <tt>-stats</tt>
49 <li>The <tt>InstVisitor</tt> template
50 <li>The general graph API
52 <li><a href="#ViewGraph">Viewing graphs while debugging code</a></li>
55 <li><a href="#datastructure">Picking the Right Data Structure for a Task</a>
57 <li><a href="#ds_sequential">Sequential Containers (std::vector, std::list, etc)</a>
59 <li><a href="#dss_arrayref">llvm/ADT/ArrayRef.h</a></li>
60 <li><a href="#dss_fixedarrays">Fixed Size Arrays</a></li>
61 <li><a href="#dss_heaparrays">Heap Allocated Arrays</a></li>
62 <li><a href="#dss_smallvector">"llvm/ADT/SmallVector.h"</a></li>
63 <li><a href="#dss_vector"><vector></a></li>
64 <li><a href="#dss_deque"><deque></a></li>
65 <li><a href="#dss_list"><list></a></li>
66 <li><a href="#dss_ilist">llvm/ADT/ilist.h</a></li>
67 <li><a href="#dss_other">Other Sequential Container Options</a></li>
69 <li><a href="#ds_set">Set-Like Containers (std::set, SmallSet, SetVector, etc)</a>
71 <li><a href="#dss_sortedvectorset">A sorted 'vector'</a></li>
72 <li><a href="#dss_smallset">"llvm/ADT/SmallSet.h"</a></li>
73 <li><a href="#dss_smallptrset">"llvm/ADT/SmallPtrSet.h"</a></li>
74 <li><a href="#dss_denseset">"llvm/ADT/DenseSet.h"</a></li>
75 <li><a href="#dss_FoldingSet">"llvm/ADT/FoldingSet.h"</a></li>
76 <li><a href="#dss_set"><set></a></li>
77 <li><a href="#dss_setvector">"llvm/ADT/SetVector.h"</a></li>
78 <li><a href="#dss_uniquevector">"llvm/ADT/UniqueVector.h"</a></li>
79 <li><a href="#dss_otherset">Other Set-Like ContainerOptions</a></li>
81 <li><a href="#ds_map">Map-Like Containers (std::map, DenseMap, etc)</a>
83 <li><a href="#dss_sortedvectormap">A sorted 'vector'</a></li>
84 <li><a href="#dss_stringmap">"llvm/ADT/StringMap.h"</a></li>
85 <li><a href="#dss_indexedmap">"llvm/ADT/IndexedMap.h"</a></li>
86 <li><a href="#dss_densemap">"llvm/ADT/DenseMap.h"</a></li>
87 <li><a href="#dss_valuemap">"llvm/ADT/ValueMap.h"</a></li>
88 <li><a href="#dss_intervalmap">"llvm/ADT/IntervalMap.h"</a></li>
89 <li><a href="#dss_map"><map></a></li>
90 <li><a href="#dss_inteqclasses">"llvm/ADT/IntEqClasses.h"</a></li>
91 <li><a href="#dss_othermap">Other Map-Like Container Options</a></li>
93 <li><a href="#ds_string">String-like containers</a>
97 <li><a href="#ds_bit">BitVector-like containers</a>
99 <li><a href="#dss_bitvector">A dense bitvector</a></li>
100 <li><a href="#dss_smallbitvector">A "small" dense bitvector</a></li>
101 <li><a href="#dss_sparsebitvector">A sparse bitvector</a></li>
105 <li><a href="#common">Helpful Hints for Common Operations</a>
107 <li><a href="#inspection">Basic Inspection and Traversal Routines</a>
109 <li><a href="#iterate_function">Iterating over the <tt>BasicBlock</tt>s
110 in a <tt>Function</tt></a> </li>
111 <li><a href="#iterate_basicblock">Iterating over the <tt>Instruction</tt>s
112 in a <tt>BasicBlock</tt></a> </li>
113 <li><a href="#iterate_institer">Iterating over the <tt>Instruction</tt>s
114 in a <tt>Function</tt></a> </li>
115 <li><a href="#iterate_convert">Turning an iterator into a
116 class pointer</a> </li>
117 <li><a href="#iterate_complex">Finding call sites: a more
118 complex example</a> </li>
119 <li><a href="#calls_and_invokes">Treating calls and invokes
120 the same way</a> </li>
121 <li><a href="#iterate_chains">Iterating over def-use &
122 use-def chains</a> </li>
123 <li><a href="#iterate_preds">Iterating over predecessors &
124 successors of blocks</a></li>
127 <li><a href="#simplechanges">Making simple changes</a>
129 <li><a href="#schanges_creating">Creating and inserting new
130 <tt>Instruction</tt>s</a> </li>
131 <li><a href="#schanges_deleting">Deleting <tt>Instruction</tt>s</a> </li>
132 <li><a href="#schanges_replacing">Replacing an <tt>Instruction</tt>
133 with another <tt>Value</tt></a> </li>
134 <li><a href="#schanges_deletingGV">Deleting <tt>GlobalVariable</tt>s</a> </li>
137 <li><a href="#create_types">How to Create Types</a></li>
139 <li>Working with the Control Flow Graph
141 <li>Accessing predecessors and successors of a <tt>BasicBlock</tt>
149 <li><a href="#threading">Threads and LLVM</a>
151 <li><a href="#startmultithreaded">Entering and Exiting Multithreaded Mode
153 <li><a href="#shutdown">Ending execution with <tt>llvm_shutdown()</tt></a></li>
154 <li><a href="#managedstatic">Lazy initialization with <tt>ManagedStatic</tt></a></li>
155 <li><a href="#llvmcontext">Achieving Isolation with <tt>LLVMContext</tt></a></li>
156 <li><a href="#jitthreading">Threads and the JIT</a></li>
160 <li><a href="#advanced">Advanced Topics</a>
162 <li><a href="#TypeResolve">LLVM Type Resolution</a>
164 <li><a href="#BuildRecType">Basic Recursive Type Construction</a></li>
165 <li><a href="#refineAbstractTypeTo">The <tt>refineAbstractTypeTo</tt> method</a></li>
166 <li><a href="#PATypeHolder">The PATypeHolder Class</a></li>
167 <li><a href="#AbstractTypeUser">The AbstractTypeUser Class</a></li>
170 <li><a href="#SymbolTable">The <tt>ValueSymbolTable</tt> and <tt>TypeSymbolTable</tt> classes</a></li>
171 <li><a href="#UserLayout">The <tt>User</tt> and owned <tt>Use</tt> classes' memory layout</a></li>
174 <li><a href="#coreclasses">The Core LLVM Class Hierarchy Reference</a>
176 <li><a href="#Type">The <tt>Type</tt> class</a> </li>
177 <li><a href="#Module">The <tt>Module</tt> class</a></li>
178 <li><a href="#Value">The <tt>Value</tt> class</a>
180 <li><a href="#User">The <tt>User</tt> class</a>
182 <li><a href="#Instruction">The <tt>Instruction</tt> class</a></li>
183 <li><a href="#Constant">The <tt>Constant</tt> class</a>
185 <li><a href="#GlobalValue">The <tt>GlobalValue</tt> class</a>
187 <li><a href="#Function">The <tt>Function</tt> class</a></li>
188 <li><a href="#GlobalVariable">The <tt>GlobalVariable</tt> class</a></li>
195 <li><a href="#BasicBlock">The <tt>BasicBlock</tt> class</a></li>
196 <li><a href="#Argument">The <tt>Argument</tt> class</a></li>
203 <div class="doc_author">
204 <p>Written by <a href="mailto:sabre@nondot.org">Chris Lattner</a>,
205 <a href="mailto:dhurjati@cs.uiuc.edu">Dinakar Dhurjati</a>,
206 <a href="mailto:ggreif@gmail.com">Gabor Greif</a>,
207 <a href="mailto:jstanley@cs.uiuc.edu">Joel Stanley</a>,
208 <a href="mailto:rspencer@x10sys.com">Reid Spencer</a> and
209 <a href="mailto:owen@apple.com">Owen Anderson</a></p>
212 <!-- *********************************************************************** -->
213 <div class="doc_section">
214 <a name="introduction">Introduction </a>
216 <!-- *********************************************************************** -->
218 <div class="doc_text">
220 <p>This document is meant to highlight some of the important classes and
221 interfaces available in the LLVM source-base. This manual is not
222 intended to explain what LLVM is, how it works, and what LLVM code looks
223 like. It assumes that you know the basics of LLVM and are interested
224 in writing transformations or otherwise analyzing or manipulating the
227 <p>This document should get you oriented so that you can find your
228 way in the continuously growing source code that makes up the LLVM
229 infrastructure. Note that this manual is not intended to serve as a
230 replacement for reading the source code, so if you think there should be
231 a method in one of these classes to do something, but it's not listed,
232 check the source. Links to the <a href="/doxygen/">doxygen</a> sources
233 are provided to make this as easy as possible.</p>
235 <p>The first section of this document describes general information that is
236 useful to know when working in the LLVM infrastructure, and the second describes
237 the Core LLVM classes. In the future this manual will be extended with
238 information describing how to use extension libraries, such as dominator
239 information, CFG traversal routines, and useful utilities like the <tt><a
240 href="/doxygen/InstVisitor_8h-source.html">InstVisitor</a></tt> template.</p>
244 <!-- *********************************************************************** -->
245 <div class="doc_section">
246 <a name="general">General Information</a>
248 <!-- *********************************************************************** -->
250 <div class="doc_text">
252 <p>This section contains general information that is useful if you are working
253 in the LLVM source-base, but that isn't specific to any particular API.</p>
257 <!-- ======================================================================= -->
258 <div class="doc_subsection">
259 <a name="stl">The C++ Standard Template Library</a>
262 <div class="doc_text">
264 <p>LLVM makes heavy use of the C++ Standard Template Library (STL),
265 perhaps much more than you are used to, or have seen before. Because of
266 this, you might want to do a little background reading in the
267 techniques used and capabilities of the library. There are many good
268 pages that discuss the STL, and several books on the subject that you
269 can get, so it will not be discussed in this document.</p>
271 <p>Here are some useful links:</p>
275 <li><a href="http://www.dinkumware.com/manuals/#Standard C++ Library">Dinkumware
276 C++ Library reference</a> - an excellent reference for the STL and other parts
277 of the standard C++ library.</li>
279 <li><a href="http://www.tempest-sw.com/cpp/">C++ In a Nutshell</a> - This is an
280 O'Reilly book in the making. It has a decent Standard Library
281 Reference that rivals Dinkumware's, and is unfortunately no longer free since the
282 book has been published.</li>
284 <li><a href="http://www.parashift.com/c++-faq-lite/">C++ Frequently Asked
287 <li><a href="http://www.sgi.com/tech/stl/">SGI's STL Programmer's Guide</a> -
289 href="http://www.sgi.com/tech/stl/stl_introduction.html">Introduction to the
292 <li><a href="http://www.research.att.com/%7Ebs/C++.html">Bjarne Stroustrup's C++
295 <li><a href="http://64.78.49.204/">
296 Bruce Eckel's Thinking in C++, 2nd ed. Volume 2 Revision 4.0 (even better, get
301 <p>You are also encouraged to take a look at the <a
302 href="CodingStandards.html">LLVM Coding Standards</a> guide which focuses on how
303 to write maintainable code more than where to put your curly braces.</p>
307 <!-- ======================================================================= -->
308 <div class="doc_subsection">
309 <a name="stl">Other useful references</a>
312 <div class="doc_text">
315 <li><a href="http://www.fortran-2000.com/ArnaudRecipes/sharedlib.html">Using
316 static and shared libraries across platforms</a></li>
321 <!-- *********************************************************************** -->
322 <div class="doc_section">
323 <a name="apis">Important and useful LLVM APIs</a>
325 <!-- *********************************************************************** -->
327 <div class="doc_text">
329 <p>Here we highlight some LLVM APIs that are generally useful and good to
330 know about when writing transformations.</p>
334 <!-- ======================================================================= -->
335 <div class="doc_subsection">
336 <a name="isa">The <tt>isa<></tt>, <tt>cast<></tt> and
337 <tt>dyn_cast<></tt> templates</a>
340 <div class="doc_text">
342 <p>The LLVM source-base makes extensive use of a custom form of RTTI.
343 These templates have many similarities to the C++ <tt>dynamic_cast<></tt>
344 operator, but they don't have some drawbacks (primarily stemming from
345 the fact that <tt>dynamic_cast<></tt> only works on classes that
346 have a v-table). Because they are used so often, you must know what they
347 do and how they work. All of these templates are defined in the <a
348 href="/doxygen/Casting_8h-source.html"><tt>llvm/Support/Casting.h</tt></a>
349 file (note that you very rarely have to include this file directly).</p>
352 <dt><tt>isa<></tt>: </dt>
354 <dd><p>The <tt>isa<></tt> operator works exactly like the Java
355 "<tt>instanceof</tt>" operator. It returns true or false depending on whether
356 a reference or pointer points to an instance of the specified class. This can
357 be very useful for constraint checking of various sorts (example below).</p>
360 <dt><tt>cast<></tt>: </dt>
362 <dd><p>The <tt>cast<></tt> operator is a "checked cast" operation. It
363 converts a pointer or reference from a base class to a derived class, causing
364 an assertion failure if it is not really an instance of the right type. This
365 should be used in cases where you have some information that makes you believe
366 that something is of the right type. An example of the <tt>isa<></tt>
367 and <tt>cast<></tt> template is:</p>
369 <div class="doc_code">
371 static bool isLoopInvariant(const <a href="#Value">Value</a> *V, const Loop *L) {
372 if (isa<<a href="#Constant">Constant</a>>(V) || isa<<a href="#Argument">Argument</a>>(V) || isa<<a href="#GlobalValue">GlobalValue</a>>(V))
375 // <i>Otherwise, it must be an instruction...</i>
376 return !L->contains(cast<<a href="#Instruction">Instruction</a>>(V)->getParent());
381 <p>Note that you should <b>not</b> use an <tt>isa<></tt> test followed
382 by a <tt>cast<></tt>, for that use the <tt>dyn_cast<></tt>
387 <dt><tt>dyn_cast<></tt>:</dt>
389 <dd><p>The <tt>dyn_cast<></tt> operator is a "checking cast" operation.
390 It checks to see if the operand is of the specified type, and if so, returns a
391 pointer to it (this operator does not work with references). If the operand is
392 not of the correct type, a null pointer is returned. Thus, this works very
393 much like the <tt>dynamic_cast<></tt> operator in C++, and should be
394 used in the same circumstances. Typically, the <tt>dyn_cast<></tt>
395 operator is used in an <tt>if</tt> statement or some other flow control
396 statement like this:</p>
398 <div class="doc_code">
400 if (<a href="#AllocationInst">AllocationInst</a> *AI = dyn_cast<<a href="#AllocationInst">AllocationInst</a>>(Val)) {
406 <p>This form of the <tt>if</tt> statement effectively combines together a call
407 to <tt>isa<></tt> and a call to <tt>cast<></tt> into one
408 statement, which is very convenient.</p>
410 <p>Note that the <tt>dyn_cast<></tt> operator, like C++'s
411 <tt>dynamic_cast<></tt> or Java's <tt>instanceof</tt> operator, can be
412 abused. In particular, you should not use big chained <tt>if/then/else</tt>
413 blocks to check for lots of different variants of classes. If you find
414 yourself wanting to do this, it is much cleaner and more efficient to use the
415 <tt>InstVisitor</tt> class to dispatch over the instruction type directly.</p>
419 <dt><tt>cast_or_null<></tt>: </dt>
421 <dd><p>The <tt>cast_or_null<></tt> operator works just like the
422 <tt>cast<></tt> operator, except that it allows for a null pointer as an
423 argument (which it then propagates). This can sometimes be useful, allowing
424 you to combine several null checks into one.</p></dd>
426 <dt><tt>dyn_cast_or_null<></tt>: </dt>
428 <dd><p>The <tt>dyn_cast_or_null<></tt> operator works just like the
429 <tt>dyn_cast<></tt> operator, except that it allows for a null pointer
430 as an argument (which it then propagates). This can sometimes be useful,
431 allowing you to combine several null checks into one.</p></dd>
435 <p>These five templates can be used with any classes, whether they have a
436 v-table or not. To add support for these templates, you simply need to add
437 <tt>classof</tt> static methods to the class you are interested casting
438 to. Describing this is currently outside the scope of this document, but there
439 are lots of examples in the LLVM source base.</p>
444 <!-- ======================================================================= -->
445 <div class="doc_subsection">
446 <a name="string_apis">Passing strings (the <tt>StringRef</tt>
447 and <tt>Twine</tt> classes)</a>
450 <div class="doc_text">
452 <p>Although LLVM generally does not do much string manipulation, we do have
453 several important APIs which take strings. Two important examples are the
454 Value class -- which has names for instructions, functions, etc. -- and the
455 StringMap class which is used extensively in LLVM and Clang.</p>
457 <p>These are generic classes, and they need to be able to accept strings which
458 may have embedded null characters. Therefore, they cannot simply take
459 a <tt>const char *</tt>, and taking a <tt>const std::string&</tt> requires
460 clients to perform a heap allocation which is usually unnecessary. Instead,
461 many LLVM APIs use a <tt>StringRef</tt> or a <tt>const Twine&</tt> for
462 passing strings efficiently.</p>
466 <!-- _______________________________________________________________________ -->
467 <div class="doc_subsubsection">
468 <a name="StringRef">The <tt>StringRef</tt> class</a>
471 <div class="doc_text">
473 <p>The <tt>StringRef</tt> data type represents a reference to a constant string
474 (a character array and a length) and supports the common operations available
475 on <tt>std:string</tt>, but does not require heap allocation.</p>
477 <p>It can be implicitly constructed using a C style null-terminated string,
478 an <tt>std::string</tt>, or explicitly with a character pointer and length.
