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">
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_tinyptrvector">"llvm/ADT/TinyPtrVector.h"</a></li>
63 <li><a href="#dss_smallvector">"llvm/ADT/SmallVector.h"</a></li>
64 <li><a href="#dss_vector"><vector></a></li>
65 <li><a href="#dss_deque"><deque></a></li>
66 <li><a href="#dss_list"><list></a></li>
67 <li><a href="#dss_ilist">llvm/ADT/ilist.h</a></li>
68 <li><a href="#dss_packedvector">llvm/ADT/PackedVector.h</a></li>
69 <li><a href="#dss_other">Other Sequential Container Options</a></li>
71 <li><a href="#ds_string">String-like containers</a>
73 <li><a href="#dss_stringref">llvm/ADT/StringRef.h</a></li>
74 <li><a href="#dss_twine">llvm/ADT/Twine.h</a></li>
75 <li><a href="#dss_smallstring">llvm/ADT/SmallString.h</a></li>
76 <li><a href="#dss_stdstring">std::string</a></li>
78 <li><a href="#ds_set">Set-Like Containers (std::set, SmallSet, SetVector, etc)</a>
80 <li><a href="#dss_sortedvectorset">A sorted 'vector'</a></li>
81 <li><a href="#dss_smallset">"llvm/ADT/SmallSet.h"</a></li>
82 <li><a href="#dss_smallptrset">"llvm/ADT/SmallPtrSet.h"</a></li>
83 <li><a href="#dss_denseset">"llvm/ADT/DenseSet.h"</a></li>
84 <li><a href="#dss_FoldingSet">"llvm/ADT/FoldingSet.h"</a></li>
85 <li><a href="#dss_set"><set></a></li>
86 <li><a href="#dss_setvector">"llvm/ADT/SetVector.h"</a></li>
87 <li><a href="#dss_uniquevector">"llvm/ADT/UniqueVector.h"</a></li>
88 <li><a href="#dss_immutableset">"llvm/ADT/ImmutableSet.h"</a></li>
89 <li><a href="#dss_otherset">Other Set-Like Container Options</a></li>
91 <li><a href="#ds_map">Map-Like Containers (std::map, DenseMap, etc)</a>
93 <li><a href="#dss_sortedvectormap">A sorted 'vector'</a></li>
94 <li><a href="#dss_stringmap">"llvm/ADT/StringMap.h"</a></li>
95 <li><a href="#dss_indexedmap">"llvm/ADT/IndexedMap.h"</a></li>
96 <li><a href="#dss_densemap">"llvm/ADT/DenseMap.h"</a></li>
97 <li><a href="#dss_valuemap">"llvm/ADT/ValueMap.h"</a></li>
98 <li><a href="#dss_intervalmap">"llvm/ADT/IntervalMap.h"</a></li>
99 <li><a href="#dss_map"><map></a></li>
100 <li><a href="#dss_inteqclasses">"llvm/ADT/IntEqClasses.h"</a></li>
101 <li><a href="#dss_immutablemap">"llvm/ADT/ImmutableMap.h"</a></li>
102 <li><a href="#dss_othermap">Other Map-Like Container Options</a></li>
104 <li><a href="#ds_bit">BitVector-like containers</a>
106 <li><a href="#dss_bitvector">A dense bitvector</a></li>
107 <li><a href="#dss_smallbitvector">A "small" dense bitvector</a></li>
108 <li><a href="#dss_sparsebitvector">A sparse bitvector</a></li>
112 <li><a href="#common">Helpful Hints for Common Operations</a>
114 <li><a href="#inspection">Basic Inspection and Traversal Routines</a>
116 <li><a href="#iterate_function">Iterating over the <tt>BasicBlock</tt>s
117 in a <tt>Function</tt></a> </li>
118 <li><a href="#iterate_basicblock">Iterating over the <tt>Instruction</tt>s
119 in a <tt>BasicBlock</tt></a> </li>
120 <li><a href="#iterate_institer">Iterating over the <tt>Instruction</tt>s
121 in a <tt>Function</tt></a> </li>
122 <li><a href="#iterate_convert">Turning an iterator into a
123 class pointer</a> </li>
124 <li><a href="#iterate_complex">Finding call sites: a more
125 complex example</a> </li>
126 <li><a href="#calls_and_invokes">Treating calls and invokes
127 the same way</a> </li>
128 <li><a href="#iterate_chains">Iterating over def-use &
129 use-def chains</a> </li>
130 <li><a href="#iterate_preds">Iterating over predecessors &
131 successors of blocks</a></li>
134 <li><a href="#simplechanges">Making simple changes</a>
136 <li><a href="#schanges_creating">Creating and inserting new
137 <tt>Instruction</tt>s</a> </li>
138 <li><a href="#schanges_deleting">Deleting <tt>Instruction</tt>s</a> </li>
139 <li><a href="#schanges_replacing">Replacing an <tt>Instruction</tt>
140 with another <tt>Value</tt></a> </li>
141 <li><a href="#schanges_deletingGV">Deleting <tt>GlobalVariable</tt>s</a> </li>
144 <li><a href="#create_types">How to Create Types</a></li>
146 <li>Working with the Control Flow Graph
148 <li>Accessing predecessors and successors of a <tt>BasicBlock</tt>
156 <li><a href="#threading">Threads and LLVM</a>
158 <li><a href="#startmultithreaded">Entering and Exiting Multithreaded Mode
160 <li><a href="#shutdown">Ending execution with <tt>llvm_shutdown()</tt></a></li>
161 <li><a href="#managedstatic">Lazy initialization with <tt>ManagedStatic</tt></a></li>
162 <li><a href="#llvmcontext">Achieving Isolation with <tt>LLVMContext</tt></a></li>
163 <li><a href="#jitthreading">Threads and the JIT</a></li>
167 <li><a href="#advanced">Advanced Topics</a>
170 <li><a href="#SymbolTable">The <tt>ValueSymbolTable</tt> class</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 <!-- *********************************************************************** -->
214 <a name="introduction">Introduction </a>
216 <!-- *********************************************************************** -->
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 <!-- *********************************************************************** -->
246 <a name="general">General Information</a>
248 <!-- *********************************************************************** -->
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>
255 <!-- ======================================================================= -->
257 <a name="stl">The C++ Standard Template Library</a>
262 <p>LLVM makes heavy use of the C++ Standard Template Library (STL),
263 perhaps much more than you are used to, or have seen before. Because of
264 this, you might want to do a little background reading in the
265 techniques used and capabilities of the library. There are many good
266 pages that discuss the STL, and several books on the subject that you
267 can get, so it will not be discussed in this document.</p>
269 <p>Here are some useful links:</p>
273 <li><a href="http://www.dinkumware.com/manuals/#Standard C++ Library">Dinkumware
274 C++ Library reference</a> - an excellent reference for the STL and other parts
275 of the standard C++ library.</li>
277 <li><a href="http://www.tempest-sw.com/cpp/">C++ In a Nutshell</a> - This is an
278 O'Reilly book in the making. It has a decent Standard Library
279 Reference that rivals Dinkumware's, and is unfortunately no longer free since the
280 book has been published.</li>
282 <li><a href="http://www.parashift.com/c++-faq-lite/">C++ Frequently Asked
285 <li><a href="http://www.sgi.com/tech/stl/">SGI's STL Programmer's Guide</a> -
287 href="http://www.sgi.com/tech/stl/stl_introduction.html">Introduction to the
290 <li><a href="http://www.research.att.com/%7Ebs/C++.html">Bjarne Stroustrup's C++
293 <li><a href="http://64.78.49.204/">
294 Bruce Eckel's Thinking in C++, 2nd ed. Volume 2 Revision 4.0 (even better, get
299 <p>You are also encouraged to take a look at the <a
300 href="CodingStandards.html">LLVM Coding Standards</a> guide which focuses on how
301 to write maintainable code more than where to put your curly braces.</p>
305 <!-- ======================================================================= -->
307 <a name="stl">Other useful references</a>
313 <li><a href="http://www.fortran-2000.com/ArnaudRecipes/sharedlib.html">Using
314 static and shared libraries across platforms</a></li>
321 <!-- *********************************************************************** -->
323 <a name="apis">Important and useful LLVM APIs</a>
325 <!-- *********************************************************************** -->
329 <p>Here we highlight some LLVM APIs that are generally useful and good to
330 know about when writing transformations.</p>
332 <!-- ======================================================================= -->
334 <a name="isa">The <tt>isa<></tt>, <tt>cast<></tt> and
335 <tt>dyn_cast<></tt> templates</a>
340 <p>The LLVM source-base makes extensive use of a custom form of RTTI.
341 These templates have many similarities to the C++ <tt>dynamic_cast<></tt>
342 operator, but they don't have some drawbacks (primarily stemming from
343 the fact that <tt>dynamic_cast<></tt> only works on classes that
344 have a v-table). Because they are used so often, you must know what they
345 do and how they work. All of these templates are defined in the <a
346 href="/doxygen/Casting_8h-source.html"><tt>llvm/Support/Casting.h</tt></a>
347 file (note that you very rarely have to include this file directly).</p>
350 <dt><tt>isa<></tt>: </dt>
352 <dd><p>The <tt>isa<></tt> operator works exactly like the Java
353 "<tt>instanceof</tt>" operator. It returns true or false depending on whether
354 a reference or pointer points to an instance of the specified class. This can
355 be very useful for constraint checking of various sorts (example below).</p>
358 <dt><tt>cast<></tt>: </dt>
360 <dd><p>The <tt>cast<></tt> operator is a "checked cast" operation. It
361 converts a pointer or reference from a base class to a derived class, causing
362 an assertion failure if it is not really an instance of the right type. This
363 should be used in cases where you have some information that makes you believe
364 that something is of the right type. An example of the <tt>isa<></tt>
365 and <tt>cast<></tt> template is:</p>
367 <div class="doc_code">
369 static bool isLoopInvariant(const <a href="#Value">Value</a> *V, const Loop *L) {
370 if (isa<<a href="#Constant">Constant</a>>(V) || isa<<a href="#Argument">Argument</a>>(V) || isa<<a href="#GlobalValue">GlobalValue</a>>(V))
373 // <i>Otherwise, it must be an instruction...</i>
374 return !L->contains(cast<<a href="#Instruction">Instruction</a>>(V)->getParent());
379 <p>Note that you should <b>not</b> use an <tt>isa<></tt> test followed
380 by a <tt>cast<></tt>, for that use the <tt>dyn_cast<></tt>
385 <dt><tt>dyn_cast<></tt>:</dt>
387 <dd><p>The <tt>dyn_cast<></tt> operator is a "checking cast" operation.
388 It checks to see if the operand is of the specified type, and if so, returns a
389 pointer to it (this operator does not work with references). If the operand is
390 not of the correct type, a null pointer is returned. Thus, this works very
391 much like the <tt>dynamic_cast<></tt> operator in C++, and should be
392 used in the same circumstances. Typically, the <tt>dyn_cast<></tt>
393 operator is used in an <tt>if</tt> statement or some other flow control
394 statement like this:</p>
396 <div class="doc_code">
398 if (<a href="#AllocationInst">AllocationInst</a> *AI = dyn_cast<<a href="#AllocationInst">AllocationInst</a>>(Val)) {
404 <p>This form of the <tt>if</tt> statement effectively combines together a call
405 to <tt>isa<></tt> and a call to <tt>cast<></tt> into one
406 statement, which is very convenient.</p>
408 <p>Note that the <tt>dyn_cast<></tt> operator, like C++'s
409 <tt>dynamic_cast<></tt> or Java's <tt>instanceof</tt> operator, can be
410 abused. In particular, you should not use big chained <tt>if/then/else</tt>
411 blocks to check for lots of different variants of classes. If you find
412 yourself wanting to do this, it is much cleaner and more efficient to use the
413 <tt>InstVisitor</tt> class to dispatch over the instruction type directly.</p>
417 <dt><tt>cast_or_null<></tt>: </dt>
419 <dd><p>The <tt>cast_or_null<></tt> operator works just like the
420 <tt>cast<></tt> operator, except that it allows for a null pointer as an
421 argument (which it then propagates). This can sometimes be useful, allowing
422 you to combine several null checks into one.</p></dd>
424 <dt><tt>dyn_cast_or_null<></tt>: </dt>
426 <dd><p>The <tt>dyn_cast_or_null<></tt> operator works just like the
427 <tt>dyn_cast<></tt> operator, except that it allows for a null pointer
428 as an argument (which it then propagates). This can sometimes be useful,
429 allowing you to combine several null checks into one.</p></dd>
433 <p>These five templates can be used with any classes, whether they have a
434 v-table or not. To add support for these templates, you simply need to add
435 <tt>classof</tt> static methods to the class you are interested casting
436 to. Describing this is currently outside the scope of this document, but there
437 are lots of examples in the LLVM source base.</p>
442 <!-- ======================================================================= -->
444 <a name="string_apis">Passing strings (the <tt>StringRef</tt>
445 and <tt>Twine</tt> classes)</a>
450 <p>Although LLVM generally does not do much string manipulation, we do have
451 several important APIs which take strings. Two important examples are the
452 Value class -- which has names for instructions, functions, etc. -- and the
453 StringMap class which is used extensively in LLVM and Clang.</p>
455 <p>These are generic classes, and they need to be able to accept strings which
456 may have embedded null characters. Therefore, they cannot simply take
457 a <tt>const char *</tt>, and taking a <tt>const std::string&</tt> requires
458 clients to perform a heap allocation which is usually unnecessary. Instead,
459 many LLVM APIs use a <tt>StringRef</tt> or a <tt>const Twine&</tt> for
460 passing strings efficiently.</p>
462 <!-- _______________________________________________________________________ -->
464 <a name="StringRef">The <tt>StringRef</tt> class</a>
469 <p>The <tt>StringRef</tt> data type represents a reference to a constant string
470 (a character array and a length) and supports the common operations available
471 on <tt>std:string</tt>, but does not require heap allocation.</p>
473 <p>It can be implicitly constructed using a C style null-terminated string,
474 an <tt>std::string</tt>, or explicitly with a character pointer and length.
