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_otherset">Other Set-Like ContainerOptions</a></li>
90 <li><a href="#ds_map">Map-Like Containers (std::map, DenseMap, etc)</a>
92 <li><a href="#dss_sortedvectormap">A sorted 'vector'</a></li>
93 <li><a href="#dss_stringmap">"llvm/ADT/StringMap.h"</a></li>
94 <li><a href="#dss_indexedmap">"llvm/ADT/IndexedMap.h"</a></li>
95 <li><a href="#dss_densemap">"llvm/ADT/DenseMap.h"</a></li>
96 <li><a href="#dss_valuemap">"llvm/ADT/ValueMap.h"</a></li>
97 <li><a href="#dss_intervalmap">"llvm/ADT/IntervalMap.h"</a></li>
98 <li><a href="#dss_map"><map></a></li>
99 <li><a href="#dss_inteqclasses">"llvm/ADT/IntEqClasses.h"</a></li>
100 <li><a href="#dss_othermap">Other Map-Like Container Options</a></li>
102 <li><a href="#ds_bit">BitVector-like containers</a>
104 <li><a href="#dss_bitvector">A dense bitvector</a></li>
105 <li><a href="#dss_smallbitvector">A "small" dense bitvector</a></li>
106 <li><a href="#dss_sparsebitvector">A sparse bitvector</a></li>
110 <li><a href="#common">Helpful Hints for Common Operations</a>
112 <li><a href="#inspection">Basic Inspection and Traversal Routines</a>
114 <li><a href="#iterate_function">Iterating over the <tt>BasicBlock</tt>s
115 in a <tt>Function</tt></a> </li>
116 <li><a href="#iterate_basicblock">Iterating over the <tt>Instruction</tt>s
117 in a <tt>BasicBlock</tt></a> </li>
118 <li><a href="#iterate_institer">Iterating over the <tt>Instruction</tt>s
119 in a <tt>Function</tt></a> </li>
120 <li><a href="#iterate_convert">Turning an iterator into a
121 class pointer</a> </li>
122 <li><a href="#iterate_complex">Finding call sites: a more
123 complex example</a> </li>
124 <li><a href="#calls_and_invokes">Treating calls and invokes
125 the same way</a> </li>
126 <li><a href="#iterate_chains">Iterating over def-use &
127 use-def chains</a> </li>
128 <li><a href="#iterate_preds">Iterating over predecessors &
129 successors of blocks</a></li>
132 <li><a href="#simplechanges">Making simple changes</a>
134 <li><a href="#schanges_creating">Creating and inserting new
135 <tt>Instruction</tt>s</a> </li>
136 <li><a href="#schanges_deleting">Deleting <tt>Instruction</tt>s</a> </li>
137 <li><a href="#schanges_replacing">Replacing an <tt>Instruction</tt>
138 with another <tt>Value</tt></a> </li>
139 <li><a href="#schanges_deletingGV">Deleting <tt>GlobalVariable</tt>s</a> </li>
142 <li><a href="#create_types">How to Create Types</a></li>
144 <li>Working with the Control Flow Graph
146 <li>Accessing predecessors and successors of a <tt>BasicBlock</tt>
154 <li><a href="#threading">Threads and LLVM</a>
156 <li><a href="#startmultithreaded">Entering and Exiting Multithreaded Mode
158 <li><a href="#shutdown">Ending execution with <tt>llvm_shutdown()</tt></a></li>
159 <li><a href="#managedstatic">Lazy initialization with <tt>ManagedStatic</tt></a></li>
160 <li><a href="#llvmcontext">Achieving Isolation with <tt>LLVMContext</tt></a></li>
161 <li><a href="#jitthreading">Threads and the JIT</a></li>
165 <li><a href="#advanced">Advanced Topics</a>
168 <li><a href="#SymbolTable">The <tt>ValueSymbolTable</tt> class</a></li>
169 <li><a href="#UserLayout">The <tt>User</tt> and owned <tt>Use</tt> classes' memory layout</a></li>
172 <li><a href="#coreclasses">The Core LLVM Class Hierarchy Reference</a>
174 <li><a href="#Type">The <tt>Type</tt> class</a> </li>
175 <li><a href="#Module">The <tt>Module</tt> class</a></li>
176 <li><a href="#Value">The <tt>Value</tt> class</a>
178 <li><a href="#User">The <tt>User</tt> class</a>
180 <li><a href="#Instruction">The <tt>Instruction</tt> class</a></li>
181 <li><a href="#Constant">The <tt>Constant</tt> class</a>
183 <li><a href="#GlobalValue">The <tt>GlobalValue</tt> class</a>
185 <li><a href="#Function">The <tt>Function</tt> class</a></li>
186 <li><a href="#GlobalVariable">The <tt>GlobalVariable</tt> class</a></li>
193 <li><a href="#BasicBlock">The <tt>BasicBlock</tt> class</a></li>
194 <li><a href="#Argument">The <tt>Argument</tt> class</a></li>
201 <div class="doc_author">
202 <p>Written by <a href="mailto:sabre@nondot.org">Chris Lattner</a>,
203 <a href="mailto:dhurjati@cs.uiuc.edu">Dinakar Dhurjati</a>,
204 <a href="mailto:ggreif@gmail.com">Gabor Greif</a>,
205 <a href="mailto:jstanley@cs.uiuc.edu">Joel Stanley</a>,
206 <a href="mailto:rspencer@x10sys.com">Reid Spencer</a> and
207 <a href="mailto:owen@apple.com">Owen Anderson</a></p>
210 <!-- *********************************************************************** -->
212 <a name="introduction">Introduction </a>
214 <!-- *********************************************************************** -->
218 <p>This document is meant to highlight some of the important classes and
219 interfaces available in the LLVM source-base. This manual is not
220 intended to explain what LLVM is, how it works, and what LLVM code looks
221 like. It assumes that you know the basics of LLVM and are interested
222 in writing transformations or otherwise analyzing or manipulating the
225 <p>This document should get you oriented so that you can find your
226 way in the continuously growing source code that makes up the LLVM
227 infrastructure. Note that this manual is not intended to serve as a
228 replacement for reading the source code, so if you think there should be
229 a method in one of these classes to do something, but it's not listed,
230 check the source. Links to the <a href="/doxygen/">doxygen</a> sources
231 are provided to make this as easy as possible.</p>
233 <p>The first section of this document describes general information that is
234 useful to know when working in the LLVM infrastructure, and the second describes
235 the Core LLVM classes. In the future this manual will be extended with
236 information describing how to use extension libraries, such as dominator
237 information, CFG traversal routines, and useful utilities like the <tt><a
238 href="/doxygen/InstVisitor_8h-source.html">InstVisitor</a></tt> template.</p>
242 <!-- *********************************************************************** -->
244 <a name="general">General Information</a>
246 <!-- *********************************************************************** -->
250 <p>This section contains general information that is useful if you are working
251 in the LLVM source-base, but that isn't specific to any particular API.</p>
253 <!-- ======================================================================= -->
255 <a name="stl">The C++ Standard Template Library</a>
260 <p>LLVM makes heavy use of the C++ Standard Template Library (STL),
261 perhaps much more than you are used to, or have seen before. Because of
262 this, you might want to do a little background reading in the
263 techniques used and capabilities of the library. There are many good
264 pages that discuss the STL, and several books on the subject that you
265 can get, so it will not be discussed in this document.</p>
267 <p>Here are some useful links:</p>
271 <li><a href="http://www.dinkumware.com/manuals/#Standard C++ Library">Dinkumware
272 C++ Library reference</a> - an excellent reference for the STL and other parts
273 of the standard C++ library.</li>
275 <li><a href="http://www.tempest-sw.com/cpp/">C++ In a Nutshell</a> - This is an
276 O'Reilly book in the making. It has a decent Standard Library
277 Reference that rivals Dinkumware's, and is unfortunately no longer free since the
278 book has been published.</li>
280 <li><a href="http://www.parashift.com/c++-faq-lite/">C++ Frequently Asked
283 <li><a href="http://www.sgi.com/tech/stl/">SGI's STL Programmer's Guide</a> -
285 href="http://www.sgi.com/tech/stl/stl_introduction.html">Introduction to the
288 <li><a href="http://www.research.att.com/%7Ebs/C++.html">Bjarne Stroustrup's C++
291 <li><a href="http://64.78.49.204/">
292 Bruce Eckel's Thinking in C++, 2nd ed. Volume 2 Revision 4.0 (even better, get
297 <p>You are also encouraged to take a look at the <a
298 href="CodingStandards.html">LLVM Coding Standards</a> guide which focuses on how
299 to write maintainable code more than where to put your curly braces.</p>
303 <!-- ======================================================================= -->
305 <a name="stl">Other useful references</a>
311 <li><a href="http://www.fortran-2000.com/ArnaudRecipes/sharedlib.html">Using
312 static and shared libraries across platforms</a></li>
319 <!-- *********************************************************************** -->
321 <a name="apis">Important and useful LLVM APIs</a>
323 <!-- *********************************************************************** -->
327 <p>Here we highlight some LLVM APIs that are generally useful and good to
328 know about when writing transformations.</p>
330 <!-- ======================================================================= -->
332 <a name="isa">The <tt>isa<></tt>, <tt>cast<></tt> and
333 <tt>dyn_cast<></tt> templates</a>
338 <p>The LLVM source-base makes extensive use of a custom form of RTTI.
339 These templates have many similarities to the C++ <tt>dynamic_cast<></tt>
340 operator, but they don't have some drawbacks (primarily stemming from
341 the fact that <tt>dynamic_cast<></tt> only works on classes that
342 have a v-table). Because they are used so often, you must know what they
343 do and how they work. All of these templates are defined in the <a
344 href="/doxygen/Casting_8h-source.html"><tt>llvm/Support/Casting.h</tt></a>
345 file (note that you very rarely have to include this file directly).</p>
348 <dt><tt>isa<></tt>: </dt>
350 <dd><p>The <tt>isa<></tt> operator works exactly like the Java
351 "<tt>instanceof</tt>" operator. It returns true or false depending on whether
352 a reference or pointer points to an instance of the specified class. This can
353 be very useful for constraint checking of various sorts (example below).</p>
356 <dt><tt>cast<></tt>: </dt>
358 <dd><p>The <tt>cast<></tt> operator is a "checked cast" operation. It
359 converts a pointer or reference from a base class to a derived class, causing
360 an assertion failure if it is not really an instance of the right type. This
361 should be used in cases where you have some information that makes you believe
362 that something is of the right type. An example of the <tt>isa<></tt>
363 and <tt>cast<></tt> template is:</p>
365 <div class="doc_code">
367 static bool isLoopInvariant(const <a href="#Value">Value</a> *V, const Loop *L) {
368 if (isa<<a href="#Constant">Constant</a>>(V) || isa<<a href="#Argument">Argument</a>>(V) || isa<<a href="#GlobalValue">GlobalValue</a>>(V))
371 // <i>Otherwise, it must be an instruction...</i>
372 return !L->contains(cast<<a href="#Instruction">Instruction</a>>(V)->getParent());
377 <p>Note that you should <b>not</b> use an <tt>isa<></tt> test followed
378 by a <tt>cast<></tt>, for that use the <tt>dyn_cast<></tt>
383 <dt><tt>dyn_cast<></tt>:</dt>
385 <dd><p>The <tt>dyn_cast<></tt> operator is a "checking cast" operation.
386 It checks to see if the operand is of the specified type, and if so, returns a
387 pointer to it (this operator does not work with references). If the operand is
388 not of the correct type, a null pointer is returned. Thus, this works very
389 much like the <tt>dynamic_cast<></tt> operator in C++, and should be
390 used in the same circumstances. Typically, the <tt>dyn_cast<></tt>
391 operator is used in an <tt>if</tt> statement or some other flow control
392 statement like this:</p>
394 <div class="doc_code">
396 if (<a href="#AllocationInst">AllocationInst</a> *AI = dyn_cast<<a href="#AllocationInst">AllocationInst</a>>(Val)) {
402 <p>This form of the <tt>if</tt> statement effectively combines together a call
403 to <tt>isa<></tt> and a call to <tt>cast<></tt> into one
404 statement, which is very convenient.</p>
406 <p>Note that the <tt>dyn_cast<></tt> operator, like C++'s
407 <tt>dynamic_cast<></tt> or Java's <tt>instanceof</tt> operator, can be
408 abused. In particular, you should not use big chained <tt>if/then/else</tt>
409 blocks to check for lots of different variants of classes. If you find
410 yourself wanting to do this, it is much cleaner and more efficient to use the
411 <tt>InstVisitor</tt> class to dispatch over the instruction type directly.</p>
415 <dt><tt>cast_or_null<></tt>: </dt>
417 <dd><p>The <tt>cast_or_null<></tt> operator works just like the
418 <tt>cast<></tt> operator, except that it allows for a null pointer as an
419 argument (which it then propagates). This can sometimes be useful, allowing
420 you to combine several null checks into one.</p></dd>
422 <dt><tt>dyn_cast_or_null<></tt>: </dt>
424 <dd><p>The <tt>dyn_cast_or_null<></tt> operator works just like the
425 <tt>dyn_cast<></tt> operator, except that it allows for a null pointer
426 as an argument (which it then propagates). This can sometimes be useful,
427 allowing you to combine several null checks into one.</p></dd>
431 <p>These five templates can be used with any classes, whether they have a
432 v-table or not. To add support for these templates, you simply need to add
433 <tt>classof</tt> static methods to the class you are interested casting
434 to. Describing this is currently outside the scope of this document, but there
435 are lots of examples in the LLVM source base.</p>
440 <!-- ======================================================================= -->
442 <a name="string_apis">Passing strings (the <tt>StringRef</tt>
443 and <tt>Twine</tt> classes)</a>
448 <p>Although LLVM generally does not do much string manipulation, we do have
449 several important APIs which take strings. Two important examples are the
450 Value class -- which has names for instructions, functions, etc. -- and the
451 StringMap class which is used extensively in LLVM and Clang.</p>
453 <p>These are generic classes, and they need to be able to accept strings which
454 may have embedded null characters. Therefore, they cannot simply take
455 a <tt>const char *</tt>, and taking a <tt>const std::string&</tt> requires
456 clients to perform a heap allocation which is usually unnecessary. Instead,
457 many LLVM APIs use a <tt>StringRef</tt> or a <tt>const Twine&</tt> for
458 passing strings efficiently.</p>
460 <!-- _______________________________________________________________________ -->
462 <a name="StringRef">The <tt>StringRef</tt> class</a>
467 <p>The <tt>StringRef</tt> data type represents a reference to a constant string
468 (a character array and a length) and supports the common operations available
469 on <tt>std:string</tt>, but does not require heap allocation.</p>
471 <p>It can be implicitly constructed using a C style null-terminated string,
472 an <tt>std::string</tt>, or explicitly with a character pointer and length.