479 For example, the <tt>StringRef</tt> find function is declared as:</p>
481 <pre class="doc_code">
482 iterator find(StringRef Key);
485 <p>and clients can call it using any one of:</p>
487 <pre class="doc_code">
488 Map.find("foo"); <i>// Lookup "foo"</i>
489 Map.find(std::string("bar")); <i>// Lookup "bar"</i>
490 Map.find(StringRef("\0baz", 4)); <i>// Lookup "\0baz"</i>
493 <p>Similarly, APIs which need to return a string may return a <tt>StringRef</tt>
494 instance, which can be used directly or converted to an <tt>std::string</tt>
495 using the <tt>str</tt> member function. See
496 "<tt><a href="/doxygen/classllvm_1_1StringRef_8h-source.html">llvm/ADT/StringRef.h</a></tt>"
497 for more information.</p>
499 <p>You should rarely use the <tt>StringRef</tt> class directly, because it contains
500 pointers to external memory it is not generally safe to store an instance of the
501 class (unless you know that the external storage will not be freed). StringRef is
502 small and pervasive enough in LLVM that it should always be passed by value.</p>
506 <!-- _______________________________________________________________________ -->
507 <div class="doc_subsubsection">
508 <a name="Twine">The <tt>Twine</tt> class</a>
511 <div class="doc_text">
513 <p>The <tt>Twine</tt> class is an efficient way for APIs to accept concatenated
514 strings. For example, a common LLVM paradigm is to name one instruction based on
515 the name of another instruction with a suffix, for example:</p>
517 <div class="doc_code">
519 New = CmpInst::Create(<i>...</i>, SO->getName() + ".cmp");
523 <p>The <tt>Twine</tt> class is effectively a
524 lightweight <a href="http://en.wikipedia.org/wiki/Rope_(computer_science)">rope</a>
525 which points to temporary (stack allocated) objects. Twines can be implicitly
526 constructed as the result of the plus operator applied to strings (i.e., a C
527 strings, an <tt>std::string</tt>, or a <tt>StringRef</tt>). The twine delays the
528 actual concatenation of strings until it is actually required, at which point
529 it can be efficiently rendered directly into a character array. This avoids
530 unnecessary heap allocation involved in constructing the temporary results of
531 string concatenation. See
532 "<tt><a href="/doxygen/classllvm_1_1Twine_8h-source.html">llvm/ADT/Twine.h</a></tt>"
533 for more information.</p>
535 <p>As with a <tt>StringRef</tt>, <tt>Twine</tt> objects point to external memory
536 and should almost never be stored or mentioned directly. They are intended
537 solely for use when defining a function which should be able to efficiently
538 accept concatenated strings.</p>
543 <!-- ======================================================================= -->
544 <div class="doc_subsection">
545 <a name="DEBUG">The <tt>DEBUG()</tt> macro and <tt>-debug</tt> option</a>
548 <div class="doc_text">
550 <p>Often when working on your pass you will put a bunch of debugging printouts
551 and other code into your pass. After you get it working, you want to remove
552 it, but you may need it again in the future (to work out new bugs that you run
555 <p> Naturally, because of this, you don't want to delete the debug printouts,
556 but you don't want them to always be noisy. A standard compromise is to comment
557 them out, allowing you to enable them if you need them in the future.</p>
559 <p>The "<tt><a href="/doxygen/Debug_8h-source.html">llvm/Support/Debug.h</a></tt>"
560 file provides a macro named <tt>DEBUG()</tt> that is a much nicer solution to
561 this problem. Basically, you can put arbitrary code into the argument of the
562 <tt>DEBUG</tt> macro, and it is only executed if '<tt>opt</tt>' (or any other
563 tool) is run with the '<tt>-debug</tt>' command line argument:</p>
565 <div class="doc_code">
567 DEBUG(errs() << "I am here!\n");
571 <p>Then you can run your pass like this:</p>
573 <div class="doc_code">
575 $ opt < a.bc > /dev/null -mypass
576 <i><no output></i>
577 $ opt < a.bc > /dev/null -mypass -debug
582 <p>Using the <tt>DEBUG()</tt> macro instead of a home-brewed solution allows you
583 to not have to create "yet another" command line option for the debug output for
584 your pass. Note that <tt>DEBUG()</tt> macros are disabled for optimized builds,
585 so they do not cause a performance impact at all (for the same reason, they
586 should also not contain side-effects!).</p>
588 <p>One additional nice thing about the <tt>DEBUG()</tt> macro is that you can
589 enable or disable it directly in gdb. Just use "<tt>set DebugFlag=0</tt>" or
590 "<tt>set DebugFlag=1</tt>" from the gdb if the program is running. If the
591 program hasn't been started yet, you can always just run it with
596 <!-- _______________________________________________________________________ -->
597 <div class="doc_subsubsection">
598 <a name="DEBUG_TYPE">Fine grained debug info with <tt>DEBUG_TYPE</tt> and
599 the <tt>-debug-only</tt> option</a>
602 <div class="doc_text">
604 <p>Sometimes you may find yourself in a situation where enabling <tt>-debug</tt>
605 just turns on <b>too much</b> information (such as when working on the code
606 generator). If you want to enable debug information with more fine-grained
607 control, you define the <tt>DEBUG_TYPE</tt> macro and the <tt>-debug</tt> only
608 option as follows:</p>
610 <div class="doc_code">
613 DEBUG(errs() << "No debug type\n");
614 #define DEBUG_TYPE "foo"
615 DEBUG(errs() << "'foo' debug type\n");
617 #define DEBUG_TYPE "bar"
618 DEBUG(errs() << "'bar' debug type\n"));
620 #define DEBUG_TYPE ""
621 DEBUG(errs() << "No debug type (2)\n");
625 <p>Then you can run your pass like this:</p>
627 <div class="doc_code">
629 $ opt < a.bc > /dev/null -mypass
630 <i><no output></i>
631 $ opt < a.bc > /dev/null -mypass -debug
636 $ opt < a.bc > /dev/null -mypass -debug-only=foo
638 $ opt < a.bc > /dev/null -mypass -debug-only=bar
643 <p>Of course, in practice, you should only set <tt>DEBUG_TYPE</tt> at the top of
644 a file, to specify the debug type for the entire module (if you do this before
645 you <tt>#include "llvm/Support/Debug.h"</tt>, you don't have to insert the ugly
646 <tt>#undef</tt>'s). Also, you should use names more meaningful than "foo" and
647 "bar", because there is no system in place to ensure that names do not
648 conflict. If two different modules use the same string, they will all be turned
649 on when the name is specified. This allows, for example, all debug information
650 for instruction scheduling to be enabled with <tt>-debug-type=InstrSched</tt>,
651 even if the source lives in multiple files.</p>
653 <p>The <tt>DEBUG_WITH_TYPE</tt> macro is also available for situations where you
654 would like to set <tt>DEBUG_TYPE</tt>, but only for one specific <tt>DEBUG</tt>
655 statement. It takes an additional first parameter, which is the type to use. For
656 example, the preceding example could be written as:</p>
659 <div class="doc_code">
661 DEBUG_WITH_TYPE("", errs() << "No debug type\n");
662 DEBUG_WITH_TYPE("foo", errs() << "'foo' debug type\n");
663 DEBUG_WITH_TYPE("bar", errs() << "'bar' debug type\n"));
664 DEBUG_WITH_TYPE("", errs() << "No debug type (2)\n");
670 <!-- ======================================================================= -->
671 <div class="doc_subsection">
672 <a name="Statistic">The <tt>Statistic</tt> class & <tt>-stats</tt>
676 <div class="doc_text">
679 href="/doxygen/Statistic_8h-source.html">llvm/ADT/Statistic.h</a></tt>" file
680 provides a class named <tt>Statistic</tt> that is used as a unified way to
681 keep track of what the LLVM compiler is doing and how effective various
682 optimizations are. It is useful to see what optimizations are contributing to
683 making a particular program run faster.</p>
685 <p>Often you may run your pass on some big program, and you're interested to see
686 how many times it makes a certain transformation. Although you can do this with
687 hand inspection, or some ad-hoc method, this is a real pain and not very useful
688 for big programs. Using the <tt>Statistic</tt> class makes it very easy to
689 keep track of this information, and the calculated information is presented in a
690 uniform manner with the rest of the passes being executed.</p>
692 <p>There are many examples of <tt>Statistic</tt> uses, but the basics of using
693 it are as follows:</p>
696 <li><p>Define your statistic like this:</p>
698 <div class="doc_code">
700 #define <a href="#DEBUG_TYPE">DEBUG_TYPE</a> "mypassname" <i>// This goes before any #includes.</i>
701 STATISTIC(NumXForms, "The # of times I did stuff");
705 <p>The <tt>STATISTIC</tt> macro defines a static variable, whose name is
706 specified by the first argument. The pass name is taken from the DEBUG_TYPE
707 macro, and the description is taken from the second argument. The variable
708 defined ("NumXForms" in this case) acts like an unsigned integer.</p></li>
710 <li><p>Whenever you make a transformation, bump the counter:</p>
712 <div class="doc_code">
714 ++NumXForms; // <i>I did stuff!</i>
721 <p>That's all you have to do. To get '<tt>opt</tt>' to print out the
722 statistics gathered, use the '<tt>-stats</tt>' option:</p>
724 <div class="doc_code">
726 $ opt -stats -mypassname < program.bc > /dev/null
727 <i>... statistics output ...</i>
731 <p> When running <tt>opt</tt> on a C file from the SPEC benchmark
732 suite, it gives a report that looks like this:</p>
734 <div class="doc_code">
736 7646 bitcodewriter - Number of normal instructions
737 725 bitcodewriter - Number of oversized instructions
738 129996 bitcodewriter - Number of bitcode bytes written
739 2817 raise - Number of insts DCEd or constprop'd
740 3213 raise - Number of cast-of-self removed
741 5046 raise - Number of expression trees converted
742 75 raise - Number of other getelementptr's formed
743 138 raise - Number of load/store peepholes
744 42 deadtypeelim - Number of unused typenames removed from symtab
745 392 funcresolve - Number of varargs functions resolved
746 27 globaldce - Number of global variables removed
747 2 adce - Number of basic blocks removed
748 134 cee - Number of branches revectored
749 49 cee - Number of setcc instruction eliminated
750 532 gcse - Number of loads removed
751 2919 gcse - Number of instructions removed
752 86 indvars - Number of canonical indvars added
753 87 indvars - Number of aux indvars removed
754 25 instcombine - Number of dead inst eliminate
755 434 instcombine - Number of insts combined
756 248 licm - Number of load insts hoisted
757 1298 licm - Number of insts hoisted to a loop pre-header
758 3 licm - Number of insts hoisted to multiple loop preds (bad, no loop pre-header)
759 75 mem2reg - Number of alloca's promoted
760 1444 cfgsimplify - Number of blocks simplified
764 <p>Obviously, with so many optimizations, having a unified framework for this
765 stuff is very nice. Making your pass fit well into the framework makes it more
766 maintainable and useful.</p>
770 <!-- ======================================================================= -->
771 <div class="doc_subsection">
772 <a name="ViewGraph">Viewing graphs while debugging code</a>
775 <div class="doc_text">
777 <p>Several of the important data structures in LLVM are graphs: for example
778 CFGs made out of LLVM <a href="#BasicBlock">BasicBlock</a>s, CFGs made out of
779 LLVM <a href="CodeGenerator.html#machinebasicblock">MachineBasicBlock</a>s, and
780 <a href="CodeGenerator.html#selectiondag_intro">Instruction Selection
781 DAGs</a>. In many cases, while debugging various parts of the compiler, it is
782 nice to instantly visualize these graphs.</p>
784 <p>LLVM provides several callbacks that are available in a debug build to do
785 exactly that. If you call the <tt>Function::viewCFG()</tt> method, for example,
786 the current LLVM tool will pop up a window containing the CFG for the function
787 where each basic block is a node in the graph, and each node contains the
788 instructions in the block. Similarly, there also exists
789 <tt>Function::viewCFGOnly()</tt> (does not include the instructions), the
790 <tt>MachineFunction::viewCFG()</tt> and <tt>MachineFunction::viewCFGOnly()</tt>,
791 and the <tt>SelectionDAG::viewGraph()</tt> methods. Within GDB, for example,
792 you can usually use something like <tt>call DAG.viewGraph()</tt> to pop
793 up a window. Alternatively, you can sprinkle calls to these functions in your
794 code in places you want to debug.</p>
796 <p>Getting this to work requires a small amount of configuration. On Unix
797 systems with X11, install the <a href="http://www.graphviz.org">graphviz</a>
798 toolkit, and make sure 'dot' and 'gv' are in your path. If you are running on
799 Mac OS/X, download and install the Mac OS/X <a
800 href="http://www.pixelglow.com/graphviz/">Graphviz program</a>, and add
801 <tt>/Applications/Graphviz.app/Contents/MacOS/</tt> (or wherever you install
802 it) to your path. Once in your system and path are set up, rerun the LLVM
803 configure script and rebuild LLVM to enable this functionality.</p>
805 <p><tt>SelectionDAG</tt> has been extended to make it easier to locate
806 <i>interesting</i> nodes in large complex graphs. From gdb, if you
807 <tt>call DAG.setGraphColor(<i>node</i>, "<i>color</i>")</tt>, then the
808 next <tt>call DAG.viewGraph()</tt> would highlight the node in the
809 specified color (choices of colors can be found at <a
810 href="http://www.graphviz.org/doc/info/colors.html">colors</a>.) More
811 complex node attributes can be provided with <tt>call
812 DAG.setGraphAttrs(<i>node</i>, "<i>attributes</i>")</tt> (choices can be
813 found at <a href="http://www.graphviz.org/doc/info/attrs.html">Graph
814 Attributes</a>.) If you want to restart and clear all the current graph
815 attributes, then you can <tt>call DAG.clearGraphAttrs()</tt>. </p>
819 <!-- *********************************************************************** -->
820 <div class="doc_section">
821 <a name="datastructure">Picking the Right Data Structure for a Task</a>
823 <!-- *********************************************************************** -->
825 <div class="doc_text">
827 <p>LLVM has a plethora of data structures in the <tt>llvm/ADT/</tt> directory,
828 and we commonly use STL data structures. This section describes the trade-offs
829 you should consider when you pick one.</p>
832 The first step is a choose your own adventure: do you want a sequential
833 container, a set-like container, or a map-like container? The most important
834 thing when choosing a container is the algorithmic properties of how you plan to
835 access the container. Based on that, you should use:</p>
838 <li>a <a href="#ds_map">map-like</a> container if you need efficient look-up
839 of an value based on another value. Map-like containers also support
840 efficient queries for containment (whether a key is in the map). Map-like
841 containers generally do not support efficient reverse mapping (values to
842 keys). If you need that, use two maps. Some map-like containers also
843 support efficient iteration through the keys in sorted order. Map-like
844 containers are the most expensive sort, only use them if you need one of
845 these capabilities.</li>
847 <li>a <a href="#ds_set">set-like</a> container if you need to put a bunch of
848 stuff into a container that automatically eliminates duplicates. Some
849 set-like containers support efficient iteration through the elements in
850 sorted order. Set-like containers are more expensive than sequential
854 <li>a <a href="#ds_sequential">sequential</a> container provides
855 the most efficient way to add elements and keeps track of the order they are
856 added to the collection. They permit duplicates and support efficient
857 iteration, but do not support efficient look-up based on a key.
860 <li>a <a href="#ds_string">string</a> container is a specialized sequential
861 container or reference structure that is used for character or byte
864 <li>a <a href="#ds_bit">bit</a> container provides an efficient way to store and
865 perform set operations on sets of numeric id's, while automatically
866 eliminating duplicates. Bit containers require a maximum of 1 bit for each
867 identifier you want to store.
872 Once the proper category of container is determined, you can fine tune the
873 memory use, constant factors, and cache behaviors of access by intelligently
874 picking a member of the category. Note that constant factors and cache behavior
875 can be a big deal. If you have a vector that usually only contains a few
876 elements (but could contain many), for example, it's much better to use
877 <a href="#dss_smallvector">SmallVector</a> than <a href="#dss_vector">vector</a>
878 . Doing so avoids (relatively) expensive malloc/free calls, which dwarf the
879 cost of adding the elements to the container. </p>
883 <!-- ======================================================================= -->
884 <div class="doc_subsection">
885 <a name="ds_sequential">Sequential Containers (std::vector, std::list, etc)</a>
888 <div class="doc_text">
889 There are a variety of sequential containers available for you, based on your
890 needs. Pick the first in this section that will do what you want.
893 <!-- _______________________________________________________________________ -->
894 <div class="doc_subsubsection">
895 <a name="dss_arrayref">llvm/ADT/ArrayRef.h</a>
898 <div class="doc_text">
899 <p>The llvm::ArrayRef class is the preferred class to use in an interface that
900 accepts a sequential list of elements in memory and just reads from them. By
901 taking an ArrayRef, the API can be passed a fixed size array, an std::vector,
902 an llvm::SmallVector and anything else that is contiguous in memory.
908 <!-- _______________________________________________________________________ -->
909 <div class="doc_subsubsection">
910 <a name="dss_fixedarrays">Fixed Size Arrays</a>
913 <div class="doc_text">
914 <p>Fixed size arrays are very simple and very fast. They are good if you know
915 exactly how many elements you have, or you have a (low) upper bound on how many
919 <!-- _______________________________________________________________________ -->
920 <div class="doc_subsubsection">
921 <a name="dss_heaparrays">Heap Allocated Arrays</a>
924 <div class="doc_text">
925 <p>Heap allocated arrays (new[] + delete[]) are also simple. They are good if
926 the number of elements is variable, if you know how many elements you will need
927 before the array is allocated, and if the array is usually large (if not,
928 consider a <a href="#dss_smallvector">SmallVector</a>). The cost of a heap
929 allocated array is the cost of the new/delete (aka malloc/free). Also note that
930 if you are allocating an array of a type with a constructor, the constructor and
931 destructors will be run for every element in the array (re-sizable vectors only
932 construct those elements actually used).</p>
935 <!-- _______________________________________________________________________ -->
936 <div class="doc_subsubsection">
937 <a name="dss_smallvector">"llvm/ADT/SmallVector.h"</a>
940 <div class="doc_text">
941 <p><tt>SmallVector<Type, N></tt> is a simple class that looks and smells
942 just like <tt>vector<Type></tt>:
943 it supports efficient iteration, lays out elements in memory order (so you can
944 do pointer arithmetic between elements), supports efficient push_back/pop_back
945 operations, supports efficient random access to its elements, etc.</p>
947 <p>The advantage of SmallVector is that it allocates space for
948 some number of elements (N) <b>in the object itself</b>. Because of this, if
949 the SmallVector is dynamically smaller than N, no malloc is performed. This can
950 be a big win in cases where the malloc/free call is far more expensive than the
951 code that fiddles around with the elements.</p>
953 <p>This is good for vectors that are "usually small" (e.g. the number of
954 predecessors/successors of a block is usually less than 8). On the other hand,
955 this makes the size of the SmallVector itself large, so you don't want to
956 allocate lots of them (doing so will waste a lot of space). As such,
957 SmallVectors are most useful when on the stack.</p>
959 <p>SmallVector also provides a nice portable and efficient replacement for
964 <!-- _______________________________________________________________________ -->
965 <div class="doc_subsubsection">
966 <a name="dss_vector"><vector></a>
969 <div class="doc_text">
971 std::vector is well loved and respected. It is useful when SmallVector isn't:
972 when the size of the vector is often large (thus the small optimization will
973 rarely be a benefit) or if you will be allocating many instances of the vector
974 itself (which would waste space for elements that aren't in the container).
975 vector is also useful when interfacing with code that expects vectors :).
978 <p>One worthwhile note about std::vector: avoid code like this:</p>
980 <div class="doc_code">
983 std::vector<foo> V;
989 <p>Instead, write this as:</p>
991 <div class="doc_code">
993 std::vector<foo> V;
1001 <p>Doing so will save (at least) one heap allocation and free per iteration of
1006 <!-- _______________________________________________________________________ -->
1007 <div class="doc_subsubsection">
1008 <a name="dss_deque"><deque></a>
1011 <div class="doc_text">
1012 <p>std::deque is, in some senses, a generalized version of std::vector. Like
1013 std::vector, it provides constant time random access and other similar
1014 properties, but it also provides efficient access to the front of the list. It
1015 does not guarantee continuity of elements within memory.</p>
1017 <p>In exchange for this extra flexibility, std::deque has significantly higher
1018 constant factor costs than std::vector. If possible, use std::vector or
1019 something cheaper.</p>
1022 <!-- _______________________________________________________________________ -->
1023 <div class="doc_subsubsection">
1024 <a name="dss_list"><list></a>
1027 <div class="doc_text">
1028 <p>std::list is an extremely inefficient class that is rarely useful.