475 For example, the <tt>StringRef</tt> find function is declared as:</p>
477 <pre class="doc_code">
478 iterator find(StringRef Key);
481 <p>and clients can call it using any one of:</p>
483 <pre class="doc_code">
484 Map.find("foo"); <i>// Lookup "foo"</i>
485 Map.find(std::string("bar")); <i>// Lookup "bar"</i>
486 Map.find(StringRef("\0baz", 4)); <i>// Lookup "\0baz"</i>
489 <p>Similarly, APIs which need to return a string may return a <tt>StringRef</tt>
490 instance, which can be used directly or converted to an <tt>std::string</tt>
491 using the <tt>str</tt> member function. See
492 "<tt><a href="/doxygen/classllvm_1_1StringRef_8h-source.html">llvm/ADT/StringRef.h</a></tt>"
493 for more information.</p>
495 <p>You should rarely use the <tt>StringRef</tt> class directly, because it contains
496 pointers to external memory it is not generally safe to store an instance of the
497 class (unless you know that the external storage will not be freed). StringRef is
498 small and pervasive enough in LLVM that it should always be passed by value.</p>
502 <!-- _______________________________________________________________________ -->
504 <a name="Twine">The <tt>Twine</tt> class</a>
509 <p>The <tt>Twine</tt> class is an efficient way for APIs to accept concatenated
510 strings. For example, a common LLVM paradigm is to name one instruction based on
511 the name of another instruction with a suffix, for example:</p>
513 <div class="doc_code">
515 New = CmpInst::Create(<i>...</i>, SO->getName() + ".cmp");
519 <p>The <tt>Twine</tt> class is effectively a
520 lightweight <a href="http://en.wikipedia.org/wiki/Rope_(computer_science)">rope</a>
521 which points to temporary (stack allocated) objects. Twines can be implicitly
522 constructed as the result of the plus operator applied to strings (i.e., a C
523 strings, an <tt>std::string</tt>, or a <tt>StringRef</tt>). The twine delays the
524 actual concatenation of strings until it is actually required, at which point
525 it can be efficiently rendered directly into a character array. This avoids
526 unnecessary heap allocation involved in constructing the temporary results of
527 string concatenation. See
528 "<tt><a href="/doxygen/classllvm_1_1Twine_8h-source.html">llvm/ADT/Twine.h</a></tt>"
529 for more information.</p>
531 <p>As with a <tt>StringRef</tt>, <tt>Twine</tt> objects point to external memory
532 and should almost never be stored or mentioned directly. They are intended
533 solely for use when defining a function which should be able to efficiently
534 accept concatenated strings.</p>
540 <!-- ======================================================================= -->
542 <a name="DEBUG">The <tt>DEBUG()</tt> macro and <tt>-debug</tt> option</a>
547 <p>Often when working on your pass you will put a bunch of debugging printouts
548 and other code into your pass. After you get it working, you want to remove
549 it, but you may need it again in the future (to work out new bugs that you run
552 <p> Naturally, because of this, you don't want to delete the debug printouts,
553 but you don't want them to always be noisy. A standard compromise is to comment
554 them out, allowing you to enable them if you need them in the future.</p>
556 <p>The "<tt><a href="/doxygen/Debug_8h-source.html">llvm/Support/Debug.h</a></tt>"
557 file provides a macro named <tt>DEBUG()</tt> that is a much nicer solution to
558 this problem. Basically, you can put arbitrary code into the argument of the
559 <tt>DEBUG</tt> macro, and it is only executed if '<tt>opt</tt>' (or any other
560 tool) is run with the '<tt>-debug</tt>' command line argument:</p>
562 <div class="doc_code">
564 DEBUG(errs() << "I am here!\n");
568 <p>Then you can run your pass like this:</p>
570 <div class="doc_code">
572 $ opt < a.bc > /dev/null -mypass
573 <i><no output></i>
574 $ opt < a.bc > /dev/null -mypass -debug
579 <p>Using the <tt>DEBUG()</tt> macro instead of a home-brewed solution allows you
580 to not have to create "yet another" command line option for the debug output for
581 your pass. Note that <tt>DEBUG()</tt> macros are disabled for optimized builds,
582 so they do not cause a performance impact at all (for the same reason, they
583 should also not contain side-effects!).</p>
585 <p>One additional nice thing about the <tt>DEBUG()</tt> macro is that you can
586 enable or disable it directly in gdb. Just use "<tt>set DebugFlag=0</tt>" or
587 "<tt>set DebugFlag=1</tt>" from the gdb if the program is running. If the
588 program hasn't been started yet, you can always just run it with
591 <!-- _______________________________________________________________________ -->
593 <a name="DEBUG_TYPE">Fine grained debug info with <tt>DEBUG_TYPE</tt> and
594 the <tt>-debug-only</tt> option</a>
599 <p>Sometimes you may find yourself in a situation where enabling <tt>-debug</tt>
600 just turns on <b>too much</b> information (such as when working on the code
601 generator). If you want to enable debug information with more fine-grained
602 control, you define the <tt>DEBUG_TYPE</tt> macro and the <tt>-debug</tt> only
603 option as follows:</p>
605 <div class="doc_code">
608 DEBUG(errs() << "No debug type\n");
609 #define DEBUG_TYPE "foo"
610 DEBUG(errs() << "'foo' debug type\n");
612 #define DEBUG_TYPE "bar"
613 DEBUG(errs() << "'bar' debug type\n"));
615 #define DEBUG_TYPE ""
616 DEBUG(errs() << "No debug type (2)\n");
620 <p>Then you can run your pass like this:</p>
622 <div class="doc_code">
624 $ opt < a.bc > /dev/null -mypass
625 <i><no output></i>
626 $ opt < a.bc > /dev/null -mypass -debug
631 $ opt < a.bc > /dev/null -mypass -debug-only=foo
633 $ opt < a.bc > /dev/null -mypass -debug-only=bar
638 <p>Of course, in practice, you should only set <tt>DEBUG_TYPE</tt> at the top of
639 a file, to specify the debug type for the entire module (if you do this before
640 you <tt>#include "llvm/Support/Debug.h"</tt>, you don't have to insert the ugly
641 <tt>#undef</tt>'s). Also, you should use names more meaningful than "foo" and
642 "bar", because there is no system in place to ensure that names do not
643 conflict. If two different modules use the same string, they will all be turned
644 on when the name is specified. This allows, for example, all debug information
645 for instruction scheduling to be enabled with <tt>-debug-type=InstrSched</tt>,
646 even if the source lives in multiple files.</p>
648 <p>The <tt>DEBUG_WITH_TYPE</tt> macro is also available for situations where you
649 would like to set <tt>DEBUG_TYPE</tt>, but only for one specific <tt>DEBUG</tt>
650 statement. It takes an additional first parameter, which is the type to use. For
651 example, the preceding example could be written as:</p>
654 <div class="doc_code">
656 DEBUG_WITH_TYPE("", errs() << "No debug type\n");
657 DEBUG_WITH_TYPE("foo", errs() << "'foo' debug type\n");
658 DEBUG_WITH_TYPE("bar", errs() << "'bar' debug type\n"));
659 DEBUG_WITH_TYPE("", errs() << "No debug type (2)\n");
667 <!-- ======================================================================= -->
669 <a name="Statistic">The <tt>Statistic</tt> class & <tt>-stats</tt>
676 href="/doxygen/Statistic_8h-source.html">llvm/ADT/Statistic.h</a></tt>" file
677 provides a class named <tt>Statistic</tt> that is used as a unified way to
678 keep track of what the LLVM compiler is doing and how effective various
679 optimizations are. It is useful to see what optimizations are contributing to
680 making a particular program run faster.</p>
682 <p>Often you may run your pass on some big program, and you're interested to see
683 how many times it makes a certain transformation. Although you can do this with
684 hand inspection, or some ad-hoc method, this is a real pain and not very useful
685 for big programs. Using the <tt>Statistic</tt> class makes it very easy to
686 keep track of this information, and the calculated information is presented in a
687 uniform manner with the rest of the passes being executed.</p>
689 <p>There are many examples of <tt>Statistic</tt> uses, but the basics of using
690 it are as follows:</p>
693 <li><p>Define your statistic like this:</p>
695 <div class="doc_code">
697 #define <a href="#DEBUG_TYPE">DEBUG_TYPE</a> "mypassname" <i>// This goes before any #includes.</i>
698 STATISTIC(NumXForms, "The # of times I did stuff");
702 <p>The <tt>STATISTIC</tt> macro defines a static variable, whose name is
703 specified by the first argument. The pass name is taken from the DEBUG_TYPE
704 macro, and the description is taken from the second argument. The variable
705 defined ("NumXForms" in this case) acts like an unsigned integer.</p></li>
707 <li><p>Whenever you make a transformation, bump the counter:</p>
709 <div class="doc_code">
711 ++NumXForms; // <i>I did stuff!</i>
718 <p>That's all you have to do. To get '<tt>opt</tt>' to print out the
719 statistics gathered, use the '<tt>-stats</tt>' option:</p>
721 <div class="doc_code">
723 $ opt -stats -mypassname < program.bc > /dev/null
724 <i>... statistics output ...</i>
728 <p> When running <tt>opt</tt> on a C file from the SPEC benchmark
729 suite, it gives a report that looks like this:</p>
731 <div class="doc_code">
733 7646 bitcodewriter - Number of normal instructions
734 725 bitcodewriter - Number of oversized instructions
735 129996 bitcodewriter - Number of bitcode bytes written
736 2817 raise - Number of insts DCEd or constprop'd
737 3213 raise - Number of cast-of-self removed
738 5046 raise - Number of expression trees converted
739 75 raise - Number of other getelementptr's formed
740 138 raise - Number of load/store peepholes
741 42 deadtypeelim - Number of unused typenames removed from symtab
742 392 funcresolve - Number of varargs functions resolved
743 27 globaldce - Number of global variables removed
744 2 adce - Number of basic blocks removed
745 134 cee - Number of branches revectored
746 49 cee - Number of setcc instruction eliminated
747 532 gcse - Number of loads removed
748 2919 gcse - Number of instructions removed
749 86 indvars - Number of canonical indvars added
750 87 indvars - Number of aux indvars removed
751 25 instcombine - Number of dead inst eliminate
752 434 instcombine - Number of insts combined
753 248 licm - Number of load insts hoisted
754 1298 licm - Number of insts hoisted to a loop pre-header
755 3 licm - Number of insts hoisted to multiple loop preds (bad, no loop pre-header)
756 75 mem2reg - Number of alloca's promoted
757 1444 cfgsimplify - Number of blocks simplified
761 <p>Obviously, with so many optimizations, having a unified framework for this
762 stuff is very nice. Making your pass fit well into the framework makes it more
763 maintainable and useful.</p>
767 <!-- ======================================================================= -->
769 <a name="ViewGraph">Viewing graphs while debugging code</a>
774 <p>Several of the important data structures in LLVM are graphs: for example
775 CFGs made out of LLVM <a href="#BasicBlock">BasicBlock</a>s, CFGs made out of
776 LLVM <a href="CodeGenerator.html#machinebasicblock">MachineBasicBlock</a>s, and
777 <a href="CodeGenerator.html#selectiondag_intro">Instruction Selection
778 DAGs</a>. In many cases, while debugging various parts of the compiler, it is
779 nice to instantly visualize these graphs.</p>
781 <p>LLVM provides several callbacks that are available in a debug build to do
782 exactly that. If you call the <tt>Function::viewCFG()</tt> method, for example,
783 the current LLVM tool will pop up a window containing the CFG for the function
784 where each basic block is a node in the graph, and each node contains the
785 instructions in the block. Similarly, there also exists
786 <tt>Function::viewCFGOnly()</tt> (does not include the instructions), the
787 <tt>MachineFunction::viewCFG()</tt> and <tt>MachineFunction::viewCFGOnly()</tt>,
788 and the <tt>SelectionDAG::viewGraph()</tt> methods. Within GDB, for example,
789 you can usually use something like <tt>call DAG.viewGraph()</tt> to pop
790 up a window. Alternatively, you can sprinkle calls to these functions in your
791 code in places you want to debug.</p>
793 <p>Getting this to work requires a small amount of configuration. On Unix
794 systems with X11, install the <a href="http://www.graphviz.org">graphviz</a>
795 toolkit, and make sure 'dot' and 'gv' are in your path. If you are running on
796 Mac OS/X, download and install the Mac OS/X <a
797 href="http://www.pixelglow.com/graphviz/">Graphviz program</a>, and add
798 <tt>/Applications/Graphviz.app/Contents/MacOS/</tt> (or wherever you install
799 it) to your path. Once in your system and path are set up, rerun the LLVM
800 configure script and rebuild LLVM to enable this functionality.</p>
802 <p><tt>SelectionDAG</tt> has been extended to make it easier to locate
803 <i>interesting</i> nodes in large complex graphs. From gdb, if you
804 <tt>call DAG.setGraphColor(<i>node</i>, "<i>color</i>")</tt>, then the
805 next <tt>call DAG.viewGraph()</tt> would highlight the node in the
806 specified color (choices of colors can be found at <a
807 href="http://www.graphviz.org/doc/info/colors.html">colors</a>.) More
808 complex node attributes can be provided with <tt>call
809 DAG.setGraphAttrs(<i>node</i>, "<i>attributes</i>")</tt> (choices can be
810 found at <a href="http://www.graphviz.org/doc/info/attrs.html">Graph
811 Attributes</a>.) If you want to restart and clear all the current graph
812 attributes, then you can <tt>call DAG.clearGraphAttrs()</tt>. </p>
814 <p>Note that graph visualization features are compiled out of Release builds
815 to reduce file size. This means that you need a Debug+Asserts or
816 Release+Asserts build to use these features.</p>
822 <!-- *********************************************************************** -->
824 <a name="datastructure">Picking the Right Data Structure for a Task</a>
826 <!-- *********************************************************************** -->
830 <p>LLVM has a plethora of data structures in the <tt>llvm/ADT/</tt> directory,
831 and we commonly use STL data structures. This section describes the trade-offs
832 you should consider when you pick one.</p>
835 The first step is a choose your own adventure: do you want a sequential
836 container, a set-like container, or a map-like container? The most important
837 thing when choosing a container is the algorithmic properties of how you plan to
838 access the container. Based on that, you should use:</p>
841 <li>a <a href="#ds_map">map-like</a> container if you need efficient look-up
842 of an value based on another value. Map-like containers also support
843 efficient queries for containment (whether a key is in the map). Map-like
844 containers generally do not support efficient reverse mapping (values to
845 keys). If you need that, use two maps. Some map-like containers also
846 support efficient iteration through the keys in sorted order. Map-like
847 containers are the most expensive sort, only use them if you need one of
848 these capabilities.</li>
850 <li>a <a href="#ds_set">set-like</a> container if you need to put a bunch of
851 stuff into a container that automatically eliminates duplicates. Some
852 set-like containers support efficient iteration through the elements in
853 sorted order. Set-like containers are more expensive than sequential
857 <li>a <a href="#ds_sequential">sequential</a> container provides
858 the most efficient way to add elements and keeps track of the order they are
859 added to the collection. They permit duplicates and support efficient
860 iteration, but do not support efficient look-up based on a key.
863 <li>a <a href="#ds_string">string</a> container is a specialized sequential
864 container or reference structure that is used for character or byte
867 <li>a <a href="#ds_bit">bit</a> container provides an efficient way to store and
868 perform set operations on sets of numeric id's, while automatically
869 eliminating duplicates. Bit containers require a maximum of 1 bit for each
870 identifier you want to store.
875 Once the proper category of container is determined, you can fine tune the
876 memory use, constant factors, and cache behaviors of access by intelligently
877 picking a member of the category. Note that constant factors and cache behavior
878 can be a big deal. If you have a vector that usually only contains a few
879 elements (but could contain many), for example, it's much better to use
880 <a href="#dss_smallvector">SmallVector</a> than <a href="#dss_vector">vector</a>
881 . Doing so avoids (relatively) expensive malloc/free calls, which dwarf the
882 cost of adding the elements to the container. </p>
884 <!-- ======================================================================= -->
886 <a name="ds_sequential">Sequential Containers (std::vector, std::list, etc)</a>
890 There are a variety of sequential containers available for you, based on your
891 needs. Pick the first in this section that will do what you want.
893 <!-- _______________________________________________________________________ -->
895 <a name="dss_arrayref">llvm/ADT/ArrayRef.h</a>
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 <!-- _______________________________________________________________________ -->
910 <a name="dss_fixedarrays">Fixed Size Arrays</a>
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 <!-- _______________________________________________________________________ -->
921 <a name="dss_heaparrays">Heap Allocated Arrays</a>
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 <!-- _______________________________________________________________________ -->
937 <a name="dss_tinyptrvector">"llvm/ADT/TinyPtrVector.h"</a>
942 <p><tt>TinyPtrVector<Type></tt> is a highly specialized collection class
943 that is optimized to avoid allocation in the case when a vector has zero or one
944 elements. It has two major restrictions: 1) it can only hold values of pointer
945 type, and 2) it cannot hold a null pointer.</p>
947 <p>Since this container is highly specialized, it is rarely used.</p>
951 <!-- _______________________________________________________________________ -->
953 <a name="dss_smallvector">"llvm/ADT/SmallVector.h"</a>
957 <p><tt>SmallVector<Type, N></tt> is a simple class that looks and smells
958 just like <tt>vector<Type></tt>:
959 it supports efficient iteration, lays out elements in memory order (so you can
960 do pointer arithmetic between elements), supports efficient push_back/pop_back
961 operations, supports efficient random access to its elements, etc.</p>
963 <p>The advantage of SmallVector is that it allocates space for
964 some number of elements (N) <b>in the object itself</b>. Because of this, if
965 the SmallVector is dynamically smaller than N, no malloc is performed. This can
966 be a big win in cases where the malloc/free call is far more expensive than the
967 code that fiddles around with the elements.</p>
969 <p>This is good for vectors that are "usually small" (e.g. the number of
970 predecessors/successors of a block is usually less than 8). On the other hand,
971 this makes the size of the SmallVector itself large, so you don't want to
972 allocate lots of them (doing so will waste a lot of space). As such,
973 SmallVectors are most useful when on the stack.</p>
975 <p>SmallVector also provides a nice portable and efficient replacement for
980 <!-- _______________________________________________________________________ -->
982 <a name="dss_vector"><vector></a>
987 std::vector is well loved and respected. It is useful when SmallVector isn't:
988 when the size of the vector is often large (thus the small optimization will
989 rarely be a benefit) or if you will be allocating many instances of the vector
990 itself (which would waste space for elements that aren't in the container).
991 vector is also useful when interfacing with code that expects vectors :).
994 <p>One worthwhile note about std::vector: avoid code like this:</p>
996 <div class="doc_code">
999 std::vector<foo> V;
1005 <p>Instead, write this as:</p>
1007 <div class="doc_code">
1009 std::vector<foo> V;
1017 <p>Doing so will save (at least) one heap allocation and free per iteration of
1022 <!-- _______________________________________________________________________ -->
1024 <a name="dss_deque"><deque></a>
1028 <p>std::deque is, in some senses, a generalized version of std::vector. Like
1029 std::vector, it provides constant time random access and other similar
1030 properties, but it also provides efficient access to the front of the list. It
1031 does not guarantee continuity of elements within memory.</p>
1033 <p>In exchange for this extra flexibility, std::deque has significantly higher
1034 constant factor costs than std::vector. If possible, use std::vector or
1035 something cheaper.</p>
1038 <!-- _______________________________________________________________________ -->
1040 <a name="dss_list"><list></a>
1044 <p>std::list is an extremely inefficient class that is rarely useful.
1045 It performs a heap allocation for every element inserted into it, thus having an
1046 extremely high constant factor, particularly for small data types. std::list
1047 also only supports bidirectional iteration, not random access iteration.</p>
1049 <p>In exchange for this high cost, std::list supports efficient access to both
1050 ends of the list (like std::deque, but unlike std::vector or SmallVector). In
1051 addition, the iterator invalidation characteristics of std::list are stronger
1052 than that of a vector class: inserting or removing an element into the list does
1053 not invalidate iterator or pointers to other elements in the list.</p>
1056 <!-- _______________________________________________________________________ -->
1058 <a name="dss_ilist">llvm/ADT/ilist.h</a>
1062 <p><tt>ilist<T></tt> implements an 'intrusive' doubly-linked list. It is
1063 intrusive, because it requires the element to store and provide access to the
1064 prev/next pointers for the list.</p>
1066 <p><tt>ilist</tt> has the same drawbacks as <tt>std::list</tt>, and additionally
1067 requires an <tt>ilist_traits</tt> implementation for the element type, but it
1068 provides some novel characteristics. In particular, it can efficiently store
1069 polymorphic objects, the traits class is informed when an element is inserted or
1070 removed from the list, and <tt>ilist</tt>s are guaranteed to support a
1071 constant-time splice operation.</p>
1073 <p>These properties are exactly what we want for things like
1074 <tt>Instruction</tt>s and basic blocks, which is why these are implemented with
1075 <tt>ilist</tt>s.</p>
1077 Related classes of interest are explained in the following subsections:
1079 <li><a href="#dss_ilist_traits">ilist_traits</a></li>
1080 <li><a href="#dss_iplist">iplist</a></li>
1081 <li><a href="#dss_ilist_node">llvm/ADT/ilist_node.h</a></li>
1082 <li><a href="#dss_ilist_sentinel">Sentinels</a></li>
1086 <!-- _______________________________________________________________________ -->
1088 <a name="dss_packedvector">llvm/ADT/PackedVector.h</a>
1093 Useful for storing a vector of values using only a few number of bits for each
1094 value. Apart from the standard operations of a vector-like container, it can
1095 also perform an 'or' set operation.
1100 <div class="doc_code">
1104 FirstCondition = 0x1,
1105 SecondCondition = 0x2,
1110 PackedVector<State, 2> Vec1;
1111 Vec1.push_back(FirstCondition);
1113 PackedVector<State, 2> Vec2;
1114 Vec2.push_back(SecondCondition);
1117 return Vec1[0]; // returns 'Both'.