473 For example, the <tt>StringRef</tt> find function is declared as:</p>
475 <pre class="doc_code">
476 iterator find(StringRef Key);
479 <p>and clients can call it using any one of:</p>
481 <pre class="doc_code">
482 Map.find("foo"); <i>// Lookup "foo"</i>
483 Map.find(std::string("bar")); <i>// Lookup "bar"</i>
484 Map.find(StringRef("\0baz", 4)); <i>// Lookup "\0baz"</i>
487 <p>Similarly, APIs which need to return a string may return a <tt>StringRef</tt>
488 instance, which can be used directly or converted to an <tt>std::string</tt>
489 using the <tt>str</tt> member function. See
490 "<tt><a href="/doxygen/classllvm_1_1StringRef_8h-source.html">llvm/ADT/StringRef.h</a></tt>"
491 for more information.</p>
493 <p>You should rarely use the <tt>StringRef</tt> class directly, because it contains
494 pointers to external memory it is not generally safe to store an instance of the
495 class (unless you know that the external storage will not be freed). StringRef is
496 small and pervasive enough in LLVM that it should always be passed by value.</p>
500 <!-- _______________________________________________________________________ -->
502 <a name="Twine">The <tt>Twine</tt> class</a>
507 <p>The <tt>Twine</tt> class is an efficient way for APIs to accept concatenated
508 strings. For example, a common LLVM paradigm is to name one instruction based on
509 the name of another instruction with a suffix, for example:</p>
511 <div class="doc_code">
513 New = CmpInst::Create(<i>...</i>, SO->getName() + ".cmp");
517 <p>The <tt>Twine</tt> class is effectively a
518 lightweight <a href="http://en.wikipedia.org/wiki/Rope_(computer_science)">rope</a>
519 which points to temporary (stack allocated) objects. Twines can be implicitly
520 constructed as the result of the plus operator applied to strings (i.e., a C
521 strings, an <tt>std::string</tt>, or a <tt>StringRef</tt>). The twine delays the
522 actual concatenation of strings until it is actually required, at which point
523 it can be efficiently rendered directly into a character array. This avoids
524 unnecessary heap allocation involved in constructing the temporary results of
525 string concatenation. See
526 "<tt><a href="/doxygen/classllvm_1_1Twine_8h-source.html">llvm/ADT/Twine.h</a></tt>"
527 for more information.</p>
529 <p>As with a <tt>StringRef</tt>, <tt>Twine</tt> objects point to external memory
530 and should almost never be stored or mentioned directly. They are intended
531 solely for use when defining a function which should be able to efficiently
532 accept concatenated strings.</p>
538 <!-- ======================================================================= -->
540 <a name="DEBUG">The <tt>DEBUG()</tt> macro and <tt>-debug</tt> option</a>
545 <p>Often when working on your pass you will put a bunch of debugging printouts
546 and other code into your pass. After you get it working, you want to remove
547 it, but you may need it again in the future (to work out new bugs that you run
550 <p> Naturally, because of this, you don't want to delete the debug printouts,
551 but you don't want them to always be noisy. A standard compromise is to comment
552 them out, allowing you to enable them if you need them in the future.</p>
554 <p>The "<tt><a href="/doxygen/Debug_8h-source.html">llvm/Support/Debug.h</a></tt>"
555 file provides a macro named <tt>DEBUG()</tt> that is a much nicer solution to
556 this problem. Basically, you can put arbitrary code into the argument of the
557 <tt>DEBUG</tt> macro, and it is only executed if '<tt>opt</tt>' (or any other
558 tool) is run with the '<tt>-debug</tt>' command line argument:</p>
560 <div class="doc_code">
562 DEBUG(errs() << "I am here!\n");
566 <p>Then you can run your pass like this:</p>
568 <div class="doc_code">
570 $ opt < a.bc > /dev/null -mypass
571 <i><no output></i>
572 $ opt < a.bc > /dev/null -mypass -debug
577 <p>Using the <tt>DEBUG()</tt> macro instead of a home-brewed solution allows you
578 to not have to create "yet another" command line option for the debug output for
579 your pass. Note that <tt>DEBUG()</tt> macros are disabled for optimized builds,
580 so they do not cause a performance impact at all (for the same reason, they
581 should also not contain side-effects!).</p>
583 <p>One additional nice thing about the <tt>DEBUG()</tt> macro is that you can
584 enable or disable it directly in gdb. Just use "<tt>set DebugFlag=0</tt>" or
585 "<tt>set DebugFlag=1</tt>" from the gdb if the program is running. If the
586 program hasn't been started yet, you can always just run it with
589 <!-- _______________________________________________________________________ -->
591 <a name="DEBUG_TYPE">Fine grained debug info with <tt>DEBUG_TYPE</tt> and
592 the <tt>-debug-only</tt> option</a>
597 <p>Sometimes you may find yourself in a situation where enabling <tt>-debug</tt>
598 just turns on <b>too much</b> information (such as when working on the code
599 generator). If you want to enable debug information with more fine-grained
600 control, you define the <tt>DEBUG_TYPE</tt> macro and the <tt>-debug</tt> only
601 option as follows:</p>
603 <div class="doc_code">
606 DEBUG(errs() << "No debug type\n");
607 #define DEBUG_TYPE "foo"
608 DEBUG(errs() << "'foo' debug type\n");
610 #define DEBUG_TYPE "bar"
611 DEBUG(errs() << "'bar' debug type\n"));
613 #define DEBUG_TYPE ""
614 DEBUG(errs() << "No debug type (2)\n");
618 <p>Then you can run your pass like this:</p>
620 <div class="doc_code">
622 $ opt < a.bc > /dev/null -mypass
623 <i><no output></i>
624 $ opt < a.bc > /dev/null -mypass -debug
629 $ opt < a.bc > /dev/null -mypass -debug-only=foo
631 $ opt < a.bc > /dev/null -mypass -debug-only=bar
636 <p>Of course, in practice, you should only set <tt>DEBUG_TYPE</tt> at the top of
637 a file, to specify the debug type for the entire module (if you do this before
638 you <tt>#include "llvm/Support/Debug.h"</tt>, you don't have to insert the ugly
639 <tt>#undef</tt>'s). Also, you should use names more meaningful than "foo" and
640 "bar", because there is no system in place to ensure that names do not
641 conflict. If two different modules use the same string, they will all be turned
642 on when the name is specified. This allows, for example, all debug information
643 for instruction scheduling to be enabled with <tt>-debug-type=InstrSched</tt>,
644 even if the source lives in multiple files.</p>
646 <p>The <tt>DEBUG_WITH_TYPE</tt> macro is also available for situations where you
647 would like to set <tt>DEBUG_TYPE</tt>, but only for one specific <tt>DEBUG</tt>
648 statement. It takes an additional first parameter, which is the type to use. For
649 example, the preceding example could be written as:</p>
652 <div class="doc_code">
654 DEBUG_WITH_TYPE("", errs() << "No debug type\n");
655 DEBUG_WITH_TYPE("foo", errs() << "'foo' debug type\n");
656 DEBUG_WITH_TYPE("bar", errs() << "'bar' debug type\n"));
657 DEBUG_WITH_TYPE("", errs() << "No debug type (2)\n");
665 <!-- ======================================================================= -->
667 <a name="Statistic">The <tt>Statistic</tt> class & <tt>-stats</tt>
674 href="/doxygen/Statistic_8h-source.html">llvm/ADT/Statistic.h</a></tt>" file
675 provides a class named <tt>Statistic</tt> that is used as a unified way to
676 keep track of what the LLVM compiler is doing and how effective various
677 optimizations are. It is useful to see what optimizations are contributing to
678 making a particular program run faster.</p>
680 <p>Often you may run your pass on some big program, and you're interested to see
681 how many times it makes a certain transformation. Although you can do this with
682 hand inspection, or some ad-hoc method, this is a real pain and not very useful
683 for big programs. Using the <tt>Statistic</tt> class makes it very easy to
684 keep track of this information, and the calculated information is presented in a
685 uniform manner with the rest of the passes being executed.</p>
687 <p>There are many examples of <tt>Statistic</tt> uses, but the basics of using
688 it are as follows:</p>
691 <li><p>Define your statistic like this:</p>
693 <div class="doc_code">
695 #define <a href="#DEBUG_TYPE">DEBUG_TYPE</a> "mypassname" <i>// This goes before any #includes.</i>
696 STATISTIC(NumXForms, "The # of times I did stuff");
700 <p>The <tt>STATISTIC</tt> macro defines a static variable, whose name is
701 specified by the first argument. The pass name is taken from the DEBUG_TYPE
702 macro, and the description is taken from the second argument. The variable
703 defined ("NumXForms" in this case) acts like an unsigned integer.</p></li>
705 <li><p>Whenever you make a transformation, bump the counter:</p>
707 <div class="doc_code">
709 ++NumXForms; // <i>I did stuff!</i>
716 <p>That's all you have to do. To get '<tt>opt</tt>' to print out the
717 statistics gathered, use the '<tt>-stats</tt>' option:</p>
719 <div class="doc_code">
721 $ opt -stats -mypassname < program.bc > /dev/null
722 <i>... statistics output ...</i>
726 <p> When running <tt>opt</tt> on a C file from the SPEC benchmark
727 suite, it gives a report that looks like this:</p>
729 <div class="doc_code">
731 7646 bitcodewriter - Number of normal instructions
732 725 bitcodewriter - Number of oversized instructions
733 129996 bitcodewriter - Number of bitcode bytes written
734 2817 raise - Number of insts DCEd or constprop'd
735 3213 raise - Number of cast-of-self removed
736 5046 raise - Number of expression trees converted
737 75 raise - Number of other getelementptr's formed
738 138 raise - Number of load/store peepholes
739 42 deadtypeelim - Number of unused typenames removed from symtab
740 392 funcresolve - Number of varargs functions resolved
741 27 globaldce - Number of global variables removed
742 2 adce - Number of basic blocks removed
743 134 cee - Number of branches revectored
744 49 cee - Number of setcc instruction eliminated
745 532 gcse - Number of loads removed
746 2919 gcse - Number of instructions removed
747 86 indvars - Number of canonical indvars added
748 87 indvars - Number of aux indvars removed
749 25 instcombine - Number of dead inst eliminate
750 434 instcombine - Number of insts combined
751 248 licm - Number of load insts hoisted
752 1298 licm - Number of insts hoisted to a loop pre-header
753 3 licm - Number of insts hoisted to multiple loop preds (bad, no loop pre-header)
754 75 mem2reg - Number of alloca's promoted
755 1444 cfgsimplify - Number of blocks simplified
759 <p>Obviously, with so many optimizations, having a unified framework for this
760 stuff is very nice. Making your pass fit well into the framework makes it more
761 maintainable and useful.</p>
765 <!-- ======================================================================= -->
767 <a name="ViewGraph">Viewing graphs while debugging code</a>
772 <p>Several of the important data structures in LLVM are graphs: for example
773 CFGs made out of LLVM <a href="#BasicBlock">BasicBlock</a>s, CFGs made out of
774 LLVM <a href="CodeGenerator.html#machinebasicblock">MachineBasicBlock</a>s, and
775 <a href="CodeGenerator.html#selectiondag_intro">Instruction Selection
776 DAGs</a>. In many cases, while debugging various parts of the compiler, it is
777 nice to instantly visualize these graphs.</p>
779 <p>LLVM provides several callbacks that are available in a debug build to do
780 exactly that. If you call the <tt>Function::viewCFG()</tt> method, for example,
781 the current LLVM tool will pop up a window containing the CFG for the function
782 where each basic block is a node in the graph, and each node contains the
783 instructions in the block. Similarly, there also exists
784 <tt>Function::viewCFGOnly()</tt> (does not include the instructions), the
785 <tt>MachineFunction::viewCFG()</tt> and <tt>MachineFunction::viewCFGOnly()</tt>,
786 and the <tt>SelectionDAG::viewGraph()</tt> methods. Within GDB, for example,
787 you can usually use something like <tt>call DAG.viewGraph()</tt> to pop
788 up a window. Alternatively, you can sprinkle calls to these functions in your
789 code in places you want to debug.</p>
791 <p>Getting this to work requires a small amount of configuration. On Unix
792 systems with X11, install the <a href="http://www.graphviz.org">graphviz</a>
793 toolkit, and make sure 'dot' and 'gv' are in your path. If you are running on
794 Mac OS/X, download and install the Mac OS/X <a
795 href="http://www.pixelglow.com/graphviz/">Graphviz program</a>, and add
796 <tt>/Applications/Graphviz.app/Contents/MacOS/</tt> (or wherever you install
797 it) to your path. Once in your system and path are set up, rerun the LLVM
798 configure script and rebuild LLVM to enable this functionality.</p>
800 <p><tt>SelectionDAG</tt> has been extended to make it easier to locate
801 <i>interesting</i> nodes in large complex graphs. From gdb, if you
802 <tt>call DAG.setGraphColor(<i>node</i>, "<i>color</i>")</tt>, then the
803 next <tt>call DAG.viewGraph()</tt> would highlight the node in the
804 specified color (choices of colors can be found at <a
805 href="http://www.graphviz.org/doc/info/colors.html">colors</a>.) More
806 complex node attributes can be provided with <tt>call
807 DAG.setGraphAttrs(<i>node</i>, "<i>attributes</i>")</tt> (choices can be
808 found at <a href="http://www.graphviz.org/doc/info/attrs.html">Graph
809 Attributes</a>.) If you want to restart and clear all the current graph
810 attributes, then you can <tt>call DAG.clearGraphAttrs()</tt>. </p>
812 <p>Note that graph visualization features are compiled out of Release builds
813 to reduce file size. This means that you need a Debug+Asserts or
814 Release+Asserts build to use these features.</p>
820 <!-- *********************************************************************** -->
822 <a name="datastructure">Picking the Right Data Structure for a Task</a>
824 <!-- *********************************************************************** -->
828 <p>LLVM has a plethora of data structures in the <tt>llvm/ADT/</tt> directory,
829 and we commonly use STL data structures. This section describes the trade-offs
830 you should consider when you pick one.</p>
833 The first step is a choose your own adventure: do you want a sequential
834 container, a set-like container, or a map-like container? The most important
835 thing when choosing a container is the algorithmic properties of how you plan to
836 access the container. Based on that, you should use:</p>
839 <li>a <a href="#ds_map">map-like</a> container if you need efficient look-up
840 of an value based on another value. Map-like containers also support
841 efficient queries for containment (whether a key is in the map). Map-like
842 containers generally do not support efficient reverse mapping (values to
843 keys). If you need that, use two maps. Some map-like containers also
844 support efficient iteration through the keys in sorted order. Map-like
845 containers are the most expensive sort, only use them if you need one of
846 these capabilities.</li>
848 <li>a <a href="#ds_set">set-like</a> container if you need to put a bunch of
849 stuff into a container that automatically eliminates duplicates. Some
850 set-like containers support efficient iteration through the elements in
851 sorted order. Set-like containers are more expensive than sequential
855 <li>a <a href="#ds_sequential">sequential</a> container provides
856 the most efficient way to add elements and keeps track of the order they are
857 added to the collection. They permit duplicates and support efficient
858 iteration, but do not support efficient look-up based on a key.
861 <li>a <a href="#ds_string">string</a> container is a specialized sequential
862 container or reference structure that is used for character or byte
865 <li>a <a href="#ds_bit">bit</a> container provides an efficient way to store and
866 perform set operations on sets of numeric id's, while automatically
867 eliminating duplicates. Bit containers require a maximum of 1 bit for each
868 identifier you want to store.
873 Once the proper category of container is determined, you can fine tune the
874 memory use, constant factors, and cache behaviors of access by intelligently
875 picking a member of the category. Note that constant factors and cache behavior
876 can be a big deal. If you have a vector that usually only contains a few
877 elements (but could contain many), for example, it's much better to use
878 <a href="#dss_smallvector">SmallVector</a> than <a href="#dss_vector">vector</a>
879 . Doing so avoids (relatively) expensive malloc/free calls, which dwarf the
880 cost of adding the elements to the container. </p>
882 <!-- ======================================================================= -->
884 <a name="ds_sequential">Sequential Containers (std::vector, std::list, etc)</a>
888 There are a variety of sequential containers available for you, based on your
889 needs. Pick the first in this section that will do what you want.
891 <!-- _______________________________________________________________________ -->
893 <a name="dss_arrayref">llvm/ADT/ArrayRef.h</a>
897 <p>The llvm::ArrayRef class is the preferred class to use in an interface that
898 accepts a sequential list of elements in memory and just reads from them. By
899 taking an ArrayRef, the API can be passed a fixed size array, an std::vector,
900 an llvm::SmallVector and anything else that is contiguous in memory.
906 <!-- _______________________________________________________________________ -->
908 <a name="dss_fixedarrays">Fixed Size Arrays</a>
912 <p>Fixed size arrays are very simple and very fast. They are good if you know
913 exactly how many elements you have, or you have a (low) upper bound on how many
917 <!-- _______________________________________________________________________ -->
919 <a name="dss_heaparrays">Heap Allocated Arrays</a>
923 <p>Heap allocated arrays (new[] + delete[]) are also simple. They are good if
924 the number of elements is variable, if you know how many elements you will need
925 before the array is allocated, and if the array is usually large (if not,
926 consider a <a href="#dss_smallvector">SmallVector</a>). The cost of a heap
927 allocated array is the cost of the new/delete (aka malloc/free). Also note that
928 if you are allocating an array of a type with a constructor, the constructor and
929 destructors will be run for every element in the array (re-sizable vectors only
930 construct those elements actually used).</p>
933 <!-- _______________________________________________________________________ -->
935 <a name="dss_tinyptrvector">"llvm/ADT/TinyPtrVector.h"</a>
940 <p><tt>TinyPtrVector<Type></tt> is a highly specialized collection class
941 that is optimized to avoid allocation in the case when a vector has zero or one
942 elements. It has two major restrictions: 1) it can only hold values of pointer
943 type, and 2) it cannot hold a null pointer.</p>
945 <p>Since this container is highly specialized, it is rarely used.</p>
949 <!-- _______________________________________________________________________ -->
951 <a name="dss_smallvector">"llvm/ADT/SmallVector.h"</a>
955 <p><tt>SmallVector<Type, N></tt> is a simple class that looks and smells
956 just like <tt>vector<Type></tt>:
957 it supports efficient iteration, lays out elements in memory order (so you can
958 do pointer arithmetic between elements), supports efficient push_back/pop_back
959 operations, supports efficient random access to its elements, etc.</p>
961 <p>The advantage of SmallVector is that it allocates space for
962 some number of elements (N) <b>in the object itself</b>. Because of this, if
963 the SmallVector is dynamically smaller than N, no malloc is performed. This can
964 be a big win in cases where the malloc/free call is far more expensive than the
965 code that fiddles around with the elements.</p>
967 <p>This is good for vectors that are "usually small" (e.g. the number of
968 predecessors/successors of a block is usually less than 8). On the other hand,
969 this makes the size of the SmallVector itself large, so you don't want to
970 allocate lots of them (doing so will waste a lot of space). As such,
971 SmallVectors are most useful when on the stack.</p>
973 <p>SmallVector also provides a nice portable and efficient replacement for
978 <!-- _______________________________________________________________________ -->
980 <a name="dss_vector"><vector></a>
985 std::vector is well loved and respected. It is useful when SmallVector isn't:
986 when the size of the vector is often large (thus the small optimization will
987 rarely be a benefit) or if you will be allocating many instances of the vector
988 itself (which would waste space for elements that aren't in the container).
989 vector is also useful when interfacing with code that expects vectors :).
992 <p>One worthwhile note about std::vector: avoid code like this:</p>
994 <div class="doc_code">
997 std::vector<foo> V;
1003 <p>Instead, write this as:</p>
1005 <div class="doc_code">
1007 std::vector<foo> V;
1015 <p>Doing so will save (at least) one heap allocation and free per iteration of
1020 <!-- _______________________________________________________________________ -->
1022 <a name="dss_deque"><deque></a>
1026 <p>std::deque is, in some senses, a generalized version of std::vector. Like
1027 std::vector, it provides constant time random access and other similar
1028 properties, but it also provides efficient access to the front of the list. It
1029 does not guarantee continuity of elements within memory.</p>
1031 <p>In exchange for this extra flexibility, std::deque has significantly higher
1032 constant factor costs than std::vector. If possible, use std::vector or
1033 something cheaper.</p>
1036 <!-- _______________________________________________________________________ -->
1038 <a name="dss_list"><list></a>
1042 <p>std::list is an extremely inefficient class that is rarely useful.
1043 It performs a heap allocation for every element inserted into it, thus having an
1044 extremely high constant factor, particularly for small data types. std::list
1045 also only supports bidirectional iteration, not random access iteration.</p>
1047 <p>In exchange for this high cost, std::list supports efficient access to both
1048 ends of the list (like std::deque, but unlike std::vector or SmallVector). In
1049 addition, the iterator invalidation characteristics of std::list are stronger
1050 than that of a vector class: inserting or removing an element into the list does
1051 not invalidate iterator or pointers to other elements in the list.</p>
1054 <!-- _______________________________________________________________________ -->
1056 <a name="dss_ilist">llvm/ADT/ilist.h</a>
1060 <p><tt>ilist<T></tt> implements an 'intrusive' doubly-linked list. It is
1061 intrusive, because it requires the element to store and provide access to the
1062 prev/next pointers for the list.</p>
1064 <p><tt>ilist</tt> has the same drawbacks as <tt>std::list</tt>, and additionally
1065 requires an <tt>ilist_traits</tt> implementation for the element type, but it
1066 provides some novel characteristics. In particular, it can efficiently store
1067 polymorphic objects, the traits class is informed when an element is inserted or
1068 removed from the list, and <tt>ilist</tt>s are guaranteed to support a
1069 constant-time splice operation.</p>
1071 <p>These properties are exactly what we want for things like
1072 <tt>Instruction</tt>s and basic blocks, which is why these are implemented with
1073 <tt>ilist</tt>s.</p>
1075 Related classes of interest are explained in the following subsections:
1077 <li><a href="#dss_ilist_traits">ilist_traits</a></li>
1078 <li><a href="#dss_iplist">iplist</a></li>
1079 <li><a href="#dss_ilist_node">llvm/ADT/ilist_node.h</a></li>
1080 <li><a href="#dss_ilist_sentinel">Sentinels</a></li>
1084 <!-- _______________________________________________________________________ -->
1086 <a name="dss_packedvector">llvm/ADT/PackedVector.h</a>
1091 Useful for storing a vector of values using only a few number of bits for each
1092 value. Apart from the standard operations of a vector-like container, it can
1093 also perform an 'or' set operation.