1029 It performs a heap allocation for every element inserted into it, thus having an
1030 extremely high constant factor, particularly for small data types. std::list
1031 also only supports bidirectional iteration, not random access iteration.</p>
1033 <p>In exchange for this high cost, std::list supports efficient access to both
1034 ends of the list (like std::deque, but unlike std::vector or SmallVector). In
1035 addition, the iterator invalidation characteristics of std::list are stronger
1036 than that of a vector class: inserting or removing an element into the list does
1037 not invalidate iterator or pointers to other elements in the list.</p>
1040 <!-- _______________________________________________________________________ -->
1041 <div class="doc_subsubsection">
1042 <a name="dss_ilist">llvm/ADT/ilist.h</a>
1045 <div class="doc_text">
1046 <p><tt>ilist<T></tt> implements an 'intrusive' doubly-linked list. It is
1047 intrusive, because it requires the element to store and provide access to the
1048 prev/next pointers for the list.</p>
1050 <p><tt>ilist</tt> has the same drawbacks as <tt>std::list</tt>, and additionally
1051 requires an <tt>ilist_traits</tt> implementation for the element type, but it
1052 provides some novel characteristics. In particular, it can efficiently store
1053 polymorphic objects, the traits class is informed when an element is inserted or
1054 removed from the list, and <tt>ilist</tt>s are guaranteed to support a
1055 constant-time splice operation.</p>
1057 <p>These properties are exactly what we want for things like
1058 <tt>Instruction</tt>s and basic blocks, which is why these are implemented with
1059 <tt>ilist</tt>s.</p>
1061 Related classes of interest are explained in the following subsections:
1063 <li><a href="#dss_ilist_traits">ilist_traits</a></li>
1064 <li><a href="#dss_iplist">iplist</a></li>
1065 <li><a href="#dss_ilist_node">llvm/ADT/ilist_node.h</a></li>
1066 <li><a href="#dss_ilist_sentinel">Sentinels</a></li>
1070 <!-- _______________________________________________________________________ -->
1071 <div class="doc_subsubsection">
1072 <a name="dss_ilist_traits">ilist_traits</a>
1075 <div class="doc_text">
1076 <p><tt>ilist_traits<T></tt> is <tt>ilist<T></tt>'s customization
1077 mechanism. <tt>iplist<T></tt> (and consequently <tt>ilist<T></tt>)
1078 publicly derive from this traits class.</p>
1081 <!-- _______________________________________________________________________ -->
1082 <div class="doc_subsubsection">
1083 <a name="dss_iplist">iplist</a>
1086 <div class="doc_text">
1087 <p><tt>iplist<T></tt> is <tt>ilist<T></tt>'s base and as such
1088 supports a slightly narrower interface. Notably, inserters from
1089 <tt>T&</tt> are absent.</p>
1091 <p><tt>ilist_traits<T></tt> is a public base of this class and can be
1092 used for a wide variety of customizations.</p>
1095 <!-- _______________________________________________________________________ -->
1096 <div class="doc_subsubsection">
1097 <a name="dss_ilist_node">llvm/ADT/ilist_node.h</a>
1100 <div class="doc_text">
1101 <p><tt>ilist_node<T></tt> implements a the forward and backward links
1102 that are expected by the <tt>ilist<T></tt> (and analogous containers)
1103 in the default manner.</p>
1105 <p><tt>ilist_node<T></tt>s are meant to be embedded in the node type
1106 <tt>T</tt>, usually <tt>T</tt> publicly derives from
1107 <tt>ilist_node<T></tt>.</p>
1110 <!-- _______________________________________________________________________ -->
1111 <div class="doc_subsubsection">
1112 <a name="dss_ilist_sentinel">Sentinels</a>
1115 <div class="doc_text">
1116 <p><tt>ilist</tt>s have another specialty that must be considered. To be a good
1117 citizen in the C++ ecosystem, it needs to support the standard container
1118 operations, such as <tt>begin</tt> and <tt>end</tt> iterators, etc. Also, the
1119 <tt>operator--</tt> must work correctly on the <tt>end</tt> iterator in the
1120 case of non-empty <tt>ilist</tt>s.</p>
1122 <p>The only sensible solution to this problem is to allocate a so-called
1123 <i>sentinel</i> along with the intrusive list, which serves as the <tt>end</tt>
1124 iterator, providing the back-link to the last element. However conforming to the
1125 C++ convention it is illegal to <tt>operator++</tt> beyond the sentinel and it
1126 also must not be dereferenced.</p>
1128 <p>These constraints allow for some implementation freedom to the <tt>ilist</tt>
1129 how to allocate and store the sentinel. The corresponding policy is dictated
1130 by <tt>ilist_traits<T></tt>. By default a <tt>T</tt> gets heap-allocated
1131 whenever the need for a sentinel arises.</p>
1133 <p>While the default policy is sufficient in most cases, it may break down when
1134 <tt>T</tt> does not provide a default constructor. Also, in the case of many
1135 instances of <tt>ilist</tt>s, the memory overhead of the associated sentinels
1136 is wasted. To alleviate the situation with numerous and voluminous
1137 <tt>T</tt>-sentinels, sometimes a trick is employed, leading to <i>ghostly
1140 <p>Ghostly sentinels are obtained by specially-crafted <tt>ilist_traits<T></tt>
1141 which superpose the sentinel with the <tt>ilist</tt> instance in memory. Pointer
1142 arithmetic is used to obtain the sentinel, which is relative to the
1143 <tt>ilist</tt>'s <tt>this</tt> pointer. The <tt>ilist</tt> is augmented by an
1144 extra pointer, which serves as the back-link of the sentinel. This is the only
1145 field in the ghostly sentinel which can be legally accessed.</p>
1148 <!-- _______________________________________________________________________ -->
1149 <div class="doc_subsubsection">
1150 <a name="dss_other">Other Sequential Container options</a>
1153 <div class="doc_text">
1154 <p>Other STL containers are available, such as std::string.</p>
1156 <p>There are also various STL adapter classes such as std::queue,
1157 std::priority_queue, std::stack, etc. These provide simplified access to an
1158 underlying container but don't affect the cost of the container itself.</p>
1163 <!-- ======================================================================= -->
1164 <div class="doc_subsection">
1165 <a name="ds_set">Set-Like Containers (std::set, SmallSet, SetVector, etc)</a>
1168 <div class="doc_text">
1170 <p>Set-like containers are useful when you need to canonicalize multiple values
1171 into a single representation. There are several different choices for how to do
1172 this, providing various trade-offs.</p>
1177 <!-- _______________________________________________________________________ -->
1178 <div class="doc_subsubsection">
1179 <a name="dss_sortedvectorset">A sorted 'vector'</a>
1182 <div class="doc_text">
1184 <p>If you intend to insert a lot of elements, then do a lot of queries, a
1185 great approach is to use a vector (or other sequential container) with
1186 std::sort+std::unique to remove duplicates. This approach works really well if
1187 your usage pattern has these two distinct phases (insert then query), and can be
1188 coupled with a good choice of <a href="#ds_sequential">sequential container</a>.
1192 This combination provides the several nice properties: the result data is
1193 contiguous in memory (good for cache locality), has few allocations, is easy to
1194 address (iterators in the final vector are just indices or pointers), and can be
1195 efficiently queried with a standard binary or radix search.</p>
1199 <!-- _______________________________________________________________________ -->
1200 <div class="doc_subsubsection">
1201 <a name="dss_smallset">"llvm/ADT/SmallSet.h"</a>
1204 <div class="doc_text">
1206 <p>If you have a set-like data structure that is usually small and whose elements
1207 are reasonably small, a <tt>SmallSet<Type, N></tt> is a good choice. This set
1208 has space for N elements in place (thus, if the set is dynamically smaller than
1209 N, no malloc traffic is required) and accesses them with a simple linear search.
1210 When the set grows beyond 'N' elements, it allocates a more expensive representation that
1211 guarantees efficient access (for most types, it falls back to std::set, but for
1212 pointers it uses something far better, <a
1213 href="#dss_smallptrset">SmallPtrSet</a>).</p>
1215 <p>The magic of this class is that it handles small sets extremely efficiently,
1216 but gracefully handles extremely large sets without loss of efficiency. The
1217 drawback is that the interface is quite small: it supports insertion, queries
1218 and erasing, but does not support iteration.</p>
1222 <!-- _______________________________________________________________________ -->
1223 <div class="doc_subsubsection">
1224 <a name="dss_smallptrset">"llvm/ADT/SmallPtrSet.h"</a>
1227 <div class="doc_text">
1229 <p>SmallPtrSet has all the advantages of <tt>SmallSet</tt> (and a <tt>SmallSet</tt> of pointers is
1230 transparently implemented with a <tt>SmallPtrSet</tt>), but also supports iterators. If
1231 more than 'N' insertions are performed, a single quadratically
1232 probed hash table is allocated and grows as needed, providing extremely
1233 efficient access (constant time insertion/deleting/queries with low constant
1234 factors) and is very stingy with malloc traffic.</p>
1236 <p>Note that, unlike <tt>std::set</tt>, the iterators of <tt>SmallPtrSet</tt> are invalidated
1237 whenever an insertion occurs. Also, the values visited by the iterators are not
1238 visited in sorted order.</p>
1242 <!-- _______________________________________________________________________ -->
1243 <div class="doc_subsubsection">
1244 <a name="dss_denseset">"llvm/ADT/DenseSet.h"</a>
1247 <div class="doc_text">
1250 DenseSet is a simple quadratically probed hash table. It excels at supporting
1251 small values: it uses a single allocation to hold all of the pairs that
1252 are currently inserted in the set. DenseSet is a great way to unique small
1253 values that are not simple pointers (use <a
1254 href="#dss_smallptrset">SmallPtrSet</a> for pointers). Note that DenseSet has
1255 the same requirements for the value type that <a
1256 href="#dss_densemap">DenseMap</a> has.
1261 <!-- _______________________________________________________________________ -->
1262 <div class="doc_subsubsection">
1263 <a name="dss_FoldingSet">"llvm/ADT/FoldingSet.h"</a>
1266 <div class="doc_text">
1269 FoldingSet is an aggregate class that is really good at uniquing
1270 expensive-to-create or polymorphic objects. It is a combination of a chained
1271 hash table with intrusive links (uniqued objects are required to inherit from
1272 FoldingSetNode) that uses <a href="#dss_smallvector">SmallVector</a> as part of
1275 <p>Consider a case where you want to implement a "getOrCreateFoo" method for
1276 a complex object (for example, a node in the code generator). The client has a
1277 description of *what* it wants to generate (it knows the opcode and all the
1278 operands), but we don't want to 'new' a node, then try inserting it into a set
1279 only to find out it already exists, at which point we would have to delete it
1280 and return the node that already exists.
1283 <p>To support this style of client, FoldingSet perform a query with a
1284 FoldingSetNodeID (which wraps SmallVector) that can be used to describe the
1285 element that we want to query for. The query either returns the element
1286 matching the ID or it returns an opaque ID that indicates where insertion should
1287 take place. Construction of the ID usually does not require heap traffic.</p>
1289 <p>Because FoldingSet uses intrusive links, it can support polymorphic objects
1290 in the set (for example, you can have SDNode instances mixed with LoadSDNodes).
1291 Because the elements are individually allocated, pointers to the elements are
1292 stable: inserting or removing elements does not invalidate any pointers to other
1298 <!-- _______________________________________________________________________ -->
1299 <div class="doc_subsubsection">
1300 <a name="dss_set"><set></a>
1303 <div class="doc_text">
1305 <p><tt>std::set</tt> is a reasonable all-around set class, which is decent at
1306 many things but great at nothing. std::set allocates memory for each element
1307 inserted (thus it is very malloc intensive) and typically stores three pointers
1308 per element in the set (thus adding a large amount of per-element space
1309 overhead). It offers guaranteed log(n) performance, which is not particularly
1310 fast from a complexity standpoint (particularly if the elements of the set are
1311 expensive to compare, like strings), and has extremely high constant factors for
1312 lookup, insertion and removal.</p>
1314 <p>The advantages of std::set are that its iterators are stable (deleting or
1315 inserting an element from the set does not affect iterators or pointers to other
1316 elements) and that iteration over the set is guaranteed to be in sorted order.
1317 If the elements in the set are large, then the relative overhead of the pointers
1318 and malloc traffic is not a big deal, but if the elements of the set are small,
1319 std::set is almost never a good choice.</p>
1323 <!-- _______________________________________________________________________ -->
1324 <div class="doc_subsubsection">
1325 <a name="dss_setvector">"llvm/ADT/SetVector.h"</a>
1328 <div class="doc_text">
1329 <p>LLVM's SetVector<Type> is an adapter class that combines your choice of
1330 a set-like container along with a <a href="#ds_sequential">Sequential
1331 Container</a>. The important property
1332 that this provides is efficient insertion with uniquing (duplicate elements are
1333 ignored) with iteration support. It implements this by inserting elements into
1334 both a set-like container and the sequential container, using the set-like
1335 container for uniquing and the sequential container for iteration.
1338 <p>The difference between SetVector and other sets is that the order of
1339 iteration is guaranteed to match the order of insertion into the SetVector.
1340 This property is really important for things like sets of pointers. Because
1341 pointer values are non-deterministic (e.g. vary across runs of the program on
1342 different machines), iterating over the pointers in the set will
1343 not be in a well-defined order.</p>
1346 The drawback of SetVector is that it requires twice as much space as a normal
1347 set and has the sum of constant factors from the set-like container and the
1348 sequential container that it uses. Use it *only* if you need to iterate over
1349 the elements in a deterministic order. SetVector is also expensive to delete
1350 elements out of (linear time), unless you use it's "pop_back" method, which is
1354 <p>SetVector is an adapter class that defaults to using std::vector and std::set
1355 for the underlying containers, so it is quite expensive. However,
1356 <tt>"llvm/ADT/SetVector.h"</tt> also provides a SmallSetVector class, which
1357 defaults to using a SmallVector and SmallSet of a specified size. If you use
1358 this, and if your sets are dynamically smaller than N, you will save a lot of
1363 <!-- _______________________________________________________________________ -->
1364 <div class="doc_subsubsection">
1365 <a name="dss_uniquevector">"llvm/ADT/UniqueVector.h"</a>
1368 <div class="doc_text">
1371 UniqueVector is similar to <a href="#dss_setvector">SetVector</a>, but it
1372 retains a unique ID for each element inserted into the set. It internally
1373 contains a map and a vector, and it assigns a unique ID for each value inserted
1376 <p>UniqueVector is very expensive: its cost is the sum of the cost of
1377 maintaining both the map and vector, it has high complexity, high constant
1378 factors, and produces a lot of malloc traffic. It should be avoided.</p>
1383 <!-- _______________________________________________________________________ -->
1384 <div class="doc_subsubsection">
1385 <a name="dss_otherset">Other Set-Like Container Options</a>
1388 <div class="doc_text">
1391 The STL provides several other options, such as std::multiset and the various
1392 "hash_set" like containers (whether from C++ TR1 or from the SGI library). We
1393 never use hash_set and unordered_set because they are generally very expensive
1394 (each insertion requires a malloc) and very non-portable.
1397 <p>std::multiset is useful if you're not interested in elimination of
1398 duplicates, but has all the drawbacks of std::set. A sorted vector (where you
1399 don't delete duplicate entries) or some other approach is almost always
1404 <!-- ======================================================================= -->
1405 <div class="doc_subsection">
1406 <a name="ds_map">Map-Like Containers (std::map, DenseMap, etc)</a>
1409 <div class="doc_text">
1410 Map-like containers are useful when you want to associate data to a key. As
1411 usual, there are a lot of different ways to do this. :)
1414 <!-- _______________________________________________________________________ -->
1415 <div class="doc_subsubsection">
1416 <a name="dss_sortedvectormap">A sorted 'vector'</a>
1419 <div class="doc_text">
1422 If your usage pattern follows a strict insert-then-query approach, you can
1423 trivially use the same approach as <a href="#dss_sortedvectorset">sorted vectors
1424 for set-like containers</a>. The only difference is that your query function
1425 (which uses std::lower_bound to get efficient log(n) lookup) should only compare
1426 the key, not both the key and value. This yields the same advantages as sorted
1431 <!-- _______________________________________________________________________ -->
1432 <div class="doc_subsubsection">
1433 <a name="dss_stringmap">"llvm/ADT/StringMap.h"</a>
1436 <div class="doc_text">
1439 Strings are commonly used as keys in maps, and they are difficult to support
1440 efficiently: they are variable length, inefficient to hash and compare when
1441 long, expensive to copy, etc. StringMap is a specialized container designed to
1442 cope with these issues. It supports mapping an arbitrary range of bytes to an
1443 arbitrary other object.</p>
1445 <p>The StringMap implementation uses a quadratically-probed hash table, where
1446 the buckets store a pointer to the heap allocated entries (and some other
1447 stuff). The entries in the map must be heap allocated because the strings are
1448 variable length. The string data (key) and the element object (value) are
1449 stored in the same allocation with the string data immediately after the element
1450 object. This container guarantees the "<tt>(char*)(&Value+1)</tt>" points
1451 to the key string for a value.</p>
1453 <p>The StringMap is very fast for several reasons: quadratic probing is very
1454 cache efficient for lookups, the hash value of strings in buckets is not
1455 recomputed when looking up an element, StringMap rarely has to touch the
1456 memory for unrelated objects when looking up a value (even when hash collisions
1457 happen), hash table growth does not recompute the hash values for strings
1458 already in the table, and each pair in the map is store in a single allocation
1459 (the string data is stored in the same allocation as the Value of a pair).</p>
1461 <p>StringMap also provides query methods that take byte ranges, so it only ever
1462 copies a string if a value is inserted into the table.</p>
1465 <!-- _______________________________________________________________________ -->
1466 <div class="doc_subsubsection">
1467 <a name="dss_indexedmap">"llvm/ADT/IndexedMap.h"</a>
1470 <div class="doc_text">
1472 IndexedMap is a specialized container for mapping small dense integers (or
1473 values that can be mapped to small dense integers) to some other type. It is
1474 internally implemented as a vector with a mapping function that maps the keys to
1475 the dense integer range.
1479 This is useful for cases like virtual registers in the LLVM code generator: they
1480 have a dense mapping that is offset by a compile-time constant (the first
1481 virtual register ID).</p>
1485 <!-- _______________________________________________________________________ -->
1486 <div class="doc_subsubsection">
1487 <a name="dss_densemap">"llvm/ADT/DenseMap.h"</a>
1490 <div class="doc_text">
1493 DenseMap is a simple quadratically probed hash table. It excels at supporting
1494 small keys and values: it uses a single allocation to hold all of the pairs that
1495 are currently inserted in the map. DenseMap is a great way to map pointers to
1496 pointers, or map other small types to each other.