1124 <!-- _______________________________________________________________________ -->
1126 <a name="dss_ilist_traits">ilist_traits</a>
1130 <p><tt>ilist_traits<T></tt> is <tt>ilist<T></tt>'s customization
1131 mechanism. <tt>iplist<T></tt> (and consequently <tt>ilist<T></tt>)
1132 publicly derive from this traits class.</p>
1135 <!-- _______________________________________________________________________ -->
1137 <a name="dss_iplist">iplist</a>
1141 <p><tt>iplist<T></tt> is <tt>ilist<T></tt>'s base and as such
1142 supports a slightly narrower interface. Notably, inserters from
1143 <tt>T&</tt> are absent.</p>
1145 <p><tt>ilist_traits<T></tt> is a public base of this class and can be
1146 used for a wide variety of customizations.</p>
1149 <!-- _______________________________________________________________________ -->
1151 <a name="dss_ilist_node">llvm/ADT/ilist_node.h</a>
1155 <p><tt>ilist_node<T></tt> implements a the forward and backward links
1156 that are expected by the <tt>ilist<T></tt> (and analogous containers)
1157 in the default manner.</p>
1159 <p><tt>ilist_node<T></tt>s are meant to be embedded in the node type
1160 <tt>T</tt>, usually <tt>T</tt> publicly derives from
1161 <tt>ilist_node<T></tt>.</p>
1164 <!-- _______________________________________________________________________ -->
1166 <a name="dss_ilist_sentinel">Sentinels</a>
1170 <p><tt>ilist</tt>s have another specialty that must be considered. To be a good
1171 citizen in the C++ ecosystem, it needs to support the standard container
1172 operations, such as <tt>begin</tt> and <tt>end</tt> iterators, etc. Also, the
1173 <tt>operator--</tt> must work correctly on the <tt>end</tt> iterator in the
1174 case of non-empty <tt>ilist</tt>s.</p>
1176 <p>The only sensible solution to this problem is to allocate a so-called
1177 <i>sentinel</i> along with the intrusive list, which serves as the <tt>end</tt>
1178 iterator, providing the back-link to the last element. However conforming to the
1179 C++ convention it is illegal to <tt>operator++</tt> beyond the sentinel and it
1180 also must not be dereferenced.</p>
1182 <p>These constraints allow for some implementation freedom to the <tt>ilist</tt>
1183 how to allocate and store the sentinel. The corresponding policy is dictated
1184 by <tt>ilist_traits<T></tt>. By default a <tt>T</tt> gets heap-allocated
1185 whenever the need for a sentinel arises.</p>
1187 <p>While the default policy is sufficient in most cases, it may break down when
1188 <tt>T</tt> does not provide a default constructor. Also, in the case of many
1189 instances of <tt>ilist</tt>s, the memory overhead of the associated sentinels
1190 is wasted. To alleviate the situation with numerous and voluminous
1191 <tt>T</tt>-sentinels, sometimes a trick is employed, leading to <i>ghostly
1194 <p>Ghostly sentinels are obtained by specially-crafted <tt>ilist_traits<T></tt>
1195 which superpose the sentinel with the <tt>ilist</tt> instance in memory. Pointer
1196 arithmetic is used to obtain the sentinel, which is relative to the
1197 <tt>ilist</tt>'s <tt>this</tt> pointer. The <tt>ilist</tt> is augmented by an
1198 extra pointer, which serves as the back-link of the sentinel. This is the only
1199 field in the ghostly sentinel which can be legally accessed.</p>
1202 <!-- _______________________________________________________________________ -->
1204 <a name="dss_other">Other Sequential Container options</a>
1208 <p>Other STL containers are available, such as std::string.</p>
1210 <p>There are also various STL adapter classes such as std::queue,
1211 std::priority_queue, std::stack, etc. These provide simplified access to an
1212 underlying container but don't affect the cost of the container itself.</p>
1217 <!-- ======================================================================= -->
1219 <a name="ds_string">String-like containers</a>
1225 There are a variety of ways to pass around and use strings in C and C++, and
1226 LLVM adds a few new options to choose from. Pick the first option on this list
1227 that will do what you need, they are ordered according to their relative cost.
1230 Note that is is generally preferred to <em>not</em> pass strings around as
1231 "<tt>const char*</tt>"'s. These have a number of problems, including the fact
1232 that they cannot represent embedded nul ("\0") characters, and do not have a
1233 length available efficiently. The general replacement for '<tt>const
1234 char*</tt>' is StringRef.
1237 <p>For more information on choosing string containers for APIs, please see
1238 <a href="#string_apis">Passing strings</a>.</p>
1241 <!-- _______________________________________________________________________ -->
1243 <a name="dss_stringref">llvm/ADT/StringRef.h</a>
1248 The StringRef class is a simple value class that contains a pointer to a
1249 character and a length, and is quite related to the <a
1250 href="#dss_arrayref">ArrayRef</a> class (but specialized for arrays of
1251 characters). Because StringRef carries a length with it, it safely handles
1252 strings with embedded nul characters in it, getting the length does not require
1253 a strlen call, and it even has very convenient APIs for slicing and dicing the
1254 character range that it represents.
1258 StringRef is ideal for passing simple strings around that are known to be live,
1259 either because they are C string literals, std::string, a C array, or a
1260 SmallVector. Each of these cases has an efficient implicit conversion to
1261 StringRef, which doesn't result in a dynamic strlen being executed.
1264 <p>StringRef has a few major limitations which make more powerful string
1265 containers useful:</p>
1268 <li>You cannot directly convert a StringRef to a 'const char*' because there is
1269 no way to add a trailing nul (unlike the .c_str() method on various stronger
1273 <li>StringRef doesn't own or keep alive the underlying string bytes.
1274 As such it can easily lead to dangling pointers, and is not suitable for
1275 embedding in datastructures in most cases (instead, use an std::string or
1276 something like that).</li>
1278 <li>For the same reason, StringRef cannot be used as the return value of a
1279 method if the method "computes" the result string. Instead, use
1282 <li>StringRef's do not allow you to mutate the pointed-to string bytes and it
1283 doesn't allow you to insert or remove bytes from the range. For editing
1284 operations like this, it interoperates with the <a
1285 href="#dss_twine">Twine</a> class.</li>
1288 <p>Because of its strengths and limitations, it is very common for a function to
1289 take a StringRef and for a method on an object to return a StringRef that
1290 points into some string that it owns.</p>
1294 <!-- _______________________________________________________________________ -->
1296 <a name="dss_twine">llvm/ADT/Twine.h</a>
1301 The Twine class is used as an intermediary datatype for APIs that want to take
1302 a string that can be constructed inline with a series of concatenations.
1303 Twine works by forming recursive instances of the Twine datatype (a simple
1304 value object) on the stack as temporary objects, linking them together into a
1305 tree which is then linearized when the Twine is consumed. Twine is only safe
1306 to use as the argument to a function, and should always be a const reference,
1311 void foo(const Twine &T);
1315 foo(X + "." + Twine(i));
1318 <p>This example forms a string like "blarg.42" by concatenating the values
1319 together, and does not form intermediate strings containing "blarg" or
1323 <p>Because Twine is constructed with temporary objects on the stack, and
1324 because these instances are destroyed at the end of the current statement,
1325 it is an inherently dangerous API. For example, this simple variant contains
1326 undefined behavior and will probably crash:</p>
1329 void foo(const Twine &T);
1333 const Twine &Tmp = X + "." + Twine(i);
1337 <p>... because the temporaries are destroyed before the call. That said,
1338 Twine's are much more efficient than intermediate std::string temporaries, and
1339 they work really well with StringRef. Just be aware of their limitations.</p>
1344 <!-- _______________________________________________________________________ -->
1346 <a name="dss_smallstring">llvm/ADT/SmallString.h</a>
1351 <p>SmallString is a subclass of <a href="#dss_smallvector">SmallVector</a> that
1352 adds some convenience APIs like += that takes StringRef's. SmallString avoids
1353 allocating memory in the case when the preallocated space is enough to hold its
1354 data, and it calls back to general heap allocation when required. Since it owns
1355 its data, it is very safe to use and supports full mutation of the string.</p>
1357 <p>Like SmallVector's, the big downside to SmallString is their sizeof. While
1358 they are optimized for small strings, they themselves are not particularly
1359 small. This means that they work great for temporary scratch buffers on the
1360 stack, but should not generally be put into the heap: it is very rare to
1361 see a SmallString as the member of a frequently-allocated heap data structure
1362 or returned by-value.
1367 <!-- _______________________________________________________________________ -->
1369 <a name="dss_stdstring">std::string</a>
1374 <p>The standard C++ std::string class is a very general class that (like
1375 SmallString) owns its underlying data. sizeof(std::string) is very reasonable
1376 so it can be embedded into heap data structures and returned by-value.
1377 On the other hand, std::string is highly inefficient for inline editing (e.g.
1378 concatenating a bunch of stuff together) and because it is provided by the
1379 standard library, its performance characteristics depend a lot of the host
1380 standard library (e.g. libc++ and MSVC provide a highly optimized string
1381 class, GCC contains a really slow implementation).
1384 <p>The major disadvantage of std::string is that almost every operation that
1385 makes them larger can allocate memory, which is slow. As such, it is better
1386 to use SmallVector or Twine as a scratch buffer, but then use std::string to
1387 persist the result.</p>
1392 <!-- end of strings -->
1396 <!-- ======================================================================= -->
1398 <a name="ds_set">Set-Like Containers (std::set, SmallSet, SetVector, etc)</a>
1403 <p>Set-like containers are useful when you need to canonicalize multiple values
1404 into a single representation. There are several different choices for how to do
1405 this, providing various trade-offs.</p>
1407 <!-- _______________________________________________________________________ -->
1409 <a name="dss_sortedvectorset">A sorted 'vector'</a>
1414 <p>If you intend to insert a lot of elements, then do a lot of queries, a
1415 great approach is to use a vector (or other sequential container) with
1416 std::sort+std::unique to remove duplicates. This approach works really well if
1417 your usage pattern has these two distinct phases (insert then query), and can be
1418 coupled with a good choice of <a href="#ds_sequential">sequential container</a>.
1422 This combination provides the several nice properties: the result data is
1423 contiguous in memory (good for cache locality), has few allocations, is easy to
1424 address (iterators in the final vector are just indices or pointers), and can be
1425 efficiently queried with a standard binary or radix search.</p>
1429 <!-- _______________________________________________________________________ -->
1431 <a name="dss_smallset">"llvm/ADT/SmallSet.h"</a>
1436 <p>If you have a set-like data structure that is usually small and whose elements
1437 are reasonably small, a <tt>SmallSet<Type, N></tt> is a good choice. This set
1438 has space for N elements in place (thus, if the set is dynamically smaller than
1439 N, no malloc traffic is required) and accesses them with a simple linear search.
1440 When the set grows beyond 'N' elements, it allocates a more expensive representation that
1441 guarantees efficient access (for most types, it falls back to std::set, but for
1442 pointers it uses something far better, <a
1443 href="#dss_smallptrset">SmallPtrSet</a>).</p>
1445 <p>The magic of this class is that it handles small sets extremely efficiently,
1446 but gracefully handles extremely large sets without loss of efficiency. The
1447 drawback is that the interface is quite small: it supports insertion, queries
1448 and erasing, but does not support iteration.</p>
1452 <!-- _______________________________________________________________________ -->
1454 <a name="dss_smallptrset">"llvm/ADT/SmallPtrSet.h"</a>
1459 <p>SmallPtrSet has all the advantages of <tt>SmallSet</tt> (and a <tt>SmallSet</tt> of pointers is
1460 transparently implemented with a <tt>SmallPtrSet</tt>), but also supports iterators. If
1461 more than 'N' insertions are performed, a single quadratically
1462 probed hash table is allocated and grows as needed, providing extremely
1463 efficient access (constant time insertion/deleting/queries with low constant
1464 factors) and is very stingy with malloc traffic.</p>
1466 <p>Note that, unlike <tt>std::set</tt>, the iterators of <tt>SmallPtrSet</tt> are invalidated
1467 whenever an insertion occurs. Also, the values visited by the iterators are not
1468 visited in sorted order.</p>
1472 <!-- _______________________________________________________________________ -->
1474 <a name="dss_denseset">"llvm/ADT/DenseSet.h"</a>
1480 DenseSet is a simple quadratically probed hash table. It excels at supporting
1481 small values: it uses a single allocation to hold all of the pairs that
1482 are currently inserted in the set. DenseSet is a great way to unique small
1483 values that are not simple pointers (use <a
1484 href="#dss_smallptrset">SmallPtrSet</a> for pointers). Note that DenseSet has
1485 the same requirements for the value type that <a
1486 href="#dss_densemap">DenseMap</a> has.
1491 <!-- _______________________________________________________________________ -->
1493 <a name="dss_FoldingSet">"llvm/ADT/FoldingSet.h"</a>
1499 FoldingSet is an aggregate class that is really good at uniquing
1500 expensive-to-create or polymorphic objects. It is a combination of a chained
1501 hash table with intrusive links (uniqued objects are required to inherit from
1502 FoldingSetNode) that uses <a href="#dss_smallvector">SmallVector</a> as part of
1505 <p>Consider a case where you want to implement a "getOrCreateFoo" method for
1506 a complex object (for example, a node in the code generator). The client has a
1507 description of *what* it wants to generate (it knows the opcode and all the
1508 operands), but we don't want to 'new' a node, then try inserting it into a set
1509 only to find out it already exists, at which point we would have to delete it
1510 and return the node that already exists.
1513 <p>To support this style of client, FoldingSet perform a query with a
1514 FoldingSetNodeID (which wraps SmallVector) that can be used to describe the
1515 element that we want to query for. The query either returns the element
1516 matching the ID or it returns an opaque ID that indicates where insertion should
1517 take place. Construction of the ID usually does not require heap traffic.</p>
1519 <p>Because FoldingSet uses intrusive links, it can support polymorphic objects
1520 in the set (for example, you can have SDNode instances mixed with LoadSDNodes).
1521 Because the elements are individually allocated, pointers to the elements are
1522 stable: inserting or removing elements does not invalidate any pointers to other
1528 <!-- _______________________________________________________________________ -->
1530 <a name="dss_set"><set></a>
1535 <p><tt>std::set</tt> is a reasonable all-around set class, which is decent at
1536 many things but great at nothing. std::set allocates memory for each element
1537 inserted (thus it is very malloc intensive) and typically stores three pointers
1538 per element in the set (thus adding a large amount of per-element space
1539 overhead). It offers guaranteed log(n) performance, which is not particularly
1540 fast from a complexity standpoint (particularly if the elements of the set are
1541 expensive to compare, like strings), and has extremely high constant factors for
1542 lookup, insertion and removal.</p>
1544 <p>The advantages of std::set are that its iterators are stable (deleting or
1545 inserting an element from the set does not affect iterators or pointers to other
1546 elements) and that iteration over the set is guaranteed to be in sorted order.
1547 If the elements in the set are large, then the relative overhead of the pointers
1548 and malloc traffic is not a big deal, but if the elements of the set are small,
1549 std::set is almost never a good choice.</p>
1553 <!-- _______________________________________________________________________ -->
1555 <a name="dss_setvector">"llvm/ADT/SetVector.h"</a>
1559 <p>LLVM's SetVector<Type> is an adapter class that combines your choice of
1560 a set-like container along with a <a href="#ds_sequential">Sequential
1561 Container</a>. The important property
1562 that this provides is efficient insertion with uniquing (duplicate elements are
1563 ignored) with iteration support. It implements this by inserting elements into
1564 both a set-like container and the sequential container, using the set-like
1565 container for uniquing and the sequential container for iteration.
1568 <p>The difference between SetVector and other sets is that the order of
1569 iteration is guaranteed to match the order of insertion into the SetVector.
1570 This property is really important for things like sets of pointers. Because
1571 pointer values are non-deterministic (e.g. vary across runs of the program on
1572 different machines), iterating over the pointers in the set will
1573 not be in a well-defined order.</p>
1576 The drawback of SetVector is that it requires twice as much space as a normal
1577 set and has the sum of constant factors from the set-like container and the
1578 sequential container that it uses. Use it *only* if you need to iterate over
1579 the elements in a deterministic order. SetVector is also expensive to delete
1580 elements out of (linear time), unless you use it's "pop_back" method, which is
1584 <p><tt>SetVector</tt> is an adapter class that defaults to
1585 using <tt>std::vector</tt> and a size 16 <tt>SmallSet</tt> for the underlying
1586 containers, so it is quite expensive. However,
1587 <tt>"llvm/ADT/SetVector.h"</tt> also provides a <tt>SmallSetVector</tt>
1588 class, which defaults to using a <tt>SmallVector</tt> and <tt>SmallSet</tt>
1589 of a specified size. If you use this, and if your sets are dynamically
1590 smaller than <tt>N</tt>, you will save a lot of heap traffic.</p>
1594 <!-- _______________________________________________________________________ -->
1596 <a name="dss_uniquevector">"llvm/ADT/UniqueVector.h"</a>
1602 UniqueVector is similar to <a href="#dss_setvector">SetVector</a>, but it
1603 retains a unique ID for each element inserted into the set. It internally
1604 contains a map and a vector, and it assigns a unique ID for each value inserted
1607 <p>UniqueVector is very expensive: its cost is the sum of the cost of
1608 maintaining both the map and vector, it has high complexity, high constant
1609 factors, and produces a lot of malloc traffic. It should be avoided.</p>
1613 <!-- _______________________________________________________________________ -->
1615 <a name="dss_immutableset">"llvm/ADT/ImmutableSet.h"</a>
1621 ImmutableSet is an immutable (functional) set implementation based on an AVL
1623 Adding or removing elements is done through a Factory object and results in the
1624 creation of a new ImmutableSet object.
1625 If an ImmutableSet already exists with the given contents, then the existing one
1626 is returned; equality is compared with a FoldingSetNodeID.
1627 The time and space complexity of add or remove operations is logarithmic in the
1628 size of the original set.
1631 There is no method for returning an element of the set, you can only check for
1637 <!-- _______________________________________________________________________ -->
1639 <a name="dss_otherset">Other Set-Like Container Options</a>
1645 The STL provides several other options, such as std::multiset and the various
1646 "hash_set" like containers (whether from C++ TR1 or from the SGI library). We
1647 never use hash_set and unordered_set because they are generally very expensive
1648 (each insertion requires a malloc) and very non-portable.