1098 <div class="doc_code">
1102 FirstCondition = 0x1,
1103 SecondCondition = 0x2,
1108 PackedVector<State, 2> Vec1;
1109 Vec1.push_back(FirstCondition);
1111 PackedVector<State, 2> Vec2;
1112 Vec2.push_back(SecondCondition);
1115 return Vec1[0]; // returns 'Both'.
1122 <!-- _______________________________________________________________________ -->
1124 <a name="dss_ilist_traits">ilist_traits</a>
1128 <p><tt>ilist_traits<T></tt> is <tt>ilist<T></tt>'s customization
1129 mechanism. <tt>iplist<T></tt> (and consequently <tt>ilist<T></tt>)
1130 publicly derive from this traits class.</p>
1133 <!-- _______________________________________________________________________ -->
1135 <a name="dss_iplist">iplist</a>
1139 <p><tt>iplist<T></tt> is <tt>ilist<T></tt>'s base and as such
1140 supports a slightly narrower interface. Notably, inserters from
1141 <tt>T&</tt> are absent.</p>
1143 <p><tt>ilist_traits<T></tt> is a public base of this class and can be
1144 used for a wide variety of customizations.</p>
1147 <!-- _______________________________________________________________________ -->
1149 <a name="dss_ilist_node">llvm/ADT/ilist_node.h</a>
1153 <p><tt>ilist_node<T></tt> implements a the forward and backward links
1154 that are expected by the <tt>ilist<T></tt> (and analogous containers)
1155 in the default manner.</p>
1157 <p><tt>ilist_node<T></tt>s are meant to be embedded in the node type
1158 <tt>T</tt>, usually <tt>T</tt> publicly derives from
1159 <tt>ilist_node<T></tt>.</p>
1162 <!-- _______________________________________________________________________ -->
1164 <a name="dss_ilist_sentinel">Sentinels</a>
1168 <p><tt>ilist</tt>s have another specialty that must be considered. To be a good
1169 citizen in the C++ ecosystem, it needs to support the standard container
1170 operations, such as <tt>begin</tt> and <tt>end</tt> iterators, etc. Also, the
1171 <tt>operator--</tt> must work correctly on the <tt>end</tt> iterator in the
1172 case of non-empty <tt>ilist</tt>s.</p>
1174 <p>The only sensible solution to this problem is to allocate a so-called
1175 <i>sentinel</i> along with the intrusive list, which serves as the <tt>end</tt>
1176 iterator, providing the back-link to the last element. However conforming to the
1177 C++ convention it is illegal to <tt>operator++</tt> beyond the sentinel and it
1178 also must not be dereferenced.</p>
1180 <p>These constraints allow for some implementation freedom to the <tt>ilist</tt>
1181 how to allocate and store the sentinel. The corresponding policy is dictated
1182 by <tt>ilist_traits<T></tt>. By default a <tt>T</tt> gets heap-allocated
1183 whenever the need for a sentinel arises.</p>
1185 <p>While the default policy is sufficient in most cases, it may break down when
1186 <tt>T</tt> does not provide a default constructor. Also, in the case of many
1187 instances of <tt>ilist</tt>s, the memory overhead of the associated sentinels
1188 is wasted. To alleviate the situation with numerous and voluminous
1189 <tt>T</tt>-sentinels, sometimes a trick is employed, leading to <i>ghostly
1192 <p>Ghostly sentinels are obtained by specially-crafted <tt>ilist_traits<T></tt>
1193 which superpose the sentinel with the <tt>ilist</tt> instance in memory. Pointer
1194 arithmetic is used to obtain the sentinel, which is relative to the
1195 <tt>ilist</tt>'s <tt>this</tt> pointer. The <tt>ilist</tt> is augmented by an
1196 extra pointer, which serves as the back-link of the sentinel. This is the only
1197 field in the ghostly sentinel which can be legally accessed.</p>
1200 <!-- _______________________________________________________________________ -->
1202 <a name="dss_other">Other Sequential Container options</a>
1206 <p>Other STL containers are available, such as std::string.</p>
1208 <p>There are also various STL adapter classes such as std::queue,
1209 std::priority_queue, std::stack, etc. These provide simplified access to an
1210 underlying container but don't affect the cost of the container itself.</p>
1215 <!-- ======================================================================= -->
1217 <a name="ds_string">String-like containers</a>
1223 There are a variety of ways to pass around and use strings in C and C++, and
1224 LLVM adds a few new options to choose from. Pick the first option on this list
1225 that will do what you need, they are ordered according to their relative cost.
1228 Note that is is generally preferred to <em>not</em> pass strings around as
1229 "<tt>const char*</tt>"'s. These have a number of problems, including the fact
1230 that they cannot represent embedded nul ("\0") characters, and do not have a
1231 length available efficiently. The general replacement for '<tt>const
1232 char*</tt>' is StringRef.
1235 <p>For more information on choosing string containers for APIs, please see
1236 <a href="#string_apis">Passing strings</a>.</p>
1239 <!-- _______________________________________________________________________ -->
1241 <a name="dss_stringref">llvm/ADT/StringRef.h</a>
1246 The StringRef class is a simple value class that contains a pointer to a
1247 character and a length, and is quite related to the <a
1248 href="#dss_arrayref">ArrayRef</a> class (but specialized for arrays of
1249 characters). Because StringRef carries a length with it, it safely handles
1250 strings with embedded nul characters in it, getting the length does not require
1251 a strlen call, and it even has very convenient APIs for slicing and dicing the
1252 character range that it represents.
1256 StringRef is ideal for passing simple strings around that are known to be live,
1257 either because they are C string literals, std::string, a C array, or a
1258 SmallVector. Each of these cases has an efficient implicit conversion to
1259 StringRef, which doesn't result in a dynamic strlen being executed.
1262 <p>StringRef has a few major limitations which make more powerful string
1263 containers useful:</p>
1266 <li>You cannot directly convert a StringRef to a 'const char*' because there is
1267 no way to add a trailing nul (unlike the .c_str() method on various stronger
1271 <li>StringRef doesn't own or keep alive the underlying string bytes.
1272 As such it can easily lead to dangling pointers, and is not suitable for
1273 embedding in datastructures in most cases (instead, use an std::string or
1274 something like that).</li>
1276 <li>For the same reason, StringRef cannot be used as the return value of a
1277 method if the method "computes" the result string. Instead, use
1280 <li>StringRef's do not allow you to mutate the pointed-to string bytes and it
1281 doesn't allow you to insert or remove bytes from the range. For editing
1282 operations like this, it interoperates with the <a
1283 href="#dss_twine">Twine</a> class.</li>
1286 <p>Because of its strengths and limitations, it is very common for a function to
1287 take a StringRef and for a method on an object to return a StringRef that
1288 points into some string that it owns.</p>
1292 <!-- _______________________________________________________________________ -->
1294 <a name="dss_twine">llvm/ADT/Twine.h</a>
1299 The Twine class is used as an intermediary datatype for APIs that want to take
1300 a string that can be constructed inline with a series of concatenations.
1301 Twine works by forming recursive instances of the Twine datatype (a simple
1302 value object) on the stack as temporary objects, linking them together into a
1303 tree which is then linearized when the Twine is consumed. Twine is only safe
1304 to use as the argument to a function, and should always be a const reference,
1309 void foo(const Twine &T);
1313 foo(X + "." + Twine(i));
1316 <p>This example forms a string like "blarg.42" by concatenating the values
1317 together, and does not form intermediate strings containing "blarg" or
1321 <p>Because Twine is constructed with temporary objects on the stack, and
1322 because these instances are destroyed at the end of the current statement,
1323 it is an inherently dangerous API. For example, this simple variant contains
1324 undefined behavior and will probably crash:</p>
1327 void foo(const Twine &T);
1331 const Twine &Tmp = X + "." + Twine(i);
1335 <p>... because the temporaries are destroyed before the call. That said,
1336 Twine's are much more efficient than intermediate std::string temporaries, and
1337 they work really well with StringRef. Just be aware of their limitations.</p>
1342 <!-- _______________________________________________________________________ -->
1344 <a name="dss_smallstring">llvm/ADT/SmallString.h</a>
1349 <p>SmallString is a subclass of <a href="#dss_smallvector">SmallVector</a> that
1350 adds some convenience APIs like += that takes StringRef's. SmallString avoids
1351 allocating memory in the case when the preallocated space is enough to hold its
1352 data, and it calls back to general heap allocation when required. Since it owns
1353 its data, it is very safe to use and supports full mutation of the string.</p>
1355 <p>Like SmallVector's, the big downside to SmallString is their sizeof. While
1356 they are optimized for small strings, they themselves are not particularly
1357 small. This means that they work great for temporary scratch buffers on the
1358 stack, but should not generally be put into the heap: it is very rare to
1359 see a SmallString as the member of a frequently-allocated heap data structure
1360 or returned by-value.
1365 <!-- _______________________________________________________________________ -->
1367 <a name="dss_stdstring">std::string</a>
1372 <p>The standard C++ std::string class is a very general class that (like
1373 SmallString) owns its underlying data. sizeof(std::string) is very reasonable
1374 so it can be embedded into heap data structures and returned by-value.
1375 On the other hand, std::string is highly inefficient for inline editing (e.g.
1376 concatenating a bunch of stuff together) and because it is provided by the
1377 standard library, its performance characteristics depend a lot of the host
1378 standard library (e.g. libc++ and MSVC provide a highly optimized string
1379 class, GCC contains a really slow implementation).
1382 <p>The major disadvantage of std::string is that almost every operation that
1383 makes them larger can allocate memory, which is slow. As such, it is better
1384 to use SmallVector or Twine as a scratch buffer, but then use std::string to
1385 persist the result.</p>
1390 <!-- end of strings -->
1394 <!-- ======================================================================= -->
1396 <a name="ds_set">Set-Like Containers (std::set, SmallSet, SetVector, etc)</a>
1401 <p>Set-like containers are useful when you need to canonicalize multiple values
1402 into a single representation. There are several different choices for how to do
1403 this, providing various trade-offs.</p>
1405 <!-- _______________________________________________________________________ -->
1407 <a name="dss_sortedvectorset">A sorted 'vector'</a>
1412 <p>If you intend to insert a lot of elements, then do a lot of queries, a
1413 great approach is to use a vector (or other sequential container) with
1414 std::sort+std::unique to remove duplicates. This approach works really well if
1415 your usage pattern has these two distinct phases (insert then query), and can be
1416 coupled with a good choice of <a href="#ds_sequential">sequential container</a>.
1420 This combination provides the several nice properties: the result data is
1421 contiguous in memory (good for cache locality), has few allocations, is easy to
1422 address (iterators in the final vector are just indices or pointers), and can be
1423 efficiently queried with a standard binary or radix search.</p>
1427 <!-- _______________________________________________________________________ -->
1429 <a name="dss_smallset">"llvm/ADT/SmallSet.h"</a>
1434 <p>If you have a set-like data structure that is usually small and whose elements
1435 are reasonably small, a <tt>SmallSet<Type, N></tt> is a good choice. This set
1436 has space for N elements in place (thus, if the set is dynamically smaller than
1437 N, no malloc traffic is required) and accesses them with a simple linear search.
1438 When the set grows beyond 'N' elements, it allocates a more expensive representation that
1439 guarantees efficient access (for most types, it falls back to std::set, but for
1440 pointers it uses something far better, <a
1441 href="#dss_smallptrset">SmallPtrSet</a>).</p>
1443 <p>The magic of this class is that it handles small sets extremely efficiently,
1444 but gracefully handles extremely large sets without loss of efficiency. The
1445 drawback is that the interface is quite small: it supports insertion, queries
1446 and erasing, but does not support iteration.</p>
1450 <!-- _______________________________________________________________________ -->
1452 <a name="dss_smallptrset">"llvm/ADT/SmallPtrSet.h"</a>
1457 <p>SmallPtrSet has all the advantages of <tt>SmallSet</tt> (and a <tt>SmallSet</tt> of pointers is
1458 transparently implemented with a <tt>SmallPtrSet</tt>), but also supports iterators. If
1459 more than 'N' insertions are performed, a single quadratically
1460 probed hash table is allocated and grows as needed, providing extremely
1461 efficient access (constant time insertion/deleting/queries with low constant
1462 factors) and is very stingy with malloc traffic.</p>
1464 <p>Note that, unlike <tt>std::set</tt>, the iterators of <tt>SmallPtrSet</tt> are invalidated
1465 whenever an insertion occurs. Also, the values visited by the iterators are not
1466 visited in sorted order.</p>
1470 <!-- _______________________________________________________________________ -->
1472 <a name="dss_denseset">"llvm/ADT/DenseSet.h"</a>
1478 DenseSet is a simple quadratically probed hash table. It excels at supporting
1479 small values: it uses a single allocation to hold all of the pairs that
1480 are currently inserted in the set. DenseSet is a great way to unique small
1481 values that are not simple pointers (use <a
1482 href="#dss_smallptrset">SmallPtrSet</a> for pointers). Note that DenseSet has
1483 the same requirements for the value type that <a
1484 href="#dss_densemap">DenseMap</a> has.
1489 <!-- _______________________________________________________________________ -->
1491 <a name="dss_FoldingSet">"llvm/ADT/FoldingSet.h"</a>
1497 FoldingSet is an aggregate class that is really good at uniquing
1498 expensive-to-create or polymorphic objects. It is a combination of a chained
1499 hash table with intrusive links (uniqued objects are required to inherit from
1500 FoldingSetNode) that uses <a href="#dss_smallvector">SmallVector</a> as part of
1503 <p>Consider a case where you want to implement a "getOrCreateFoo" method for
1504 a complex object (for example, a node in the code generator). The client has a
1505 description of *what* it wants to generate (it knows the opcode and all the
1506 operands), but we don't want to 'new' a node, then try inserting it into a set
1507 only to find out it already exists, at which point we would have to delete it
1508 and return the node that already exists.
1511 <p>To support this style of client, FoldingSet perform a query with a
1512 FoldingSetNodeID (which wraps SmallVector) that can be used to describe the
1513 element that we want to query for. The query either returns the element
1514 matching the ID or it returns an opaque ID that indicates where insertion should
1515 take place. Construction of the ID usually does not require heap traffic.</p>
1517 <p>Because FoldingSet uses intrusive links, it can support polymorphic objects
1518 in the set (for example, you can have SDNode instances mixed with LoadSDNodes).
1519 Because the elements are individually allocated, pointers to the elements are
1520 stable: inserting or removing elements does not invalidate any pointers to other
1526 <!-- _______________________________________________________________________ -->
1528 <a name="dss_set"><set></a>
1533 <p><tt>std::set</tt> is a reasonable all-around set class, which is decent at
1534 many things but great at nothing. std::set allocates memory for each element
1535 inserted (thus it is very malloc intensive) and typically stores three pointers
1536 per element in the set (thus adding a large amount of per-element space
1537 overhead). It offers guaranteed log(n) performance, which is not particularly
1538 fast from a complexity standpoint (particularly if the elements of the set are
1539 expensive to compare, like strings), and has extremely high constant factors for
1540 lookup, insertion and removal.</p>
1542 <p>The advantages of std::set are that its iterators are stable (deleting or
1543 inserting an element from the set does not affect iterators or pointers to other
1544 elements) and that iteration over the set is guaranteed to be in sorted order.
1545 If the elements in the set are large, then the relative overhead of the pointers
1546 and malloc traffic is not a big deal, but if the elements of the set are small,
1547 std::set is almost never a good choice.</p>
1551 <!-- _______________________________________________________________________ -->
1553 <a name="dss_setvector">"llvm/ADT/SetVector.h"</a>
1557 <p>LLVM's SetVector<Type> is an adapter class that combines your choice of
1558 a set-like container along with a <a href="#ds_sequential">Sequential
1559 Container</a>. The important property
1560 that this provides is efficient insertion with uniquing (duplicate elements are
1561 ignored) with iteration support. It implements this by inserting elements into
1562 both a set-like container and the sequential container, using the set-like
1563 container for uniquing and the sequential container for iteration.
1566 <p>The difference between SetVector and other sets is that the order of
1567 iteration is guaranteed to match the order of insertion into the SetVector.
1568 This property is really important for things like sets of pointers. Because
1569 pointer values are non-deterministic (e.g. vary across runs of the program on
1570 different machines), iterating over the pointers in the set will
1571 not be in a well-defined order.</p>
1574 The drawback of SetVector is that it requires twice as much space as a normal
1575 set and has the sum of constant factors from the set-like container and the
1576 sequential container that it uses. Use it *only* if you need to iterate over
1577 the elements in a deterministic order. SetVector is also expensive to delete
1578 elements out of (linear time), unless you use it's "pop_back" method, which is
1582 <p><tt>SetVector</tt> is an adapter class that defaults to
1583 using <tt>std::vector</tt> and a size 16 <tt>SmallSet</tt> for the underlying
1584 containers, so it is quite expensive. However,
1585 <tt>"llvm/ADT/SetVector.h"</tt> also provides a <tt>SmallSetVector</tt>
1586 class, which defaults to using a <tt>SmallVector</tt> and <tt>SmallSet</tt>
1587 of a specified size. If you use this, and if your sets are dynamically
1588 smaller than <tt>N</tt>, you will save a lot of heap traffic.</p>
1592 <!-- _______________________________________________________________________ -->
1594 <a name="dss_uniquevector">"llvm/ADT/UniqueVector.h"</a>
1600 UniqueVector is similar to <a href="#dss_setvector">SetVector</a>, but it
1601 retains a unique ID for each element inserted into the set. It internally
1602 contains a map and a vector, and it assigns a unique ID for each value inserted
1605 <p>UniqueVector is very expensive: its cost is the sum of the cost of
1606 maintaining both the map and vector, it has high complexity, high constant
1607 factors, and produces a lot of malloc traffic. It should be avoided.</p>
1612 <!-- _______________________________________________________________________ -->
1614 <a name="dss_otherset">Other Set-Like Container Options</a>
1620 The STL provides several other options, such as std::multiset and the various
1621 "hash_set" like containers (whether from C++ TR1 or from the SGI library). We
1622 never use hash_set and unordered_set because they are generally very expensive
1623 (each insertion requires a malloc) and very non-portable.