1500 There are several aspects of DenseMap that you should be aware of, however. The
1501 iterators in a densemap are invalidated whenever an insertion occurs, unlike
1502 map. Also, because DenseMap allocates space for a large number of key/value
1503 pairs (it starts with 64 by default), it will waste a lot of space if your keys
1504 or values are large. Finally, you must implement a partial specialization of
1505 DenseMapInfo for the key that you want, if it isn't already supported. This
1506 is required to tell DenseMap about two special marker values (which can never be
1507 inserted into the map) that it needs internally.</p>
1511 <!-- _______________________________________________________________________ -->
1512 <div class="doc_subsubsection">
1513 <a name="dss_valuemap">"llvm/ADT/ValueMap.h"</a>
1516 <div class="doc_text">
1519 ValueMap is a wrapper around a <a href="#dss_densemap">DenseMap</a> mapping
1520 Value*s (or subclasses) to another type. When a Value is deleted or RAUW'ed,
1521 ValueMap will update itself so the new version of the key is mapped to the same
1522 value, just as if the key were a WeakVH. You can configure exactly how this
1523 happens, and what else happens on these two events, by passing
1524 a <code>Config</code> parameter to the ValueMap template.</p>
1528 <!-- _______________________________________________________________________ -->
1529 <div class="doc_subsubsection">
1530 <a name="dss_intervalmap">"llvm/ADT/IntervalMap.h"</a>
1533 <div class="doc_text">
1535 <p> IntervalMap is a compact map for small keys and values. It maps key
1536 intervals instead of single keys, and it will automatically coalesce adjacent
1537 intervals. When then map only contains a few intervals, they are stored in the
1538 map object itself to avoid allocations.</p>
1540 <p> The IntervalMap iterators are quite big, so they should not be passed around
1541 as STL iterators. The heavyweight iterators allow a smaller data structure.</p>
1545 <!-- _______________________________________________________________________ -->
1546 <div class="doc_subsubsection">
1547 <a name="dss_map"><map></a>
1550 <div class="doc_text">
1553 std::map has similar characteristics to <a href="#dss_set">std::set</a>: it uses
1554 a single allocation per pair inserted into the map, it offers log(n) lookup with
1555 an extremely large constant factor, imposes a space penalty of 3 pointers per
1556 pair in the map, etc.</p>
1558 <p>std::map is most useful when your keys or values are very large, if you need
1559 to iterate over the collection in sorted order, or if you need stable iterators
1560 into the map (i.e. they don't get invalidated if an insertion or deletion of
1561 another element takes place).</p>
1565 <!-- _______________________________________________________________________ -->
1566 <div class="doc_subsubsection">
1567 <a name="dss_inteqclasses">"llvm/ADT/IntEqClasses.h"</a>
1570 <div class="doc_text">
1572 <p>IntEqClasses provides a compact representation of equivalence classes of
1573 small integers. Initially, each integer in the range 0..n-1 has its own
1574 equivalence class. Classes can be joined by passing two class representatives to
1575 the join(a, b) method. Two integers are in the same class when findLeader()
1576 returns the same representative.</p>
1578 <p>Once all equivalence classes are formed, the map can be compressed so each
1579 integer 0..n-1 maps to an equivalence class number in the range 0..m-1, where m
1580 is the total number of equivalence classes. The map must be uncompressed before
1581 it can be edited again.</p>
1585 <!-- _______________________________________________________________________ -->
1586 <div class="doc_subsubsection">
1587 <a name="dss_othermap">Other Map-Like Container Options</a>
1590 <div class="doc_text">
1593 The STL provides several other options, such as std::multimap and the various
1594 "hash_map" like containers (whether from C++ TR1 or from the SGI library). We
1595 never use hash_set and unordered_set because they are generally very expensive
1596 (each insertion requires a malloc) and very non-portable.</p>
1598 <p>std::multimap is useful if you want to map a key to multiple values, but has
1599 all the drawbacks of std::map. A sorted vector or some other approach is almost
1604 <!-- ======================================================================= -->
1605 <div class="doc_subsection">
1606 <a name="ds_string">String-like containers</a>
1609 <div class="doc_text">
1612 TODO: const char* vs stringref vs smallstring vs std::string. Describe twine,
1613 xref to #string_apis.
1618 <!-- ======================================================================= -->
1619 <div class="doc_subsection">
1620 <a name="ds_bit">Bit storage containers (BitVector, SparseBitVector)</a>
1623 <div class="doc_text">
1624 <p>Unlike the other containers, there are only two bit storage containers, and
1625 choosing when to use each is relatively straightforward.</p>
1627 <p>One additional option is
1628 <tt>std::vector<bool></tt>: we discourage its use for two reasons 1) the
1629 implementation in many common compilers (e.g. commonly available versions of
1630 GCC) is extremely inefficient and 2) the C++ standards committee is likely to
1631 deprecate this container and/or change it significantly somehow. In any case,
1632 please don't use it.</p>
1635 <!-- _______________________________________________________________________ -->
1636 <div class="doc_subsubsection">
1637 <a name="dss_bitvector">BitVector</a>
1640 <div class="doc_text">
1641 <p> The BitVector container provides a dynamic size set of bits for manipulation.
1642 It supports individual bit setting/testing, as well as set operations. The set
1643 operations take time O(size of bitvector), but operations are performed one word
1644 at a time, instead of one bit at a time. This makes the BitVector very fast for
1645 set operations compared to other containers. Use the BitVector when you expect
1646 the number of set bits to be high (IE a dense set).
1650 <!-- _______________________________________________________________________ -->
1651 <div class="doc_subsubsection">
1652 <a name="dss_smallbitvector">SmallBitVector</a>
1655 <div class="doc_text">
1656 <p> The SmallBitVector container provides the same interface as BitVector, but
1657 it is optimized for the case where only a small number of bits, less than
1658 25 or so, are needed. It also transparently supports larger bit counts, but
1659 slightly less efficiently than a plain BitVector, so SmallBitVector should
1660 only be used when larger counts are rare.
1664 At this time, SmallBitVector does not support set operations (and, or, xor),
1665 and its operator[] does not provide an assignable lvalue.
1669 <!-- _______________________________________________________________________ -->
1670 <div class="doc_subsubsection">
1671 <a name="dss_sparsebitvector">SparseBitVector</a>
1674 <div class="doc_text">
1675 <p> The SparseBitVector container is much like BitVector, with one major
1676 difference: Only the bits that are set, are stored. This makes the
1677 SparseBitVector much more space efficient than BitVector when the set is sparse,
1678 as well as making set operations O(number of set bits) instead of O(size of
1679 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
1680 (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).
1684 <!-- *********************************************************************** -->
1685 <div class="doc_section">
1686 <a name="common">Helpful Hints for Common Operations</a>
1688 <!-- *********************************************************************** -->
1690 <div class="doc_text">
1692 <p>This section describes how to perform some very simple transformations of
1693 LLVM code. This is meant to give examples of common idioms used, showing the
1694 practical side of LLVM transformations. <p> Because this is a "how-to" section,
1695 you should also read about the main classes that you will be working with. The
1696 <a href="#coreclasses">Core LLVM Class Hierarchy Reference</a> contains details
1697 and descriptions of the main classes that you should know about.</p>
1701 <!-- NOTE: this section should be heavy on example code -->
1702 <!-- ======================================================================= -->
1703 <div class="doc_subsection">
1704 <a name="inspection">Basic Inspection and Traversal Routines</a>
1707 <div class="doc_text">
1709 <p>The LLVM compiler infrastructure have many different data structures that may
1710 be traversed. Following the example of the C++ standard template library, the
1711 techniques used to traverse these various data structures are all basically the
1712 same. For a enumerable sequence of values, the <tt>XXXbegin()</tt> function (or
1713 method) returns an iterator to the start of the sequence, the <tt>XXXend()</tt>
1714 function returns an iterator pointing to one past the last valid element of the
1715 sequence, and there is some <tt>XXXiterator</tt> data type that is common
1716 between the two operations.</p>
1718 <p>Because the pattern for iteration is common across many different aspects of
1719 the program representation, the standard template library algorithms may be used
1720 on them, and it is easier to remember how to iterate. First we show a few common
1721 examples of the data structures that need to be traversed. Other data
1722 structures are traversed in very similar ways.</p>
1726 <!-- _______________________________________________________________________ -->
1727 <div class="doc_subsubsection">
1728 <a name="iterate_function">Iterating over the </a><a
1729 href="#BasicBlock"><tt>BasicBlock</tt></a>s in a <a
1730 href="#Function"><tt>Function</tt></a>
1733 <div class="doc_text">
1735 <p>It's quite common to have a <tt>Function</tt> instance that you'd like to
1736 transform in some way; in particular, you'd like to manipulate its
1737 <tt>BasicBlock</tt>s. To facilitate this, you'll need to iterate over all of
1738 the <tt>BasicBlock</tt>s that constitute the <tt>Function</tt>. The following is
1739 an example that prints the name of a <tt>BasicBlock</tt> and the number of
1740 <tt>Instruction</tt>s it contains:</p>
1742 <div class="doc_code">
1744 // <i>func is a pointer to a Function instance</i>
1745 for (Function::iterator i = func->begin(), e = func->end(); i != e; ++i)
1746 // <i>Print out the name of the basic block if it has one, and then the</i>
1747 // <i>number of instructions that it contains</i>
1748 errs() << "Basic block (name=" << i->getName() << ") has "
1749 << i->size() << " instructions.\n";
1753 <p>Note that i can be used as if it were a pointer for the purposes of
1754 invoking member functions of the <tt>Instruction</tt> class. This is
1755 because the indirection operator is overloaded for the iterator
1756 classes. In the above code, the expression <tt>i->size()</tt> is
1757 exactly equivalent to <tt>(*i).size()</tt> just like you'd expect.</p>
1761 <!-- _______________________________________________________________________ -->
1762 <div class="doc_subsubsection">
1763 <a name="iterate_basicblock">Iterating over the </a><a
1764 href="#Instruction"><tt>Instruction</tt></a>s in a <a
1765 href="#BasicBlock"><tt>BasicBlock</tt></a>
1768 <div class="doc_text">
1770 <p>Just like when dealing with <tt>BasicBlock</tt>s in <tt>Function</tt>s, it's
1771 easy to iterate over the individual instructions that make up
1772 <tt>BasicBlock</tt>s. Here's a code snippet that prints out each instruction in
1773 a <tt>BasicBlock</tt>:</p>
1775 <div class="doc_code">
1777 // <i>blk is a pointer to a BasicBlock instance</i>
1778 for (BasicBlock::iterator i = blk->begin(), e = blk->end(); i != e; ++i)
1779 // <i>The next statement works since operator<<(ostream&,...)</i>
1780 // <i>is overloaded for Instruction&</i>
1781 errs() << *i << "\n";
1785 <p>However, this isn't really the best way to print out the contents of a
1786 <tt>BasicBlock</tt>! Since the ostream operators are overloaded for virtually
1787 anything you'll care about, you could have just invoked the print routine on the
1788 basic block itself: <tt>errs() << *blk << "\n";</tt>.</p>
1792 <!-- _______________________________________________________________________ -->
1793 <div class="doc_subsubsection">
1794 <a name="iterate_institer">Iterating over the </a><a
1795 href="#Instruction"><tt>Instruction</tt></a>s in a <a
1796 href="#Function"><tt>Function</tt></a>
1799 <div class="doc_text">
1801 <p>If you're finding that you commonly iterate over a <tt>Function</tt>'s
1802 <tt>BasicBlock</tt>s and then that <tt>BasicBlock</tt>'s <tt>Instruction</tt>s,
1803 <tt>InstIterator</tt> should be used instead. You'll need to include <a
1804 href="/doxygen/InstIterator_8h-source.html"><tt>llvm/Support/InstIterator.h</tt></a>,
1805 and then instantiate <tt>InstIterator</tt>s explicitly in your code. Here's a
1806 small example that shows how to dump all instructions in a function to the standard error stream:<p>
1808 <div class="doc_code">
1810 #include "<a href="/doxygen/InstIterator_8h-source.html">llvm/Support/InstIterator.h</a>"
1812 // <i>F is a pointer to a Function instance</i>
1813 for (inst_iterator I = inst_begin(F), E = inst_end(F); I != E; ++I)
1814 errs() << *I << "\n";
1818 <p>Easy, isn't it? You can also use <tt>InstIterator</tt>s to fill a
1819 work list with its initial contents. For example, if you wanted to
1820 initialize a work list to contain all instructions in a <tt>Function</tt>
1821 F, all you would need to do is something like:</p>
1823 <div class="doc_code">
1825 std::set<Instruction*> worklist;
1826 // or better yet, SmallPtrSet<Instruction*, 64> worklist;
1828 for (inst_iterator I = inst_begin(F), E = inst_end(F); I != E; ++I)
1829 worklist.insert(&*I);
1833 <p>The STL set <tt>worklist</tt> would now contain all instructions in the
1834 <tt>Function</tt> pointed to by F.</p>
1838 <!-- _______________________________________________________________________ -->
1839 <div class="doc_subsubsection">
1840 <a name="iterate_convert">Turning an iterator into a class pointer (and
1844 <div class="doc_text">
1846 <p>Sometimes, it'll be useful to grab a reference (or pointer) to a class
1847 instance when all you've got at hand is an iterator. Well, extracting
1848 a reference or a pointer from an iterator is very straight-forward.
1849 Assuming that <tt>i</tt> is a <tt>BasicBlock::iterator</tt> and <tt>j</tt>
1850 is a <tt>BasicBlock::const_iterator</tt>:</p>
1852 <div class="doc_code">
1854 Instruction& inst = *i; // <i>Grab reference to instruction reference</i>
1855 Instruction* pinst = &*i; // <i>Grab pointer to instruction reference</i>
1856 const Instruction& inst = *j;
1860 <p>However, the iterators you'll be working with in the LLVM framework are
1861 special: they will automatically convert to a ptr-to-instance type whenever they
1862 need to. Instead of dereferencing the iterator and then taking the address of
1863 the result, you can simply assign the iterator to the proper pointer type and
1864 you get the dereference and address-of operation as a result of the assignment
1865 (behind the scenes, this is a result of overloading casting mechanisms). Thus
1866 the last line of the last example,</p>
1868 <div class="doc_code">
1870 Instruction *pinst = &*i;
1874 <p>is semantically equivalent to</p>
1876 <div class="doc_code">
1878 Instruction *pinst = i;
1882 <p>It's also possible to turn a class pointer into the corresponding iterator,
1883 and this is a constant time operation (very efficient). The following code
1884 snippet illustrates use of the conversion constructors provided by LLVM
1885 iterators. By using these, you can explicitly grab the iterator of something
1886 without actually obtaining it via iteration over some structure:</p>
1888 <div class="doc_code">
1890 void printNextInstruction(Instruction* inst) {
1891 BasicBlock::iterator it(inst);
1892 ++it; // <i>After this line, it refers to the instruction after *inst</i>
1893 if (it != inst->getParent()->end()) errs() << *it << "\n";
1898 <p>Unfortunately, these implicit conversions come at a cost; they prevent
1899 these iterators from conforming to standard iterator conventions, and thus
1900 from being usable with standard algorithms and containers. For example, they
1901 prevent the following code, where <tt>B</tt> is a <tt>BasicBlock</tt>,
1904 <div class="doc_code">
1906 llvm::SmallVector<llvm::Instruction *, 16>(B->begin(), B->end());
1910 <p>Because of this, these implicit conversions may be removed some day,
1911 and <tt>operator*</tt> changed to return a pointer instead of a reference.</p>
1915 <!--_______________________________________________________________________-->
1916 <div class="doc_subsubsection">
1917 <a name="iterate_complex">Finding call sites: a slightly more complex
1921 <div class="doc_text">
1923 <p>Say that you're writing a FunctionPass and would like to count all the
1924 locations in the entire module (that is, across every <tt>Function</tt>) where a
1925 certain function (i.e., some <tt>Function</tt>*) is already in scope. As you'll
1926 learn later, you may want to use an <tt>InstVisitor</tt> to accomplish this in a
1927 much more straight-forward manner, but this example will allow us to explore how
1928 you'd do it if you didn't have <tt>InstVisitor</tt> around. In pseudo-code, this
1929 is what we want to do:</p>
1931 <div class="doc_code">
1933 initialize callCounter to zero
1934 for each Function f in the Module
1935 for each BasicBlock b in f
1936 for each Instruction i in b
1937 if (i is a CallInst and calls the given function)
1938 increment callCounter
1942 <p>And the actual code is (remember, because we're writing a
1943 <tt>FunctionPass</tt>, our <tt>FunctionPass</tt>-derived class simply has to
1944 override the <tt>runOnFunction</tt> method):</p>
1946 <div class="doc_code">
1948 Function* targetFunc = ...;
1950 class OurFunctionPass : public FunctionPass {
1952 OurFunctionPass(): callCounter(0) { }
1954 virtual runOnFunction(Function& F) {
1955 for (Function::iterator b = F.begin(), be = F.end(); b != be; ++b) {
1956 for (BasicBlock::iterator i = b->begin(), ie = b->end(); i != ie; ++i) {
1957 if (<a href="#CallInst">CallInst</a>* callInst = <a href="#isa">dyn_cast</a><<a
1958 href="#CallInst">CallInst</a>>(&*i)) {
1959 // <i>We know we've encountered a call instruction, so we</i>
1960 // <i>need to determine if it's a call to the</i>
1961 // <i>function pointed to by m_func or not.</i>
1962 if (callInst->getCalledFunction() == targetFunc)
1970 unsigned callCounter;
1977 <!--_______________________________________________________________________-->
1978 <div class="doc_subsubsection">
1979 <a name="calls_and_invokes">Treating calls and invokes the same way</a>
1982 <div class="doc_text">
1984 <p>You may have noticed that the previous example was a bit oversimplified in
1985 that it did not deal with call sites generated by 'invoke' instructions. In
1986 this, and in other situations, you may find that you want to treat
1987 <tt>CallInst</tt>s and <tt>InvokeInst</tt>s the same way, even though their
1988 most-specific common base class is <tt>Instruction</tt>, which includes lots of
1989 less closely-related things. For these cases, LLVM provides a handy wrapper
1991 href="http://llvm.org/doxygen/classllvm_1_1CallSite.html"><tt>CallSite</tt></a>.
1992 It is essentially a wrapper around an <tt>Instruction</tt> pointer, with some
1993 methods that provide functionality common to <tt>CallInst</tt>s and
1994 <tt>InvokeInst</tt>s.</p>
1996 <p>This class has "value semantics": it should be passed by value, not by
1997 reference and it should not be dynamically allocated or deallocated using
1998 <tt>operator new</tt> or <tt>operator delete</tt>. It is efficiently copyable,
1999 assignable and constructable, with costs equivalents to that of a bare pointer.
2000 If you look at its definition, it has only a single pointer member.</p>
2004 <!--_______________________________________________________________________-->
2005 <div class="doc_subsubsection">
2006 <a name="iterate_chains">Iterating over def-use & use-def chains</a>
2009 <div class="doc_text">
2011 <p>Frequently, we might have an instance of the <a
2012 href="/doxygen/classllvm_1_1Value.html">Value Class</a> and we want to
2013 determine which <tt>User</tt>s use the <tt>Value</tt>. The list of all
2014 <tt>User</tt>s of a particular <tt>Value</tt> is called a <i>def-use</i> chain.