1651 <p>std::multiset is useful if you're not interested in elimination of
1652 duplicates, but has all the drawbacks of std::set. A sorted vector (where you
1653 don't delete duplicate entries) or some other approach is almost always
1660 <!-- ======================================================================= -->
1662 <a name="ds_map">Map-Like Containers (std::map, DenseMap, etc)</a>
1666 Map-like containers are useful when you want to associate data to a key. As
1667 usual, there are a lot of different ways to do this. :)
1669 <!-- _______________________________________________________________________ -->
1671 <a name="dss_sortedvectormap">A sorted 'vector'</a>
1677 If your usage pattern follows a strict insert-then-query approach, you can
1678 trivially use the same approach as <a href="#dss_sortedvectorset">sorted vectors
1679 for set-like containers</a>. The only difference is that your query function
1680 (which uses std::lower_bound to get efficient log(n) lookup) should only compare
1681 the key, not both the key and value. This yields the same advantages as sorted
1686 <!-- _______________________________________________________________________ -->
1688 <a name="dss_stringmap">"llvm/ADT/StringMap.h"</a>
1694 Strings are commonly used as keys in maps, and they are difficult to support
1695 efficiently: they are variable length, inefficient to hash and compare when
1696 long, expensive to copy, etc. StringMap is a specialized container designed to
1697 cope with these issues. It supports mapping an arbitrary range of bytes to an
1698 arbitrary other object.</p>
1700 <p>The StringMap implementation uses a quadratically-probed hash table, where
1701 the buckets store a pointer to the heap allocated entries (and some other
1702 stuff). The entries in the map must be heap allocated because the strings are
1703 variable length. The string data (key) and the element object (value) are
1704 stored in the same allocation with the string data immediately after the element
1705 object. This container guarantees the "<tt>(char*)(&Value+1)</tt>" points
1706 to the key string for a value.</p>
1708 <p>The StringMap is very fast for several reasons: quadratic probing is very
1709 cache efficient for lookups, the hash value of strings in buckets is not
1710 recomputed when looking up an element, StringMap rarely has to touch the
1711 memory for unrelated objects when looking up a value (even when hash collisions
1712 happen), hash table growth does not recompute the hash values for strings
1713 already in the table, and each pair in the map is store in a single allocation
1714 (the string data is stored in the same allocation as the Value of a pair).</p>
1716 <p>StringMap also provides query methods that take byte ranges, so it only ever
1717 copies a string if a value is inserted into the table.</p>
1720 <!-- _______________________________________________________________________ -->
1722 <a name="dss_indexedmap">"llvm/ADT/IndexedMap.h"</a>
1727 IndexedMap is a specialized container for mapping small dense integers (or
1728 values that can be mapped to small dense integers) to some other type. It is
1729 internally implemented as a vector with a mapping function that maps the keys to
1730 the dense integer range.
1734 This is useful for cases like virtual registers in the LLVM code generator: they
1735 have a dense mapping that is offset by a compile-time constant (the first
1736 virtual register ID).</p>
1740 <!-- _______________________________________________________________________ -->
1742 <a name="dss_densemap">"llvm/ADT/DenseMap.h"</a>
1748 DenseMap is a simple quadratically probed hash table. It excels at supporting
1749 small keys and values: it uses a single allocation to hold all of the pairs that
1750 are currently inserted in the map. DenseMap is a great way to map pointers to
1751 pointers, or map other small types to each other.
1755 There are several aspects of DenseMap that you should be aware of, however. The
1756 iterators in a DenseMap are invalidated whenever an insertion occurs, unlike
1757 map. Also, because DenseMap allocates space for a large number of key/value
1758 pairs (it starts with 64 by default), it will waste a lot of space if your keys
1759 or values are large. Finally, you must implement a partial specialization of
1760 DenseMapInfo for the key that you want, if it isn't already supported. This
1761 is required to tell DenseMap about two special marker values (which can never be
1762 inserted into the map) that it needs internally.</p>
1765 DenseMap's find_as() method supports lookup operations using an alternate key
1766 type. This is useful in cases where the normal key type is expensive to
1767 construct, but cheap to compare against. The DenseMapInfo is responsible for
1768 defining the appropriate comparison and hashing methods for each alternate
1774 <!-- _______________________________________________________________________ -->
1776 <a name="dss_valuemap">"llvm/ADT/ValueMap.h"</a>
1782 ValueMap is a wrapper around a <a href="#dss_densemap">DenseMap</a> mapping
1783 Value*s (or subclasses) to another type. When a Value is deleted or RAUW'ed,
1784 ValueMap will update itself so the new version of the key is mapped to the same
1785 value, just as if the key were a WeakVH. You can configure exactly how this
1786 happens, and what else happens on these two events, by passing
1787 a <code>Config</code> parameter to the ValueMap template.</p>
1791 <!-- _______________________________________________________________________ -->
1793 <a name="dss_intervalmap">"llvm/ADT/IntervalMap.h"</a>
1798 <p> IntervalMap is a compact map for small keys and values. It maps key
1799 intervals instead of single keys, and it will automatically coalesce adjacent
1800 intervals. When then map only contains a few intervals, they are stored in the
1801 map object itself to avoid allocations.</p>
1803 <p> The IntervalMap iterators are quite big, so they should not be passed around
1804 as STL iterators. The heavyweight iterators allow a smaller data structure.</p>
1808 <!-- _______________________________________________________________________ -->
1810 <a name="dss_map"><map></a>
1816 std::map has similar characteristics to <a href="#dss_set">std::set</a>: it uses
1817 a single allocation per pair inserted into the map, it offers log(n) lookup with
1818 an extremely large constant factor, imposes a space penalty of 3 pointers per
1819 pair in the map, etc.</p>
1821 <p>std::map is most useful when your keys or values are very large, if you need
1822 to iterate over the collection in sorted order, or if you need stable iterators
1823 into the map (i.e. they don't get invalidated if an insertion or deletion of
1824 another element takes place).</p>
1828 <!-- _______________________________________________________________________ -->
1830 <a name="dss_inteqclasses">"llvm/ADT/IntEqClasses.h"</a>
1835 <p>IntEqClasses provides a compact representation of equivalence classes of
1836 small integers. Initially, each integer in the range 0..n-1 has its own
1837 equivalence class. Classes can be joined by passing two class representatives to
1838 the join(a, b) method. Two integers are in the same class when findLeader()
1839 returns the same representative.</p>
1841 <p>Once all equivalence classes are formed, the map can be compressed so each
1842 integer 0..n-1 maps to an equivalence class number in the range 0..m-1, where m
1843 is the total number of equivalence classes. The map must be uncompressed before
1844 it can be edited again.</p>
1848 <!-- _______________________________________________________________________ -->
1850 <a name="dss_immutablemap">"llvm/ADT/ImmutableMap.h"</a>
1856 ImmutableMap is an immutable (functional) map implementation based on an AVL
1858 Adding or removing elements is done through a Factory object and results in the
1859 creation of a new ImmutableMap object.
1860 If an ImmutableMap already exists with the given key set, then the existing one
1861 is returned; equality is compared with a FoldingSetNodeID.
1862 The time and space complexity of add or remove operations is logarithmic in the
1863 size of the original map.
1867 <!-- _______________________________________________________________________ -->
1869 <a name="dss_othermap">Other Map-Like Container Options</a>
1875 The STL provides several other options, such as std::multimap and the various
1876 "hash_map" like containers (whether from C++ TR1 or from the SGI library). We
1877 never use hash_set and unordered_set because they are generally very expensive
1878 (each insertion requires a malloc) and very non-portable.</p>
1880 <p>std::multimap is useful if you want to map a key to multiple values, but has
1881 all the drawbacks of std::map. A sorted vector or some other approach is almost
1888 <!-- ======================================================================= -->
1890 <a name="ds_bit">Bit storage containers (BitVector, SparseBitVector)</a>
1894 <p>Unlike the other containers, there are only two bit storage containers, and
1895 choosing when to use each is relatively straightforward.</p>
1897 <p>One additional option is
1898 <tt>std::vector<bool></tt>: we discourage its use for two reasons 1) the
1899 implementation in many common compilers (e.g. commonly available versions of
1900 GCC) is extremely inefficient and 2) the C++ standards committee is likely to
1901 deprecate this container and/or change it significantly somehow. In any case,
1902 please don't use it.</p>
1904 <!-- _______________________________________________________________________ -->
1906 <a name="dss_bitvector">BitVector</a>
1910 <p> The BitVector container provides a dynamic size set of bits for manipulation.
1911 It supports individual bit setting/testing, as well as set operations. The set
1912 operations take time O(size of bitvector), but operations are performed one word
1913 at a time, instead of one bit at a time. This makes the BitVector very fast for
1914 set operations compared to other containers. Use the BitVector when you expect
1915 the number of set bits to be high (IE a dense set).
1919 <!-- _______________________________________________________________________ -->
1921 <a name="dss_smallbitvector">SmallBitVector</a>
1925 <p> The SmallBitVector container provides the same interface as BitVector, but
1926 it is optimized for the case where only a small number of bits, less than
1927 25 or so, are needed. It also transparently supports larger bit counts, but
1928 slightly less efficiently than a plain BitVector, so SmallBitVector should
1929 only be used when larger counts are rare.
1933 At this time, SmallBitVector does not support set operations (and, or, xor),
1934 and its operator[] does not provide an assignable lvalue.
1938 <!-- _______________________________________________________________________ -->
1940 <a name="dss_sparsebitvector">SparseBitVector</a>
1944 <p> The SparseBitVector container is much like BitVector, with one major
1945 difference: Only the bits that are set, are stored. This makes the
1946 SparseBitVector much more space efficient than BitVector when the set is sparse,
1947 as well as making set operations O(number of set bits) instead of O(size of
1948 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
1949 (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).
1957 <!-- *********************************************************************** -->
1959 <a name="common">Helpful Hints for Common Operations</a>
1961 <!-- *********************************************************************** -->
1965 <p>This section describes how to perform some very simple transformations of
1966 LLVM code. This is meant to give examples of common idioms used, showing the
1967 practical side of LLVM transformations. <p> Because this is a "how-to" section,
1968 you should also read about the main classes that you will be working with. The
1969 <a href="#coreclasses">Core LLVM Class Hierarchy Reference</a> contains details
1970 and descriptions of the main classes that you should know about.</p>
1972 <!-- NOTE: this section should be heavy on example code -->
1973 <!-- ======================================================================= -->
1975 <a name="inspection">Basic Inspection and Traversal Routines</a>
1980 <p>The LLVM compiler infrastructure have many different data structures that may
1981 be traversed. Following the example of the C++ standard template library, the
1982 techniques used to traverse these various data structures are all basically the
1983 same. For a enumerable sequence of values, the <tt>XXXbegin()</tt> function (or
1984 method) returns an iterator to the start of the sequence, the <tt>XXXend()</tt>
1985 function returns an iterator pointing to one past the last valid element of the
1986 sequence, and there is some <tt>XXXiterator</tt> data type that is common
1987 between the two operations.</p>
1989 <p>Because the pattern for iteration is common across many different aspects of
1990 the program representation, the standard template library algorithms may be used
1991 on them, and it is easier to remember how to iterate. First we show a few common
1992 examples of the data structures that need to be traversed. Other data
1993 structures are traversed in very similar ways.</p>
1995 <!-- _______________________________________________________________________ -->
1997 <a name="iterate_function">Iterating over the </a><a
1998 href="#BasicBlock"><tt>BasicBlock</tt></a>s in a <a
1999 href="#Function"><tt>Function</tt></a>
2004 <p>It's quite common to have a <tt>Function</tt> instance that you'd like to
2005 transform in some way; in particular, you'd like to manipulate its
2006 <tt>BasicBlock</tt>s. To facilitate this, you'll need to iterate over all of
2007 the <tt>BasicBlock</tt>s that constitute the <tt>Function</tt>. The following is
2008 an example that prints the name of a <tt>BasicBlock</tt> and the number of
2009 <tt>Instruction</tt>s it contains:</p>
2011 <div class="doc_code">
2013 // <i>func is a pointer to a Function instance</i>
2014 for (Function::iterator i = func->begin(), e = func->end(); i != e; ++i)
2015 // <i>Print out the name of the basic block if it has one, and then the</i>
2016 // <i>number of instructions that it contains</i>
2017 errs() << "Basic block (name=" << i->getName() << ") has "
2018 << i->size() << " instructions.\n";
2022 <p>Note that i can be used as if it were a pointer for the purposes of
2023 invoking member functions of the <tt>Instruction</tt> class. This is
2024 because the indirection operator is overloaded for the iterator
2025 classes. In the above code, the expression <tt>i->size()</tt> is
2026 exactly equivalent to <tt>(*i).size()</tt> just like you'd expect.</p>
2030 <!-- _______________________________________________________________________ -->
2032 <a name="iterate_basicblock">Iterating over the </a><a
2033 href="#Instruction"><tt>Instruction</tt></a>s in a <a
2034 href="#BasicBlock"><tt>BasicBlock</tt></a>
2039 <p>Just like when dealing with <tt>BasicBlock</tt>s in <tt>Function</tt>s, it's
2040 easy to iterate over the individual instructions that make up
2041 <tt>BasicBlock</tt>s. Here's a code snippet that prints out each instruction in
2042 a <tt>BasicBlock</tt>:</p>
2044 <div class="doc_code">
2046 // <i>blk is a pointer to a BasicBlock instance</i>
2047 for (BasicBlock::iterator i = blk->begin(), e = blk->end(); i != e; ++i)
2048 // <i>The next statement works since operator<<(ostream&,...)</i>
2049 // <i>is overloaded for Instruction&</i>
2050 errs() << *i << "\n";
2054 <p>However, this isn't really the best way to print out the contents of a
2055 <tt>BasicBlock</tt>! Since the ostream operators are overloaded for virtually
2056 anything you'll care about, you could have just invoked the print routine on the
2057 basic block itself: <tt>errs() << *blk << "\n";</tt>.</p>
2061 <!-- _______________________________________________________________________ -->
2063 <a name="iterate_institer">Iterating over the </a><a
2064 href="#Instruction"><tt>Instruction</tt></a>s in a <a
2065 href="#Function"><tt>Function</tt></a>
2070 <p>If you're finding that you commonly iterate over a <tt>Function</tt>'s
2071 <tt>BasicBlock</tt>s and then that <tt>BasicBlock</tt>'s <tt>Instruction</tt>s,
2072 <tt>InstIterator</tt> should be used instead. You'll need to include <a
2073 href="/doxygen/InstIterator_8h-source.html"><tt>llvm/Support/InstIterator.h</tt></a>,
2074 and then instantiate <tt>InstIterator</tt>s explicitly in your code. Here's a
2075 small example that shows how to dump all instructions in a function to the standard error stream:<p>
2077 <div class="doc_code">
2079 #include "<a href="/doxygen/InstIterator_8h-source.html">llvm/Support/InstIterator.h</a>"
2081 // <i>F is a pointer to a Function instance</i>
2082 for (inst_iterator I = inst_begin(F), E = inst_end(F); I != E; ++I)
2083 errs() << *I << "\n";
2087 <p>Easy, isn't it? You can also use <tt>InstIterator</tt>s to fill a
2088 work list with its initial contents. For example, if you wanted to
2089 initialize a work list to contain all instructions in a <tt>Function</tt>
2090 F, all you would need to do is something like:</p>
2092 <div class="doc_code">
2094 std::set<Instruction*> worklist;
2095 // or better yet, SmallPtrSet<Instruction*, 64> worklist;
2097 for (inst_iterator I = inst_begin(F), E = inst_end(F); I != E; ++I)
2098 worklist.insert(&*I);
2102 <p>The STL set <tt>worklist</tt> would now contain all instructions in the
2103 <tt>Function</tt> pointed to by F.</p>
2107 <!-- _______________________________________________________________________ -->
2109 <a name="iterate_convert">Turning an iterator into a class pointer (and
2115 <p>Sometimes, it'll be useful to grab a reference (or pointer) to a class
2116 instance when all you've got at hand is an iterator. Well, extracting
2117 a reference or a pointer from an iterator is very straight-forward.
2118 Assuming that <tt>i</tt> is a <tt>BasicBlock::iterator</tt> and <tt>j</tt>
2119 is a <tt>BasicBlock::const_iterator</tt>:</p>
2121 <div class="doc_code">
2123 Instruction& inst = *i; // <i>Grab reference to instruction reference</i>
2124 Instruction* pinst = &*i; // <i>Grab pointer to instruction reference</i>
2125 const Instruction& inst = *j;
2129 <p>However, the iterators you'll be working with in the LLVM framework are
2130 special: they will automatically convert to a ptr-to-instance type whenever they
2131 need to. Instead of dereferencing the iterator and then taking the address of
2132 the result, you can simply assign the iterator to the proper pointer type and
2133 you get the dereference and address-of operation as a result of the assignment
2134 (behind the scenes, this is a result of overloading casting mechanisms). Thus
2135 the last line of the last example,</p>
2137 <div class="doc_code">
2139 Instruction *pinst = &*i;
2143 <p>is semantically equivalent to</p>
2145 <div class="doc_code">
2147 Instruction *pinst = i;
2151 <p>It's also possible to turn a class pointer into the corresponding iterator,
2152 and this is a constant time operation (very efficient). The following code
2153 snippet illustrates use of the conversion constructors provided by LLVM
2154 iterators. By using these, you can explicitly grab the iterator of something
2155 without actually obtaining it via iteration over some structure:</p>
2157 <div class="doc_code">
2159 void printNextInstruction(Instruction* inst) {
2160 BasicBlock::iterator it(inst);
2161 ++it; // <i>After this line, it refers to the instruction after *inst</i>
2162 if (it != inst->getParent()->end()) errs() << *it << "\n";
2167 <p>Unfortunately, these implicit conversions come at a cost; they prevent
2168 these iterators from conforming to standard iterator conventions, and thus
2169 from being usable with standard algorithms and containers. For example, they
2170 prevent the following code, where <tt>B</tt> is a <tt>BasicBlock</tt>,
2173 <div class="doc_code">
2175 llvm::SmallVector<llvm::Instruction *, 16>(B->begin(), B->end());
2179 <p>Because of this, these implicit conversions may be removed some day,
2180 and <tt>operator*</tt> changed to return a pointer instead of a reference.</p>
2184 <!--_______________________________________________________________________-->
2186 <a name="iterate_complex">Finding call sites: a slightly more complex
2192 <p>Say that you're writing a FunctionPass and would like to count all the
2193 locations in the entire module (that is, across every <tt>Function</tt>) where a
2194 certain function (i.e., some <tt>Function</tt>*) is already in scope. As you'll
2195 learn later, you may want to use an <tt>InstVisitor</tt> to accomplish this in a
2196 much more straight-forward manner, but this example will allow us to explore how
2197 you'd do it if you didn't have <tt>InstVisitor</tt> around. In pseudo-code, this
2198 is what we want to do:</p>
2200 <div class="doc_code">
2202 initialize callCounter to zero
2203 for each Function f in the Module
2204 for each BasicBlock b in f
2205 for each Instruction i in b
2206 if (i is a CallInst and calls the given function)
2207 increment callCounter
2211 <p>And the actual code is (remember, because we're writing a
2212 <tt>FunctionPass</tt>, our <tt>FunctionPass</tt>-derived class simply has to
2213 override the <tt>runOnFunction</tt> method):</p>
2215 <div class="doc_code">
2217 Function* targetFunc = ...;
2219 class OurFunctionPass : public FunctionPass {
2221 OurFunctionPass(): callCounter(0) { }
2223 virtual runOnFunction(Function& F) {
2224 for (Function::iterator b = F.begin(), be = F.end(); b != be; ++b) {
2225 for (BasicBlock::iterator i = b->begin(), ie = b->end(); i != ie; ++i) {
2226 if (<a href="#CallInst">CallInst</a>* callInst = <a href="#isa">dyn_cast</a><<a
2227 href="#CallInst">CallInst</a>>(&*i)) {
2228 // <i>We know we've encountered a call instruction, so we</i>
2229 // <i>need to determine if it's a call to the</i>
2230 // <i>function pointed to by m_func or not.</i>
2231 if (callInst->getCalledFunction() == targetFunc)
2239 unsigned callCounter;
2246 <!--_______________________________________________________________________-->
2248 <a name="calls_and_invokes">Treating calls and invokes the same way</a>
2253 <p>You may have noticed that the previous example was a bit oversimplified in
2254 that it did not deal with call sites generated by 'invoke' instructions. In
2255 this, and in other situations, you may find that you want to treat
2256 <tt>CallInst</tt>s and <tt>InvokeInst</tt>s the same way, even though their
2257 most-specific common base class is <tt>Instruction</tt>, which includes lots of
2258 less closely-related things. For these cases, LLVM provides a handy wrapper
2260 href="http://llvm.org/doxygen/classllvm_1_1CallSite.html"><tt>CallSite</tt></a>.