1626 <p>std::multiset is useful if you're not interested in elimination of
1627 duplicates, but has all the drawbacks of std::set. A sorted vector (where you
1628 don't delete duplicate entries) or some other approach is almost always
1635 <!-- ======================================================================= -->
1637 <a name="ds_map">Map-Like Containers (std::map, DenseMap, etc)</a>
1641 Map-like containers are useful when you want to associate data to a key. As
1642 usual, there are a lot of different ways to do this. :)
1644 <!-- _______________________________________________________________________ -->
1646 <a name="dss_sortedvectormap">A sorted 'vector'</a>
1652 If your usage pattern follows a strict insert-then-query approach, you can
1653 trivially use the same approach as <a href="#dss_sortedvectorset">sorted vectors
1654 for set-like containers</a>. The only difference is that your query function
1655 (which uses std::lower_bound to get efficient log(n) lookup) should only compare
1656 the key, not both the key and value. This yields the same advantages as sorted
1661 <!-- _______________________________________________________________________ -->
1663 <a name="dss_stringmap">"llvm/ADT/StringMap.h"</a>
1669 Strings are commonly used as keys in maps, and they are difficult to support
1670 efficiently: they are variable length, inefficient to hash and compare when
1671 long, expensive to copy, etc. StringMap is a specialized container designed to
1672 cope with these issues. It supports mapping an arbitrary range of bytes to an
1673 arbitrary other object.</p>
1675 <p>The StringMap implementation uses a quadratically-probed hash table, where
1676 the buckets store a pointer to the heap allocated entries (and some other
1677 stuff). The entries in the map must be heap allocated because the strings are
1678 variable length. The string data (key) and the element object (value) are
1679 stored in the same allocation with the string data immediately after the element
1680 object. This container guarantees the "<tt>(char*)(&Value+1)</tt>" points
1681 to the key string for a value.</p>
1683 <p>The StringMap is very fast for several reasons: quadratic probing is very
1684 cache efficient for lookups, the hash value of strings in buckets is not
1685 recomputed when looking up an element, StringMap rarely has to touch the
1686 memory for unrelated objects when looking up a value (even when hash collisions
1687 happen), hash table growth does not recompute the hash values for strings
1688 already in the table, and each pair in the map is store in a single allocation
1689 (the string data is stored in the same allocation as the Value of a pair).</p>
1691 <p>StringMap also provides query methods that take byte ranges, so it only ever
1692 copies a string if a value is inserted into the table.</p>
1695 <!-- _______________________________________________________________________ -->
1697 <a name="dss_indexedmap">"llvm/ADT/IndexedMap.h"</a>
1702 IndexedMap is a specialized container for mapping small dense integers (or
1703 values that can be mapped to small dense integers) to some other type. It is
1704 internally implemented as a vector with a mapping function that maps the keys to
1705 the dense integer range.
1709 This is useful for cases like virtual registers in the LLVM code generator: they
1710 have a dense mapping that is offset by a compile-time constant (the first
1711 virtual register ID).</p>
1715 <!-- _______________________________________________________________________ -->
1717 <a name="dss_densemap">"llvm/ADT/DenseMap.h"</a>
1723 DenseMap is a simple quadratically probed hash table. It excels at supporting
1724 small keys and values: it uses a single allocation to hold all of the pairs that
1725 are currently inserted in the map. DenseMap is a great way to map pointers to
1726 pointers, or map other small types to each other.
1730 There are several aspects of DenseMap that you should be aware of, however. The
1731 iterators in a densemap are invalidated whenever an insertion occurs, unlike
1732 map. Also, because DenseMap allocates space for a large number of key/value
1733 pairs (it starts with 64 by default), it will waste a lot of space if your keys
1734 or values are large. Finally, you must implement a partial specialization of
1735 DenseMapInfo for the key that you want, if it isn't already supported. This
1736 is required to tell DenseMap about two special marker values (which can never be
1737 inserted into the map) that it needs internally.</p>
1741 <!-- _______________________________________________________________________ -->
1743 <a name="dss_valuemap">"llvm/ADT/ValueMap.h"</a>
1749 ValueMap is a wrapper around a <a href="#dss_densemap">DenseMap</a> mapping
1750 Value*s (or subclasses) to another type. When a Value is deleted or RAUW'ed,
1751 ValueMap will update itself so the new version of the key is mapped to the same
1752 value, just as if the key were a WeakVH. You can configure exactly how this
1753 happens, and what else happens on these two events, by passing
1754 a <code>Config</code> parameter to the ValueMap template.</p>
1758 <!-- _______________________________________________________________________ -->
1760 <a name="dss_intervalmap">"llvm/ADT/IntervalMap.h"</a>
1765 <p> IntervalMap is a compact map for small keys and values. It maps key
1766 intervals instead of single keys, and it will automatically coalesce adjacent
1767 intervals. When then map only contains a few intervals, they are stored in the
1768 map object itself to avoid allocations.</p>
1770 <p> The IntervalMap iterators are quite big, so they should not be passed around
1771 as STL iterators. The heavyweight iterators allow a smaller data structure.</p>
1775 <!-- _______________________________________________________________________ -->
1777 <a name="dss_map"><map></a>
1783 std::map has similar characteristics to <a href="#dss_set">std::set</a>: it uses
1784 a single allocation per pair inserted into the map, it offers log(n) lookup with
1785 an extremely large constant factor, imposes a space penalty of 3 pointers per
1786 pair in the map, etc.</p>
1788 <p>std::map is most useful when your keys or values are very large, if you need
1789 to iterate over the collection in sorted order, or if you need stable iterators
1790 into the map (i.e. they don't get invalidated if an insertion or deletion of
1791 another element takes place).</p>
1795 <!-- _______________________________________________________________________ -->
1797 <a name="dss_inteqclasses">"llvm/ADT/IntEqClasses.h"</a>
1802 <p>IntEqClasses provides a compact representation of equivalence classes of
1803 small integers. Initially, each integer in the range 0..n-1 has its own
1804 equivalence class. Classes can be joined by passing two class representatives to
1805 the join(a, b) method. Two integers are in the same class when findLeader()
1806 returns the same representative.</p>
1808 <p>Once all equivalence classes are formed, the map can be compressed so each
1809 integer 0..n-1 maps to an equivalence class number in the range 0..m-1, where m
1810 is the total number of equivalence classes. The map must be uncompressed before
1811 it can be edited again.</p>
1815 <!-- _______________________________________________________________________ -->
1817 <a name="dss_othermap">Other Map-Like Container Options</a>
1823 The STL provides several other options, such as std::multimap and the various
1824 "hash_map" like containers (whether from C++ TR1 or from the SGI library). We
1825 never use hash_set and unordered_set because they are generally very expensive
1826 (each insertion requires a malloc) and very non-portable.</p>
1828 <p>std::multimap is useful if you want to map a key to multiple values, but has
1829 all the drawbacks of std::map. A sorted vector or some other approach is almost
1836 <!-- ======================================================================= -->
1838 <a name="ds_bit">Bit storage containers (BitVector, SparseBitVector)</a>
1842 <p>Unlike the other containers, there are only two bit storage containers, and
1843 choosing when to use each is relatively straightforward.</p>
1845 <p>One additional option is
1846 <tt>std::vector<bool></tt>: we discourage its use for two reasons 1) the
1847 implementation in many common compilers (e.g. commonly available versions of
1848 GCC) is extremely inefficient and 2) the C++ standards committee is likely to
1849 deprecate this container and/or change it significantly somehow. In any case,
1850 please don't use it.</p>
1852 <!-- _______________________________________________________________________ -->
1854 <a name="dss_bitvector">BitVector</a>
1858 <p> The BitVector container provides a dynamic size set of bits for manipulation.
1859 It supports individual bit setting/testing, as well as set operations. The set
1860 operations take time O(size of bitvector), but operations are performed one word
1861 at a time, instead of one bit at a time. This makes the BitVector very fast for
1862 set operations compared to other containers. Use the BitVector when you expect
1863 the number of set bits to be high (IE a dense set).
1867 <!-- _______________________________________________________________________ -->
1869 <a name="dss_smallbitvector">SmallBitVector</a>
1873 <p> The SmallBitVector container provides the same interface as BitVector, but
1874 it is optimized for the case where only a small number of bits, less than
1875 25 or so, are needed. It also transparently supports larger bit counts, but
1876 slightly less efficiently than a plain BitVector, so SmallBitVector should
1877 only be used when larger counts are rare.
1881 At this time, SmallBitVector does not support set operations (and, or, xor),
1882 and its operator[] does not provide an assignable lvalue.
1886 <!-- _______________________________________________________________________ -->
1888 <a name="dss_sparsebitvector">SparseBitVector</a>
1892 <p> The SparseBitVector container is much like BitVector, with one major
1893 difference: Only the bits that are set, are stored. This makes the
1894 SparseBitVector much more space efficient than BitVector when the set is sparse,
1895 as well as making set operations O(number of set bits) instead of O(size of
1896 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
1897 (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).
1905 <!-- *********************************************************************** -->
1907 <a name="common">Helpful Hints for Common Operations</a>
1909 <!-- *********************************************************************** -->
1913 <p>This section describes how to perform some very simple transformations of
1914 LLVM code. This is meant to give examples of common idioms used, showing the
1915 practical side of LLVM transformations. <p> Because this is a "how-to" section,
1916 you should also read about the main classes that you will be working with. The
1917 <a href="#coreclasses">Core LLVM Class Hierarchy Reference</a> contains details
1918 and descriptions of the main classes that you should know about.</p>
1920 <!-- NOTE: this section should be heavy on example code -->
1921 <!-- ======================================================================= -->
1923 <a name="inspection">Basic Inspection and Traversal Routines</a>
1928 <p>The LLVM compiler infrastructure have many different data structures that may
1929 be traversed. Following the example of the C++ standard template library, the
1930 techniques used to traverse these various data structures are all basically the
1931 same. For a enumerable sequence of values, the <tt>XXXbegin()</tt> function (or
1932 method) returns an iterator to the start of the sequence, the <tt>XXXend()</tt>
1933 function returns an iterator pointing to one past the last valid element of the
1934 sequence, and there is some <tt>XXXiterator</tt> data type that is common
1935 between the two operations.</p>
1937 <p>Because the pattern for iteration is common across many different aspects of
1938 the program representation, the standard template library algorithms may be used
1939 on them, and it is easier to remember how to iterate. First we show a few common
1940 examples of the data structures that need to be traversed. Other data
1941 structures are traversed in very similar ways.</p>
1943 <!-- _______________________________________________________________________ -->
1945 <a name="iterate_function">Iterating over the </a><a
1946 href="#BasicBlock"><tt>BasicBlock</tt></a>s in a <a
1947 href="#Function"><tt>Function</tt></a>
1952 <p>It's quite common to have a <tt>Function</tt> instance that you'd like to
1953 transform in some way; in particular, you'd like to manipulate its
1954 <tt>BasicBlock</tt>s. To facilitate this, you'll need to iterate over all of
1955 the <tt>BasicBlock</tt>s that constitute the <tt>Function</tt>. The following is
1956 an example that prints the name of a <tt>BasicBlock</tt> and the number of
1957 <tt>Instruction</tt>s it contains:</p>
1959 <div class="doc_code">
1961 // <i>func is a pointer to a Function instance</i>
1962 for (Function::iterator i = func->begin(), e = func->end(); i != e; ++i)
1963 // <i>Print out the name of the basic block if it has one, and then the</i>
1964 // <i>number of instructions that it contains</i>
1965 errs() << "Basic block (name=" << i->getName() << ") has "
1966 << i->size() << " instructions.\n";
1970 <p>Note that i can be used as if it were a pointer for the purposes of
1971 invoking member functions of the <tt>Instruction</tt> class. This is
1972 because the indirection operator is overloaded for the iterator
1973 classes. In the above code, the expression <tt>i->size()</tt> is
1974 exactly equivalent to <tt>(*i).size()</tt> just like you'd expect.</p>
1978 <!-- _______________________________________________________________________ -->
1980 <a name="iterate_basicblock">Iterating over the </a><a
1981 href="#Instruction"><tt>Instruction</tt></a>s in a <a
1982 href="#BasicBlock"><tt>BasicBlock</tt></a>
1987 <p>Just like when dealing with <tt>BasicBlock</tt>s in <tt>Function</tt>s, it's
1988 easy to iterate over the individual instructions that make up
1989 <tt>BasicBlock</tt>s. Here's a code snippet that prints out each instruction in
1990 a <tt>BasicBlock</tt>:</p>
1992 <div class="doc_code">
1994 // <i>blk is a pointer to a BasicBlock instance</i>
1995 for (BasicBlock::iterator i = blk->begin(), e = blk->end(); i != e; ++i)
1996 // <i>The next statement works since operator<<(ostream&,...)</i>
1997 // <i>is overloaded for Instruction&</i>
1998 errs() << *i << "\n";
2002 <p>However, this isn't really the best way to print out the contents of a
2003 <tt>BasicBlock</tt>! Since the ostream operators are overloaded for virtually
2004 anything you'll care about, you could have just invoked the print routine on the
2005 basic block itself: <tt>errs() << *blk << "\n";</tt>.</p>
2009 <!-- _______________________________________________________________________ -->
2011 <a name="iterate_institer">Iterating over the </a><a
2012 href="#Instruction"><tt>Instruction</tt></a>s in a <a
2013 href="#Function"><tt>Function</tt></a>
2018 <p>If you're finding that you commonly iterate over a <tt>Function</tt>'s
2019 <tt>BasicBlock</tt>s and then that <tt>BasicBlock</tt>'s <tt>Instruction</tt>s,
2020 <tt>InstIterator</tt> should be used instead. You'll need to include <a
2021 href="/doxygen/InstIterator_8h-source.html"><tt>llvm/Support/InstIterator.h</tt></a>,
2022 and then instantiate <tt>InstIterator</tt>s explicitly in your code. Here's a
2023 small example that shows how to dump all instructions in a function to the standard error stream:<p>
2025 <div class="doc_code">
2027 #include "<a href="/doxygen/InstIterator_8h-source.html">llvm/Support/InstIterator.h</a>"
2029 // <i>F is a pointer to a Function instance</i>
2030 for (inst_iterator I = inst_begin(F), E = inst_end(F); I != E; ++I)
2031 errs() << *I << "\n";
2035 <p>Easy, isn't it? You can also use <tt>InstIterator</tt>s to fill a
2036 work list with its initial contents. For example, if you wanted to
2037 initialize a work list to contain all instructions in a <tt>Function</tt>
2038 F, all you would need to do is something like:</p>
2040 <div class="doc_code">
2042 std::set<Instruction*> worklist;
2043 // or better yet, SmallPtrSet<Instruction*, 64> worklist;
2045 for (inst_iterator I = inst_begin(F), E = inst_end(F); I != E; ++I)
2046 worklist.insert(&*I);
2050 <p>The STL set <tt>worklist</tt> would now contain all instructions in the
2051 <tt>Function</tt> pointed to by F.</p>
2055 <!-- _______________________________________________________________________ -->
2057 <a name="iterate_convert">Turning an iterator into a class pointer (and
2063 <p>Sometimes, it'll be useful to grab a reference (or pointer) to a class
2064 instance when all you've got at hand is an iterator. Well, extracting
2065 a reference or a pointer from an iterator is very straight-forward.
2066 Assuming that <tt>i</tt> is a <tt>BasicBlock::iterator</tt> and <tt>j</tt>
2067 is a <tt>BasicBlock::const_iterator</tt>:</p>
2069 <div class="doc_code">
2071 Instruction& inst = *i; // <i>Grab reference to instruction reference</i>
2072 Instruction* pinst = &*i; // <i>Grab pointer to instruction reference</i>
2073 const Instruction& inst = *j;
2077 <p>However, the iterators you'll be working with in the LLVM framework are
2078 special: they will automatically convert to a ptr-to-instance type whenever they
2079 need to. Instead of dereferencing the iterator and then taking the address of
2080 the result, you can simply assign the iterator to the proper pointer type and
2081 you get the dereference and address-of operation as a result of the assignment
2082 (behind the scenes, this is a result of overloading casting mechanisms). Thus
2083 the last line of the last example,</p>
2085 <div class="doc_code">
2087 Instruction *pinst = &*i;
2091 <p>is semantically equivalent to</p>
2093 <div class="doc_code">
2095 Instruction *pinst = i;
2099 <p>It's also possible to turn a class pointer into the corresponding iterator,
2100 and this is a constant time operation (very efficient). The following code
2101 snippet illustrates use of the conversion constructors provided by LLVM
2102 iterators. By using these, you can explicitly grab the iterator of something
2103 without actually obtaining it via iteration over some structure:</p>
2105 <div class="doc_code">
2107 void printNextInstruction(Instruction* inst) {
2108 BasicBlock::iterator it(inst);
2109 ++it; // <i>After this line, it refers to the instruction after *inst</i>
2110 if (it != inst->getParent()->end()) errs() << *it << "\n";
2115 <p>Unfortunately, these implicit conversions come at a cost; they prevent
2116 these iterators from conforming to standard iterator conventions, and thus
2117 from being usable with standard algorithms and containers. For example, they
2118 prevent the following code, where <tt>B</tt> is a <tt>BasicBlock</tt>,
2121 <div class="doc_code">
2123 llvm::SmallVector<llvm::Instruction *, 16>(B->begin(), B->end());
2127 <p>Because of this, these implicit conversions may be removed some day,
2128 and <tt>operator*</tt> changed to return a pointer instead of a reference.</p>
2132 <!--_______________________________________________________________________-->
2134 <a name="iterate_complex">Finding call sites: a slightly more complex
2140 <p>Say that you're writing a FunctionPass and would like to count all the
2141 locations in the entire module (that is, across every <tt>Function</tt>) where a
2142 certain function (i.e., some <tt>Function</tt>*) is already in scope. As you'll
2143 learn later, you may want to use an <tt>InstVisitor</tt> to accomplish this in a
2144 much more straight-forward manner, but this example will allow us to explore how
2145 you'd do it if you didn't have <tt>InstVisitor</tt> around. In pseudo-code, this
2146 is what we want to do:</p>
2148 <div class="doc_code">
2150 initialize callCounter to zero
2151 for each Function f in the Module
2152 for each BasicBlock b in f
2153 for each Instruction i in b
2154 if (i is a CallInst and calls the given function)
2155 increment callCounter
2159 <p>And the actual code is (remember, because we're writing a
2160 <tt>FunctionPass</tt>, our <tt>FunctionPass</tt>-derived class simply has to
2161 override the <tt>runOnFunction</tt> method):</p>
2163 <div class="doc_code">
2165 Function* targetFunc = ...;
2167 class OurFunctionPass : public FunctionPass {
2169 OurFunctionPass(): callCounter(0) { }
2171 virtual runOnFunction(Function& F) {
2172 for (Function::iterator b = F.begin(), be = F.end(); b != be; ++b) {
2173 for (BasicBlock::iterator i = b->begin(), ie = b->end(); i != ie; ++i) {
2174 if (<a href="#CallInst">CallInst</a>* callInst = <a href="#isa">dyn_cast</a><<a
2175 href="#CallInst">CallInst</a>>(&*i)) {
2176 // <i>We know we've encountered a call instruction, so we</i>
2177 // <i>need to determine if it's a call to the</i>
2178 // <i>function pointed to by m_func or not.</i>
2179 if (callInst->getCalledFunction() == targetFunc)
2187 unsigned callCounter;
2194 <!--_______________________________________________________________________-->
2196 <a name="calls_and_invokes">Treating calls and invokes the same way</a>
2201 <p>You may have noticed that the previous example was a bit oversimplified in
2202 that it did not deal with call sites generated by 'invoke' instructions. In
2203 this, and in other situations, you may find that you want to treat
2204 <tt>CallInst</tt>s and <tt>InvokeInst</tt>s the same way, even though their
2205 most-specific common base class is <tt>Instruction</tt>, which includes lots of
2206 less closely-related things. For these cases, LLVM provides a handy wrapper
2208 href="http://llvm.org/doxygen/classllvm_1_1CallSite.html"><tt>CallSite</tt></a>.