2015 For example, let's say we have a <tt>Function*</tt> named <tt>F</tt> to a
2016 particular function <tt>foo</tt>. Finding all of the instructions that
2017 <i>use</i> <tt>foo</tt> is as simple as iterating over the <i>def-use</i> chain
2020 <div class="doc_code">
2024 for (Value::use_iterator i = F->use_begin(), e = F->use_end(); i != e; ++i)
2025 if (Instruction *Inst = dyn_cast<Instruction>(*i)) {
2026 errs() << "F is used in instruction:\n";
2027 errs() << *Inst << "\n";
2032 <p>Note that dereferencing a <tt>Value::use_iterator</tt> is not a very cheap
2033 operation. Instead of performing <tt>*i</tt> above several times, consider
2034 doing it only once in the loop body and reusing its result.</p>
2036 <p>Alternatively, it's common to have an instance of the <a
2037 href="/doxygen/classllvm_1_1User.html">User Class</a> and need to know what
2038 <tt>Value</tt>s are used by it. The list of all <tt>Value</tt>s used by a
2039 <tt>User</tt> is known as a <i>use-def</i> chain. Instances of class
2040 <tt>Instruction</tt> are common <tt>User</tt>s, so we might want to iterate over
2041 all of the values that a particular instruction uses (that is, the operands of
2042 the particular <tt>Instruction</tt>):</p>
2044 <div class="doc_code">
2046 Instruction *pi = ...;
2048 for (User::op_iterator i = pi->op_begin(), e = pi->op_end(); i != e; ++i) {
2055 <p>Declaring objects as <tt>const</tt> is an important tool of enforcing
2056 mutation free algorithms (such as analyses, etc.). For this purpose above
2057 iterators come in constant flavors as <tt>Value::const_use_iterator</tt>
2058 and <tt>Value::const_op_iterator</tt>. They automatically arise when
2059 calling <tt>use/op_begin()</tt> on <tt>const Value*</tt>s or
2060 <tt>const User*</tt>s respectively. Upon dereferencing, they return
2061 <tt>const Use*</tt>s. Otherwise the above patterns remain unchanged.</p>
2065 <!--_______________________________________________________________________-->
2066 <div class="doc_subsubsection">
2067 <a name="iterate_preds">Iterating over predecessors &
2068 successors of blocks</a>
2071 <div class="doc_text">
2073 <p>Iterating over the predecessors and successors of a block is quite easy
2074 with the routines defined in <tt>"llvm/Support/CFG.h"</tt>. Just use code like
2075 this to iterate over all predecessors of BB:</p>
2077 <div class="doc_code">
2079 #include "llvm/Support/CFG.h"
2080 BasicBlock *BB = ...;
2082 for (pred_iterator PI = pred_begin(BB), E = pred_end(BB); PI != E; ++PI) {
2083 BasicBlock *Pred = *PI;
2089 <p>Similarly, to iterate over successors use
2090 succ_iterator/succ_begin/succ_end.</p>
2095 <!-- ======================================================================= -->
2096 <div class="doc_subsection">
2097 <a name="simplechanges">Making simple changes</a>
2100 <div class="doc_text">
2102 <p>There are some primitive transformation operations present in the LLVM
2103 infrastructure that are worth knowing about. When performing
2104 transformations, it's fairly common to manipulate the contents of basic
2105 blocks. This section describes some of the common methods for doing so
2106 and gives example code.</p>
2110 <!--_______________________________________________________________________-->
2111 <div class="doc_subsubsection">
2112 <a name="schanges_creating">Creating and inserting new
2113 <tt>Instruction</tt>s</a>
2116 <div class="doc_text">
2118 <p><i>Instantiating Instructions</i></p>
2120 <p>Creation of <tt>Instruction</tt>s is straight-forward: simply call the
2121 constructor for the kind of instruction to instantiate and provide the necessary
2122 parameters. For example, an <tt>AllocaInst</tt> only <i>requires</i> a
2123 (const-ptr-to) <tt>Type</tt>. Thus:</p>
2125 <div class="doc_code">
2127 AllocaInst* ai = new AllocaInst(Type::Int32Ty);
2131 <p>will create an <tt>AllocaInst</tt> instance that represents the allocation of
2132 one integer in the current stack frame, at run time. Each <tt>Instruction</tt>
2133 subclass is likely to have varying default parameters which change the semantics
2134 of the instruction, so refer to the <a
2135 href="/doxygen/classllvm_1_1Instruction.html">doxygen documentation for the subclass of
2136 Instruction</a> that you're interested in instantiating.</p>
2138 <p><i>Naming values</i></p>
2140 <p>It is very useful to name the values of instructions when you're able to, as
2141 this facilitates the debugging of your transformations. If you end up looking
2142 at generated LLVM machine code, you definitely want to have logical names
2143 associated with the results of instructions! By supplying a value for the
2144 <tt>Name</tt> (default) parameter of the <tt>Instruction</tt> constructor, you
2145 associate a logical name with the result of the instruction's execution at
2146 run time. For example, say that I'm writing a transformation that dynamically
2147 allocates space for an integer on the stack, and that integer is going to be
2148 used as some kind of index by some other code. To accomplish this, I place an
2149 <tt>AllocaInst</tt> at the first point in the first <tt>BasicBlock</tt> of some
2150 <tt>Function</tt>, and I'm intending to use it within the same
2151 <tt>Function</tt>. I might do:</p>
2153 <div class="doc_code">
2155 AllocaInst* pa = new AllocaInst(Type::Int32Ty, 0, "indexLoc");
2159 <p>where <tt>indexLoc</tt> is now the logical name of the instruction's
2160 execution value, which is a pointer to an integer on the run time stack.</p>
2162 <p><i>Inserting instructions</i></p>
2164 <p>There are essentially two ways to insert an <tt>Instruction</tt>
2165 into an existing sequence of instructions that form a <tt>BasicBlock</tt>:</p>
2168 <li>Insertion into an explicit instruction list
2170 <p>Given a <tt>BasicBlock* pb</tt>, an <tt>Instruction* pi</tt> within that
2171 <tt>BasicBlock</tt>, and a newly-created instruction we wish to insert
2172 before <tt>*pi</tt>, we do the following: </p>
2174 <div class="doc_code">
2176 BasicBlock *pb = ...;
2177 Instruction *pi = ...;
2178 Instruction *newInst = new Instruction(...);
2180 pb->getInstList().insert(pi, newInst); // <i>Inserts newInst before pi in pb</i>
2184 <p>Appending to the end of a <tt>BasicBlock</tt> is so common that
2185 the <tt>Instruction</tt> class and <tt>Instruction</tt>-derived
2186 classes provide constructors which take a pointer to a
2187 <tt>BasicBlock</tt> to be appended to. For example code that
2190 <div class="doc_code">
2192 BasicBlock *pb = ...;
2193 Instruction *newInst = new Instruction(...);
2195 pb->getInstList().push_back(newInst); // <i>Appends newInst to pb</i>
2201 <div class="doc_code">
2203 BasicBlock *pb = ...;
2204 Instruction *newInst = new Instruction(..., pb);
2208 <p>which is much cleaner, especially if you are creating
2209 long instruction streams.</p></li>
2211 <li>Insertion into an implicit instruction list
2213 <p><tt>Instruction</tt> instances that are already in <tt>BasicBlock</tt>s
2214 are implicitly associated with an existing instruction list: the instruction
2215 list of the enclosing basic block. Thus, we could have accomplished the same
2216 thing as the above code without being given a <tt>BasicBlock</tt> by doing:
2219 <div class="doc_code">
2221 Instruction *pi = ...;
2222 Instruction *newInst = new Instruction(...);
2224 pi->getParent()->getInstList().insert(pi, newInst);
2228 <p>In fact, this sequence of steps occurs so frequently that the
2229 <tt>Instruction</tt> class and <tt>Instruction</tt>-derived classes provide
2230 constructors which take (as a default parameter) a pointer to an
2231 <tt>Instruction</tt> which the newly-created <tt>Instruction</tt> should
2232 precede. That is, <tt>Instruction</tt> constructors are capable of
2233 inserting the newly-created instance into the <tt>BasicBlock</tt> of a
2234 provided instruction, immediately before that instruction. Using an
2235 <tt>Instruction</tt> constructor with a <tt>insertBefore</tt> (default)
2236 parameter, the above code becomes:</p>
2238 <div class="doc_code">
2240 Instruction* pi = ...;
2241 Instruction* newInst = new Instruction(..., pi);
2245 <p>which is much cleaner, especially if you're creating a lot of
2246 instructions and adding them to <tt>BasicBlock</tt>s.</p></li>
2251 <!--_______________________________________________________________________-->
2252 <div class="doc_subsubsection">
2253 <a name="schanges_deleting">Deleting <tt>Instruction</tt>s</a>
2256 <div class="doc_text">
2258 <p>Deleting an instruction from an existing sequence of instructions that form a
2259 <a href="#BasicBlock"><tt>BasicBlock</tt></a> is very straight-forward: just
2260 call the instruction's eraseFromParent() method. For example:</p>
2262 <div class="doc_code">
2264 <a href="#Instruction">Instruction</a> *I = .. ;
2265 I->eraseFromParent();
2269 <p>This unlinks the instruction from its containing basic block and deletes
2270 it. If you'd just like to unlink the instruction from its containing basic
2271 block but not delete it, you can use the <tt>removeFromParent()</tt> method.</p>
2275 <!--_______________________________________________________________________-->
2276 <div class="doc_subsubsection">
2277 <a name="schanges_replacing">Replacing an <tt>Instruction</tt> with another
2281 <div class="doc_text">
2283 <p><i>Replacing individual instructions</i></p>
2285 <p>Including "<a href="/doxygen/BasicBlockUtils_8h-source.html">llvm/Transforms/Utils/BasicBlockUtils.h</a>"
2286 permits use of two very useful replace functions: <tt>ReplaceInstWithValue</tt>
2287 and <tt>ReplaceInstWithInst</tt>.</p>
2289 <h4><a name="schanges_deleting">Deleting <tt>Instruction</tt>s</a></h4>
2292 <li><tt>ReplaceInstWithValue</tt>
2294 <p>This function replaces all uses of a given instruction with a value,
2295 and then removes the original instruction. The following example
2296 illustrates the replacement of the result of a particular
2297 <tt>AllocaInst</tt> that allocates memory for a single integer with a null
2298 pointer to an integer.</p>
2300 <div class="doc_code">
2302 AllocaInst* instToReplace = ...;
2303 BasicBlock::iterator ii(instToReplace);
2305 ReplaceInstWithValue(instToReplace->getParent()->getInstList(), ii,
2306 Constant::getNullValue(PointerType::getUnqual(Type::Int32Ty)));
2309 <li><tt>ReplaceInstWithInst</tt>
2311 <p>This function replaces a particular instruction with another
2312 instruction, inserting the new instruction into the basic block at the
2313 location where the old instruction was, and replacing any uses of the old
2314 instruction with the new instruction. The following example illustrates
2315 the replacement of one <tt>AllocaInst</tt> with another.</p>
2317 <div class="doc_code">
2319 AllocaInst* instToReplace = ...;
2320 BasicBlock::iterator ii(instToReplace);
2322 ReplaceInstWithInst(instToReplace->getParent()->getInstList(), ii,
2323 new AllocaInst(Type::Int32Ty, 0, "ptrToReplacedInt"));
2327 <p><i>Replacing multiple uses of <tt>User</tt>s and <tt>Value</tt>s</i></p>
2329 <p>You can use <tt>Value::replaceAllUsesWith</tt> and
2330 <tt>User::replaceUsesOfWith</tt> to change more than one use at a time. See the
2331 doxygen documentation for the <a href="/doxygen/classllvm_1_1Value.html">Value Class</a>
2332 and <a href="/doxygen/classllvm_1_1User.html">User Class</a>, respectively, for more
2335 <!-- Value::replaceAllUsesWith User::replaceUsesOfWith Point out:
2336 include/llvm/Transforms/Utils/ especially BasicBlockUtils.h with:
2337 ReplaceInstWithValue, ReplaceInstWithInst -->
2341 <!--_______________________________________________________________________-->
2342 <div class="doc_subsubsection">
2343 <a name="schanges_deletingGV">Deleting <tt>GlobalVariable</tt>s</a>
2346 <div class="doc_text">
2348 <p>Deleting a global variable from a module is just as easy as deleting an
2349 Instruction. First, you must have a pointer to the global variable that you wish
2350 to delete. You use this pointer to erase it from its parent, the module.
2353 <div class="doc_code">
2355 <a href="#GlobalVariable">GlobalVariable</a> *GV = .. ;
2357 GV->eraseFromParent();
2363 <!-- ======================================================================= -->
2364 <div class="doc_subsection">
2365 <a name="create_types">How to Create Types</a>
2368 <div class="doc_text">
2370 <p>In generating IR, you may need some complex types. If you know these types
2371 statically, you can use <tt>TypeBuilder<...>::get()</tt>, defined
2372 in <tt>llvm/Support/TypeBuilder.h</tt>, to retrieve them. <tt>TypeBuilder</tt>
2373 has two forms depending on whether you're building types for cross-compilation
2374 or native library use. <tt>TypeBuilder<T, true></tt> requires
2375 that <tt>T</tt> be independent of the host environment, meaning that it's built
2377 the <a href="/doxygen/namespacellvm_1_1types.html"><tt>llvm::types</tt></a>
2378 namespace and pointers, functions, arrays, etc. built of
2379 those. <tt>TypeBuilder<T, false></tt> additionally allows native C types
2380 whose size may depend on the host compiler. For example,</p>
2382 <div class="doc_code">
2384 FunctionType *ft = TypeBuilder<types::i<8>(types::i<32>*), true>::get();
2388 <p>is easier to read and write than the equivalent</p>
2390 <div class="doc_code">
2392 std::vector<const Type*> params;
2393 params.push_back(PointerType::getUnqual(Type::Int32Ty));
2394 FunctionType *ft = FunctionType::get(Type::Int8Ty, params, false);
2398 <p>See the <a href="/doxygen/TypeBuilder_8h-source.html#l00001">class
2399 comment</a> for more details.</p>
2403 <!-- *********************************************************************** -->
2404 <div class="doc_section">
2405 <a name="threading">Threads and LLVM</a>
2407 <!-- *********************************************************************** -->
2409 <div class="doc_text">
2411 This section describes the interaction of the LLVM APIs with multithreading,
2412 both on the part of client applications, and in the JIT, in the hosted
2417 Note that LLVM's support for multithreading is still relatively young. Up
2418 through version 2.5, the execution of threaded hosted applications was
2419 supported, but not threaded client access to the APIs. While this use case is
2420 now supported, clients <em>must</em> adhere to the guidelines specified below to
2421 ensure proper operation in multithreaded mode.
2425 Note that, on Unix-like platforms, LLVM requires the presence of GCC's atomic
2426 intrinsics in order to support threaded operation. If you need a
2427 multhreading-capable LLVM on a platform without a suitably modern system
2428 compiler, consider compiling LLVM and LLVM-GCC in single-threaded mode, and
2429 using the resultant compiler to build a copy of LLVM with multithreading
2434 <!-- ======================================================================= -->
2435 <div class="doc_subsection">
2436 <a name="startmultithreaded">Entering and Exiting Multithreaded Mode</a>
2439 <div class="doc_text">
2442 In order to properly protect its internal data structures while avoiding
2443 excessive locking overhead in the single-threaded case, the LLVM must intialize
2444 certain data structures necessary to provide guards around its internals. To do
2445 so, the client program must invoke <tt>llvm_start_multithreaded()</tt> before
2446 making any concurrent LLVM API calls. To subsequently tear down these
2447 structures, use the <tt>llvm_stop_multithreaded()</tt> call. You can also use
2448 the <tt>llvm_is_multithreaded()</tt> call to check the status of multithreaded
2453 Note that both of these calls must be made <em>in isolation</em>. That is to
2454 say that no other LLVM API calls may be executing at any time during the
2455 execution of <tt>llvm_start_multithreaded()</tt> or <tt>llvm_stop_multithreaded
2456 </tt>. It's is the client's responsibility to enforce this isolation.
2460 The return value of <tt>llvm_start_multithreaded()</tt> indicates the success or
2461 failure of the initialization. Failure typically indicates that your copy of
2462 LLVM was built without multithreading support, typically because GCC atomic
2463 intrinsics were not found in your system compiler. In this case, the LLVM API
2464 will not be safe for concurrent calls. However, it <em>will</em> be safe for
2465 hosting threaded applications in the JIT, though <a href="#jitthreading">care
2466 must be taken</a> to ensure that side exits and the like do not accidentally
2467 result in concurrent LLVM API calls.
2471 <!-- ======================================================================= -->
2472 <div class="doc_subsection">
2473 <a name="shutdown">Ending Execution with <tt>llvm_shutdown()</tt></a>
2476 <div class="doc_text">
2478 When you are done using the LLVM APIs, you should call <tt>llvm_shutdown()</tt>
2479 to deallocate memory used for internal structures. This will also invoke
2480 <tt>llvm_stop_multithreaded()</tt> if LLVM is operating in multithreaded mode.
2481 As such, <tt>llvm_shutdown()</tt> requires the same isolation guarantees as
2482 <tt>llvm_stop_multithreaded()</tt>.
2486 Note that, if you use scope-based shutdown, you can use the
2487 <tt>llvm_shutdown_obj</tt> class, which calls <tt>llvm_shutdown()</tt> in its
2491 <!-- ======================================================================= -->
2492 <div class="doc_subsection">
2493 <a name="managedstatic">Lazy Initialization with <tt>ManagedStatic</tt></a>
2496 <div class="doc_text">
2498 <tt>ManagedStatic</tt> is a utility class in LLVM used to implement static
2499 initialization of static resources, such as the global type tables. Before the
2500 invocation of <tt>llvm_shutdown()</tt>, it implements a simple lazy
2501 initialization scheme. Once <tt>llvm_start_multithreaded()</tt> returns,
2502 however, it uses double-checked locking to implement thread-safe lazy
2507 Note that, because no other threads are allowed to issue LLVM API calls before
2508 <tt>llvm_start_multithreaded()</tt> returns, it is possible to have
2509 <tt>ManagedStatic</tt>s of <tt>llvm::sys::Mutex</tt>s.
2513 The <tt>llvm_acquire_global_lock()</tt> and <tt>llvm_release_global_lock</tt>
2514 APIs provide access to the global lock used to implement the double-checked
2515 locking for lazy initialization. These should only be used internally to LLVM,
2516 and only if you know what you're doing!
2520 <!-- ======================================================================= -->
2521 <div class="doc_subsection">
2522 <a name="llvmcontext">Achieving Isolation with <tt>LLVMContext</tt></a>
2525 <div class="doc_text">
2527 <tt>LLVMContext</tt> is an opaque class in the LLVM API which clients can use
2528 to operate multiple, isolated instances of LLVM concurrently within the same
2529 address space. For instance, in a hypothetical compile-server, the compilation
2530 of an individual translation unit is conceptually independent from all the
2531 others, and it would be desirable to be able to compile incoming translation
2532 units concurrently on independent server threads. Fortunately,
2533 <tt>LLVMContext</tt> exists to enable just this kind of scenario!
2537 Conceptually, <tt>LLVMContext</tt> provides isolation. Every LLVM entity
2538 (<tt>Module</tt>s, <tt>Value</tt>s, <tt>Type</tt>s, <tt>Constant</tt>s, etc.)
2539 in LLVM's in-memory IR belongs to an <tt>LLVMContext</tt>. Entities in
2540 different contexts <em>cannot</em> interact with each other: <tt>Module</tt>s in
2541 different contexts cannot be linked together, <tt>Function</tt>s cannot be added
2542 to <tt>Module</tt>s in different contexts, etc. What this means is that is is
2543 safe to compile on multiple threads simultaneously, as long as no two threads
2544 operate on entities within the same context.
2548 In practice, very few places in the API require the explicit specification of a
2549 <tt>LLVMContext</tt>, other than the <tt>Type</tt> creation/lookup APIs.
2550 Because every <tt>Type</tt> carries a reference to its owning context, most
2551 other entities can determine what context they belong to by looking at their
2552 own <tt>Type</tt>. If you are adding new entities to LLVM IR, please try to
2553 maintain this interface design.
2557 For clients that do <em>not</em> require the benefits of isolation, LLVM
2558 provides a convenience API <tt>getGlobalContext()</tt>. This returns a global,
2559 lazily initialized <tt>LLVMContext</tt> that may be used in situations where
2560 isolation is not a concern.
2564 <!-- ======================================================================= -->
2565 <div class="doc_subsection">
2566 <a name="jitthreading">Threads and the JIT</a>
2569 <div class="doc_text">
2571 LLVM's "eager" JIT compiler is safe to use in threaded programs. Multiple
2572 threads can call <tt>ExecutionEngine::getPointerToFunction()</tt> or
2573 <tt>ExecutionEngine::runFunction()</tt> concurrently, and multiple threads can
2574 run code output by the JIT concurrently. The user must still ensure that only
2575 one thread accesses IR in a given <tt>LLVMContext</tt> while another thread
2576 might be modifying it. One way to do that is to always hold the JIT lock while
2577 accessing IR outside the JIT (the JIT <em>modifies</em> the IR by adding
2578 <tt>CallbackVH</tt>s). Another way is to only
2579 call <tt>getPointerToFunction()</tt> from the <tt>LLVMContext</tt>'s thread.
2582 <p>When the JIT is configured to compile lazily (using
2583 <tt>ExecutionEngine::DisableLazyCompilation(false)</tt>), there is currently a
2584 <a href="http://llvm.org/bugs/show_bug.cgi?id=5184">race condition</a> in
2585 updating call sites after a function is lazily-jitted. It's still possible to
2586 use the lazy JIT in a threaded program if you ensure that only one thread at a
2587 time can call any particular lazy stub and that the JIT lock guards any IR
2588 access, but we suggest using only the eager JIT in threaded programs.