2261 It is essentially a wrapper around an <tt>Instruction</tt> pointer, with some
2262 methods that provide functionality common to <tt>CallInst</tt>s and
2263 <tt>InvokeInst</tt>s.</p>
2265 <p>This class has "value semantics": it should be passed by value, not by
2266 reference and it should not be dynamically allocated or deallocated using
2267 <tt>operator new</tt> or <tt>operator delete</tt>. It is efficiently copyable,
2268 assignable and constructable, with costs equivalents to that of a bare pointer.
2269 If you look at its definition, it has only a single pointer member.</p>
2273 <!--_______________________________________________________________________-->
2275 <a name="iterate_chains">Iterating over def-use & use-def chains</a>
2280 <p>Frequently, we might have an instance of the <a
2281 href="/doxygen/classllvm_1_1Value.html">Value Class</a> and we want to
2282 determine which <tt>User</tt>s use the <tt>Value</tt>. The list of all
2283 <tt>User</tt>s of a particular <tt>Value</tt> is called a <i>def-use</i> chain.
2284 For example, let's say we have a <tt>Function*</tt> named <tt>F</tt> to a
2285 particular function <tt>foo</tt>. Finding all of the instructions that
2286 <i>use</i> <tt>foo</tt> is as simple as iterating over the <i>def-use</i> chain
2289 <div class="doc_code">
2293 for (Value::use_iterator i = F->use_begin(), e = F->use_end(); i != e; ++i)
2294 if (Instruction *Inst = dyn_cast<Instruction>(*i)) {
2295 errs() << "F is used in instruction:\n";
2296 errs() << *Inst << "\n";
2301 <p>Note that dereferencing a <tt>Value::use_iterator</tt> is not a very cheap
2302 operation. Instead of performing <tt>*i</tt> above several times, consider
2303 doing it only once in the loop body and reusing its result.</p>
2305 <p>Alternatively, it's common to have an instance of the <a
2306 href="/doxygen/classllvm_1_1User.html">User Class</a> and need to know what
2307 <tt>Value</tt>s are used by it. The list of all <tt>Value</tt>s used by a
2308 <tt>User</tt> is known as a <i>use-def</i> chain. Instances of class
2309 <tt>Instruction</tt> are common <tt>User</tt>s, so we might want to iterate over
2310 all of the values that a particular instruction uses (that is, the operands of
2311 the particular <tt>Instruction</tt>):</p>
2313 <div class="doc_code">
2315 Instruction *pi = ...;
2317 for (User::op_iterator i = pi->op_begin(), e = pi->op_end(); i != e; ++i) {
2324 <p>Declaring objects as <tt>const</tt> is an important tool of enforcing
2325 mutation free algorithms (such as analyses, etc.). For this purpose above
2326 iterators come in constant flavors as <tt>Value::const_use_iterator</tt>
2327 and <tt>Value::const_op_iterator</tt>. They automatically arise when
2328 calling <tt>use/op_begin()</tt> on <tt>const Value*</tt>s or
2329 <tt>const User*</tt>s respectively. Upon dereferencing, they return
2330 <tt>const Use*</tt>s. Otherwise the above patterns remain unchanged.</p>
2334 <!--_______________________________________________________________________-->
2336 <a name="iterate_preds">Iterating over predecessors &
2337 successors of blocks</a>
2342 <p>Iterating over the predecessors and successors of a block is quite easy
2343 with the routines defined in <tt>"llvm/Support/CFG.h"</tt>. Just use code like
2344 this to iterate over all predecessors of BB:</p>
2346 <div class="doc_code">
2348 #include "llvm/Support/CFG.h"
2349 BasicBlock *BB = ...;
2351 for (pred_iterator PI = pred_begin(BB), E = pred_end(BB); PI != E; ++PI) {
2352 BasicBlock *Pred = *PI;
2358 <p>Similarly, to iterate over successors use
2359 succ_iterator/succ_begin/succ_end.</p>
2365 <!-- ======================================================================= -->
2367 <a name="simplechanges">Making simple changes</a>
2372 <p>There are some primitive transformation operations present in the LLVM
2373 infrastructure that are worth knowing about. When performing
2374 transformations, it's fairly common to manipulate the contents of basic
2375 blocks. This section describes some of the common methods for doing so
2376 and gives example code.</p>
2378 <!--_______________________________________________________________________-->
2380 <a name="schanges_creating">Creating and inserting new
2381 <tt>Instruction</tt>s</a>
2386 <p><i>Instantiating Instructions</i></p>
2388 <p>Creation of <tt>Instruction</tt>s is straight-forward: simply call the
2389 constructor for the kind of instruction to instantiate and provide the necessary
2390 parameters. For example, an <tt>AllocaInst</tt> only <i>requires</i> a
2391 (const-ptr-to) <tt>Type</tt>. Thus:</p>
2393 <div class="doc_code">
2395 AllocaInst* ai = new AllocaInst(Type::Int32Ty);
2399 <p>will create an <tt>AllocaInst</tt> instance that represents the allocation of
2400 one integer in the current stack frame, at run time. Each <tt>Instruction</tt>
2401 subclass is likely to have varying default parameters which change the semantics
2402 of the instruction, so refer to the <a
2403 href="/doxygen/classllvm_1_1Instruction.html">doxygen documentation for the subclass of
2404 Instruction</a> that you're interested in instantiating.</p>
2406 <p><i>Naming values</i></p>
2408 <p>It is very useful to name the values of instructions when you're able to, as
2409 this facilitates the debugging of your transformations. If you end up looking
2410 at generated LLVM machine code, you definitely want to have logical names
2411 associated with the results of instructions! By supplying a value for the
2412 <tt>Name</tt> (default) parameter of the <tt>Instruction</tt> constructor, you
2413 associate a logical name with the result of the instruction's execution at
2414 run time. For example, say that I'm writing a transformation that dynamically
2415 allocates space for an integer on the stack, and that integer is going to be
2416 used as some kind of index by some other code. To accomplish this, I place an
2417 <tt>AllocaInst</tt> at the first point in the first <tt>BasicBlock</tt> of some
2418 <tt>Function</tt>, and I'm intending to use it within the same
2419 <tt>Function</tt>. I might do:</p>
2421 <div class="doc_code">
2423 AllocaInst* pa = new AllocaInst(Type::Int32Ty, 0, "indexLoc");
2427 <p>where <tt>indexLoc</tt> is now the logical name of the instruction's
2428 execution value, which is a pointer to an integer on the run time stack.</p>
2430 <p><i>Inserting instructions</i></p>
2432 <p>There are essentially two ways to insert an <tt>Instruction</tt>
2433 into an existing sequence of instructions that form a <tt>BasicBlock</tt>:</p>
2436 <li>Insertion into an explicit instruction list
2438 <p>Given a <tt>BasicBlock* pb</tt>, an <tt>Instruction* pi</tt> within that
2439 <tt>BasicBlock</tt>, and a newly-created instruction we wish to insert
2440 before <tt>*pi</tt>, we do the following: </p>
2442 <div class="doc_code">
2444 BasicBlock *pb = ...;
2445 Instruction *pi = ...;
2446 Instruction *newInst = new Instruction(...);
2448 pb->getInstList().insert(pi, newInst); // <i>Inserts newInst before pi in pb</i>
2452 <p>Appending to the end of a <tt>BasicBlock</tt> is so common that
2453 the <tt>Instruction</tt> class and <tt>Instruction</tt>-derived
2454 classes provide constructors which take a pointer to a
2455 <tt>BasicBlock</tt> to be appended to. For example code that
2458 <div class="doc_code">
2460 BasicBlock *pb = ...;
2461 Instruction *newInst = new Instruction(...);
2463 pb->getInstList().push_back(newInst); // <i>Appends newInst to pb</i>
2469 <div class="doc_code">
2471 BasicBlock *pb = ...;
2472 Instruction *newInst = new Instruction(..., pb);
2476 <p>which is much cleaner, especially if you are creating
2477 long instruction streams.</p></li>
2479 <li>Insertion into an implicit instruction list
2481 <p><tt>Instruction</tt> instances that are already in <tt>BasicBlock</tt>s
2482 are implicitly associated with an existing instruction list: the instruction
2483 list of the enclosing basic block. Thus, we could have accomplished the same
2484 thing as the above code without being given a <tt>BasicBlock</tt> by doing:
2487 <div class="doc_code">
2489 Instruction *pi = ...;
2490 Instruction *newInst = new Instruction(...);
2492 pi->getParent()->getInstList().insert(pi, newInst);
2496 <p>In fact, this sequence of steps occurs so frequently that the
2497 <tt>Instruction</tt> class and <tt>Instruction</tt>-derived classes provide
2498 constructors which take (as a default parameter) a pointer to an
2499 <tt>Instruction</tt> which the newly-created <tt>Instruction</tt> should
2500 precede. That is, <tt>Instruction</tt> constructors are capable of
2501 inserting the newly-created instance into the <tt>BasicBlock</tt> of a
2502 provided instruction, immediately before that instruction. Using an
2503 <tt>Instruction</tt> constructor with a <tt>insertBefore</tt> (default)
2504 parameter, the above code becomes:</p>
2506 <div class="doc_code">
2508 Instruction* pi = ...;
2509 Instruction* newInst = new Instruction(..., pi);
2513 <p>which is much cleaner, especially if you're creating a lot of
2514 instructions and adding them to <tt>BasicBlock</tt>s.</p></li>
2519 <!--_______________________________________________________________________-->
2521 <a name="schanges_deleting">Deleting <tt>Instruction</tt>s</a>
2526 <p>Deleting an instruction from an existing sequence of instructions that form a
2527 <a href="#BasicBlock"><tt>BasicBlock</tt></a> is very straight-forward: just
2528 call the instruction's eraseFromParent() method. For example:</p>
2530 <div class="doc_code">
2532 <a href="#Instruction">Instruction</a> *I = .. ;
2533 I->eraseFromParent();
2537 <p>This unlinks the instruction from its containing basic block and deletes
2538 it. If you'd just like to unlink the instruction from its containing basic
2539 block but not delete it, you can use the <tt>removeFromParent()</tt> method.</p>
2543 <!--_______________________________________________________________________-->
2545 <a name="schanges_replacing">Replacing an <tt>Instruction</tt> with another
2551 <p><i>Replacing individual instructions</i></p>
2553 <p>Including "<a href="/doxygen/BasicBlockUtils_8h-source.html">llvm/Transforms/Utils/BasicBlockUtils.h</a>"
2554 permits use of two very useful replace functions: <tt>ReplaceInstWithValue</tt>
2555 and <tt>ReplaceInstWithInst</tt>.</p>
2557 <h5><a name="schanges_deleting">Deleting <tt>Instruction</tt>s</a></h5>
2560 <li><tt>ReplaceInstWithValue</tt>
2562 <p>This function replaces all uses of a given instruction with a value,
2563 and then removes the original instruction. The following example
2564 illustrates the replacement of the result of a particular
2565 <tt>AllocaInst</tt> that allocates memory for a single integer with a null
2566 pointer to an integer.</p>
2568 <div class="doc_code">
2570 AllocaInst* instToReplace = ...;
2571 BasicBlock::iterator ii(instToReplace);
2573 ReplaceInstWithValue(instToReplace->getParent()->getInstList(), ii,
2574 Constant::getNullValue(PointerType::getUnqual(Type::Int32Ty)));
2577 <li><tt>ReplaceInstWithInst</tt>
2579 <p>This function replaces a particular instruction with another
2580 instruction, inserting the new instruction into the basic block at the
2581 location where the old instruction was, and replacing any uses of the old
2582 instruction with the new instruction. The following example illustrates
2583 the replacement of one <tt>AllocaInst</tt> with another.</p>
2585 <div class="doc_code">
2587 AllocaInst* instToReplace = ...;
2588 BasicBlock::iterator ii(instToReplace);
2590 ReplaceInstWithInst(instToReplace->getParent()->getInstList(), ii,
2591 new AllocaInst(Type::Int32Ty, 0, "ptrToReplacedInt"));
2595 <p><i>Replacing multiple uses of <tt>User</tt>s and <tt>Value</tt>s</i></p>
2597 <p>You can use <tt>Value::replaceAllUsesWith</tt> and
2598 <tt>User::replaceUsesOfWith</tt> to change more than one use at a time. See the
2599 doxygen documentation for the <a href="/doxygen/classllvm_1_1Value.html">Value Class</a>
2600 and <a href="/doxygen/classllvm_1_1User.html">User Class</a>, respectively, for more
2603 <!-- Value::replaceAllUsesWith User::replaceUsesOfWith Point out:
2604 include/llvm/Transforms/Utils/ especially BasicBlockUtils.h with:
2605 ReplaceInstWithValue, ReplaceInstWithInst -->
2609 <!--_______________________________________________________________________-->
2611 <a name="schanges_deletingGV">Deleting <tt>GlobalVariable</tt>s</a>
2616 <p>Deleting a global variable from a module is just as easy as deleting an
2617 Instruction. First, you must have a pointer to the global variable that you wish
2618 to delete. You use this pointer to erase it from its parent, the module.
2621 <div class="doc_code">
2623 <a href="#GlobalVariable">GlobalVariable</a> *GV = .. ;
2625 GV->eraseFromParent();
2633 <!-- ======================================================================= -->
2635 <a name="create_types">How to Create Types</a>
2640 <p>In generating IR, you may need some complex types. If you know these types
2641 statically, you can use <tt>TypeBuilder<...>::get()</tt>, defined
2642 in <tt>llvm/Support/TypeBuilder.h</tt>, to retrieve them. <tt>TypeBuilder</tt>
2643 has two forms depending on whether you're building types for cross-compilation
2644 or native library use. <tt>TypeBuilder<T, true></tt> requires
2645 that <tt>T</tt> be independent of the host environment, meaning that it's built
2647 the <a href="/doxygen/namespacellvm_1_1types.html"><tt>llvm::types</tt></a>
2648 namespace and pointers, functions, arrays, etc. built of
2649 those. <tt>TypeBuilder<T, false></tt> additionally allows native C types
2650 whose size may depend on the host compiler. For example,</p>
2652 <div class="doc_code">
2654 FunctionType *ft = TypeBuilder<types::i<8>(types::i<32>*), true>::get();
2658 <p>is easier to read and write than the equivalent</p>
2660 <div class="doc_code">
2662 std::vector<const Type*> params;
2663 params.push_back(PointerType::getUnqual(Type::Int32Ty));
2664 FunctionType *ft = FunctionType::get(Type::Int8Ty, params, false);
2668 <p>See the <a href="/doxygen/TypeBuilder_8h-source.html#l00001">class
2669 comment</a> for more details.</p>
2675 <!-- *********************************************************************** -->
2677 <a name="threading">Threads and LLVM</a>
2679 <!-- *********************************************************************** -->
2683 This section describes the interaction of the LLVM APIs with multithreading,
2684 both on the part of client applications, and in the JIT, in the hosted
2689 Note that LLVM's support for multithreading is still relatively young. Up
2690 through version 2.5, the execution of threaded hosted applications was
2691 supported, but not threaded client access to the APIs. While this use case is
2692 now supported, clients <em>must</em> adhere to the guidelines specified below to
2693 ensure proper operation in multithreaded mode.
2697 Note that, on Unix-like platforms, LLVM requires the presence of GCC's atomic
2698 intrinsics in order to support threaded operation. If you need a
2699 multhreading-capable LLVM on a platform without a suitably modern system
2700 compiler, consider compiling LLVM and LLVM-GCC in single-threaded mode, and
2701 using the resultant compiler to build a copy of LLVM with multithreading
2705 <!-- ======================================================================= -->
2707 <a name="startmultithreaded">Entering and Exiting Multithreaded Mode</a>
2713 In order to properly protect its internal data structures while avoiding
2714 excessive locking overhead in the single-threaded case, the LLVM must intialize
2715 certain data structures necessary to provide guards around its internals. To do
2716 so, the client program must invoke <tt>llvm_start_multithreaded()</tt> before
2717 making any concurrent LLVM API calls. To subsequently tear down these
2718 structures, use the <tt>llvm_stop_multithreaded()</tt> call. You can also use
2719 the <tt>llvm_is_multithreaded()</tt> call to check the status of multithreaded
2724 Note that both of these calls must be made <em>in isolation</em>. That is to
2725 say that no other LLVM API calls may be executing at any time during the
2726 execution of <tt>llvm_start_multithreaded()</tt> or <tt>llvm_stop_multithreaded
2727 </tt>. It's is the client's responsibility to enforce this isolation.