2209 It is essentially a wrapper around an <tt>Instruction</tt> pointer, with some
2210 methods that provide functionality common to <tt>CallInst</tt>s and
2211 <tt>InvokeInst</tt>s.</p>
2213 <p>This class has "value semantics": it should be passed by value, not by
2214 reference and it should not be dynamically allocated or deallocated using
2215 <tt>operator new</tt> or <tt>operator delete</tt>. It is efficiently copyable,
2216 assignable and constructable, with costs equivalents to that of a bare pointer.
2217 If you look at its definition, it has only a single pointer member.</p>
2221 <!--_______________________________________________________________________-->
2223 <a name="iterate_chains">Iterating over def-use & use-def chains</a>
2228 <p>Frequently, we might have an instance of the <a
2229 href="/doxygen/classllvm_1_1Value.html">Value Class</a> and we want to
2230 determine which <tt>User</tt>s use the <tt>Value</tt>. The list of all
2231 <tt>User</tt>s of a particular <tt>Value</tt> is called a <i>def-use</i> chain.
2232 For example, let's say we have a <tt>Function*</tt> named <tt>F</tt> to a
2233 particular function <tt>foo</tt>. Finding all of the instructions that
2234 <i>use</i> <tt>foo</tt> is as simple as iterating over the <i>def-use</i> chain
2237 <div class="doc_code">
2241 for (Value::use_iterator i = F->use_begin(), e = F->use_end(); i != e; ++i)
2242 if (Instruction *Inst = dyn_cast<Instruction>(*i)) {
2243 errs() << "F is used in instruction:\n";
2244 errs() << *Inst << "\n";
2249 <p>Note that dereferencing a <tt>Value::use_iterator</tt> is not a very cheap
2250 operation. Instead of performing <tt>*i</tt> above several times, consider
2251 doing it only once in the loop body and reusing its result.</p>
2253 <p>Alternatively, it's common to have an instance of the <a
2254 href="/doxygen/classllvm_1_1User.html">User Class</a> and need to know what
2255 <tt>Value</tt>s are used by it. The list of all <tt>Value</tt>s used by a
2256 <tt>User</tt> is known as a <i>use-def</i> chain. Instances of class
2257 <tt>Instruction</tt> are common <tt>User</tt>s, so we might want to iterate over
2258 all of the values that a particular instruction uses (that is, the operands of
2259 the particular <tt>Instruction</tt>):</p>
2261 <div class="doc_code">
2263 Instruction *pi = ...;
2265 for (User::op_iterator i = pi->op_begin(), e = pi->op_end(); i != e; ++i) {
2272 <p>Declaring objects as <tt>const</tt> is an important tool of enforcing
2273 mutation free algorithms (such as analyses, etc.). For this purpose above
2274 iterators come in constant flavors as <tt>Value::const_use_iterator</tt>
2275 and <tt>Value::const_op_iterator</tt>. They automatically arise when
2276 calling <tt>use/op_begin()</tt> on <tt>const Value*</tt>s or
2277 <tt>const User*</tt>s respectively. Upon dereferencing, they return
2278 <tt>const Use*</tt>s. Otherwise the above patterns remain unchanged.</p>
2282 <!--_______________________________________________________________________-->
2284 <a name="iterate_preds">Iterating over predecessors &
2285 successors of blocks</a>
2290 <p>Iterating over the predecessors and successors of a block is quite easy
2291 with the routines defined in <tt>"llvm/Support/CFG.h"</tt>. Just use code like
2292 this to iterate over all predecessors of BB:</p>
2294 <div class="doc_code">
2296 #include "llvm/Support/CFG.h"
2297 BasicBlock *BB = ...;
2299 for (pred_iterator PI = pred_begin(BB), E = pred_end(BB); PI != E; ++PI) {
2300 BasicBlock *Pred = *PI;
2306 <p>Similarly, to iterate over successors use
2307 succ_iterator/succ_begin/succ_end.</p>
2313 <!-- ======================================================================= -->
2315 <a name="simplechanges">Making simple changes</a>
2320 <p>There are some primitive transformation operations present in the LLVM
2321 infrastructure that are worth knowing about. When performing
2322 transformations, it's fairly common to manipulate the contents of basic
2323 blocks. This section describes some of the common methods for doing so
2324 and gives example code.</p>
2326 <!--_______________________________________________________________________-->
2328 <a name="schanges_creating">Creating and inserting new
2329 <tt>Instruction</tt>s</a>
2334 <p><i>Instantiating Instructions</i></p>
2336 <p>Creation of <tt>Instruction</tt>s is straight-forward: simply call the
2337 constructor for the kind of instruction to instantiate and provide the necessary
2338 parameters. For example, an <tt>AllocaInst</tt> only <i>requires</i> a
2339 (const-ptr-to) <tt>Type</tt>. Thus:</p>
2341 <div class="doc_code">
2343 AllocaInst* ai = new AllocaInst(Type::Int32Ty);
2347 <p>will create an <tt>AllocaInst</tt> instance that represents the allocation of
2348 one integer in the current stack frame, at run time. Each <tt>Instruction</tt>
2349 subclass is likely to have varying default parameters which change the semantics
2350 of the instruction, so refer to the <a
2351 href="/doxygen/classllvm_1_1Instruction.html">doxygen documentation for the subclass of
2352 Instruction</a> that you're interested in instantiating.</p>
2354 <p><i>Naming values</i></p>
2356 <p>It is very useful to name the values of instructions when you're able to, as
2357 this facilitates the debugging of your transformations. If you end up looking
2358 at generated LLVM machine code, you definitely want to have logical names
2359 associated with the results of instructions! By supplying a value for the
2360 <tt>Name</tt> (default) parameter of the <tt>Instruction</tt> constructor, you
2361 associate a logical name with the result of the instruction's execution at
2362 run time. For example, say that I'm writing a transformation that dynamically
2363 allocates space for an integer on the stack, and that integer is going to be
2364 used as some kind of index by some other code. To accomplish this, I place an
2365 <tt>AllocaInst</tt> at the first point in the first <tt>BasicBlock</tt> of some
2366 <tt>Function</tt>, and I'm intending to use it within the same
2367 <tt>Function</tt>. I might do:</p>
2369 <div class="doc_code">
2371 AllocaInst* pa = new AllocaInst(Type::Int32Ty, 0, "indexLoc");
2375 <p>where <tt>indexLoc</tt> is now the logical name of the instruction's
2376 execution value, which is a pointer to an integer on the run time stack.</p>
2378 <p><i>Inserting instructions</i></p>
2380 <p>There are essentially two ways to insert an <tt>Instruction</tt>
2381 into an existing sequence of instructions that form a <tt>BasicBlock</tt>:</p>
2384 <li>Insertion into an explicit instruction list
2386 <p>Given a <tt>BasicBlock* pb</tt>, an <tt>Instruction* pi</tt> within that
2387 <tt>BasicBlock</tt>, and a newly-created instruction we wish to insert
2388 before <tt>*pi</tt>, we do the following: </p>
2390 <div class="doc_code">
2392 BasicBlock *pb = ...;
2393 Instruction *pi = ...;
2394 Instruction *newInst = new Instruction(...);
2396 pb->getInstList().insert(pi, newInst); // <i>Inserts newInst before pi in pb</i>
2400 <p>Appending to the end of a <tt>BasicBlock</tt> is so common that
2401 the <tt>Instruction</tt> class and <tt>Instruction</tt>-derived
2402 classes provide constructors which take a pointer to a
2403 <tt>BasicBlock</tt> to be appended to. For example code that
2406 <div class="doc_code">
2408 BasicBlock *pb = ...;
2409 Instruction *newInst = new Instruction(...);
2411 pb->getInstList().push_back(newInst); // <i>Appends newInst to pb</i>
2417 <div class="doc_code">
2419 BasicBlock *pb = ...;
2420 Instruction *newInst = new Instruction(..., pb);
2424 <p>which is much cleaner, especially if you are creating
2425 long instruction streams.</p></li>
2427 <li>Insertion into an implicit instruction list
2429 <p><tt>Instruction</tt> instances that are already in <tt>BasicBlock</tt>s
2430 are implicitly associated with an existing instruction list: the instruction
2431 list of the enclosing basic block. Thus, we could have accomplished the same
2432 thing as the above code without being given a <tt>BasicBlock</tt> by doing:
2435 <div class="doc_code">
2437 Instruction *pi = ...;
2438 Instruction *newInst = new Instruction(...);
2440 pi->getParent()->getInstList().insert(pi, newInst);
2444 <p>In fact, this sequence of steps occurs so frequently that the
2445 <tt>Instruction</tt> class and <tt>Instruction</tt>-derived classes provide
2446 constructors which take (as a default parameter) a pointer to an
2447 <tt>Instruction</tt> which the newly-created <tt>Instruction</tt> should
2448 precede. That is, <tt>Instruction</tt> constructors are capable of
2449 inserting the newly-created instance into the <tt>BasicBlock</tt> of a
2450 provided instruction, immediately before that instruction. Using an
2451 <tt>Instruction</tt> constructor with a <tt>insertBefore</tt> (default)
2452 parameter, the above code becomes:</p>
2454 <div class="doc_code">
2456 Instruction* pi = ...;
2457 Instruction* newInst = new Instruction(..., pi);
2461 <p>which is much cleaner, especially if you're creating a lot of
2462 instructions and adding them to <tt>BasicBlock</tt>s.</p></li>
2467 <!--_______________________________________________________________________-->
2469 <a name="schanges_deleting">Deleting <tt>Instruction</tt>s</a>
2474 <p>Deleting an instruction from an existing sequence of instructions that form a
2475 <a href="#BasicBlock"><tt>BasicBlock</tt></a> is very straight-forward: just
2476 call the instruction's eraseFromParent() method. For example:</p>
2478 <div class="doc_code">
2480 <a href="#Instruction">Instruction</a> *I = .. ;
2481 I->eraseFromParent();
2485 <p>This unlinks the instruction from its containing basic block and deletes
2486 it. If you'd just like to unlink the instruction from its containing basic
2487 block but not delete it, you can use the <tt>removeFromParent()</tt> method.</p>
2491 <!--_______________________________________________________________________-->
2493 <a name="schanges_replacing">Replacing an <tt>Instruction</tt> with another
2499 <p><i>Replacing individual instructions</i></p>
2501 <p>Including "<a href="/doxygen/BasicBlockUtils_8h-source.html">llvm/Transforms/Utils/BasicBlockUtils.h</a>"
2502 permits use of two very useful replace functions: <tt>ReplaceInstWithValue</tt>
2503 and <tt>ReplaceInstWithInst</tt>.</p>
2505 <h5><a name="schanges_deleting">Deleting <tt>Instruction</tt>s</a></h5>
2508 <li><tt>ReplaceInstWithValue</tt>
2510 <p>This function replaces all uses of a given instruction with a value,
2511 and then removes the original instruction. The following example
2512 illustrates the replacement of the result of a particular
2513 <tt>AllocaInst</tt> that allocates memory for a single integer with a null
2514 pointer to an integer.</p>
2516 <div class="doc_code">
2518 AllocaInst* instToReplace = ...;
2519 BasicBlock::iterator ii(instToReplace);
2521 ReplaceInstWithValue(instToReplace->getParent()->getInstList(), ii,
2522 Constant::getNullValue(PointerType::getUnqual(Type::Int32Ty)));
2525 <li><tt>ReplaceInstWithInst</tt>
2527 <p>This function replaces a particular instruction with another
2528 instruction, inserting the new instruction into the basic block at the
2529 location where the old instruction was, and replacing any uses of the old
2530 instruction with the new instruction. The following example illustrates
2531 the replacement of one <tt>AllocaInst</tt> with another.</p>
2533 <div class="doc_code">
2535 AllocaInst* instToReplace = ...;
2536 BasicBlock::iterator ii(instToReplace);
2538 ReplaceInstWithInst(instToReplace->getParent()->getInstList(), ii,
2539 new AllocaInst(Type::Int32Ty, 0, "ptrToReplacedInt"));
2543 <p><i>Replacing multiple uses of <tt>User</tt>s and <tt>Value</tt>s</i></p>
2545 <p>You can use <tt>Value::replaceAllUsesWith</tt> and
2546 <tt>User::replaceUsesOfWith</tt> to change more than one use at a time. See the
2547 doxygen documentation for the <a href="/doxygen/classllvm_1_1Value.html">Value Class</a>
2548 and <a href="/doxygen/classllvm_1_1User.html">User Class</a>, respectively, for more
2551 <!-- Value::replaceAllUsesWith User::replaceUsesOfWith Point out:
2552 include/llvm/Transforms/Utils/ especially BasicBlockUtils.h with:
2553 ReplaceInstWithValue, ReplaceInstWithInst -->
2557 <!--_______________________________________________________________________-->
2559 <a name="schanges_deletingGV">Deleting <tt>GlobalVariable</tt>s</a>
2564 <p>Deleting a global variable from a module is just as easy as deleting an
2565 Instruction. First, you must have a pointer to the global variable that you wish
2566 to delete. You use this pointer to erase it from its parent, the module.
2569 <div class="doc_code">
2571 <a href="#GlobalVariable">GlobalVariable</a> *GV = .. ;
2573 GV->eraseFromParent();
2581 <!-- ======================================================================= -->
2583 <a name="create_types">How to Create Types</a>
2588 <p>In generating IR, you may need some complex types. If you know these types
2589 statically, you can use <tt>TypeBuilder<...>::get()</tt>, defined
2590 in <tt>llvm/Support/TypeBuilder.h</tt>, to retrieve them. <tt>TypeBuilder</tt>
2591 has two forms depending on whether you're building types for cross-compilation
2592 or native library use. <tt>TypeBuilder<T, true></tt> requires
2593 that <tt>T</tt> be independent of the host environment, meaning that it's built
2595 the <a href="/doxygen/namespacellvm_1_1types.html"><tt>llvm::types</tt></a>
2596 namespace and pointers, functions, arrays, etc. built of
2597 those. <tt>TypeBuilder<T, false></tt> additionally allows native C types
2598 whose size may depend on the host compiler. For example,</p>
2600 <div class="doc_code">
2602 FunctionType *ft = TypeBuilder<types::i<8>(types::i<32>*), true>::get();
2606 <p>is easier to read and write than the equivalent</p>
2608 <div class="doc_code">
2610 std::vector<const Type*> params;
2611 params.push_back(PointerType::getUnqual(Type::Int32Ty));
2612 FunctionType *ft = FunctionType::get(Type::Int8Ty, params, false);
2616 <p>See the <a href="/doxygen/TypeBuilder_8h-source.html#l00001">class
2617 comment</a> for more details.</p>
2623 <!-- *********************************************************************** -->
2625 <a name="threading">Threads and LLVM</a>
2627 <!-- *********************************************************************** -->
2631 This section describes the interaction of the LLVM APIs with multithreading,
2632 both on the part of client applications, and in the JIT, in the hosted
2637 Note that LLVM's support for multithreading is still relatively young. Up
2638 through version 2.5, the execution of threaded hosted applications was
2639 supported, but not threaded client access to the APIs. While this use case is
2640 now supported, clients <em>must</em> adhere to the guidelines specified below to
2641 ensure proper operation in multithreaded mode.
2645 Note that, on Unix-like platforms, LLVM requires the presence of GCC's atomic
2646 intrinsics in order to support threaded operation. If you need a
2647 multhreading-capable LLVM on a platform without a suitably modern system
2648 compiler, consider compiling LLVM and LLVM-GCC in single-threaded mode, and
2649 using the resultant compiler to build a copy of LLVM with multithreading
2653 <!-- ======================================================================= -->
2655 <a name="startmultithreaded">Entering and Exiting Multithreaded Mode</a>
2661 In order to properly protect its internal data structures while avoiding
2662 excessive locking overhead in the single-threaded case, the LLVM must intialize
2663 certain data structures necessary to provide guards around its internals. To do
2664 so, the client program must invoke <tt>llvm_start_multithreaded()</tt> before
2665 making any concurrent LLVM API calls. To subsequently tear down these
2666 structures, use the <tt>llvm_stop_multithreaded()</tt> call. You can also use
2667 the <tt>llvm_is_multithreaded()</tt> call to check the status of multithreaded
2672 Note that both of these calls must be made <em>in isolation</em>. That is to
2673 say that no other LLVM API calls may be executing at any time during the
2674 execution of <tt>llvm_start_multithreaded()</tt> or <tt>llvm_stop_multithreaded
2675 </tt>. It's is the client's responsibility to enforce this isolation.