2592 <!-- *********************************************************************** -->
2593 <div class="doc_section">
2594 <a name="advanced">Advanced Topics</a>
2596 <!-- *********************************************************************** -->
2598 <div class="doc_text">
2600 This section describes some of the advanced or obscure API's that most clients
2601 do not need to be aware of. These API's tend manage the inner workings of the
2602 LLVM system, and only need to be accessed in unusual circumstances.
2606 <!-- ======================================================================= -->
2607 <div class="doc_subsection">
2608 <a name="TypeResolve">LLVM Type Resolution</a>
2611 <div class="doc_text">
2614 The LLVM type system has a very simple goal: allow clients to compare types for
2615 structural equality with a simple pointer comparison (aka a shallow compare).
2616 This goal makes clients much simpler and faster, and is used throughout the LLVM
2621 Unfortunately achieving this goal is not a simple matter. In particular,
2622 recursive types and late resolution of opaque types makes the situation very
2623 difficult to handle. Fortunately, for the most part, our implementation makes
2624 most clients able to be completely unaware of the nasty internal details. The
2625 primary case where clients are exposed to the inner workings of it are when
2626 building a recursive type. In addition to this case, the LLVM bitcode reader,
2627 assembly parser, and linker also have to be aware of the inner workings of this
2632 For our purposes below, we need three concepts. First, an "Opaque Type" is
2633 exactly as defined in the <a href="LangRef.html#t_opaque">language
2634 reference</a>. Second an "Abstract Type" is any type which includes an
2635 opaque type as part of its type graph (for example "<tt>{ opaque, i32 }</tt>").
2636 Third, a concrete type is a type that is not an abstract type (e.g. "<tt>{ i32,
2642 <!-- ______________________________________________________________________ -->
2643 <div class="doc_subsubsection">
2644 <a name="BuildRecType">Basic Recursive Type Construction</a>
2647 <div class="doc_text">
2650 Because the most common question is "how do I build a recursive type with LLVM",
2651 we answer it now and explain it as we go. Here we include enough to cause this
2652 to be emitted to an output .ll file:
2655 <div class="doc_code">
2657 %mylist = type { %mylist*, i32 }
2662 To build this, use the following LLVM APIs:
2665 <div class="doc_code">
2667 // <i>Create the initial outer struct</i>
2668 <a href="#PATypeHolder">PATypeHolder</a> StructTy = OpaqueType::get();
2669 std::vector<const Type*> Elts;
2670 Elts.push_back(PointerType::getUnqual(StructTy));
2671 Elts.push_back(Type::Int32Ty);
2672 StructType *NewSTy = StructType::get(Elts);
2674 // <i>At this point, NewSTy = "{ opaque*, i32 }". Tell VMCore that</i>
2675 // <i>the struct and the opaque type are actually the same.</i>
2676 cast<OpaqueType>(StructTy.get())-><a href="#refineAbstractTypeTo">refineAbstractTypeTo</a>(NewSTy);
2678 // <i>NewSTy is potentially invalidated, but StructTy (a <a href="#PATypeHolder">PATypeHolder</a>) is</i>
2679 // <i>kept up-to-date</i>
2680 NewSTy = cast<StructType>(StructTy.get());
2682 // <i>Add a name for the type to the module symbol table (optional)</i>
2683 MyModule->addTypeName("mylist", NewSTy);
2688 This code shows the basic approach used to build recursive types: build a
2689 non-recursive type using 'opaque', then use type unification to close the cycle.
2690 The type unification step is performed by the <tt><a
2691 href="#refineAbstractTypeTo">refineAbstractTypeTo</a></tt> method, which is
2692 described next. After that, we describe the <a
2693 href="#PATypeHolder">PATypeHolder class</a>.
2698 <!-- ______________________________________________________________________ -->
2699 <div class="doc_subsubsection">
2700 <a name="refineAbstractTypeTo">The <tt>refineAbstractTypeTo</tt> method</a>
2703 <div class="doc_text">
2705 The <tt>refineAbstractTypeTo</tt> method starts the type unification process.
2706 While this method is actually a member of the DerivedType class, it is most
2707 often used on OpaqueType instances. Type unification is actually a recursive
2708 process. After unification, types can become structurally isomorphic to
2709 existing types, and all duplicates are deleted (to preserve pointer equality).
2713 In the example above, the OpaqueType object is definitely deleted.
2714 Additionally, if there is an "{ \2*, i32}" type already created in the system,
2715 the pointer and struct type created are <b>also</b> deleted. Obviously whenever
2716 a type is deleted, any "Type*" pointers in the program are invalidated. As
2717 such, it is safest to avoid having <i>any</i> "Type*" pointers to abstract types
2718 live across a call to <tt>refineAbstractTypeTo</tt> (note that non-abstract
2719 types can never move or be deleted). To deal with this, the <a
2720 href="#PATypeHolder">PATypeHolder</a> class is used to maintain a stable
2721 reference to a possibly refined type, and the <a
2722 href="#AbstractTypeUser">AbstractTypeUser</a> class is used to update more
2723 complex datastructures.
2728 <!-- ______________________________________________________________________ -->
2729 <div class="doc_subsubsection">
2730 <a name="PATypeHolder">The PATypeHolder Class</a>
2733 <div class="doc_text">
2735 PATypeHolder is a form of a "smart pointer" for Type objects. When VMCore
2736 happily goes about nuking types that become isomorphic to existing types, it
2737 automatically updates all PATypeHolder objects to point to the new type. In the
2738 example above, this allows the code to maintain a pointer to the resultant
2739 resolved recursive type, even though the Type*'s are potentially invalidated.
2743 PATypeHolder is an extremely light-weight object that uses a lazy union-find
2744 implementation to update pointers. For example the pointer from a Value to its
2745 Type is maintained by PATypeHolder objects.
2750 <!-- ______________________________________________________________________ -->
2751 <div class="doc_subsubsection">
2752 <a name="AbstractTypeUser">The AbstractTypeUser Class</a>
2755 <div class="doc_text">
2758 Some data structures need more to perform more complex updates when types get
2759 resolved. To support this, a class can derive from the AbstractTypeUser class.
2761 allows it to get callbacks when certain types are resolved. To register to get
2762 callbacks for a particular type, the DerivedType::{add/remove}AbstractTypeUser
2763 methods can be called on a type. Note that these methods only work for <i>
2764 abstract</i> types. Concrete types (those that do not include any opaque
2765 objects) can never be refined.
2770 <!-- ======================================================================= -->
2771 <div class="doc_subsection">
2772 <a name="SymbolTable">The <tt>ValueSymbolTable</tt> and
2773 <tt>TypeSymbolTable</tt> classes</a>
2776 <div class="doc_text">
2777 <p>The <tt><a href="http://llvm.org/doxygen/classllvm_1_1ValueSymbolTable.html">
2778 ValueSymbolTable</a></tt> class provides a symbol table that the <a
2779 href="#Function"><tt>Function</tt></a> and <a href="#Module">
2780 <tt>Module</tt></a> classes use for naming value definitions. The symbol table
2781 can provide a name for any <a href="#Value"><tt>Value</tt></a>.
2782 The <tt><a href="http://llvm.org/doxygen/classllvm_1_1TypeSymbolTable.html">
2783 TypeSymbolTable</a></tt> class is used by the <tt>Module</tt> class to store
2784 names for types.</p>
2786 <p>Note that the <tt>SymbolTable</tt> class should not be directly accessed
2787 by most clients. It should only be used when iteration over the symbol table
2788 names themselves are required, which is very special purpose. Note that not
2790 <tt><a href="#Value">Value</a></tt>s have names, and those without names (i.e. they have
2791 an empty name) do not exist in the symbol table.
2794 <p>These symbol tables support iteration over the values/types in the symbol
2795 table with <tt>begin/end/iterator</tt> and supports querying to see if a
2796 specific name is in the symbol table (with <tt>lookup</tt>). The
2797 <tt>ValueSymbolTable</tt> class exposes no public mutator methods, instead,
2798 simply call <tt>setName</tt> on a value, which will autoinsert it into the
2799 appropriate symbol table. For types, use the Module::addTypeName method to
2800 insert entries into the symbol table.</p>
2806 <!-- ======================================================================= -->
2807 <div class="doc_subsection">
2808 <a name="UserLayout">The <tt>User</tt> and owned <tt>Use</tt> classes' memory layout</a>
2811 <div class="doc_text">
2812 <p>The <tt><a href="http://llvm.org/doxygen/classllvm_1_1User.html">
2813 User</a></tt> class provides a basis for expressing the ownership of <tt>User</tt>
2814 towards other <tt><a href="http://llvm.org/doxygen/classllvm_1_1Value.html">
2815 Value</a></tt>s. The <tt><a href="http://llvm.org/doxygen/classllvm_1_1Use.html">
2816 Use</a></tt> helper class is employed to do the bookkeeping and to facilitate <i>O(1)</i>
2817 addition and removal.</p>
2819 <!-- ______________________________________________________________________ -->
2820 <div class="doc_subsubsection">
2821 <a name="Use2User">Interaction and relationship between <tt>User</tt> and <tt>Use</tt> objects</a>
2824 <div class="doc_text">
2826 A subclass of <tt>User</tt> can choose between incorporating its <tt>Use</tt> objects
2827 or refer to them out-of-line by means of a pointer. A mixed variant
2828 (some <tt>Use</tt>s inline others hung off) is impractical and breaks the invariant
2829 that the <tt>Use</tt> objects belonging to the same <tt>User</tt> form a contiguous array.
2834 We have 2 different layouts in the <tt>User</tt> (sub)classes:
2837 The <tt>Use</tt> object(s) are inside (resp. at fixed offset) of the <tt>User</tt>
2838 object and there are a fixed number of them.</p>
2841 The <tt>Use</tt> object(s) are referenced by a pointer to an
2842 array from the <tt>User</tt> object and there may be a variable
2846 As of v2.4 each layout still possesses a direct pointer to the
2847 start of the array of <tt>Use</tt>s. Though not mandatory for layout a),
2848 we stick to this redundancy for the sake of simplicity.
2849 The <tt>User</tt> object also stores the number of <tt>Use</tt> objects it
2850 has. (Theoretically this information can also be calculated
2851 given the scheme presented below.)</p>
2853 Special forms of allocation operators (<tt>operator new</tt>)
2854 enforce the following memory layouts:</p>
2857 <li><p>Layout a) is modelled by prepending the <tt>User</tt> object by the <tt>Use[]</tt> array.</p>
2860 ...---.---.---.---.-------...
2861 | P | P | P | P | User
2862 '''---'---'---'---'-------'''
2865 <li><p>Layout b) is modelled by pointing at the <tt>Use[]</tt> array.</p>
2877 <i>(In the above figures '<tt>P</tt>' stands for the <tt>Use**</tt> that
2878 is stored in each <tt>Use</tt> object in the member <tt>Use::Prev</tt>)</i>
2880 <!-- ______________________________________________________________________ -->
2881 <div class="doc_subsubsection">
2882 <a name="Waymarking">The waymarking algorithm</a>
2885 <div class="doc_text">
2887 Since the <tt>Use</tt> objects are deprived of the direct (back)pointer to
2888 their <tt>User</tt> objects, there must be a fast and exact method to
2889 recover it. This is accomplished by the following scheme:</p>
2892 A bit-encoding in the 2 LSBits (least significant bits) of the <tt>Use::Prev</tt> allows to find the
2893 start of the <tt>User</tt> object:
2895 <li><tt>00</tt> —> binary digit 0</li>
2896 <li><tt>01</tt> —> binary digit 1</li>
2897 <li><tt>10</tt> —> stop and calculate (<tt>s</tt>)</li>
2898 <li><tt>11</tt> —> full stop (<tt>S</tt>)</li>
2901 Given a <tt>Use*</tt>, all we have to do is to walk till we get
2902 a stop and we either have a <tt>User</tt> immediately behind or
2903 we have to walk to the next stop picking up digits
2904 and calculating the offset:</p>
2906 .---.---.---.---.---.---.---.---.---.---.---.---.---.---.---.---.----------------
2907 | 1 | s | 1 | 0 | 1 | 0 | s | 1 | 1 | 0 | s | 1 | 1 | s | 1 | S | User (or User*)
2908 '---'---'---'---'---'---'---'---'---'---'---'---'---'---'---'---'----------------
2909 |+15 |+10 |+6 |+3 |+1
2912 | | |______________________>
2913 | |______________________________________>
2914 |__________________________________________________________>
2917 Only the significant number of bits need to be stored between the
2918 stops, so that the <i>worst case is 20 memory accesses</i> when there are
2919 1000 <tt>Use</tt> objects associated with a <tt>User</tt>.</p>
2921 <!-- ______________________________________________________________________ -->
2922 <div class="doc_subsubsection">
2923 <a name="ReferenceImpl">Reference implementation</a>
2926 <div class="doc_text">
2928 The following literate Haskell fragment demonstrates the concept:</p>
2931 <div class="doc_code">
2933 > import Test.QuickCheck
2935 > digits :: Int -> [Char] -> [Char]
2936 > digits 0 acc = '0' : acc
2937 > digits 1 acc = '1' : acc
2938 > digits n acc = digits (n `div` 2) $ digits (n `mod` 2) acc
2940 > dist :: Int -> [Char] -> [Char]
2943 > dist 1 acc = let r = dist 0 acc in 's' : digits (length r) r
2944 > dist n acc = dist (n - 1) $ dist 1 acc
2946 > takeLast n ss = reverse $ take n $ reverse ss
2948 > test = takeLast 40 $ dist 20 []
2953 Printing <test> gives: <tt>"1s100000s11010s10100s1111s1010s110s11s1S"</tt></p>
2955 The reverse algorithm computes the length of the string just by examining
2956 a certain prefix:</p>
2958 <div class="doc_code">
2960 > pref :: [Char] -> Int
2962 > pref ('s':'1':rest) = decode 2 1 rest
2963 > pref (_:rest) = 1 + pref rest
2965 > decode walk acc ('0':rest) = decode (walk + 1) (acc * 2) rest
2966 > decode walk acc ('1':rest) = decode (walk + 1) (acc * 2 + 1) rest
2967 > decode walk acc _ = walk + acc
2972 Now, as expected, printing <pref test> gives <tt>40</tt>.</p>
2974 We can <i>quickCheck</i> this with following property:</p>
2976 <div class="doc_code">
2978 > testcase = dist 2000 []
2979 > testcaseLength = length testcase
2981 > identityProp n = n > 0 && n <= testcaseLength ==> length arr == pref arr
2982 > where arr = takeLast n testcase
2987 As expected <quickCheck identityProp> gives:</p>
2990 *Main> quickCheck identityProp
2991 OK, passed 100 tests.
2994 Let's be a bit more exhaustive:</p>
2996 <div class="doc_code">
2999 > deepCheck p = check (defaultConfig { configMaxTest = 500 }) p
3004 And here is the result of <deepCheck identityProp>:</p>
3007 *Main> deepCheck identityProp
3008 OK, passed 500 tests.
3011 <!-- ______________________________________________________________________ -->
3012 <div class="doc_subsubsection">
3013 <a name="Tagging">Tagging considerations</a>
3017 To maintain the invariant that the 2 LSBits of each <tt>Use**</tt> in <tt>Use</tt>
3018 never change after being set up, setters of <tt>Use::Prev</tt> must re-tag the
3019 new <tt>Use**</tt> on every modification. Accordingly getters must strip the
3022 For layout b) instead of the <tt>User</tt> we find a pointer (<tt>User*</tt> with LSBit set).
3023 Following this pointer brings us to the <tt>User</tt>. A portable trick ensures
3024 that the first bytes of <tt>User</tt> (if interpreted as a pointer) never has
3025 the LSBit set. (Portability is relying on the fact that all known compilers place the
3026 <tt>vptr</tt> in the first word of the instances.)</p>
3030 <!-- *********************************************************************** -->
3031 <div class="doc_section">
3032 <a name="coreclasses">The Core LLVM Class Hierarchy Reference </a>
3034 <!-- *********************************************************************** -->
3036 <div class="doc_text">
3037 <p><tt>#include "<a href="/doxygen/Type_8h-source.html">llvm/Type.h</a>"</tt>
3038 <br>doxygen info: <a href="/doxygen/classllvm_1_1Type.html">Type Class</a></p>
3040 <p>The Core LLVM classes are the primary means of representing the program
3041 being inspected or transformed. The core LLVM classes are defined in
3042 header files in the <tt>include/llvm/</tt> directory, and implemented in
3043 the <tt>lib/VMCore</tt> directory.</p>
3047 <!-- ======================================================================= -->
3048 <div class="doc_subsection">
3049 <a name="Type">The <tt>Type</tt> class and Derived Types</a>
3052 <div class="doc_text">
3054 <p><tt>Type</tt> is a superclass of all type classes. Every <tt>Value</tt> has
3055 a <tt>Type</tt>. <tt>Type</tt> cannot be instantiated directly but only
3056 through its subclasses. Certain primitive types (<tt>VoidType</tt>,
3057 <tt>LabelType</tt>, <tt>FloatType</tt> and <tt>DoubleType</tt>) have hidden
3058 subclasses. They are hidden because they offer no useful functionality beyond
3059 what the <tt>Type</tt> class offers except to distinguish themselves from
3060 other subclasses of <tt>Type</tt>.</p>
3061 <p>All other types are subclasses of <tt>DerivedType</tt>. Types can be
3062 named, but this is not a requirement. There exists exactly
3063 one instance of a given shape at any one time. This allows type equality to
3064 be performed with address equality of the Type Instance. That is, given two
3065 <tt>Type*</tt> values, the types are identical if the pointers are identical.
3069 <!-- _______________________________________________________________________ -->
3070 <div class="doc_subsubsection">
3071 <a name="m_Type">Important Public Methods</a>
3074 <div class="doc_text">
3077 <li><tt>bool isIntegerTy() const</tt>: Returns true for any integer type.</li>
3079 <li><tt>bool isFloatingPointTy()</tt>: Return true if this is one of the five
3080 floating point types.</li>
3082 <li><tt>bool isAbstract()</tt>: Return true if the type is abstract (contains
3083 an OpaqueType anywhere in its definition).</li>
3085 <li><tt>bool isSized()</tt>: Return true if the type has known size. Things
3086 that don't have a size are abstract types, labels and void.</li>
3091 <!-- _______________________________________________________________________ -->
3092 <div class="doc_subsubsection">
3093 <a name="derivedtypes">Important Derived Types</a>
3095 <div class="doc_text">
3097 <dt><tt>IntegerType</tt></dt>
3098 <dd>Subclass of DerivedType that represents integer types of any bit width.
3099 Any bit width between <tt>IntegerType::MIN_INT_BITS</tt> (1) and
3100 <tt>IntegerType::MAX_INT_BITS</tt> (~8 million) can be represented.
3102 <li><tt>static const IntegerType* get(unsigned NumBits)</tt>: get an integer
3103 type of a specific bit width.</li>
3104 <li><tt>unsigned getBitWidth() const</tt>: Get the bit width of an integer
3108 <dt><tt>SequentialType</tt></dt>
3109 <dd>This is subclassed by ArrayType and PointerType
3111 <li><tt>const Type * getElementType() const</tt>: Returns the type of each
3112 of the elements in the sequential type. </li>
3115 <dt><tt>ArrayType</tt></dt>
3116 <dd>This is a subclass of SequentialType and defines the interface for array
3119 <li><tt>unsigned getNumElements() const</tt>: Returns the number of
3120 elements in the array. </li>
3123 <dt><tt>PointerType</tt></dt>
3124 <dd>Subclass of SequentialType for pointer types.</dd>
3125 <dt><tt>VectorType</tt></dt>
3126 <dd>Subclass of SequentialType for vector types. A
3127 vector type is similar to an ArrayType but is distinguished because it is
3128 a first class type whereas ArrayType is not. Vector types are used for
3129 vector operations and are usually small vectors of of an integer or floating
3131 <dt><tt>StructType</tt></dt>
3132 <dd>Subclass of DerivedTypes for struct types.</dd>
3133 <dt><tt><a name="FunctionType">FunctionType</a></tt></dt>
3134 <dd>Subclass of DerivedTypes for function types.