2731 The return value of <tt>llvm_start_multithreaded()</tt> indicates the success or
2732 failure of the initialization. Failure typically indicates that your copy of
2733 LLVM was built without multithreading support, typically because GCC atomic
2734 intrinsics were not found in your system compiler. In this case, the LLVM API
2735 will not be safe for concurrent calls. However, it <em>will</em> be safe for
2736 hosting threaded applications in the JIT, though <a href="#jitthreading">care
2737 must be taken</a> to ensure that side exits and the like do not accidentally
2738 result in concurrent LLVM API calls.
2742 <!-- ======================================================================= -->
2744 <a name="shutdown">Ending Execution with <tt>llvm_shutdown()</tt></a>
2749 When you are done using the LLVM APIs, you should call <tt>llvm_shutdown()</tt>
2750 to deallocate memory used for internal structures. This will also invoke
2751 <tt>llvm_stop_multithreaded()</tt> if LLVM is operating in multithreaded mode.
2752 As such, <tt>llvm_shutdown()</tt> requires the same isolation guarantees as
2753 <tt>llvm_stop_multithreaded()</tt>.
2757 Note that, if you use scope-based shutdown, you can use the
2758 <tt>llvm_shutdown_obj</tt> class, which calls <tt>llvm_shutdown()</tt> in its
2762 <!-- ======================================================================= -->
2764 <a name="managedstatic">Lazy Initialization with <tt>ManagedStatic</tt></a>
2769 <tt>ManagedStatic</tt> is a utility class in LLVM used to implement static
2770 initialization of static resources, such as the global type tables. Before the
2771 invocation of <tt>llvm_shutdown()</tt>, it implements a simple lazy
2772 initialization scheme. Once <tt>llvm_start_multithreaded()</tt> returns,
2773 however, it uses double-checked locking to implement thread-safe lazy
2778 Note that, because no other threads are allowed to issue LLVM API calls before
2779 <tt>llvm_start_multithreaded()</tt> returns, it is possible to have
2780 <tt>ManagedStatic</tt>s of <tt>llvm::sys::Mutex</tt>s.
2784 The <tt>llvm_acquire_global_lock()</tt> and <tt>llvm_release_global_lock</tt>
2785 APIs provide access to the global lock used to implement the double-checked
2786 locking for lazy initialization. These should only be used internally to LLVM,
2787 and only if you know what you're doing!
2791 <!-- ======================================================================= -->
2793 <a name="llvmcontext">Achieving Isolation with <tt>LLVMContext</tt></a>
2798 <tt>LLVMContext</tt> is an opaque class in the LLVM API which clients can use
2799 to operate multiple, isolated instances of LLVM concurrently within the same
2800 address space. For instance, in a hypothetical compile-server, the compilation
2801 of an individual translation unit is conceptually independent from all the
2802 others, and it would be desirable to be able to compile incoming translation
2803 units concurrently on independent server threads. Fortunately,
2804 <tt>LLVMContext</tt> exists to enable just this kind of scenario!
2808 Conceptually, <tt>LLVMContext</tt> provides isolation. Every LLVM entity
2809 (<tt>Module</tt>s, <tt>Value</tt>s, <tt>Type</tt>s, <tt>Constant</tt>s, etc.)
2810 in LLVM's in-memory IR belongs to an <tt>LLVMContext</tt>. Entities in
2811 different contexts <em>cannot</em> interact with each other: <tt>Module</tt>s in
2812 different contexts cannot be linked together, <tt>Function</tt>s cannot be added
2813 to <tt>Module</tt>s in different contexts, etc. What this means is that is is
2814 safe to compile on multiple threads simultaneously, as long as no two threads
2815 operate on entities within the same context.
2819 In practice, very few places in the API require the explicit specification of a
2820 <tt>LLVMContext</tt>, other than the <tt>Type</tt> creation/lookup APIs.
2821 Because every <tt>Type</tt> carries a reference to its owning context, most
2822 other entities can determine what context they belong to by looking at their
2823 own <tt>Type</tt>. If you are adding new entities to LLVM IR, please try to
2824 maintain this interface design.
2828 For clients that do <em>not</em> require the benefits of isolation, LLVM
2829 provides a convenience API <tt>getGlobalContext()</tt>. This returns a global,
2830 lazily initialized <tt>LLVMContext</tt> that may be used in situations where
2831 isolation is not a concern.
2835 <!-- ======================================================================= -->
2837 <a name="jitthreading">Threads and the JIT</a>
2842 LLVM's "eager" JIT compiler is safe to use in threaded programs. Multiple
2843 threads can call <tt>ExecutionEngine::getPointerToFunction()</tt> or
2844 <tt>ExecutionEngine::runFunction()</tt> concurrently, and multiple threads can
2845 run code output by the JIT concurrently. The user must still ensure that only
2846 one thread accesses IR in a given <tt>LLVMContext</tt> while another thread
2847 might be modifying it. One way to do that is to always hold the JIT lock while
2848 accessing IR outside the JIT (the JIT <em>modifies</em> the IR by adding
2849 <tt>CallbackVH</tt>s). Another way is to only
2850 call <tt>getPointerToFunction()</tt> from the <tt>LLVMContext</tt>'s thread.
2853 <p>When the JIT is configured to compile lazily (using
2854 <tt>ExecutionEngine::DisableLazyCompilation(false)</tt>), there is currently a
2855 <a href="http://llvm.org/bugs/show_bug.cgi?id=5184">race condition</a> in
2856 updating call sites after a function is lazily-jitted. It's still possible to
2857 use the lazy JIT in a threaded program if you ensure that only one thread at a
2858 time can call any particular lazy stub and that the JIT lock guards any IR
2859 access, but we suggest using only the eager JIT in threaded programs.
2865 <!-- *********************************************************************** -->
2867 <a name="advanced">Advanced Topics</a>
2869 <!-- *********************************************************************** -->
2873 This section describes some of the advanced or obscure API's that most clients
2874 do not need to be aware of. These API's tend manage the inner workings of the
2875 LLVM system, and only need to be accessed in unusual circumstances.
2879 <!-- ======================================================================= -->
2881 <a name="SymbolTable">The <tt>ValueSymbolTable</tt> class</a>
2885 <p>The <tt><a href="http://llvm.org/doxygen/classllvm_1_1ValueSymbolTable.html">
2886 ValueSymbolTable</a></tt> class provides a symbol table that the <a
2887 href="#Function"><tt>Function</tt></a> and <a href="#Module">
2888 <tt>Module</tt></a> classes use for naming value definitions. The symbol table
2889 can provide a name for any <a href="#Value"><tt>Value</tt></a>.
2892 <p>Note that the <tt>SymbolTable</tt> class should not be directly accessed
2893 by most clients. It should only be used when iteration over the symbol table
2894 names themselves are required, which is very special purpose. Note that not
2896 <tt><a href="#Value">Value</a></tt>s have names, and those without names (i.e. they have
2897 an empty name) do not exist in the symbol table.
2900 <p>Symbol tables support iteration over the values in the symbol
2901 table with <tt>begin/end/iterator</tt> and supports querying to see if a
2902 specific name is in the symbol table (with <tt>lookup</tt>). The
2903 <tt>ValueSymbolTable</tt> class exposes no public mutator methods, instead,
2904 simply call <tt>setName</tt> on a value, which will autoinsert it into the
2905 appropriate symbol table.</p>
2911 <!-- ======================================================================= -->
2913 <a name="UserLayout">The <tt>User</tt> and owned <tt>Use</tt> classes' memory layout</a>
2917 <p>The <tt><a href="http://llvm.org/doxygen/classllvm_1_1User.html">
2918 User</a></tt> class provides a basis for expressing the ownership of <tt>User</tt>
2919 towards other <tt><a href="http://llvm.org/doxygen/classllvm_1_1Value.html">
2920 Value</a></tt>s. The <tt><a href="http://llvm.org/doxygen/classllvm_1_1Use.html">
2921 Use</a></tt> helper class is employed to do the bookkeeping and to facilitate <i>O(1)</i>
2922 addition and removal.</p>
2924 <!-- ______________________________________________________________________ -->
2927 Interaction and relationship between <tt>User</tt> and <tt>Use</tt> objects
2933 A subclass of <tt>User</tt> can choose between incorporating its <tt>Use</tt> objects
2934 or refer to them out-of-line by means of a pointer. A mixed variant
2935 (some <tt>Use</tt>s inline others hung off) is impractical and breaks the invariant
2936 that the <tt>Use</tt> objects belonging to the same <tt>User</tt> form a contiguous array.
2940 We have 2 different layouts in the <tt>User</tt> (sub)classes:
2943 The <tt>Use</tt> object(s) are inside (resp. at fixed offset) of the <tt>User</tt>
2944 object and there are a fixed number of them.</p>
2947 The <tt>Use</tt> object(s) are referenced by a pointer to an
2948 array from the <tt>User</tt> object and there may be a variable
2952 As of v2.4 each layout still possesses a direct pointer to the
2953 start of the array of <tt>Use</tt>s. Though not mandatory for layout a),
2954 we stick to this redundancy for the sake of simplicity.
2955 The <tt>User</tt> object also stores the number of <tt>Use</tt> objects it
2956 has. (Theoretically this information can also be calculated
2957 given the scheme presented below.)</p>
2959 Special forms of allocation operators (<tt>operator new</tt>)
2960 enforce the following memory layouts:</p>
2963 <li><p>Layout a) is modelled by prepending the <tt>User</tt> object by the <tt>Use[]</tt> array.</p>
2966 ...---.---.---.---.-------...
2967 | P | P | P | P | User
2968 '''---'---'---'---'-------'''
2971 <li><p>Layout b) is modelled by pointing at the <tt>Use[]</tt> array.</p>
2983 <i>(In the above figures '<tt>P</tt>' stands for the <tt>Use**</tt> that
2984 is stored in each <tt>Use</tt> object in the member <tt>Use::Prev</tt>)</i>
2988 <!-- ______________________________________________________________________ -->
2990 <a name="Waymarking">The waymarking algorithm</a>
2995 Since the <tt>Use</tt> objects are deprived of the direct (back)pointer to
2996 their <tt>User</tt> objects, there must be a fast and exact method to
2997 recover it. This is accomplished by the following scheme:</p>
2999 A bit-encoding in the 2 LSBits (least significant bits) of the <tt>Use::Prev</tt> allows to find the
3000 start of the <tt>User</tt> object:
3002 <li><tt>00</tt> —> binary digit 0</li>
3003 <li><tt>01</tt> —> binary digit 1</li>
3004 <li><tt>10</tt> —> stop and calculate (<tt>s</tt>)</li>
3005 <li><tt>11</tt> —> full stop (<tt>S</tt>)</li>
3008 Given a <tt>Use*</tt>, all we have to do is to walk till we get
3009 a stop and we either have a <tt>User</tt> immediately behind or
3010 we have to walk to the next stop picking up digits
3011 and calculating the offset:</p>
3013 .---.---.---.---.---.---.---.---.---.---.---.---.---.---.---.---.----------------
3014 | 1 | s | 1 | 0 | 1 | 0 | s | 1 | 1 | 0 | s | 1 | 1 | s | 1 | S | User (or User*)
3015 '---'---'---'---'---'---'---'---'---'---'---'---'---'---'---'---'----------------
3016 |+15 |+10 |+6 |+3 |+1
3019 | | |______________________>
3020 | |______________________________________>
3021 |__________________________________________________________>
3024 Only the significant number of bits need to be stored between the
3025 stops, so that the <i>worst case is 20 memory accesses</i> when there are
3026 1000 <tt>Use</tt> objects associated with a <tt>User</tt>.</p>
3030 <!-- ______________________________________________________________________ -->
3032 <a name="ReferenceImpl">Reference implementation</a>
3037 The following literate Haskell fragment demonstrates the concept:</p>
3039 <div class="doc_code">
3041 > import Test.QuickCheck
3043 > digits :: Int -> [Char] -> [Char]
3044 > digits 0 acc = '0' : acc
3045 > digits 1 acc = '1' : acc
3046 > digits n acc = digits (n `div` 2) $ digits (n `mod` 2) acc
3048 > dist :: Int -> [Char] -> [Char]
3051 > dist 1 acc = let r = dist 0 acc in 's' : digits (length r) r
3052 > dist n acc = dist (n - 1) $ dist 1 acc
3054 > takeLast n ss = reverse $ take n $ reverse ss
3056 > test = takeLast 40 $ dist 20 []
3061 Printing <test> gives: <tt>"1s100000s11010s10100s1111s1010s110s11s1S"</tt></p>
3063 The reverse algorithm computes the length of the string just by examining
3064 a certain prefix:</p>
3066 <div class="doc_code">
3068 > pref :: [Char] -> Int
3070 > pref ('s':'1':rest) = decode 2 1 rest
3071 > pref (_:rest) = 1 + pref rest
3073 > decode walk acc ('0':rest) = decode (walk + 1) (acc * 2) rest
3074 > decode walk acc ('1':rest) = decode (walk + 1) (acc * 2 + 1) rest
3075 > decode walk acc _ = walk + acc
3080 Now, as expected, printing <pref test> gives <tt>40</tt>.</p>
3082 We can <i>quickCheck</i> this with following property:</p>
3084 <div class="doc_code">
3086 > testcase = dist 2000 []
3087 > testcaseLength = length testcase
3089 > identityProp n = n > 0 && n <= testcaseLength ==> length arr == pref arr
3090 > where arr = takeLast n testcase
3095 As expected <quickCheck identityProp> gives:</p>
3098 *Main> quickCheck identityProp
3099 OK, passed 100 tests.
3102 Let's be a bit more exhaustive:</p>
3104 <div class="doc_code">
3107 > deepCheck p = check (defaultConfig { configMaxTest = 500 }) p
3112 And here is the result of <deepCheck identityProp>:</p>
3115 *Main> deepCheck identityProp
3116 OK, passed 500 tests.
3121 <!-- ______________________________________________________________________ -->
3123 <a name="Tagging">Tagging considerations</a>
3129 To maintain the invariant that the 2 LSBits of each <tt>Use**</tt> in <tt>Use</tt>
3130 never change after being set up, setters of <tt>Use::Prev</tt> must re-tag the
3131 new <tt>Use**</tt> on every modification. Accordingly getters must strip the
3134 For layout b) instead of the <tt>User</tt> we find a pointer (<tt>User*</tt> with LSBit set).
3135 Following this pointer brings us to the <tt>User</tt>. A portable trick ensures
3136 that the first bytes of <tt>User</tt> (if interpreted as a pointer) never has
3137 the LSBit set. (Portability is relying on the fact that all known compilers place the
3138 <tt>vptr</tt> in the first word of the instances.)</p>
3146 <!-- *********************************************************************** -->
3148 <a name="coreclasses">The Core LLVM Class Hierarchy Reference </a>
3150 <!-- *********************************************************************** -->
3153 <p><tt>#include "<a href="/doxygen/Type_8h-source.html">llvm/Type.h</a>"</tt>
3154 <br>doxygen info: <a href="/doxygen/classllvm_1_1Type.html">Type Class</a></p>
3156 <p>The Core LLVM classes are the primary means of representing the program
3157 being inspected or transformed. The core LLVM classes are defined in
3158 header files in the <tt>include/llvm/</tt> directory, and implemented in
3159 the <tt>lib/VMCore</tt> directory.</p>
3161 <!-- ======================================================================= -->
3163 <a name="Type">The <tt>Type</tt> class and Derived Types</a>
3168 <p><tt>Type</tt> is a superclass of all type classes. Every <tt>Value</tt> has
3169 a <tt>Type</tt>. <tt>Type</tt> cannot be instantiated directly but only
3170 through its subclasses. Certain primitive types (<tt>VoidType</tt>,
3171 <tt>LabelType</tt>, <tt>FloatType</tt> and <tt>DoubleType</tt>) have hidden
3172 subclasses. They are hidden because they offer no useful functionality beyond
3173 what the <tt>Type</tt> class offers except to distinguish themselves from
3174 other subclasses of <tt>Type</tt>.</p>
3175 <p>All other types are subclasses of <tt>DerivedType</tt>. Types can be
3176 named, but this is not a requirement. There exists exactly
3177 one instance of a given shape at any one time. This allows type equality to
3178 be performed with address equality of the Type Instance. That is, given two
3179 <tt>Type*</tt> values, the types are identical if the pointers are identical.
3182 <!-- _______________________________________________________________________ -->
3184 <a name="m_Type">Important Public Methods</a>
3190 <li><tt>bool isIntegerTy() const</tt>: Returns true for any integer type.</li>
3192 <li><tt>bool isFloatingPointTy()</tt>: Return true if this is one of the five
3193 floating point types.</li>
3195 <li><tt>bool isSized()</tt>: Return true if the type has known size. Things
3196 that don't have a size are abstract types, labels and void.</li>
3201 <!-- _______________________________________________________________________ -->
3203 <a name="derivedtypes">Important Derived Types</a>
3207 <dt><tt>IntegerType</tt></dt>
3208 <dd>Subclass of DerivedType that represents integer types of any bit width.
3209 Any bit width between <tt>IntegerType::MIN_INT_BITS</tt> (1) and
3210 <tt>IntegerType::MAX_INT_BITS</tt> (~8 million) can be represented.
3212 <li><tt>static const IntegerType* get(unsigned NumBits)</tt>: get an integer
3213 type of a specific bit width.</li>
3214 <li><tt>unsigned getBitWidth() const</tt>: Get the bit width of an integer
3218 <dt><tt>SequentialType</tt></dt>
3219 <dd>This is subclassed by ArrayType, PointerType and VectorType.
3221 <li><tt>const Type * getElementType() const</tt>: Returns the type of each
3222 of the elements in the sequential type. </li>
3225 <dt><tt>ArrayType</tt></dt>
3226 <dd>This is a subclass of SequentialType and defines the interface for array
3229 <li><tt>unsigned getNumElements() const</tt>: Returns the number of
3230 elements in the array. </li>
3233 <dt><tt>PointerType</tt></dt>
3234 <dd>Subclass of SequentialType for pointer types.</dd>
3235 <dt><tt>VectorType</tt></dt>
3236 <dd>Subclass of SequentialType for vector types. A
3237 vector type is similar to an ArrayType but is distinguished because it is
3238 a first class type whereas ArrayType is not. Vector types are used for
3239 vector operations and are usually small vectors of of an integer or floating
3241 <dt><tt>StructType</tt></dt>
3242 <dd>Subclass of DerivedTypes for struct types.</dd>
3243 <dt><tt><a name="FunctionType">FunctionType</a></tt></dt>
3244 <dd>Subclass of DerivedTypes for function types.