2679 The return value of <tt>llvm_start_multithreaded()</tt> indicates the success or
2680 failure of the initialization. Failure typically indicates that your copy of
2681 LLVM was built without multithreading support, typically because GCC atomic
2682 intrinsics were not found in your system compiler. In this case, the LLVM API
2683 will not be safe for concurrent calls. However, it <em>will</em> be safe for
2684 hosting threaded applications in the JIT, though <a href="#jitthreading">care
2685 must be taken</a> to ensure that side exits and the like do not accidentally
2686 result in concurrent LLVM API calls.
2690 <!-- ======================================================================= -->
2692 <a name="shutdown">Ending Execution with <tt>llvm_shutdown()</tt></a>
2697 When you are done using the LLVM APIs, you should call <tt>llvm_shutdown()</tt>
2698 to deallocate memory used for internal structures. This will also invoke
2699 <tt>llvm_stop_multithreaded()</tt> if LLVM is operating in multithreaded mode.
2700 As such, <tt>llvm_shutdown()</tt> requires the same isolation guarantees as
2701 <tt>llvm_stop_multithreaded()</tt>.
2705 Note that, if you use scope-based shutdown, you can use the
2706 <tt>llvm_shutdown_obj</tt> class, which calls <tt>llvm_shutdown()</tt> in its
2710 <!-- ======================================================================= -->
2712 <a name="managedstatic">Lazy Initialization with <tt>ManagedStatic</tt></a>
2717 <tt>ManagedStatic</tt> is a utility class in LLVM used to implement static
2718 initialization of static resources, such as the global type tables. Before the
2719 invocation of <tt>llvm_shutdown()</tt>, it implements a simple lazy
2720 initialization scheme. Once <tt>llvm_start_multithreaded()</tt> returns,
2721 however, it uses double-checked locking to implement thread-safe lazy
2726 Note that, because no other threads are allowed to issue LLVM API calls before
2727 <tt>llvm_start_multithreaded()</tt> returns, it is possible to have
2728 <tt>ManagedStatic</tt>s of <tt>llvm::sys::Mutex</tt>s.
2732 The <tt>llvm_acquire_global_lock()</tt> and <tt>llvm_release_global_lock</tt>
2733 APIs provide access to the global lock used to implement the double-checked
2734 locking for lazy initialization. These should only be used internally to LLVM,
2735 and only if you know what you're doing!
2739 <!-- ======================================================================= -->
2741 <a name="llvmcontext">Achieving Isolation with <tt>LLVMContext</tt></a>
2746 <tt>LLVMContext</tt> is an opaque class in the LLVM API which clients can use
2747 to operate multiple, isolated instances of LLVM concurrently within the same
2748 address space. For instance, in a hypothetical compile-server, the compilation
2749 of an individual translation unit is conceptually independent from all the
2750 others, and it would be desirable to be able to compile incoming translation
2751 units concurrently on independent server threads. Fortunately,
2752 <tt>LLVMContext</tt> exists to enable just this kind of scenario!
2756 Conceptually, <tt>LLVMContext</tt> provides isolation. Every LLVM entity
2757 (<tt>Module</tt>s, <tt>Value</tt>s, <tt>Type</tt>s, <tt>Constant</tt>s, etc.)
2758 in LLVM's in-memory IR belongs to an <tt>LLVMContext</tt>. Entities in
2759 different contexts <em>cannot</em> interact with each other: <tt>Module</tt>s in
2760 different contexts cannot be linked together, <tt>Function</tt>s cannot be added
2761 to <tt>Module</tt>s in different contexts, etc. What this means is that is is
2762 safe to compile on multiple threads simultaneously, as long as no two threads
2763 operate on entities within the same context.
2767 In practice, very few places in the API require the explicit specification of a
2768 <tt>LLVMContext</tt>, other than the <tt>Type</tt> creation/lookup APIs.
2769 Because every <tt>Type</tt> carries a reference to its owning context, most
2770 other entities can determine what context they belong to by looking at their
2771 own <tt>Type</tt>. If you are adding new entities to LLVM IR, please try to
2772 maintain this interface design.
2776 For clients that do <em>not</em> require the benefits of isolation, LLVM
2777 provides a convenience API <tt>getGlobalContext()</tt>. This returns a global,
2778 lazily initialized <tt>LLVMContext</tt> that may be used in situations where
2779 isolation is not a concern.
2783 <!-- ======================================================================= -->
2785 <a name="jitthreading">Threads and the JIT</a>
2790 LLVM's "eager" JIT compiler is safe to use in threaded programs. Multiple
2791 threads can call <tt>ExecutionEngine::getPointerToFunction()</tt> or
2792 <tt>ExecutionEngine::runFunction()</tt> concurrently, and multiple threads can
2793 run code output by the JIT concurrently. The user must still ensure that only
2794 one thread accesses IR in a given <tt>LLVMContext</tt> while another thread
2795 might be modifying it. One way to do that is to always hold the JIT lock while
2796 accessing IR outside the JIT (the JIT <em>modifies</em> the IR by adding
2797 <tt>CallbackVH</tt>s). Another way is to only
2798 call <tt>getPointerToFunction()</tt> from the <tt>LLVMContext</tt>'s thread.
2801 <p>When the JIT is configured to compile lazily (using
2802 <tt>ExecutionEngine::DisableLazyCompilation(false)</tt>), there is currently a
2803 <a href="http://llvm.org/bugs/show_bug.cgi?id=5184">race condition</a> in
2804 updating call sites after a function is lazily-jitted. It's still possible to
2805 use the lazy JIT in a threaded program if you ensure that only one thread at a
2806 time can call any particular lazy stub and that the JIT lock guards any IR
2807 access, but we suggest using only the eager JIT in threaded programs.
2813 <!-- *********************************************************************** -->
2815 <a name="advanced">Advanced Topics</a>
2817 <!-- *********************************************************************** -->
2821 This section describes some of the advanced or obscure API's that most clients
2822 do not need to be aware of. These API's tend manage the inner workings of the
2823 LLVM system, and only need to be accessed in unusual circumstances.
2827 <!-- ======================================================================= -->
2829 <a name="SymbolTable">The <tt>ValueSymbolTable</tt> class</a>
2833 <p>The <tt><a href="http://llvm.org/doxygen/classllvm_1_1ValueSymbolTable.html">
2834 ValueSymbolTable</a></tt> class provides a symbol table that the <a
2835 href="#Function"><tt>Function</tt></a> and <a href="#Module">
2836 <tt>Module</tt></a> classes use for naming value definitions. The symbol table
2837 can provide a name for any <a href="#Value"><tt>Value</tt></a>.
2840 <p>Note that the <tt>SymbolTable</tt> class should not be directly accessed
2841 by most clients. It should only be used when iteration over the symbol table
2842 names themselves are required, which is very special purpose. Note that not
2844 <tt><a href="#Value">Value</a></tt>s have names, and those without names (i.e. they have
2845 an empty name) do not exist in the symbol table.
2848 <p>Symbol tables support iteration over the values in the symbol
2849 table with <tt>begin/end/iterator</tt> and supports querying to see if a
2850 specific name is in the symbol table (with <tt>lookup</tt>). The
2851 <tt>ValueSymbolTable</tt> class exposes no public mutator methods, instead,
2852 simply call <tt>setName</tt> on a value, which will autoinsert it into the
2853 appropriate symbol table.</p>
2859 <!-- ======================================================================= -->
2861 <a name="UserLayout">The <tt>User</tt> and owned <tt>Use</tt> classes' memory layout</a>
2865 <p>The <tt><a href="http://llvm.org/doxygen/classllvm_1_1User.html">
2866 User</a></tt> class provides a basis for expressing the ownership of <tt>User</tt>
2867 towards other <tt><a href="http://llvm.org/doxygen/classllvm_1_1Value.html">
2868 Value</a></tt>s. The <tt><a href="http://llvm.org/doxygen/classllvm_1_1Use.html">
2869 Use</a></tt> helper class is employed to do the bookkeeping and to facilitate <i>O(1)</i>
2870 addition and removal.</p>
2872 <!-- ______________________________________________________________________ -->
2875 Interaction and relationship between <tt>User</tt> and <tt>Use</tt> objects
2881 A subclass of <tt>User</tt> can choose between incorporating its <tt>Use</tt> objects
2882 or refer to them out-of-line by means of a pointer. A mixed variant
2883 (some <tt>Use</tt>s inline others hung off) is impractical and breaks the invariant
2884 that the <tt>Use</tt> objects belonging to the same <tt>User</tt> form a contiguous array.
2888 We have 2 different layouts in the <tt>User</tt> (sub)classes:
2891 The <tt>Use</tt> object(s) are inside (resp. at fixed offset) of the <tt>User</tt>
2892 object and there are a fixed number of them.</p>
2895 The <tt>Use</tt> object(s) are referenced by a pointer to an
2896 array from the <tt>User</tt> object and there may be a variable
2900 As of v2.4 each layout still possesses a direct pointer to the
2901 start of the array of <tt>Use</tt>s. Though not mandatory for layout a),
2902 we stick to this redundancy for the sake of simplicity.
2903 The <tt>User</tt> object also stores the number of <tt>Use</tt> objects it
2904 has. (Theoretically this information can also be calculated
2905 given the scheme presented below.)</p>
2907 Special forms of allocation operators (<tt>operator new</tt>)
2908 enforce the following memory layouts:</p>
2911 <li><p>Layout a) is modelled by prepending the <tt>User</tt> object by the <tt>Use[]</tt> array.</p>
2914 ...---.---.---.---.-------...
2915 | P | P | P | P | User
2916 '''---'---'---'---'-------'''
2919 <li><p>Layout b) is modelled by pointing at the <tt>Use[]</tt> array.</p>
2931 <i>(In the above figures '<tt>P</tt>' stands for the <tt>Use**</tt> that
2932 is stored in each <tt>Use</tt> object in the member <tt>Use::Prev</tt>)</i>
2936 <!-- ______________________________________________________________________ -->
2938 <a name="Waymarking">The waymarking algorithm</a>
2943 Since the <tt>Use</tt> objects are deprived of the direct (back)pointer to
2944 their <tt>User</tt> objects, there must be a fast and exact method to
2945 recover it. This is accomplished by the following scheme:</p>
2947 A bit-encoding in the 2 LSBits (least significant bits) of the <tt>Use::Prev</tt> allows to find the
2948 start of the <tt>User</tt> object:
2950 <li><tt>00</tt> —> binary digit 0</li>
2951 <li><tt>01</tt> —> binary digit 1</li>
2952 <li><tt>10</tt> —> stop and calculate (<tt>s</tt>)</li>
2953 <li><tt>11</tt> —> full stop (<tt>S</tt>)</li>
2956 Given a <tt>Use*</tt>, all we have to do is to walk till we get
2957 a stop and we either have a <tt>User</tt> immediately behind or
2958 we have to walk to the next stop picking up digits
2959 and calculating the offset:</p>
2961 .---.---.---.---.---.---.---.---.---.---.---.---.---.---.---.---.----------------
2962 | 1 | s | 1 | 0 | 1 | 0 | s | 1 | 1 | 0 | s | 1 | 1 | s | 1 | S | User (or User*)
2963 '---'---'---'---'---'---'---'---'---'---'---'---'---'---'---'---'----------------
2964 |+15 |+10 |+6 |+3 |+1
2967 | | |______________________>
2968 | |______________________________________>
2969 |__________________________________________________________>
2972 Only the significant number of bits need to be stored between the
2973 stops, so that the <i>worst case is 20 memory accesses</i> when there are
2974 1000 <tt>Use</tt> objects associated with a <tt>User</tt>.</p>
2978 <!-- ______________________________________________________________________ -->
2980 <a name="ReferenceImpl">Reference implementation</a>
2985 The following literate Haskell fragment demonstrates the concept:</p>
2987 <div class="doc_code">
2989 > import Test.QuickCheck
2991 > digits :: Int -> [Char] -> [Char]
2992 > digits 0 acc = '0' : acc
2993 > digits 1 acc = '1' : acc
2994 > digits n acc = digits (n `div` 2) $ digits (n `mod` 2) acc
2996 > dist :: Int -> [Char] -> [Char]
2999 > dist 1 acc = let r = dist 0 acc in 's' : digits (length r) r
3000 > dist n acc = dist (n - 1) $ dist 1 acc
3002 > takeLast n ss = reverse $ take n $ reverse ss
3004 > test = takeLast 40 $ dist 20 []
3009 Printing <test> gives: <tt>"1s100000s11010s10100s1111s1010s110s11s1S"</tt></p>
3011 The reverse algorithm computes the length of the string just by examining
3012 a certain prefix:</p>
3014 <div class="doc_code">
3016 > pref :: [Char] -> Int
3018 > pref ('s':'1':rest) = decode 2 1 rest
3019 > pref (_:rest) = 1 + pref rest
3021 > decode walk acc ('0':rest) = decode (walk + 1) (acc * 2) rest
3022 > decode walk acc ('1':rest) = decode (walk + 1) (acc * 2 + 1) rest
3023 > decode walk acc _ = walk + acc
3028 Now, as expected, printing <pref test> gives <tt>40</tt>.</p>
3030 We can <i>quickCheck</i> this with following property:</p>
3032 <div class="doc_code">
3034 > testcase = dist 2000 []
3035 > testcaseLength = length testcase
3037 > identityProp n = n > 0 && n <= testcaseLength ==> length arr == pref arr
3038 > where arr = takeLast n testcase
3043 As expected <quickCheck identityProp> gives:</p>
3046 *Main> quickCheck identityProp
3047 OK, passed 100 tests.
3050 Let's be a bit more exhaustive:</p>
3052 <div class="doc_code">
3055 > deepCheck p = check (defaultConfig { configMaxTest = 500 }) p
3060 And here is the result of <deepCheck identityProp>:</p>
3063 *Main> deepCheck identityProp
3064 OK, passed 500 tests.
3069 <!-- ______________________________________________________________________ -->
3071 <a name="Tagging">Tagging considerations</a>
3077 To maintain the invariant that the 2 LSBits of each <tt>Use**</tt> in <tt>Use</tt>
3078 never change after being set up, setters of <tt>Use::Prev</tt> must re-tag the
3079 new <tt>Use**</tt> on every modification. Accordingly getters must strip the
3082 For layout b) instead of the <tt>User</tt> we find a pointer (<tt>User*</tt> with LSBit set).
3083 Following this pointer brings us to the <tt>User</tt>. A portable trick ensures
3084 that the first bytes of <tt>User</tt> (if interpreted as a pointer) never has
3085 the LSBit set. (Portability is relying on the fact that all known compilers place the
3086 <tt>vptr</tt> in the first word of the instances.)</p>
3094 <!-- *********************************************************************** -->
3096 <a name="coreclasses">The Core LLVM Class Hierarchy Reference </a>
3098 <!-- *********************************************************************** -->
3101 <p><tt>#include "<a href="/doxygen/Type_8h-source.html">llvm/Type.h</a>"</tt>
3102 <br>doxygen info: <a href="/doxygen/classllvm_1_1Type.html">Type Class</a></p>
3104 <p>The Core LLVM classes are the primary means of representing the program
3105 being inspected or transformed. The core LLVM classes are defined in
3106 header files in the <tt>include/llvm/</tt> directory, and implemented in
3107 the <tt>lib/VMCore</tt> directory.</p>
3109 <!-- ======================================================================= -->
3111 <a name="Type">The <tt>Type</tt> class and Derived Types</a>
3116 <p><tt>Type</tt> is a superclass of all type classes. Every <tt>Value</tt> has
3117 a <tt>Type</tt>. <tt>Type</tt> cannot be instantiated directly but only
3118 through its subclasses. Certain primitive types (<tt>VoidType</tt>,
3119 <tt>LabelType</tt>, <tt>FloatType</tt> and <tt>DoubleType</tt>) have hidden
3120 subclasses. They are hidden because they offer no useful functionality beyond
3121 what the <tt>Type</tt> class offers except to distinguish themselves from
3122 other subclasses of <tt>Type</tt>.</p>
3123 <p>All other types are subclasses of <tt>DerivedType</tt>. Types can be
3124 named, but this is not a requirement. There exists exactly
3125 one instance of a given shape at any one time. This allows type equality to
3126 be performed with address equality of the Type Instance. That is, given two
3127 <tt>Type*</tt> values, the types are identical if the pointers are identical.
3130 <!-- _______________________________________________________________________ -->
3132 <a name="m_Type">Important Public Methods</a>
3138 <li><tt>bool isIntegerTy() const</tt>: Returns true for any integer type.</li>
3140 <li><tt>bool isFloatingPointTy()</tt>: Return true if this is one of the five
3141 floating point types.</li>
3143 <li><tt>bool isSized()</tt>: Return true if the type has known size. Things
3144 that don't have a size are abstract types, labels and void.</li>
3149 <!-- _______________________________________________________________________ -->
3151 <a name="derivedtypes">Important Derived Types</a>
3155 <dt><tt>IntegerType</tt></dt>
3156 <dd>Subclass of DerivedType that represents integer types of any bit width.
3157 Any bit width between <tt>IntegerType::MIN_INT_BITS</tt> (1) and
3158 <tt>IntegerType::MAX_INT_BITS</tt> (~8 million) can be represented.
3160 <li><tt>static const IntegerType* get(unsigned NumBits)</tt>: get an integer
3161 type of a specific bit width.</li>
3162 <li><tt>unsigned getBitWidth() const</tt>: Get the bit width of an integer
3166 <dt><tt>SequentialType</tt></dt>
3167 <dd>This is subclassed by ArrayType, PointerType and VectorType.
3169 <li><tt>const Type * getElementType() const</tt>: Returns the type of each
3170 of the elements in the sequential type. </li>
3173 <dt><tt>ArrayType</tt></dt>
3174 <dd>This is a subclass of SequentialType and defines the interface for array
3177 <li><tt>unsigned getNumElements() const</tt>: Returns the number of
3178 elements in the array. </li>
3181 <dt><tt>PointerType</tt></dt>
3182 <dd>Subclass of SequentialType for pointer types.</dd>
3183 <dt><tt>VectorType</tt></dt>
3184 <dd>Subclass of SequentialType for vector types. A
3185 vector type is similar to an ArrayType but is distinguished because it is
3186 a first class type whereas ArrayType is not. Vector types are used for
3187 vector operations and are usually small vectors of of an integer or floating
3189 <dt><tt>StructType</tt></dt>
3190 <dd>Subclass of DerivedTypes for struct types.</dd>
3191 <dt><tt><a name="FunctionType">FunctionType</a></tt></dt>
3192 <dd>Subclass of DerivedTypes for function types.