3136 <li><tt>bool isVarArg() const</tt>: Returns true if it's a vararg
3138 <li><tt> const Type * getReturnType() const</tt>: Returns the
3139 return type of the function.</li>
3140 <li><tt>const Type * getParamType (unsigned i)</tt>: Returns
3141 the type of the ith parameter.</li>
3142 <li><tt> const unsigned getNumParams() const</tt>: Returns the
3143 number of formal parameters.</li>
3146 <dt><tt>OpaqueType</tt></dt>
3147 <dd>Sublcass of DerivedType for abstract types. This class
3148 defines no content and is used as a placeholder for some other type. Note
3149 that OpaqueType is used (temporarily) during type resolution for forward
3150 references of types. Once the referenced type is resolved, the OpaqueType
3151 is replaced with the actual type. OpaqueType can also be used for data
3152 abstraction. At link time opaque types can be resolved to actual types
3153 of the same name.</dd>
3159 <!-- ======================================================================= -->
3160 <div class="doc_subsection">
3161 <a name="Module">The <tt>Module</tt> class</a>
3164 <div class="doc_text">
3167 href="/doxygen/Module_8h-source.html">llvm/Module.h</a>"</tt><br> doxygen info:
3168 <a href="/doxygen/classllvm_1_1Module.html">Module Class</a></p>
3170 <p>The <tt>Module</tt> class represents the top level structure present in LLVM
3171 programs. An LLVM module is effectively either a translation unit of the
3172 original program or a combination of several translation units merged by the
3173 linker. The <tt>Module</tt> class keeps track of a list of <a
3174 href="#Function"><tt>Function</tt></a>s, a list of <a
3175 href="#GlobalVariable"><tt>GlobalVariable</tt></a>s, and a <a
3176 href="#SymbolTable"><tt>SymbolTable</tt></a>. Additionally, it contains a few
3177 helpful member functions that try to make common operations easy.</p>
3181 <!-- _______________________________________________________________________ -->
3182 <div class="doc_subsubsection">
3183 <a name="m_Module">Important Public Members of the <tt>Module</tt> class</a>
3186 <div class="doc_text">
3189 <li><tt>Module::Module(std::string name = "")</tt></li>
3192 <p>Constructing a <a href="#Module">Module</a> is easy. You can optionally
3193 provide a name for it (probably based on the name of the translation unit).</p>
3196 <li><tt>Module::iterator</tt> - Typedef for function list iterator<br>
3197 <tt>Module::const_iterator</tt> - Typedef for const_iterator.<br>
3199 <tt>begin()</tt>, <tt>end()</tt>
3200 <tt>size()</tt>, <tt>empty()</tt>
3202 <p>These are forwarding methods that make it easy to access the contents of
3203 a <tt>Module</tt> object's <a href="#Function"><tt>Function</tt></a>
3206 <li><tt>Module::FunctionListType &getFunctionList()</tt>
3208 <p> Returns the list of <a href="#Function"><tt>Function</tt></a>s. This is
3209 necessary to use when you need to update the list or perform a complex
3210 action that doesn't have a forwarding method.</p>
3212 <p><!-- Global Variable --></p></li>
3218 <li><tt>Module::global_iterator</tt> - Typedef for global variable list iterator<br>
3220 <tt>Module::const_global_iterator</tt> - Typedef for const_iterator.<br>
3222 <tt>global_begin()</tt>, <tt>global_end()</tt>
3223 <tt>global_size()</tt>, <tt>global_empty()</tt>
3225 <p> These are forwarding methods that make it easy to access the contents of
3226 a <tt>Module</tt> object's <a
3227 href="#GlobalVariable"><tt>GlobalVariable</tt></a> list.</p></li>
3229 <li><tt>Module::GlobalListType &getGlobalList()</tt>
3231 <p>Returns the list of <a
3232 href="#GlobalVariable"><tt>GlobalVariable</tt></a>s. This is necessary to
3233 use when you need to update the list or perform a complex action that
3234 doesn't have a forwarding method.</p>
3236 <p><!-- Symbol table stuff --> </p></li>
3242 <li><tt><a href="#SymbolTable">SymbolTable</a> *getSymbolTable()</tt>
3244 <p>Return a reference to the <a href="#SymbolTable"><tt>SymbolTable</tt></a>
3245 for this <tt>Module</tt>.</p>
3247 <p><!-- Convenience methods --></p></li>
3253 <li><tt><a href="#Function">Function</a> *getFunction(const std::string
3254 &Name, const <a href="#FunctionType">FunctionType</a> *Ty)</tt>
3256 <p>Look up the specified function in the <tt>Module</tt> <a
3257 href="#SymbolTable"><tt>SymbolTable</tt></a>. If it does not exist, return
3258 <tt>null</tt>.</p></li>
3260 <li><tt><a href="#Function">Function</a> *getOrInsertFunction(const
3261 std::string &Name, const <a href="#FunctionType">FunctionType</a> *T)</tt>
3263 <p>Look up the specified function in the <tt>Module</tt> <a
3264 href="#SymbolTable"><tt>SymbolTable</tt></a>. If it does not exist, add an
3265 external declaration for the function and return it.</p></li>
3267 <li><tt>std::string getTypeName(const <a href="#Type">Type</a> *Ty)</tt>
3269 <p>If there is at least one entry in the <a
3270 href="#SymbolTable"><tt>SymbolTable</tt></a> for the specified <a
3271 href="#Type"><tt>Type</tt></a>, return it. Otherwise return the empty
3274 <li><tt>bool addTypeName(const std::string &Name, const <a
3275 href="#Type">Type</a> *Ty)</tt>
3277 <p>Insert an entry in the <a href="#SymbolTable"><tt>SymbolTable</tt></a>
3278 mapping <tt>Name</tt> to <tt>Ty</tt>. If there is already an entry for this
3279 name, true is returned and the <a
3280 href="#SymbolTable"><tt>SymbolTable</tt></a> is not modified.</p></li>
3286 <!-- ======================================================================= -->
3287 <div class="doc_subsection">
3288 <a name="Value">The <tt>Value</tt> class</a>
3291 <div class="doc_text">
3293 <p><tt>#include "<a href="/doxygen/Value_8h-source.html">llvm/Value.h</a>"</tt>
3295 doxygen info: <a href="/doxygen/classllvm_1_1Value.html">Value Class</a></p>
3297 <p>The <tt>Value</tt> class is the most important class in the LLVM Source
3298 base. It represents a typed value that may be used (among other things) as an
3299 operand to an instruction. There are many different types of <tt>Value</tt>s,
3300 such as <a href="#Constant"><tt>Constant</tt></a>s,<a
3301 href="#Argument"><tt>Argument</tt></a>s. Even <a
3302 href="#Instruction"><tt>Instruction</tt></a>s and <a
3303 href="#Function"><tt>Function</tt></a>s are <tt>Value</tt>s.</p>
3305 <p>A particular <tt>Value</tt> may be used many times in the LLVM representation
3306 for a program. For example, an incoming argument to a function (represented
3307 with an instance of the <a href="#Argument">Argument</a> class) is "used" by
3308 every instruction in the function that references the argument. To keep track
3309 of this relationship, the <tt>Value</tt> class keeps a list of all of the <a
3310 href="#User"><tt>User</tt></a>s that is using it (the <a
3311 href="#User"><tt>User</tt></a> class is a base class for all nodes in the LLVM
3312 graph that can refer to <tt>Value</tt>s). This use list is how LLVM represents
3313 def-use information in the program, and is accessible through the <tt>use_</tt>*
3314 methods, shown below.</p>
3316 <p>Because LLVM is a typed representation, every LLVM <tt>Value</tt> is typed,
3317 and this <a href="#Type">Type</a> is available through the <tt>getType()</tt>
3318 method. In addition, all LLVM values can be named. The "name" of the
3319 <tt>Value</tt> is a symbolic string printed in the LLVM code:</p>
3321 <div class="doc_code">
3323 %<b>foo</b> = add i32 1, 2
3327 <p><a name="nameWarning">The name of this instruction is "foo".</a> <b>NOTE</b>
3328 that the name of any value may be missing (an empty string), so names should
3329 <b>ONLY</b> be used for debugging (making the source code easier to read,
3330 debugging printouts), they should not be used to keep track of values or map
3331 between them. For this purpose, use a <tt>std::map</tt> of pointers to the
3332 <tt>Value</tt> itself instead.</p>
3334 <p>One important aspect of LLVM is that there is no distinction between an SSA
3335 variable and the operation that produces it. Because of this, any reference to
3336 the value produced by an instruction (or the value available as an incoming
3337 argument, for example) is represented as a direct pointer to the instance of
3339 represents this value. Although this may take some getting used to, it
3340 simplifies the representation and makes it easier to manipulate.</p>
3344 <!-- _______________________________________________________________________ -->
3345 <div class="doc_subsubsection">
3346 <a name="m_Value">Important Public Members of the <tt>Value</tt> class</a>
3349 <div class="doc_text">
3352 <li><tt>Value::use_iterator</tt> - Typedef for iterator over the
3354 <tt>Value::const_use_iterator</tt> - Typedef for const_iterator over
3356 <tt>unsigned use_size()</tt> - Returns the number of users of the
3358 <tt>bool use_empty()</tt> - Returns true if there are no users.<br>
3359 <tt>use_iterator use_begin()</tt> - Get an iterator to the start of
3361 <tt>use_iterator use_end()</tt> - Get an iterator to the end of the
3363 <tt><a href="#User">User</a> *use_back()</tt> - Returns the last
3364 element in the list.
3365 <p> These methods are the interface to access the def-use
3366 information in LLVM. As with all other iterators in LLVM, the naming
3367 conventions follow the conventions defined by the <a href="#stl">STL</a>.</p>
3369 <li><tt><a href="#Type">Type</a> *getType() const</tt>
3370 <p>This method returns the Type of the Value.</p>
3372 <li><tt>bool hasName() const</tt><br>
3373 <tt>std::string getName() const</tt><br>
3374 <tt>void setName(const std::string &Name)</tt>
3375 <p> This family of methods is used to access and assign a name to a <tt>Value</tt>,
3376 be aware of the <a href="#nameWarning">precaution above</a>.</p>
3378 <li><tt>void replaceAllUsesWith(Value *V)</tt>
3380 <p>This method traverses the use list of a <tt>Value</tt> changing all <a
3381 href="#User"><tt>User</tt>s</a> of the current value to refer to
3382 "<tt>V</tt>" instead. For example, if you detect that an instruction always
3383 produces a constant value (for example through constant folding), you can
3384 replace all uses of the instruction with the constant like this:</p>
3386 <div class="doc_code">
3388 Inst->replaceAllUsesWith(ConstVal);
3396 <!-- ======================================================================= -->
3397 <div class="doc_subsection">
3398 <a name="User">The <tt>User</tt> class</a>
3401 <div class="doc_text">
3404 <tt>#include "<a href="/doxygen/User_8h-source.html">llvm/User.h</a>"</tt><br>
3405 doxygen info: <a href="/doxygen/classllvm_1_1User.html">User Class</a><br>
3406 Superclass: <a href="#Value"><tt>Value</tt></a></p>
3408 <p>The <tt>User</tt> class is the common base class of all LLVM nodes that may
3409 refer to <a href="#Value"><tt>Value</tt></a>s. It exposes a list of "Operands"
3410 that are all of the <a href="#Value"><tt>Value</tt></a>s that the User is
3411 referring to. The <tt>User</tt> class itself is a subclass of
3414 <p>The operands of a <tt>User</tt> point directly to the LLVM <a
3415 href="#Value"><tt>Value</tt></a> that it refers to. Because LLVM uses Static
3416 Single Assignment (SSA) form, there can only be one definition referred to,
3417 allowing this direct connection. This connection provides the use-def
3418 information in LLVM.</p>
3422 <!-- _______________________________________________________________________ -->
3423 <div class="doc_subsubsection">
3424 <a name="m_User">Important Public Members of the <tt>User</tt> class</a>
3427 <div class="doc_text">
3429 <p>The <tt>User</tt> class exposes the operand list in two ways: through
3430 an index access interface and through an iterator based interface.</p>
3433 <li><tt>Value *getOperand(unsigned i)</tt><br>
3434 <tt>unsigned getNumOperands()</tt>
3435 <p> These two methods expose the operands of the <tt>User</tt> in a
3436 convenient form for direct access.</p></li>
3438 <li><tt>User::op_iterator</tt> - Typedef for iterator over the operand
3440 <tt>op_iterator op_begin()</tt> - Get an iterator to the start of
3441 the operand list.<br>
3442 <tt>op_iterator op_end()</tt> - Get an iterator to the end of the
3444 <p> Together, these methods make up the iterator based interface to
3445 the operands of a <tt>User</tt>.</p></li>
3450 <!-- ======================================================================= -->
3451 <div class="doc_subsection">
3452 <a name="Instruction">The <tt>Instruction</tt> class</a>
3455 <div class="doc_text">
3457 <p><tt>#include "</tt><tt><a
3458 href="/doxygen/Instruction_8h-source.html">llvm/Instruction.h</a>"</tt><br>
3459 doxygen info: <a href="/doxygen/classllvm_1_1Instruction.html">Instruction Class</a><br>
3460 Superclasses: <a href="#User"><tt>User</tt></a>, <a
3461 href="#Value"><tt>Value</tt></a></p>
3463 <p>The <tt>Instruction</tt> class is the common base class for all LLVM
3464 instructions. It provides only a few methods, but is a very commonly used
3465 class. The primary data tracked by the <tt>Instruction</tt> class itself is the
3466 opcode (instruction type) and the parent <a
3467 href="#BasicBlock"><tt>BasicBlock</tt></a> the <tt>Instruction</tt> is embedded
3468 into. To represent a specific type of instruction, one of many subclasses of
3469 <tt>Instruction</tt> are used.</p>
3471 <p> Because the <tt>Instruction</tt> class subclasses the <a
3472 href="#User"><tt>User</tt></a> class, its operands can be accessed in the same
3473 way as for other <a href="#User"><tt>User</tt></a>s (with the
3474 <tt>getOperand()</tt>/<tt>getNumOperands()</tt> and
3475 <tt>op_begin()</tt>/<tt>op_end()</tt> methods).</p> <p> An important file for
3476 the <tt>Instruction</tt> class is the <tt>llvm/Instruction.def</tt> file. This
3477 file contains some meta-data about the various different types of instructions
3478 in LLVM. It describes the enum values that are used as opcodes (for example
3479 <tt>Instruction::Add</tt> and <tt>Instruction::ICmp</tt>), as well as the
3480 concrete sub-classes of <tt>Instruction</tt> that implement the instruction (for
3481 example <tt><a href="#BinaryOperator">BinaryOperator</a></tt> and <tt><a
3482 href="#CmpInst">CmpInst</a></tt>). Unfortunately, the use of macros in
3483 this file confuses doxygen, so these enum values don't show up correctly in the
3484 <a href="/doxygen/classllvm_1_1Instruction.html">doxygen output</a>.</p>
3488 <!-- _______________________________________________________________________ -->
3489 <div class="doc_subsubsection">
3490 <a name="s_Instruction">Important Subclasses of the <tt>Instruction</tt>
3493 <div class="doc_text">
3495 <li><tt><a name="BinaryOperator">BinaryOperator</a></tt>
3496 <p>This subclasses represents all two operand instructions whose operands
3497 must be the same type, except for the comparison instructions.</p></li>
3498 <li><tt><a name="CastInst">CastInst</a></tt>
3499 <p>This subclass is the parent of the 12 casting instructions. It provides
3500 common operations on cast instructions.</p>
3501 <li><tt><a name="CmpInst">CmpInst</a></tt>
3502 <p>This subclass respresents the two comparison instructions,
3503 <a href="LangRef.html#i_icmp">ICmpInst</a> (integer opreands), and
3504 <a href="LangRef.html#i_fcmp">FCmpInst</a> (floating point operands).</p>
3505 <li><tt><a name="TerminatorInst">TerminatorInst</a></tt>
3506 <p>This subclass is the parent of all terminator instructions (those which
3507 can terminate a block).</p>
3511 <!-- _______________________________________________________________________ -->
3512 <div class="doc_subsubsection">
3513 <a name="m_Instruction">Important Public Members of the <tt>Instruction</tt>
3517 <div class="doc_text">
3520 <li><tt><a href="#BasicBlock">BasicBlock</a> *getParent()</tt>
3521 <p>Returns the <a href="#BasicBlock"><tt>BasicBlock</tt></a> that
3522 this <tt>Instruction</tt> is embedded into.</p></li>
3523 <li><tt>bool mayWriteToMemory()</tt>
3524 <p>Returns true if the instruction writes to memory, i.e. it is a
3525 <tt>call</tt>,<tt>free</tt>,<tt>invoke</tt>, or <tt>store</tt>.</p></li>
3526 <li><tt>unsigned getOpcode()</tt>
3527 <p>Returns the opcode for the <tt>Instruction</tt>.</p></li>
3528 <li><tt><a href="#Instruction">Instruction</a> *clone() const</tt>
3529 <p>Returns another instance of the specified instruction, identical
3530 in all ways to the original except that the instruction has no parent
3531 (ie it's not embedded into a <a href="#BasicBlock"><tt>BasicBlock</tt></a>),
3532 and it has no name</p></li>
3537 <!-- ======================================================================= -->
3538 <div class="doc_subsection">
3539 <a name="Constant">The <tt>Constant</tt> class and subclasses</a>
3542 <div class="doc_text">
3544 <p>Constant represents a base class for different types of constants. It
3545 is subclassed by ConstantInt, ConstantArray, etc. for representing
3546 the various types of Constants. <a href="#GlobalValue">GlobalValue</a> is also
3547 a subclass, which represents the address of a global variable or function.
3552 <!-- _______________________________________________________________________ -->
3553 <div class="doc_subsubsection">Important Subclasses of Constant </div>
3554 <div class="doc_text">
3556 <li>ConstantInt : This subclass of Constant represents an integer constant of
3559 <li><tt>const APInt& getValue() const</tt>: Returns the underlying
3560 value of this constant, an APInt value.</li>
3561 <li><tt>int64_t getSExtValue() const</tt>: Converts the underlying APInt
3562 value to an int64_t via sign extension. If the value (not the bit width)
3563 of the APInt is too large to fit in an int64_t, an assertion will result.
3564 For this reason, use of this method is discouraged.</li>
3565 <li><tt>uint64_t getZExtValue() const</tt>: Converts the underlying APInt
3566 value to a uint64_t via zero extension. IF the value (not the bit width)
3567 of the APInt is too large to fit in a uint64_t, an assertion will result.
3568 For this reason, use of this method is discouraged.</li>
3569 <li><tt>static ConstantInt* get(const APInt& Val)</tt>: Returns the
3570 ConstantInt object that represents the value provided by <tt>Val</tt>.
3571 The type is implied as the IntegerType that corresponds to the bit width
3572 of <tt>Val</tt>.</li>
3573 <li><tt>static ConstantInt* get(const Type *Ty, uint64_t Val)</tt>:
3574 Returns the ConstantInt object that represents the value provided by
3575 <tt>Val</tt> for integer type <tt>Ty</tt>.</li>
3578 <li>ConstantFP : This class represents a floating point constant.