3246 <li><tt>bool isVarArg() const</tt>: Returns true if it's a vararg
3248 <li><tt> const Type * getReturnType() const</tt>: Returns the
3249 return type of the function.</li>
3250 <li><tt>const Type * getParamType (unsigned i)</tt>: Returns
3251 the type of the ith parameter.</li>
3252 <li><tt> const unsigned getNumParams() const</tt>: Returns the
3253 number of formal parameters.</li>
3261 <!-- ======================================================================= -->
3263 <a name="Module">The <tt>Module</tt> class</a>
3269 href="/doxygen/Module_8h-source.html">llvm/Module.h</a>"</tt><br> doxygen info:
3270 <a href="/doxygen/classllvm_1_1Module.html">Module Class</a></p>
3272 <p>The <tt>Module</tt> class represents the top level structure present in LLVM
3273 programs. An LLVM module is effectively either a translation unit of the
3274 original program or a combination of several translation units merged by the
3275 linker. The <tt>Module</tt> class keeps track of a list of <a
3276 href="#Function"><tt>Function</tt></a>s, a list of <a
3277 href="#GlobalVariable"><tt>GlobalVariable</tt></a>s, and a <a
3278 href="#SymbolTable"><tt>SymbolTable</tt></a>. Additionally, it contains a few
3279 helpful member functions that try to make common operations easy.</p>
3281 <!-- _______________________________________________________________________ -->
3283 <a name="m_Module">Important Public Members of the <tt>Module</tt> class</a>
3289 <li><tt>Module::Module(std::string name = "")</tt></li>
3292 <p>Constructing a <a href="#Module">Module</a> is easy. You can optionally
3293 provide a name for it (probably based on the name of the translation unit).</p>
3296 <li><tt>Module::iterator</tt> - Typedef for function list iterator<br>
3297 <tt>Module::const_iterator</tt> - Typedef for const_iterator.<br>
3299 <tt>begin()</tt>, <tt>end()</tt>
3300 <tt>size()</tt>, <tt>empty()</tt>
3302 <p>These are forwarding methods that make it easy to access the contents of
3303 a <tt>Module</tt> object's <a href="#Function"><tt>Function</tt></a>
3306 <li><tt>Module::FunctionListType &getFunctionList()</tt>
3308 <p> Returns the list of <a href="#Function"><tt>Function</tt></a>s. This is
3309 necessary to use when you need to update the list or perform a complex
3310 action that doesn't have a forwarding method.</p>
3312 <p><!-- Global Variable --></p></li>
3318 <li><tt>Module::global_iterator</tt> - Typedef for global variable list iterator<br>
3320 <tt>Module::const_global_iterator</tt> - Typedef for const_iterator.<br>
3322 <tt>global_begin()</tt>, <tt>global_end()</tt>
3323 <tt>global_size()</tt>, <tt>global_empty()</tt>
3325 <p> These are forwarding methods that make it easy to access the contents of
3326 a <tt>Module</tt> object's <a
3327 href="#GlobalVariable"><tt>GlobalVariable</tt></a> list.</p></li>
3329 <li><tt>Module::GlobalListType &getGlobalList()</tt>
3331 <p>Returns the list of <a
3332 href="#GlobalVariable"><tt>GlobalVariable</tt></a>s. This is necessary to
3333 use when you need to update the list or perform a complex action that
3334 doesn't have a forwarding method.</p>
3336 <p><!-- Symbol table stuff --> </p></li>
3342 <li><tt><a href="#SymbolTable">SymbolTable</a> *getSymbolTable()</tt>
3344 <p>Return a reference to the <a href="#SymbolTable"><tt>SymbolTable</tt></a>
3345 for this <tt>Module</tt>.</p>
3347 <p><!-- Convenience methods --></p></li>
3353 <li><tt><a href="#Function">Function</a> *getFunction(const std::string
3354 &Name, const <a href="#FunctionType">FunctionType</a> *Ty)</tt>
3356 <p>Look up the specified function in the <tt>Module</tt> <a
3357 href="#SymbolTable"><tt>SymbolTable</tt></a>. If it does not exist, return
3358 <tt>null</tt>.</p></li>
3360 <li><tt><a href="#Function">Function</a> *getOrInsertFunction(const
3361 std::string &Name, const <a href="#FunctionType">FunctionType</a> *T)</tt>
3363 <p>Look up the specified function in the <tt>Module</tt> <a
3364 href="#SymbolTable"><tt>SymbolTable</tt></a>. If it does not exist, add an
3365 external declaration for the function and return it.</p></li>
3367 <li><tt>std::string getTypeName(const <a href="#Type">Type</a> *Ty)</tt>
3369 <p>If there is at least one entry in the <a
3370 href="#SymbolTable"><tt>SymbolTable</tt></a> for the specified <a
3371 href="#Type"><tt>Type</tt></a>, return it. Otherwise return the empty
3374 <li><tt>bool addTypeName(const std::string &Name, const <a
3375 href="#Type">Type</a> *Ty)</tt>
3377 <p>Insert an entry in the <a href="#SymbolTable"><tt>SymbolTable</tt></a>
3378 mapping <tt>Name</tt> to <tt>Ty</tt>. If there is already an entry for this
3379 name, true is returned and the <a
3380 href="#SymbolTable"><tt>SymbolTable</tt></a> is not modified.</p></li>
3387 <!-- ======================================================================= -->
3389 <a name="Value">The <tt>Value</tt> class</a>
3394 <p><tt>#include "<a href="/doxygen/Value_8h-source.html">llvm/Value.h</a>"</tt>
3396 doxygen info: <a href="/doxygen/classllvm_1_1Value.html">Value Class</a></p>
3398 <p>The <tt>Value</tt> class is the most important class in the LLVM Source
3399 base. It represents a typed value that may be used (among other things) as an
3400 operand to an instruction. There are many different types of <tt>Value</tt>s,
3401 such as <a href="#Constant"><tt>Constant</tt></a>s,<a
3402 href="#Argument"><tt>Argument</tt></a>s. Even <a
3403 href="#Instruction"><tt>Instruction</tt></a>s and <a
3404 href="#Function"><tt>Function</tt></a>s are <tt>Value</tt>s.</p>
3406 <p>A particular <tt>Value</tt> may be used many times in the LLVM representation
3407 for a program. For example, an incoming argument to a function (represented
3408 with an instance of the <a href="#Argument">Argument</a> class) is "used" by
3409 every instruction in the function that references the argument. To keep track
3410 of this relationship, the <tt>Value</tt> class keeps a list of all of the <a
3411 href="#User"><tt>User</tt></a>s that is using it (the <a
3412 href="#User"><tt>User</tt></a> class is a base class for all nodes in the LLVM
3413 graph that can refer to <tt>Value</tt>s). This use list is how LLVM represents
3414 def-use information in the program, and is accessible through the <tt>use_</tt>*
3415 methods, shown below.</p>
3417 <p>Because LLVM is a typed representation, every LLVM <tt>Value</tt> is typed,
3418 and this <a href="#Type">Type</a> is available through the <tt>getType()</tt>
3419 method. In addition, all LLVM values can be named. The "name" of the
3420 <tt>Value</tt> is a symbolic string printed in the LLVM code:</p>
3422 <div class="doc_code">
3424 %<b>foo</b> = add i32 1, 2
3428 <p><a name="nameWarning">The name of this instruction is "foo".</a> <b>NOTE</b>
3429 that the name of any value may be missing (an empty string), so names should
3430 <b>ONLY</b> be used for debugging (making the source code easier to read,
3431 debugging printouts), they should not be used to keep track of values or map
3432 between them. For this purpose, use a <tt>std::map</tt> of pointers to the
3433 <tt>Value</tt> itself instead.</p>
3435 <p>One important aspect of LLVM is that there is no distinction between an SSA
3436 variable and the operation that produces it. Because of this, any reference to
3437 the value produced by an instruction (or the value available as an incoming
3438 argument, for example) is represented as a direct pointer to the instance of
3440 represents this value. Although this may take some getting used to, it
3441 simplifies the representation and makes it easier to manipulate.</p>
3443 <!-- _______________________________________________________________________ -->
3445 <a name="m_Value">Important Public Members of the <tt>Value</tt> class</a>
3451 <li><tt>Value::use_iterator</tt> - Typedef for iterator over the
3453 <tt>Value::const_use_iterator</tt> - Typedef for const_iterator over
3455 <tt>unsigned use_size()</tt> - Returns the number of users of the
3457 <tt>bool use_empty()</tt> - Returns true if there are no users.<br>
3458 <tt>use_iterator use_begin()</tt> - Get an iterator to the start of
3460 <tt>use_iterator use_end()</tt> - Get an iterator to the end of the
3462 <tt><a href="#User">User</a> *use_back()</tt> - Returns the last
3463 element in the list.
3464 <p> These methods are the interface to access the def-use
3465 information in LLVM. As with all other iterators in LLVM, the naming
3466 conventions follow the conventions defined by the <a href="#stl">STL</a>.</p>
3468 <li><tt><a href="#Type">Type</a> *getType() const</tt>
3469 <p>This method returns the Type of the Value.</p>
3471 <li><tt>bool hasName() const</tt><br>
3472 <tt>std::string getName() const</tt><br>
3473 <tt>void setName(const std::string &Name)</tt>
3474 <p> This family of methods is used to access and assign a name to a <tt>Value</tt>,
3475 be aware of the <a href="#nameWarning">precaution above</a>.</p>
3477 <li><tt>void replaceAllUsesWith(Value *V)</tt>
3479 <p>This method traverses the use list of a <tt>Value</tt> changing all <a
3480 href="#User"><tt>User</tt>s</a> of the current value to refer to
3481 "<tt>V</tt>" instead. For example, if you detect that an instruction always
3482 produces a constant value (for example through constant folding), you can
3483 replace all uses of the instruction with the constant like this:</p>
3485 <div class="doc_code">
3487 Inst->replaceAllUsesWith(ConstVal);
3497 <!-- ======================================================================= -->
3499 <a name="User">The <tt>User</tt> class</a>
3505 <tt>#include "<a href="/doxygen/User_8h-source.html">llvm/User.h</a>"</tt><br>
3506 doxygen info: <a href="/doxygen/classllvm_1_1User.html">User Class</a><br>
3507 Superclass: <a href="#Value"><tt>Value</tt></a></p>
3509 <p>The <tt>User</tt> class is the common base class of all LLVM nodes that may
3510 refer to <a href="#Value"><tt>Value</tt></a>s. It exposes a list of "Operands"
3511 that are all of the <a href="#Value"><tt>Value</tt></a>s that the User is
3512 referring to. The <tt>User</tt> class itself is a subclass of
3515 <p>The operands of a <tt>User</tt> point directly to the LLVM <a
3516 href="#Value"><tt>Value</tt></a> that it refers to. Because LLVM uses Static
3517 Single Assignment (SSA) form, there can only be one definition referred to,
3518 allowing this direct connection. This connection provides the use-def
3519 information in LLVM.</p>
3521 <!-- _______________________________________________________________________ -->
3523 <a name="m_User">Important Public Members of the <tt>User</tt> class</a>
3528 <p>The <tt>User</tt> class exposes the operand list in two ways: through
3529 an index access interface and through an iterator based interface.</p>
3532 <li><tt>Value *getOperand(unsigned i)</tt><br>
3533 <tt>unsigned getNumOperands()</tt>
3534 <p> These two methods expose the operands of the <tt>User</tt> in a
3535 convenient form for direct access.</p></li>
3537 <li><tt>User::op_iterator</tt> - Typedef for iterator over the operand
3539 <tt>op_iterator op_begin()</tt> - Get an iterator to the start of
3540 the operand list.<br>
3541 <tt>op_iterator op_end()</tt> - Get an iterator to the end of the
3543 <p> Together, these methods make up the iterator based interface to
3544 the operands of a <tt>User</tt>.</p></li>
3551 <!-- ======================================================================= -->
3553 <a name="Instruction">The <tt>Instruction</tt> class</a>
3558 <p><tt>#include "</tt><tt><a
3559 href="/doxygen/Instruction_8h-source.html">llvm/Instruction.h</a>"</tt><br>
3560 doxygen info: <a href="/doxygen/classllvm_1_1Instruction.html">Instruction Class</a><br>
3561 Superclasses: <a href="#User"><tt>User</tt></a>, <a
3562 href="#Value"><tt>Value</tt></a></p>
3564 <p>The <tt>Instruction</tt> class is the common base class for all LLVM
3565 instructions. It provides only a few methods, but is a very commonly used
3566 class. The primary data tracked by the <tt>Instruction</tt> class itself is the
3567 opcode (instruction type) and the parent <a
3568 href="#BasicBlock"><tt>BasicBlock</tt></a> the <tt>Instruction</tt> is embedded
3569 into. To represent a specific type of instruction, one of many subclasses of
3570 <tt>Instruction</tt> are used.</p>
3572 <p> Because the <tt>Instruction</tt> class subclasses the <a
3573 href="#User"><tt>User</tt></a> class, its operands can be accessed in the same
3574 way as for other <a href="#User"><tt>User</tt></a>s (with the
3575 <tt>getOperand()</tt>/<tt>getNumOperands()</tt> and
3576 <tt>op_begin()</tt>/<tt>op_end()</tt> methods).</p> <p> An important file for
3577 the <tt>Instruction</tt> class is the <tt>llvm/Instruction.def</tt> file. This
3578 file contains some meta-data about the various different types of instructions
3579 in LLVM. It describes the enum values that are used as opcodes (for example
3580 <tt>Instruction::Add</tt> and <tt>Instruction::ICmp</tt>), as well as the
3581 concrete sub-classes of <tt>Instruction</tt> that implement the instruction (for
3582 example <tt><a href="#BinaryOperator">BinaryOperator</a></tt> and <tt><a
3583 href="#CmpInst">CmpInst</a></tt>). Unfortunately, the use of macros in
3584 this file confuses doxygen, so these enum values don't show up correctly in the
3585 <a href="/doxygen/classllvm_1_1Instruction.html">doxygen output</a>.</p>
3587 <!-- _______________________________________________________________________ -->
3589 <a name="s_Instruction">
3590 Important Subclasses of the <tt>Instruction</tt> class
3595 <li><tt><a name="BinaryOperator">BinaryOperator</a></tt>
3596 <p>This subclasses represents all two operand instructions whose operands
3597 must be the same type, except for the comparison instructions.</p></li>
3598 <li><tt><a name="CastInst">CastInst</a></tt>
3599 <p>This subclass is the parent of the 12 casting instructions. It provides
3600 common operations on cast instructions.</p>
3601 <li><tt><a name="CmpInst">CmpInst</a></tt>
3602 <p>This subclass respresents the two comparison instructions,
3603 <a href="LangRef.html#i_icmp">ICmpInst</a> (integer opreands), and
3604 <a href="LangRef.html#i_fcmp">FCmpInst</a> (floating point operands).</p>
3605 <li><tt><a name="TerminatorInst">TerminatorInst</a></tt>
3606 <p>This subclass is the parent of all terminator instructions (those which
3607 can terminate a block).</p>
3611 <!-- _______________________________________________________________________ -->
3613 <a name="m_Instruction">
3614 Important Public Members of the <tt>Instruction</tt> class
3621 <li><tt><a href="#BasicBlock">BasicBlock</a> *getParent()</tt>
3622 <p>Returns the <a href="#BasicBlock"><tt>BasicBlock</tt></a> that
3623 this <tt>Instruction</tt> is embedded into.</p></li>
3624 <li><tt>bool mayWriteToMemory()</tt>
3625 <p>Returns true if the instruction writes to memory, i.e. it is a
3626 <tt>call</tt>,<tt>free</tt>,<tt>invoke</tt>, or <tt>store</tt>.</p></li>
3627 <li><tt>unsigned getOpcode()</tt>
3628 <p>Returns the opcode for the <tt>Instruction</tt>.</p></li>
3629 <li><tt><a href="#Instruction">Instruction</a> *clone() const</tt>
3630 <p>Returns another instance of the specified instruction, identical
3631 in all ways to the original except that the instruction has no parent
3632 (ie it's not embedded into a <a href="#BasicBlock"><tt>BasicBlock</tt></a>),
3633 and it has no name</p></li>
3640 <!-- ======================================================================= -->
3642 <a name="Constant">The <tt>Constant</tt> class and subclasses</a>
3647 <p>Constant represents a base class for different types of constants. It
3648 is subclassed by ConstantInt, ConstantArray, etc. for representing
3649 the various types of Constants. <a href="#GlobalValue">GlobalValue</a> is also
3650 a subclass, which represents the address of a global variable or function.
3653 <!-- _______________________________________________________________________ -->
3654 <h4>Important Subclasses of Constant</h4>
3657 <li>ConstantInt : This subclass of Constant represents an integer constant of
3660 <li><tt>const APInt& getValue() const</tt>: Returns the underlying
3661 value of this constant, an APInt value.</li>
3662 <li><tt>int64_t getSExtValue() const</tt>: Converts the underlying APInt
3663 value to an int64_t via sign extension. If the value (not the bit width)
3664 of the APInt is too large to fit in an int64_t, an assertion will result.
3665 For this reason, use of this method is discouraged.</li>
3666 <li><tt>uint64_t getZExtValue() const</tt>: Converts the underlying APInt
3667 value to a uint64_t via zero extension. IF the value (not the bit width)
3668 of the APInt is too large to fit in a uint64_t, an assertion will result.
3669 For this reason, use of this method is discouraged.</li>
3670 <li><tt>static ConstantInt* get(const APInt& Val)</tt>: Returns the
3671 ConstantInt object that represents the value provided by <tt>Val</tt>.
3672 The type is implied as the IntegerType that corresponds to the bit width
3673 of <tt>Val</tt>.</li>
3674 <li><tt>static ConstantInt* get(const Type *Ty, uint64_t Val)</tt>:
3675 Returns the ConstantInt object that represents the value provided by
3676 <tt>Val</tt> for integer type <tt>Ty</tt>.</li>
3679 <li>ConstantFP : This class represents a floating point constant.
3681 <li><tt>double getValue() const</tt>: Returns the underlying value of
3682 this constant. </li>
3685 <li>ConstantArray : This represents a constant array.
3687 <li><tt>const std::vector<Use> &getValues() const</tt>: Returns
3688 a vector of component constants that makeup this array. </li>
3691 <li>ConstantStruct : This represents a constant struct.
3693 <li><tt>const std::vector<Use> &getValues() const</tt>: Returns
3694 a vector of component constants that makeup this array. </li>
3697 <li>GlobalValue : This represents either a global variable or a function. In
3698 either case, the value is a constant fixed address (after linking).