3194 <li><tt>bool isVarArg() const</tt>: Returns true if it's a vararg
3196 <li><tt> const Type * getReturnType() const</tt>: Returns the
3197 return type of the function.</li>
3198 <li><tt>const Type * getParamType (unsigned i)</tt>: Returns
3199 the type of the ith parameter.</li>
3200 <li><tt> const unsigned getNumParams() const</tt>: Returns the
3201 number of formal parameters.</li>
3209 <!-- ======================================================================= -->
3211 <a name="Module">The <tt>Module</tt> class</a>
3217 href="/doxygen/Module_8h-source.html">llvm/Module.h</a>"</tt><br> doxygen info:
3218 <a href="/doxygen/classllvm_1_1Module.html">Module Class</a></p>
3220 <p>The <tt>Module</tt> class represents the top level structure present in LLVM
3221 programs. An LLVM module is effectively either a translation unit of the
3222 original program or a combination of several translation units merged by the
3223 linker. The <tt>Module</tt> class keeps track of a list of <a
3224 href="#Function"><tt>Function</tt></a>s, a list of <a
3225 href="#GlobalVariable"><tt>GlobalVariable</tt></a>s, and a <a
3226 href="#SymbolTable"><tt>SymbolTable</tt></a>. Additionally, it contains a few
3227 helpful member functions that try to make common operations easy.</p>
3229 <!-- _______________________________________________________________________ -->
3231 <a name="m_Module">Important Public Members of the <tt>Module</tt> class</a>
3237 <li><tt>Module::Module(std::string name = "")</tt></li>
3240 <p>Constructing a <a href="#Module">Module</a> is easy. You can optionally
3241 provide a name for it (probably based on the name of the translation unit).</p>
3244 <li><tt>Module::iterator</tt> - Typedef for function list iterator<br>
3245 <tt>Module::const_iterator</tt> - Typedef for const_iterator.<br>
3247 <tt>begin()</tt>, <tt>end()</tt>
3248 <tt>size()</tt>, <tt>empty()</tt>
3250 <p>These are forwarding methods that make it easy to access the contents of
3251 a <tt>Module</tt> object's <a href="#Function"><tt>Function</tt></a>
3254 <li><tt>Module::FunctionListType &getFunctionList()</tt>
3256 <p> Returns the list of <a href="#Function"><tt>Function</tt></a>s. This is
3257 necessary to use when you need to update the list or perform a complex
3258 action that doesn't have a forwarding method.</p>
3260 <p><!-- Global Variable --></p></li>
3266 <li><tt>Module::global_iterator</tt> - Typedef for global variable list iterator<br>
3268 <tt>Module::const_global_iterator</tt> - Typedef for const_iterator.<br>
3270 <tt>global_begin()</tt>, <tt>global_end()</tt>
3271 <tt>global_size()</tt>, <tt>global_empty()</tt>
3273 <p> These are forwarding methods that make it easy to access the contents of
3274 a <tt>Module</tt> object's <a
3275 href="#GlobalVariable"><tt>GlobalVariable</tt></a> list.</p></li>
3277 <li><tt>Module::GlobalListType &getGlobalList()</tt>
3279 <p>Returns the list of <a
3280 href="#GlobalVariable"><tt>GlobalVariable</tt></a>s. This is necessary to
3281 use when you need to update the list or perform a complex action that
3282 doesn't have a forwarding method.</p>
3284 <p><!-- Symbol table stuff --> </p></li>
3290 <li><tt><a href="#SymbolTable">SymbolTable</a> *getSymbolTable()</tt>
3292 <p>Return a reference to the <a href="#SymbolTable"><tt>SymbolTable</tt></a>
3293 for this <tt>Module</tt>.</p>
3295 <p><!-- Convenience methods --></p></li>
3301 <li><tt><a href="#Function">Function</a> *getFunction(const std::string
3302 &Name, const <a href="#FunctionType">FunctionType</a> *Ty)</tt>
3304 <p>Look up the specified function in the <tt>Module</tt> <a
3305 href="#SymbolTable"><tt>SymbolTable</tt></a>. If it does not exist, return
3306 <tt>null</tt>.</p></li>
3308 <li><tt><a href="#Function">Function</a> *getOrInsertFunction(const
3309 std::string &Name, const <a href="#FunctionType">FunctionType</a> *T)</tt>
3311 <p>Look up the specified function in the <tt>Module</tt> <a
3312 href="#SymbolTable"><tt>SymbolTable</tt></a>. If it does not exist, add an
3313 external declaration for the function and return it.</p></li>
3315 <li><tt>std::string getTypeName(const <a href="#Type">Type</a> *Ty)</tt>
3317 <p>If there is at least one entry in the <a
3318 href="#SymbolTable"><tt>SymbolTable</tt></a> for the specified <a
3319 href="#Type"><tt>Type</tt></a>, return it. Otherwise return the empty
3322 <li><tt>bool addTypeName(const std::string &Name, const <a
3323 href="#Type">Type</a> *Ty)</tt>
3325 <p>Insert an entry in the <a href="#SymbolTable"><tt>SymbolTable</tt></a>
3326 mapping <tt>Name</tt> to <tt>Ty</tt>. If there is already an entry for this
3327 name, true is returned and the <a
3328 href="#SymbolTable"><tt>SymbolTable</tt></a> is not modified.</p></li>
3335 <!-- ======================================================================= -->
3337 <a name="Value">The <tt>Value</tt> class</a>
3342 <p><tt>#include "<a href="/doxygen/Value_8h-source.html">llvm/Value.h</a>"</tt>
3344 doxygen info: <a href="/doxygen/classllvm_1_1Value.html">Value Class</a></p>
3346 <p>The <tt>Value</tt> class is the most important class in the LLVM Source
3347 base. It represents a typed value that may be used (among other things) as an
3348 operand to an instruction. There are many different types of <tt>Value</tt>s,
3349 such as <a href="#Constant"><tt>Constant</tt></a>s,<a
3350 href="#Argument"><tt>Argument</tt></a>s. Even <a
3351 href="#Instruction"><tt>Instruction</tt></a>s and <a
3352 href="#Function"><tt>Function</tt></a>s are <tt>Value</tt>s.</p>
3354 <p>A particular <tt>Value</tt> may be used many times in the LLVM representation
3355 for a program. For example, an incoming argument to a function (represented
3356 with an instance of the <a href="#Argument">Argument</a> class) is "used" by
3357 every instruction in the function that references the argument. To keep track
3358 of this relationship, the <tt>Value</tt> class keeps a list of all of the <a
3359 href="#User"><tt>User</tt></a>s that is using it (the <a
3360 href="#User"><tt>User</tt></a> class is a base class for all nodes in the LLVM
3361 graph that can refer to <tt>Value</tt>s). This use list is how LLVM represents
3362 def-use information in the program, and is accessible through the <tt>use_</tt>*
3363 methods, shown below.</p>
3365 <p>Because LLVM is a typed representation, every LLVM <tt>Value</tt> is typed,
3366 and this <a href="#Type">Type</a> is available through the <tt>getType()</tt>
3367 method. In addition, all LLVM values can be named. The "name" of the
3368 <tt>Value</tt> is a symbolic string printed in the LLVM code:</p>
3370 <div class="doc_code">
3372 %<b>foo</b> = add i32 1, 2
3376 <p><a name="nameWarning">The name of this instruction is "foo".</a> <b>NOTE</b>
3377 that the name of any value may be missing (an empty string), so names should
3378 <b>ONLY</b> be used for debugging (making the source code easier to read,
3379 debugging printouts), they should not be used to keep track of values or map
3380 between them. For this purpose, use a <tt>std::map</tt> of pointers to the
3381 <tt>Value</tt> itself instead.</p>
3383 <p>One important aspect of LLVM is that there is no distinction between an SSA
3384 variable and the operation that produces it. Because of this, any reference to
3385 the value produced by an instruction (or the value available as an incoming
3386 argument, for example) is represented as a direct pointer to the instance of
3388 represents this value. Although this may take some getting used to, it
3389 simplifies the representation and makes it easier to manipulate.</p>
3391 <!-- _______________________________________________________________________ -->
3393 <a name="m_Value">Important Public Members of the <tt>Value</tt> class</a>
3399 <li><tt>Value::use_iterator</tt> - Typedef for iterator over the
3401 <tt>Value::const_use_iterator</tt> - Typedef for const_iterator over
3403 <tt>unsigned use_size()</tt> - Returns the number of users of the
3405 <tt>bool use_empty()</tt> - Returns true if there are no users.<br>
3406 <tt>use_iterator use_begin()</tt> - Get an iterator to the start of
3408 <tt>use_iterator use_end()</tt> - Get an iterator to the end of the
3410 <tt><a href="#User">User</a> *use_back()</tt> - Returns the last
3411 element in the list.
3412 <p> These methods are the interface to access the def-use
3413 information in LLVM. As with all other iterators in LLVM, the naming
3414 conventions follow the conventions defined by the <a href="#stl">STL</a>.</p>
3416 <li><tt><a href="#Type">Type</a> *getType() const</tt>
3417 <p>This method returns the Type of the Value.</p>
3419 <li><tt>bool hasName() const</tt><br>
3420 <tt>std::string getName() const</tt><br>
3421 <tt>void setName(const std::string &Name)</tt>
3422 <p> This family of methods is used to access and assign a name to a <tt>Value</tt>,
3423 be aware of the <a href="#nameWarning">precaution above</a>.</p>
3425 <li><tt>void replaceAllUsesWith(Value *V)</tt>
3427 <p>This method traverses the use list of a <tt>Value</tt> changing all <a
3428 href="#User"><tt>User</tt>s</a> of the current value to refer to
3429 "<tt>V</tt>" instead. For example, if you detect that an instruction always
3430 produces a constant value (for example through constant folding), you can
3431 replace all uses of the instruction with the constant like this:</p>
3433 <div class="doc_code">
3435 Inst->replaceAllUsesWith(ConstVal);
3445 <!-- ======================================================================= -->
3447 <a name="User">The <tt>User</tt> class</a>
3453 <tt>#include "<a href="/doxygen/User_8h-source.html">llvm/User.h</a>"</tt><br>
3454 doxygen info: <a href="/doxygen/classllvm_1_1User.html">User Class</a><br>
3455 Superclass: <a href="#Value"><tt>Value</tt></a></p>
3457 <p>The <tt>User</tt> class is the common base class of all LLVM nodes that may
3458 refer to <a href="#Value"><tt>Value</tt></a>s. It exposes a list of "Operands"
3459 that are all of the <a href="#Value"><tt>Value</tt></a>s that the User is
3460 referring to. The <tt>User</tt> class itself is a subclass of
3463 <p>The operands of a <tt>User</tt> point directly to the LLVM <a
3464 href="#Value"><tt>Value</tt></a> that it refers to. Because LLVM uses Static
3465 Single Assignment (SSA) form, there can only be one definition referred to,
3466 allowing this direct connection. This connection provides the use-def
3467 information in LLVM.</p>
3469 <!-- _______________________________________________________________________ -->
3471 <a name="m_User">Important Public Members of the <tt>User</tt> class</a>
3476 <p>The <tt>User</tt> class exposes the operand list in two ways: through
3477 an index access interface and through an iterator based interface.</p>
3480 <li><tt>Value *getOperand(unsigned i)</tt><br>
3481 <tt>unsigned getNumOperands()</tt>
3482 <p> These two methods expose the operands of the <tt>User</tt> in a
3483 convenient form for direct access.</p></li>
3485 <li><tt>User::op_iterator</tt> - Typedef for iterator over the operand
3487 <tt>op_iterator op_begin()</tt> - Get an iterator to the start of
3488 the operand list.<br>
3489 <tt>op_iterator op_end()</tt> - Get an iterator to the end of the
3491 <p> Together, these methods make up the iterator based interface to
3492 the operands of a <tt>User</tt>.</p></li>
3499 <!-- ======================================================================= -->
3501 <a name="Instruction">The <tt>Instruction</tt> class</a>
3506 <p><tt>#include "</tt><tt><a
3507 href="/doxygen/Instruction_8h-source.html">llvm/Instruction.h</a>"</tt><br>
3508 doxygen info: <a href="/doxygen/classllvm_1_1Instruction.html">Instruction Class</a><br>
3509 Superclasses: <a href="#User"><tt>User</tt></a>, <a
3510 href="#Value"><tt>Value</tt></a></p>
3512 <p>The <tt>Instruction</tt> class is the common base class for all LLVM
3513 instructions. It provides only a few methods, but is a very commonly used
3514 class. The primary data tracked by the <tt>Instruction</tt> class itself is the
3515 opcode (instruction type) and the parent <a
3516 href="#BasicBlock"><tt>BasicBlock</tt></a> the <tt>Instruction</tt> is embedded
3517 into. To represent a specific type of instruction, one of many subclasses of
3518 <tt>Instruction</tt> are used.</p>
3520 <p> Because the <tt>Instruction</tt> class subclasses the <a
3521 href="#User"><tt>User</tt></a> class, its operands can be accessed in the same
3522 way as for other <a href="#User"><tt>User</tt></a>s (with the
3523 <tt>getOperand()</tt>/<tt>getNumOperands()</tt> and
3524 <tt>op_begin()</tt>/<tt>op_end()</tt> methods).</p> <p> An important file for
3525 the <tt>Instruction</tt> class is the <tt>llvm/Instruction.def</tt> file. This
3526 file contains some meta-data about the various different types of instructions
3527 in LLVM. It describes the enum values that are used as opcodes (for example
3528 <tt>Instruction::Add</tt> and <tt>Instruction::ICmp</tt>), as well as the
3529 concrete sub-classes of <tt>Instruction</tt> that implement the instruction (for
3530 example <tt><a href="#BinaryOperator">BinaryOperator</a></tt> and <tt><a
3531 href="#CmpInst">CmpInst</a></tt>). Unfortunately, the use of macros in
3532 this file confuses doxygen, so these enum values don't show up correctly in the
3533 <a href="/doxygen/classllvm_1_1Instruction.html">doxygen output</a>.</p>
3535 <!-- _______________________________________________________________________ -->
3537 <a name="s_Instruction">
3538 Important Subclasses of the <tt>Instruction</tt> class
3543 <li><tt><a name="BinaryOperator">BinaryOperator</a></tt>
3544 <p>This subclasses represents all two operand instructions whose operands
3545 must be the same type, except for the comparison instructions.</p></li>
3546 <li><tt><a name="CastInst">CastInst</a></tt>
3547 <p>This subclass is the parent of the 12 casting instructions. It provides
3548 common operations on cast instructions.</p>
3549 <li><tt><a name="CmpInst">CmpInst</a></tt>
3550 <p>This subclass respresents the two comparison instructions,
3551 <a href="LangRef.html#i_icmp">ICmpInst</a> (integer opreands), and
3552 <a href="LangRef.html#i_fcmp">FCmpInst</a> (floating point operands).</p>
3553 <li><tt><a name="TerminatorInst">TerminatorInst</a></tt>
3554 <p>This subclass is the parent of all terminator instructions (those which
3555 can terminate a block).</p>
3559 <!-- _______________________________________________________________________ -->
3561 <a name="m_Instruction">
3562 Important Public Members of the <tt>Instruction</tt> class
3569 <li><tt><a href="#BasicBlock">BasicBlock</a> *getParent()</tt>
3570 <p>Returns the <a href="#BasicBlock"><tt>BasicBlock</tt></a> that
3571 this <tt>Instruction</tt> is embedded into.</p></li>
3572 <li><tt>bool mayWriteToMemory()</tt>
3573 <p>Returns true if the instruction writes to memory, i.e. it is a
3574 <tt>call</tt>,<tt>free</tt>,<tt>invoke</tt>, or <tt>store</tt>.</p></li>
3575 <li><tt>unsigned getOpcode()</tt>
3576 <p>Returns the opcode for the <tt>Instruction</tt>.</p></li>
3577 <li><tt><a href="#Instruction">Instruction</a> *clone() const</tt>
3578 <p>Returns another instance of the specified instruction, identical
3579 in all ways to the original except that the instruction has no parent
3580 (ie it's not embedded into a <a href="#BasicBlock"><tt>BasicBlock</tt></a>),
3581 and it has no name</p></li>
3588 <!-- ======================================================================= -->
3590 <a name="Constant">The <tt>Constant</tt> class and subclasses</a>
3595 <p>Constant represents a base class for different types of constants. It
3596 is subclassed by ConstantInt, ConstantArray, etc. for representing
3597 the various types of Constants. <a href="#GlobalValue">GlobalValue</a> is also
3598 a subclass, which represents the address of a global variable or function.
3601 <!-- _______________________________________________________________________ -->
3602 <h4>Important Subclasses of Constant</h4>
3605 <li>ConstantInt : This subclass of Constant represents an integer constant of
3608 <li><tt>const APInt& getValue() const</tt>: Returns the underlying
3609 value of this constant, an APInt value.</li>
3610 <li><tt>int64_t getSExtValue() const</tt>: Converts the underlying APInt
3611 value to an int64_t via sign extension. If the value (not the bit width)
3612 of the APInt is too large to fit in an int64_t, an assertion will result.
3613 For this reason, use of this method is discouraged.</li>
3614 <li><tt>uint64_t getZExtValue() const</tt>: Converts the underlying APInt
3615 value to a uint64_t via zero extension. IF the value (not the bit width)
3616 of the APInt is too large to fit in a uint64_t, an assertion will result.
3617 For this reason, use of this method is discouraged.</li>
3618 <li><tt>static ConstantInt* get(const APInt& Val)</tt>: Returns the
3619 ConstantInt object that represents the value provided by <tt>Val</tt>.
3620 The type is implied as the IntegerType that corresponds to the bit width
3621 of <tt>Val</tt>.</li>
3622 <li><tt>static ConstantInt* get(const Type *Ty, uint64_t Val)</tt>:
3623 Returns the ConstantInt object that represents the value provided by
3624 <tt>Val</tt> for integer type <tt>Ty</tt>.</li>
3627 <li>ConstantFP : This class represents a floating point constant.
3629 <li><tt>double getValue() const</tt>: Returns the underlying value of
3630 this constant. </li>
3633 <li>ConstantArray : This represents a constant array.
3635 <li><tt>const std::vector<Use> &getValues() const</tt>: Returns
3636 a vector of component constants that makeup this array. </li>
3639 <li>ConstantStruct : This represents a constant struct.