3580 <li><tt>double getValue() const</tt>: Returns the underlying value of
3581 this constant. </li>
3584 <li>ConstantArray : This represents a constant array.
3586 <li><tt>const std::vector<Use> &getValues() const</tt>: Returns
3587 a vector of component constants that makeup this array. </li>
3590 <li>ConstantStruct : This represents a constant struct.
3592 <li><tt>const std::vector<Use> &getValues() const</tt>: Returns
3593 a vector of component constants that makeup this array. </li>
3596 <li>GlobalValue : This represents either a global variable or a function. In
3597 either case, the value is a constant fixed address (after linking).
3603 <!-- ======================================================================= -->
3604 <div class="doc_subsection">
3605 <a name="GlobalValue">The <tt>GlobalValue</tt> class</a>
3608 <div class="doc_text">
3611 href="/doxygen/GlobalValue_8h-source.html">llvm/GlobalValue.h</a>"</tt><br>
3612 doxygen info: <a href="/doxygen/classllvm_1_1GlobalValue.html">GlobalValue
3614 Superclasses: <a href="#Constant"><tt>Constant</tt></a>,
3615 <a href="#User"><tt>User</tt></a>, <a href="#Value"><tt>Value</tt></a></p>
3617 <p>Global values (<a href="#GlobalVariable"><tt>GlobalVariable</tt></a>s or <a
3618 href="#Function"><tt>Function</tt></a>s) are the only LLVM values that are
3619 visible in the bodies of all <a href="#Function"><tt>Function</tt></a>s.
3620 Because they are visible at global scope, they are also subject to linking with
3621 other globals defined in different translation units. To control the linking
3622 process, <tt>GlobalValue</tt>s know their linkage rules. Specifically,
3623 <tt>GlobalValue</tt>s know whether they have internal or external linkage, as
3624 defined by the <tt>LinkageTypes</tt> enumeration.</p>
3626 <p>If a <tt>GlobalValue</tt> has internal linkage (equivalent to being
3627 <tt>static</tt> in C), it is not visible to code outside the current translation
3628 unit, and does not participate in linking. If it has external linkage, it is
3629 visible to external code, and does participate in linking. In addition to
3630 linkage information, <tt>GlobalValue</tt>s keep track of which <a
3631 href="#Module"><tt>Module</tt></a> they are currently part of.</p>
3633 <p>Because <tt>GlobalValue</tt>s are memory objects, they are always referred to
3634 by their <b>address</b>. As such, the <a href="#Type"><tt>Type</tt></a> of a
3635 global is always a pointer to its contents. It is important to remember this
3636 when using the <tt>GetElementPtrInst</tt> instruction because this pointer must
3637 be dereferenced first. For example, if you have a <tt>GlobalVariable</tt> (a
3638 subclass of <tt>GlobalValue)</tt> that is an array of 24 ints, type <tt>[24 x
3639 i32]</tt>, then the <tt>GlobalVariable</tt> is a pointer to that array. Although
3640 the address of the first element of this array and the value of the
3641 <tt>GlobalVariable</tt> are the same, they have different types. The
3642 <tt>GlobalVariable</tt>'s type is <tt>[24 x i32]</tt>. The first element's type
3643 is <tt>i32.</tt> Because of this, accessing a global value requires you to
3644 dereference the pointer with <tt>GetElementPtrInst</tt> first, then its elements
3645 can be accessed. This is explained in the <a href="LangRef.html#globalvars">LLVM
3646 Language Reference Manual</a>.</p>
3650 <!-- _______________________________________________________________________ -->
3651 <div class="doc_subsubsection">
3652 <a name="m_GlobalValue">Important Public Members of the <tt>GlobalValue</tt>
3656 <div class="doc_text">
3659 <li><tt>bool hasInternalLinkage() const</tt><br>
3660 <tt>bool hasExternalLinkage() const</tt><br>
3661 <tt>void setInternalLinkage(bool HasInternalLinkage)</tt>
3662 <p> These methods manipulate the linkage characteristics of the <tt>GlobalValue</tt>.</p>
3665 <li><tt><a href="#Module">Module</a> *getParent()</tt>
3666 <p> This returns the <a href="#Module"><tt>Module</tt></a> that the
3667 GlobalValue is currently embedded into.</p></li>
3672 <!-- ======================================================================= -->
3673 <div class="doc_subsection">
3674 <a name="Function">The <tt>Function</tt> class</a>
3677 <div class="doc_text">
3680 href="/doxygen/Function_8h-source.html">llvm/Function.h</a>"</tt><br> doxygen
3681 info: <a href="/doxygen/classllvm_1_1Function.html">Function Class</a><br>
3682 Superclasses: <a href="#GlobalValue"><tt>GlobalValue</tt></a>,
3683 <a href="#Constant"><tt>Constant</tt></a>,
3684 <a href="#User"><tt>User</tt></a>,
3685 <a href="#Value"><tt>Value</tt></a></p>
3687 <p>The <tt>Function</tt> class represents a single procedure in LLVM. It is
3688 actually one of the more complex classes in the LLVM hierarchy because it must
3689 keep track of a large amount of data. The <tt>Function</tt> class keeps track
3690 of a list of <a href="#BasicBlock"><tt>BasicBlock</tt></a>s, a list of formal
3691 <a href="#Argument"><tt>Argument</tt></a>s, and a
3692 <a href="#SymbolTable"><tt>SymbolTable</tt></a>.</p>
3694 <p>The list of <a href="#BasicBlock"><tt>BasicBlock</tt></a>s is the most
3695 commonly used part of <tt>Function</tt> objects. The list imposes an implicit
3696 ordering of the blocks in the function, which indicate how the code will be
3697 laid out by the backend. Additionally, the first <a
3698 href="#BasicBlock"><tt>BasicBlock</tt></a> is the implicit entry node for the
3699 <tt>Function</tt>. It is not legal in LLVM to explicitly branch to this initial
3700 block. There are no implicit exit nodes, and in fact there may be multiple exit
3701 nodes from a single <tt>Function</tt>. If the <a
3702 href="#BasicBlock"><tt>BasicBlock</tt></a> list is empty, this indicates that
3703 the <tt>Function</tt> is actually a function declaration: the actual body of the
3704 function hasn't been linked in yet.</p>
3706 <p>In addition to a list of <a href="#BasicBlock"><tt>BasicBlock</tt></a>s, the
3707 <tt>Function</tt> class also keeps track of the list of formal <a
3708 href="#Argument"><tt>Argument</tt></a>s that the function receives. This
3709 container manages the lifetime of the <a href="#Argument"><tt>Argument</tt></a>
3710 nodes, just like the <a href="#BasicBlock"><tt>BasicBlock</tt></a> list does for
3711 the <a href="#BasicBlock"><tt>BasicBlock</tt></a>s.</p>
3713 <p>The <a href="#SymbolTable"><tt>SymbolTable</tt></a> is a very rarely used
3714 LLVM feature that is only used when you have to look up a value by name. Aside
3715 from that, the <a href="#SymbolTable"><tt>SymbolTable</tt></a> is used
3716 internally to make sure that there are not conflicts between the names of <a
3717 href="#Instruction"><tt>Instruction</tt></a>s, <a
3718 href="#BasicBlock"><tt>BasicBlock</tt></a>s, or <a
3719 href="#Argument"><tt>Argument</tt></a>s in the function body.</p>
3721 <p>Note that <tt>Function</tt> is a <a href="#GlobalValue">GlobalValue</a>
3722 and therefore also a <a href="#Constant">Constant</a>. The value of the function
3723 is its address (after linking) which is guaranteed to be constant.</p>
3726 <!-- _______________________________________________________________________ -->
3727 <div class="doc_subsubsection">
3728 <a name="m_Function">Important Public Members of the <tt>Function</tt>
3732 <div class="doc_text">
3735 <li><tt>Function(const </tt><tt><a href="#FunctionType">FunctionType</a>
3736 *Ty, LinkageTypes Linkage, const std::string &N = "", Module* Parent = 0)</tt>
3738 <p>Constructor used when you need to create new <tt>Function</tt>s to add
3739 the the program. The constructor must specify the type of the function to
3740 create and what type of linkage the function should have. The <a
3741 href="#FunctionType"><tt>FunctionType</tt></a> argument
3742 specifies the formal arguments and return value for the function. The same
3743 <a href="#FunctionType"><tt>FunctionType</tt></a> value can be used to
3744 create multiple functions. The <tt>Parent</tt> argument specifies the Module
3745 in which the function is defined. If this argument is provided, the function
3746 will automatically be inserted into that module's list of
3749 <li><tt>bool isDeclaration()</tt>
3751 <p>Return whether or not the <tt>Function</tt> has a body defined. If the
3752 function is "external", it does not have a body, and thus must be resolved
3753 by linking with a function defined in a different translation unit.</p></li>
3755 <li><tt>Function::iterator</tt> - Typedef for basic block list iterator<br>
3756 <tt>Function::const_iterator</tt> - Typedef for const_iterator.<br>
3758 <tt>begin()</tt>, <tt>end()</tt>
3759 <tt>size()</tt>, <tt>empty()</tt>
3761 <p>These are forwarding methods that make it easy to access the contents of
3762 a <tt>Function</tt> object's <a href="#BasicBlock"><tt>BasicBlock</tt></a>
3765 <li><tt>Function::BasicBlockListType &getBasicBlockList()</tt>
3767 <p>Returns the list of <a href="#BasicBlock"><tt>BasicBlock</tt></a>s. This
3768 is necessary to use when you need to update the list or perform a complex
3769 action that doesn't have a forwarding method.</p></li>
3771 <li><tt>Function::arg_iterator</tt> - Typedef for the argument list
3773 <tt>Function::const_arg_iterator</tt> - Typedef for const_iterator.<br>
3775 <tt>arg_begin()</tt>, <tt>arg_end()</tt>
3776 <tt>arg_size()</tt>, <tt>arg_empty()</tt>
3778 <p>These are forwarding methods that make it easy to access the contents of
3779 a <tt>Function</tt> object's <a href="#Argument"><tt>Argument</tt></a>
3782 <li><tt>Function::ArgumentListType &getArgumentList()</tt>
3784 <p>Returns the list of <a href="#Argument"><tt>Argument</tt></a>s. This is
3785 necessary to use when you need to update the list or perform a complex
3786 action that doesn't have a forwarding method.</p></li>
3788 <li><tt><a href="#BasicBlock">BasicBlock</a> &getEntryBlock()</tt>
3790 <p>Returns the entry <a href="#BasicBlock"><tt>BasicBlock</tt></a> for the
3791 function. Because the entry block for the function is always the first
3792 block, this returns the first block of the <tt>Function</tt>.</p></li>
3794 <li><tt><a href="#Type">Type</a> *getReturnType()</tt><br>
3795 <tt><a href="#FunctionType">FunctionType</a> *getFunctionType()</tt>
3797 <p>This traverses the <a href="#Type"><tt>Type</tt></a> of the
3798 <tt>Function</tt> and returns the return type of the function, or the <a
3799 href="#FunctionType"><tt>FunctionType</tt></a> of the actual
3802 <li><tt><a href="#SymbolTable">SymbolTable</a> *getSymbolTable()</tt>
3804 <p> Return a pointer to the <a href="#SymbolTable"><tt>SymbolTable</tt></a>
3805 for this <tt>Function</tt>.</p></li>
3810 <!-- ======================================================================= -->
3811 <div class="doc_subsection">
3812 <a name="GlobalVariable">The <tt>GlobalVariable</tt> class</a>
3815 <div class="doc_text">
3818 href="/doxygen/GlobalVariable_8h-source.html">llvm/GlobalVariable.h</a>"</tt>
3820 doxygen info: <a href="/doxygen/classllvm_1_1GlobalVariable.html">GlobalVariable
3822 Superclasses: <a href="#GlobalValue"><tt>GlobalValue</tt></a>,
3823 <a href="#Constant"><tt>Constant</tt></a>,
3824 <a href="#User"><tt>User</tt></a>,
3825 <a href="#Value"><tt>Value</tt></a></p>
3827 <p>Global variables are represented with the (surprise surprise)
3828 <tt>GlobalVariable</tt> class. Like functions, <tt>GlobalVariable</tt>s are also
3829 subclasses of <a href="#GlobalValue"><tt>GlobalValue</tt></a>, and as such are
3830 always referenced by their address (global values must live in memory, so their
3831 "name" refers to their constant address). See
3832 <a href="#GlobalValue"><tt>GlobalValue</tt></a> for more on this. Global
3833 variables may have an initial value (which must be a
3834 <a href="#Constant"><tt>Constant</tt></a>), and if they have an initializer,
3835 they may be marked as "constant" themselves (indicating that their contents
3836 never change at runtime).</p>
3839 <!-- _______________________________________________________________________ -->
3840 <div class="doc_subsubsection">
3841 <a name="m_GlobalVariable">Important Public Members of the
3842 <tt>GlobalVariable</tt> class</a>
3845 <div class="doc_text">
3848 <li><tt>GlobalVariable(const </tt><tt><a href="#Type">Type</a> *Ty, bool
3849 isConstant, LinkageTypes& Linkage, <a href="#Constant">Constant</a>
3850 *Initializer = 0, const std::string &Name = "", Module* Parent = 0)</tt>
3852 <p>Create a new global variable of the specified type. If
3853 <tt>isConstant</tt> is true then the global variable will be marked as
3854 unchanging for the program. The Linkage parameter specifies the type of
3855 linkage (internal, external, weak, linkonce, appending) for the variable.
3856 If the linkage is InternalLinkage, WeakAnyLinkage, WeakODRLinkage,
3857 LinkOnceAnyLinkage or LinkOnceODRLinkage, then the resultant
3858 global variable will have internal linkage. AppendingLinkage concatenates
3859 together all instances (in different translation units) of the variable
3860 into a single variable but is only applicable to arrays. See
3861 the <a href="LangRef.html#modulestructure">LLVM Language Reference</a> for
3862 further details on linkage types. Optionally an initializer, a name, and the
3863 module to put the variable into may be specified for the global variable as
3866 <li><tt>bool isConstant() const</tt>
3868 <p>Returns true if this is a global variable that is known not to
3869 be modified at runtime.</p></li>
3871 <li><tt>bool hasInitializer()</tt>
3873 <p>Returns true if this <tt>GlobalVariable</tt> has an intializer.</p></li>
3875 <li><tt><a href="#Constant">Constant</a> *getInitializer()</tt>
3877 <p>Returns the initial value for a <tt>GlobalVariable</tt>. It is not legal
3878 to call this method if there is no initializer.</p></li>
3884 <!-- ======================================================================= -->
3885 <div class="doc_subsection">
3886 <a name="BasicBlock">The <tt>BasicBlock</tt> class</a>
3889 <div class="doc_text">
3892 href="/doxygen/BasicBlock_8h-source.html">llvm/BasicBlock.h</a>"</tt><br>
3893 doxygen info: <a href="/doxygen/classllvm_1_1BasicBlock.html">BasicBlock
3895 Superclass: <a href="#Value"><tt>Value</tt></a></p>
3897 <p>This class represents a single entry single exit section of the code,
3898 commonly known as a basic block by the compiler community. The
3899 <tt>BasicBlock</tt> class maintains a list of <a
3900 href="#Instruction"><tt>Instruction</tt></a>s, which form the body of the block.
3901 Matching the language definition, the last element of this list of instructions
3902 is always a terminator instruction (a subclass of the <a
3903 href="#TerminatorInst"><tt>TerminatorInst</tt></a> class).</p>
3905 <p>In addition to tracking the list of instructions that make up the block, the
3906 <tt>BasicBlock</tt> class also keeps track of the <a
3907 href="#Function"><tt>Function</tt></a> that it is embedded into.</p>
3909 <p>Note that <tt>BasicBlock</tt>s themselves are <a
3910 href="#Value"><tt>Value</tt></a>s, because they are referenced by instructions
3911 like branches and can go in the switch tables. <tt>BasicBlock</tt>s have type
3916 <!-- _______________________________________________________________________ -->
3917 <div class="doc_subsubsection">
3918 <a name="m_BasicBlock">Important Public Members of the <tt>BasicBlock</tt>
3922 <div class="doc_text">
3925 <li><tt>BasicBlock(const std::string &Name = "", </tt><tt><a
3926 href="#Function">Function</a> *Parent = 0)</tt>
3928 <p>The <tt>BasicBlock</tt> constructor is used to create new basic blocks for
3929 insertion into a function. The constructor optionally takes a name for the new
3930 block, and a <a href="#Function"><tt>Function</tt></a> to insert it into. If
3931 the <tt>Parent</tt> parameter is specified, the new <tt>BasicBlock</tt> is
3932 automatically inserted at the end of the specified <a
3933 href="#Function"><tt>Function</tt></a>, if not specified, the BasicBlock must be
3934 manually inserted into the <a href="#Function"><tt>Function</tt></a>.</p></li>
3936 <li><tt>BasicBlock::iterator</tt> - Typedef for instruction list iterator<br>
3937 <tt>BasicBlock::const_iterator</tt> - Typedef for const_iterator.<br>
3938 <tt>begin()</tt>, <tt>end()</tt>, <tt>front()</tt>, <tt>back()</tt>,
3939 <tt>size()</tt>, <tt>empty()</tt>
3940 STL-style functions for accessing the instruction list.
3942 <p>These methods and typedefs are forwarding functions that have the same
3943 semantics as the standard library methods of the same names. These methods
3944 expose the underlying instruction list of a basic block in a way that is easy to
3945 manipulate. To get the full complement of container operations (including
3946 operations to update the list), you must use the <tt>getInstList()</tt>
3949 <li><tt>BasicBlock::InstListType &getInstList()</tt>
3951 <p>This method is used to get access to the underlying container that actually
3952 holds the Instructions. This method must be used when there isn't a forwarding
3953 function in the <tt>BasicBlock</tt> class for the operation that you would like
3954 to perform. Because there are no forwarding functions for "updating"
3955 operations, you need to use this if you want to update the contents of a
3956 <tt>BasicBlock</tt>.</p></li>
3958 <li><tt><a href="#Function">Function</a> *getParent()</tt>
3960 <p> Returns a pointer to <a href="#Function"><tt>Function</tt></a> the block is
3961 embedded into, or a null pointer if it is homeless.</p></li>
3963 <li><tt><a href="#TerminatorInst">TerminatorInst</a> *getTerminator()</tt>
3965 <p> Returns a pointer to the terminator instruction that appears at the end of
3966 the <tt>BasicBlock</tt>. If there is no terminator instruction, or if the last
3967 instruction in the block is not a terminator, then a null pointer is
3975 <!-- ======================================================================= -->
3976 <div class="doc_subsection">
3977 <a name="Argument">The <tt>Argument</tt> class</a>
3980 <div class="doc_text">
3982 <p>This subclass of Value defines the interface for incoming formal
3983 arguments to a function. A Function maintains a list of its formal
3984 arguments. An argument has a pointer to the parent Function.</p>
3988 <!-- *********************************************************************** -->
3991 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
3992 src="http://jigsaw.w3.org/css-validator/images/vcss-blue" alt="Valid CSS"></a>
3993 <a href="http://validator.w3.org/check/referer"><img
3994 src="http://www.w3.org/Icons/valid-html401" alt="Valid HTML 4.01 Strict"></a>
3996 <a href="mailto:dhurjati@cs.uiuc.edu">Dinakar Dhurjati</a> and
3997 <a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
3998 <a href="http://llvm.org/">The LLVM Compiler Infrastructure</a><br>
3999 Last modified: $Date$