3705 <!-- ======================================================================= -->
3707 <a name="GlobalValue">The <tt>GlobalValue</tt> class</a>
3713 href="/doxygen/GlobalValue_8h-source.html">llvm/GlobalValue.h</a>"</tt><br>
3714 doxygen info: <a href="/doxygen/classllvm_1_1GlobalValue.html">GlobalValue
3716 Superclasses: <a href="#Constant"><tt>Constant</tt></a>,
3717 <a href="#User"><tt>User</tt></a>, <a href="#Value"><tt>Value</tt></a></p>
3719 <p>Global values (<a href="#GlobalVariable"><tt>GlobalVariable</tt></a>s or <a
3720 href="#Function"><tt>Function</tt></a>s) are the only LLVM values that are
3721 visible in the bodies of all <a href="#Function"><tt>Function</tt></a>s.
3722 Because they are visible at global scope, they are also subject to linking with
3723 other globals defined in different translation units. To control the linking
3724 process, <tt>GlobalValue</tt>s know their linkage rules. Specifically,
3725 <tt>GlobalValue</tt>s know whether they have internal or external linkage, as
3726 defined by the <tt>LinkageTypes</tt> enumeration.</p>
3728 <p>If a <tt>GlobalValue</tt> has internal linkage (equivalent to being
3729 <tt>static</tt> in C), it is not visible to code outside the current translation
3730 unit, and does not participate in linking. If it has external linkage, it is
3731 visible to external code, and does participate in linking. In addition to
3732 linkage information, <tt>GlobalValue</tt>s keep track of which <a
3733 href="#Module"><tt>Module</tt></a> they are currently part of.</p>
3735 <p>Because <tt>GlobalValue</tt>s are memory objects, they are always referred to
3736 by their <b>address</b>. As such, the <a href="#Type"><tt>Type</tt></a> of a
3737 global is always a pointer to its contents. It is important to remember this
3738 when using the <tt>GetElementPtrInst</tt> instruction because this pointer must
3739 be dereferenced first. For example, if you have a <tt>GlobalVariable</tt> (a
3740 subclass of <tt>GlobalValue)</tt> that is an array of 24 ints, type <tt>[24 x
3741 i32]</tt>, then the <tt>GlobalVariable</tt> is a pointer to that array. Although
3742 the address of the first element of this array and the value of the
3743 <tt>GlobalVariable</tt> are the same, they have different types. The
3744 <tt>GlobalVariable</tt>'s type is <tt>[24 x i32]</tt>. The first element's type
3745 is <tt>i32.</tt> Because of this, accessing a global value requires you to
3746 dereference the pointer with <tt>GetElementPtrInst</tt> first, then its elements
3747 can be accessed. This is explained in the <a href="LangRef.html#globalvars">LLVM
3748 Language Reference Manual</a>.</p>
3750 <!-- _______________________________________________________________________ -->
3752 <a name="m_GlobalValue">
3753 Important Public Members of the <tt>GlobalValue</tt> class
3760 <li><tt>bool hasInternalLinkage() const</tt><br>
3761 <tt>bool hasExternalLinkage() const</tt><br>
3762 <tt>void setInternalLinkage(bool HasInternalLinkage)</tt>
3763 <p> These methods manipulate the linkage characteristics of the <tt>GlobalValue</tt>.</p>
3766 <li><tt><a href="#Module">Module</a> *getParent()</tt>
3767 <p> This returns the <a href="#Module"><tt>Module</tt></a> that the
3768 GlobalValue is currently embedded into.</p></li>
3775 <!-- ======================================================================= -->
3777 <a name="Function">The <tt>Function</tt> class</a>
3783 href="/doxygen/Function_8h-source.html">llvm/Function.h</a>"</tt><br> doxygen
3784 info: <a href="/doxygen/classllvm_1_1Function.html">Function Class</a><br>
3785 Superclasses: <a href="#GlobalValue"><tt>GlobalValue</tt></a>,
3786 <a href="#Constant"><tt>Constant</tt></a>,
3787 <a href="#User"><tt>User</tt></a>,
3788 <a href="#Value"><tt>Value</tt></a></p>
3790 <p>The <tt>Function</tt> class represents a single procedure in LLVM. It is
3791 actually one of the more complex classes in the LLVM hierarchy because it must
3792 keep track of a large amount of data. The <tt>Function</tt> class keeps track
3793 of a list of <a href="#BasicBlock"><tt>BasicBlock</tt></a>s, a list of formal
3794 <a href="#Argument"><tt>Argument</tt></a>s, and a
3795 <a href="#SymbolTable"><tt>SymbolTable</tt></a>.</p>
3797 <p>The list of <a href="#BasicBlock"><tt>BasicBlock</tt></a>s is the most
3798 commonly used part of <tt>Function</tt> objects. The list imposes an implicit
3799 ordering of the blocks in the function, which indicate how the code will be
3800 laid out by the backend. Additionally, the first <a
3801 href="#BasicBlock"><tt>BasicBlock</tt></a> is the implicit entry node for the
3802 <tt>Function</tt>. It is not legal in LLVM to explicitly branch to this initial
3803 block. There are no implicit exit nodes, and in fact there may be multiple exit
3804 nodes from a single <tt>Function</tt>. If the <a
3805 href="#BasicBlock"><tt>BasicBlock</tt></a> list is empty, this indicates that
3806 the <tt>Function</tt> is actually a function declaration: the actual body of the
3807 function hasn't been linked in yet.</p>
3809 <p>In addition to a list of <a href="#BasicBlock"><tt>BasicBlock</tt></a>s, the
3810 <tt>Function</tt> class also keeps track of the list of formal <a
3811 href="#Argument"><tt>Argument</tt></a>s that the function receives. This
3812 container manages the lifetime of the <a href="#Argument"><tt>Argument</tt></a>
3813 nodes, just like the <a href="#BasicBlock"><tt>BasicBlock</tt></a> list does for
3814 the <a href="#BasicBlock"><tt>BasicBlock</tt></a>s.</p>
3816 <p>The <a href="#SymbolTable"><tt>SymbolTable</tt></a> is a very rarely used
3817 LLVM feature that is only used when you have to look up a value by name. Aside
3818 from that, the <a href="#SymbolTable"><tt>SymbolTable</tt></a> is used
3819 internally to make sure that there are not conflicts between the names of <a
3820 href="#Instruction"><tt>Instruction</tt></a>s, <a
3821 href="#BasicBlock"><tt>BasicBlock</tt></a>s, or <a
3822 href="#Argument"><tt>Argument</tt></a>s in the function body.</p>
3824 <p>Note that <tt>Function</tt> is a <a href="#GlobalValue">GlobalValue</a>
3825 and therefore also a <a href="#Constant">Constant</a>. The value of the function
3826 is its address (after linking) which is guaranteed to be constant.</p>
3828 <!-- _______________________________________________________________________ -->
3830 <a name="m_Function">
3831 Important Public Members of the <tt>Function</tt> class
3838 <li><tt>Function(const </tt><tt><a href="#FunctionType">FunctionType</a>
3839 *Ty, LinkageTypes Linkage, const std::string &N = "", Module* Parent = 0)</tt>
3841 <p>Constructor used when you need to create new <tt>Function</tt>s to add
3842 the the program. The constructor must specify the type of the function to
3843 create and what type of linkage the function should have. The <a
3844 href="#FunctionType"><tt>FunctionType</tt></a> argument
3845 specifies the formal arguments and return value for the function. The same
3846 <a href="#FunctionType"><tt>FunctionType</tt></a> value can be used to
3847 create multiple functions. The <tt>Parent</tt> argument specifies the Module
3848 in which the function is defined. If this argument is provided, the function
3849 will automatically be inserted into that module's list of
3852 <li><tt>bool isDeclaration()</tt>
3854 <p>Return whether or not the <tt>Function</tt> has a body defined. If the
3855 function is "external", it does not have a body, and thus must be resolved
3856 by linking with a function defined in a different translation unit.</p></li>
3858 <li><tt>Function::iterator</tt> - Typedef for basic block list iterator<br>
3859 <tt>Function::const_iterator</tt> - Typedef for const_iterator.<br>
3861 <tt>begin()</tt>, <tt>end()</tt>
3862 <tt>size()</tt>, <tt>empty()</tt>
3864 <p>These are forwarding methods that make it easy to access the contents of
3865 a <tt>Function</tt> object's <a href="#BasicBlock"><tt>BasicBlock</tt></a>
3868 <li><tt>Function::BasicBlockListType &getBasicBlockList()</tt>
3870 <p>Returns the list of <a href="#BasicBlock"><tt>BasicBlock</tt></a>s. This
3871 is necessary to use when you need to update the list or perform a complex
3872 action that doesn't have a forwarding method.</p></li>
3874 <li><tt>Function::arg_iterator</tt> - Typedef for the argument list
3876 <tt>Function::const_arg_iterator</tt> - Typedef for const_iterator.<br>
3878 <tt>arg_begin()</tt>, <tt>arg_end()</tt>
3879 <tt>arg_size()</tt>, <tt>arg_empty()</tt>
3881 <p>These are forwarding methods that make it easy to access the contents of
3882 a <tt>Function</tt> object's <a href="#Argument"><tt>Argument</tt></a>
3885 <li><tt>Function::ArgumentListType &getArgumentList()</tt>
3887 <p>Returns the list of <a href="#Argument"><tt>Argument</tt></a>s. This is
3888 necessary to use when you need to update the list or perform a complex
3889 action that doesn't have a forwarding method.</p></li>
3891 <li><tt><a href="#BasicBlock">BasicBlock</a> &getEntryBlock()</tt>
3893 <p>Returns the entry <a href="#BasicBlock"><tt>BasicBlock</tt></a> for the
3894 function. Because the entry block for the function is always the first
3895 block, this returns the first block of the <tt>Function</tt>.</p></li>
3897 <li><tt><a href="#Type">Type</a> *getReturnType()</tt><br>
3898 <tt><a href="#FunctionType">FunctionType</a> *getFunctionType()</tt>
3900 <p>This traverses the <a href="#Type"><tt>Type</tt></a> of the
3901 <tt>Function</tt> and returns the return type of the function, or the <a
3902 href="#FunctionType"><tt>FunctionType</tt></a> of the actual
3905 <li><tt><a href="#SymbolTable">SymbolTable</a> *getSymbolTable()</tt>
3907 <p> Return a pointer to the <a href="#SymbolTable"><tt>SymbolTable</tt></a>
3908 for this <tt>Function</tt>.</p></li>
3915 <!-- ======================================================================= -->
3917 <a name="GlobalVariable">The <tt>GlobalVariable</tt> class</a>
3923 href="/doxygen/GlobalVariable_8h-source.html">llvm/GlobalVariable.h</a>"</tt>
3925 doxygen info: <a href="/doxygen/classllvm_1_1GlobalVariable.html">GlobalVariable
3927 Superclasses: <a href="#GlobalValue"><tt>GlobalValue</tt></a>,
3928 <a href="#Constant"><tt>Constant</tt></a>,
3929 <a href="#User"><tt>User</tt></a>,
3930 <a href="#Value"><tt>Value</tt></a></p>
3932 <p>Global variables are represented with the (surprise surprise)
3933 <tt>GlobalVariable</tt> class. Like functions, <tt>GlobalVariable</tt>s are also
3934 subclasses of <a href="#GlobalValue"><tt>GlobalValue</tt></a>, and as such are
3935 always referenced by their address (global values must live in memory, so their
3936 "name" refers to their constant address). See
3937 <a href="#GlobalValue"><tt>GlobalValue</tt></a> for more on this. Global
3938 variables may have an initial value (which must be a
3939 <a href="#Constant"><tt>Constant</tt></a>), and if they have an initializer,
3940 they may be marked as "constant" themselves (indicating that their contents
3941 never change at runtime).</p>
3943 <!-- _______________________________________________________________________ -->
3945 <a name="m_GlobalVariable">
3946 Important Public Members of the <tt>GlobalVariable</tt> class
3953 <li><tt>GlobalVariable(const </tt><tt><a href="#Type">Type</a> *Ty, bool
3954 isConstant, LinkageTypes& Linkage, <a href="#Constant">Constant</a>
3955 *Initializer = 0, const std::string &Name = "", Module* Parent = 0)</tt>
3957 <p>Create a new global variable of the specified type. If
3958 <tt>isConstant</tt> is true then the global variable will be marked as
3959 unchanging for the program. The Linkage parameter specifies the type of
3960 linkage (internal, external, weak, linkonce, appending) for the variable.
3961 If the linkage is InternalLinkage, WeakAnyLinkage, WeakODRLinkage,
3962 LinkOnceAnyLinkage or LinkOnceODRLinkage, then the resultant
3963 global variable will have internal linkage. AppendingLinkage concatenates
3964 together all instances (in different translation units) of the variable
3965 into a single variable but is only applicable to arrays. See
3966 the <a href="LangRef.html#modulestructure">LLVM Language Reference</a> for
3967 further details on linkage types. Optionally an initializer, a name, and the
3968 module to put the variable into may be specified for the global variable as
3971 <li><tt>bool isConstant() const</tt>
3973 <p>Returns true if this is a global variable that is known not to
3974 be modified at runtime.</p></li>
3976 <li><tt>bool hasInitializer()</tt>
3978 <p>Returns true if this <tt>GlobalVariable</tt> has an intializer.</p></li>
3980 <li><tt><a href="#Constant">Constant</a> *getInitializer()</tt>
3982 <p>Returns the initial value for a <tt>GlobalVariable</tt>. It is not legal
3983 to call this method if there is no initializer.</p></li>
3990 <!-- ======================================================================= -->
3992 <a name="BasicBlock">The <tt>BasicBlock</tt> class</a>
3998 href="/doxygen/BasicBlock_8h-source.html">llvm/BasicBlock.h</a>"</tt><br>
3999 doxygen info: <a href="/doxygen/classllvm_1_1BasicBlock.html">BasicBlock
4001 Superclass: <a href="#Value"><tt>Value</tt></a></p>
4003 <p>This class represents a single entry single exit section of the code,
4004 commonly known as a basic block by the compiler community. The
4005 <tt>BasicBlock</tt> class maintains a list of <a
4006 href="#Instruction"><tt>Instruction</tt></a>s, which form the body of the block.
4007 Matching the language definition, the last element of this list of instructions
4008 is always a terminator instruction (a subclass of the <a
4009 href="#TerminatorInst"><tt>TerminatorInst</tt></a> class).</p>
4011 <p>In addition to tracking the list of instructions that make up the block, the
4012 <tt>BasicBlock</tt> class also keeps track of the <a
4013 href="#Function"><tt>Function</tt></a> that it is embedded into.</p>
4015 <p>Note that <tt>BasicBlock</tt>s themselves are <a
4016 href="#Value"><tt>Value</tt></a>s, because they are referenced by instructions
4017 like branches and can go in the switch tables. <tt>BasicBlock</tt>s have type
4020 <!-- _______________________________________________________________________ -->
4022 <a name="m_BasicBlock">
4023 Important Public Members of the <tt>BasicBlock</tt> class
4030 <li><tt>BasicBlock(const std::string &Name = "", </tt><tt><a
4031 href="#Function">Function</a> *Parent = 0)</tt>
4033 <p>The <tt>BasicBlock</tt> constructor is used to create new basic blocks for
4034 insertion into a function. The constructor optionally takes a name for the new
4035 block, and a <a href="#Function"><tt>Function</tt></a> to insert it into. If
4036 the <tt>Parent</tt> parameter is specified, the new <tt>BasicBlock</tt> is
4037 automatically inserted at the end of the specified <a
4038 href="#Function"><tt>Function</tt></a>, if not specified, the BasicBlock must be
4039 manually inserted into the <a href="#Function"><tt>Function</tt></a>.</p></li>
4041 <li><tt>BasicBlock::iterator</tt> - Typedef for instruction list iterator<br>
4042 <tt>BasicBlock::const_iterator</tt> - Typedef for const_iterator.<br>
4043 <tt>begin()</tt>, <tt>end()</tt>, <tt>front()</tt>, <tt>back()</tt>,
4044 <tt>size()</tt>, <tt>empty()</tt>
4045 STL-style functions for accessing the instruction list.
4047 <p>These methods and typedefs are forwarding functions that have the same
4048 semantics as the standard library methods of the same names. These methods
4049 expose the underlying instruction list of a basic block in a way that is easy to
4050 manipulate. To get the full complement of container operations (including
4051 operations to update the list), you must use the <tt>getInstList()</tt>
4054 <li><tt>BasicBlock::InstListType &getInstList()</tt>
4056 <p>This method is used to get access to the underlying container that actually
4057 holds the Instructions. This method must be used when there isn't a forwarding
4058 function in the <tt>BasicBlock</tt> class for the operation that you would like
4059 to perform. Because there are no forwarding functions for "updating"
4060 operations, you need to use this if you want to update the contents of a
4061 <tt>BasicBlock</tt>.</p></li>
4063 <li><tt><a href="#Function">Function</a> *getParent()</tt>
4065 <p> Returns a pointer to <a href="#Function"><tt>Function</tt></a> the block is
4066 embedded into, or a null pointer if it is homeless.</p></li>
4068 <li><tt><a href="#TerminatorInst">TerminatorInst</a> *getTerminator()</tt>
4070 <p> Returns a pointer to the terminator instruction that appears at the end of
4071 the <tt>BasicBlock</tt>. If there is no terminator instruction, or if the last
4072 instruction in the block is not a terminator, then a null pointer is
4081 <!-- ======================================================================= -->
4083 <a name="Argument">The <tt>Argument</tt> class</a>
4088 <p>This subclass of Value defines the interface for incoming formal
4089 arguments to a function. A Function maintains a list of its formal
4090 arguments. An argument has a pointer to the parent Function.</p>
4096 <!-- *********************************************************************** -->
4099 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
4100 src="http://jigsaw.w3.org/css-validator/images/vcss-blue" alt="Valid CSS"></a>
4101 <a href="http://validator.w3.org/check/referer"><img
4102 src="http://www.w3.org/Icons/valid-html401" alt="Valid HTML 4.01 Strict"></a>
4104 <a href="mailto:dhurjati@cs.uiuc.edu">Dinakar Dhurjati</a> and
4105 <a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
4106 <a href="http://llvm.org/">The LLVM Compiler Infrastructure</a><br>
4107 Last modified: $Date$