3641 <li><tt>const std::vector<Use> &getValues() const</tt>: Returns
3642 a vector of component constants that makeup this array. </li>
3645 <li>GlobalValue : This represents either a global variable or a function. In
3646 either case, the value is a constant fixed address (after linking).
3653 <!-- ======================================================================= -->
3655 <a name="GlobalValue">The <tt>GlobalValue</tt> class</a>
3661 href="/doxygen/GlobalValue_8h-source.html">llvm/GlobalValue.h</a>"</tt><br>
3662 doxygen info: <a href="/doxygen/classllvm_1_1GlobalValue.html">GlobalValue
3664 Superclasses: <a href="#Constant"><tt>Constant</tt></a>,
3665 <a href="#User"><tt>User</tt></a>, <a href="#Value"><tt>Value</tt></a></p>
3667 <p>Global values (<a href="#GlobalVariable"><tt>GlobalVariable</tt></a>s or <a
3668 href="#Function"><tt>Function</tt></a>s) are the only LLVM values that are
3669 visible in the bodies of all <a href="#Function"><tt>Function</tt></a>s.
3670 Because they are visible at global scope, they are also subject to linking with
3671 other globals defined in different translation units. To control the linking
3672 process, <tt>GlobalValue</tt>s know their linkage rules. Specifically,
3673 <tt>GlobalValue</tt>s know whether they have internal or external linkage, as
3674 defined by the <tt>LinkageTypes</tt> enumeration.</p>
3676 <p>If a <tt>GlobalValue</tt> has internal linkage (equivalent to being
3677 <tt>static</tt> in C), it is not visible to code outside the current translation
3678 unit, and does not participate in linking. If it has external linkage, it is
3679 visible to external code, and does participate in linking. In addition to
3680 linkage information, <tt>GlobalValue</tt>s keep track of which <a
3681 href="#Module"><tt>Module</tt></a> they are currently part of.</p>
3683 <p>Because <tt>GlobalValue</tt>s are memory objects, they are always referred to
3684 by their <b>address</b>. As such, the <a href="#Type"><tt>Type</tt></a> of a
3685 global is always a pointer to its contents. It is important to remember this
3686 when using the <tt>GetElementPtrInst</tt> instruction because this pointer must
3687 be dereferenced first. For example, if you have a <tt>GlobalVariable</tt> (a
3688 subclass of <tt>GlobalValue)</tt> that is an array of 24 ints, type <tt>[24 x
3689 i32]</tt>, then the <tt>GlobalVariable</tt> is a pointer to that array. Although
3690 the address of the first element of this array and the value of the
3691 <tt>GlobalVariable</tt> are the same, they have different types. The
3692 <tt>GlobalVariable</tt>'s type is <tt>[24 x i32]</tt>. The first element's type
3693 is <tt>i32.</tt> Because of this, accessing a global value requires you to
3694 dereference the pointer with <tt>GetElementPtrInst</tt> first, then its elements
3695 can be accessed. This is explained in the <a href="LangRef.html#globalvars">LLVM
3696 Language Reference Manual</a>.</p>
3698 <!-- _______________________________________________________________________ -->
3700 <a name="m_GlobalValue">
3701 Important Public Members of the <tt>GlobalValue</tt> class
3708 <li><tt>bool hasInternalLinkage() const</tt><br>
3709 <tt>bool hasExternalLinkage() const</tt><br>
3710 <tt>void setInternalLinkage(bool HasInternalLinkage)</tt>
3711 <p> These methods manipulate the linkage characteristics of the <tt>GlobalValue</tt>.</p>
3714 <li><tt><a href="#Module">Module</a> *getParent()</tt>
3715 <p> This returns the <a href="#Module"><tt>Module</tt></a> that the
3716 GlobalValue is currently embedded into.</p></li>
3723 <!-- ======================================================================= -->
3725 <a name="Function">The <tt>Function</tt> class</a>
3731 href="/doxygen/Function_8h-source.html">llvm/Function.h</a>"</tt><br> doxygen
3732 info: <a href="/doxygen/classllvm_1_1Function.html">Function Class</a><br>
3733 Superclasses: <a href="#GlobalValue"><tt>GlobalValue</tt></a>,
3734 <a href="#Constant"><tt>Constant</tt></a>,
3735 <a href="#User"><tt>User</tt></a>,
3736 <a href="#Value"><tt>Value</tt></a></p>
3738 <p>The <tt>Function</tt> class represents a single procedure in LLVM. It is
3739 actually one of the more complex classes in the LLVM hierarchy because it must
3740 keep track of a large amount of data. The <tt>Function</tt> class keeps track
3741 of a list of <a href="#BasicBlock"><tt>BasicBlock</tt></a>s, a list of formal
3742 <a href="#Argument"><tt>Argument</tt></a>s, and a
3743 <a href="#SymbolTable"><tt>SymbolTable</tt></a>.</p>
3745 <p>The list of <a href="#BasicBlock"><tt>BasicBlock</tt></a>s is the most
3746 commonly used part of <tt>Function</tt> objects. The list imposes an implicit
3747 ordering of the blocks in the function, which indicate how the code will be
3748 laid out by the backend. Additionally, the first <a
3749 href="#BasicBlock"><tt>BasicBlock</tt></a> is the implicit entry node for the
3750 <tt>Function</tt>. It is not legal in LLVM to explicitly branch to this initial
3751 block. There are no implicit exit nodes, and in fact there may be multiple exit
3752 nodes from a single <tt>Function</tt>. If the <a
3753 href="#BasicBlock"><tt>BasicBlock</tt></a> list is empty, this indicates that
3754 the <tt>Function</tt> is actually a function declaration: the actual body of the
3755 function hasn't been linked in yet.</p>
3757 <p>In addition to a list of <a href="#BasicBlock"><tt>BasicBlock</tt></a>s, the
3758 <tt>Function</tt> class also keeps track of the list of formal <a
3759 href="#Argument"><tt>Argument</tt></a>s that the function receives. This
3760 container manages the lifetime of the <a href="#Argument"><tt>Argument</tt></a>
3761 nodes, just like the <a href="#BasicBlock"><tt>BasicBlock</tt></a> list does for
3762 the <a href="#BasicBlock"><tt>BasicBlock</tt></a>s.</p>
3764 <p>The <a href="#SymbolTable"><tt>SymbolTable</tt></a> is a very rarely used
3765 LLVM feature that is only used when you have to look up a value by name. Aside
3766 from that, the <a href="#SymbolTable"><tt>SymbolTable</tt></a> is used
3767 internally to make sure that there are not conflicts between the names of <a
3768 href="#Instruction"><tt>Instruction</tt></a>s, <a
3769 href="#BasicBlock"><tt>BasicBlock</tt></a>s, or <a
3770 href="#Argument"><tt>Argument</tt></a>s in the function body.</p>
3772 <p>Note that <tt>Function</tt> is a <a href="#GlobalValue">GlobalValue</a>
3773 and therefore also a <a href="#Constant">Constant</a>. The value of the function
3774 is its address (after linking) which is guaranteed to be constant.</p>
3776 <!-- _______________________________________________________________________ -->
3778 <a name="m_Function">
3779 Important Public Members of the <tt>Function</tt> class
3786 <li><tt>Function(const </tt><tt><a href="#FunctionType">FunctionType</a>
3787 *Ty, LinkageTypes Linkage, const std::string &N = "", Module* Parent = 0)</tt>
3789 <p>Constructor used when you need to create new <tt>Function</tt>s to add
3790 the the program. The constructor must specify the type of the function to
3791 create and what type of linkage the function should have. The <a
3792 href="#FunctionType"><tt>FunctionType</tt></a> argument
3793 specifies the formal arguments and return value for the function. The same
3794 <a href="#FunctionType"><tt>FunctionType</tt></a> value can be used to
3795 create multiple functions. The <tt>Parent</tt> argument specifies the Module
3796 in which the function is defined. If this argument is provided, the function
3797 will automatically be inserted into that module's list of
3800 <li><tt>bool isDeclaration()</tt>
3802 <p>Return whether or not the <tt>Function</tt> has a body defined. If the
3803 function is "external", it does not have a body, and thus must be resolved
3804 by linking with a function defined in a different translation unit.</p></li>
3806 <li><tt>Function::iterator</tt> - Typedef for basic block list iterator<br>
3807 <tt>Function::const_iterator</tt> - Typedef for const_iterator.<br>
3809 <tt>begin()</tt>, <tt>end()</tt>
3810 <tt>size()</tt>, <tt>empty()</tt>
3812 <p>These are forwarding methods that make it easy to access the contents of
3813 a <tt>Function</tt> object's <a href="#BasicBlock"><tt>BasicBlock</tt></a>
3816 <li><tt>Function::BasicBlockListType &getBasicBlockList()</tt>
3818 <p>Returns the list of <a href="#BasicBlock"><tt>BasicBlock</tt></a>s. This
3819 is necessary to use when you need to update the list or perform a complex
3820 action that doesn't have a forwarding method.</p></li>
3822 <li><tt>Function::arg_iterator</tt> - Typedef for the argument list
3824 <tt>Function::const_arg_iterator</tt> - Typedef for const_iterator.<br>
3826 <tt>arg_begin()</tt>, <tt>arg_end()</tt>
3827 <tt>arg_size()</tt>, <tt>arg_empty()</tt>
3829 <p>These are forwarding methods that make it easy to access the contents of
3830 a <tt>Function</tt> object's <a href="#Argument"><tt>Argument</tt></a>
3833 <li><tt>Function::ArgumentListType &getArgumentList()</tt>
3835 <p>Returns the list of <a href="#Argument"><tt>Argument</tt></a>s. This is
3836 necessary to use when you need to update the list or perform a complex
3837 action that doesn't have a forwarding method.</p></li>
3839 <li><tt><a href="#BasicBlock">BasicBlock</a> &getEntryBlock()</tt>
3841 <p>Returns the entry <a href="#BasicBlock"><tt>BasicBlock</tt></a> for the
3842 function. Because the entry block for the function is always the first
3843 block, this returns the first block of the <tt>Function</tt>.</p></li>
3845 <li><tt><a href="#Type">Type</a> *getReturnType()</tt><br>
3846 <tt><a href="#FunctionType">FunctionType</a> *getFunctionType()</tt>
3848 <p>This traverses the <a href="#Type"><tt>Type</tt></a> of the
3849 <tt>Function</tt> and returns the return type of the function, or the <a
3850 href="#FunctionType"><tt>FunctionType</tt></a> of the actual
3853 <li><tt><a href="#SymbolTable">SymbolTable</a> *getSymbolTable()</tt>
3855 <p> Return a pointer to the <a href="#SymbolTable"><tt>SymbolTable</tt></a>
3856 for this <tt>Function</tt>.</p></li>
3863 <!-- ======================================================================= -->
3865 <a name="GlobalVariable">The <tt>GlobalVariable</tt> class</a>
3871 href="/doxygen/GlobalVariable_8h-source.html">llvm/GlobalVariable.h</a>"</tt>
3873 doxygen info: <a href="/doxygen/classllvm_1_1GlobalVariable.html">GlobalVariable
3875 Superclasses: <a href="#GlobalValue"><tt>GlobalValue</tt></a>,
3876 <a href="#Constant"><tt>Constant</tt></a>,
3877 <a href="#User"><tt>User</tt></a>,
3878 <a href="#Value"><tt>Value</tt></a></p>
3880 <p>Global variables are represented with the (surprise surprise)
3881 <tt>GlobalVariable</tt> class. Like functions, <tt>GlobalVariable</tt>s are also
3882 subclasses of <a href="#GlobalValue"><tt>GlobalValue</tt></a>, and as such are
3883 always referenced by their address (global values must live in memory, so their
3884 "name" refers to their constant address). See
3885 <a href="#GlobalValue"><tt>GlobalValue</tt></a> for more on this. Global
3886 variables may have an initial value (which must be a
3887 <a href="#Constant"><tt>Constant</tt></a>), and if they have an initializer,
3888 they may be marked as "constant" themselves (indicating that their contents
3889 never change at runtime).</p>
3891 <!-- _______________________________________________________________________ -->
3893 <a name="m_GlobalVariable">
3894 Important Public Members of the <tt>GlobalVariable</tt> class
3901 <li><tt>GlobalVariable(const </tt><tt><a href="#Type">Type</a> *Ty, bool
3902 isConstant, LinkageTypes& Linkage, <a href="#Constant">Constant</a>
3903 *Initializer = 0, const std::string &Name = "", Module* Parent = 0)</tt>
3905 <p>Create a new global variable of the specified type. If
3906 <tt>isConstant</tt> is true then the global variable will be marked as
3907 unchanging for the program. The Linkage parameter specifies the type of
3908 linkage (internal, external, weak, linkonce, appending) for the variable.
3909 If the linkage is InternalLinkage, WeakAnyLinkage, WeakODRLinkage,
3910 LinkOnceAnyLinkage or LinkOnceODRLinkage, then the resultant
3911 global variable will have internal linkage. AppendingLinkage concatenates
3912 together all instances (in different translation units) of the variable
3913 into a single variable but is only applicable to arrays. See
3914 the <a href="LangRef.html#modulestructure">LLVM Language Reference</a> for
3915 further details on linkage types. Optionally an initializer, a name, and the
3916 module to put the variable into may be specified for the global variable as
3919 <li><tt>bool isConstant() const</tt>
3921 <p>Returns true if this is a global variable that is known not to
3922 be modified at runtime.</p></li>
3924 <li><tt>bool hasInitializer()</tt>
3926 <p>Returns true if this <tt>GlobalVariable</tt> has an intializer.</p></li>
3928 <li><tt><a href="#Constant">Constant</a> *getInitializer()</tt>
3930 <p>Returns the initial value for a <tt>GlobalVariable</tt>. It is not legal
3931 to call this method if there is no initializer.</p></li>
3938 <!-- ======================================================================= -->
3940 <a name="BasicBlock">The <tt>BasicBlock</tt> class</a>
3946 href="/doxygen/BasicBlock_8h-source.html">llvm/BasicBlock.h</a>"</tt><br>
3947 doxygen info: <a href="/doxygen/classllvm_1_1BasicBlock.html">BasicBlock
3949 Superclass: <a href="#Value"><tt>Value</tt></a></p>
3951 <p>This class represents a single entry single exit section of the code,
3952 commonly known as a basic block by the compiler community. The
3953 <tt>BasicBlock</tt> class maintains a list of <a
3954 href="#Instruction"><tt>Instruction</tt></a>s, which form the body of the block.
3955 Matching the language definition, the last element of this list of instructions
3956 is always a terminator instruction (a subclass of the <a
3957 href="#TerminatorInst"><tt>TerminatorInst</tt></a> class).</p>
3959 <p>In addition to tracking the list of instructions that make up the block, the
3960 <tt>BasicBlock</tt> class also keeps track of the <a
3961 href="#Function"><tt>Function</tt></a> that it is embedded into.</p>
3963 <p>Note that <tt>BasicBlock</tt>s themselves are <a
3964 href="#Value"><tt>Value</tt></a>s, because they are referenced by instructions
3965 like branches and can go in the switch tables. <tt>BasicBlock</tt>s have type
3968 <!-- _______________________________________________________________________ -->
3970 <a name="m_BasicBlock">
3971 Important Public Members of the <tt>BasicBlock</tt> class
3978 <li><tt>BasicBlock(const std::string &Name = "", </tt><tt><a
3979 href="#Function">Function</a> *Parent = 0)</tt>
3981 <p>The <tt>BasicBlock</tt> constructor is used to create new basic blocks for
3982 insertion into a function. The constructor optionally takes a name for the new
3983 block, and a <a href="#Function"><tt>Function</tt></a> to insert it into. If
3984 the <tt>Parent</tt> parameter is specified, the new <tt>BasicBlock</tt> is
3985 automatically inserted at the end of the specified <a
3986 href="#Function"><tt>Function</tt></a>, if not specified, the BasicBlock must be
3987 manually inserted into the <a href="#Function"><tt>Function</tt></a>.</p></li>
3989 <li><tt>BasicBlock::iterator</tt> - Typedef for instruction list iterator<br>
3990 <tt>BasicBlock::const_iterator</tt> - Typedef for const_iterator.<br>
3991 <tt>begin()</tt>, <tt>end()</tt>, <tt>front()</tt>, <tt>back()</tt>,
3992 <tt>size()</tt>, <tt>empty()</tt>
3993 STL-style functions for accessing the instruction list.
3995 <p>These methods and typedefs are forwarding functions that have the same
3996 semantics as the standard library methods of the same names. These methods
3997 expose the underlying instruction list of a basic block in a way that is easy to
3998 manipulate. To get the full complement of container operations (including
3999 operations to update the list), you must use the <tt>getInstList()</tt>
4002 <li><tt>BasicBlock::InstListType &getInstList()</tt>
4004 <p>This method is used to get access to the underlying container that actually
4005 holds the Instructions. This method must be used when there isn't a forwarding
4006 function in the <tt>BasicBlock</tt> class for the operation that you would like
4007 to perform. Because there are no forwarding functions for "updating"
4008 operations, you need to use this if you want to update the contents of a
4009 <tt>BasicBlock</tt>.</p></li>
4011 <li><tt><a href="#Function">Function</a> *getParent()</tt>
4013 <p> Returns a pointer to <a href="#Function"><tt>Function</tt></a> the block is
4014 embedded into, or a null pointer if it is homeless.</p></li>
4016 <li><tt><a href="#TerminatorInst">TerminatorInst</a> *getTerminator()</tt>
4018 <p> Returns a pointer to the terminator instruction that appears at the end of
4019 the <tt>BasicBlock</tt>. If there is no terminator instruction, or if the last
4020 instruction in the block is not a terminator, then a null pointer is
4029 <!-- ======================================================================= -->
4031 <a name="Argument">The <tt>Argument</tt> class</a>
4036 <p>This subclass of Value defines the interface for incoming formal
4037 arguments to a function. A Function maintains a list of its formal
4038 arguments. An argument has a pointer to the parent Function.</p>
4044 <!-- *********************************************************************** -->
4047 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
4048 src="http://jigsaw.w3.org/css-validator/images/vcss-blue" alt="Valid CSS"></a>
4049 <a href="http://validator.w3.org/check/referer"><img
4050 src="http://www.w3.org/Icons/valid-html401" alt="Valid HTML 4.01 Strict"></a>
4052 <a href="mailto:dhurjati@cs.uiuc.edu">Dinakar Dhurjati</a> and
4053 <a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
4054 <a href="http://llvm.org/">The LLVM Compiler Infrastructure</a><br>
4055 Last modified: $Date$